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 firom 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 corner and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white puOiOgiiipuit punis me available fui any pliuiographs or ill us ir allons appearing in this copy for an additional charge. Contact UMI directly to order.

UMI University Microfilms international A Beil & Howell information Company 300 Nortfi ZeeO Roao. Ann Aroor, Ml 48106-1346 USA

313 , 761-4700 800 521-0600

Order Number 9238160

Chinese morphology and its interface with the syntax

Dai, Xiang-ling, Ph.D.

The Ohio State University, 1992

Copyright ©1992 by Dai, Xiang-ling. All rights reserved.

UMI 300 N. Zeeb Rd. Ann Arbor, MI 4X106

CHINESE MORPHOLOGY AND ITS INTERFACE WITH THE SYNTAX

DISSERTATION

Presented in Partial Fulfillment of tfie Requirement for

tfie Degree Doctor of Philosoptiy in the Graduate

Scchool of The Ohio State University

by

Xiang-ling Dai, B.Sc., M.A.

*****

The Ohio State University

1992

Dissertation Committee: Approved by

Catherine A. Callaghan

Brian D. Joseph

Carl Pollard A- Advisor Jam es H-Y. Tai Department of Linguistics Copyright by Xiang-ling Dai 1992 ACKNOWLEDGMENTS

First of all, I wish to express my deepest gratitude to my academic advisor Arnold Zwicky, who serves as the chairperson of my dissertation committee. His theoretical framework has enormously influenced me in forming the program of this thesis. Arnold tirelessly read, re-read and re-re-read all of my papers and this dissertation, making constructive criticisms, raising insightful questions and giving valuable suggestions. Without his constant academic and moral support, this work couid not possibiy appear in the current form. It is my pleasure to complete this thesis during the year that Arnold serves as the President of the Linguistic Society of America

Thanks also go to the other committee members: Brian Joseph, Carl Pollard, Jam es Tai and Catherine Callaghan. Brian Joseph has been unfailingly encouraging, understanding and supportive of me in his five busy years of service as the chairperson of the Ohio State University

(OSU) Department of Linguistics. I am greatly indebted to him for his feedback on my papers and this thesis and for his service as my advisor in the first two years of my four-year Ph.D tenure.

It was his lecture on historical morphology and his (and Richard Janda’s) claim that no language can possibly lack both affixation and compounding that aroused my interest in Chinese morphology, leading to my first few publications in linguistics.

I am immensely grateful to Carl Pollard for his extremely useful criticism of and detailed comments on early presentations of my papers and this dissertation. Special thanks go to James

Tai for reading quite a few of my papers and chapters of this thesis, and encouraging my research in Chinese morphology, pointing out to me that this area had been virtually a gap in the literature of Chinese linguistics, it was his suggestion that led me to decide on the thesis’s topic.

Catherine Callaghan has always been encouraging and supportive throughout my six-year study at OSU, reading and commenting on several of my papers and chapters of this thesis, even after her retirement. I am very thankful to her.

My thanks also belong to the faculty members at OSU who taught me linguistics; Mary

Beckman, Wayne Cowart, David Dowty, Tom Ernst, Mike Geis, Robert Levine, Joel Nevis and

David Odden.

To the OSU Graduate School, I offer my gratitude for honoring me as one of the

Presidential Fellows in my final, dissertation, year. To the OSU Department of Linguistics, I owe much for the privileges and opportunities provided to me, including numerous graduate teaching, research and administrative asscciateships, as well as travel funds for me to attend linguistics conferences domestically and abroad. Thanks are also due to Marlene Payha and Chriss Large for their highly efficient and supportive administration of the Department office. I am also grateful to the OSU College of Humanities for the special grants which enabled me to present my papers at the Twenty-Third International Conference on Sino-Tibetan Languages and Linguistics held in

Arlington, Texas in 1990 and the First International Conference on Chinese Linguistics held in

Singapore in 1992.

For their friendship and help, I thank the following people around OSU’s Cunz Hall and

Oxiey Hall: Benjamin Ao, Jill Beckman, Christie Block, Hee-Rahk Chae, Marge Chan, Young Hee

Chung, Monica Crabtree, Jason Frank, Qian Gao, Brad Getz, Jeanette Hoipp, Joy Hoyte, Wen-zhe

Hu, Beth Hume, Kang Hyeonseok, Ken de Jong, Sun-Ah Jun, Andreas Kathol, No-Ju Kim, Claudia

Kurz, Gina Lee, Sook-Hyang Lee, Use Lehiste, Mark Libucha, Philip Miller, Yong-Kyoon No,

Frederick Parkinson, Craige Roberts, Ruth Roberts, Robert Sanders, Ying-yu Sheu, Uma

Subramanian, Halyna Sydorenko, JianTang, Hideo Tomita, Lianqing Wang, Don Winford, Darryl

Wylie, Chuck Yocom, Zheng-sheng Zhang, Zhiming Zhao, Guohe Zheng, and Ke Zou. Among them, I wish to single out Jason Frank for his thorough job of proofreading of my papers and this

dissertation. As for people outside the OSU circle, I am grateful to Konrad Koerner, James McCawley and William Wang for recommending me for the OSU Presidential Fellowship, and their constant academic and moral support. I should also mention the following nam es for their help or support:

Claire Chang, Matthew Dryer, San Duanmu, Li Fang, One-Soon Her, Chu-Ren Huang, Tom Hukari,

Richard Janda, Zixin Jiang, Cheung Shing Leung, Joan Maling, KP. Mohanan, Jerry Packard, Eric

Schiller, Chilin Shih, Richard Sproat, and Shi Zhang. Jam es Huang should be credited for his pioneering work on Chinese linguistics, which in turn stimulated my interest in the area.

I owe a great deal to Syracuse University for sponsoring my graduate study in the U.S. and providing me with graduate scholarship and teaching assistantship, and to Tej Bhatia, Jaklin

Komfilt, Mary Edwards, Jeannette Mocera, Patricia Moody, Beulah Rohrlich, William Ritchie and

Martha Wright for teaching me introductory linguistics and other courses at Syracuse. Special thanks go to Tom Chang and Robert Laubach, for without their co-sponsorship, I would have been unable to come to the U.S. for study. Mei-lan Chai, Wei Fang and Ding Tan deserve my gratitute for teaching me English in Nanjing Institute of Meteorology, .

Finally, I would like to thank my wife (Janet) Qing-fang Guo and my son (Mike) Meng Dai, who began to understand what I have been doing for these years for a living. I apologize to them for my being so often missing in their daily life, and dedicate this work to them.

IV VITA

July 24, 19 5 0...... Born - Nanjing, Jiangsu Province, China

1982 ...... B.Sc., Nanjing Institute of Meteorology, China

1982-1984 ...... Instructor of English, Nanjing Instit'jte of Meteorology

1985-198 6 ...... instructor of Chinese, The U.S.-Sino Friendship Association,Syracuse,NewYork,USA;Graduate Teaching Assistant, Syracuse University

1986 ...... M.A., Syracuse University

1986-1988 ...... Graduate Research and Administrative Associate, TheOhioStateUniversity, Columbus, Ohio, USA

1988 ...... M.A., The Ohio State University

1988-1991 ...... Graduate Teaching Associate, The Ohio State University

PUBUCATIONS

The Head in M/O PAO DE KUAI. Journal of Chinese Linguistics 20:1. 84-119. (1992)

Rethinking Case Theory for Constituency and Word Order in Chinese. Journal of the Teachers Association 27:1. (1992)

Derivational-Inflectional Disitinction and Levels in Chinese Morphology. Chicago Linguistic Society 28:2. (1992)

Fundamental Concepts in Chinese Morphology, international Symposium on Chinese Languages and Linguistics 3. National Tsing Hua University. (1992)

The Resultative DE as an Inflectional Morpheme in Chinese. Eastern States Conference on Linguistics 7. 67-78. (1991)

The Negator BU and a Morphosyntactic Analysis of A-not-A Questions in Chinese. Chicago Linguistic Society 27:2. (1991) Historical Morphoiogization of Syntactic Structures: Evidence from Derived Verbs in Chinese. Diachronica Vll:1. 9-47. (1990)

Some Issues on A-not-A Questions in Chinese. Journal of Chinese Linguistics 18:2. 285-316. (1990)

Syntactic Constructions in Serial Verb Expressions in Chinese. The Ohio State University Working Papers in Linguistics 39. 316-339. (1990)

FIELDS OF STUDY

Major Field: Linguistics

Studies in Syntax, Morphology, Historical Linguistics, Chinese Linguistics, English Linguistics, English as a Second Language

VI TABLES OF CONTENTS

ACKNOWLEDGMENTS ...... ii

VITA ...... V

CHAPTER PAGE

I. INTRODUCTION ...... 1

1.0. Three Issues Surrounding the Components of Grammar ...... 1 1.1. Misconceptions about Chinese and Reconceptualization of Chinese Morphology ...... 3 1.2. Statement of Goal ...... 6 1.3. The Approach Taken; Theoretical Framework ...... 7 1.4. The Organization of the Dissertation ...... 10

II. CONCEPTS IN CHINESE MORPHOLOGY ...... 13

2.0. Introduction ...... 13 2.1. The Notion of Word in Natural Languages ...... 14 2.2. The Notion of Word in Chinese ...... 16 2.2.1. Previous Research ...... 17 2.2.2. The Syntactic Word ...... 20 2.2.3. The Phonological Word ...... 23 2.2.4. The Morphological Word ...... 26 2.3. A Methodology for Locating Word and Morpheme Boundaries ...... 28 2.3.1. Method 1 : Syntactic Independence ...... 29 2.3.2. Method 2: Movement ...... 30 2.3.3. Method 3: Pronominalization ...... 30 2.3.4. Method 4; Slot in Syntactic Constructions ...... 31 ir r*. — —. % : « ..W .w r . I «...... O l 2.3.6. Method 6: Modification ...... 32 2.3.7. Method 7: Expansion ...... 32 2.3.8. Problems of Coordination ...... 37 2.3.9. Summary ...... 38 2.4. The Division between Syntax and Morphology ...... 39 2.4.1. Bound Morphemes within Words ...... 39 2.4.2. Bound Stems ...... 40 2.4.3. Bound Stems and the Delimitation of Morpheme Boundaries ...... 41 2.4.4. The Syntax-Morphology Division ...... 45 2.5. A Fragment of Chinese ...... 56

VII III. COMPOUNDING ...... 64

3.0. Introduction ...... 64 3.1. A Review of the Literature and Issues in Chinese Compounding ...... 67 3.2. Compound Like Bimorphemic Lexemes ...... 75 3.3. The Lexicalization of Some Verb-Object (VO) Structures ...... 77 3.4. The Resultative Compounds ...... 92 3.5. N+N and A+N Compoundings ...... 99 3.5.1. Compounding of N+N ...... 101 3.5.2. Compounding of A+N ...... 106 3.6. Coordination in Compound Morphology ...... 109 3.7. Summary of Theoretical Positions ...... 122

IV. DERIVATIONAL MORPHOLOGY ...... 126

4.0. Introduction ...... 126 4.1. Reduplication ...... 127 4.1.1. Delimitative of Volitional Verbs ...... 127 4.1.2. Vividative of Adjectives ...... 137 4.1.3. Distributive Measures ...... 141 4.2. Affixation ...... 145 4.3. Summary ...... 150

V. INFLECTIONAL MORPHOLOGY: CONCEPTS ...... 152

5.0. Introduction ...... 152 5.1. The Division between Derivational Morphology and Inflectional Morphology .. 153 5.2. The Measurative TA as an Inflectional Morpheme ...... 160 5.2.1. TA as an Affix ...... 163 5.2.2. TA as an Inflectional Morpheme ...... 165 5.3. The Resultative DE as an Inflectional Morpheme ...... 170 5.3.1. DE as an Affix ...... 172 5.3.2. DE as an Inflectional Morpheme ...... 162 5.4. The Aspect Markers as Inflectional Morphemes ...... 188 5.5. The Deverbal Nominal-fou as an Infectional Morpheme ...... 197 5.5.1. The Deservative-fou as a Suffix ...... 198 5.5.2. Morphosyntactic Government:-fou as an Inflectional Morpheme ...... 202 5.5.3. -Tou as a Deverbal Nominal Inflection ...... 205 5.5.4. Grammatical Relations and the Bare Deverbal ...... 207 5.6. Summary ...... 2 n

VI. INFLECTIONAL MORPHOLOGY: CONSEQUENCES ...... 212

6.0. Introduction ...... 212 6.1. The Head in the Resultative Construction ...... 215 6.2. The Chinese Phrase Structure Condition as an Artifact ...... 217 6.3. Rethinking Case Theory for Chinese Word Order ...... 218 6.4. Head-Marking vs. Dependent-Marking Grammar: A Correlating Factor to Chinese Word O rder ...... 224 6.5. An Inflectional Analysis of the Reduced A-not-A Questions ...... 231 6.5.1. Against Phonological Reduplication ...... 232 6.5.2. Against Morphological Reduplication ...... 234

v i i i 6.5.3. An Inflectional Analysis ...... 241 6.5.4. Consequences ...... 246 6.6. The Deiivatlon-lnflection Distinction as Obscured in Lexical Morphology ...... 254 6.6.1. The Issues ...... 255 6.6.2. Derivations Occurring at the Last Level and Inflections at the Previous Levels...... 258 6.6 3. The Issue of Productivity ...... 260 6.6.4. Theoretical Implications ...... 261 6.7. Summary ...... 265

VII. CLITICS...... 266

7.0. Introduction ...... 266 7.1. Sentential Particles as Words but Not Clitics ...... 268 7.2. Discussions ...... 277

VIII. CONCLUSIONS...... 279

8.0. Introduction ...... 279 8.1. Word, Affix and Clitic ...... 279 8.2. The Methodology of Identifying Morpheme Boundaries within a Word ...... 280 8.3. The Distinction between Syntax and Morphology ...... 280 8.4. Compounding and Syntax-Morphology Interface ...... 281 8.5. The Distinction between Derivational and Inflectional Morphology ...... 282 8.6. Chinese Inflectional Morphology and Its Theoretical C onsequences ...... 283 8.7. Concluding Remarks ...... 284

BiBLIOGRAPHY...... 286

IX CHINESE MORPHOLOGY AND ITS INTERFACE WITH THE SYNTAX

By

Xiang-ling Dai, Ph.D

The Ohio State University, 1992

Professor Arnold M. Zwicky, Advisor

This work studies the word structure of Chinese to justify three distinctions and posit them as language universels; syntax vs. morphology, compounds vs. phrases, and derivational morphology vs. inflectional morphology. Chinese is chosen because linguistic tradition holds that the language has little morphology except for compounding, that the internal structure of its compounds is simply a reflection of its phrasal syntax, and that Chinese has no inflection.

With universal notions developed for affixes, words, com pounds and clitics, an investigation of compounding, derivation and inflection in Chinese is conducted. This study shows that principles governing word structure and phrase structure in Chinese are distinct, with respect to alternative ordering of constituents, the ability to change lexical categories, reference to phonology, acceptance of optional dependents, suppression of constituents, and tolerance of exceptions to rules.

While the verb-object and subject-predicate "compounds' are shown to be non-compound words, some "NPs" are demonstrated to be compounds, on a par with the resultative compounds and coordinative compounds. A compound behaves differently from a phrase in that a compound may exhibit lexical, semantic or phonological idiosyncrasies and its constituents observe lexical integrity. 2

The measurative -ta, resultative -de, deservative -tou, and three aspect suffixes are argued to be inflectional morphemes, with -tou a morphosyntactically governed form similar to case- marking in European languages. These inflections bear on several unsolved theoretical issues in Chinese syntax: (1) The verb inflected with -de is demonstrably the head of the resultative construction, prohibiting its subcategorized complement from co-occurring with an NP object. (2)

Inflectional morphology does not support the C ase account of word order in Chinese as it does in European languages, because NP-fou shows no morphological case alternations in different distributions; instead, the exceptional head-initial word order in Chinese correlates with a morphological head-marking pattern within verbal subcategorizations. (3) An inflectional truncation analysis of reduced A-not-A questions is shown to be superior to various reduplication analyses. (4) It is argued that the theory of Lexical Morphology cannot capture the derivational- inflectional distinction in Chinese, since it unavoidably causes an ordering paradox. CHAPTER I

INTRODUCTION

1.0. Three Issues Surrounding the Conri|3onents of Grammar

Does a grammar of language consist of distinct modules, referred to here as components

(and subcomponents), or do a set of general principles govern the operations of all aspects of language (sound, meaning, and mechnisms that associate sound with meaning at various levels)?

A generally accepted modular view of grammar, as expressed implicitly in Chomsky (1965,1970) and explicitly in Zwicky (1984:365), assum es that a grammar is com posed of a number of autonomous components (syntax, phonology, morphology, etc.). These components are distinct from and independent of each other, interacting with one another in limited ways. Based on this assumption, some linguists struggle to understand the nature of and constraints on the interaction of these grammatical com ponents in the description of language. For example, Zwicky & Pullum

(1986:63-91) argue for a distinction between syntax and phonology, positing the Principle of

Phonology-Free Syntax, which states that phonology plays no part in the operation of a syntactic rule. Anderson (1988:148-50) provides reasons as to why morphology is distinct from syntax on the one hand, and from phonology on the other hand.

However, as noted by Zwicky & Pullum (1986:86), this modular view of grammar has been challenged on many occasions by linguists from different theoretical frameworks, who have explicitly or implicitly assum ed a merger of one of the (sub)components with another. For example, the syntax-phonology distinction has been rejected in Baker & Brame (1972:54), Hall &

Hall (1976:115), Hudson (1976:115), Awbery (1975:24), Tegey (1975:571), Rivero & Walker (1976),

R. Lakoff (1974:40) and Traugott (1977:90).

1 2

Similarly, the assumption of a syntax morphology distinction has been abandoned or

relaxed to varying degrees by Fabb (1984), Hoeksema (1984), Marantz (1984), Sproat (1985) and

Baker (1987), bringing in one of the hotly disputed issues in the past decade as to whether a distinction should be made between syntax and morphology in universal grammar. Zwicky

(1989a:159,1992a:348-56) is largely a negative reply to the rejection or relaxation of the syntax-

morphology distinction. Put differently, the question is whether syntax (plus phonology) alone is

both necessary and sufficient for the description of the word structure of human languages, or

given that the syntax is justified, whether the morphology is indeed needed as an independent

component in universal grammar.

A related issue questions whether a distinction should be made between compounds and

phrases. For example, Anderson (1988:188) suggests that the assimilation of compounding to

other mechanisms of word formation is exaggerated and that the formation of compounds seems

to involve a genuinely syntactic combination of lexical items below the level of the word. In other

words, the inquiry is made as to whether syntax, without morphology, can independently describe

compounds in addition to phrases.

For linguists who recognize morphology as an autonomous component, a third issue asks

whether a distinction should be further made between derivational morphology and inflectional

morphology. Although this distinction is traditional, a number of proposals deny explicitly that

there is any significant difference between inflectional and derivational morphology. The

morphological structure presented in Lapointe (1980), for example, assigns information about both

inflected and derived forms to the lexicon. Moreover, the literature on the theory of Lexical

Morphology and Phonology, following Lieber (1983), generally assumes that both inflection and

derivation are 'in the lexicon", based on the obsenration that both invoke the sam e set of formal

operations. Each of the proposals, though, is criticized by Anderson (1988:169-71).

The purpose of this work is to study in depth the word structure of

’Also called , the grammar of which is based on the Beijing dialect 3

(or Chinese morphology) in order to justify these three distinctions as universels of grammar: syntax vs. morphology, compounds vs. phrases, and dervational vs. inflectional morphology.

1.1. Misconceptions about Chinese and Reconceptualization of Chinese Morphology

Chinese, with its long history and large population of speakers, seem s to be one of the most throroughly studied natural languages. Grammarians have long taken for granted that the language has some properties that set it apart from Indo-European and other languages.

Chinese is the object language in this study because the tradition, among both Western linguists and Sinologists, holds that Chinese has little or no morphology of any kind, except for compounding (e.g., from Sapir 1921 up to Jensen 1990 and S-K. Zhang 1957). Therefore, not surprisingly, Chinese has been cited in the linguistic literature as among the most isolating or analytical languages of the world, i.e., with a putative one-to-one mapping between morphemes and words.

As pointed out by Kratochvil (1968:73), traditional grammarians further believe that the internal structure of compounds in Chinese is simply a reflection of the structure of its phrasal syntax. For example, the so-called resultative (R) compound kan-DE-qing [see-R-clear] ‘see clearly' and the resultative phrase kan-DE hen qingchu [see R very clear] see very clearly’ appear to assume the same syntactic structure, including syntactic category (verb for kan and adjective for qing(chu)), subcategory {kan as a transitive verb taking a resultative complement, and (hen) qing(chu) as an adjective phrase), ordering (V preceding AP), and the form (the resulative DE suffxed on the head).

Finally, Chinese is widely assum ed to have no inflectional morphology at all (e.g., Hockett

1958:210, Li & Thompson 1981:134, Zwicky 1987:642, Sun & Cole 1991:44).

If these conceptions about Chinese were true, then the grammar of Chinese would not need a morphological component, let alone inflectional morphology: and Chinese syntax (plus phonology) would be both necessary and sufficient for describing not only its phrasal structure 4 but its word structure as well. As a net result, one could claim that the distinctions between syntax and morphology, between compounds and phrases, and between derivation and Inflection, are at least not absolute language universels.

However, linguistic study on a well-known language still yields surprises. The discovery of indirect and subtle evidence sometimes leads to an unexpected analysis.^ A profound

Investigation of Chinese word structure can demonstrate that the putative poverty of Chinese morphology Is simply an Illusion. For example, ShIh (1986:141), citing Zhu (1956), argues that bisyllabic adjective (A) and noun (N) combinations are morphological compounds rather than, as often assumed, phrases, because they have typical lexical correlates which are not true of the corresponding phrasal category. In particular, they have arbitrary gaps. For instance, while zang- shou ‘dirty hand’ Is grammatical, *zang-tang ‘dirty candy’ Is hardly acceptable. But with Insertion of the semantically dummy phrasal modifier marker de, the resulting zhang de tang ‘dirty candy"

Is grammatical. In addition, A-hN bisyilables (e.g., bai-zhi ‘white paper’) tolerate a contrary syntactic phrasal modifier {hui de ‘grayish’), as in hui de beû-zhi ‘grayish white-paper"; but in the

presence of the phrasal marker de between A and N, the string Is unacceptable: *hui de bai de zhi ‘îgrayish white paper’. This suggests that Chinese morphological compounding does not

entirely mimic the behavior of Its phrasal syntax.

More significantly. Dal (1991a:67 & 1992a:98) advances several arguments for the claim

that the resultative (R) marker DE is an Inflectional verbal suffix, because, among other

morphological characteristics, the syntax has to allow the V-DE to take a resultative complement

(VP or AP) In (1), but disallow It In other constructions, as when followed by a prepositional

phrase, even though the PP Is semantically a resultative complement, as In (2).

^For Instance, Zwicky & Puiium (1983:502) present six lines of evidence for taking the English - n't In He isn't going to school as an Inflectional morpheme, rather than a clitic as commonly assumed. They also cite Mallng’s (1983) evidence for near in near the wall as an adjective taking an NP complement. Instead of a preposition, and for like and worth as prepositions Instead of adjectives. (1) Ta pao-DE kiiai= (cf. *Ta pao kuai.)

he run DE fast

He runs fast.’

(2) *Ta tiao-DE zai ma shang. (cf. Ta tiao zai ma shang.)

he jump DE at horse on

He jumped down onto the horse."

Based on this and other evidence, one can conclude that DE is not a particle word or an enclitic as commonly assumed.

What I call the deservative -tou in (3) has never attracted linguists’ serious attention. As

I will detail in this study, not only is -tou an inflectional suffix attaching to deverbal nominals, but

also the fou-form is morphosyntactically governed by the main verb you, on a par with the well-

known facts of morphosyntactic government in morphological case markings in German and a

vast number of other languages. This is because the syntax requires that -tou co-occur with the

deservative verb you ‘worth’. Such co-occurrence entails that the main verb you is also obligatory

in the construction, as shown in (4), where the ungrammaticality arises when even a synonymous

verb zhide ‘worth’ replaces you. In the absence of -tou, however, you is not required; in fact, it

is expressly forbidden, as illustrated in (5).

(3) Zhe dianying you kan-tou.

this movie worth see deservative

‘This movie is worth seeing.’

(4) *Zhe dianying (zhide) kan-tou.

this movie (worth) have see deservative (5) Zhe dianying zhide kan (*-tou).

this movie worth see (deservative)

This movie is worth seeing.’

These two sets of examples, together with numerous phenomena to t>e discussed in this work, further suggest that Chinese possesses more interesting morp>hology, including inflection, than traditionally believed. It is hoped that research on the morphology of such a highly analytical language will contribute to a better understanding of the organization of the grammar of human languages in general and the three distinctions in modular grammar in particular. This is ultimate goal of this dissertation.

1.2. Statement of Goal

The goal of this thesis is to clarify basic concepts of the morphology of Chinese and to explore the consequences of certain reconceptualizations. Descriptively, I will systematically investigate compounding, derivational morphology and inflectional morphology in the language.

As previous research on Chinese linguistics has focused almost exclusively on the grammatical components of either syntax or phonology, I hope to show that solutions to some of the hotly debated issues in Chinese syntax follow naturally from the recognition of morphology in the language and of its interface with the syntax.

The consequences of this study are not limited to Chinese, but can contribute to general

linguistics as well. First of all, the predominant view that Chinese has meager morphology will be

rejected in the course of demonstrating that Chinese possesses most or all of the varieties of

morphology in universal grammar - derivation and inflection in addition to compounding, like a

"standard aveiage European" language such as English.

The existence of morphology in Chinese necessitates rejecting a popular theoretical

position that there is no division between syntax and morphology in grammar, as well as the view 7 of morphology as "the syntax of words" (e.g., Selkirk 1982^, Lieber 1992). A highly analytical

language perhaps should be cited as most convincing evidence in support of a single unified

morpho-syntactic component in a grammar. However, linguistic facts from Chinese will support

instead a division between word grammar and phrasal grammar even for an ostensibly highly

analytical language.

The syntax-morphology division nevertheless does not necessarily deny the existence of

a relationship between the two components of grammar, and this thesis argues for certain

principles of the interaction, or more technically, the interface, between syntax and morphology

as a language universal.

I will also consider the following much-disputed and unsolved problems in the literature

on Chinese linguistics, most of these assumed to be syntactic in nature. Does there exist a

Chinese Phrase Structure Condition which prohibits a verb from being followed by two

constituents (C-T. Huang 1982, 1984); does the Case account of Chinese word order have

morphological support as it does for other languages (Y-H. U 1990, Dai 1992b); do som e A-not-A

questions delete parts of a word (C-T. Huang 1989, Dai 1990b); which verb is the main verb of

the resultative construction (C-T. Huang 1988, C-R. Huang 1990, Dai 1992a); and is there a

separation of the parts of a compound (Chao 1968, C-T. Huang 1984)? It will be argued that

these puzzles can be solved most satisfactorily in the domain of morphology and its interface with

syntax, especially after a recognition of inflectional morphology in the language. This research

will also bear on the notions of affix, word, and clitic, and the theory of Lexical Morphology and

Phonology, as they apply to Chinese.

1.3. The Approach Taken: Theoretical Framework

Below, I sketch some important assumptions and concepts of morphology and syntax

^Although Selkirk apparently recognizes that syntax and morphology are separate components of grammar, she assumes that tho major mechanism for describing syntactic structures is both necessary and sufficient for describing morphological structures. 8 adopted in my theoretical framework to describe the structure of Chinese and other languages; all of them will be further discussed in the following chapters.

The modular view of grammar leads to two restrictions which are particularly relevant to this study. One is the Lexical Integrity Hypothesis (henceforth, UN, Jackendoff 1972), an oversimplified version of which states that phrasal rules cannot refer to parts of a word. The UH is also known as the Lexicalist position (Chomsky 1970), the Strong Lexicaiist Hypothesis (see the summary in Scalise 1984:101-2), Atomicity (Di Sciuilo & Williams 1987:ch.3), and the Principle of

Morphology-Free Syntax (PMFS) (Zwicky 1987a:654). The PMFS is a more elaborated version of the UH, stating that syntax is blind to morphology, i.e., syntactic rules cannot make reference to the internal morphological composition of words or to the particular rules involved in their morphological description. The only properties of a lexeme accessible to syntax are its syntactic category, subcategory, and morphosyntactic properties, such as gender, number, case, person, etc. (Zwicky 1989a: 161).

The other restriction is the Principle of Phonology-Free Syntax, stating that the syntactic module has no recourse to phonology, and phonological information plays no part in the operation of a syntactic rule (Zwicky 1984:365).

Modularity entails a syntax-morphology division, running counter to the notion of morphology as th e syntax of words' (Selkirk 1982, Ueber 1992). Anderson (1988) and Zwicky

(1992a) have presented convincing evidence for such a division: principles governing word- structure and phrase-structure are distinct, with respect to, inter alia, the order of constituents, the formal operations of rules, locality conditions, reference to phonology, and agreement between heads and their arguments.

Morphology is assumed to be the study of the structure of the word, but there is no unified notion of word consistent with all components of grammar. Words will be viewed here as certain domains in which rules of different grammatical components may apply (Dai 1990d:333), and at least three distinct notions should be identified in grammar: syntactic word, phonological 9 word and morphological word. A modular grammar entails that the three notions of word are independent of one another, and therefore it is important to note that the matching among the three in a linguistic expression may not be perfect, even though the default relationship among them is one-to-one mapping (Lyons 1968:69, Anderson 1985:152, Zwicky 1990:201).

Anderson (1982:587,1988:168) and Zwicky (1987a:655,1989a:144) adopt a fundamental theoretical assumption that derivational morphology and inflectional morphology constitute separate (sub)components of grammar. Inflectional morphology consists exactly of those aspects of word structure that are syntactically relevant, i.e., determined by or accessible to syntactic rules.^ Derivational morphology, on the other hand, relates the stems of different lexemes, in contrast to inflectional morphology, which relates the stem of one lexeme to its forms.

Hockett (1954) raises the issue of the choice between two competing approaches to morphological description: Item-and-Arrangement and Item-and-Process. There is no a priori reason to expect affixation to be the only operation in morphology, though it is the predominant one. Since the morphology of natural languages does employ, in addition to affixation, operations like alternation of vowels or consonants, interdigitation, reduplication, metathesis and subtraction,

I adopt a processual morphology rather than a morphology based strictly on concatenation, foliowing the lead of Matthews (1965,1972), Aronoff (1976), Anderson (1977,1982, 1984,1988),

Janda (1983), Hoeksema & Janda (1988), and Zwicky (1985b, 1988). I also maintain, with Zwicky

(1988), that morphological rules are distinguished from, and select, morphological operations listed in the inventory of universal grammar.

Morphological rules in this thesis are viewed as redundancy predictors rather than generators. This view presupposes a static condition but not a derivational analysis in

‘’Syntactic relevance of inflectional morphology does not violate the PMFS. A syntactic construction may require specific features for one or more of its immediate constituents. This feature is sometimes realized as a certain morphological form (e.g., the past participle) of the head word of a constituent. However, this morphosyntactic property cannot refer to particular morphological operations involved, such as suffixation, prefixation, or the shapes of -ed, -en, or suppletion {have worked/known/gone). 10 morphology, and a full lexicon, including both the list of lexemes and the set of morphological rules which together provide an account of the fuil inventory of lexemes (Zwicky 1992a;339).^

There are thus no Word-Formation Rules (WFR) in Aronoff s (1976:21) sense. A WFR is replaced by a (static) derivational^ rule which predicts and describes the relationship between sources and derivatives. This view reflects AronofTs (1976:31) distinction between WFR and analysis of existing words. For him, the two matters are similar, and yet different, since already existing words tend to kre idiosyncratic, and resistant to any system which derives their properties by general WFR rules. Rules which relate derivatwes to sources for deriving a new word can be treated as redundancy rules (Jackendoff 1975) analyzing a word into morphological constituents

(Aronoff 1976:30).

As for the syntax of grammar, although a nontransformational monostratal framework, such as Generalized Phrase Structure Grammar (GPSG, as presented in Gazdar et al. (1985)),

is demonstrably more restrictive, classical transformational terms such as movement, deletion,

pronominalization, etc., will be frequently used for ease of exposition.

1.4. The Organization of the Dissertation

The contents of this dissertation are arranged as follows. Chapter 2 establishes the object

of interest, i.e., the motivation of morphology in Chinese. I elaborate the notions of word, develop

^ h c rôCürôiïci icSô ô* âùi nc i i iûi pl iûlûÿiûci! i ulcS i iiây yivc i iScà iOicA icui i u i ii lUl liit; s L e . 11 lib goes against a popular generative view of the lexicon as minimal and redundancy-free (so that morphological rules can generate what is predictable within the lexicon). Given a unification of morphology and lexicon (Zwicky 1992a:339), the lexicon is simply the domain of linguistic expression types that morphology concerns itself with. Thus the lexicon contains both finite morphological rules and a list of all lexemes the rules describe. There can be no issue of the choice between the rules or the lexeme list. For theoretical and metatheoretical considerations, this thesis adopts a static framework, and it is not a concern here as to whether operations of the lexicon/morphology coincide with the language acquisition device, or what a language learner has in mind, in the generative sense.

^ h e words derive, deriving, or derivation{a[j should be understood as static or nontransformational in this thesis. They do not imply generation but rather describe generalizations about the relationships between lexemes. But for convenience, I will continue to use these words. 11 a methodology for locating word boundaries in a phrase vs. morpheme t)Oundaries in a word, and justify keeping the distinction between syntax and morphology, with presentations of some word grammar and a fragment of the language.

In Chapter 3 ,1 investigate Chinese compounding, justifying a notion of compounding and analyzing a large portion of historical compounds which lack synchronic justification. Notions of vertj-object compounds are critically reviewed with respect to the (in)separability of their parts.

Productive and synchronic compoundings, e.g., the resultative and the nominal compounds, are introduced. Universal consequences are outlined in terms of the syntax-morphology division, interface, and co-determination with respect to compounds. Of particular interest here is the constraint on syntactic coordination in compound morphology.

Chapter 4 addresses Chinese derivational morphology. Reduplication and its constraints will be discussed, with an eye to whether the rules invoking reduplication are syntactic or morphological, and if morphological, whether they are syntactically relevant. I reanalyze some stem s as affixes and thus expand the family of affixes in the language, which in turn opens the way for recognizing a much wider range of affixes and bound stems and a broader role for them

in the grammar of Chinese.

Chapter 5 presents the inflectional morphology of Chinese. Arguments are advanced for the claim that the measurative TA, the resultative DE, the aspect markers -le, -zhe and -guo, and the deservative -tou are inflectional morphemes in Chinese.

Empirical, methodological and theoretical consequences of these instances of inflectional

morphology are presented in Chapter 6. I demonstrate that the recognition of inflectional

morphology sheds light on such hotly debated issues in Chinese linguistics as the wordhood test

for verbal expressions, the justification of parts of speech, the notion of head in the resultative

construction, the constraint on Chinese phrase structure, the relation behween Case theory and

word order, the analysis of reduced A-not-A questions, and the arguments against an analysis

of Chinese word structure in the framework of Lexical Morphology and Phonology. 12

Chapter 7 deals with the notion of clitics. A summary of the tfieoretical positions and conclusions from this thesis is given in Chapter 8. CHAPTER II

CONCEPTS IN CHINESE MORPHOLOGY

2.0. Introduction

This chapter attempts to justify the object of interest here - the word grammar, i.e., morphology, of Chinese. As mentioned in 1.1, this language has been widely assumed to have little or no morphological complexity. Textbooks written by modem linguists, e.g., Hockett

(1958:181), Lyons (1968:188), Bauer (1988:167), and Norman (1988:10), to mention just a few, continue to cite Chinese as a standard example of an 'analytical" or "isolating" language, in which word is coterminous with morpheme. Works by Lu (1964) and Kratochivil (1963:ch.3), who seem to have been the first pioneers to deal seriously with the morphology of the language, have been for the most part unacknowledged. Later reference grammars of Chinese by Chao (1968:ch.3 &

4) and Li & Thompson (1981:ch.3) present a fairly long list of affixation and compounding processes in the language. But despite the insights of these works, more empirical justification, conceptual clarification, and theoretical analysis are needed for Chinese to contribute properly to morphological theory. My previous work (see Bibliography) notes the state of research in this area and begins to explore compounding, derivational and inflectional morphology of Chinese.

But a systematic description and theoretical approach is still lacking.

Confirming the existence of morphology in Chinese requires a demonstration that the

grammar of Chinese, like that of other natural languages, needs to refer to the notion of word in

addition to phrase, clause and sentence; that a Chinese word may be further analyzed into

meaningful morphemes; and that these morphemes recur and are not themselves words allowing further analysis. Thus starting from the foregoing studies, 2.1 and 2.2 motivate the notion of word

13 14 cross-linguisticalty and for Chinese; 2.3 develops a methodology for differentiating word tx)undaries in a phrase or sentence from morpheme boundaries in a word; 2.4 argues for a distinction between syntax and morphology in the language; and 2.5 presents a fragment of

Chinese.

2.1. The Notion of Word in Natural Languages

The notion of word seem s to be well-motivated in natural languages. For Bloomfield

(1933:178), word is purely a syntactic notion, defined as a minimal freeform. I maintain, however, that the word is not a unified construct throughout grammar but is rather characterized as certain domains within which rules of different grammatical com ponents may apply (c.f. 1.3).

A word may be the object of syntax, thus the term SYNTAGMATIC unit (Sadock 1985, Di

Sciullo & Williams 1987, Zwicky 1990:201). I call it the syntactic word. Thus, cat’s in (1 b) contains two syntactic words, cat, which is the head N of the subject NP, a n d ’s, which is the auxiliary verb of the predicate VP, parallel to cat is in (la).

(1)a The cat is going to eat.

b. The cat’s going to eat.

A phonological word is a word-like PROSODIC domain in which certain automatic phonological rules^ may apply. These phonological rules are traditionally called internal sandhi

^They apply within prosodic (phonological) domains, subject only to phonological conditions (Nespor & Vogel 1986), e.g., flapping alveolar stops between a stressed and an unstressed vowels, or assimilation of nasals in place of articulation with the following consonant in English. By contrast, nonautomatic rules are morphologically or syntactically conditioned (e.g., the k-s alternation in electric and electricity, or the alternation of the indefinite article a-an in English). To me, nonautomatic rules are morphological rules resorting to phonological operations (cf. Joseph a Janda 1988, Zwicky 1988). 15 rules, as opposed to external (or phrasal) sandhi rules, which may apply across word boundaries.^ Therefore, cat’s in (1b) is one phonological word, evidenced by the applicability of a word-internal sandhi: /s/ in cat’s is voiceless, as opposed to Is/ in cat is, which is voiced. The

phonological rule of devoicing assimilation applies within a word, the same domain as the voicing alternations of the plural -s, possessive -'s, and third person singular present tense Note that cat’s as one phonological word corresponds to two syntactic words, overriding the default one-to- one mapping between them. In many other languages, a word is phonologically marked. For

example, a Polish word is penultimately accented; in Czech, the accent is on the word-initial syllable.^

A word may also be the object of morphology. A morphological definition of word can also be used for highly inflecting languages (Anderson 1985:153). The obligatory inflectional

morphemes always close the word, indicating the word class. In Latin, nouns are marked with case and number, and verbs with tense, mood and . For instance, in bono.s iibro:s ‘good

books’ (accusative plural), two morphological words could be defined: bono:s consists of the base bon and an ending of case, number and gender of adjectives; libro.s is composed of the base libr and an ending of case and number. The term word is thus used in the

morphological/PARADIGMATiC sense, parallel to the forms of a lexeme. I call it the morphological word. Thus for the remaider of this thesis I characterize three notions of word as follows (cf. Dai

1990d:333).

(2) A SYNTACTIC WORD (W) is a minimal constituent to which syntactic rules may refer; a

PHONOLOGICAL WORD is a certain prosodic domain in which internal phonological rules

^Thus the flapping rule is an internal sandhi, and the nasal place assimilation an external sandhi. For nonautomatic phonology, the k-s alternation {electric vs. electricity) applies word- internally, and the the article a-an alternation externally.

^However, phonological features are not necessarily distributed exclusively within prosodic domains. For instance, word accent in Polish also serves a morphological purpose, marking morphological words. 16

may apply (as opposed to external or phrasal sandhi rules); and a MORPHOLOGICAL

WORD (w) is a maximal domain in which morphological rules may apply.

Again, there is no reason to believe that the default one-to-one mapping is absolute between syntactic word, phonological word and morphological word. From now on, "word" will be used only in a generic or a loose sense, indicating the default convergence of the three notions of word. "Lexeme" is an abstract notion for a different form of an inflectional paradigm

(Lyons 1968:197). Thus dog and dogs are the same morphological word or lexeme. Technical terms syntactic word (abbreviated to W), morphological word (abbreviated to w) and phonological word are employed, as they apply to the descriptions of staictures.

The notions of W and w are of major relevance to the interface between syntax and morphology. It is regarding the paradigmatic notion w rather than the syntagmatic notion W that the lexical integrity of a word is respected by phrasal syntactic rules (cf. an oversimplified version of the UN: phrasal rules cannot refer to parts of a word, and an elatx>rated version in terms of the

Principle of Morphology-Free Syntax in 1.3). The morphological unit w does not in general tolerate syntactic operations on its components except for special cases such as compounding, cliticization or serialization where Ws occur within a w. In these cases, syntactic rules refer to W but not to phrases and sentences. Following Zwicky (1990:206), I maintain that there is no general principle of "W intergrity", on a par with w integrity. Just as a phrase may consist of other phrases, a W may be viewed as containing Ws and phrases, giving rise to the term "phrase-word".

When a phrase is placed in a postion which syntax requires to be a W, syntactic rules may refer to the internal structure within this phrase-W. There is no violation of lexical integrity in this case, since the phrase-W is not a w {ibid, and to be discussed shortly).

2.2. The Notion of Word in Chinese

The predominant view that Chinese has little or no word-internal structure seems to have 17 originated from what is called the Monosyllabic Myth of the language (DeFrancis 1984:ch.8), i.e., that word, morpheme, syllable and written character (a) are all the same in the language. This is presumably true of Classical Chinese. But because of the phonological changes of weakening and loss of coda stops and nasals leading to increasingly large numbers of homophones, the language has been developing quantities of bi- and poly syllabic expressions through compounding over its history (Chen & Wang 1975). In fact, it has been claimed that the majority of the lexical entries of the language are bisyliabic (Lü 1963). The central issue is whether these bisyllabic expressions should be analyzed as words or phrases in syntax, and heated debates and much confusion on the issue can be found in the literature. Whether or not Chinese has a structural level of word as do other languages becomes immediately relevant. Previous reseachers have motivated the notion from different angles, with limited success.

2.2.1. Previous Research

Kratochvil (1968:92-4) seem s to agree with Bloomfield’s definition. By reducing what

Kratochvil calls th e maximal expression"/mf/an tianqi hen hao ‘The weather is very nice today’ to its six corresponding "intermediate and minimal expressions", namely, jintian hen hao ‘Today it is very nice’, jintian hao Today it is nice’, tianqi hen hao ‘The weather is very nice’, tianqi hao ‘The weather is nice’, hen hao ‘It is very nice’, and hao ‘It is nice’, he argues that each of the units preceding hao in the reduced expressions is a word, the smallest unit which may function as an immediate constituent of a Chinese phrase. It is not clear however why those units preceding hao cannot possibly be phrases. As Zwicky (1990:205) shows, languages have "word-phrases", i.e., phrases that comprise only a single head word constituent, e.g., English NP girls, VP vanished,

AdvP quickly in Girls vanished quickly. Likewise, the bisyllabic jintian ‘today’ in jintian hao could well be phrasal, since it occupies the same position as other NPs do, e.g., Beijing de dongbu ‘the

eastern part of Beijing’; and tianqi ‘weather’ in tianqi hao fills the position that other NPs like

women de laoshi ‘our teachers’ can fill. In fact, the two words in question are analyzed as 18 phrases in most, if not all, current syntactic theories. Moreover, Kratochvil does not say why jintian or tianqi should be treated as minimal units in a phrase, as each syllable of these bisyllables is also a meaningful unit recurring elsewhere in the language.

Chao (1968:ch.3) adopts both phonological and syntactic approaches to the word. He concludes that tone (sandhi) and stress are relatively unimportant in marking syntactic words but potential pauses are good tests for word boundaries. Thus for him, if a speaker for any reason hesitates in the middle of a polysyllabic word, he does not continue from there, but repeats by starting from the beginning of the word; hence he introduces the notion of uninterruptability of the syntactic word (p. 154-5). I will evaluate the relative merits of this phonological approach shortly.

In his syntactic approach, Chao, citing Lu (1964), ioosens Bloomfield's definition by adding a methodological consideration: a word can fill the functional frame of a typical syntactic structure (Chao 1968:161, cf. 2.3.4). Dai (1990a:11-2) further motivates this approach, pointing out that Bloomfield’s definition is too drastic, since it is known that not every word can occur alone

(cf. English articles a, the, prepositions to, along etc, Chinese sentential particles ne, ma,

prepositions cong ‘from’, yan ‘along’, etc.). However, a free verb like zhidao ‘know’ can typically fill the transitive verb position in a veitr-object construction, as in zhidao zhe renlzhe shujzhe shi

‘know this person/this book/this thing", etc. Similarly, English articles and prepositions are words since they can occur in a certain position in NPs or PPs respectively. Any minimal units

corresponding in meaning to those words which cannot fill these syntactic frames are not words,

but bound morphemes. I should point out that while this method may effectively separate bound

morphemes from free words,"* it cannot differentiate t>etween words and phrases, as phrases

may also be positioned in a syntactic frame. For example, an NP, rather than a noun (N), typically

fills the subject position in natural languages. Worse still, it is possible to analyze a phrase as a

W, hence the term phrase-word (Bloomfield 1933, Zwicky 1990:206). In English, the phrasal

^his approach is employed by Dai (1990a) for the description of the historical morphologization of syntactic phrases in the language. 19 expression give Mary may be analyzed as a W (or a transitive verb) in give Mary a book, i.e., as is transitive verb discard in discard a book. For some speakers, give Mary can be conjoined with a VW, i.e., V or verbal W rank, as opposed to VP, as in /’// either give Maty or discard the book.

Similarly, in Chinese, the VP containing a verb plus a postveital measure phrase, ti-le yi jiao can be said to function syntactically as a word-like transitive verb ti-le taking the object (da-men), as in Ta [ti-le yi jiao] da-men [he kick perf. one foot big-door] ‘He kicked the gate once.’.

The notion of word is also motivated in Sheu (1990:15-31). She gives three reasons for positing the word level in Chinese grammar. First, she takes for granted that Chinese has no inflectional morphology and suggests that word equals lexeme. Second, she assum es Chao’s phonological pause test for syntactic wordhood. Third, drawing on Dowty’s (1978 & 1979) distinction between 'syntactic rules' and 'lexical rules', she t>elieves that a distinction between word and phrase in Chinese can explain why syntactic rules tend to be productive and semantically compositional, while lexical rules tend to be unproductive and semantically less compositional.

I have a few reservations for accepting her argumentation on empirical, methodological and theoretical grounds. First of all, whether or not Chinese has inflectional morphology at all poses both empirical and theoretical questions. If the inflectional morphology for Chinese I posit in chapter 5 can stand up to scrutiny, then word cannot be on a par with the notion of lexeme in Chinese, since the latter will be reserved as an abstract notion for the inflectional paradigms of verbs or nouns. Second, using the possibility of phonological pause as a test for syntactic wordhood is methodologically and empirically suspect. Syntactic wordhood should be motivated on syntactic rather than phonological grounds, because of mismatches like cat is and cat's, even though the default relationship t>etween syntactic and phonological words is one-to-one mapping.

Further, certain syntactic phrases cannot be phonologically interrupted either. At least for some syntactic combinations in Chinese (which will later be shown to be phrasal but not lexical), if the string is interrupted, the speaker must re-start from the very beginning of the phrase, e.g., the 20

adverb+adjective structures hen hao Very good’ and gang huai ‘still worse’, and the verb-object

structure da ta ‘beat him’ and hen tamen ‘hate them’, in which the object is a pronoun. I suspect the same is true of English phrases such as a boy, eat them up, and so on. The point here is that

phonological pause is most relevant in motivating phonological wordhood or prosodic affiliation,

but unreliable as a test for syntactic wordhood.

It is also problematic to use productivity in distinguishing word from phrase, as

productivity itself is a relative notion (Dai 1990a;28) in morphology. In addition, it is not always

clear how to draw the line between syntactic productivity and morphological or lexical

unproductivity. For example, idiom chunks are parasitic on syntactic constructions (Zwicky

1989b), but unproductive (cf. The cat was let out the bag/*sack Speakers - lose the idiomatic

reading when bag is replaced by the synonym sack.), while the derivational rules of nominal

gerunds from verbs (finding, going) and of adverbs from adjectives (happiiy, anxiousiy) are

productive.® In Chinese, the syntactic ba- and be/-constructions are not very productive because

of semantic and pragmatic restrictions (Li & Thompson 1981 :ch.15 & 16), while the morphological

rule of DE suffixation in the resultative construction is productive (to be discussed in chapter 5).

In summary, no previous study on Chinese grammar draws clear distinctions amongst the

different notions of word in syntax, phonology and morphology. Below, I present each of these

notions in the context of Chinese.

2.2.2. The Syntactic Word

As in other languages, the syntactic word in Chinese is a minimal constituent that

syntactic rules may refer to. This universal definition assumes a division between syntax, the

component of grammar dealing with structural levels at or above W (phrases, clauses and

sentences), and morphology, the component of grammar dealing with the internal structure of

words. Phasal syntax cannot look inside a w. For instance, such syntactic operations as

®There are still lexical gaps; *ajarly, *bluely, *smally, etc. (Pollard, p.c.). 21 movement/extraction, deletion/gapping, insertion/expansion, and pronominalization/replacement, to use transformational terms, do not apply within a w. In other words, morphological operations and rules are distinct from those in phrasal syntax. Below I introduce a syntactic rule in Chinese which calls for a lexical constituent, or W, just a s a phrase structure rule in English refers to W, e.g., in PP -> P NP {in the garden), P is lexical or W® and NP is phrasai.

As shown in Dai (1992a:91 -2), the Chinese verti-object (henceforth V0-) construction (VP -

> V NP), like its English counterpart {kissed the girl), requires its first immediate constituent (V) to be lexical (or precisely, a VW) rather than phrasal (VP).^ Although this structural requirement on the verbal lexical item (VW) seem s to have been taken for granted, I would still like to demonstrate that this is true, as illustrated in (3). The VO-construction (VP) contains the verb

(VW) xiu ‘repair’ and the object NP qiche ‘car’.

(3) Ta [vp[y xiu] [^p qiche]].

he repair car

‘He repairs cars.’

Example (4a) is an expansion of (3). It is obvious that the object is phrasal (NP), not only

“A pnrase-iike unit such as right in may apparently till the slot in PP > _ NP giving right in the garden so that the putative preposition position in PP does not have to be W. However, right in is not a constituent in the PP, as evidenced in the pronominalization of the PP, giving right there, which shows that in the garden is constituent independent of right. The phrase right in as an answer to the question Is the doctor in the garden? cannot be the evidence for the constituency for right in in the PP, since when right in is in isolation, in is either an adverb or an intransitive preposition, which is a different word from the transitive in. Therefore PP -> AdvP PP is the correct rule for right in the garden, but PP > PP NP is not.

^As a reminder, in English, the slot for VW in VP - > V NP may be filled by a VP or phrase- word, as in [gave Mary] a book and [sent away] the money (Zwicky 1990:206). The sam e is true of the corresponding slot in Chinese, since a V + measure-P (x/u san tian ‘repair three days’) can fill the slot. The phenomenon requires an adequate grammar to allow rank shift {ibid & 4.1.2), i.e., a phrase may function as a word, but it does not entail that VP -> V NP is not a rule in both languages. 22 because the adjectival/subordinative de® intervenes between women 'we' and qiche ‘car’, but also because the expanded NP object can move to sentence-initial position as a constituent, leaving the verb xiu "repair" behind, as in (4b). The preverbal adverbial, manman de, is taken as the modifier of the inner VP rather than of its head verb, because the inner VP can be also displaced as a constituent sentence-initially, as in (4c). This indicates that xiu is isolated to some degree from both the postverbal object NP and the preveital adverbial, consistent with its lexical assignment in (3).

(4)a. Ta [^p manman de [^p xiu] [^p wo men de qiche]]].

he slow de repair we de car

‘He repaired our car slowly.’

b. Women de qiche, ta manman de xiu.

c. Xiu women de qiche, ta zongshi [always] manman de.

Besides, the inner VP in (4a) participates in the deletion in (5a) and in the pronominalizations by gan or zuo ‘do’, independent of the preverbal adverbial in (5b).

(5)a. Zhangsan manman de xiu wo-men de qiche, Lisi daoshi hen kuai de.

Zhangsan slow de repair we de car Lisi but very fast de

‘Zhangsan repaired our car slowly but Lisi fast.’

b. Zhangsan manman de xiu wo men de qiche, Lisi daoshi hen kuai de gan de.

Zhangsan slow de repair we de car Lisi but very fast de do de

®The grammatical status of de will be discussed in chapter 5. 23

‘Zhangsan repaired our car slowly but Lisi did so fast.’

All these show that the VO-construction contains a lexical verb, a VW forming no immediate constituent with preverbal elements and being isolated somehow from postverbal elements. Other rules calling for syntactic words can be found throughout the syntax of Chinese (cf. 2.5).

2.2.3. The Phonological Word

The phonology of Chinese also utilizes the notion of word, which is a certain domain certain phonological rules may refer to, hence the term phonological word. Those phonological rules apply only within a word-like unit, as opposed to the rules which may apply across word

boundaries.® Note that the word-like unit here is a purely prosodic domain, which is independent of, but often coincides with, a syntactic or morphological word. An example of word internal sandhi in English is the devoicing assimilation of /-si with the preceding segment in the auxiliary

contraction of -’s, the plural -s, possessive -’s and third person singular present tense ^ (cf. 2.1).

Another is the deaspiration of a voiceless stop following /s/ within a word, as in the pairs pace- space, top-stop, and kill-skill. Besides, vowel harmony as a word internal rule can be found in

Turkish and Finnish (Nevis 1988 & references there).

For at least three rules in Chinese, previous studies fail to refer to the distinction between

automatic and nonautomatic phonology, and between word-internal and external domains.

Cheng (1973:34) posits a phonological rule Final Elision (FE), which optionally deletes the

rime of a second syllable and resyllabifies its bilabial nasal onset as the coda of the first syllable,

as in (6a-d).^° Dai (1990c & d:332) extends the application of the FE to all bilabial stops as the

®As mentioned 2.1, the two types of phonological rules are traditionally, and somewhat confusingly, called word internal sandhi vs. word external sandhi.

’°Cf. Chao (1968:254), which perhaps stems from the research of others. 24 onset of the second syllable, as in (6e/l), and argues explicitly that the FE is a word-internal rule,^^ for while the rule applies within a word in (6), it is blocked across a word boundary, as in (7), taken from Dai (1990c).

(6)a. wo-men - > worn b. ta-men - > tarn

I PL ‘we’ he PI ‘they’

c. shen-me -> shem d. zen-me -> zem

what f. ‘what’ how f. ‘how’ (f. = formative)

e. ba-ba -> bap f. jiu bu qiche -> jiup qiche

dad ‘dad’ nine Measure car ‘nine cars’

(7) Ta meng le tou. -> *Tam le tou.

he mask perf. head

‘He has masked his head.’

One crucial phonological condition on the application of FE is that the target syllable must be prosodically weak (unstressed and atonic). FE fails to apply in xiao-men {*xiaom) ‘school-gate’ or guang-bo (*guangp) ‘broadcast’, because the second syllables are strong (stressed and tonic).

Otherwise, FE would have to be a morphological rule trecause of lacking phonological generality.

This condition apparently obscures the word-internal nature of the rule, for in (7) meng is strong.

Since word-initial syllables, including prefixes, are generally strong in the language, FE cannot

’^Additional conditions must be placed on the application of the FE besides those obsarved in the traditional literature and in Dai (1990c). For example, FE does not seem to work if the vowel of the second syllable is a , nor if the second verb in the /a/-construction (to be discussed shortly) is bisyllabic. I won’t explore these conditions in detail, since they do not affect my argument here. 25 apply across word boundaries. After a phrasal intonation is assigned and a certain syntactic word such as the measure word bu in (6f) is pronounced as weak, FE applies across the syntactic word boundary to define a phonological word.^^

The extended FE is relevant in analyzing two constructions in the language, and also bears on whether sentential particles are clitics or not (to be discussed in chapter 7). In the reduced A-not-A q u e s tio n s ,th e FE optionally applies in xi bu xihuan. Therefore, xi bu is a phonological word, as in (8).

(8) xi bu xihuan? - > xip xihuan?

like not like ‘Do you like it or not?’

In what I call the serial verb construction, or the /a/-construction (Dai 1990d;332-3), for those instances where the onset of the second verb is a bilabial, FE may apply. This rule ignores the syntactic demarcation between the first verb (V1) and second verb (V2), and V1-FV2 instantiates a single phonological word, as in (9).

(9)a. qu bu yi-fu -> qup yi-fu b. qu bu yu -> qup yu

go mend clothes go catch fish

'go mend clothes’ go catch fish'

c. lai pu chuang -> laip chuang

come make bed

^ ^Readers may have noticed some circularity in characterizing phonological word. To avoid this, phonological word should be treated as a sort of primitive notion with respect to syntax and morphology.

^ ^ h e A-not-A question is one of the hotly-debated issues in Chinese syntax (for more discussion, see C-T. Huang 1989, Dai 1990b, c, 1991c, and ch.6 of this thesis). 26

‘come make bed’

The point of interest in (6f), (8) and (9) is that two syntactic words are pronounced as one

phonological word, and that the default one-to-one correspondence between syntactic word and

phonological word is overridden (cf. the English case in (1)).

The notion of phonological word is relevant in at least two more cases. One is the

diminutive Retroflex Suffixation (Cheng 1973:24 and references therein), which deletes the vowel

in the second syllable sr (the diminutive marker etymolcgically related to er ‘son’) in a compound

and resyllabifies -r as the coda of the first syllable {dao-er -> daor ‘knife’, pai-er -> per

’signboard ). The other case is that a nominal formative starting v.ith a , such as -zi, s i

or -fu can phonologicaliy attach to its root syllable with deletion of the vowel in the formative {er-zi

-> erz 'son', laoshi -> laosh ‘teacher’, shi-fu -> shif ’master’).A g ain , both of these word-

internal rules are similar to FE in that they apply optionally and require that their tragets be weak.

2.2.4. The Morphological Word

Recall that word boundaries in highly inflecting languages like Latin and Miwok can be

determined by locating obligatory inflectional morphemes which close a word.^® This

traditionally motivates a third notion of word, the morphological word (w), the maximal domain to

which morphological rules may refer. As w is the maximal domain for morphological purposes

in general, the notion should not be narrowly understood as one used exclusively by inflectional

morphology (or for paradigmatic purposes). For instance, parts of an uninflected word also

^'’Chao (1968:141) thinks that the vowel in question is not missing but is pronounced as voiceless: thus the formative is still syllabic. He also suspects that the rule is caused by the falling tone in preceding (root) syllable. However, i observe that tonality is irrelevant to many speakers. At this point I am not sure whether resyllabification or devoicing is closer to being representative and leave the burden of proof to the phoneticians. Fortunately, neither choice will affect the main concern here, namely, the rule is a word-internal one and cannot apply across word boundaries.

^ ^except for function words (often called "particles”) that do not inflect. 27 demonstrably observe the lexical integrity of the whole. Inflection and w are often mentioned together in introductory textbooks, since inflectional morphemes tend to close a word and be located at the edge of the domain; inflectional morphemes can thus be used as a rule of thumb to locate w boundaries and to define the maximal demain of that w.

In traditional grammar, "morphological word" would be an irrelavant notion to Chinese, since the language has been assumed to have little or no inflections. This view of Chinese I refute in chapter 5. Let us take for granted the conclusion there that aspect markers such as -le or -zhe in Chinese are inflectional morphemes which close a word (verb), although they are not obligatory in every construction. Now the notion of w becomes relevant in analyzing the lai- construction. The data below indicates that -/e can independently attach to the first verb (VI) lai or the second verb (V2) chang when either occurs alone, as in (1 Oa/b) respectively. When V1 and V2 are in series, however, only V2, but not V1, can be so suffixed, a s in (lOc/d). It follows from the notion of w that there is no morphological word boundary between VI and V2 in serialization and that V1+V2 forms a single morphological word.’®

(10)a Ta lai-le liangci.

he come Perf. twice

He cam e twice."

b. Ta chang-le liangci.

he sing Perf. twice

‘He sang twice."

c. Ta lai chang-le liangci.

’®As pointed out in Dai (1990d:332), V1 -r-V2 is not a compound. This is because no syntactic rules (including senring as a short answer to questions) may refer to this combination; it is not a minimal syntactic constituent or W. 28

he come sing Perf. twice

‘He came and sang twice.’

d. *Ta lai-le chang liangci.

he come Perf. sing twice

Here one morphological word instantiates two syntactic words.

2.3. A Methodology for Locating Word and Morpheme Boundaries

Having justified the notion of word in the different components of Chinese grammar, I turn now to the main object of this study; the morphological word. In most cases, a w matches a

The first thing is to design an effective and relatively easy method to differentiate W t)oundaries in a phrase from morpheme boundaries in a w, or, hopefully, to single out each W in syntax. Based on the concepts developed so far, it is safe to assume that if a (phrasal) syntactic

rule refers to a constituent, that constituent is not part of a w.^® However, one cannot further

state that if some syntactic rule cannot apply to a constituent, that constituent must t>e part of a w, or a bound morpheme. This is because a rule may be structure- or language-specific. In

order for a rule to apply, all of its preconditions, and not just structural levels and ranks alone,

must be satisfied. But one plausible way of identifying bound morphemes within a word is to try

applying some syntactic rules to them, i.e., the unexpected inapplicability of some syntactic rule

would suggest the possibility of their tjeing morphologically bound. The question is which of the

syntactic rules or operations should be tried, and how strong a claim can be made from each

’^Marked cases in which a w is a not W are the serial verb constructions in Chinese (Dai 19G0d) and in English, the portmanteau in French, and the clitic groups in English (Zwicky 1990).

^ ^Certain syntactic operations such as coordination and anaphora may addtionally refer to Ws in compounding (see ch.3). 29 syntactic test applied. I take the passing of one of the following tests as a sufficient condition on a free form/^ namely, the rank of the constituent in question is at or above, the structural level of W.'w, but not below.

2.3.1. Method 1: Syntactic Independence

If a constituent can stand alone, it may not be part of a w; hence the notion of syntactic independence of words (cf. Lyons 1968:203).

(11)a Do you like cars or wagons? - Cars.

b. Ni xihuan bu xihuan qiche? - Bu.

you like not like car no

'Do you like cars or not?' - 'No.'

The constituent cars in (11a) is not a bound form in the yes-no question, nor is the negator bu in the A-not-A question in (11b), because each can stand free as a short answer to some relevant question (Dai 1S91c). No stronger claim can be made here, though. On the one hand, (11) tells nothing about the rank of cars and bu, namely, whether they are Ws, phrases or clauses, since the latter two are also free forms. Intuitively, cars is a word-phrase, bu a W/w. On the other hand, particle lexemes such as prepositions/postpositions, articles, etc, cannot generally occur alone

in natural languages, but by other tests (to be discussed), they are free Ws.

’^As a reminder, a free form cannot necessarily occur alone. In addition, I will follow a traditional principle In this thesis, i.e., an item is considered free if it is sometimes free. For instance, go and black are free words in They go to school and The board is black respectively, but bound morphemes in They are going to school and The blackboard is there. But if an item is said to be bound, it is however always bound, e.g., -ing and -ness are always bound suffixes in English. 30

2.3.2. Method 2: Movement

If a constituent can be moved or displaced in a related construction, it cannot be bound as a part of a w (cf. positional movability of words (Lyons 1968:203)).

(12)a. I like pizza vs. Pizza, I like.

b. Wo da le ta. vs. Wo ba ta da le.

I beat-perf. he vs. ! BA he beat-perf.

I beat him.’

Here, pizza and the pronoun ta is not a bound morpheme. The movement test also has its limitations, since syntactic nonconstituents cannot be moved either, and not all syntactic consituents can be moved, i.e., movement rules are largely language- and construction-specific.

Such limitations are also true of Methods 3-5 below.

2.3.3. Method 3: Pronominalization

If a constituent can be replaced by a proform, it cannot be a part or a bound morpheme of a w.

(13)a. I like John/him.

b. Wo xihuan jiejie/ta.

I like elder-sister/she

"I like my elder sister.’ 31

John and jiejie are not bound morphemes, since each can be pronominalized.

2.3.4. Method 4: Slot in Syntactic Constructions

If a constituent can fill in a slot in a syntactic construction, it cannot be a bound morpheme of aw . This test is utilized in Dai (1990a: 12), where I dem onstrate that a tx)und form

{-xi ‘practice’) cannot syntactically function in the same way as its corresponding free word (xue-x/

‘learn’). For instance, on the assumption that a verb typically fills the VW slot in the VO- construction, the following example indicates that pro- and -xi are not Ws but parts of w’s.

(14)a He proclaimed/claimed/*pro-ed the victory.

b. Zhangsan xue-xi/xue/*xi ying-yu.

Zhangsan learn English

‘Zhangsan learns English.’

2.3.5. Method 5: Deletion

If a constituent is deleted or missing later in discourse, or in a non-coordinate structure

(see 2.3.8 for reasons), it cannot be a txaund morpheme of a w. Only a Chinese example is

provided for the deietion test, as English has a clear disinction between transitive and intransitive

verbs, and English can delete W/w’s within a compound {He likes the Chomsky theory, I like the

Bloomfieid.), whereas Chinese cannot.

(15) Zhangsan zhe ge ren hao he jiu, wo yi-dianr ye bu xihuan.

Zhangsan this M(easure) person love drink wine I a-bit not like

‘Zhangsan is an achoholic. I don’t like him at all.’ 32

Here Zhangsan as the object is missing in the second clause of the discourse, and it must be at or above word-rank in the first clause. For the same reason, numerals in the languages are assumed to t>e words, because at least one of them, yi ‘one’, is deletable even without any context: lai yi ge ren [come one M person] vs. lai ge ren ‘A person came.’.^

2.3.6. Method 6: Modification

If a constituent can be modified at the phrasal level, it cannot be a part of a w. For example, hong 'red' and black in (16) are not part of a w, since the adverbial modifications {hen

‘very’ and very) operate at the phrasal level in Chinese and English, By contrast, the corresponding constituents in (17) reject the same modifications, indicating that they are part of a w.

(16)a. hen hong de hua

very red de flower

‘very red flowers’

b. very black board

(17)a. *hen hong-hua

very red flower

b. *very blackboard

2.3.7. Method 7: Expansion

^°Whether the missing elements in these two cases are due to deletion or are simply phonetically unrealized is not the point at issue. What matters is that no null element can refer to a bound morpheme in a w. 33

If a constituent tolerates the insertion of a phrase (or a word functioning as a phrase - hence word phrase) on both its left and right side, it cannot be a bound morpheme of a w. The rationale here is that a phrase may be inserted between two W s but not between two morphemes within one w. This type of expansion is customarily appealed to in textbook discussions of criteria for (syntagmatic) word-hood (Bloomfield 1933, Lu 1964, Lyons 1968), and adopted in Dai (1991a

& b). At this point, I assum e that the inserted phrase must kre a dependent (i.e., either a modifier or argument) of the preceding constituent or the following constituent. The method has wide

empirical support. For content W/w (N, V, A, P) may allow optional phrasal dependents by default

in the syntax of natural languages,^^ as opposed to rules of movemenL deletion,

pronominalization, etc., which are usually marked by language- or construction-specific conditions,

although each of these is a sufficient, but not necessary, condition on wordhood.

I will illustrate this test with English compounds. The compound television tabie is one

W/w, for neither the dependents (arguments or modifers) of the first element, teievision, nor those

of the second, table, may interrupt the whole, as in *television from Japan table and *television big table (Zwicky 1990:212). The ungrammaticality of the former is caused by the intervention of

the postmodifier from Japan of the first component, the latter by the premodifier big of the second

component. Similarly, in- in incorrect is not a W/w because no W can be inserted on its right

side, e.g., *in-absolutely-correct (cf. absolutely incorrect), since affixes are proper parts of w, and

no Ws can separate an affix from its stem. However, in is a W/w in They sat in chairs because

some syntactic words, e.g., the postmodifier of sat, comfonable, or the premodifier of chairs, big,

can be inserted, giving They sat comfortably in big chairs. Hence neither sat in nor in chairs is

a W/w here.

Applying the expansion test to Chinese, I take ting ji in ting ji tian in (18a) as two words,

since either a postargument of ting, (i.e., ke), or a premodifer of ji (i.e., hao), or both can

Language- and structure-specific restrictions on expansion can still be found. For instance, in English, no word may intervene between the verb see and its pronominal object NP it in see it. 34 intervene, as in (18b/c/d).

(18)a Ting ji tian

listen severai-day

to listen several days’

b. Ting ke ji tian

listen class several day

to listen to lectures several days'

c. Ting hao ji tian

listen good several day

‘to listen for quite a few days’

d. Ting ke hao ji tian

listen class good several day

to listen to lectures for quite a few days’

In contrast, the three aspect markers perfective -le, durative -zhe, and experiential -guo are not words but morphemes bound to the stems they attach to, since they fail the test. For instance, ting-guo in ting-guo ji tian in (19a) is a w containing the stem ting and the suffix -guo, since the same postargument of ting cannot intervene, as in (19b), although the postargument, the

premodifier of ji tian, or both, can occur after -guo, as in (19c/d/e).

(19)a. Ting-guo ji tian

listen exp. several day 35

‘listened for several days’

b. *Tlng ke-guo ji tian

listen class exp. several day

c. Ting-guo ke ji tian

listen exp. class several day

‘listened to lectures for several days’

d. Ting-guo hao ji tian

listen exp. class several day

‘listened for quite a few days’

e. Ting-guo ke hao ji tian

listen exp. class several day

‘listened to lectures for quite a few days’

Similarly, the expansion test justifies the traditional assumption that the aspectual -le and -zhe are bound morphemes. The morpheme -le is independent of and distinguished from the sentence- final particle le, as in (20a). Le is a W/w, because a phrase can occur before it, a s in (20b), in constrast to -le in (20c). (The right expansion is irrelevant to the sentence-final position for le.)

(20)a. Wo chi le fan le.

I eat perf. rice le

I had a meal.’ 36

b. Wo chi le fan san xiaohsi le.

I eat perf. rice three hour le

‘It has been three hours since I had a meal.’

c. *Wo chi fan-le san xiaoshi (le).

The experiential -zhe also fails to pass the expansion test, as (21 b), which contrasts with another durativa marker zai in the preverbal position, as in (22a). Since zai can pass the tests of expansion on both sides, a s in (22b), it is a W/w.

(21 )a. Wo chi-zhe fan.

I eat dur. rice

I am having a meal.’

b. *Wo chi (dur.) fan-zhe.

(22)a. Wo zai chi fan.

i dur. eat rice

I am having a meal.’

b. Wo yong kuaizi zai hen kuai de chi fan.

I use chopstick dur. very fast eat rice

‘I am having a meal very fast with chopsticks.’

In the same way, the adverbs hen ‘very’, feichang ‘extraordinarily’, geng ‘even more’, tai

‘too’, ye also’, zui ’most’, etc. can be shown to be W/w’s, since the word bu not’ can intervene 37 between an adverb and the modified verbal expression following it: hen hao very good’ vs. hen bu hao ‘very bad’. The negator bu not’ is a W/w, since not only can it occur in isolation, but it can also pass the expansion test: bu du shu [not read book] ‘do not read books’ vs. bu ronzhen

(de) du shu [not serious (de) read book] do not read books seriously’. In addition, in one pattern of A-not-A questions, all the elements to the right of bu can be missing: Ta xihuan zhe ben shu buT [he like this M book not] ‘Does he like this book or not?’ (Dai 1990b:287). This clearly shows bu is an end-free item, to adopt Chao’s (1968:144) term. By my account, bu is a W/w, because the sister constituent immediately following bu is demonstrably phrasal but not lexical and can be missing. The same point can be made with respect to its immediate preceding constituent.

2.3.8. Problems of Coordination

Before concluding this section, I should show why coodination is an unreliable test for morphological wordhood. Recall that in an expansion test, the Ws inserted between A and B must be either the dependents (modifiers or arguments) of A or of B. This is because syntactic material coordinate to A or B may intervene in compounds (see ch.3) and even in some putative cases of affixation, as in (23a/b), but cf. (23c).

(23)a. television table: television and VCR table, television shelf and table

b. pro-democracy: pro- and anti-democracy, pro-democracy and capitalism

c. jumping: "jump- and running, ‘ jumped and -ing

It seems that in Chinese not all compounds allow their components to be conjoined with others’, as in (24a/b): and most of the affixes allow coordination of the hosts, giving rise to the notion of

■phrasal affix', as in (24c) (to be discussed in 3.6). 38

(24)a. huo-che [fire-vehicle] ‘train’, qi-che [vapor/gas-vehicle] ’car’

*huo-qi-che train and car’

b. Zhongguo-renmin [China-people] ‘Chinese people’

Chaoxian-renmin [Kcrea-peopie] ‘Korean people‘

Zhongguo (he [and]) Chaoxian renmin ‘Chinese and Korean people’

c. !acshi-men [teacher-plural (PL)] ‘teachers’

tongxue-men [student-PL] ‘students’

laoshi (he [and]) tongxue-men ‘teachers and students’

Looking back on the deletion test in 2.3.5, one may also wonder whether the result of (24b/c) is due to coordination, or to the deletion. Thus the deletion test should be administered in discourse or in a non-coordinate structure.

Since both compounding and affixation belong to morphology, the question here is why coordination, which is considered a syntactic operation, can affect certain domains of morphology.

For my purposes, it is more important to investigate to what degree coordination figures in morphology and to set universal constraints on its operation in this component, rather than taking its application in morphology as evidence for erasing the distinction between morphology and syntax. These theoretical issues are discussed in chapter 3.

2.3.9. Summary

Seven syntactic tests or methods have been introduced, each serving as a sufficient, but not necessary, condition on placing a syntactic item at or above the rank of word. These seven are syntactic independence, movement, pronominalization, distribution in syntactic frames, deletion, modification and expansion. The expansion test has been shown to have wider 39 applicability cross-linguistically and will be extensively employed in this thesis. Simply by modus tollens, if a constituent is bound as part of a w, it cannot pass any of the seven tests. These seven rules of thumb are based on the assumption that these operations refer to rank phrase or higher rank in Chinese, and phrasal rules cannot apply within a w to break its lexical integrity.

The result of the coordination test is indeterminate, since coordination may in addition refer to Ws in Chinese (and English), conjoining Ws within a compound (cf. 3.6).

2.4. The Dh/ision between Syntax and Morphology

Whether or not there is a syntax-morphology division in Chinese depends first on identifying morphology in the language. So far I have developed the notion of word and a methodology for determining it (or its boundaries). In this section I will show that a Chinese word does have internal structure as in other languages, and justify the syntax-morphology division in

Chinese by arguing that some issues in Chinese syntax cannot be satisfactorily accounted for without referring to the morphology of the language.

2.4.1. Bound Morphemes within Words

Up to this point, I have already presented a few word-internal structures. For instance,

in (14), xue-xi ‘learn’ is composed of a free formxue and a bound form -xi. The aspect markers

in (17)-(19) are parts of verbs. Admittedly, these observations are hardly original - credit goes to Lu (1964), Kratochvil (1968;ch.2 & 3) and Chao (1968:ch.3 & 4), who give lists of compounding

and affixation in Chinese. In a functionally oriented reference grammar, Li & Thompson

(1981:ch.3) adapt their predecessors’ work and give a condensed presentation of Chinese

morphology in a way more accessible to Western linguists. I refer readers to their grammar for

basic information. Below, I concentrate on an analysis of bound stems, as affixation will be

discussed in chapter 4. 40

2.4.2. Bound Stems

One grammatically significant class of bound forms constitutes the great majority of

morpheme entries in a dictionary. I call them BOUND STEMS.^ In processual morphology, a stem is the phonological material, or the base, on which derivational and inflectional rules to operate. A stem is associated with a meaning, as the lexeme it represents has meaning. For

instance, bag is the stem of lexeme BAG and is taken as the base for the inflectional plural rule to apply, giving bags. A stem may be built on a stem, e.g., direct-ion-al-iti-es. In Chinese, bound

stems are stems of lexemes which have special properties. Bound stems are distinguished from

(ordinary) stems, since bound stems, as indicated by their name, are always bound in the

language (cf. the two bound stems In zhi-dao [know-way] 'know' and the stem jin in jin-zi [gold-

norminal suffix] 'gold'.), whereas other stems are free. A bound stem can be attached by a

derivational affix, or combine with another bound stem or a free word. More crucially, a

combination of a bound stem and an inflectional affix is ungrammatical (*zhi-le) in contrast to a

stem-inflection combination (^hi-dao-le ‘knew’). Finally, bound stem s cannot be analyzed as

affixes, as they show few or none of the characteristic properties of affixes (cf. 4.2). This is why

I employ the term bound stem.

Take the bisyllabic expressions in (25) as an example, where F means free, and B means

bound. They are all "compounds" in Chao’s term, but their components are bound.

(25)a. FF: kai-guan [open-close] 'switch'

sheng-zhang [born-grow] 'grow'

b. FB: li-chun [establish-spring] 'Beginning of Spring’

you-qing [right-tendency] 'Right deviation’

^They correspond to "root words" in Chao (1968:145). Standard textbooks (e.g., Hockett 1958:240-2, Bauer 1988:252-3) define a root as the obligatory and unanalyzable part of a word when all inflectional and derivational affixes have been removed. I wiil not use "root" in this thesis. 41

C . BF: you-hao [friend-good] friendly’

wei-xian [danger-danger] ’danger"

d. BB: zhi-dao [know-principle] ‘know’

jue-wu [feei-conscious] ‘be conscious’

All of the B’s in (25) are bound stems. There is no question that all of the words in (25) are histrically phrases, and that via compounding, one or both of the oomponents of many of them

(e.g., those in (23b-d)) have been morphologized to tXDund stems (Dai 1990a;9). Bound stems will be further discussed in 3.2.

2.4.3. Bound Stems and the Delimitation of Morpheme Boundaries

By the structuralist definition, a morpheme is a minimal arbitrary constant union of sound and meaning. I take "meaning" here as either semantic (board in blackboard) or grammatical (- ness in kindness), or both (-s in boys). Aronoff (1976:15) takes this one step further. For him, a morpheme is a phonetic string that can be connected to a linguistic entity outside that string, and constitutes what a speaker is able to recognize. To be sure, in modern Chinese, morpheme boundaries are much more difficult to locate in words than word boundaries in phrases, because of the existence of bound stems, which are etymologically related to free words in Classical

Chinese. There is no consensus among speakers on whether a bisyllabic word is further analyzable into morphemes, when it contains one or two bound stems. In particular, are the words in (23b-d) mono- or bi-morphemic? They appear to be bimorphemic among speakers v/ith

substantial literacy and knowledge of Classical Chinese, as the Classical and dictionary meanings

are provided in the brackets in (23). This seems to be the position taken in Chao (1968:142), as

he suggests taking the maximal analysis of literate speakers as opposed to speakers of varying

degree of literacy for a better chance of reaching a consensus. For illiterate speakers or those 42 with little or no knowledge of Classical Chinese, however, the words in (23b-d) might well be monomorphemic, especially in the case of (23d), for which both syllables are bound in the modem language. Thus it appears that whether an individual word can be further segm ented into morphemes largely depends on the individual’s level of literacy or knowledge of Classical

Chinese. Because of this and the fact that the attachment of bound stems are not easily describable in a productive affixation (or morphological) rule, Sheu (1990:26-8) adheres to the other extreme, considering the internal boundaries within words like those in (23b-d) as not exactly morphemic in nature, i.e., these words are taken in her analysis a s internally monomorphemic and unanalyzable. The same is true of the words in (23a), for although their components are free, the resultant meanings from the combinations are not semantically compositional. My position is just the opposite: bound stems are morphemes within words in general, and so are those B’s in (23) in particular; all of the words in (23) are morphologically segmentable and hence bimorphemic.

For me, the grammar of Chinese is ideally a description of the speakers’ linguistic knowledge of the language. Here "speakers' does not equate Chao’s "literate speakers", and the degree of literacy and knowledge of Classical Chinese play little role in my synchronic analysis of Chinese word structure. I maintain that whether or not a bisyllabic word is further analyzable into morphemes depends on the size of vocabulary of the speaker, specifically, on whether a putative morpheme segmented from one word recurs in another in his lexicon. Some of the

recurrences of those B-syllables, or bound stems, in (25b-d) are listed below.

(25)’b. CHUN in li-chun

chun-tian [spring-day] ‘spring’, chun-feng [spring-wind], kai-chun [open-spring] ‘starting

of spring’, ying-chun [welcome-spring] ‘welcome spring’... 43 b. QING in you-qing

zuo-qing [left-tendency] ‘Light deviation’, qing-xiang [tendency-direction] ‘tendency’, qing-

xie [tendency-tilt] ‘tilted’...

c. YOU in you-hao

you-yi [friend-friendly] ‘friendship’, you-ai [friend-love] ‘friendly’, hao-you [good-friend]

‘good friend‘, peng-you [friend-friend] ‘friend’...

0 . WEI in wei-xian

wei-hai [danger-harmj ‘endanger’, wei-ji [danger-situation] ‘crisis’, bing-wei [sick-danger]

‘critically ill’...

d. ZHI in zhi-dao

zhi-shi [know-recognize] knowledge’, zhi-zu [know-enough] be content with one’s lot’,

tong-zhi [through-know] ‘notify’, wu-zhi [no-know] ‘ignorance’...

d. DAO in zhi-dao

dao-li [principle-reason] ‘principle’, dao-lu [way-road] road’, men-dao [door way] ‘way to

do something’, gan-dao [feel-principle] ‘sen se’...

d. JUE in jue-wu

jue-de [feel get] ‘feel’, jue-xing [feel-wake] ‘awaken’, gan-jue [feel-perceive] ‘feel’, shi-jue

[see-perceive] ‘sight ...

d. WU in jue-wu

ling-wu [appreciate-conscious] ‘appreciate’, xing-wu [wake-conscious] ‘wake up to 44

reality’...

Suppose that all of the words in (25)' are within the vocabulary range of an average speaker of

Chinese, whether or not (s)he is literate. The point of interest of (25)’ is that the bound forms in

(25b-d) do recur in the language. Though they vary idiosyncratically with respect to the degree of combinability with other morphemes and the limitation on left- or right-boundness, they do have consistent meanings and pronunciations. Ignoring such recurrences and saying that words in

(25b-d) are monomorphemic constitutes loss of a significant generalization, implying that the speaker, and hence the grammar of Chinese, does not relate these syllables semantically or/and grammatically at all in different words. In English, this would amount to saying that cranberry and cranapple are monomorphemic, and cran- is not a morpheme in the two words. Just a s speakers of English know that cran- in cranapple has to do with cranberry, and cranberry is a kind of berry, speakers of Chinese know that zhi in zhi-dao and tong-zhi in (25’d) has the same pronunciation and the consistent meaning ’know’, hence the same morpheme. For Aronoff (1976:12-5), even the recurring forms which are referred by a rule may indicate morphemic status in cases where

meanings are opaque but not divergent, such as -mit in permit, transmit, commit, etc. or -take in undertake and partake, and -stand in understand and withstand. He argues that -take and -stand

are individual morphemes, because they are consistent in the phonological operations (- take/took/taken, and -stand/stoodlstood) called by the morphological rules to create the past tense

and past participle forms respectively. The same is true of -mit, witness perm/f/permission,

transmit/transmission, commit/commission, etc.

The morphemic status of zhi in zhi-dao and tong-zhi contrasts sharply with the non-

morphemic status of knowledge in acknowledge in English, given that knowledge has no

synchronic relation (including meaning) in these words at all. Nothing can be drawn beyond the

homonymy of knowledge in the two words (Pollard, p.c.), and hence it is not a morpheme in

acknowledge, which is in turn monomorphemic. 45

To push the argument further, if one syllable in a bisyllabic word is recognized as a morpheme (zh/), so must the other syllable (dao), even though the latter may be a totally isolated bound stem in the language. The same point can made for boysen- in boysenberry. In conclusion, zhi-dao in (23d), for example, is by all means a bimorphemic word, even though one could argue that dao does not really recur elsewhere. This is because zhi- in zhi-dao clearly recurs in the lexicon, as in (25’d).^

Before closing this subsection, I would like to point out that a proper recognition or analysis of the bound stems of Chinese contributes to the understanding of Chinese morphology not merely in quality, but also in quantity. The fact that the language abounds in bound stems makes highly untenable the traditional view that Chinese has little or no morphology.

2.4.4. The Syntax-Morphology Division

Morphology is viewed by some linguists as "the syntax of words' (Selkirk 1982, Lieber

1992), because tx)th syntax and morphology seem to be concerned with requirements on constituency and ordering in certain constructs. Thus, according to Selkirk (1982:2), the word lies on the interface between syntactic representation of two varieties of structure, defined by different sets of grammatical principles; and yet, word structure has the same general properties as syntactic structure and is generated by the same rule systems. In particular, this view seems to assume that hierarchical structure (X+Y -> Z) and linear ordering (X precedes Y, prefixation or suffixation) are tjoth necessary and sufficient to describe word-structure.

The virtue of this view lies in the fact that most morphological rules in natural langauges involve only affixation. Also as in syntax, a context-free rewriting system seem s to be able to assign a labeled tree to all words in a language (e.g., in English, V -> V Affix {work-ed, edit-ed)).

The phrase structure rules not only capture the internal constituency and categorization of words.

^^Note that bi- or poly-syllabic words which are monomorphemic do occur but are rare in Chinese (e.g., boli ‘glass’, zhizhu ‘spider’, tanke ‘tank’, shafa ‘sofa’, etc.). Some of these are clearly borrowed words. 46 but also are recursive, and hence there is no upper bound on the length of words, e.g., given compounding N > N+N, there is ...bathroom towel rack designer training program...; or the derivation in re-re-re-read. Furthermore, certain notions of the X-t)ar theory are required to characterize word structure, in order to capture the 'headedness' of word structure. One can suppose that a suffix is the head in a word, and thus, for regular-ity, the X-bar syntax would be

N - > A N[Affix], where -ity is the head N[Affix], determining the category of the mother as N.

With the head notion extended from syntax to morphology, -ity is treated as the "bottom head" of the maximal syntactic projection the regularity.

This treatment of morphology as isomorphic with syntax is criticized by several authors, most notably by Anderson (1988) and Zwicky (1992:348-54). True, word structure is similar to phrase and sentence structure in the ways above, but affixation does not exhaustively describe morphology. There is a wide range of morphoiogical processes which are not purely affixal in the above sense: infixation, vowel or consonant shift, reduplication, metathesis, subtraction and interdigitation (for a summary, see Zwicky 1988 and references therein). These processes are not easily describable by the X-bar theory. Thus at least non-concatenation morphology does not look like "the syntax of words", but rather, it seems to be more like "the phonology of words" in

Janda’s (1983) sense.

The notion of head in syntax is summarized in Bauer (1990:2), based on Zwicky (1985b) and Hudson (1987), in terms of the criteria that a head of a construction typically fulfills: semantic functor, subcategorizand, morphosyntactic locus, governor, distributional equivalent and

obligatory element. As demonstrated in Zwicky (1985b:1) and Bauer (1990:1), the introduction

of the notion of head from syntax into morphology is unjustified in linguistic theory. For Zwicky

(1985b: 10), in syntax, the morphosyntactic locus is the head in syntactic percolation in the sense

of the Head Feature Convention in GPSG, the constituent bearing morphosyntactic (inflectional)

features and determining the construct as a whole relevant to the external syntax. But the head

in morphological percolation would have a double role: the morphosyntactic locus, and the 47 category determinant on the category of its host and the category of the resultant word.^^ If the head/percolatlon notion In morphology parallels that In syntax, then the head in syntax must also combine both morphosyntactic locus and syntactic determinant. However, the two do not coincide In syntax.^®

For a cross-linguistic perspective, I will show that a division between syntax and morphology should be maintained even In Chinese, one of the most analytical languages In the world. It will be demonstrated that, as In other languages, significantly different principles govern word structure and phrase structure in Chinese. Such a differentiation will hopefully be of some theoretical and metatheoretical Interest, since the grammar of an analytical language would otherwise provide the most convincing evidence for a theory resorting to a single unified component covering both syntax and morphology.

Alternative ordering of constituents Is common in syntax. A constituent may occur in an unexpected position. Suppose that the default word order In Chinese is SVC, as in (26a).

However, In the corresponding ba-construction in (26b), the logical object appears before the verb

(SOV): and in the be/-construction in (26c), the object occurs before the subject (CSV).

(26) a. Wo da ta le.

I hit he Particle (prt.)

‘I hit him.’

b. Wo ba ta da le.

I BA he hit prt.

^'*For example, -ity attaches to an adjective {regular) and determines the nominal category of the combination {regularity).

^^For instance, in an NR containing Determiner + N {the boys) and an S’ containing Complementizer + S {that she is a student), Determiner and Complementizer are syntactic determinants but N and S’ morphosyntactic loci respectively. 48

‘I hit him.’

c. Ta bei wo da le.

he BEI I hit prt.

‘He was hit by me.'

By contrast, the ordering of morphemes within a word is strict, and no alternative ordering is available in morphology (^ue-xi ‘learn’ vs. *xi-xue, zou-xhe ‘vvalking’ vs. *zhe-zou, ha/-zr ‘child’ vs.

*zi-hai, etc.). In other words, movement rules are absent in morphology. Morphological rules are

"local", but syntactic rules may not be so. Sometimes, the latter are even unbounded (cf.

Topicalization: Zhangsan, wo renwei Lisi bu xihuan ‘Zhangsan, I think Lisi doesn’t like.’). In addition, a bound stem apparently in coordination with another in a word cannot interchange positions (BF: wei-xian [danger-danger] ‘danger’ vs. *xian-wei), but two constituents in a syntactic coordination can {laoshi he tongxue ‘teacher and pupil’ vs. tongxue he laoshi).

A morphological rule may change the lexical category of its host (cf. suffixation of -xing

’-ity’ to an adjective jianding ‘steadfast’ to derive a noun jianding-xing steadfastness’); but syntactic rules do not have this ability. For instance, in the ba- and be/-constructions in (24), the head of the displaced NPs remains an N.

Both syntactic rules and morphological rules can have exceptions, but the exceptions in morphology are mostly arbitrary. Thus, in the morphological reduplication for the vividitive of adjectives, *meimeiliii (from meili ‘beautiful’) is an arbitrary gap, since its synonym piaoliang

‘pretty’ can be reduplicated to piaopiaoliangliang (cf. 4.1). In syntax, however, the exceptions are by no means accidental, but, among other things, determined by the semantic and/or pragmatic requirements of the relevent construction. The unproductivity of idiom chunks, such as gua yang tou, mai gou rou [hang a goat’s head, sell a dog’s meat] ‘inconsistency between what one has said and what he is doing’ (C-T. Huang 1984:63), are explained by the non-compositional 49 semantics of the idiom. Similarly, ai ‘love’ or xiang ‘miss’ cannot support the Pa-construction and hen ‘hate’ or taoyan ‘be sick of cannot occur in the Pe;-construction, because these verbs do not signal disposal, a pragmatic condition which is required by the two constructions (Li & Thompson

1981:467 & 501).

Morphological rules may refer to phonology, but syntactic rules may not For instance, although the delimitative reduplication (cf. 4.1) can apply to both mono- and bi-syliable words, only the derivatives from monosyllables can be further infixed with -yi- one’ (xue ‘learn’ -> xue-xue and xue-yi-xue ‘learn a little’ vs. xue-xi ‘learn’ -> xue-xi-xue-xi ‘learn a little‘ but *xue-xi-yi-xue-xi).

However, there is no syntactic rule I know of in Chinese which puts phonological constraints on the material within its domain, such as a topicalized NP must be monosyllabic, or begin with an obstruent (cf. 1.3, the Principle of Phonology-Free Syntax).

Syntactic constituents normally accept optional dependents (arguments or modifiers) in

Chinese. This is related to the syntactic expansion test for wordhood (cf. 2.3.7). But no

morphological constituent accepts optional modifiers (lao-tou [old-norminal affix] ‘old man’ but

*hen-lao-tou and *lao-DE-hen-tou ‘very old man’).

Syntactic constructions may require some immediate constituent to be specific lexeme(s).

Thus the lexemes shi and de are required in the sh/...de-construction below to nominalize the verbal constituent in between (Li & Thompson 1981:587).

(27) Ta S h i zuotian lai de.

he be yesterday come de

‘The situation is that he came yesterday.’

But the idea "strict cyclicity" (Kiparsky 1982) presupposes there could not be such a lexeme

requirement in a morphological rule, since the morphological composition of a lexeme is not

available to morphological rules building on that lexeme (Zwicky 1992:349). 50

There is generally no way to delete a constituent of a word in the non-coordination morphology, but in syntax, deletion, or equivalently, interpretation of a phonetically unrealized element, abounds (cf. 2.3.5). For instance, in (15), the second clause has a missing object

Zhangsan. Such phonological suppression is disallowed in compounds, let alone affixed non­ compound words. For example, neither of the guantou's in Wo xihuan shuiguo-guantou ([fruit- can]), taxihuan sucal-guantou ([vegetable-can]) I like cans of fruit, and he likes cans of vegetable’ can possibly t>e missing in the respective compounds.^

Needless to say, affixes and bound stems cannot stand alone (cf. 2.3.1), and neither of them can be pronominalized (2.3.3); but independent occurrences and pronominalizations of syntactic constituents are common in Chinese.

The above list of distinctions between syntax and morphology does not exhaust all

possibilities, but these are all universal distinctions not specific to Chinese. Again, readers are

referred to Anderson (1988) and Zwicky (1992) for discussion of the universality of the distinctions.

Now I turn to some language-specific points to further motivate the syntax-morphology distinction

in Chinese. If syntax (together with phonology) alone is believed to be sufficient to describe the

structure of Chinese, then many issues on structural properties cannot be satisfactorily

addressed. Below, I mention just a few of them.

Although many items intuitively belong to a major syntactic category [V] or [N], they

cannot freely be placed in a proper slot in a syntactic construction where the corresponding

syntactic words occur, such as those slots reserved for subjects, predicate verbs or objects. For

instance, neither of the components of jue-wu 'be conscious" in (25d) can function as a main verb

in a construction as jue-wu can, even though jue and wu are intuitively [V] categories, according

to their meanings.

^Note again that similar suppression in English compounding is possible (/ adopted the Bloomfield theory, but he adopted the Chomsky.). 51

(28)a *Wo jue tou teng.

I feel head ache

I feel a headache.

b. *Ta wu cuo le.

he know wrong prt.

‘He knows that he was wrong.'

c. Ta jue-wu le.

he be conscious prt,

‘He was conscious (of something).’

Recall that jue and wu are bound stems. Bound stems are syntactically different from their corresponding words in being unable to fill syntactic slots (cf. 2.3.4). Likewise, bound forms cannot undergo certain morphological processes such as the delimitative reduplication, as do their free counterparts: *jue-jue, *wu-wu vs. jue-wu-jue-wu. For details, see Dai (1990a: 13).

In addition, some putative syntactic deletion rules seem to involve components of words, which would violate the Lexicalist position. A well-known structure is the reduced A-not-A questions (Wang 1967, C-T. Huang 1989, Dai 1990b, c & 1991c) where the first syllable is part of the word xihuan ‘like’: xi- bu xihuan [like not like] ‘like or not’. At least three other constructions exhibit this property: xi- shi xihuan, xi- dou xihuan, xi- hai xihuan ‘like be/all/be like’. This class of constructions is discussed in chapter 6 based on a morphosyntactic analysis by Dai (1991c).

Moreover, syntactic movement or expansion rules are unexpectedly prevented from applying to objects of certain intuitive VO structures (dan xin [load-mind] ‘worry’, Xin, ta bu dan.

‘He is not worried.’, but *Xin, ta bu dan Lisi. ‘he doesn’t worry about Lisi.’). The phenomenon is called the "ionization" of certain compounds by Chao (1968:426), a concept which I will discuss 52 in chapter 3.

The notion of prosodic domain formation in phonology may refer to morphology. For one thing, the prosodic formation rule for bisyllabic rhythm respects lexical integrity (Dai 1990b:297).

More precisely, although a bisyllablic foot can freely form at the left edge of certain constructions, such bisyllables are ill-formed if one syllable happens to be part of a word which as a whole cannot be in that bisyllable in forming the prosodic domain. This interaction between phonology and morphology is taken to account adequately for one of the superficially syntactic distinctions bet-Afesn A-not-A questions and alternative questions observed in C-T. Huang (1989:143-50). For example, the grammaticality distinction between the A-not-A question in (29a) and the alternative question (with an extra haishi ‘or’) in (29b) is due not to the syntactic difference between the two types of questions, but because in forming an obligatory bisyllabic prosodic domain ([..]) at the initial position of the predicate VP, the second syllable (hai) cannot be included in that domain since it is part of the word haishi. In other words, the lexical integrity of haishi cannot be broken for the prosodic domain formation. Thus (29b) is ruled out on phonological grounds. By contrast, no such problem arises with bu in (29a), because it does not form a word with the following cong.

(29)a. Ni [cong bu] cong zher chu-qu?

you from not from here go-out

‘Are you going out from here or not?

b. *Ni [cong hai-]shi bu cong zher chu-qu?

you from or not from here go-out

‘Are you going out from here or not?

Here is another case of morphology influencing prosodic domain formation. Shih

(1986:136-46) finds that some putative phrasal categories show word-internal prosodic behavior. 53

For example, the A(djective)+N(oun) structure (especially when A and N are monosyllabic) like hong hua ‘red flower* or hao jiu (good-wine] 'good wine' appears to be a phrasal category NP, which looks both syntactically and semantically transparent. Shih appeals to lexical integrity to explain why one of the expected tone sandhi patterns is unacceptable with slow speech for x/ao gou-bing-gan "small dog-biscuit’. The expected tone sandhi would change the first syllable’s

(x/ao) underlying low tone to low-high, because of the low tone of the second syllable (gou). But this is not the case. According to Shih, x/ao and gou do not form a phonological foot to which the sandhi might apply, since the latter is part of the lexical item gau-bing-gan (/b/d.:141). Citing

Zhu (1956), she presents the following additional lexical correlates to the A-t-N bisyllables which are not true of the corresponding phrasal category. These lexical properties would be accounted for naturally if the A+N is treated as a compound.

Arbitrary gaps are abundant for an A-kN combination, e.g., zang-shou 'dirty hand’ vs.

*zang tang 'dirty candy’. However, grammaticality returns when the function word de (to be discussed in chapter 3 & 5) intervenes to turn the combination into a phrase: zang de tang.

The A-kN combination tolerates a contrary syntactic modifier, but only without de intervening between A and N: hui de bai (*de) zhi 'grayish white paper’. The same is true for the reduplication of an adjective: x/ao de x/ao (*de) hao-zi a tiny little mouse".

Monosyllabic adjectives without de are forbidden in a position before an NP containing determiner-numeral-classifer-N, a clear position for phrasal categories, indicating these adjectives are bound somewhere else.

(30)a. yi duo hong (de) hua

one classifier red (de) flower

b. *hong yi duo hua

c. hong de yi duo hua 54

By a compounding account, the arbitrary gaps are expected lexical gaps. The insertion of de would give a phrasal NP -> AP de N, and an additional modifying AP contrary to the original

AP will impose two contrary properties on the N, resulting in semantic and pragmatic anomaly.

A bare A is banned in pre-NP position, since bare A s only combine with N’s to form a compound

(3.4.2). Combining with NP does nothing but break lexical integrity, a s the phrase yi duo ‘one’ would be part of a compound word, as in (30b).

To strengthen my argument for the lexicalization (or compounding in my term) of A+N,

I add here a fourth lexical correlate: no adverbial modifiers to A are allowed (cf. 2.3.6).^^ For example, lexical integrity bans a modification by an outside adverbial of the A in a compound of

A+N, by assuming that adverbs are words and that the combination of an adverb and an

adjective is a phrasal process (cf. 2.3.7 & Lu 1964:55).

(31 )a. hong hua vs. *hen hong hua vs. "feichang hong hua

‘red flower’ vs. ‘very red flower’ vs. ‘very red flower’

cf. hong de hua vs. hen hong de hua vs. feichang hong de hua

b. piaoliang yishang vs. "hen piaoliang yishang vs. "feichang piaoliang yishang

‘pretty clothes’ vs. ‘very pretty clothes’ vs. ‘very pretty clothes’

cf. piaoliang de yishang vs. hen piaoliang de yishang vs. feichang piaoliang de yishang

These examples shows that components of an A + N compound are not restricted to

monosyllabic words as implicitly assum ed in Shih and Zhu, for (31) involves a monosyllabic

adjective, noun and adverbial (hong, hua and hen respectively), and corresponding bisyllables

(piaoliang, yishang and feichang respectively). For more bisyllabic examples, see 3.5.

^^This is implicit in Norman (1988:161) and explicit in Sheu (1990:139), although Sheu does not explcitly relate it to the LIH, perhaps because she is a sort of anti-strict-Lexicalist (see the discussion of the verb-object phrases vs. compounds in 3.3). 55

Furthermore, my framework predicts that a postadjectival modifier cannot intervene between A and N, but a prenominal modifier without de can. This is because the former would intervene as a syntactic phrase in the compound, while the latter would be merely an adjectival component of the inner A+N compound, showing the recursiveness of this rule.

(32)a. *hong-DE xiang huo yiyang hua

hong-DE xiang huo yiyang de hua

red DE like fire sam e de flower

‘flower red like fire’

b. hong xian (*de) hua

red fresh (de) flower

‘fresh red flower'

However, when the compounding gets recursive, there may be more lexical gaps.

To summarize, the notion that 'morphology is the syntax of words' seems more misleading than enlightening. With Chinese as a target language for study, the distinction between syntax and morphology is too striking to ignore, with respect to constituent order, the ability to change parts of speech, reference to phonology in rule applications, acceptability of modifiers, requirement of specific lexemes, and allowance for missing constituents. Chinese, frequently cited as a typical isolating or analytical language, should be an ideal language for the proponents of the view of morphology as the syntax of words. However, too many differences between the two components in Chinese grammar mitigate the relatively few similarities between them, suggesting just the opposite, a universal division between syntax and morphology. This is one of the fundamental theoretical assumptions on the organization of grammar governing the course of this thesis. 56

In addition, I have presented some classical problems in Chinese linguistics. These include the reduced A-not-A questions, bound stems, restrictions on some prosodic domain formation and tone sandhi rules, and lexical properties of an A-i-N combination. It should be pointed out that any attempt to resort to (automatic) phonology to solve these problems is doomed to failure, since none of these can be properly explained by independently motivated phonological rules in the language. These and many other phenomena indicate that Chinese grammar has a morphological component, which I taelieve is the case for every natural language.

Having motivated the notions of word throughout the major components of Chinese grammar, developed a methodology of accessing word txiundaries in a phrase and morpheme boundaries in a word, and justified a division between the syntax and the morphology of the language, I present a fragment of Chinese below before sketching a description of the morphology of Chinese in the traditional terms of compounding, derivation and inflection in the following chapters.

2.5. A Fragment of Chinese

The following fragment of the language only provides the constructions considered in this thesis. For other constructions, readers are referred to Chao (1968) and Li & Thompson (1981).

I adopt the GPSG ID (Immediate Dominance) rule system, without the incorporation of the UP

(Linear Precedence) statement. H(ead) means that the daughter node inherits the relevant syntactic features from the mother, via the Head Feature Convention. In addition to the major

syntactic categories in terms of N, V, A, P, the head features include [SUBJ, SPE, R, M].

Subcategorization numbers placed in square brackets indicate lexical categories (Ws). (H =

HEAD, S = SENTENCE/CLAUSE, P = PHRASE, W = WORD-RANK, XP = NP, VP, AP or PP, Adv

= ADVERB, SUBJ = SUBJECT, CONJ = CONJUNCT, CONJFORM = CONJUCTION FORM, SPE

= SPECIFIER, DEM = DEMONSTRATIVE, NUM = NUMERAL, [+V] = V or A, [-t-N] = N or A, M

= MEASURE, R = RRESULTATIVE/DESCRIPTIVE, VFORM = VERB FORM, PFORM = 57

PREPOSITION FORM).

S - > XP HP[-SUBJ]

8 - > XP[TOPIC] HS

X - > H[CONJ]+ H[GONJ, {CONJFORM: he 'a n d '...)]

NP -> NP[+SPE] H

NP ~> XP de H

NP - > (XP de) HW[1] {shu ‘book’, ren ‘person’ ...)

NP[+SPE] -> DEMW (NUMW) HW[2] (classifiers: ge, tiao, p i ...)

NP[+SPE] -> (NP[+SPE]) HW[3] {Zhongguo China’, nali ‘that place’, wo T ...)

NP[+M] - > NUMW HW[4] NP (classifiers: ge, tiao, pi, ci, xia ...)

XP[+V] - > AdvP (de) HP

XP[+V] > AdvW[5] H {bu, mei ‘not’ ...)

XP[+V] > AdvW[6] H {hen ‘very’, geng ‘more’)

XP[+V] > H[7] (youyong ‘swim’, meili ‘beautiful’ ...)

VP - > HW[8] NP {hai ‘love’, ta ‘beat’ ...)

VP - > HW{9] NP[+M] {deng ‘wait’, kan ‘look’ ...)

VP > HW[10] NP NP (ge/ ‘give’, song ‘send’ ...)

VP - > HW[11] NP VP {quan persuade", pai ‘appoint’ ...)

VP - > HW[12] S (y/ang ‘think’, zhidao ‘know’ ...)

VP - > HW[13] (NP) 8 {gaosu ‘tell’, tongzhi ‘notify’ ...)

VP > HW[14] VP {neng can", yinggai ‘should’ ...)

VP > HW[15] VP {yao ‘want’, xiwang ‘hope’ ...) 58

VP > HW[16] VP {lai ‘come’, qu ‘go’ ...)

VP[+R] - > VW[17, VFORM:DE] XP[+V] (pao ‘run’, lei ‘tired’ ...)

VP[+Mj - > VW[18, VFORM:TA] NP[+M] {qie ‘cut’, xie ‘write’ ...)

VP > HW(19] PP (zuo ‘sit’, tang ‘lie’ ...)

VP - > PP[PFORM:da] HW[20] (NP) (da ‘beat’, mai‘sell’ ...)

VP - > PP[PFORM;de/] HW[21] (NP) {kan ‘look’, tou ‘steal’ ...)

PP - > HW[22] NP (za; ‘at', cong ‘from’ ...)

NW - > XW[+N] HW[23] {shu ‘tree’, guantou ‘can’ ...)

VW[+R] - > HW[24] XP[+V] (zuo ‘do’, xie ‘write’ ...)

VW[+R] - > HW[25, VFORM:DE] XP[+V] (zuo do’, xie ‘write’ ...)

VW - > HW[26] NP[+M] {deng ‘wait’, kan ‘look’ ...)

VW - > HW[27] NP {ti ‘kick’, kan ‘look’ ...)

The first three rules license the basic pattern of subject + VP, topic + comment S, and coordination which conjoins constituents of the same rank, with a possible conjunction forming a constituent with the last H. Constituents with [+SPE] in the NP rules occupy the leftmost position in an NP, similar to English determiners (a, the, that etc). A specifier may be a phrase in Chinese. It does not have to be the leftmost constituent in a maximal NP, since an NP with a specifier can serve as the head of a maximal NP containing it which has a leftmost non-specifier modifying constituent.

XP[-HV] unifies AP and VP, since the two share a number of important features, e.g., being able to be modified by adverbials including the negators and attached by aspect markers, and being the loci of A-not-A questions. In the literature, A and V are distinguished in terms of stative vs. non-stative (action) verbs, and the boundary is often not easy to draw. The ability to take 59 various complements differentiates most of the Vs from A s, as seen in the set of VP rules.

VP rule [8] licenses a VO-construction. A duration or frequency adverbial NP, marked with [+M] can also be the complement of V, as indicated in rule [9]. Constructions similar to the

English double object and subject-Equi structures are given in [10] and [11]. It has been suggested in the literature that the first complement in these two constructions be analyzed as forming a single constituent with the preceding head V (Y-H. Li 1990), or that the two postverbal complements form a constituent (Dai 1992b); but these are peripheral concerns at this point.

Verbs in [12] and [13] take an S as a complement, and subcategory [13] can in addition has an indirect object. Verbs in [14], [15] and [16] are auxiliary, subject-control, and serial verbs respectively. The resultative construction in [17] and the measurative constructions in [18] will be discussed in detail in chapter 5, the [VFORM]’s of which indicate the existence of inflectional morphemes in the language. Verbal subcategory [19] can take a PP complement. The often- discussed ba- and be/-constructions are licensed by rules [20] and [21]. Implicit here is a denial of one assumption that the two constructions are derived from the VO-constructions. Alternative orderings of constituents, i.e., movements, are not represented by the ID rule system, as they are licensed by gap licensing process specified elsewhere.

The last five W rules license syntactic compounds. The matrix is or functions as a W in syntax. Internally it is syntactically composite, i.e., composed of Ws and Ps. Subcategorization

[23] describe the A+N compound {IQ shu ‘green tree’) and the N+N compound {shuiguo guantou

‘fruit can’). The so-called resultative compounds are licensed by rule [24]. Subcategorization [25] is the potential form of resultative compounds {shou-DE qing, be able to say clear ). Rules [26] and [27] license a constituent of a V plus a measure or object NP which can fill the transitive VW position in VP rule [8] (qiao yixia men [knock once door] ‘knock the door once ) and in rule [9]

{qiao-le men sand [knock perf. door thrice] knocked at the door thrice ).

In sketching the morphology of Chinese, I assume divisions among compounding, derivational and inflectional rules. Morphological compounding, in contrast to syntactic 60 compounding, is illustrated as w]], a morphological word (w) which is morphologically composite, made of two or more w’s. Following Zwicky (1988), I adopt the term operation type.

Thus concatenation of lexemes, or w’s, is one of the operation types.

Operation type 1 : concatenation of w’s (CR = compounding rule)

CR1: concatenation of Xw[+NJ + Nw[23) (X is the modifier or adjunct to Nw; the output is an

Nw: lû-shu ‘gree tree’, rou-guantau [meat-can] ‘can of meat"...)

CR2: concatenation of Vw[24j + Xw[+V] (X denotes the result of Vw; the output is a transitive

Vw[+R]: xie-qingchu ‘write clearly’ ...)

CR3; concatenation of Vw[25, VFORM:DE] + Xw[+V] (X denotes the result of Vw; the output is a transitive Vw[+R] in a potential mode: xie-DE-chingchu ‘be unable to write clearly‘ ...)

CR4: concatenation of wjCONJ]'*’ wjCONJ, (CONJFORM: he ‘and’ ...)] (The output is Xw; the

CONJ’s share sufficient semantic features so that the Xw falls into a semantic category: sucai-(he)- shuiguo ‘vegetable and fruit’)

A compound is licensed when one of the CRs matches the corresponding syntactic (W) compounding rules [23], [24], [25] and the coordination rule conjoining Ws respectively.

Derivation is different from compounding in that a derived w is not morphologically composed of lexemes; it may be internally complex, though, by affixation and other phonological operations such as reduplications. A derivation rule applies to a stem of a lexeme. 61

Operation type 2: affixation (opNo. = a particular operation in the type list, DR = derivational rule)

opi : suffix -tou, op2; suffix -xing, op3: suffix -/e, op5: suffix -zhe, op6: suffix -guo, op7: suffix -DE, op8: suffix -TA ... op23: prefix /ao-, op24: prefix ke- ...

DR 1 : N-stem + opi (The output is an Nw; shi-tou {stone -noun formative] stone', mu-tou

[wood-noun formative] ‘wood’ ...)

DR 2: V-stem + op2 (The output is an Nw: jiandin-xing [firm-ity] ‘firmness’, geming-xing [revolve-

ity] ‘revolutionality’ ...)

DR 23: N-stem + op23 (The output is an Nw: lao-shi [noun formative-teacher] ‘teacher’, lao-hu

(noun formative-tiger] ‘tiger’ ...)

DR 24: V-stem + op24 (The output is an stative Vw: ke-ai [able-Iove] ‘lovable’, ke-kao [able-

depend] ‘dependable’ ...)

Operation type 3: reduplication

DR 25: reduplication of V-stem (The input is an action V; the output is the delimitative: chang

‘taste’ vs. chang-chang [taste-taste] ‘taste a little’, xuexi ‘learn’ vs. xuexi-xuexi [learn-leam] ‘learn

a little’...)

DR 26: reduplication of each syllable of V-stem (The input is stative; the output denotes

vividness: jieshi ‘solid’ vs. jiejieshishi ‘strong and solid’, qingchu ‘clear’ vs. qingqingchuchu

‘perfectly clear’...) 62

DR 27: reduplication of measure-stern (The input is a measure N; the output is the distributive measures: ge-ge a great many’ (of the classifier ge))

Inflection is differentiated from compounding and derivation in that inflected forms of a lexeme are syntactically relevant. An inflectional rule (IR) may call for any operation in the type list. One common morphological feature of Chinese verbs is that they may take an aspect marker

(-lelzhelguo) in a VP rule, except in the (R- and M-) constructions where they are required to assum e other VFCRMs.

IR 3: V-stem + opS (V-perfectiva aspect: chi-le fan [eat-perf. rice] ‘ate meal’, zuo-le shi [do-perf. thing] ‘did something’ ...)

IR 4: V-stem -f op4 (V-durative aspect: chi-zhe fan [eat-dur. rice] ‘eating meals’, zuo-zhe sfii [do- dur. thing] ‘doing something’ ...)

IR 5: V-stem + op5 (V-experiential aspect: chi-guo fan [eat-exp. rice] ‘once ate m eals’, zuo-guo stii [do-exp. thing] ‘once did something’ ...)

IR 6: V-stem + op6 (V-resultative: chi-DE tiao [eat-R well] ‘ate well’, zuo-DE kuai [do-R fast] ‘did fast’ ...)

IR 7: V-stem + op6 (V-potential resultative: kan-DE jian [look-Potential-see] ‘be able to see’, xie-

DE-qingchu [write-Potential clear] ’be able to write clear’ ...) 63

IR 8: V-stem + op7 (V-measurative: chi-TA san wan [eat-M three bowl] eat bowls of food’, zuo-

TA san tian [do-M three day] ‘did three days’ ...)

IR 9: N-stem -+- op1 (The input is a deveibal N; N-deservative: you chi-tou [eat-tou] ‘worth eating’, mei zuo-tou [do-tou] ‘not worth doing’ ...)

As no general compounding, derivational and inflectional rules can describe the irregular reoccurrences of bound stems of the lexemes in the lexicon, I appeal to lexical redundancy rules

(Jackendoff 1975) to relate the stems in the different lexemes containing them. For instance, one rule relates the meaning and pronunciation of the first stem in zhi-dao [know-principle] ‘know’ to the last in wu-zhi [without-know] ignorance’. CHAPTER III

COMPOUNDING

3.0. Introduction

The purpose of this chapter is to explore some new concepts of compounding in Chinese, and to examine some universai consequences of these concepts. Although the notion of compounding is straightforward in the traditional literature, i.e., a compound is a word com posed of two or more existing words, I reserve the term compound/compounding in a more restricted way. Based on the notion of word for different grammatical components, a compound is a syntactic word (W) composed of two (or more) Ws (as opposed to rank phrase or sentence), and

correspondingly, it is a morphological word (w) composed of two (or more) w’s. This is illustrated for rank W/w"^ W/w] by the English compounds blackboard and television table}

More precisely, a compound is composite in both syntax and morphology. Syntactic

composition means that the internal structure of the matrix compound W is licensed by a syntactic

rule in a language, in terms of headedness, (sub)categories, syntactic features for inflected forms,

ordering, grammatical relations, and so forth. By morphological composition, I mean that the

immediate constituents of the matrix compound w is licensed by a morphological rule

concatenating nonaffixal items and referring to their (sub)categories, forms, and ordering.^

’The component W/w’s are free in the sense that they can pass one or more of the five syntactic tests (cf. 2.3) elsewhere in the language, but are bound in compounds.

^Such constructs find independent motivation in phrasal syntax and affixal morphology anyway. As an example of the latter, that the morphological rules licensing regular-iti-es specify, among other things, that the base regular is an adjective (category), that the internal morphemes are arranged in terms of prefixation or suffixation (order), and that the result is a plural (inflectional form).

64 65

Concomitant syntactic and morphological compounding has a special status In a grammar, because such compounds are licensed by the interface between syntax and morphology, but not by either of the grammatical components alone.

If compounding were describable by morphology alone, then problems would arise from those compound Ws {blackboard) containing two or more Ws (black + board), the combination of which is prima facie syntactic. On the other hand, if syntax by itself could license a compound, then it would be hard to account for those syntactic concatenations of component Ws (black + board in blackboard) which may exhibit lexical gaps (like *blijeboard), semantic ncn- compositionality (e.g., A blackboard is not necessarily a black board) and/or phonological

idiosyncrasies (e.g., the primary stress of backboard is on black but not on board). The nature

of these idiosyncrasies is typically morphological. More importantly, all components of a

compound observe the lexical integrity of a word.

In order to explain both syntactic and morphological properties of compounding, I

reiterate two crucial criteria for determining a compound: its internal structure must be a syntactic

concatenation of Ws, and the whole compound must be a morphological unit w.

The syntactic-morphological co-licensing of a compound predicts at least the following,

if other conditions are met: a) syntactic operations referring to rank W may refer to the internal

Ws of a compound; b) compounding rules may be as productive and recursive as phrasal rules;

c) there may be grammatical relations (subject, object, modifier, etc.) between the component

W/w’s of a compound; d) compounding may have arbitrary lexical gaps and semantic and/or

phonological idiosyncrasies, and e) components of a compound respect its lexical integrity.

Because of the internal syntactic concatenations of Ws, prediction a) is straightforward.

For instance, since coordination in English may refer to rank W besides phrase and sentence,

coordination is expected to be possible in compounds (e.g., television and VCR table). By

contrast, since movement, modification or pronominalization operate universally at phrasal levels,

Ws within a compound are not accessible to them according the UH, as stated in e). Likewise, 66 b) is expected, since syntactic concatenations elsewhere are usually productive and recursive

(e.g., the N+N compounding is recursive: television table factory). As a result, a syntactic compounding rule predicts or licenses an open-ended list of compounds in that category. The

N -> N+N syntactic compounding^ rule in English is a case in point, as it predicts an open- ended list (TV table, girl friend, computer program...). By contrast, the sporadic "P > P+P* {into, onto) is not a syntactic compounding rule in English, since *oveno, *underto, etc. are imposssible.

Predication c) follows because grammatical relations (subject, object, modifier, etc.) are constructs of syntax (see examples (1)’-(5)’ later).

The role of the matrix w enforces prediction d) in addition to e), i.e., since a compound as a whole is a morphological unit, or a lexical item, idiosyncratic properties may be assigned to it. Thus a), b) and c) differentiate compounding rules from typical affixation rules in morphology,^ and d) and e) distinguish compounding from phrasal rules in syntax.

The remaining sections of this chapter are arranged as follows. In 3 .1 ,1 differentiate my

notion of compounding from those in the literature and then outline the main issues surrounding

Chinese compounding. Section 3.2 sketches some compound-like bimorphemic lexemes in the

language by employing the notion of bound stem and lexical redundancy rules. The so-called verb-object compounds will be examined with respect to the separability of their components

(3.3). Most verb-object "compounds" should be classified as phrases, some of them lexicalized

according to their morphological behavior in marked syntactic constructions. However, such an

isolated lexicalization is not describable by a general compounding rule, in contrast to the general

rules licensing the class of resultative compounds (3.4) and the A+N and N+N compounds (3.5).

^In syntax, it ought not to be strange that a W may contain Ws, e.g. N -> N+N, giving girl friend, since the immediate constituents of a phrase may contain phrases (at the same rank) or sentences (above the phrasal rank), e.g., VP -> VP+S, giving [^p [^p studied English][Q wtien he was a child]].

'^Except that inflectional rules are in general productive, since the syntactic rules accessible to them are structurally general. For example, the inflectional past participle form -en is required for every instance of the productive passive construction in English. 67

Of particular interest are the properties of syntactic coordination in morphology in terms of semantic equation between the coordinants, among other things, which leads to the proposal that coordination in morphology creates compounds (3.6). The results of this investigation will hopefully not only solve some problems concerning compounds in Chinese, but also yield certain interesting consequences. For Chinese, compounds are distinguished from noncompound words and syntactic phrases, and the magnitude and relative productivity of compounding are likewise assessed. For Universal Grammar, this chapter lends further supports for the division between syntax and morphology, as well as their determination of compoundhccd, their interface in describing compounds, and the constraints on coordination in compounding. These theoretical positions are summarized in 3.7.

3.1. A Review of the Literature and Issues in Chinese Compounding

I first review the definitions of compounding used in the literature. I do not adopt them for this thesis, not because they are inadequate for describing compound-like units in individual languages, but because they do not serve to demonstrate how syntax and morphology interact to describe a subset of the compound-like units, the compounds considered in this framework.

Compound is used as a cover term for a collection of related, but not necessarily identical, phenomena in the literature.^ For some linguists (e.g., Bloomfield 1933:227, Bauer

1988:239, Anderson 1988:187), a compound consists of the combination of two or more existing words into a new word. By this definition, blackboard is a compound in English, as opoosed to transmit, which is not a compound, since neither trans- nor -mit is a word. But the definition does not specify which of the notions of word is in question. For other linguists, however, the term compounding may be used in a broader sense. Hockett (1958:241) uses the term "stem compound" for a highly inflecting language like Latin. Agricola ‘farmer’ has the stem agricol-,

^Sometimes the term is so taken for granted that it need not even be defined in some works on compounding (e.g., see Bates (1988) for compoundings in English and Y-F. Li (1990) for Chinese.). 68 composed of the stem agr- in ager ‘field’ and the stem col- of colere ‘to cultivate’. Since

(uninflected) stems are always bound in the language, a Latin compound is not composed of existing words but of stems. Hockett also uses the term "root compound' for telegraph, phonograph or gramophone, and the term "phrasal compound' for blackboard, bluebird or barefoot in English. Matthews (1974:40) refers to the Latin case as 'compound roof or

'compound stem'.® Note that the morphological component alone seems adequate to describe the root compounds and stem compounds in the sense of Hockett and Matthews, since the nature of their components’ combination is strictly morphological.

The term compound as used by Sinologists represents an even broader concept.

Practically any polysyllabic word (written with two or more characters) is a compound. Li &

Thompson (1981 ;46) take this position for pedagogical purposes. For them, compounds are polysyllabic units that have properties of single words and can be analyzed into two or more meaningful elements, or morphemes, even if these morphemes cannot occur independently.

Chao (1968:ch.6) assum es a relatively restricted definition. He excludes affixes as components of a compound. For him, the items in (1)-(5) below are all compounds. They are traditionally classified according to their 'internal syntax' (summarized in C-T. Huang 1984:61-4). Below, I provide some bisyllabic examples. I use F to indicate a free component or W, and 8 for bound.

(1) Goordinative Compounds

a. FF: deng-huo [lamp-fire] ‘illumination’

b. FB: jiao-shou (teach-transmit] ‘professor’

c. BF: ran-shao [burn-burn] ‘burn’

d. BB: fu-mu [father-mother] ‘parent’

®As a reminder, I assume that a stem is the phonological material used as a base for rules of inflection or derivation. A stem may be built on a stem, in which case the root is the innermost stem. 69

(2) Subject-Predicate Compounds

a FF: tou-teng [head ache] ‘headache’

b. FB: hai-xiao [sea-scream] ‘tidal wave’

c. BF: mu-ai [mother-love] ’maternal love’

d. BB: dong-zhi [winter-come] ‘winter solstice’

(3) Subordinative Compounds

a FF: hei-ban [biack-board] ’blackboard’

b. FB: shou-ce [hand-book] ’manual’

c. BF: liang-xin [good-heart] ‘conscience’

d. BB: deng-hao [equal-sign] ‘equality sign’

(4) Verb-Object Compound

a FF: dan-xin [ioad-mind] ’worry about’

b. FB: ru-wu [enter-rank] ‘join the army’

c. BF: zhi-yin [know-sound] ‘bosom buddy’

d. BB: yan-shi [tire-world] ‘world weary’

(5) Resultative Compounds

a. FF: da-bai [beat-lose] ’defeat’

b. FB: shou-ming [say-clear] ‘illustrate’

c. BF: ju-jue [resist-stop] ‘reject’ 70

d. BB: fa-xian [find(?)-appear] ‘discover"

For Chao, any of the following criteria is sufficient for compoundhood;

(6) a Part of the item is bound.

b. Part of it is in neutral tone.

c. The meaning of the whole is not derivable from the meanings of its parts.

d. The internal structure is exocentric.

e. The parts are inseparable from each other.

The criteria in (6) strike me as a set of unrelated stipulations. Only (6e) follows automatically from the LIH of w. Criterion (6a) presupposes that the components of a compound do not have to be

Ws, thus approximating the Latin type of "compounding", which can be described by morphology alone and not the main concern here (but see 3.2).

C-T. Huang (1984:63-4) excludes condition (6c) as sufficient for compoundhood and subsumes the rest of (6) into the UH. Criterion (6c) is excluded because it is generally agreed that idioms (e.g., kick the bucket in English, and gua yang-tou mai gou-rou ‘hang the goat’s head, but sell the dog’s meat’ in Chinese) are not compounds but phrases; in fact most idioms are built on syntactic constructions (Zwicky 1989b). But Huang seems to have neglected the fact that some (or many) compounds do have non-compositional readings (e.g., VJhite House). In the current framework, this characteristic behavior of compounds follows from an assignment of w to the matrix W and then from a (possible) assignment of a special meaning to a compound lexical item. Another natural consequence of this approach is the exocentricity of compounding in (6d), on the assumption that parts of speech are also assigned in the lexicon, independently of those assigned elsewhere to the components. For instance, barefoot can be an adverb, but neither bare nor foot can. 71

Criterion (6b) wrongly assumes that a compound-like unit in phonology must instantiate a compound-like unit in syntax and morphology: I have already otaserved that there are rank mismatches across different components of grammar (cf. 2.2). Huang reduces (6b) to (6e) by assuming that neutrai-tone elements are usually bound. However, in Chinese, tonality hardly ever determines structural levels reliably. For instance, the noun formative prefix lao- in lao-shu

‘mouse’ or lao-hu tiger’ bears a full tone (Low/Tone 3) but is nevertheless bound. The adjectival de and the adverbial de in wo de shu ‘my book’ and manman de pao ‘slowly run’, on the other hand, are toneless but Ws, since the constituents on either side can be expanded (cf. ch.5). The same is true of the sentential particle words (ma, ne, a, etc., cf. ch.7). Worse still, a neutral-tone element in a putative trochaic compound like /re-sou ‘give a cough’ is separable to ke sheme sou

[give what cough] ‘why cough?’. After separation, sou becomes tonal (Lu 1964:85). Therefore,

I propose that tonality is neither a necessary nor a sufficient condition for compoundhood in

Chinese. Neutral-tone components of compounds are possible or expected, again because phonological idiosyncrasies are assigned in the lexicon. Thus the syntactic and morphological approach to compounding captures and unifies the insights made in (6). I present additional arguments for this non-phonological/semantic approach in section 3.3.

Proceeding from here, the notion of compound must be distinguished from at least four types of compound-like units. The first one has a matrix W/w, but its internal structure is morphologically composite while syntactically simple. This a W/w is composed of two (bound) stems but not Ws, as illustrated by stem"*" stem] in the d-forms in (1)-(5) for Chinese, and for

Latin “compounds'. The second compound-like unit is the b- and c-forms in (1)-(5), in which a

W/w comprises a W/w and a bound stem, a s illustrated by W/w stem] or stem W/w].

The third type represents a phrase-like unit functioning as a W, as illustrated by W/w"*" W/w] for the term "phrase-word" (Bloomfield 1933, Zwicky 1990). For example, gave Mary functions as a transitive verb (e.g., burned) taking a direct object {the book) in gave Mary a book (cf. 3.4). Note the matrix W here is not a w. The last type is a w which is composed of W/w’s, but the matrix w 72 is not a W, as illustrated by W/w"*" W/w] for English clitic groups containing contracted auxiliaries (they’ve) and the serial verb constructions (go see my doctor) in English (Zwicky

1990:212). It is worth mentioning one common characteristic of the above compound-like phenomena in English, Latin and Chinese: each of the complex words is morphologically composite, containing at least two morphemes which linguists are reluctant to call affixes. It is this composite nature of word, but not the label "compounding", that is of some significance to grammar. An adequate grammar should be able to describe each of these phenomena

Having reviewed and clarified the notions of compounding, ! turn to Chinese compounding. Chinese compounding should be of some interest for general linguistics, because it bears on a major issue concerning the organization of Universal Grammar: whether or not a syntax-morphology division should be maintained in compounds. Previous research assumes, either explicitly or implicitly, that such a division is not needed, contrary to the position taken in this paper. For traditional grammarians, the internal structure of compound-like items is a reflection of the organization of words in corresponding syntactic phrases of the language. For example, all types of "internal syntax" in the compound-like units in (1)-(5) also occur in corresponding phrases, as below, and therefore can be said to be describable by the same syntactic rules that license (1)’-(5)’.

(1)’ Coordination

dongtian de xue chuntian de hua

winter de snow spring de flower

‘snows in winter and flowers in spring' 73

(2)’ Subject-Predicate Construction

Nage ren zou le.

that person walk pit.

That man was gone.’

(3)’ Subordination

meiii de hua

beautiful de flower

‘beautiful flowers’

(4)’ Verb-Object Construction

hua-le henduo de qian

spend much de money

‘spend a lot of money’

(5)’ Resultative Construction

pao-DE hen lei

run DE very tired

‘run with the result that one is tired’

Kratochvil (1968:73-4) notes that this identity in the internal grammatical relations has led previous grammarians to disregard the difference between phrases and compound-like lexemes. He also 74 conjectures that such structural resemblance has blurred the borderline between syntax and morphology, in the absence of well-defined criteria for distinguishing phrases from compound-like lexemes. In fact, positing a mixed morpho-syntactic component describing both compounds and phrases has undoubtedly been one of the most vexing problems in both general linguistics and

Chinese grammar. Chinese could be cited as one of the languages whose syntax can describe everything, and whose morphology retreats to "the syntax of words' (Selkirk 1982, Lieber 1992).

In addition, correctly analyzing the apparent resemblance between compounds and phrases is also one of the most urgent areas of research because of its practical implications: the alphabetization of Chinese, lexicography, language reform, and language teaching (Norman

1988:156).

In view of these problems, this study focuses on generalizations between compound words and syntactic phrases, seeking to describe and explain their formal similarities, and account for their differences. Concerning the current notion of compounding, I am also obliged to re-assess the range and degree of productivity of compounding in the language. The

predominant view holds that except for compound-like lexemes, Chinese is a morphology-free

language. Thus one would expect compounding in the traditional sense to have a special place

in Chinese morphology. Lu (1964), Chao (1968:ch.6), Kratochvil (1968:sec.3.5), and Li &

Thompson (1981 :sec.3.2) present various types of processes similar to compounding, which give

rise to the impression that this putative morphology-poor language is nevertheless rich in

"compounding". The richness of Chinese compound-like lexemes would be unquestionable, if one

believes that every natural language should have a minimal amount of morphology, and no

natural language can possibly lack both affixation and compounding (Joseph & Janda 1988).

Since the current notion of compounding is more restrictive, it is conceivable that Chinese

compounding may not be as rich as traditionally assumed. This would pose two further

questions: a) how to describe the compound-like lexemes which are not compounds in my

framework? and b) are there any compounding processes that have been ignored or taken as 75 phrasal syntactic processes? Below I start with the compound-like lexemes in (1)-(5) which cannot be compounds in my framework.

3.2. Compound-Like Bimorphemic Lexemes

Contrary to almost all previous analyses, I hold that examples in (1 )-(5) are not necessarily compounds, though they can be shown to be W/w’s (via the tests in 2.3). According to the current framework, none of the l>, c-, and d-forms are compounds, since each of them contains at least one bound (B) element which is not a W. Those B-eiements belong to no syntactic

(sub)categories. For instance, the "verbal" loound form in (4c), zhi-, never functions as a free verb in a construction. They are stem s of a class of lexemes with special properties, which I call bound stems (cf. 2.4.2). The distributions of many such stems in a W are not rule-governed. There is no general derivational rule I know of taking zhi- as a base/stem to form a new lexeme with another stem: *zhi-li [know-principle], *you-zhi [have-know], etc. Further, they cannot serve as the base for an inflectional rule to give a w as other stem s can: *zh/-DE, -TA, -ie, -zhe, -guo, or

-tou. All these observations concerning zhi- are true of the other B-elements in the b-, c- and d- forms. The most significant characteristic of these B-elements or bound stems, as argued in

2.4.3, is their morphemic status. Since they are stems of some lexemes, they recur as meaningful elements in the language. They are like those morphemes in cranberry, cranapple, boysenberry, and even gramophone, photostat or telegram in English (cf. Hockett’s (1958:241) term root compound). In order to achieve the W/w status, such a stem needs to combine with another

(bound or free) stem, as in (1)-(5), or serve as a base to which some derivational rule applies

(with a suffix as in yi-zi 'chair', or with a prefix as in lao-hu 'tiger').^ I call the words in (1)-(5)

^As the concatenation of a bound stem to another stem is not rule-governed, these derivational affixation rules are by no means general, because the bound stems they operate on form no natural classes (cf. shi-zi vs. *lao-shi 'lion', *shu-zi vs lao-shu 'mouse'). 76 bimorphemic lexemes.® These used to be phrases, with their parts morphologized via compounding at a certain point of in the history of Chinese.

Because those bimorphemeic lexemes are not compounds, Chinese compounding does not seem so rich as it is in the traditional grammar. I will discuss an alternative later by discovering some productive synchronic compoundings in the language. The major problem, though, is howto account for the internal structure of those compound-like non-compound words, given that the distributions of their components are not describable by general rules and that the components recur as morphemes with some predictability in pronunciation and meaning.

Obviously, the predictable properties should be specified in the lexicon. With some hesitation,

I further suggest that these lexical properties are covered by lexical redundancy rules. Such rules are also needed to predict lexical redundancy properties of English photograph, gramophone, photostat or telegram, whose internal structures cannot be covered by general rules, either.

As the lexicon lists the properties of these bimorphemic lexemes, the morphology need not state that, for instance, BB; yan-shi [tire-world] ‘world weary’ is generated from two existing words tao-yan ‘hate’ and shi-jie ‘world’ by truncation of the first syllable of the first word and the second syllable of the second word.® Such a process is hardly justifiable, since yan- could be so derived from many other "underlying" existing words: yan-fan [tire-be-sick-of] be sick of, yan- juan [tire-sleepy] ‘be weary of, yan-wu [tire-hate] ‘detest’, etc, and -shi could be also derived from shi-dao [world-priciple] ‘the manners and morals of the time’, shi-gu [world-wise] ‘worldly-wise’, shi-su [world-custom] ‘common customs’, etc. Instead, with a full lexicon in which the lexemes like tao-yan, shi-jie, yan-shi and others are all listed a redundancy rule predicts that the lexicon

®Bimorphemic lexemes in Chinese are similar to the compound-like units of the Latin type in that morphology alone can describe both; the Chinese type is different from the Latin one in that the former can contain a free (F) form.

®There is so little literature in Chinese generative morphology that I know of no one who has actually made such a proposal before. But such an approach is implicit in Thompson (1973:365), who claims that a lexical rule uses an existing resultative compound {ia-kai [pull-open]) to derive its potential form {ia-DE/bu-kai [pull-can/cannot-open] ‘can/cannot pull (something) open ) by insertion of -DE/bu-. 77 should include properties of these stems and relate the same stems to one another in all the exisiting W/w’s containing them.

To recapitulate, by the notion of compounding proposed here, most of the bimorphemic lexemes which were historically compounds are no longer synchronic compounds in Chinese.

They are lexemes containing stems of different lexemes.

What remains for me to consider is the a-forms in (f) (5). Since their components are free, they are subject to a compound analysis, pending their syntactic and morphological behaviors. Next, ! turn to the a-forms by focusing on the VO-type dan-xin in (4a). involved in the

"ionization" phenomenon, this type of expression has instigated much debate as to whether they are compounds or phrases.

3.3. The Lexicalization of Some Verb-Object (VO) Structures

The term ionization is first used in Chao (1968:159, 426-34), who observes that the components of some of the putative VO-compounds are separable, that is, they can pass the expansion and movement tests just as VO-phrases can. The phenomenon is similar to chemical ionization in which a part of a chemical compound floats around in the sam e solution as the other parts. Now consider dan-xin [Ioad-mind] ‘feel anxious, worry’’° in (7).

(7)a. Ta tian-tian dan xin.

he day-day load mind

‘He was worried day after day.’

b. Ta dan-xin zhejian shi.

he worry this matter

^°Xin ‘mind’ is etymologically related toxin ‘heart’, since in Chinese folklore (as reflected in the language), people think by "heart", as well as by "brain/mind". The interpretation of "load mind" means th e mind is burdened (by something)", hence the English translation ‘worry, feel anxious’. 78

‘He worried about this matter.’

c. Ta dan-le bantian de xin.

he load perf. half-day de mind

‘He worried for a while.’

d. Xin, ta conglai mei dan.

mind he up-tc-now not load

‘He has never been worried."

e. *Ta dan-le bantian de xin zhejian shi.

he load perf. half-day de mind this matter

f. *Xin, ta conglai mai dan zhejian shi.

mind he up-to-now not load this matter

These examples illustrate that dan xin in (7a) might be a compound, since both of its compronents are W/w’s elsewhere, and more importantly, it can take the object zhejian shi like a transitive V,

as in (7b). However, a time modifier appears to syntactically interrupt this compound, as in (7c),

and xin as part of the word may be fronted, as in (7d). Traditional grammar, as implicit in Chao

(1968), thus treats parts of certain compounds as separable.

Such a treatment is criticized by C-T. Huang (1984), since it constitutes an obvious

violation of the LIH, as the syntactic operations of expansion and movement refer to proper parts

of a w. More importantly, what needs to be explained is the observation that when dan-xin takes

an object, a similar separation of its parts is impossible, as in (7e) and (7f). Adhering strictly to

the LIH, C-T. Huang (1984:64) proposes that these VO-expressions are idiomatic phrases, which 79 are lexicalized to compounds when followed by an object, enforced by his Chinese Phrase

Structure Condition (PSC) in (8).

(8) The PSC

Within a given sentence in Chinese, the head (here, the verb or VP) may branch to the

left only once, and only on the lowest level of expansion (p.54).

Stating this in other terms, the PSC could be understood as ’a verb in Chinese may be followed by at most one constituent" (p.54). According to Huang, the PSC "requires" (p.69) that the phrase dan xin in (7a) be lexicalized into a single word in (7b), since otherwise the verb dan would be followed by two constituents xin and zhejian shi, in violation of the PSC. And in accordance with the UH, the lexicalized dan-xin in (7b) cannot be separated, as illustrated in (7e) and (7f).

Sheu (1990:ch.2) takes issue with Huang, criticizing him for not formalizing the rule of lexicalization. She also argues that Huang’s analysis is implausible in two ways (p.40-2). First, it lacks motivation for classifying these bisyllabic VO-expressions as idiomatic verb phrases in the lexicon. Second, Huang’s lexicalization rule must be configurationally conditioned by the PSC and by whether the expression in question is a transitive verb followed by an object. To solve these problems, Sheu proposes instead a derivation in the opposite direction: the "ionized" VO- phrases are derived via a "reanalysis" rule which inserts an adverbial phrase into the lexeme to give a phrase. That is, when dan-xin is an intransitive verb in (7a), the reanalysis rule, for instance, sandwiches the phrase bantian da between the two syllables, giving (7c); but if the word is transitive as in (7b), then the reanalysis rule does not apply, and hence the ungrammaticality of (7e). Presumably, a similar reanalysis can be made to account for the contrast between (7d) and (7f) with respect to the possibility of fronting the object xin. More crucially, Sheu takes her reanalysis rule as evidence for relaxing the strict UH, since word-internal information must remain accessible to certain syntactic rules (p.43). 80

Such a reanalysis rule cannot be adopted in this thesis, since the syntactic rule obviously

refers to the internal structure of words, though Sheu’s critique of C-T. Huang is worth

considering.

I submit that listing those VO-phrases as idioms, which Sheu takes as a vulnerable

position on metatheoretical grounds, is not the crucial point at issue, since (at least) the "idioms'

here are formally built on the VO-construction. They participate in the construction, allowing for the expansion and extraction of the object, as in (7c/d).^^ Alternatively, these VO-expressions

may illustrate syntactic selection (Bach 1984) (make/*dol*arrangeI*handle the bed), something

between a fully productive construction (kiss the girt) and an isolated idiom (kick the bucket), both

parasitic on the VO-construction. By this view, the meaning of the combination of verb (dan) and

object noun (x/n) is not necessarily non-compositional, and their collocation is listed in the lexical

entries of dan and xin, which is in turn invoked by the syntactic VO-rule.

Indeed, the main points in C-T. Huang (1984) must be preserved: the maintenance of a

strict Lexicalist position and a desire to capture a generalization in relationship between VO-

phrases and their corresponding VO-compounds. The main flaw of Huang is the introduction of

the problematic PSC into his analysis which should have resorted solely to the UH for an

explanation. I assume that the LIH is a theoretical universal, slight varaints of which underlie most

current linguistic theories, as opposed the parochial and stipulative PSC, which has been hotly

disputed in the literature (cf. 6.2). Discarding the PSC would certainly improve discussion of the

issue. This does not mean that the phrase structure of Chinese is unconstrained. The point at

issue here is whether a putative constraint should be stipulated on a language-specific basis as

in the PSC or made to follow from some more general principles of grammar.^ ^ For instance,

’^Cf. Some English idioms can undergo the passive rule: I let the cat cut of the bag vs. The cat was let out of the bag.

^^C-T. Huang (1984:74) is obviously aware of this issue, suggesting his PSC may be made as a theorem of the Case theory in GB (Chomsky 1982). But the notion of syntactic Case is highly theory-dependent, and it is controversial over whether the assumption underlying Case theory is universal (Nichols 1986:114) and whether abstract Case is a plausibie construct in 81 the left-headedness and binary branching parameters can certainly be made to follow from the principle governing the position and number of immediate constituents with repect to the head within individual subcategorization frames. One conceivable tendency, though not necessarily a condition of Chinese grammar, would be that a veital head subcategorizes for only one complement (Dai 1992a: 112. cf. 5.4 & 6.2), except in some marked subcategorizations analogous to the English double-object and Equi-constructions (Dai 1992b). Such a tendency would give rise to the binary-branching 'condition' as stipulated in the PSC. Below, I give my analysis for the ionization phenomenon without appealing to the PSC, although the result is consistent with the binary-branching tendency exhibited by default subcategorizations.

To begin with, I do not presuppose that the first point at issue is whether dan xin is a compound or not; but rather whether or not it is a lexical item (W/w). Only after its wordhood and then its internal resemblance to some syntactic structure are confirmed will the possibility of a compound analysis be open. This is because compoundhood depends also on whether or not its internal parts are describable by a general syntactic rule, like the one for boyfriend or blackboard, but not for into in English.

The leading idea states that an apparent phrase-like constituent can be analyzed as a lexical item according to its syntactic environment and morphological behavior. The key difference between the PSC and my program is that the former is a putative language-specific condition whereas the latter will be demonstrated to be universally valid for distinguishing between a lexical item and its corresponding phrase. In particular, the phrase-like dan xin IS a W/w if some syntactic constructions refer to it as a W and if the internal structure of this W observes the lexical

integrity of w.

In my analysis, dan xin is a VO-phrase in (7a) because it passes the expansion and

movement tests in (7c/d), as other VO-phrases do. However, this putative VO-phrase may or

should be analyzed as a W in a marked construction, that is, if dan xin can be inserted in a

Chinese (Dai 1992b, cf. 6.3). 82

position required by a syntactic construction for word-rank, it should be (re)ana!yzed as a W, or

bar-zero (X°), whichever terms one prefers. I contend that dan-xin is a W, or at least functions as

a W, in a matrix VO-construction in (7b), since the construction calls for word-rank V, i.e.,

apparently dan-xin is the V and zhejian shi the NP object. Dan-xin not only occupies the head verb postion and subcategories for an object in the VO-subcategorization frame as would a

monomorphemic verb like zuo 'do' in (7b), but can also be conjoined with an unquestionable

bimorphemic lexeme (W/w) you-lû ‘anxious’ in the construction (Ni bu-bi dan-xin he you-lû zhejian shi [you not-must worry and anxious this matter] ‘You don’t have to be worried and anxious about this matter.’. I therefore conclude that dan-xin is a W in the syntax of (7b).

As pointed out by Zwicky (1990:206), though, functioning as word syntactically does not

entail that the constituent is a word in morphology.^^ Independent morphological evidence is

needed for dan-xin to be considered one syntactic and morphological word in (7b). Such

evidence does in fact exist. It is exactly in this lexical position that dan-xin is not expandable or

extractable as it is elsewhere, owing to the methodology of testing morphological wordhood. The

morphological wordhood (w) of dan-xin in (7b) is justified empirically, for the same expansion and

movement are unexpectedly blocked in (7e) and (7f), both being sufficient conditions to determine

the structural level above the word, if other conditions are met, as in (7c) and (7d). Below, I

reinforce the position that the lexical status of a phrase-like unit is determined by both syntactic

in English, me pnrase-iiKe expression give Mary runaions as a woro, i.e., as a transitive verb read, in give Mary a book. For some more liberal speakers, it can be conjoined with a VW (I'll either give Mary or discard the book). But give Mary is not a word in the morphological sense, since its parts are not subject to lexical integrity. They are movable, i.e., referred to by a syntactic passive rule, e.g., Mary was given a book. Similarly, in Chinese, the VP containing a verb plus a postverbal measure phrase, ti-le yi jiao in (i), can be said to function syntactically as a word-like transitive verb ti-le in (ii), taking the object (da-men), but the VP has no lexical status at all owing to the separability of its parts, as in (iii).

i. Ta ti-le yi jiao da-men. he kick perf. one foot big-door ‘He kicked the gate once.'

ii. Ta ti-le da-men. iii. Ta ti-le da-men yi jiao. 83 environment and morphological behavior, which I call the syntax-morphology co-determination of wordhood.

Sheu (1990:42) points out that it is odd that a VO-phrase is turned into a verb if followed by an NP, as proposed by Huang (1984), but does not state the reason for this.’^ On the contrary, I consider Huang’s underlying syntactic assumption universal,’® though Huang’s condition is stated in language-specific terms and has empirical problems.

For instance, the wordhood of dan-xin does not syntactically depend on whether dan is followed by b«vo constituents, as "required" by Huang’s PSC, since two constituents can immediately follow dan elsewhere, as in

(9) Ta dan-le wo de xin bantian.

he load perf. I de mind half-day

‘He worried about me for quite a while.’

Nor is the wordhood of dan-xin due to the fact that in a VP two NP objects are disallowed. The real reason is that although dan ‘load’ is a verb in Chinese, it is not ditransitive, taking double objects. Dan behaves very differently from ditransitive verbs like song send’ and gei ‘give’ with respect to the fronting of the indirect object, and the attachment of aspect markers, as in (10) and

(11), as well as in semantic argument structure.

(10)a. Ta dan-xin zhejian shi.

he worry this matter

‘He worried about this matter.’

’^1 do not see a substantial difference in the awkwardness and parochiality between Sheu’s requirement of verbal intransitivity vs. transtivity for her "reanalysis" and Huang’s condition of a single postverbal object on his "lexicalization".

’®That is, the share of the burdern that syntax holds in the co-determination of wordhood. 84

b. *Xin, ta dan zhejian shi.

c. *Ta dan-le/guo-xin zhejian shi.

d. Ta dan-xin-le/guo zhejian shi.

(11)a. Wo gei Zhangsan naben shu.

I give perf. Zhangsan that book

‘I will give that book to Zhangsan.’

b. Zhangsan, wo gei naben shu.

c. Wo gei-le/guo Zhangsan naben shu.

d. *Wo gei Zhangsan-le/guo naben shu.

The distinction in the b-forms show that Zhangsan but not -xin is the indirect object NP in a ditransitive construction. The c-forms illustrate that gei but not dan- is the outermost morphological word (w) which fills the W position as a ditransitive verb. By contrast, the d-forms demonstrate that it is dan-xin but not gei Zhangsan that is the outermost w filling the VW position in the VO-construction.

Moreover, it is a false generalization that dan xin is lexicalized if followed by an object (or a constituent), so as to obey the PSC. While (9) demonstrates this superficial effect of the PSC,

(12) illustrates that lexicalization takes place even if no constituents follow dan-xin, as evidenced

by the inseparability of it parts, as in (12b/c).

(12)a. Ta hen dan-xin.

he very worry

‘He was very worried.’ 85

b. *Ta hen dan-le bantian de xin.

he very load perf. half-day de mind

‘He was very worried for a while.’

c. *Xin, ta yizhi hen dan.

mind he continuous very load

‘He has been worried.'

Sheu’s (1990) account also makes a wrong prediction here, since by her analysis, dan-xin in (12a) is a prima facie intransitive verb, to which one expects the 'reanalysis' rule to apply to 'ionize' the parts of the verb as the rule does in (7c/d). In the current framework, by contrast, dan-xin in (11 a) is a lexeme, since its parts are subject to lexical integrity, i.e., unexpandable and unextractable, as in (12a/b), as opposed to (7c/d). Further supporting evidence comes from the syntactic fact that the adverb hen ‘very’ can modify an adjective {hen hong ‘very red ) or a stative verb {hen xihuan ‘very fond of), but not a non-stative verb {*hen da to beat very much ). Note that when dan xin is a VO-phrase, dan is an action V, literally meaning ‘load (mind)’. When hen is present, however, a stative reading ‘worry’ is forced on dan plus xin. This further confirms that the hen- construction plays a crucial role (i.e., dan-xin is a W/w after hen) in the lexicalization. Thus syntactic determination should not be narrowly understood as relevant only to certain following constituents.’®

The resultative construction (see ch.5) also supports the syntactic and morphological principle proposed. Here dan-xin is a W too, in the lexical head position, suffixed by the resultative DE and subcategorizing for a VP in the construction (Dai 1991a:76), as in (13a). It is unexpandable, as in (13b). Mobility of its parts is shown by some speakers, as in (13c/d). I have

’®lt should be pointed out that without hen, (12a) would be structurally ambiguous as to phrasal or W/w assignment. 86 no formal account for (13c/d), but suspect that such an extraction of xin involves a different construction. Here dan may be analyzed as an independent verb, andx/n as a base-generated topic in situ,^^ not necessarily as "moved" from the position in (13a).

(13)a. Ta dan-xin-DE lian fan dou chi bu xia.

he worry DE even meal all eat not down

‘He was so worried that he couldn't eat.’

b. *Ta dan bantian de xin-DE lian fan dou chi bu xia.

he carry halfday de heart DE even meal all eat down

c. ?Xin, ta dan-DE lian fan dou chi bu xia.

d. ?Ta xin dan-DE lian fan dou chi bu xia

Thus the universal principle that the lexical status of a "phrase" is syntactically and morphologically determined renders some language-specific conditions (e.g., the PSC) and rules

(e.g., the "reanalysis") superficial, although lexicalization is expected to comply with the principles governing individual subcategorizations. For instance, if dan-xin in (7b) and (13a) were a VO- phrase, the putative verb dan would occur in an illegal frame subcategorizing for two complements, since it is not marked as ditransitive like gei ‘give’ in the lexicon.

It follows from syntactic and morphological co-determination that an expression or a sequence of Ws in a language may be a phrase in one structure but a W/w in another. English bare back and bareback are a case in point. It should be pointed out, though, that the syntactic

^ ^Perhaps it is because the corresponding "verb copying" construction Ta dan xin dan-DE lian fan dou chi bu xia is fine. Since here the second dan{-DE) is used alone as the action verb and interpreted as referential to the first dan, it is predicted that dan xin is phrasal. This is borne out, because a prenominal modifier to xin is possible, as in dan ta didi de xin ‘worry about his younger brother'. 87 determination of W/w status is often obscured by pfionoiogical and/or semantic factors. For example, there is a meaning difference between the phrase bang tui ‘wrap legs’ and the W/w bang-tui ‘puttee’; the second syllable of the phrase gai huo ‘cover fire’ is tonal (Low/Tone 3), as opposed to that in the W/w gai-huo ‘stove cover’, which is toneless, in English, the afterstressed phrase bare back in The horse has a bare back vs. the forestressed bareback in She rode bareback and without any reins are often taken as sufficient for determining the wordhood of bareback, and the syntactic determination of W status in the second construction retreats to secondary importance. But an overlooked syntactic fact is that if bareback were an NP as bare back containing an AP and an N, then other similar NPs could also occur in that position (*She rode (a) black bike and without any seats.). In Chinese, tou teng [head-ache] can be analyzed as a phrase^® or a W/w according to their syntactic properties (Li & Thompson 1981:70), e.g..

Ta de tou (hen) teng [he de head (very) ache] ‘He has a severe headache.’ vs. Ta (hen) tou-teng

[he-(very)-headache] ‘He has a severe headache.’. In the latter sentence, tou-teng is a W because the adverbial hen only modifies an adjectival W in the hen-construction. But this overtly syntactic determination is obscured by an apparent semantic factor, that is, the W has an alternative idiosyncractic meaning "bothered by a tough problem" and the whole sentence means

‘He is having a tough time’.

The current framework leads one to expect idiosyncratic phonological and semantic

properties in individual lexemes and not in the corresponding phrases, but there is no convincing

evidence to show that such distinctions exist for every pair. That is, a W/w and the corresponding

phrase must have distinctions in phonology or semantics, or both. The examples in (7) are a

case in point, where there are no obvious phonological and semantic criteria to differentiate

between the phrase and the corresponding W/w.’® In such a case, the syntactic determination

’^Actually, it is a two Ws forming no syntactic constituent in this case.

’®There are some subtle differences. For example, xin can be prosodically weak and non- referential in (7b) but not in (7a). 88 of the status of dan-xin is salient. As mentioned earlier, the crucial difference loetween (7a) and

(7b) is that in the latter case dan-xin is in the lexical head position and takes an object The morphological wordhood of dan-xin is confirmed, and determined, by the fact that its parts in (7e) and (7f) respect the UH: this inseparabitty is otherwise unexplainable.

I do not deny the fact that semantics/phonology may play a role in determining the lexical status of an expression; but my point is that syntax-morphology provides both necessary and sufficient conditions on wordhood, and the possible idiosyncrasy in semantics/phonology follows only as a consequence of the lexical and morphological properties of that expression. As some putative language-specific conditions (e.g., the PSC) fail to differentiate phrases from the corresponding W/w’s and thus lack explanatory adequacy, I raise this syntax-morphology co­ determination of wordhood to the status of language universal, subject to further falsification by other languages.

Like all previous analyses, my analysis captures the generalization in similarities and differences between the lexical dan-xin and the phrasal dan xin. One could simply say that dan- xin (like roll in English) is listed both as an intransitive verb, to account for (12a), and as a transitive verb, to account for (7a/b/e/f). To cover (7c/d), VW dan and NW xin should be also listed in the lexicon, invoked by an independent VO-construction, giving [yp [y dan][f.^p xin]\.

Unfortunately, listing three separate lexical items presupposes that danxin does not have to bear any relationship to dan and xin, therefore wrongly implying that the striking syntactic, semantic and phonological similarities among the three expressions are accidental. But this very relationship among the three expressions is exactly the generalization that Chao, Huang, Sheu, and other linguists try to capture.

A desirable proposal would, for instance, posit dan-xin as a compound, corresponding to the phrasal dan xin. If this were the case, then the internal syntax of dan-xin would be licensed by a general syntactic rule VW ~> HW NW, giving (yyy dan][^jy/Xfn]], which is similar to the

VO-phrasal rule for [yp [y dan][,^p x/n]j, except for the differences in syntactic rank and 89 morphological requirements. By this account, the generalization would be captured in terms of the relationship between the VW-rule and the VP-rule. Unfortunately, this cannot be the case, since Chinese demonstrably does not have such a VW rule.

Admittedly, the lexicalization of dan xin is not an isolated instance in Chinese. There are quite a number of such VO expressions. Their parts are usually separable in one way or another, but not when the expressions are in a VW position taking an object. This group contains bang mang [help busy] ‘help’, bao mi [keep secrecy] ‘keep (something) secret’, bao xian [assure danger] ‘buy insurance (for)’, cao xin [rack brain] ‘worry (about)’, dao luan [smash disorder]

‘make trouble’, fan xin/shen [vex mind] ’feel vexed (about)’, fei xin/shen [spend mind] ‘give consideration (to)’, fuze [take responsibility] ‘be responsible (for)’, guanxin [show mind] ’show concern (for)’, Hu xin/shen [leave mind] ‘be careful (about)’, ou qi [have anger] ‘sulk (over)’, ping fan [level opposition] ‘rehabilitate’, qi hong [rise uproar] ’gather together to create a disturbance

(of)’, qing shi [request instruction] ’ask (someone) for instructions (about/of)’, tou ji [invest chance] ‘be opportunistic (about)’, z/7uy/ [pay attention] ‘pay attention (to)’, etc. Liberal speakers may include others, such as chu ban [issue edition] ‘publish’, de zui [get offense] ‘offend’, qi cao

[rise draft] ‘draft a document (on)’, zhu ce [register book-classifier] ‘register’, etc. in the list.

The relevant point, however, is not length of this list, but whether the list is open-ended, thus demanding a general syntactic rule VW > HW NW to describe it. Such a rule would meet morphological conditions so that a general VO-compoundinq rule is warranted. Unfortunately, the list of VO-lexicalizations is not open-ended, and not all VO-phrases in the language can be lexicalized in the way that dan xin does. In other words, speakers cannot creatively use this lexicalization process. A review of all of the examples in the list reveals that they are all bisyllabic, the collocation between V and 0 is strict, and more importantly, their meanings, lexemes or phrases, all involve a certain degree of idiomaticity. Therefore, it is safe to say that they are isolated historical residues calling for no VO-compounding rules. There are many syntactically

and semantically transparent VO phrases like kan dianying see movies’, xuexi shuxue learn math’. 90 or yun dongxi ‘carry things’ which cannot become compounds in the same structure. Worse still, even those cases attributable to historical residue similar to dan xin fail here. Although they may be understood to be semantically transitive, they never take external syntactic objects (C-T. Huang

1984:71 ) or resultative complements as dan-xin does. They are always VO-phrases/constructions, ranging from productive VO-combinations (e.g., buo pi [peel skin] ‘peel’) to idioms built on the

VO-construction (e.g., xiao bian [small convenience] ’urinate’, chi doufu [eat tjeancurd] tease’, sheng qi [rise air] ‘angry’), with syntactic selections in between (e.g., su ku [tell bitterness]

‘complain’, kai dao [open knife] ‘operate on’, li fa [arrange hair] ‘give a haircut’, jie hun [join marriage] ‘marry’).

Since the syntax of a compounding rule must refer at least in part to categories (V, N, etc.) but not exclusively to item-specific lexemes {dan, xin, etc.), the generalization between these phrases and lexemes is that the dan xin type of idioms which are formally built on the VO- construction can be used as W/w’s - so far only some twenty of them. Such a W/w (dan-xin) is morphologically composite in that it contains two w’s (dan and xin), which are the same w’s which syntax can refer to as Ws in the phrasal dan xin. Preserved in the lexicalization analysis are the semantic interpretation and most of the syntactic requirements of the VO idiomatic construction, which accounts for the similarity between dan-xin and dan xin. However, the W/w (dan-xin) is not syntactically composite. Dan-xin appears to fit the compounding rank schema [^/^ W/w"*" W/w] owing to the default one-to-one correspondence between W and w in Chinese (and other natural languages). But since its components can hardly be referred to categorially in syntax, dan-xin assumes the structure [y^/^ w"*" w] instead. This case is isolated, or an "isolated compound". The putative subject-predicate compounds in (2) also belong to this type. Cases of "isolated compounds" exist in other languages. For example, English has no general compounding rule for the lexicalization of verb + particle: giveaway vs. *givein, sit-in vs. * sit-over, pushover vs.

*pushoff, send-off vs. *send-away, etc. Note that all such cases involve a certain degree of idiomaticity, even for their phrasal counterparts. 91

In summary, I find that the lexical analysis adopting the syntax-morphology approach is the optimal one for accounting for the ionization phenomenon. As compared with C-T. Huang

(1984) and Sheu (1990), the present analysis not only captures the generalization between lexical dan-xin and phrasal dan xin via their realization by the same morphological words, but also maintains the universal UH and abandons the problematic PSC. The study justifies a principle in universal grammar that whether or not a phrase-like constituent (or a sequence of W/w’s) should be analyzed as a lexical item depends on both its syntactic and morphological contexts: the constituent is positioned where syntax requires a W and its parts are subject to the lexical integrity of w. The lexical status of a phrase-like unit does not necessarily give it compound status. The dan-xin type of VO-expressions has shown this compound-like structure w"*" w] to be merely morphologically but not syntactically composite. Although widely called VO- compound, this structure does not correspond to the structure I label as compounding in this study.

Before closing this section, I would like to point out that the ionization phenomenon is in some sense a reflection of diachronic morphologization or syntacticization in the language. After the transformation of certain historical phrases {dan xin) into words (dan-xin) via compounding, speakers are beginning to reanalyze them as phrases by manipulating them through syntactic operations of expansion and extraction, hence the term diachronic "restoration" which turns a disyllabic word into a phrase (Jin 1991:45). Thus Chao (1968) and Sheu (1990) are certainlv correct in their approach on strictly historical grounds. Clear evidence comes from some innovative speakers who even reanalyze the transliteration of the English word humor as a VO-

phrase you-mo tease". The demorphologization in question is a relatively rare example of

diachronic syntacticization of the internal structure of words, as compared with the cross-linguistic frequency of morphologization of phrases (Joseph & Janda 1988).^ However, for an optimal

^°ln addition to a few cases of demorphologization discussed there, Callaghan (1987a:337) reports that the copula stem -?e- in Plains Miwok is developed from a verbalizer suffix. 92 synchronic analysis of the ionization phenomenon, I opt for the opposite direction, that is, apparent syntactic phrases (dan xin) are analyzed as words {dan-xin) in marked constructions, in other words, the fact that the separability of these expressions is "a case of folk reanalysis'

(Sheu 1990:41) is irrelevant to a synchronic analysis; the essential point for a grammar to capture is that the 'reanalysis' exists across most or all of the speakers. This suggests that a synchronic analysis may yield dramatically different predictions from the corresponding historical process.

3.4. The Resultative Compounds

While the VO (and the subject-predicate) lexicalization has been excluded from the

(productive) compounding family, it remains as an empirical question as to whether Chinese has general compounding rules like those for blackboard and TV set in English. In this section, I will demonstrate that a subclass of the resultative (A) compounds can be covered by a general compounding rule relating to the resultative (A) construction.

Thompson (1973:361-79) presents several lexical rules to generate the R-compound verbs. I will discuss just two of them, showing their relationships with the phrasai A-construction.

Very roughly translated to my framework, an R-compound is composed of a VW indicating an action, followed by an AP or VP as the result of the action, as in (14a), matched by the morphological rule in (14b).

(14)a. VWj-t-R] - > HW[24] XP(+V] i^ie-qingchu ‘write clearly )

b. CA2: concatenation of Vw[24j + Xw[-(-V]

An R-compound forms its 'potential' mode or form by 'infixing' -DE- or its negation -bu-, adding to the meaning whether the action is achievable or not - "be able to' or 'be unable to" (Chao

1968:457, Thompson 1973:365, Li & Thompson 1981:39). The rules for the potential form are distinct from the set in (14), as in (15). 93

(15)a. VW[+R] - > HW[25,VFORM:DE] XP[+V] {^ie-DE-qingchu ‘be able to write clearly’)

b. CR3: concatenation of Vw[25, VFORM:DE] + Xw[+V]

This is because the negation of (14) has met ‘haven’t’ before the HW (me/ xie-qingchu ‘didn’t write clearly’). By contrast, the negation of the potential form in (15) has bu ‘not’ between the HW and the XP with a loss of -DE i^ie-bu-qingchu ‘be unable to write clearly’ vs. Im ei xie-DE-qingchu and l*b u xie-DE-qingchu‘be not able to write clearly ). Since the potential DE form is required by a particular syntactic rule (15a), it is inflectional by its syntactic relevance (cf. 6.6.4).

All previous studies assume that -DE- and -bu- are infixes in (15). This need not be the case, since DE can be treated as a suffix to the head in (15), and its negation can be stipulated by inserting bu with the loss of -DE. Also implicit in the previous studies is the complementary distribution of -DE and bu. But such a complementary distribution is merely semantic, not structural, since DE always forms a constituent with the preceding HW (see 5.3) and bu with the following XW. If the two constituencies were the same, there would be no explantion as to why conjunction reduction is possible with DE but impossible with bu.

(16)a. Ni kan-DE (he) ting-DE qingchu dianshi ma?

you see DE (and) listen DE clear TV Q-prt.

‘Are you able to see and listen to the TV program clearly?’

b. *Ta kan-bu (he/yie) ting-bu qingchu dianshi.

you see DE (and/also) listen DE clear TV

‘He was unable to see and listen to the TV program clearly.’ 94

These details are not of cruicia! importance at this moment, since the external syntax of both

potential and nonpotential forms are the same. The relevant point here is that the R-

compoundings (assumed to be unified under a single rule) license an indefintely long list of the

R-compounds in the language (cf. Thompson 1973:369).

(17) sh a/da/ti... (-DE/bu)-si ‘(be (un)able to) kill/beat/kick ... -die)’, kan/shuo/ting/jiang/)ieshi

... (-DE/bu)-qingchu ‘(be (un)able) to see/speak/listerVtalk/explain ... -clear),

xi/caynong/qingli... (-DE/bu)-ganjing ‘(be (un)able to) wash/erase/handle/clear... -clean',

bai/nong/fang/cunfang ... (-DE/bu)-zhengqi ‘(be (un)able to put/handle/place/store ... -

orderly, etc.

Like the lexicalized dan-xin, syntax is able to place an R-compound in the transitive VW

position of the VO-construction, taking an object, as in (18).

(18)a. Ta kan-DE-qingchu naben shu.

he see-DE-clear that book

'He was able to see that book clearly.’

b. Ni xi-bu-ganjing zhejian yifu.

you wash-not-clean this coat

‘You are unable to wash this coat clean.’

c. Qing bai-zhengqi yizi zai zuo.

please put-in-order then go

'Please put the chairs in order before leaving.’ 95

Independently, the R-construction, as discussed in Dai (1991a:67, b & 1992a:103) and chapter 5, is licensed by the following phrasal-rule;

(19) VP[+R] - > VW[17, VFORM:DE] XP[+V]

The first constituent of the VP[+R] is a VW inflected in the DE-form; the second constituent is an

AP, VP or S denoting the result or description of the head VW. Of course, VP[+R] can be

realized by sequences of Ws which happen to realize one of the compounds in (17) if all other conditions are satisfied, as shown in (20).

(2Q)a. Ta de yanjing kan-DE qingchu.

he de eye see DE clear

'His eyes see cleariy.’

b. Zhejian yifu xi-DE ganjing.

this coat wash DE clean

‘This coat was w ashed clean."

c. Yizi yao bai-DE zhengqi.

chair must put DE in-order

‘The chairs should be arranged neatly.’

But since the XP[-fV] is phrasal in the R-construction, it need not be a word-phrase (qingchu,

ganjing and zhengqi) in (20), but can be expanded one way or another, as in (21). 96

(21)a Ta de yanjing kan-DE (hen qingchu].

he de eye see DE very clear

‘His eyes see very clearly.’

b. Zhejian yifu xi-DE [zhen ganjing].

this clothing wash DE really clean

‘This article of clothing was washed really clean.’

c. Yizi yao bai-DE [bi dengzi hai zhengqi].

chair must put DE than bench more in-order

‘The chairs should be arranged in more orderly fashion than the benches.’

Such expandability clearly demonstrates that the [4-V] complement in (19) has phrase-rank rather than W-rank, as captured by the category assignment of XP[+V].

Next I establish the compoundhood of those in (18), again via the syntactic and morphological co-determination of wordwood. The syntactic wordhood of the R-structures in (18) is already borne out, since they function as a transitive VW in the VO-construction; the evidence for their morphological wordhood (w) stems from obsen/ing of the lexical integrity of the concatenation. Parallel to the dan-xin case, the same expansions as in (21) are unexpectedly blocked, as illustrated in (22).

(22)a. *Ta kan-DE-[hen]-qingchu naben shu.

he see-DE-very-clear that book

‘He was able to see that book very clearly." 97

b. *Ni xi-bu-[zhen]-ganjing zhejian yifu.

you wash-bu-really clean this coat

‘You are unable to wash this coat really clean.'

c. *Qing bai-[(DE) bi dengzi hai]-zhengqi yizi zai zuo.

please put-(DE) than bench more -in-order chair then go

‘Please put the chairs in more orderly fashion than the benches before leaving.’

Since compounds constitute a subset of W/w, it can be said that the compoundhood of an apparent syntactic phrase is both syntactically and morphologically determined, i.e., the syntax places it in a position calling for rank-W, and the morphology imposes lexical integrity on it.

Of theoretical interest are the striking structural similarités between the R-compound and the R-construction, which an adequate theoretical framework should capture. In the static framework adapted in this paper, the generalization is stated not as derivational or transformational, but as a relationship among rules in universal grammar.

By assuming that the basic semantic interpretations associated with the syntactic feature

[+Rj in both W- and phrasal-rules are the same (i.e., in an R-matrix, the second constituent denotes the result of the first one), it is predicted that the verbal subcategory class [24/25] in compounding rules (14) and (15) overlaps the subcategory [14] in phrasal rule (19), to at least a significantly large extent. This is because subcategorization classes are determined chiefly by semantics.^^ This is borne out by the facts, as almost all those in subcategories [24/25] can

^^This view predicts that give and walk in English does not share the same subcategory of verbs (/ gave you a book vs. *1 walked the dog a bone), since their semantic argument requirements are very different. However, this semantic view would wrongly predicte that give and donate are in one subcategory (/ gavel*donated the church ten dollars). To account for this, one extreme position would be that subcategory division is arbitrary, or purely syntactic. But this view misses the generalization that subcategory classes are largely predictable on semantic grounds. The position taken here is moderate; subcategory classes are semantically determined but with lexical exceptions. 98 be used as subcategory [17]. The overlapping of the lexemes of two subcategories thus anticipates the formal similarities between the two syntactic rules. I will describe the compounding first.

In terms of a syntax-morphology interface program (Zwicky 1990:208), it is the ultimate, and minimal, constituents of Ws, l.e., kan(-DE) 'see' and qingchu ‘clear’ in the compound kan(-

DE)-qingchu, which serve as the interface between syntax and morphology. Morphology provides the list of w's available for instantiating a sequence of Ws. In order to be instantiable as an expression, a (compound) morphoiogicai word must have properties compatible with those required by the syntactic rules in (14a) or (15a) licensing the compound, including a syntactic category (V for kan and [-HV] for qingchu), subcategory {kan as a transitive V taking resultative or object complement, and qingchu as an intransitive or stative VP), ordering (VW preceding XP), and the forms (the potential DE form on the head).

Note that all of these structural properties are inherent in syntactic rule (19) licensing the corresponding R-phrase, except that in the compounding rule, the head VW may take the DE-form with a potential-mode reading. By contrast, in the R-construction, the head VW is obligatorily inflected by -DE. Thus the striking similarities between the R-compounds and R-phrases - explicated in traditional terms, that the internal structure of compounds is a reflection of the organization of words in corresponding syntactic phrases - are captured by the similarity among the individual syntactic rules describing each group.

However, the similarities may easily obscure the differences between the R-compounds

and phrases, which are frequently ignored by traditional grammarians. The crucial difference,

besides those mentioned in the syntactic rules, is that the R-compound but not the R-phrase is

a morphological word. The lexical integrity of the matrix w allows no phrasal properties for its

components (hence the unexpandabiliiy in (22)). Thus the offending phrase-rank of the

complement (XP) in the syntactic compounding (15a) can be accommodated only when the XP

is an apparent word-phrase, losing all its phrasal properties in the strictly morphological context 99 in (15b). The second difference is the semantic idiosyncrasy in the compounding rule: while the

DE is semantically a dummy or means to such an extent that* in the phrasal case, it must be interpreted as the potential mode in the compound case. All these differences fall out naturally from the morphological nature of compounding assumed so far.

Summarizing, the study of the R-compounds leads to the position of syntactic and morphological co-determination of compoundhood for an apparent syntactic phrase. The interface link between morphology and syntax is a type of syntactic condition of a compound morphological rule, with a morphological constraint on a syntactic concatenation within the compound W/w. The relationship between a compound and its corresponding phrase, i.e., their formal similarities and differences, is captured by the relationship between the compounding rule and the phrasal rule, stemming from the overlapping of the head subcategories in the two rules.

3.5. N-i-N and A+N Compoundings

This section will demonstrate that a class of synchronic combinations of N with N and A with N which are prima facie phrases are actually compounds, describable by general rules. In

these compounds, the first element modifies the second, and the result is an N. Compounds with

an internal modifying-modified relationship are traditionally called subordinative compounds (of.

examples in (3)).

For methodological reasons, I only consider bisyllabic N’s or A s as components of the

structure, even though I have shown that the number of syllables (to be represented as $) is

irrelevant in the A-FN compounding in 2.4. For those cases, an A-fN structure with monosyllabic

components is demonstrated as compound in traditional sense. I believe that the same is true

of N4-N structures, e.g., gong-ren [work-person] ‘factory worker’. In addition, expressions like

$$4-$ or $+$$, in which one of the components is a monosyllable, have been generally taken as

compounds in the literature, presumably because of the specialization of the resultant meanings

from some of them, e.g., N4-N: shizi-gou [lion-dog] ‘Pekingese’, ma-weiba [horse-tail] ‘horse tail". 100

A+N: zhi-xingzi [straight-temper] ‘straightforward chap’, mingbai-ren [clear-person] ‘sensible person’, etc. Examples with components of more than two syllables are rare and the results may have even more specialized meanings, e.g., N+N: youji-huahewu [[have-machine]-[mett- compound-substance]] ‘organic compound’. All the above can be shown to be compounds in the current framework.

Bisyllabic components are chosen here not only because the oven^/elming majority of words are bisyllabic in the modern language (Lù 1963, Dai 1990b:294-5), but also for phonological

(but still methodological) reasons, in traditional grammar (e.g.. Lu 1964 or Chao 1968), a neutral tone or weak syllable is allowed to properly define a subpart of a compound. There is no such opportunity for a combination of two bisyllabic Ws, because the Bisyllabic Rhythm Rule (Dai

1990b:297-8) applies to make each bisyllabic syntactic word an independent phonological unit,

and it is virtually impossible for both syllables to be weak in each bisyllabic component of a

$$+$$ compound. However, a combination of two Ws with a total of three syllables may be viewed as a compound if the monosyllabic W happens to be pronounced weak or atonic.

Although the approach to compounding is both syntactic and morphological in this thesis, I try to avoid contradicting the traditional prosodic approach. My case would be made stronger or at

least less complicated by considering prosody, since even in the classical analysis, there are no

purely phonological criteria for treating a combination of two (bisyllabic) independent phonological

units as a compound. In fact, combinations of two bisyllabic Ws are widely considered to be

phrases in the literature, especially when the meaning is fairly compositional.

Recall that adjectival and adverbial de's are (function) Ws, since they pass the expansion

test (Dai 1991a:71 & ch.5). Therefore, an expression with a de in it is unquestionably a phrase

and not a W/w.^ The same can be said of adverbs like bu ‘not’, hen ‘very’, feichang ‘very’ or

geng ‘even, still’. They are W/w’s, canonically modifying an A or a stative verb. In their presence,

^This point is raised in Sheu (1990:121), who argues that adjectival expressions cannot modify nouns in the absence of de. 101 the result is also a phrase (cf. 2.3.7 & Lu 1964:55).

One major claim here is that an N+N or A+N expression without an intervening de is a compound.^^ The syntax of the two compoundings can be described by one rule: NW >

XW[+N] HW[23], matched by the morphological rule (CR1): concatenation of Xw[+N] + Nw[23].

Before proceeding, I take two things for granted and set them aside: the function of these compounds as NWs in the syntax and the corresponding similarity between the syntactic rules

NW - > XW[+N] HW[23] and NP - > XP de HW[1]. Thus I can focus on their differences - the morphological properties of compounds which are uncharacterisitic of phrases. Some of the arguments follow a similar line to those given for A+N com pounds with monosyllabic components

(cf. 2.4.4).

3.5.1. Compounding of N+N

Consider the combinations of N+N below. These are usually considered phrases but I will demonstrate them to be compounds in my framework. The structure contains two bisyllabic nouns (disregarding the internal structure of these nouns):

(23) zhongguo-renmin [China-people] ‘Chinese people’, ertong-ieyuan [child-paradise]

'children’s playground’, naiyou-binggan [butter-cracker] ‘butter crackers’, chengren-

xuexiao [adult school] ‘an adult school’, wanju-gongchang [toy-factory] 'a factory making

toys’, tudi-geming [land-revolution] ‘agrarian revolution’, shuiguo-guantou [fruit-can] ‘can

of fruit’, suanpan-naodai [abacus-brain] a math whiz’, hongmu-zhuozi [padauk-table] ‘a

padauk table’, etc.

The internal structure of the above expressions does not behave like a phrasal

^^his is the position taken by Chao (1968:285), but he gives no real justification, and most of his examples are of odd syllables with specialized meanings. 102 combination, since, first of all, there are "selectional restrictions’ on this combination, i.e., the process is not structurally general as would be expected from a phrasal rule. However, with the intervention of a (semantically dummy) de, the lack of generality disappears, a s illustrated below, ruling out a possible pragmatic explanation for the gaps. On the other hand, if the combination is morphological compounding, then the oddness can be naturally attributed to lexical exceptions/gaps; since compounding is a lexical process, lexical gaps are expected^'* (cf. White

Housenwhite-building/white building, blackbird/?greenbird/green bird, etc.).

(24)a. hongmu-zhuozi [padauk-table] ‘a padauk table’

?shanmu-zhuozi [China fir-table] a China fir table'

shanmu de zhuozi [China fir de table] ‘a table made of China fir’

b. ertong-ieyuan [child-paradise] "children playground’

?daren-leyuan [adult-paradise] ‘adult playground’

daren de leyuan [adult de paradise] 'playground or paradise for adults’

c. chengren-xuexiao [adult school] ‘an adult school’

?xiaohai'Xuexiao [child school]‘a children’s school’

xiaohai de xuexiao [child de school]a school for children’

Secondly, some of the expressions in (23) have highly specialized meanings. An intervening de causes obligatory loss of the specialized meaning, allowing only a compositional meaning. For instance, one reading of tudi-geming [land-revolution] ‘agrarian revolution’ refers

^'‘Admittedly, the degree of the (un)acceptability for the chunks marked with ’?" varies among speakers. To say the least, however, a consensus can be reached that the acceptability is low when the chunks occur to the speakers for the first time. The reason for this will be discussed later. 103 to the civil war period of 1927-1937, when the Communists fought with the Nationalists for land on behalf of the peasants. However, tudi de geming only gives a generic reading ‘revolution of land'. Sometimes the intervention of de results in an anomalous reading, but the structure is well- formed. cf. naiyou-binggan [butter-cracker] ‘butter crackers’ vs. naiyou de binggan a butter’s cracker", wanju-gongchang [toy-factory] a factory making toys’ vs. wanju de gongchang a toy's factory’, etc. Suanpan-naodai [abacus-brain] ‘math whiz’ may refer to a person who is good at calculating. But with de, the anomalous meaning is the brain of the abacus". The contrast between constructions with de and without de are quite unexpected from the phrasal account.

Appealing to idioms is implausible. This is because, although the meaning in question is less compositional, it is certainly not entirely noncompositional as is kick the bucket. More importantly, the "idiomaticity" does not hold across-the-board in the N+N combinations. By contrast, a compound analysis yields a natural explanation, since a special meaning may be assigned in the lexicon, e.g., the English compound the White House vs. the phrase the white house.

More significantly, an external adjectival phrase with de may not modify the first N (N1) in the expression, as in (25b) ("#" designates an impossible reading.). The modification of the second N (N2) is apparent, since N2 is the ultimate modified head of the expression.

Consequently, an external modifier pragmatically compatible with N1 but not N2 yields an anomalous reading, represented as "?" in (25c). With de inserted in the compound, the anomaly is gone, as in (25d). This means that only when N1 is m ade phrasal (via being followed by de) can it take a modifier. Also, an external modifier incompatible with N1 but compatible with N2 will not yield an anomaly, as in (25e).

(25)a. zhongguo-renmin [China-people] ‘Chinese people’

b. weida de zhongguo-renmin [great de China-people]

‘great Chinese people’ 104

# ‘people in the great country China’

c. liaokuo de zhongguo-renmin [broad de China-people]

?‘broad Chinese people’

# ‘people in the broad country China’

d. liaokuo de zhongguo de renmin [broad de China de people]

‘people in the broad country China’

- #?‘broad Chinese people’

e. qinlao de zhongguo-renmin [hard-working de China-people]

‘hard-working Chinese people’

#?'people in the hard-working country China’

A similar set of examples are given below:

(26) a wanju-gongchang [toy-factory] a factory making toys’

b. Shanghai de wanju-gongchang [Shanghai de toy-factory]

a toy tactory in Shanghai'

# ‘a factory making toys of Shanghai-style’

c. ertong de wanju-gongchang [child de toy-factory]

?‘a children’s play factory’

# ‘a factory making toys for children’ 105

d. ertong de wanju de gongchang [child de toy-factory]

‘a factory making toys for children'

#?‘a toy factory played by children'

0 . juda de wanju-gongchang [large de toy-factory]

‘a large factory making toys'

# ‘a factory making big toys'

The blocking of an external modifier remains unexplained by the phrasal account, since either of

N1 and N2 can receive the corresponding modifier with de in isolation, cf. liaokuo de zhongguo

'the broad country China', Shanghai de wanju ‘toys made in Shanghai'. By the morphological

compounding account, however, the solution follows from the UH. Since such modification works

at the phrasal level, it cannot penetrate into a W/w to break its integrity. Expectedly, a phrasal

modifer to N2 cannot be placed between N1 and N2, as in (27a/b), but if N1 is made phrasal by

de, then grammaticality returns, as in (27c).

(27) a hongmu-zhuozi [padauk-table] ‘padauk table’

b. *hongmu- yuan de zhuozi [padauk- round de table] ‘round padauk table’

c. hongmu de yuan de zhuozi [padauk de round de table] ‘round padauk table'

It seems possible to interrupt zhongguo-renmin ‘Chinese people’ as in zhongguo weida

de renmin ‘the great people in China' (Q. Gao, p.c.). However this is illusory, because the

construction cannot be derived by placing the AP weida de in the compound, but is instead

licensed by an independent syntactic rule NP -> NP[-»-SPE] NP (cf. 2.5), where zhongguo

occupies the NP[-I-SPE] position and weida de renmin constitutes the second NP. Supporting

evidence comes from the fact that in the "interrupted" case, the nationality reading ‘Chinese 106 people’ from compounding is absent. For Zhongguo there only has a generic SPECIFIER reading, denoting a location ‘in China’, on a par with zhege guojia ‘that country’ in the same [+SPE] position in zhege guojia weida de renmin 'the great people in that country’. This also explains why not all N+N compounds can be so "expanded", because not all Ns can occur in [+SPE] position in syntax, as demonstrated by the ungrammatical case in (27).

For these reasons, I conclude that the N+N structure is a compound rather than a phrase in Chinese.

3.S.2. Compounding of A+N

For similar reasons, A+N combinations like those below should be considered compounds but not phrases.

(28) xinxian-sucai [fresh-vegetable] ‘fresh vegetable’, fengfu-wuchan [rich-resource] ‘rich

resources’, youhao-shizhe [friendship-envoy] ‘a friendship envoy’, pinqiong-diqu [poor-

area] a poor area’, fada-guojia [developed-countries] ‘developed countries’, jianku-

douzheng [hard-struggle] ‘hard struggle’, youxiu-xuesheng [excellent-student] ‘an

excellent student’, jinyao-guantou [critical-moment] ‘a critical moment’, shizu-nianling [full-

age] ‘the exact age’, anquan-huochai [safe-match] ‘safety matches’, gaoji-fayuan [high-

court] ‘the supreme court', etc.

Most adjectives do not readily participate in this compounding process. For instance, the antonyms of the adjective in the first three compounds in (28) cannot be the A in A+N compounds: Ifuian-shucai [rotten-vegetable], ?pinfa-wuchan [poor-resource] and Ididui-shizhe

[hostile-envoyj. But with de, the result is fine. However, this may be irrelevant to compoundhood, since a phrasal account could also specify that some adjectives in the language must take de when concatenated with a noun, while others do not have to, just as double objects are allowed 107 for send {They sent a letter to him/They sent him a letter) but not for explain {They explained the question to me/*They explained me the question) in English. But the fact that an A can directly attach to only certain N’s remians unexplained. Note the oddness of the following combinations.

(29) ?xinxian-xu 0 shui [fresh-snow-water] "fresh snow water’, ?fengfu-chanpin [rich-product]

‘rich products’, ?youhao-qinqi [friendly-relative] ‘friendly relatives’, ?pinqiong-zhishi [poor-

knowledge] ‘poor knowledge’, ?fada-gongye [developed-industry] ‘developed industry’,

?iianku-changzheng [hard-long-march] ‘a hard long march’, ?youxiu-haizi [excellent-child]

’an excellent child’, ?]inyao-gongchen [critical-project] ‘a critical project’, ?shizu-daxlao

[full-size] ‘the exact size’, ?anquan-difang [safe-place] ‘a safe place’, ?gaoji-jiaoshi [high-

classroom] ‘a high-quality classroom’, etc.

No semantic or pragmatic explanation seems to be available for the awkward "phrasal" combinations in (29) in contrast to those in (28), since the awkwardness is removed if de is inserted. Note that an account falls out automatically from the current approach, which states that they are lexical exceptions/gaps in morphological compounding.

Meaning specialization is observed in some of the compounds. Take the last three cases in (28) as examples. Shizu-nianling ‘the exact age’ means the age which is officially counted from the date of birth, as opposed to a lunar counting by which a baby is already considered one year old upon his birth, so the lunar age of a person is always one year more than the full and exact' age. Anquan-huochai ‘safety matches’ are safe to use as opposed to the relatively unsafe matches of the old days, when they might have caused a spontaneous fire. GaojI-fayuan ‘the supreme court’ refers to the supreme court of a province or of the nation. If de is inserted in these three compounds, then the specialized meaning is lost, often giving rise to anomaly.

As protected by lexical integrity, these compounds may accept an adjectival modifier, the

meaning of which is contrary or even contradictory to the A within the A-fN compound, as in the 108 b-forms below. But such contradiction is disallowed in the corresponding syntactic construction with de, as in the a-forms (cf. English examples: a white blackboard vs. ?a white black board).

(30)a. ?bu shizu de shizu de nianling

b. bu shizu de shizu-nianling

not full de full-age

‘an Exact Age that is not fully counted’

(31)a ?bu anquan de anquan de huochai

b. bu anquan de anquan-huochai

not save de safe-match

a safety match which is not safe’

(32) a ?bu gaoji de gaoji de fayuan

b. bu gaoji de gaoji-fayuan

not high de high-court

a supreme court which is not of high quality’

The UH also prevents an adverbial from modifying the A within the compound, as in the b-forms below. The ungrammaticality stems from the fact that adverbs cannot modify nouns, in contrast to the acceptabilty of the corresponding phrases, as in the c-forms (cf. 2.3.6). The a-forms give the canonical position in which these adverbials occur in syntax.

(33) a. bu xinxian ‘not fresh’

b. *bu xinxian-sucai

c. bu xinxian de sucai 109

not fresh vegetable

‘unfresh vegetables’

(34)a. hen fengfu ‘very rich’

b. *hen fengfu-wuchan

c. hen fengfu de wuchan

very rich de resource

‘very rich resources'

(35)a. feichang youhao ‘very friendly’

b. *feichang youhao-shizhe

c. feichang youhao de shizhe

very friendly de envoy

a very friendly envoy’

It would be very difficult to explain all the phenomena above if the A-t-N strings in (28) were syntactic phrases. Therefore I reach the conclusion that those A+N’s are compound words in Chinese.

3.6. Coordination in Compound Morphology

One may wonder why coordination as a syntactic process may occur in morphological

compounding (cf. 2.3.8). But in the current framework, which assumes that a compound is

composed of Ws, coordination in compounding is possible and even expected. This is because

at least for Chinese (as well as English), syntactic coordination may refer to rank W in addition to phrases and clauses. If other conditions are satisfied, there is no reason why the W’s within

a compound cannot be conjoined. I proceed with an analysis of conjoined nominate and submit 110 that instances of W-/?e-W (W-and-W) in a marked structure are subject to a compounding analysis. Note that the conjunction he conjoining word-phrases is common in phrasal syntax, as below.

(36) Ta mai-le [,^p [,^p sucai] he [^p shuiguo]].

he buy perf. vegetable and fruit

‘He bought vegetables and fruit’

Given that compoundhood of Ws is syntactically determined and that [NP Njj (as opposed to [|^p [NP de Njj) is an impossible structure in Chinese, the "same" N-/7e-N in (37) is subject to, or even requires, a compounding analysis, on a par with the first component N in the N+N compound.

(37) (n [n sucai] he [,,j shuiguo]] [^ guantou]]

vegetable and fruit can

‘cans of vegetable and cans of fruit’

Indeed, I will argue for the structure of a compound within a compound given above, and show that coordination in compounding has properties different from phrasal coordinations, attributing the peculiarities of coordination in compounds to the morphological structure of compounds.

I proceed with the "conjuction reduction" reading obtained in (37). As conjunction reduction is treated universally in syntax (Sag et al. 1985, Tai 1972), the current framework allows for it to refer to W in a compound (cf. TV and VCR tables in English and (37)). It follows that a purely morphological concatenation is immune from the reduction {*worked and -ing, *sing- and dancing, and *lao-hu he -shu [N-formative-tiger and -mouse] ‘tiger and mouse). The coordinations of word-like units in (38a/b/c) are also purely morphological, since the whole is a 111 semantic mixture of the two parts. Their internal structures are syntactically simple in that the coordination is not categorially describable by syntactic W-rules. Since they are only concatenations of w’s, they are not compounds, and thus the conjunction reduction reading is not available.

(38)a. [father-mother’s] conference

‘parents conference’

^'father's conference and mother's conference'

b. [deng-huo] wanhui (cf. (la))

lamp fire evening

‘evening of illumination’

# ‘evening of lamps and evening of fire’

c. [naiyou danhuang] binggan

butter yolk cracker

‘butter-yolk cracker’

# ’butter crackers and eggyolk crackers’

d. [naiyou he danhuang] binggan

butter and yolk cracker

# ’butter-yolk cracker’

‘butter crackers and eggyolk crackers'

Conjuction reduction is usually possible in the Chinese case where he intervenes betweem each

bisyllabic conjunct, as in (38d). 112

As the (in)applicability of conjunction reduction differentiates compounds from purely morphological combinations, it also distinguishes between compounds and phrases. To begin with, Bates (1988:225-8) points out that the semantics of conjunction inside compounds is not simply the semantics of conjunction as it is inside phrases. Conjunction in compounds assumes an extra semantic element of "equation" inside compounds. She defines an equation roughly as the sharing of a significant number of semantic features, where "signficant number" might vary from compound to compound {ibid:227). As in English,^® coordination and equation can distinguish he ‘and’ in compounding, where conjuncts must share enough semantic features, from

coordination in syntax, which is not so constrained. Thus wanju and suda can be conjoined in

a phrase, as in (39a), but not so readily in compounding morphology if a conjunction reduction

reading is intended in (39b).

(39) a. Ta mai-le wanju he suda.

He buy perf. toy and soda

'He bought toys and soda.’

b. * [wanju he suda] binggan

toy and soda cracker

‘toy crackers and soda crackers’

The string in (39b) may be grammatical, but must have the phrasal reading "toys and soda

crackers". This is because ‘toy’ and ‘soda’ in (39b) do not share as many semantic features in

making crackers as vegetable and fruit do in (37) in filling cans. Toy crackers are those shaped

iike toys (to attract children) but soda crackers are those whose ingredient contains soda. A

^^Cf. meat-and-potato eater vs. *water-and-snow skier (ibid:228), television and VCR tables vs. *television and ping-pong tables. 113 conjunction reduction reading is possible when the structure is made phrasal, i.e., with de following the coordination in (39b).

As expected, an imbalance will arise if modifiers of any sort occur on only one of the conjuncts. This accounts for the unacceptability of the intended compound readings in (40b/c) based on (40a):

(40)3. (n [n sucai] he shuiguo]] guantou]]

vegetable and fruit can

‘cans of vegetables and cans of fruit’

b. *(n [n [n sucai] he redai-shuiguo]] guantou]]

vegetable and tropical fruit can

‘cans of vegetables and cans of tropical fruit’

c. *[|,j [,^ [,^ tuoshui-sucai] he shuiguo]] [|^ guantou]]

dehydration vegetable and fruit can

cans of dehydrated vegetables and cans of fruit’

d. ??(&, [m (m tuoshui-sucai] he (m redai-shuiguo]] (m guantou]]

dehydration vegetable and tropical fruit can

‘cans of dehydrated vegetables and cans of tropical fruit’

e. ?[^ beifang-sucai] he [,^ nanfang-shuiguo]] [fg guantou]]

dehydration vegetable and tropical fruit can

‘cans of northern vegetables and cans of southern fruit’ 114

Examples in (40b/c) are acceptable only on phrasal readings ‘vegetables and cans of tropical fruit’ and ‘dehydrated vegetables and cans of fruit' respectively. Although (40d) is expected to

be good since each conjunct has a modifier, its acceptabilty is very low, because the semantic features of one modifier (fuosui ‘dehydrated’) are dramatically different from those of the others

{redai ‘tropical’). By contrast, (40e) is much better, since the two modifiers {beifang ‘northern’ and nanfang ‘southern’) share a significant number of semantic features. The string is still not

completely grammatical, maybe due to a general constraint that a W/w may not be too long. If

coordination in compounds were the same as in phrases, there would be no explanation as to

why the corresponding coordinative phrases with de are possible.

(41)a. [sucai he redai-shuiguo] de guantou

vegetable and tropical fruit de can

‘cans of vegetables and cans of tropical fruit’

b. [tuoshui-sucai he shuiguo] de guantou

dehydration vegetable and fruit can

‘cans of dehydrated vegetables and cans of fruit’

Such unbalanced coordinative readings occur elsewhere in phrasal syntax.

(42)a. Ta xihuan [sucai he redai-shuiguo]

he like vegetable and tropical fruit

‘He likes vegetables and tropical fruit’

b. Ta xihuan [tuoshui-sucai he shuiguo]

he like dehydrated vegetable and fruit 115

‘He likes dehydrated vegetables and fruit'

This shows that coordination is not as constrained by considerations of equation or parallelism in phrasal syntax as it is in compound morphology.

A phrasal modifier obligatorily alters the compound structure in (43a) and its interpretation.

(43)a. [sucai he shuiguo] guantou

vegetable and fruit de can

‘cans of vegetables and cans of fruit’

b. daliang de sucai he shuiguo guantou

many de vegetable and fruit can

‘cans of fruit and many vegetables'

‘many cans of vegetables and many cans of fruit’

# ‘cans of fruit and many cans of vegetables’

c. sucai he daliang de shuiguo guantou

vegetable and many de fruit can

'vegetables and many cans of fruit'

# ‘cans of vegetables and many cans of fruit’

d. cai mai de sucai he youren de shuiguo guantou

just buy de vegetable and attract de fruit can

'the vegetables just bought and attractive cans of fruit’

#'the cans of vegetables just bought and the attractive cans of fruit’ 116

To recapitulate, coordination in compounds is different from coordination in piirases in tfiat only the former obeys the extra equational constraint on conjuncts and respects the lexical integrity of the matrix compound. Because the UH precludes there being a coordinate phrase in a compound, my contention here is that N-he-N is itself a compound within the matrix compound. Such a coordination must satisfy all compounding requirements. The whole chunk is a W/w, its constituents are W/w’s but not phrases, and it is licensed by a general syntactic rule

XW - > HjCONJ]^ H[CONJ, (CONJFORM: he)] and a morphological rule (CR4): concatenation of XwjCONJj^ XwjCONJ, (CONJFORM: he)], it aiso bears a special semantic requirement: its conjuncts must be equational and therefore the whole coordination refers to a single entity or relation. These formal requirements differentiate a coordinate compound (i.e., the coordination a? a whole is a compound) from components of other compounds, since a component of a compound is not in general itself a compound (of. girl-friend and examples in 3.5). On the other hand, coordinate compounds are differentiated from coordinate phrases, since they occupy W- positions in syntax and observe the lexical integrity of w in morphology. Below I discuss how a distinction can be made between coordination in compounds and in phrases when structural detemination is hardly discernible, as, for example, when an apparent coordinate compound occurs in a phrasal position. Should such an entity be analyzed as a compound or a phrase?

Here the equational semantics is a rule of thumb.

Given the equational semantics of coordinate compounds, I hypothesize that the coordinations in (44) may be considered as compounds in a marked structure, but those in (45) are very unlikely to be so.

(44) xiansheng he nüshi ‘lady and gentleman'

laoshi he tongxue ‘teacher and student’

gongren he nongmin ‘factory-worker and peasant’

baba he mama ‘dad and mom’ 117

shushu he ayi ‘uncle and aunt’

shuidao he xiaomai 'rice and wheat’

sucai he shuiguo ‘vegetable and fruit’

(45) xiansheng he lingdui ‘gentleman and team-leader’

laoshi he qiangdao ‘teacher and robber’

gongren he nigu ‘factory-worker and Buddhist nun’

baba he ayi dad and aunt’

shushu he jiejie ‘uncle and elder-sister’

shuidao he shuxue ‘rice and mathematics’

sucai he dapao ‘vegetable and cannon’

The hypothesis is consistent with the fact that a compound interpretation is available from an expression in (44) but not for one in (45), after a conjunction reduction.

(46)a. [shuidao he xiaomai] yanjiu the research on rice and wheat’

[sucai he shuiguo] chanliang ‘the output of vegetables and fruit’

b. * [shuidao he shuxue] yianjiu the research on rice and mathematics’

*[suicai he dapao] chanliang ‘the output of vegetables and cannons’

This is because the string shuxue yianjiu ‘math research’ in (46b) imposes a compound structure,

and if the conjunction reduction reading is intended, then shuidao ‘rice’ has to be conjoined with shuxue "math" to form the first constituent N m the N+N compound. However, shuidao and shuxue, when so conjoined, do not satisfy the equational condition, since they share few semantic features. A conjunction reduction reading is available only if de intervenes before the last word 118 in (46b) making the coordination phrasal.

This hypothesis is also congruent with the classical treatment of the plural suffix -men as a word affix (Lu 1964:130, Chao 1968:244, Kratochvil 1968:70, Li & Thompson 1981:40, etc.).

However, Lu (1964:130-2, also cf. 2.3.8) observes that -men (and other suffixes) may also attach to "a coordinate phrase" (jiaoyuan, xuesheng-men ‘teachers, students’), giving rise to the notion of "phrasal affix". One way of maintaining the word affixhood of -men is to demonstrate that the so-called "coordinate phrase" to which -men is suffixed is simply a coordinate compound word, just as the -s attaches to the matrix compound blackbird in blackbirds. This is lx>me out indeed.

It is difficult to get a plural reading associated with the first conjunct from the items in (45), but such a reading is easy to get from items in (44).

(47)a. [xiansheng he nüshij-men ‘ladies and gentlemen’

[laoshi he tongxue]-men ‘teachers and students’

(gongren he nongmin]-men ‘factory-workers and peasants’

b. ? [xiansheng he lingdui]-men ‘gentlemen and team-leaders’

? [laoshi he qiangdao]-men ‘teachers and robbers’

? [gongren he nigu]-men ‘factory-workers and Buddhist nuns’

Expectedly, a plural reading for the first conjunct is difficult to get if a phrase occurs tietween the two conjuncts in (47a), as in (48).

(48)a. xiansheng he dai-zhe maozi de nüshi-men

gentleman and wear dur. hat de lady pi.

‘a gentleman and the ladies wearing hats’

# ‘gentlemen and the ladies wearing hats’ 119

b. laoshi he san nianji de tongxue-men

teacher and three grade de student

'a teacher and third year students’

# ‘teachers and third year students’

In (46)-(48), the presense of -men facilitates the detemination of a compound structure.

A tricky case, however, is where structural determination is not immediately discernable. For instance, the coordination in (44) may also occur in a phrasal position, such as subject position.

Interestingly, the current framework predicts that in such a case ambiguity arises as to whether the coordination has an equational compound reading in (49a) or a generic phrasal reading in

(49b).

(49)a. [|,jp [^1 Sucai] he shuiguo]]] hen you yingyang.

vegetable and fruit very have nutrition

‘Vegetables and fruit (as one of the food groups) are very nutritious.’

b. [np [np [n Sucai]] he [^p shuiguo]]] hen you yingyang.

vegetable and fruit very have nutrition

‘Vegetables and fruits are very nutritious.’

The structural ambiguity is supported by two facts. First, coordinations in (45) can hardly have a compound reading in this position, as in (50a). Second, unbalanced conjuncts may cause a loss of the compound reading, as in (50b).

(50)a. [Sucai he dapao] bu jian le.

vegetable and cannon not see prt. 120

‘Vegetables and cannons (?as a group) were missing.’

b. [Sucai he redai-shuiguo] hen you yingyang.

vegetable and tropical fruit very have nutrition

‘Vegetables and tropical fruits (?as one of the food groups) are very nutritious.’

Concerning the two special properties of coordinate compounds, I have no formal and satisfactory explanation for the equational requirement at this point, although the syntactic

unmodifiability and non-expandability follow from the UH. On the one hand, there is certainly idiolectal variation for the grammaticality judgements of coordinate compoundings among speakers, i.e., some of the syntactic non-nequational coordinations are not quite acceptable, and some of the unequational coordinations in compounding may be acceptable (including the

English examples in note 25);^® on the other hand, there is at least a difference in degree

between coordinate compounds and coordinate phrases with respect to the equational

requirement on the conjuncts of each group.^^ The following only serves as an informal and

conjectural account of such a distinction or degree difference.

Conjuncts in conjunction reduction are in general expected to have some parallel meaning

concerning their common external syntax for pragmatic reasons; otherwise, speakers need not

resort to the reduction. Coordinate phrases are largely licenced by syntax, e.g.. He is [[a

Republican] and [proud of ft]] vs. */ [[[am] and [expect]] to go]]. Put simply, a coordinate phrase

[x [y and z]] is grammatical only if the rule licensing it is the sam e as the rules licensing [x y] and

[x z]. The first English example is allowed, since the linking verb subcategorizes for any

2®lt seem s that acceptability varies with the ability of speakers to perceive the equation or parallelism in a given context (Pollard, p.c.).

^^As a suggested tendency, the more deeply a coordination is embedded in syntax, the more strictly the equational requirement would be imposed on each conjunct (Zwicky, p.c.). Since the components of a compound are located at or near the lexical level of a phrase-structure tree, coordination in compounding must mostly observe the semantic equation. 121 predicative phrasal categories XP (NP, AP, VP, or PP) (Sag et al. 1985). The second one is ungrammatical, since the rules licensing am to go and expect to go are distinct. Note that in the first case, the conjuncts a Republican and proud of it are intuitively very different in semantics, yet the coordination is fine.

it should be noted that the semantic parallelism requirement becomes more strict when conjunction reduction takes place In a compound. I suggest attributing this to morphology.

Presumably, a compound denotes a semantic entity or relation. If equational semantics in a compound is relaxed to certain extent, then the semantic integrity of the compound will be destroyed and it is hard to get a compound reading. Thus each component of a coordinate compound [x and y] ought to bear an equationally comparable relation in semantics to some other component [z] in the matrix compound. In other words, if the compounds x+z and y+z are in different semantic entities, the matrix compound [[x and y]+z] is unacceptable even though the syntactic categories of x and y are the same. By contrast, coordinate phrases do not (strictly) impose such a semantic requirement, presumably because phrases do not necessarily or exclusively express semantic entities.

Summarizing, the argument for the N+N and A+N as compounds leads to the contention that coordination in morphology merely reduces to compounding and thus creates a compound.^® In no way can two affixes be conjoined to attach to a word. Thus, the

phenomenon that a suffix may attach to a coordinate phrase is in fact a process by which a suffix

^®The A he A]+N is not discussed, since good examples of an A-he-A as a compound are hard to find. The reason may be twofold. As noted in 3.5.2, most A s in Chinese are not available for participating in the A+N compounding yet, and if they are, the equational condition is contextually difficult to satisfy between two A s. One such example is

i. [n (a U wennuan] he shirunj] j,^ diquj] warm and moist area ‘a warm and moist area’ # ‘a warm area and a moist area’

However, most speakers cannot get the conjunction reduction reading but only get the mixture reading, parallel to hei-bai dianshi and its translation black-and-white television in English. 122 attaches to a compound word.^ In addition to particular syntactic and morphological requirements, coordinate compounds are distinguished from coordinate phrases in that the former must meet the semantic equational condition on each conjunct.

3.7. Summary of Theoretical Positions

This chapter has outlined a notion of compounding which occupies a special status in universal grammar. A compound is both syntactically and morphologically composite, i.e., a syntactic/morphoiogicai word containing two or more such words. From this position, the investigation of Chinese compound-like structures has led to some theoretical conclusions concerning the organization of grammar.

The analysis of some VO-idioms with respect to the (in)separability of their comp>onents not only supersedes C-T. Huang's PSC and maintains the universal UH but also yields a corollary hypothesis that lexical status of a sequence of W/w’s is determined by both syntax and morphology. This hypothesis is further substantiated by the study of the R-compound and the

R-construction. I summarize the relevant theoretical positions or assumptions as follows, and propose that they be taken as linguistic universals, subject to further scrutiny in Chinese and other human languages.

(51) Position 1: The compoundhood of a concatenation of Ws is both syntactically and

morphologically licensed. A sequence of Ws is a compound if syntax is able to

categorially describe its internal structure and to place the sequence in a position calling

for rank W, and morphology imposes lexical integrity on it.

^Given the notions of word and syntactic rank, it is doubtful that "phrasal affix' has a status in universal grammar, cf. Zwicky’s (1987b) analysis of the possessive -’s in the boy's toy as a word inflectional affix in English, as opposed to Nevis’s (1988) phrasal affix treatment, and the argument for the Chinese sentential particles as words rather than phrasal affixes in chapter 7. 123

(52) Position 2: While a compound is describable in terms of the interface between syntax and

morphology, the formal similarities and differences between the compound and its

putative corresponding phrase are captured by rule relationships, l.e., the degree to

which the formal requirements of the syntactic rule licensing the compound overlaps with

and deviates from the requirements of the rule licensing the phrase.

I take Position 1 as providing both necessary and sufficient conditions on compounding, enforcing the significant generalization in Position 2. Any possibie iexicai gaps and semantic or phonological idiosyncrasies follow from the morphological properties of compounding.

The synchronic quadrisyllabic N+N and A+N combinations, traditionally taken for granted

as phrases, have been shown to be compounds. They have lexical gaps and meaning specializations, with their components unseparable and unable to receive phrasal modifiers.

Although the components are restricted to bisyllabic words in this study, the methodology is valid

in other cases. The investigation leads to a position on coordinations.

(53) Position 3: Coordination as a syntactic operation cannot apply in morphology except for

compounding. Coordination in morphology creates a compound within a compound, and

conjunction in compounds consists semantically of coordination plus equation of

conjuncts.^

Little evidence is available for Chinese compounding as the reflection of the organization

of its phrasal syntax, presumably in support of the view of morphology as the syntax of words.

The idea is misleading, since it downplays two crucial differences between compounds and

phrases, as stated below.

^A problem for this universal is whether anti- and pro- in anti- and pro-democracy is a compound, and whether the result Is also a compound in English. 124

(54) Position 4: Syntactic words in a compound behave differently from those in a phrase in

that the former must observe the lexical integrity of the compound, and a compound as

a morphological word may exhibit lexical, semantic and/or phonological idiosyncrasies.

Most of the Chinese bimorphemic lexemes were historically compounds but are no longer compounds in the modern language, since their components are bound. I have submitted that they may consist of more than one stem. Lexical redundancy rules relate the stems in different lexemes containing them.

Before closing this chapter, I would like to consider the range of compoundings in

Chinese. As mentioned earlier, the R-compounding licenses an open-ended list of compounds.

It is noted that the N-l-N and A4-N compoundings are not only productive but r'so recursive:

N-t-[N-t-N] Shanghai-shuiguo-guantou [Shanghai-fruit-can] 'a can of fruit made in Shanghai’,

N+[A+N] redai-xinxian-shucai [tropics-fresh-vegetable] tropical fresh vegetable", [N+N]+N: shuiguo-guantou-zhanlan [fruit-can-exhibition] ‘an exhibition of cans of fruit", etc. Furthermore, the variation in acceptability of the so-called lexical gaps suggests only that speakers do not usually admit a putative combination when they first encounter it. For instance, the A+N Ixinxian-xueshui

[fresh-snow-water] "fresh snow water" is an awkward compound at first glance in (29). However, it can legitimately occur as a compound-copy of the preceding phrase xinxian de xueshui in the discource below.

(55) Xinxian de xueshui hen shao, keshi Zhangsan zuotian zai shanshang que zhaodao-le

yizhong xinxian-xueshui.

fresh de snow-water very rare but Zhangsan yesterday at hill yet find a-kind-of fresh snow­

water

‘Fresh snow water is rare, and yet Zhangsan found some type of it on the hill yesterday." 125

Thus those "unacceptable" compounds are readily accommodated provided suitable contexts, as are those "unacceptable" coordinate compounds.

The productivity and recursiveness of compounding thus "produces" an unlimited number of compounds in a language. This could pose a problem for the point of view of full lexicon, the size of which would be unboundedly large. However, this need not be the case. First, consider the same problem in phrasal syntax. Since the number of possible rule-governed syntactic concatenations is infinite, they are not listed in the syntax. However, idioms and perhaps syntactic selections should be iisted, presumably either in the syntax or in the specific lexical entries which the relevant syntactic rules invoke. The same is true of compounding. On the one hand, since compounding is partly licensed by syntactic W-rules, the productive/recursive compounds associated with compositional semantics need not be listed in the lexicon, on a par with phrases in the syntax. On the other hand, compounds involving semantic/phonological idiosyncrasies are listed as lexemes, parallel to the listed idiomatic phrases and syntactic selections.

This metaconsideration bears on the question of the degree of range and productivity of

Chinese compounding, which was raised in 3.1. Although I have excluded many compound-like

expressions, thus substantially reducing the number of compounds in the traditional sense, the

newly discovered productive compounding of nominals indicates that Chinese is still rich in

compounds. CHAPTER IV

DERIVATIONAL MORPHOLOGY

4.0. Introduction

Derivational morphology as defined in 1.3 relates the stems of different lexemes, in contrast to inflectional morphology, which relates the stem of one lexeme with its forms. Hence derivational morphology is of lexical function, relating a new word to (or deriving it from) the source word, as with nation-al vs. nation in English. The crucial difference between derivation and inflection is that syntax does not require the presence of a derivational morpheme in a word as it may an inflectional morpheme.^

Derivational morphology in Chinese has been treated under the general heading of morphology in the literature (see Lu 1964:113-33, Kratochvil 1968:68-9, Chao 1968:194-257, and

Li & Thompson 1981:28-44). This seems to be a sub-area where the classical view holds firmly:

Chinese has little or no morphology. Chinese has few affixes (Li & Thompson 1981:36), and they are not "interesting" because of a lack of morphophonological alternations. This chapter discusses two types of morphological operations: reduplication (4.1) and affixation (4.2). Three rules involving reduplication are the verbal delimitative, the adjectival vividative and the distributive measures. I will investigate whether reduplication is morphological or syntactic, and if

^As a reminder, cases of inflectional morphology do not violate the Principle of Morphology- Free Syntax, since, more precisely, syntax requires only the presence of certain features located somewhere in a construction, and these features are sometimes realized as inflections on words. For instance, the English passive construction {be kissed) requires the head of the second constituent VP to be in the past participle form. However, whether the form is realized by suffixation of -ed {kissed) or -en {known), or by a vowel change {hung), or by both {written), or by assuming the base form {cut), is irrelevant to the syntactic rule and determined only by the morphology.

126 127 morphological, whether it is syntactically relevant or not. Of some parochial importance is a constraint on regular reduplications, which must copy the entire phonological material of a stem.

Affixation is characterized according to five universal tendencies, and the inventory of affixes in

Chinese, as well as the role they play in the grammar, is consequently expanded.

Morphophonological alternations resulting from reduplication or affxation, primarily tonal changes, will not be discussed, and readers are referred to Chao’s (1968) work.

4.1. Reduplication

Following Z-S. Zhang (1987:367-88), I assum e that reduplication is a type of operation a morphological rule may invoke. While tl le existence of syntactic reduplications in Chinese or other languages is an empirical question, I will examine three cases of reduplication in Chinese to show that the operation is not syntactic but morphological.

Reduplication for some Ws is so sporadic that no one general rule can be said to account for all cases of reduplication, such as hong-tong-tong [red-?-?] 'red', lü-you-you [green-?-?]

‘green’, etc. Since îong and you are bound stems, this relation should be stated in the morphology. The same can be said of the kinship terms like ba-ba ‘father’, ma-ma ‘motheri, mei- mei ‘younger sister‘, etc. Though the occurrence of reduplication for kinship terms is quite general, the reduplicated components are bound for most speakers. Reduplication is invoked by three other morphological rules: the delimitative of volitional verbs, the vividative of adjectives, and the distributive of measures.

4.1.1. Delimitative of Volitional Verbs

Reduplication is invoked by the rule relating a volitional verb (more precisely, a verbal stem) to its delimitative aspect (Chao 1968:204, U & Thompson 1981:29), which adds the meaning "a little', or more precisely, for a transient period", e.g., chang ‘taste’ vs. chang-chang

[taste-taste] ‘taste a little‘, xue-xue [learn-learn] learn a little’, xie-xie [write-write] ‘write a little’, pao- 128 pao [run-run] ‘run a little’, etc. According to traditional grammar, the putative morpheme yi ‘one’ is optionally inserted between the original and its duplicate without changing the meaning, e.g., chang-yi-chang [taste-one-taste] ‘taste a little’, xue-yi-xue [learn-one-learn] learn a little', etc.^

The rule also applies to bisyllabic verbs, but with two restrictions: the infixal yi is not allowed and the whole stem must be reduplicated, e.g., zhuyi-zhuyi [pay-attention-pay-attention]

‘pay a little attention’, vs. *zhuyi-yi-zhuyi, *zhu-zhuyi, *zhuyi-yi, taolun-taolun [discuss-discuss]

‘discuss a little’ vs. *taolun-yi-taolun, *tao-taolun, *taolun-lun, shangliang-shangliang [consult- consuit] ‘consult a little’ vs. *shangliang-yi-siiangliang, -shang-shangiiang, -shangliang-liang, wenhou-wenhou [greet-greet] ’greet a little’ vs. *wenhou-yi-wenhou, *wen-wenhou, *wenhou-hou, etc.

All previous studies take it for granted that reduplication is somehow morphological, but it should be noted that the door to a syntactic analysis is not firmly closed. This is because it is the whole stem, or VW (= verb or verbal rank W, as opposed to VP), not just part of it, that is copied. In addition, Chao (1968:205) regards the delimitative as straddling the border between

morphological and syntactic constructions, since the distribution of the second copy -kan, as in

(1), is the same as those for what he calls (phrasal) 'cognate objects' yixia, san ge zhongtou and yi yan, as in (2).^ Here I take a transitive verb kan ’look’ as an example, collecting as many cognate objects as possible. These cognate objects in fact function as a sort of measure

adverbial expression in Chinese. They are indicated with a set of square brackets enclosing the

measure adverbial along with its preceding verb.

(1)a. kan-kan nei ben shu

look-look that M book

^For convenience, I temporarily assume that yi is the numeral W 'one', although I later argue that this is not the case.

3He also treats yi-kan in (1c) as a cognate object. 129

‘take a look at that book’

b. kan-le-kan nei ben shu'*

c. kan-yi-kan nei ben shu

(2)a. (kan yixia] nei ben shu

look once that M book

'take a look at that book’

b. [kan-le san ge zhongtou] (de) shu

look perf. three M hour (de) book

‘read the book for three hours’

0 . [kan yi yan] nei ben shu

look one eye that M book

‘take a glance at that book’

The goal here is to show that similar distributions between the second copy of the delimitatives in (1) and cognate objects in (2) are apparent. Specifically, the combination of a cognate object and its verb is a syntactic concatenation, but the combination of a reduplicate and its preceding verb cannot be syntactic and must be morphological.

I start with an analysis of the cognate construction in (2) by assuming position (3), taken from Zwicky (1990:204).

'*The perfective -le is an inflectional suffix elsewhere (cf. 5.4). I will consider its infixation status in 6.0. 130

(3) An adequate grammar must allow for rank-shift (i.e., among word, phrase and clause)®

in particular languages, according to which one rank is stipulated as functioning as

another which is prototypically associated with the construction.

As a pretheoretical rewording of (3), words can function as phrases and vice versa. In Chinese

as well as in English, the word rank wo T always occupies the position which is prototypically filled by a (noun) phrase like xuexiao de laoshi ‘the school’s teacher", hence the term "word-

phrase" (ibid).^ Likewise, but in an opposite direction, a relative clause, as in (4a), is in a position

prototypically occupied by an adjectival phrase, as in (4b).

(4)a. nei ge [zuotian lai de] xuesheng

that M yesterday come de student

‘the student who came yesterday’

b. nei ge [youxiu de] xuesheng

that M excellent de student

‘that excellent student'

Similarly, in the resuitative construction (cf. ch. 5), a clause, as in (5b), can function as a phrase

which occurs prototypically in the complement position of the construction, as in (5a).

(5)a. Ta ku-DE [hen shang xin].

he cry DE very wound heart

^Current syntactic theories regard clauses as a subtype of phrases. In such a phrase, the lexical head is semantically predicative and all its obligatory argum ents are realized by constituents within the phrase.

®Note that a formal theory would posit that the phrase has a single constituent W. 131

‘He cried very sadly.'

b. Ta ku-DE [women hen shang xin].

he cry DE we very wound heart

‘He cried so that we were very sad.'

Given such motivation for rank-shift in Chinese, I submit that the bracketed verb-cognate object Cv'-CO) combination in (2) functions as a verb in the sense that the syntax pieces the V-CO in the sam e VW position as in the verb-object (V0-) construction. Here the V-CO, like a VW, takes a direct object, and hence the term "phrase-word" (Bloomfield 1933:180, Zwicky 1990:205). The

V-CO has the internal constituent structure of a VP but external syntax of a VW. Expected from this analysis are similar syntactic behaviors between a V-CO and a prototypical bare VW in the

VO-construction on the one hand, and between the V-CO-object and the prototypical VO- construction on the other hand. Consistent with the VO-construction (cf. 2.2.2), the direct object in the V-CO-object construction can be extracted, as in (6a), on the example of (2b). But the combination of CO and the object is not extractable, as in (6b). This shows that the CO-object

string is not a constituent even in the presence of de.

(6)a. Shu, ta kan-le san ge zhongtou.

book he look perf. three M hour

‘He read the book for three hours.'

b. *San ge zhongtou (de) shu, ta kan-le .

Moreover, the V-CO-object can be displaced as a constituent like the preposed VO in the VO-

construction, as in (7b), in the presence of a preverbal adverbial, as in (7a). 132

(7) a. Ta zai nar kan-le san ge zhongtou (de) shu.

he at there look perf. three M hour (de) book

‘He read the book for three hours there.’

b. Kan-le san ge zhongtou (de) shu, ta zai nar.^

Finally, the structure participates in deletion and pronominalization rules as the VO-structure does, as in (8).

(8)a. Zhangsan mamahuhu de kan-le san ge zhongtou (de) shu, Lisi daoshi hen renzhen de.

Zhangsan careless de look perf. three fvl hour (de) book Lisi but-be very serious de

‘Zhangsan read carelessly for three hours, but Lisi did so seriously.’

b. Zhangsan mamahuhu de kan-le san ge zhongtou (de) shu, Lisi daoshi renzhen de zuo de.

Zhangsan careless de look perf. three M hour (de) book Lisi but-be serious de do de

‘Zhangsan read carelessly for three hours, bu Lisi did seriously.’

All these indicate that V-CO, like V in the VO-construction, is isolated to some degree from

both the postverbal direct object NP (via (6)) and the preverbal adverbial (see (7) and (8)), and

that V-CO-0 is a constituent (cf. (7)). Moreover, since V-CO can serve as an answer to a

question, it is itself a constituent. Thus V-CO behaves like and functions as a VW in that it fills

the VW position in the VO-construction, and the V-CO-0 as [^p V-CO] Ojj can be treated as

a subtype of the VO-construction.

Recall however that functioning as a W does not necessarily entail morphological word

^Since ta zai nar [he at there] is not a constituent in (7a), a rightward displacement is untenable. 133 status (w). Insofar as a phrase-word internally meets the conditions for a syntactic construction, it can participate in that construction (Zwicky 1990:206). For instance, CO can be displaced, as in (9a) cf. (2c), and V-CO can be expanded, as in (9b) cf. (2b).

(9)a. kan-le nei ben shu yi yan

look perf. that M book one eye

‘take a glance at that book’

b. kan-le zhengzheng san ge zhongtou (de) shu

look perf. whole three M hour (de) book

‘read the book for a whole three hours’

c. kan-le yi yan nei ben shu

d. *kan yi yan-le nei ben shu

This demonstrates that the V-CO in (2) is not a word or a compound in the morphological sense at all, because the UH is not satisfied for V-CO, as illustrated in (9), even though the V-CO functions as a W syntactically (cf. 3.3). It follows that -le as a verbal inflection attaches to the V kan but not to V-CO *kan yi yan-le, as in (9c/d), because it is the morphological w kan, rather than the syntactic W kan yi yan, that constitutes a morphological domain for the inflectional operation.

Now I return to the delimitative reduplicates in (1), repeated here, kan-(le/yi)-kan nei ben shu [look-look that M book] ‘take a look at that book’. The structure of V-reduplicate-object in (1) is similar to the V-CO-object construction in (2) in the following ways: a) the direct object can be fronted; b) the second copy of the reduplicate plus the direct object cannot be fronted; c) the whole structure can be fronted if there is a preverbal adverbial expression, or it can at least occur alone as a constituent; and d) the whole structure can be deleted or pronominalized in discourse. 134

(10)a. Nei ben shu, wo kan-kan

that M book I look-look

‘Let me take a look at that book’

b. *kan nei ben shu, wo kan-

look that M book I look

'take a look at that book’

c. ?kan-le-kan nei ben shu, ta bu naifan de

look perf. look that M book he not patient

’He took a look at that book impatiently.’

d. Zhangsan mamahuhu de kan-le-kan nei ben shu, Lisi daoshi hen renzhen de.

Zhangsan careless de look perf. that M book Lisi but-be very serious de

’Zhangsan took a look at that book carelessly, but Lisi seriously.’

Zhangsan mamahuhu de kan-le-kan nei ben shu, Lisi daoshi renzhen de zuo de.

Zhangsan careless de look perf. that M book Lisi but-be very serious de do de

’Zhangsan took a look at that book carelessly, but Lisi did seriously.’

Even though the V-reduplicate has apparently the same linear position and yields the same

results on the constituency tests as the V-CO, as illustrated in (10), a crucial difference between the two structures is that the V-reduplicate must respect the UH, since their parts are inseparable.

See the ungrammaticality caused by displacement in (11 a) and by expansion in (11b), in contrast with the good cases for V-CO in (9). 135

(11)a. *kan-(le) nei ben shu -(yi-)kan

look (perf.) that M book (one) look

b. *kan-le zhengzheng (de) -(yi-)kan nei ben shu

look (perf.) whole (de) (one) look that M book

The conclusion I reach is that the reduplicated delimitath/es are not phrase-words, but rather v/crds in both a syntactic and a morphological sense.

Note that y; would have to be infixed properly in kan-yi-kan. However, if an item is said to be an affix, it must occur always bound; but the numeral yi ‘one’ is a free W elsewhere (cf.

2.3.5). There is one way out, however, involving a denial of the putative synchronic connection between yi in the reduplicates and the numeral yi. I propose that a homophone -yi- is stipulated as an affix, independent of the numeral yi, in the grammar. The fact that the affix -yi- has its historical origin in the numeral yi and that the two are etymologically related for most speakers® are not the point at issue. There is no compelling reason why -yi- must mean ‘one’ in order to contribute semantically to the delimitative "a little" or '"for a transient period", let alone the fact that the existence of the putative ‘one’ is optional. Moreover, such a stipulation is independently needed, because there are dozens or hundreds of chengyu ("established phrases" or "set phrases") which contain yi: yibazhua ‘take everything into one’s hands’, yichang-yihe ‘sing a duet with someone’, yidao ‘together’, yigai ‘without exception', etc.® It is difficult to compare these "set phrases" with any similar phenomenon in European languages; the closest scenario would be the

Latin phrase prima facie used in modern English, according to Kratochvil (1968:81-2). Chengyu

®This is due to the homophony and homography of the two items for historical reasons.

^Hundreds of these can be found in a standard dictionary. For example, over 250 entries beginning with yi- are listed in Xiandai Hanyu Cidian [Modern Chinese Dictionary], Zhongguo Shehui-Kexueyuan Yuyan-Yanjiusuo Cidian-Bianjishi [Dictionary Compling Section, Linguistics Institute, the Chinese Academy of Social Sciences], Beijing: Comercial Press (1979). 136

Should be analyzed morphologically, and most of the yPs in question are etymologically related to, but bleached of the meaning of, 'one'. Therefore, the proposal that -yi- is an independent bound morpheme (affix) in the delimitative is not an unreasonable one.^°

My generalization involves two rules. The first one states that a delimitative verbal stem contains a reduplicate of its corresponding non-delimitative stem; the second one has an additional monosyllabic condition on the target stem and the operation of y/-infixation. This is a static or declarative version correspondingto a dynamic or procedural rule such as: reduplicate a verbal stem with an optional -yi- if the stem is monosyllabic. Recall that a stem is generaiiy a

W/w in the language since inflections are not required in every construction. Thus both the input and the output of the rule is a W/w in the language.

The observation that reduplication works at the stem-level but not at the sub-stem level bears on whether a given VO or verb-resultative expression is a phrase or a lexeme. Recall that in the discussion of the ionization phenomenon (3.3), the VO phrase dan x/n [carry-heart] ‘worry’ must be analyzed as a W/w in a marked construction (e.g., typically when it takes an object), as in (12a). The current analysis thus predicts that as a W/w its first part cannot be reduplicated for the delimitative, as in (12b). Instead, the whole W/w must be copied, as in (12c).

(12)a. Ni yinggai dan-xin zhe-jian shi.

you should worry this matter

‘You should worry about this matter."

b. *Ni yinggai dan-dan xin zhe-jian shi.

c. Ni yinggai dan-xin-dan-xin zhe-jian shi.

^ ^ h e case is on a par with words ship or like vs. the corresponding affixes -ship (friendship) or -like (mother-like) in English. 137

By constrast, the verbal component of a corresponding VO phrase in (13a) can participate in reduplication alone, by virtue of its wordhood, as in (13b). The whole "VO" can also be copied, as in (13c), which enforces a (word) stative reading.

(13)a. Ni yinggai wei Zhangsan dan xin.

you should for Zhangsan worry heart

‘You should worry about Zhangsan.'

b. Ni yinggai wei Zhangsan dan-dan xin.

c. Ni yinggai wei Zhangsan danxin-danxin.

Finally, the delimitative should be considered as derivational rather than inflectional morphology by default, because syntax never requires its presence. Implicit here is the proposal that the delimitative is not a syntactic feature in the language. It is understood only semantically, since Chinese has other means of expressing "a little" or fo r a transient period", as in (2a) and

(2c).

4.1.2. Vividative of Adjectives

Adjectives may have vivid reduplicates (Chao 1968:205), which 1 call the vividative for short; jieshi 'solid' vs. jiejieshishi strong and solid’, qingchu 'clear' vs. qingqingchuchu 'perfectly clear’, kuaihuo 'happy' vs. kuaikuaihuohuo 'happy and cheerful", pingchang 'usual' vs. pingpingchangchang 'quite usual and ordinary', guiju 'well-behaved' vs. guiguijuju 'well-behaved and well-mannered', ganjing 'clean' vs. ganganjingjing 'perfectly clean', zhengqi 'orderly' vs. zhengzhengqiqi 'orderly and neat', etc.^’

^ ^ It is known that reduplication often has iconic function for communication, mostly associated with plurality, intensification, distributive, iterative, etc. However, the two reduplication cases discussed here do not function iconically. Simliar cases are indeed found in other languages, 138

It should be noted that in the copying process of the vividative, each syllable is reduplicated: *jieshijieishi, *qingchuqingchu, etcJ^ Therefore the vividative must be XXYY,^^ but not XYXY as the delimitative found in 4.1.1. Furthermore, the whole source word must be copied, not merely part of it: *jiejieishi, *jieshishi, *qingqingchu, *qingchuchu, etc. In other words, both of the syllables X and Y must be copied as XXYY; the reduplication of one of them will not do: *XXY and *XYY. This distinction sheds light on the status of resultative compounds and phrases. It is observed that the complement in the R-phrase can be reduplicated as a vividative owing to its wordhood, as in (14), but not in the R-compound, because parts of a W/w cannot be reduplicated, as in (15).

(14)a. Ta de yianjing kan-DE qingchu.

he de eye see DE clear

‘His eyes see clearly.'

b. Ta de yianjing kan-DE qing-qing-chu-chu.

(15)a. Ta kan-DE-qingchu naben shu.

he see-DE-clear that book

e.g., the verbal jaytil-y- to get to be twilight, evening', yakkyyl-y- to look up' vs. the nominal jaatija7iit-y- ‘Evening Star’, ?akkal?akaal-y- ‘roof of the mouth’ in Northern Sierra Miwok (Callaghan 1987b:301).

^^In a transformational framework, the vividative could be described as a word formation derivation: XY > XXYY, namely, each of the syllables of the bisyllabic source word is reduplicated. The notion of "underlying forms' (XY) of the vividative is unwarranted: as first observed by Chao (1968), in cases like guiguisuisui ‘sly and stealthy’ vs. *guisui, and kukutiti ‘with sobs and tears’ vs. *kuti, the source words do not exist. In a stative framework, a generalization is stated in the presence of a source word, but there is no need to pose a non-existent underlying form.

^^1 won’t discuss the XXY reduplicates like bengbengcui ’cranklingly crispy’ and XYY reduplicates like lengbingbing ‘cold as ice’, as well as monosyllabic adverbial expressions XX changchang ‘often’ and yuanyuan ‘from afar’. This is because tfiey are too rare for a general rule to describe. 139

‘He was able to see that book clearly.’

b. *Ta kan-DE-qing-qing-chu-chu naben shu.

Now I discuss whether the reduplication serves to realize some syntactic feature.

Kratochvil (1968:78) believes in the "grammatical function" of the vividative reduplication, although without giving reasons. Chao (1968:207-9) observes its "syntactic limitations". It is argued here that none of these limitations are syntactic, or they are at least not syntactically relevant enough to warrant an inflectional analysis of the vividative.

Illustrating his point, Chao notes that the output of the reduplication is restricted to adjectives (or adverbs). This is nothing unusual, since every morphological rule is restricted to some part of speech (cf. Aronoff’s (1976:47) Unitary Base Hypothesis, i.e., the base to which an affix may attach belongs to a single part of speech, and the result from the affixation also normally belongs to a single part of speech.). What qualifies a morpheme as an inflection is that its presence (or absence) is accessible to syntax. Therefore the point at issue is whether or not the reduplication is the realization of certain syntactic features. Chao observes that the vividative adjectives are distributed in all of the five syntactic constructions where non-reduplicated adjectives occur. Moreover, it is hard to imagine that reduplication could be the only way of expressing vividness in a language. Therefore such a morphological reduplication is not syntactically relevant.

Chao (1968:209) also obsen/es that the vividative does not usually take an adverb of degree. However, such an incompatibility is not syntactic, but semantic, pragmatic or functional.

First, there are exceptions to the restriction, indicating that the restriction is not syntactically general. As Chao himself notes, tai manmantengteng(-le) ‘too slow and sluggish" is good, and

I believe that zui ganganjingjing(-le) ‘most clean and neat’ is grammatical as well. But the vividative is incompatible with the adverbs hen ‘very’ and gang ‘even more’. By taking gaoxing 140

‘glad’ as an example, I illustrate the distinction: hen gaoxing ‘very glad’ vs. *hen gaogaoxingxing and gang gaoxing ‘even gladder’ vs. *geng gaogaoxingxing. However, such an incompatibility seems to have nothing to do with syntax, since it arises even in discourse, as in (16).

(16)a. 7Ta jintian gaogaoxingxing, erqie hen gaoxing.

he today glad-and-cheerful and very glad

b. TTa jintian xian gaogaoxingxing, houiai gang gaoxing.

he today first glad-and-cheerful later even gladder

The vividative apparently cannot co-occur with the negator bu, hence bu gaoxing ’not glad’ vs.

*bu gaogaoxingxing. But as Chao observes, this is because bu precedes a reduplicate only when the former carries a contrastive sense rather than an assertive predication, and hence (17a) and

(17c) are grammatical, while (17b) is not.

(17) a. Wo bu gaogaoxingxing, nandao hai kukutiti?

I not glad-and-happy can’t still cry-and-weep

If I did not appear to be happy, should I be sobbing instead?’

b. *Wo bu gaogaoxingxing.

c. Wo bu gaoxing.

It should be noted that the so-called contrastive predication should be understood pragmatically.

It is not a syntactic feature because the the pattern illustrated by the second sentence in (17a)

is not limited to particular syntactic constructions, and as for the effect of contrastiveness, the first

sentence does not have to contain a reduplicated vividative. 141

Since the presence of reduplication is irrelevant to syntax, the vividative reduplication belongs to the domain of derivational morphology. This conclusion is consistent with the unproductivity of the process. For example, while the following adjectives can be reduplicated: jiandan ‘simple’, chengshi "honest", guiju ‘well behaved', pingfan ‘commonplace", their antonyms do not have the vividative: fuza ‘complex’, y/ao/jua ‘sly’, yeman ‘barbarous’, zhongyao ‘important’.

It is this sort of exception that leads Li & Thompson (1981:33) to believe that there may be no rule determining which adjectives can be reduplicated and which adjectives cannot. This is confirmed by comparing synonyms: piaoliang ‘pretty’ has the reduplicate, but me/// beautiful’ does not.

Exceptions are usually unproblematic for inflections, but if there are too many exceptions, then the rule invoking the vividative reduplication is probably not inflectional, or accessible by a general syntactic rule.

4.1.3. Distributive Measures

Measures (M) or classifiers are words, since in the numeral-measure-noun combination, numeral yi ’one’ may be missing and phrases can intervene after the measures (Lu 1964:16).

(18) you yi tiao yu vs. you tiao yu vs. you yi tiao da de yu

have one M fish vs. have M fish vs. have one M big de fish

‘have a fish’ vs. have a big fish’ vs. have a big fish’

Although the deletion of yi here is a diachronic process, synchronically, (18) shows that the

measure morpheme tiao can pass the expansion test (cf. 2.3.7).

Reduplication is called on regularly by measure words to express distributivity (Chao

1968:202): ge vs. ge-ge (a generic classifer), tiao vs. tiao-tiao (for long things), // vs. //-// (for grain­

like things). They denotes every member of the whole NP class. Some etymologically related

nouns can also be thus reduplicated, if they are functionally measures (or classifiers): jia ‘home’ 142

VS. jia-jia ‘every home’, ren ‘person’ vs. ren-ren ‘every person’, tian ‘day’ vs. tian-tian ‘every day’.

In this sub-section, I will argue for the reduplication as a morphological process on the one hand, and argue against the distribution of reduplicated measures as syntactically governed on the other hand.

The reduplication of measures is assum ed to be a morphological process in the literature.

For instance, distributive reduplicates are stated as compounds in Kratochvil (1968:78). In the current framework, the operation must be morphological not only because the reduplicates occur where the syntax caii for a measure word, but also because the components of the reduplicates are inseparable. More importantly, since bisyllabic measures cannot be reduplicated: jialun

‘gallon’ vs. mei jalun ‘every gallon’ vs. *jialunjialun, gongjin ‘kilogram’ vs. mei gongjin ‘every kilogram’ vs. *gongjingongin (Chao 1968:203), a syntactic account for these combinations is untenable, for no syntactic rules are known to place phonological preconditions on its input (cf. the Principle of Phonology-Free Syntax in 1.3), e.g., that a topicalized NP must be monosyllabic or begin with a bilabial sound.

Having justified that this type of reduplication is morphological, I will argue that the distribution of reduplicated measures are not syntactically governed. My proposal contrasts with

Kratochvil (1968:78) and Chao (1968:77, 204), who think that the distributive m easures bear grammatical relevance. While Kratochvil gives no description, Chao observes that the reduplicates must have definite reference and never occur at the end of the sentence, as in (19a).

(19)a. *Wo rende ge-ge xuesheng

I know M-M (every) student

b. Wo ge-ge xuesheng rende

c. Ge-ge xuesheng wo rende. 143

But Chao may not be correct on this point, since the reduplicates in (19a) are not really in the final position of the sentence. Moreover, the reduplicates below could be said to be in a position comparable to those in (19a), and yet are fully grammatical.

(20) Wo rende ge-ge xuesheng de shubao.

I know M-M student de bag

I know every student's bag.’

The first point at issue is how to account for the contrast between (19a) and (19b/c) on the one hand, and between (19a) and (20) on the other hand. It is noted that ge-ge does not directly modify the object ‘bag’ in (20). Syntactically, it appears that if the reduplicates directly

modify a postverbal object, the result is ungrammatical, as in (19a). Therefore, grammaticality would be restored if the postverbal object for (19a) occurs preverbally with some markedness, as

in (19b/c). Example (20) is grammatical, presumably because the distributive does not directly

modify the postverbal object ‘bag’.

The second point at issue is whether the syntactic distributions of the reduplicated

measures just stated are correct. Unfortunately, this putative syntactic generalization is false, as

one can find counter-examples.

(21) Ta shi-le zhong-zhong banfa.

he try perf. all-kind way

‘He tried each way.’

Some speakers also allow dao-dao shangang [range-hili] in that position. Furthermore, the

ungrammaticality of (19a) may be due to its incompleteness for discourse purposes. If (19a)

becomes part of the sentence, then the result is acceptable. 144

(22) Wo rende ge-ge xuesheng, danshi meiyuo yi ge xuesheng rende wo.

I know M-M student but not-have one M student know I

'I know every student, but none of them knows me.’

There is one more complication. If the reduplicates are preceded by the numeral yi ‘one’, the result is grammatical.

(23) Wo rende yi ge-ge xuesheng.

I know one M-M student

"I know every student.’

One might say that in the presence of ’one’, the NP is indefinite so that the reduplicates can occur

in that position. This in turn bears on the putative correlation of definiteness to a preverbal

position, as implicit in Chao. However, as argued in Light (1979:150-2), definiteness does not

determine the position of an NP relative to the verb in syntax. Since definite NPs can also occur

postverbally, previous studies do not show that definiteness is a syntactic feature in Chinese.

The best evidence for the sensitivity of reduplicated distributive measures to syntactic

positions would be that (19b) and (19c) are distinct syntactic constructions requiring the explicit

presence of the reduplicates. However, since mei ’every’, also definite, can occupy the same

position, as in (24b/c) below, it appears that syntax does not uniquely call on the reduplicates.

(24)a. Wo rende mei ge xuesheng

1 know every M student

b. Wo mei ge xuesheng dou (all] rende

c. Mei ge xuesheng wo dou [all] rende. 145

In summary, while the reduplication of measures is a morphological process, the distribution of such reduplicates is not syntactically determined. It is not clear to me at this point why mei can be in the postverbal position in (24a) but ge-ge cannot, as in (19a). Perhaps it is due to some particular discourse function of ge-ge. Whatever reasons there might be, a syntactic distributional account for (19a) is untenable. Therefore, the distributive reduplication cannot be analyzed as inflectional on a par with those morphemes discussed in chapter 5, since syntax has no access to it.

4.2. Affixation

The predominant view that Chinese has little morphology is perhaps due to the misconception that the language is poor in affixes. This is untrue. For example, Chao (1968:211-

57) presents dozens of affixes in Chinese, quantitatively comparable with those in English.

However, when the term affix is used, a conceptual problem arises which is not restricted to

Chinese, but is of some cross-linguistic significance. Dai (1990a:27-9), in arguing that some elements which are historically words can be analyzed as affixes in modern Chinese, notes that the notion affix is often taken for granted rather than being strictly defined in the literature.^'*

First, I observe five tendencies which could be used to distinguish affixes from sterns.*^ Then

I proceed from there to point out problems encountered and to characterize the term affix.

1) Affixes are (always) BOUND. The affixes in Chinese assum ed in the standard texts are

always bound, e.g., lao-, a noun formative (n.f.), in lao-hu [n.f.-tiger] tig e r, lao-shu [n.f.-mouse]

"mouse", iao-shi [n.f.-teacher], and -xing "nature, -ity" in ren-xing [person-nature] "humanity", kexue-

xing [science-nature] "scientificity", etc. The problem of using boundedness to characterize affixes

is two fold. On the one hand, it is not clear whether mono- and bi- in mono- and bi-syllabic, pro-

*'*For instance, see standard textbooks written by Bloomfield (1933), Hockett (1958), Matthews (1974) or Bauer (1988).

*®A similar but independently proposed defintion of affixes in Chinese can be found in Packard (1990:36), which I believe to be vague and not sufficiently elaborated. 146 and anti- in pro- and anti-democracy in English are affixes. If they are affixes, then the free stem s are missing in the first conjuncts, and the affixes mono- and pro- are free in the first conjuncts.

But there is possibility of analyzing mono- and bi- as a coordinate compound within the compound mono- and bi-syllabic (cf. 3.6). I do not further discuss the issue here but simply note the problem for this criterion.

The other side of the story is that boundedness is not unique to affixes. There do exist bound stems (cf. 2.4.2 & 3.2), such as mit in transmit, permit, commit; fer in prefer, confer, refer and duce in deduce, produce, induce in English, etc. There are also bound stem s in Latin, like agr- of ager- field" and col- of coiere- f o cultivate" in agricola farmer" (Hockett 1958:241).

2) The attachment of affixes may be PRODUCTIVE. However, productivity is a relative notion. The affixation of inflectional morphemes is known to be most productive, though some dervational morphemes are also highly productive, like -ness in English can be suffixed to happy, giving happiness, and a great many other English adjectives, but cf. *formerness, *ajarness,

*conjugabieness, etc. (Pollard, p.c.). Like English -th in fifth, the ordinal morpheme di- in Chinese can be productively prefixed to any cardinal number to form a corrresponding ordinal, e.g., di-yi

[one] first" or di-wu [five] fifth". Similarly, the suffix -xing ‘-ity’ can be attached to most of the polysyllabic action verbs (pohuai-xing [destroy-ity] ‘destructiveness’). But the suffixation of English

-th in length, width, warmth, strength, etc. is not very productive (or only relatively productive), since *highth, *coolth, *weakth etc. do not occur. Although the attachment of lao- in Chinese is not productive, it is much more productive than the attachment of English afc)-, which occurs perhaps only in acknowledge. Because of this, one should be reluctant to recognize a(c)- as a prefix, since no generalization can be drawn concerning its productivity.

What makes productivity as a characteristic of affixes unreliable is the productivity of syntactic concatenations, especially those affix-like Ws under the labels of clitics (the contracted auxiliary verbs in English), particles (Chinese sentential particles ne, ma, etc.), function words (de) and phrasal affixes (the possessive -’s in English). The labels assigned to these Ws may vary 147 according to different theories, but each of the attachments in syntax is known to be fully productive.

3) The MEANING of affixes tends to be empty, abstract and/or generalized. Thus -ity in activity in English is empty or abstract in meaning, arguably serving only to mark a derived noun; and the meanings of in- in incorrect and dis- in dislike are less so, but still not as concrete as the meanings of the stem content words walk and come. The sam e is true of those affixes in

Chinese, e.g., the emptiness in lao- and abstractness in -xing. Moreover, the meaning of the latter has been generalized from Temper* to "nature*.

It should be noted however that the meanings of most of the affix-like Ws (Chinese de and

English possessive - s) are also abstract.

4) Affixes tend to perform certain GRAMMATICAL FUNCTIONS. Thus the inflectional morpheme -s in English defines the syntactic agreement relationship between the subject and the verb it attaches to in terms of third person, singular and present tense. Some derivational affixes may change the part of speech of the stems they attach to, e.g., nation which is a noun vs. nation-al which is an adjective. Likewise, in Chinese, lao- and -xing are indicative of noun formatives. The suffix -xing can presumably change the part of speech of the stem, e.g., dong- yao [move-wave] TIuctuate" is a verb, but dong-yao-xing "instability* is a noun.

Like the other characteristics mentioned, the exhibition of grammatical functions is not unique to affixes, as languages have function words such as de and prepositions (cong from ).

5) Affixes tend to ATTACH TO FREE FORMS. For instance, since mit in permit is bound, some linguists treat permit as monomorphemic, i.e., per- is not an affix at all for them. The

"modern* affixes in Chinese described by Chao (1968:225-8) tend to attach to a free form, as the attachment of -xing "-ity".

However, it is known that an affix (-tou [noun formative]) may attach to a bound stem

(-mu- [wood]) to form another stem (mu-tou 'wood'); also see -ist and -a/ suffxed to a bound form dent- in dentist and dental. And needless to say, those affix-like Ws attach to free forms in syntax. 148

As stated in Dai (1990a:29), the above criteria only outline tendencies an affix may show.

No single criteria could be the determining factor, and admittedly some may -specific.

Roughly speaking, 1) would differentiate word affixes from those affix-like Ws which are free in being able to pass the expansion test (2.3.7), and 2)-4) would separate affixes from bound stems and free stems. The latter groups are unproductive in attachment, lack generalized or abstract meanings, and exhibit no grammatical functions. Tendency 5) cross-cuts both stems and affix-like

Ws and should be used with caution as a rule of thumb.

Thus it seems that, at best, one can say that if a morpheme is recognized as an affix, it is likely to exhibit most or all of the above described behaviors. According to this multi­ dimensional consideration of affixation, Dai (1990a:29) analyzes-x/ inxue-x/ [learn-practice] ‘learn’ which is traditionally considered to be a bound stem or root as a suffix. This is because its attachment is relatively productive (fu-xi [repetition-practice] "review* jian-xi [see-practice]

"practice", yan-xi [perform-practice] "rehearse" lian-xi [exercise-practice] "exercise", bu-xi [make- up-practice] "make up"...). The meaning of -xi is generalized to "practice". Its grammatical function is to change the part of speech of its host. Before the suffixation, the stem is exclusively verbal; with the suffixation, the result can be a nominal as well. Finally, -xi is usually attached to a free form.

The methodology here is important in examining the word structure of Chinese, and my main goal here is to reanalyze more such bound stems as affixes. Below I present som e of them.

(25) -yu beforehand, in advance, pre-’: yu-bao [pre-report] ‘forecast’, yu-ce [pre-measure]

‘calculate’, yu-chanqi [pre-birth-time] ‘expected date of childbirth’, yu-ding [pre-book]

‘subscribe’, yu-ding [pre determine] ‘predetermine’, yu-duan [pre-judge] ’prejudge’, yu-

fang [pre-guard-against] ‘prevent’, yu-gao [pre-tell] ‘announce’, yu-gou [pre-purchase]

‘purchase in advance’, yu-ji [pre-account] ‘estimate’, yu-jian [pre-see] ‘foresee’, yu-qi [pre­

period] ‘anticipate’, yu-re [pre-heat] ‘preheat’, yu-sai [pre-contest] ’eliminary contest’, yu- 149

Shi [pre-indicate] ‘betoken’, yu-suan [pre-account] ‘budget’, yu-xian [pre-beforehand] ‘in

advance’, yu-xiang [pre-think] ’anticipate’, yu-yan [pre-perform] ‘preview (of a

performance’, yu- [pre-appointment] ‘make an appointment‘, yu-zhan [pre-exhibition]

‘preview (of an exhibition)‘, yu-zhi [pre-make] ‘prefabricate’, yu-bei [beforehand-prepare]

‘prepare’, yu-gan [pre-feeling] ‘premonition’, yu-liao [pre-expect] ‘predict’, yu-mou [pre­

plan] ‘premeditate’, yu-xi [pre-view] ‘preview’, yu-yan [pre-word] ‘prophesy’, yu-zhao [pre­

sign] ‘omen’, etc.

The morpheme yu- is analyzed as a prefix, since it can be shown to be bound. The prefixation

process is relatively productive, as (25) does not exhaust the list. It meaning is generalized to

"pre-". The grammatical function of yu- lies in its ability to change the part of speech of the free word to which it attaches. Most of the monosyllabic stems in (25) can only be verbs; by

prefixation, they can function as nouns as well. Finally, yu- is usually prefixed to a free word.

Only in the last six is the host a bound stem. Following Dai (1990a:32), -yu can be said to have

a dual status. In the last six cases, it is a bound stem; for the rest, a prefix. The same can apply

to the two suffixal items below, except that the suffixed words are nouns only.

(26) -qi ‘implement, utensil, machine’: qi-qi [paint-thing] ’lacquenware’, ci-qi [china-untesil]

‘porcelain’, yu-qi [jade-article] ‘jade article’, shi-qi [stone-inmplement] ‘stone implement’,

tie-qi [iron-implement] ‘iron implement‘, tong-qi [bronze implement] ‘bronze implement’,

jin-qi [gold-implement] ’gold implement’, ying-qi [silver-implement] ‘silver implement’, mu-

qi [wood-implement] ’wood implement’, ji-qi [machine-implement] ‘machine’, dian-qi

[eletricity-device] ‘electric appliance’, jisuan-qi [calculate-machine] ‘calculator’, liang-qi

[measuring-device] ‘measure instrument‘, jishi-qi [timing-machine] ‘timer’, jingbao-qi

[alarm-machine] ‘siren’, feixing-qi [fly-machine] ’aircraft’, shengzhi-qi [generative-organ]

‘genital’, etc. 150

(27) -ti ‘body’: sheng-ti [body-body] ‘human body’, ren-ti [human-body] ‘human body’, wu-ti

[thing-body] ‘substance’, qi-ti [gas-state] ‘gas’, gu-ti [solid-state] ‘solid’, ye ti [liquid-state]

‘liquid’, wen-ti [article-style] ‘literary style’, zi-ti [character-style] ‘style of written characters’,

zheng-ti [politics-body] ‘system of government’, guo-ti [nation-body] state system’, youji-ti

[organic-body] ‘organism’, bingyuan-ti [pathogeny-body] ‘pathogen’, kang-ti [resist-body]

‘antibody’, chuan-ti [boat-body] ‘hull’, etc.

These three forms, yu-, qi-, and ti-, are sufficient to illustrate how some stem-like morphemes can be analyzed as affixes in Chinese. Therefore, based on the five cross-linguistic behavioral characteristics of affixes, the number of affixes in Chinese is increased. The expansion of the family of affixes in the language, together with the development of the notion of bound stem, allows for a recognition of a much wider range of affixes and bound stems and their broader roles in the grammar of Chinese, but not necessarily for the argument that Chinese is quantitatively rich in morphology, though it does show that Chinese is not as poor in morphology as traditionally assumed.

4.3. Summary

This chapter has studied derivational morphology in Chinese. The rules creating the

delimitative verbs, the vividative adjectives and the distributive measure call for morphological

operations of reduplication. With a recasting of the syntax-morphology co-determination of

wordhood or compounding, the phrase-word analysis of the delimitative is discarded, since the

second reduplicate is not syntactically separable, unlike a cognate object or a measure phrase

in apparently the same syntactic position. I maintain that reguair reduplications operate only at

the stem level. If the whole stem is made of XY, it must be reduplicated as XYXY or XXYY but not

*XXY or *XYY. This bears on whether a given expression should be analyzed as a phrase or a

compound. The various types of reduplication are of lexical function. Each relates a new word 151 with its corresponding source word; and no syntactic rule requires the presence of the reduplicated form.

In reexamining affixation operations, a methodology is adopted which resorts to five universal tendencies exhibited by affixes: boundedness, productivity, generalized meaning, grammatical function and attachment to a free form. Although a net parochial result is the enlargement of the affix inventory of Chinese, it is hoped that these tendencies could serve as a starting point toward a universal notion of affix in human languages.

Ail affixation and reduplication as morphological operations investigated thus far have no access to syntax and are thus derivational. It should be noted that individual operations can be invoked by an inflectional rule to realize syntactic features as well. I now turn to this topic and present inflectional morphology of Chinese. CHAPTER V

INFLECTIONAL MORPHOLOGY: CONCEPTS

5.0. Introduction

It might seem that inflectional morphology is a topic of limited interest to linguistics, since so many languages are cited as having little or no inflectional morphology. Chinese is generally assum ed to be among those languages, and there are no serious analyses of Chinese inflectional morphology in the literature. Thus Chinese has not only made no contribution to the understanding of inflectional morphology in linguistic theory, but has also been the source of constant confusion. For instance, Chinese is assum ed to have no verbal inflection at all (e.g.,

Hockett 1958:210, Li & Thompson 1981:134, Zwicky 1987a:642, Sun & Cole 1991:44).

Nevertheless, lip service has been paid to the claim that Chinese is in the process of gaining

inflections (Bauer 1988:84), and it is taken for granted that the plural marker -men (Anderson

1985:174, Packard 1990:35) and some verbal aspectual markers like the durative -zhe and

perfective -le (Dai 1990d:333) are inflectional morphemes. The dominant view seem s nevertheless

to be that Chinese has no interesting inflectional morphology to study.

It is the goal of this chapter to refute this predominant view. After justifying a theoretical

division between derivation and inflection (5.1), I will formally demonstrate that the measurative

TA (5.2), the resultative DE (5.3), the aspect markers -zhe, -le and -guo (5.4), and the deservative -

tou (5.5) are inflectional morphemes, undeniably relevant to the syntax of the language. Empirical,

methodological and theoretical consequences of this analysis will be discussed in the next

chapter.

152 153

5.1. The Division between Derivational Morphology and Inflectional Morphology

The traditional definition of inflectional morphology derives from Bloomfield (1933:208).

According to him, inflectional languages merge semantically distinct features in either a single bound form or in closely united bound forms. For example, the suffix -o: in a Latin form amo: ‘I love’ indicates first person, singular and present tense. A paradigmatic set consists of different inflected forms of the same word. By contrast, derivational morphology relates paradigms of different words {ibid:223, Hockett 1958:210). It is perhaps by this definition that Chinese may have come to be classifed as having little inflectional morphology.

Despite the traditional sharp distinction between inflection and derivation and a reasonable consensus as to how to classify particular items, there is no accepted defintion of inflection. For instance, Matthews (1972) and Anderson (1982:585) note the inadequacy of

defining inflection in terms of putative characteristics such as productivity, inability to change parts

of speech, marginal position, or some inventory of affected features (person, number, case,

gender, tense).

As research advances, inflectional morphology has come to be viewed as the part of

morphology which is relevant to syntax (Anderson 1982:587). Following the lead of Anderson

(1982, 1988:168), Scalise (1988:561-81) and Zwicky (1987a:654, 1989a:145), I adopt the

fundamental theoretical assumption that derivational morphology and inflectional morphology

constitute separate (sub)components of grammar, even though the apparent distinction between

the two in formal operations is minimal. Both deal with word structures, and operations such as

affixation, vowel or consonant mutation, reduplication, metathesis, etc. are available to both. The

distinction, however, can only be motivated on theory-internal grounds: inflectional morphology

is of syntactic function, consisting of exactly those aspects of word structure that are determined

by or accessible to syntactic rules, such as stipulating the past participle form [PSP] -en of the

English word work as work-ed in has worked. 154

Derivational morphology, on the other hand, is of lexical function, deriving a new word, such as careful-fy from the source word careful. Specifically, the derivational -ty cannot be considered to be syntactically mandated. The contrast in He drives carefullyl*careful is due not to the syntactic requirement of the presence/absence of Wy, but to the requirement that the postverbal constituent in this sentence be an adverbial rather than an adjectival expression.

Whether or not that adverbial is suffixed with -ly is not the point at issue, as shown in He drives fast/with care. By contrast, the syntax of the perfective construction requires not only a VP as the complement but also the presence of the inflection -en on the head of the VP, as in (He) has worked (for two hours)-, its absence (*has work) or replacement by any other suffix (*has working) yields ungrammaticality (or a different construction).

The current notion of inflection is consistent with the corresponding traditional notion, for the behavioral characteristics distinguishing the two grammatical subcomponents observed in traditional grammar (e.g., Bloomfield 1933:222-4) fall out naturally.

First, inflectional morphemes do not change parts of speech. Thus the present participle

[PRP] -ing and third person singular present tense [PRS SG 3] -s do not change the part of speech of work, in He is working now and He works everyday. But derivational morphemes are not similarly restricted, e.g., nation is a noun, but nation-ai, in which -al is derivational, is an adjective.

Second, inflectional morphemes apparently indicate the syntactic relationship between the host and another word in a phrase or a sentence. Derivational morphemes are irrelevant in this regard. Thus -ing reflects the relation between work and is in the examples above, as opposed to *He has working now.

Third, inflectional morphemes are morphologically general, or productive. Thus almost every English verb has -en and -ing forms, whereas *country-al does not exist, despite the occurrence of nation-al in derivations. 155

Fourth, inflectional morphemes tend to occur at the margins of words, outside of derivational morphemes, hence the relative positions of inflectional plural form [PL] -s and derivational -ment in govern-ment-s vs. *govern-s-ment.

The above characteristics are by no m eans exhaustive, but all follow naturally from the current notion of inflection. Since inflectional morphology is syntactically relevant, or manipulated by syntactic rules, an inflectional morpheme should not change the part of speech, which would yield an entirely different construction. Since syntactic requirements involve more than inflection on a construction, an inflectional morpheme may apparently mark thé syntactic relationship between its host and some other constituent in the construction, most notably in terms of morphosyntactic government or agreement. Inflectional processes are productive simply because the syntactic rules they have access to are structurally general. An inflection often, if not always, closes a word, because of the fact that lexical insertion into a given syntactic structure depends on a full range of lexical properties of an item: thus derivational operations must all be completed before the lexical insertion. It is only after lexical insertion that inflectional rules apply, determining morphosyntactic representations (Anderson 1988:180).

I now discuss three verbal aspect markers in Chinese that have been much studied in the literature, to see why they are widely considered inflectional morphemes and what kind of further evidence is needed. They are the perfective -/e, the durative -zhe and the experiential -guo (e.g.,

Chao 1968:219-56, Li & Thompson 1981:184-237), as in (1).’

(1)a. Wo chi le fan.

I eat perf. rice

I had a meal.’

\ l a ) and (1b) may be regarded as "incomplete" sentences. For instance, Wo chi-zhe fan kan dianshi ‘While eating a meal, I watch TV’ would be considered "complete" in contrast to (1b), as the former provides some context. However, I assume that (1 a) and (1 b) (and similar examples in this study) are grammatical because they are fully acceptable when contexts are provided by discourse. 156

b. Wo chi-zhe fan.

I eat dur. rice

‘I am having a meal.’

c. Wo chi-guo fan.

I eat exp. rice

‘I once had a meal.'

Linguists take these aspectual suffixes as inflections for several reasons. First of all, aspect markers in most European languages are inflectional morphemes, as English -en and and

-ing suffixes marking the perfective and progressive in (2).

(2)a. They have worked for three hours.

b. They are working in the office.

But identification of inflection cannot be m ade solely on the basis of the grammatical categories

involved, since a grammatical category which is inflectional in one language may be derivational

in another. The nominal categories of diminutive and augmentative, for example, are fully

inflectional in Pula, as shown by their participation in syntactic agreement, but diminutives -chen

or -lein in German and -y or -let in English are derivational (Anderson 1985:162). The plural form

-s in English is inflectional, but its counterpart -men in Chinese will later be shown to be

derivational.

More significantly, the aspect markers exhibit most of the behavioral characteristics of

inflectional morphemes. In particular, they never change the parts of speech of the words to

which they attach, as the resultant words are still verbs in (1). In addition, aspect markers can

form a verbal paradigm (Tai, p.c.) analogous to those in most European languages, a fact related 157 to the notion of productivity. Like a typical inflectional morpheme, the attachment of the aspect markers is fully productive. Virtually all verbs can be so suffixed, except for some semantically and pragmatically incompatible ones, such as certain combinations with the durative.^ Finally, aspect markers occur in word-final position, closing a verb. Indeed, using these markers to identify verb boundaries seem s a reliable method (Dai 1990d:333-4).^

However, none of the above characteristics can determine the inflectional status of the aspect markers. First, some derivational morphemes do not change the parts of speech of their hosts, e.g., the English prefixes: in-correct, dis-iike, re-view, etc. Moreover, ailowing morphological productivity as a test for inflection stems from theory-internal assum ptions, i.e., that productivity is relevant only when the syntactic rules calling for inflections exist and are highly productive. And productivity alone is not sufficient to ensure inflectional status, as exceptional examples are found.

For instance, the English derivational classes of -ing verbal nominalization for gerunds {find-ing, think-ing, etc.) and -iy adverbial formation from adjectives {carefui-iy, happi-ly, etc.) seem completely productive (with a few lexical exceptions), but the suffixation of inflectional classes of

plural comparative -er and supeifative -est are not universally productive {*informations,

*usefui-er, *useful-est). The former is productive only for count nouns, and the latter seems to

be partially conditioned by the number of the syllables of the stem. Therefore, productivity may

not directly indicate inflection. Finally, using the edge position to identify inflections turns out to

be unreliable in some cases. For instance, the English derivational -ly can also be said to close

the resulting adverb in that no further suffixation is possible, as in careful-ly. Worse yet, words

do not have to be inflected, e.g., not all English plural nouns take -s, and the verbs may be in an

uninflected infinitive form. In such a case, it is absurd to claim that the word-final morpheme must

^According to U & Thompson (1981:218-24), nonactivity verbs {*zhidao-zhe know') and verbs that describe instantaneuos activities {*si-zhe ‘in the process of dying’, *chuxian-zhe ‘in the process of emerging ) do not take -zhe.

^here are complications. Infixed aspectual markers are found in structures like kan-le-kan ‘had a look' (cf. 4.1.1, to discussed in 6.0). Also the perfective -le is distinguished from the sentence-final particle le. The latter cannot be used to locate verb boundaries. 158 be inflectional. Moreover, derivational rules have the option of being built on a stipulated inflectional form of a lexeme rather than on a stem (Zwicky 1989a:144), an apparent violation of the orthodox ordering of derivation and inflection. In fact, these apparent inflections no longer function as inflections in that they are not required by the syntax. For example, the English derivational -ness occurs outside of the "inflectional" -ed in relat-ed-ness. In Spanish, ADVs suffixed with -mente are derived from the "inflectional" feminine singular form [FEM, SING] of an

ADJ lexeme (perfectamente ‘perfectly’), rather than from the stem of the lexeme. Similarly, in

French, the FEM form of an ADJ is used in the derivation of an ADV {fausse-ment falsely’ from fausse, the FEM form of faux ‘false’). In Modern Mongolian, the derivational suffix -xi forms ADJ from ADV, and can be added to the dative/locative form of N, e.g., gerth-xi ‘domestical’ from the locative gertfi of ger ‘house’ (Anderson 1988). In addition, it is conceivable for an inflectional infix to occur inside the derivational affixes of a word. For example, Nevis & Joseph (1992) observe that the Lithuanian inflectional -si (reflexive) can occur in a position inside of verbal derivational prefixes if the host has one or more prefixes {be-si-kalbant ‘while conversing’), otherwise, it is placed at the end of the word {kalbuo-si ‘I converse’).

Thus a null hypothesis based on the syntactic markedness of inflections would specify the three aspect markers as derivational, or at least non-inflectional, if there were no evidence from the syntax. However, while I emphasize that the formal requirement of an inflection is its syntactic relevance, further qualification is in order. In the current theory, syntactic relevance is not necessarily limited to overt morphosyntactic markings via government and agreement. In the traditional sense, the [PRP] form -ing in English is inflectional in be going, because the form is said to be "governed" by the verb be in the construction, similar to object NPs in various case forms governed by verbs and prepositions in German. Government is distingushed from agreement. Thus, in French, Italian, German, Russian, Latin and many other languages of the world, the adjective must agree in certain morphoiogical features with the head noun in an NP, and in many languages, the verb must agree with its subject (typically in number, gender or 159 person). It is not surprising that linguists as recent as Jensen (1990) seem to equate syntactic relevance with syntactic linkage only via agreement and goverment (e.g., Hockett 1958:ch.25), since the two are clear cases of apparent syntactic relationships between the host word with some other word in the same local domain.

However, there is no a priori reason to believe that agreement and government are the only phenomena of syntactic relevance for inflectional morphology. A case in point is the English

[PAST] tense morpheme -ed, which can be said to indicate no syntactic relationship between the host verb and some other sister constituent. Without the presence of a past temporal adverbial,

(3a) is still grammatical, with an interpretation of past action, in constrast with -ing, which depends on the presence of the lexeme be in that construction, as in (3b).

(3)a. He work-ed very hard (yesterday/two days ago/in his college days),

b. He is working/*He working/*He has working.

Nevertheless, [PAST] is syntactically required in some subjunctive constructions, though the form does not express semantic past, e.g.. It is high time that we started or If John worked here, he would tell a lot of jokes. The main reason that [PAST] is inflectional lies in its interactions with other tense and aspect forms which are demonstrably inflectional in English. Morphologically, the relevant morpheme slot for a verb seems to be reserved for only one inflectional morpheme, and

[PAST] occupies the same morpheme slot as other inflections ([PRES SG 3] for third person singular present tense), [PAS] for the passive, or [PSP] for the perfective, or -ing [PRP] for the progressive). These morphemes all compete for the same slot. Only by comparing and contrasting the form [PAST] -edwith other inflections [PSP] in (4a), [PAS] in (4b), or [PRP] in (4c) can one see the status of the [PAST] form, as those in the braces. Because of frequent homophony inform between [PAST] and [PSP/PAS], the grammaticality contrast would be more obvious for those verbs with irregular forms below. 160

(4)a. He had {worked,driven,gone}/*{worked,drove,went} for two hours.

b. The plane was {washed,driven,flown}/*'{washed,drove,flew} by him.

c. The boss was {working,waiking,smiling}/*{worked,walked,smiled}.

The slot competition gives rise to the syntactic relevance of [PAST}, i.e., syntactically, it cannot occur in the three marked constructions in (4). Thus the notion of syntactic relevance of a morphological form is taken in this thesis as its presence ([PRP] -ing) and/or absence ([PAST] - sd) as required by individual syntactic constructions. The domain of inflection certainly includes, but is not restricted to, government and agreement, and this will bear on the inflectional status of the aspect markers in Chinese via their interactions with two morphemes in the language never before considered inflectional, the measurative TA and the resultative DE.

5.2. The Measurative TA as an Inflectional Morpheme

TA is segmentally homophonous with the personal pronoun ta ‘he, she, it’, except it is always unstressed, toneless and phonologically attached to the preceding verb, as in (5).

Example (6) shows that the presence of TA seems optional.

(5) Qie-TAsan jin rou.**

cut TA three pound meat

Cut three pounds of meat.’

(6) Qie san jin rou.

Used non-pronominally and non-referentially, TA in (5) may not be understood as the indirect

'’Many speakers allow -ge to replace TA, or accept the TA-ge sequence. Presumably, -ge is historically derived from the homophonous measure word. 161 object ‘him’ of qie, since an independent indirect object can be added, as in (7). Moreover, qie cannot be immediately followed by an indirect object without gei, as in (8a), which would otherwise force a referential and possessive reading in (8b).

(7) Qie-TA san jin rou gei ta.

cut TA three pound meat give he

‘Cut three pounds of meat for him.’

(8)a. *Qie wo/ni san jin rou.

b. Qie wo/ni san jin rou.

cut I/you three pound meat

‘Cut three pounds of my/his meat.’

No serious investigation of TA has been carried out except for Chen (1987:86). Calling TA a non-

referentiai pronoun used in some idioms limited to certain patterns, Chen gives the following

examples without further discussion.

(9) Jinnian xian zhong-TA jimu shiyan tian, qude jingyan hou zai da mianji tuiguang.

this-year first plant TA a few acre experiment field get experience after then big area oromotion

‘(We) are first going to try a few acres for experiment this year, then promote the results after

getting the experience.’

(10) Wo buguan-TA san-qi-ershiyi, qu-le zai shuo.

I ignore TA three-by-seven-(is)-twenty-one go perf. then say

I will go and see anyway in spite of what will happen.’ 162

I will use upper case letters to represent measurative TA, in contrast with the pronoun ta. To be sure, some of the examples would be ambiguous, namely, TA in a sentence could also be understood as the referential and pronominal fa, because of their phonological similarity, except for occasional differences in stress and tone. But for some cases to be acceptable at all, TA must replace fa, as below.

(11 )a. Wo song ?TA/*ta san jin rou gei ni.

I send TA/he three pound meat to him

b. Ni song ?TA/*ta san jin rou gei wo.

you send TA/he three pound meat to me

The ungrammaticality here is caused by a conflict of interpretation between a second person and a third person recipient or indirect object. Substituting TA for fa brings back grammaticality, but because of the marked pragmatic value associated with the construction (to be discussed), speakers rarely use it.

There is hardly any synchronic connection between TA and fa. Two factors may give rise to the apparent pronominal status of TA. It is unquestionable that TA has its historical origin in the pronoun fa. Their orthographic forms are identical. However, the written form, being linguistically secondary, does not constitute legitimate evidence for its pronoun-hood. Moreover,

TA is segmentally the same as fa. But as pointed out before, the two are prosodically distinct.

TA is always weak, i.e., unstressed and toneless. For instance, if TA is stressed or tonic as fa in

(5), then one acceptable reading might contain a possesive pronominal, ‘Cut three pounds of his meat',® in contrast with (12) where fa, subject or object, can be weak or strong (stressed and

®One may suggest that (5) could have a partitive reading Cut three pounds of meat from it’ and thus the partitive, now referential, could be characteristic of TA. However, for a partitive reading, "TA" must be stressed, indicating pronominal fa. Moreover, the partitive cannot be part 163 tonic) with the pronominal reading.

(12) Ni gaosu-le ta, ta you gaosu-le wo.

you tell perf. he he then tell perf. I

‘You told him, and he in turn told me.’

I will now argue that TA should be analyzed as an inflectional morpheme. The distinction

between the non-referential, non-pronominai TA and referential, pronominal ia should always be

preserved. Any confusion of one with the other would possibly give rise to a very different

analysis, or even lead to a different grammaticality judgement.

5.2.1. TA as an Affix

Like the aspect markers, TA is a bound morpheme, subject to left-expansion, as in (13),

although not to right-expansion.

(13) a. Song-TA san jin rou gei zhe-xie ren.

give TA three pound meat give these people

‘Give three pounds of meat to these people.’

b. Song-TA yi-kuai san jin de rou gei zhe-xie ren. (right expansion)

give TA one-measure three pound de meat give these people

‘Give a three-pound piece of meat to these people.’

of a general pattern. It cannot be substituted for TA below and some other examples in this paper.

i. Wo yao shui-TA santian sanyie. I want sleep TA three-day three-night ‘I want to sleep three days and nights.’ 164

c. *Song zhe-xie ren-TA san jin rou. (left expansion) but cf.

d. Song zhe-xie ren san jin rou.

give these people three pound meat

‘Give these people three pounds of meat.’

It should be emphasized that TA in (13a) cannot be interpreted as a pronominal indirect object, since the indirect object zhe-xie ren ‘these people" is given. TA cannot be considered the

'copy' of zhe-xie ran, either, for the latter is semantically plural, whereas TA is presumably singular. The notion of "copying" is methodologically and theoretically suspect, even though number agreement is satisfied by replacing zhe-xie ren by the singular proper noun Zhangsan in

(13a). This is because one could not explain, except by stipulation, why only the third person singular can be so "copied".

(14)a Song (*ni) san jin rou gei ni zhe-ge dang laoshi de.

give you three pound meat give you this be teacher de

‘Give three pounds of meat to you, as a teacher."

b. Song (*ta-men) san jin rou gei zhe-xie ren.

give he-PL three pound meat give these people

‘Give three pounds of meat to these people."

Besides, in (15a), either the pronominal or non-pronominal reading is possible given an appropriate context. However, switching the positions of yixia and fa/-TA causes an obligatory loss of the non-pronominal reading. This clearly shows that a postverbal ta is a free word, but

TA must be bound to the verb. 165

(15)a. Ni jiu gei wo song ta/-TA yixia ba.

you just for I see-off he/TA a-little prt.

‘You just see fiim off a little for me/give a little send-off for me.’

b. Ni jiu gei wo song yixia ta b a

c. *Ni jiu gei wo song yixia-TA ba.

Given that TA is a iert-bound morpheme, is it possibly an enciiticized word like the modifier marker de or sentential mood particles ma, ne, etc.? While no evidence can be found in favor of its clitic-hood, there is at least one piece of evidence for its affix-hood. TA attaches only to transitive verbs, not other categories like nouns (cf. *Fangzi-TA san jian ‘house TA three rooms’), thus showing a high degree of selection for its host, as do other affixes. By contrast, the host for a clitic can be any class of word. The differences between affixes and clitics will be considered in detail in 5.3.1 and 5.5.1. Since there is no obvious reason to assign marked clitic status (Zwicky 1985a:289) to TA, 1 assum e TA to be a verbal suffix.

5.2.2. TA as an Inflectional Morpheme

The crucial evidence for TA as an inflection lies in its syntactic relevance. When TA is attached to a verb, the verb requires an obligatory quantified NP complement, as in (16a). This is achieved by taking a quantified modifier as the head N in the NP, such as san ‘three’ in (16b).®

®Apparent exceptions to this stipulation include the following examples. i. Da-TA leng bu fang, hit TA cold not prevent ‘launch a surprise attack’ ii. Jiao-TA tian hun di han. stir TA sky faint ground dark ‘make situation unclear and complicated’

However, there are reasons to believe that these may involve a different construction. If a 166

For convenience, I mark the NP [+M(easure)]7 If the relevant NP is marked [-M], i.e., lacking a quantified modifier, then the result is ungrammatical, a s in (16c), in contrast to (16d), which is fine without the presence of TA.

(16)a. Wo qu qiel/*qie-TAI

I go cut cut TA

Til go to cut (it).’

b. Wo qu qie-TA san jin rou.

I go cut TA three pound meat

Til go to cut three pounds of meat.’

c. *Wo qu qie-TA rou.

d. Wo qu qie rou.

I go cut meat

‘I go to cut meat.’

I call the structure in (16b) the M-construction. The syntactic relevance of TA lies in the fact that its presence is licensed by the syntax of the M-construction. The M-construction stipulates three conditions. Its first immediate constituent must be a verb; the second constituent must be an NP

quantifier plus a measure NP is added after TA, these sentences are still good, and seem more natural (cf. the example in the note 8). However, only yi ‘one’, but not other numerals, can be added in. Furthermore, the constituent follov/ing TA must be an idiomatic expression, as in (i) and (ii). Finally, there are no corresponding grammatical sentences without TA, i.e., TA is obligatory as opposed to being apparently optional in other TA cases.

^[-f Q(uantifier)] would be more appropriate than [-fM], because quantifiers hen duo ‘very many’, hen shao ‘very few', etc. can also mark the NP[4-M], and [-FM] is a subset of [4-0]. 167 containing a quantified expression;® and the verb must be suffixed by TA. Thus TA is an inflection in that it apparently marks the syntactic relationship between the host verb and its NP complement.

The implicit claim here is that the presence of TA in (5) and its absence in (6) characterize two distinct syntactic constructions and that TA in the M-construction is mandatory. The formal difference between the M-construction and the corresponding [-M] construction is the presence, due to TA, of the quantified NP, as in (16b-d). Without TA suffixation, the object NP can be either missing or does not have to be quantified. This generalization is across-the-board. A monomorphemic verb may take a quantified NP as V-TA does, as in (17a), but the resulting construction is different from the one whose head verb is TA-suffixed, a s in (17b). Both sentences have an indirect object introduced by gei. However, the "dative-movemenf of Zhangsan and the topicalization of san jin rou are not allowed in the M-construction (17b) as they are in the corresponding [-M] construction (17a), as illustrated in (17c-e) and (17f-g) respectively.

(17)a. Song san jin rou gei Zhangsan.

b. Song-TA san jin rou gei Zhangsan.

give TA three pound meat give Zhangsan

‘Give three pounds of meat to Zhangsan.’

c. Song (gei) Zhangsan san jin rou.

®The unstressed numeral yi ‘one’ may be missing. Although this is a general feature of the language, in the M-construction, the classifier or measure word must be ge. i. Da-TA (yi) ge leng bu fang. hit TA (one) Measure cold not prevent ‘launch a surprise attack’ ii. Zhua-TA yi-tiao/*tiao yu lai. capture TA a Measure/Measure fish come ‘go capture a fish.’ 168

d. ?Song-TA (*gei) Zhangsan san jin rou.

e. ?Song gei-TA Zhangsan san jin rou.

f. San jin rou song gei Zhangsan.

g. *San jin rou song-TA gei Zhangsan.

This structural distinction is further supported by distinct pragmatic values associated with the M- and non M-constructions. it is true ti iat TA has no semantic analogue since it makes no contribution to the compositional semantics of the M-construction, i.e., (5) and (6) have the same literal meaning. However, as noted earlier, the first or second person as an indirect object is usually not allowed. An anomaly would arise if zhe-xie ren these people’ were replaced by wo

'!' or ni ‘you’ in (13a); also see (11). This apparent syntactic restriction is closely related to the marked pragmatics of the M-construction.

The M-construction in (5) tends to be reserved for informal conversational discourse, while the corresponding [-M] construction in (6) is stylistically neutral. Another pragmatic feature of the

M-construction is that such expressions are not meant to be polite. The speaker using it conveys a condescending kind of attitude to the theme and recipient. Consequently, the measure NP following TA often, if not always, implies a small amount or insignificance of the object theme.

Thus an NP denoting large quantities would often lead to unacceptability, e.g., *Qie-TA hao ji jin

rou (where hao ji means "quite a few", cf. (5)), ‘Cut quite a few pounds of meat’. Since the

condescending attitude would insult the second person indirect object (i.e., the hearer as the

recipient), it is generally disallowed in the construction. Similarly, the first person is rarely used

as indirect object, since otherwise the self-respect of the speaker (as the recipient) would be

damaged. Because of the marked pragmatic value, it is hard to imagine that the M-construction

could ever be used in formal speech, as in a lecture, a command, or a serious suggestion. 169

To conclude this section, it should t>e pointed out that the M-construction cannot be regarded as an idiom chunk, simply because all of the morpho-syntactic constituents involved, except for the suffix TA, are fully catégoriel, rather than item-specific. Put another way, what is invoived here is not an isolated chunk listed in the lexicon, but a fully productive syntactic pattern.

Its first constituent can be virtually any of the transitive verbs denoting activity or motion, and its second constituent can be any quantified NP in the language. If the M-construction were an idiom chunk, then theba- and be/-constructions in Chinese, and the infinitive construction marked by io or the perfective construction marked by have in Engiish, wouid also have to be idiom chunks. More examples of the M-construction are provided below.

(18)a. Zou-TA wu Ii lu.

walk TA five mile road

‘Walk five miles.’

b. Chi-TA liang wan fan.

eat TA two bowl rice

‘Eat two bowls of rice.’

c. Fu-xi-TA yi tian yi ye.

review TA one day one night

‘Review one day plus one night."

d. Piaoliang-TA yizhen.

pretty TA a-while

‘look pretty a while’ 170

5.3. The Resultative DE as an inflectional Morpheme

Having justified the inflectional status of TA, I now turn to the much disputed resultative

DE. There are several homophonous de’s in Chinese. C-R. Huang (1989 and references therein) gives a succinct survey of their distributions and typology, summarized (and revised) below (with capital letters for the resultative DE as opposed to other de’s in Chinese). He also assigns to each of them a structural level in terms of affix or clitic.

(19) a. An adverbial affix:

Ta manman-de zou lu.

he slow de walk road

‘He walks slowly.’

b. An adjectival clitic:

zoutian lai de ren

yesterday come de person

‘the person who came yesterday’

c. A verbal clitic:

Ta [yp (v, ku DE](vp 2 hen shangxinj]

he cry DE very hurt-heart

‘He cries sadly." 171

The contention here is that DE In (19c) is in fact an inflectional suffixal morpheme, In contrast to the so-called adverbial de In (19a) and adjectival de In (19b), which will be shown to be words. More examples of DE and de are given below, with the resultative DE In (20) and (21), the adjectival de In (22), and the adverbial de In (23).

(20) Ta [yp tiao-DE][yp 2 kual]].

he jump DE fast

He jumps fast.'

(21) Ta [yp han-DE](g sangzi dou ya-le]].

he shout DE throat all hoarse perf.

‘He shouted so much that his voice was getting hoarse.’

(22) Wo zai kan zuotian Lisi mai de shu.

I at look yesterday LisI sell de book

I am reading the book Li sold yesterday.’

(23) Ta da sheng de chang ge.

he big sound de sing song

‘He sang loudly.’

Before proceeding, several observations are in order. First, (20) is traditionally designated as a

"descriptive complement construction" in that VP2 describes VI, and thus VP2 may turn out to be a prototypical adjective phrase (AP). Second, (19c) and (21) are traditionally referred to as

"resultative complement constructions" because VP2 in (19c) denotes the result of VI (C-T. Huang

1988:274-5). Third, VP2 after DE as in (19c) can be replaced by a clause (S), as in (21). And 172

APs like hen lei Very tired' can also replace VP2/S in (19c) and (21). There seem s to be no problem representing such categorial overlappings here, for [+V] can be used for referring to both adjectivals and verbal categories, and (double bar), or simply VP, for both VP and S, either approach common in linguistic practice. Furthermore, I will use the same DE in (19c)/(21) vs. (20), even though the two are different in subtle (semantic) ways, and call both of them the resultative (henceforth R-) construction, as opposed to de’s for adjectival or adverbial markers in

(19a), (19b), (22) and (23), the non R-constructions. A unified R-construction has some empirical support. Descriptive and resultative complements can be conjoined, as in (24), where the first

VP after DE is semantically descriptive, and the second is resultative. Such coordination possibilities strongly argue against Y-H. Li's (1990:51 ) treatment of the resultative and descriptive sentences as two distinct and unrelated constructions.

(24) Ta pao-DE ji [[^p kuai] you [^p bu lei]],

he run DE both fast and not tired

‘He ran fast but not was not tired.'

The status of DE has been controversial: it is treated as a particle word in Zhu (1961), as a COMP introducing VP2 or S in C-T. Huang (1982) and Ross (1984), as a clitic to the preceding verb in C-R. Huang (1989), or implicitly as a derivational suffix to the verb in C-R. Huang &

Manjione (1985) and C-T. Huang (1988). In order to argue for DE as an inflectional morpheme, it is necessary to show that its properties are unlike those of any of these designations previously put forward, but rather like those of an inflectional affix bearinfg syntactic relevance.

5.3.1. DE as an Affix

DE cannot be a word of category N, V or A. But can it possibly be a grammatical function word like the English prepositions in, over or under, or determiners a or the, or the COMP that. 173 to name just a few? DE does not behave like a syntactic word, because although words can be inserted on its right side, they cannot be inserted on its left side, as (25) and (26), which are expanded from (20) and (21) respectively, show.

(25)a *Ta tiao sheng-DE kuai. (left expansion)

he jump rope [rope-skipping] DE fast

b. Ta tiao-DE hen kuai. (right expansion)

he jump DE very fast

‘He jumps very fast.'

(26)a. *Ta han ren-DE sangzi dou ya-le. (left expansion)

he shout person DE throat all hoarse perf.

b. Ta han-DE ta de sangzi dou ya-le. (right expansion)

he shout DE he de throat all hoarse perf.

‘He shouted so much that his voice was getting hoarse.’

By contrast, syntactic expansion can apply to de in (22) and (23) on both sides, as in (27) and

(28), which demonstrate that the adjectival and adverbial de’s are syntactic (function) words by

my criteria.

(27)a. Wo zai kan zuotian Lisi mai gei meimei de shu. (left expansion)

I at look yesterday Lisi sell to younger-sister de book

‘I am reading the book Lisi sold to my sister yesterday.’ 174

b. Wo zai kan zuotian Lisi mai de neiben shu. (right expansion)

I at look yesterday Lisi sell de that book

'I am reading the book Lisi sold yesterday.’

(28) a. Ta da sheng xiang da lei yiyang de chang ge. (left expansion)

he big sound like thunder like de sing song

'He sang as loud as thunder.’

b. Ta da sheng de zai wu Ii chang ge. (right expansion)

he big sound de at room inside sing song

’He sang in the room loudly.’

A straightforward explanation of these data follows from the claim that the immediately preceding and following constituents of de are both phrasal categories (XP, or bar-two). However, although the constituent immediately following DE is also phrasal, its immediately preceding constituent must be a lexical category (X°, bar-zero, or word), typically a verb. This indicates that DE is a bound morpheme to its left host, which gives rise to a possible analysis of DE as the suffix to a verb, not necessarily entailing that DE cannot be a particle word or a clitic. By assuming word status for DE, the syntax could legitimately stipulate a verb as its immediately preceding constituent and impose a strict intervention constraint between them, accounting as well for the grammaticality contrast in (25) and (26). Thus, the expansion test is consistent with DE being a suffix, but does not force this conclusion.

Before resolving the issue, I would like to point out that DE behaves differently from a

"standard" COMP assumed in C-T. Huang (1982), like that in English. Example (29) would have the structure indicated if DE were a COMP, comparable with the COMP position in English in (30). 175

(29) Ta han [g. [comp DE][g sangzi dou ya-le]].

he shout DE throat all hoarse perf.

‘He shouted so much that his voice was getting hoarse.’

(30) We thought [g. (comp ^hat] [g he was a good student]].

The English COMP that In (30) forms a syntactic constituent with the following S it introduces. This constituency seems to be confirmed by the topicalization of S plus COMP in

(31 a), compared with the ungrammaticality of (31b).®

(31) a. That he w as a good student, we knew,

b. *He was a good student, we knew that.

The structure In (29) for which DE Is In COMP contravenes the claim that DE forms a morphosyntactic constituent with the preceding verb. Although the topicalization In (32b) might sound a bit unnatural. It contrasts sharply with the ungrammatical (32c) where DE is also fronted.

The only alternative Is to stipulate that DE Is phonologically depeendent on the preceding constituent (verb), rather than on the following word, since a parenthetical remark {hmn) or a phonological pause can Intervene only after, but not before, the DE, as In (32d) and (32e).

(32)a. Ta [yp [V han-DE][g sangzi dou ya le]].

b. ?Sangzl dou ya le, ta han-DE.

c. *DE sangzi dou ya le, ta han.

®This constituency explanation appears to be theory-neutral, and methodologically preferable to some popular theory-dependent explanations, namely, either that Is not the head of that + S and therefore cannot license a gap, or It does not assign a semantic theta-role to S and therefore falls to properly govern S or the gap. Also note that (31 b) would be grammatical. If It were considered a left-dlslocation construction. 176

d. Ta han-DE, hmn, sangzi dou ya le.

e. *Ta han, hmn, -DE sangzi dou ya le.

The constituency test seems to argue for the constituency of V-DE in (32a), and against the structure of (29) where DE is in COMP. For additional evidence, see C-R. Huang (1985a: 139).

Besides, the notion of COMP itself is both theory-dependent and language-specific. Not all syntacticians adopt COMP in their theories, and whether or not Chinese has COMP remains an empirical and theoretical question. For all these reasons, I do not consider DE to be COMP.

Provided that DE is bound somehow, there are two possible structures in which DE might lean on its left host: [y[STEM][SUFFIX:DE]] and [yp[V][PARTICLE:DE]]. The former contains one word in which DE is a suffix to a verb. In the latter, DE is a particle word (Zhu 1961), but cliticized to its left word to form a prosodic unit (C-R. Huang 1989:24), one phonological word. The clitic- hood of DE seem s well justified, since DE is always prosodically weak and must therefore phonologically lean on a host.^° Let us define clitics loosely as syntactic words, phonologically weak, and dependent on and bound to an adjacent word to form a prosodic unit (but cf. 7.0), like the contracted AUX -’s in He's going. I shall now evaluate the relative merits of the suffix analysis and particle-clitic analysis of DE.

The clitic analysis suffers first of all on metatheoretical grounds. The implicit claim about the general human ability for language is that clitics are more marked than inflectional affixes

(Zwicky 1985a:289). Cross-linguistically, affixes are statistically more common than clitics. To take one extreme position, given that the structural levels of word (syntactic and phonological) and affix are independently motivated in universal grammar, it is doubtful that grammar should ever need the notion of clitic (cf. ch.7). In Chinese, clear evidence is lacking for a resolution, since suffixes are also prosodically weak, toneless and unstressed in general (Chao 1968:219),

^°This analysis is also assumed by COMP proponents: DE as a COMP forms a syntactic constituent with its following element, but is cliticized to its preceding element, thereby explaining the grammaticality constrast in (32). 177 indicating that prosodic weakness is not a sufficient condition for clitichood. Therefore, in the absence of clear evidence classifying DE as an affix or clitic, one should assume it to be an affix.

However, metacriteria should constitute the last and weakest resort for explanation, and may never convincingly decide a case. Therefore, I turn to some theoretical assumptions and empirical evidence, which will lead to the conclusion that the suffix analysis should be favored over the clitic analysis, for the latter would make wrong predictions.

Among the six lines of cross-linguistic evidence which separate clitics and inflectional morphemes (Zwicky & Puiium 1983:502), three favor DE as an affix rather than a clitic - specifically, the degree of selection between the dependent morpheme and the word to which it attaches, the existence of arbitrary lexical gaps, and the possibility of syntactic operations affecting the combinations. These three tendencies follow from the assumption that word-clitic combinability is largely governed by syntactic considerations. The conditions governing the combinability of stems and affixes are, however, morphological and/or lexical in character

(ibid:503)V If one takes some random affix, it probably exhibits a high degree of selection for its host. For instance, -ing only attaches to a verb in English, as opposed to the clitic AUX -'s, whose host can be any category {The catI saw's going to eat). In Chinese, the aspect markers are affixes, since they attach only to verbs. By contrast, the sentential particles {le, ma, ba, etc.) are viewed as clitics (Chao 1968:149), since they can prosodically be attached to any category

(C-R. Huang 1985b). Likewise, there are few or no arbitrary gaps in the set of word-clitic combinations, e.g., there is no single unexpected failure of such combinations between a contracted AUX and its host in English, and between a sentential particle and its host in Chinese.

However, arbitrary gaps abound in derivational morphology, and even in relatively productive inflectional morphology. Thus the English verb stride anomalously lacks a past participle

{stridden) for many speakers (Zwicky & Pullum 1983:505). The suffix -men denotes the PL form

^^This does not mean that syntactic conditions are totally irrelevant to affixation. See the inflection-derivation distinction discussed above. 178 of Chinese pronouns. However, the honorific nin [PERSON 2 SG] idiosyncraticaiiy has no PL form

(Chao 1968:244). Furthermore, syntactic rules can affect affixed words, but usually not clitic groups. This is because only an affixed word is a (minimal) constituent to which syntactic rules may refer. A clitic is syntactically attached to a phrase but phonologically attached to the first or last word in that phrase, and therefore the constituency of a word-clitic combination cannot be verified, a s illustrated by saw ’s in The cat I saw’s going to eat. As expected, no syntactic rules refer to a sentential clitic (e.g., =ma) and its host in Chinese. For example, the topicalization rule cannot front the clitic group as it can the affixed word, as we see below.

(33)a. Ni xihuan ta-m en=ma?

you like he-PL pit.

‘Do you like them ?’

b. Ta-men ni xihuan= ma?

c. *Ta-men=ma ni xihuan?

The affixhood of DE is first supported by the high degree of selection for its hosts. The host of DE must be a [4-V] category (verbs or adjectives), as in (34). By contrast, the adjectival de, as in (35) below, may attach to nouns, and exhibits a low degree of selection.

(34) a. [-t-V] hosts

pao-DE kuai zou-DE dong liang-DE yao yan kun-DE yao ming

run DE fast walk DE move bright DE strike eye sleepy DE want die

‘run fast' ‘able to walk’ ‘too bright for eyes' ‘dead tired' 179

b. Noun hosts

*yu-DE da *yifu-DE hao kan

rain DE big clothes DE pretty

(35) a. [+V] hosts

pao de dongzuo zou de yangzi liang de xing kun de mianrong

run de action walk de posture bright de star tired de expression

‘action of running’ ‘way of walking’ ‘bright star’ ‘tired expression’

b. Noun hosts

xia yu de rizi yifu de yangzi

down rain de day clothes de look

‘rainy day’ ‘appearance of clothes’

Likewise, the category of the host of the adverbial de is an adjective in (19a) and (23) and an

adverb in (28a). In (36), however, the host is a noun, although de seems optional here.

(36) Ni yao fangkai sangmen de han, ta cai neng tingjian.

you should let-open voice de shout he only can hear

‘You should shout with open voice in order for him to hear you.’

The hosts for DE are restricted even further to a subset of [-t-V] categories. Auxiliary

verbs {neng ‘can’, yinggai ‘should’, keyi ‘may’, gen ‘dare’, etc.) do not host DE. Therefore, DE

behaves like an affix in its high degree of selection for hosts. The low degree of selection of the

other de’s would qualify them as clitics in the language. 180

If DE is a verbal suffix, arbitrary gaps should come as no surprise, since the combinability of the stem-affix is largely morphological and/or lexical in nature/^ Consider the unexpected failure of DE to attach to yiwei ‘presume’ and renwei ‘opine’, though both are in the same verb subcategory as jianyi ‘suggest’ and xiang ‘think’, taking a prepositional S complement. Since each verb can take a preverbal modifer, as in (37a/c/e), there is no semantic or pragmatic reason as to why for those instances where the modifer converts to a postverbal descriptive following DE, only yiwei and rewei fail, as in (37b/d/f).

(37)a. Ta shifen renzhen de jianyi qu Shanghai,

he very serious de suggest go Shanghai

‘He very seriously suggested going to Shanghai.'

b. Ta jianyi qu Shanghai, jianyi-DE shifen renzhen

he suggest go Shanghai suggest DE very serious

‘He suggested going to Shanghai, and he did so very seriously.’

c. Ta tianzhen de xiang qian hen rongyi zhuan-dao.

he naive de think money very easy earn

‘He thought naively that money was easy to earn.’

d. Ta xiang qian hen rongyi zhuan-dao, erqie xiang-DE shifen tianzhen.

he think money very easy earn and think DE very naive

‘He thought that money was easy to earn, and he thought so naively.’

^^Again, I won’t deny the assumption that inflectional morphology is of syntactic function, as opposed to the lexical function of derivational morphology. However, since inflections are realized within words (by morphology), lexical exceptions seem unavoidable. On the other hand, the number of inflectional gaps is expected to be lower than the number of derivational gaps, because derivation is purely lexical, and therefore less productive. 181

e. Ta tianzhen de yiwei/renwei qian hen rongyi zhuan-dao.

he naive de presume/opine money very easy earn

'He presumed/opined naively that money was easy to earn.'

f. Ta yiwei/renwei qian hen rongyi zhuan-dao, erqie *yiwei-/*renwei-DE shifen tianzhen.

he presume/opine money very easy earn and presume/opine DE very naive

‘He presumed/opined that money was easy to earn, and he did so naively.’

Thus DE behaves like an affix in that it shows arbitrary gaps, though relatively few of them, in contrast to clitics like the de’s and sentential particles, which, as long as the syntactic conditions are satisfied, do not exhibit lexical exceptions for attachment to their hosts.

Finally, syntactic operations of movement can affect the combination/constituency of DE and its host, a behavior which is very natural in the suffix analysis but would be quite unexpected in the clitic analysis (cf. (33)). By assuming that (38a) and (38b) are syntactically related and zhege shiyan ’this experiment’ is the TOPIC, zuo-DE in (38a) can be extracted to the post-topic position, as in (38b). It seem s unlikely that the rightward movement of Han zui you naixin de ren

’even the most patient persons’ is involved, since the string is not a constituent in (38a).

Alternatively, if (38a) is assum ed to be derived from (38b), then a rightward movement of zuo-DE rather than a leftward movement of Han zui you naixin de ren dou is involved, because the latter is demonstrably not a constituent in (38b). Rightward movement seems plausible, since the topic can also right-dislocated, as in (38c).

(38)a. Zhege shiyan, lian zui you naixin de ren dou zuo-DE bu naifan le.

this experiment even most have patience de persons all do DE not patient prt.

Doing this experiment even caused the most patient persons to be impatient.’ 182

b. Zhege shiyan, zuo-DE lian zui you naixin de ren dou bu naifan le.

c. Lian zui you naixin de ren, zhege shiyan zuo-DE dou bu naifan le.

To summarize, the failure in the left expansion test shows that DE is a left bound morpheme, a result which gives rise to two competing hypotheses - either DE is a suffix or an enclitic. At least three lines of evidence, the degree of selection for hosts, arbitrary lexical gaps, and syntactic operations on combinations, suggest that a suffix analysis would be preferable to a clitic analysis.^® Tills conclusion is surprising perhaps, but satisifying nonetheless, not only on empirical grounds but also on metateoretical grounds. An unnecessary analysis has been avoided which would lead to a relatively marked situation, and Chinese grammar does not have to stipulate that DE as a word is prosodically weak, and therefore phonologically bound to the word on its left. I do not deny that the modern suffix DE is historically derived from some free word, probably with an intermediate stage of cliticization; but I maintain that it has been re­ analyzed as an affix, on a par with most affixes in Modern Chinese.

5.3.2. DE as an Inflectional Morpheme

Having established the affixhood of DE, I will now argue that DE must be analyzed as an inflectional morpheme rather than a derivational morpheme, because its presence or absence is required by syntax (cf. syntactic relevance of inflection). Specifically, if the rule of DE-affixation were derivational in nature, operating in the lexicon apart from syntax, a verb with DE would be a separate lexical item related to its source word without DE, shown in (39a), just like the output of other derivational rules, as in (39b-d).

^^he other three lines of evidence have not been found to favor either of the analyses. They are phonological idiosyncrasies, semantic idiosycrasies, and restrictions on the combinability of clitics with inflectional affixes (Zwicky & Pullum 1983). 183

(39) SOURCE WORD OPERATIONS DERIVED WORD

a Verb: tiao ‘jump’ DE-sufTixation Verb: tiao-DE ‘jump’(?) b. Noun: ran ‘person’ Men-suffixation Noun: ren-men ‘persons’ c. Verb: ai ‘love’ Ke-prefixation Adj.: ke-ai ‘lovable’ d. Verb: keneng ‘possible’ Xing-suffixation Noun: keneng-xing ‘possibility’

Since derivational rules are independently needed in the language, then what problems arise from treating verb-DE as a derived lexical item? One theoretical flaw is the loss of the distinction between (39a), which is of syntactic function, and (39b-d), which are of lexical function. For the syntax has to allow the "derived" V-DE in the syntactic environment in (40), but disallow it in (41), if one wants a correct syntactic description of Chinese.

(40)a. Ta tiao-DE kuai. (*Ta tiao kuai.)

he jump DE fast

‘He jumps fast.’

b. Ta chang-DE hao ting. (*Ta chang hao ting.)

he sing DE good hear

‘He sings pleasantly.'

c. Ta chao-DE women shui bu zhao. (*Ta chao women shui bu zhao.)

he make-noise DE we sleep not sound

‘He made such a big noise that we couldn’t get to sleep.’ 184

d. Ta jidong-DE shuo bu chu hua. (*Ta jidong shuo bu chu hua)

he be-moved speak not out expression

‘He was so moved that he couldn't speak.’

(41)a. *Ta tiao-DE sheng. (Ta tiao sheng.)

he jump DE rope

‘He jumped rope.'

b. *Ta tiao-DE zai ma shang. (Ta tiao zai ma shang.)

he jump DE at horse up

‘He jumped down onto the horse.’

By contrast, the derivational (39b) yields plural nouns or pronouns In Chinese. Plurality

is no doubt a semantic feature; but It Is by no means a syntactic feature In Chinese, for there is

no evidence that any syntactic constructions require the presence of -men on nouns, like number

agreement on verbs in English, or on adjectives In German. In Chinese, the subject In Zhexie

ren(-men) zai chang ge ‘These people are singing' Is understood as plural, but the presence of

-men Is optional. Therefore -men is not of syntactic relevance but is instead a derivational

morpheme.

Likewise, no syntactic rule has access to ke-affixation rule in (39c), either. The syntactic

distribution of ke-V is the same as other adjectivals, e.g., ke-ai de hai-zi a lovable child', ai de hai-

zi ‘the child someone loves', zhide ai de hai-zi ‘a child deserving love', da de hai-zi a big child'

all serve as NP's with adjectival modifiers, independent of ke-prefixation. Similarly, there appears

to be no syntactic rule which calls for a noun which must be suffixed with -xing in (39d).

However, one might assum e that (39a) converts a V to an adverb and maintain, as implicit

in Huang & Mangione (1985:85), that DE is a derivational affix, since It induces a change in part 185 of speech. Thus the R-construction with DE in (19c) would be similar to the adverbial construction with de in (19a) so that a unified treatment of de could indicate that the suffixed word is a modifier. By this account, (19c) would mean 'He is sad in crying’. Further, such a unified construction does not have to be suffixed with DE for a resultative reading, as below.

(42) Ta kai che (*DE) hen kuai.

he drive car (DE) very fast

‘He drives (cars) vast.’

There are several flaws with this proposal. First, constructions with adverbial de in (19a) and with resultative DE in (19c) are by no means structurally similar. For instance, it is not possible to conjoin the first or second constituent of each with those of the other as it is for the resultative and descriptive complements in the R-construction (cf. (25)); and (19a) has no resultative reading at all. I do not deny that there is possibly more than one way to express semantic resultatives in the language, as seen in (42), but 1 maintain that the R-construction

morphologically marked with DE and associated with a resulative reading is a distinct pattern.

The resultative without DE in (42) has different external syntax and semantics from the DE-

resultative, as seen in its awkward coordinations with the R-construction in (43).

(43)a. *Ta kai che he pao-DE hen kuai.

he drive car and run DE very fast

He drives and runs very fast.’

*Ta pao-DE he kai che hen kuai. 186

b. *Ta kai che he pao-DE hen lei.

he drive car and run DE very tired

He drove and ran and as a result was very tired.’

*Ta pao-DE he kai che hen lei.

Put simply, the constructions with de, with DE, and without either show too few similarities for a syntactic or semantic rule to unify them in Chinese, even though each may give rise to a resultative interpretation.

Worse still, no prediction is available from the derivational-adverbial account as to where

DE/de is disallowed, as in (42), optional, as in (19a) and (36), or obligatory, as in (19c). More problematic is the fact that wherever DE is required, the derivational-adverbial account destroys the virtue of a foundational consideration in grammar (Anderson 1988, Scalise 1988, Zwicky 1989), namely, that syntax never requires proper presence or absence of a derivational morpheme as it always does for DE in (40) and (41). The derivational account also entails that the complement

VP rather than V-DE is the head of the R-construction, which I will show to be incorrect in the next chapter. Finally, the proposal makes no prediction regarding the productivity of DE-affixation, which on this account would be just as accidental as with the English -ly in careful-ly. By the inflectional account, however, the productivity is entailed by the general syntactic rule sketched below.

For convenience, I assume that the whole resultative construction is marked by a feature

[+R],^‘* and the R-construction sets up three stipulations: its first immediate constituent must be a bare verb (VI), and the second must be an XP which is a [+V] category (VP2), and VI must take the morphological DE-form. All three requirements are met in the good sentences in (40),

^'^As a reminder, the feature [R] has a broader application which includes [DESCRIPTIVE], as I have unified the two subconstructions. 187 and the bad sentences arise from the missing DE. On the other hand, the sentences in (41) are unacceptable since the XPs are [-V] categories, i.e., an NP in (41a) and a PP in (41b). Without

DE, to be sure, they are fine, because the construction is then Assuming this connection, it is not implausible to submit a subcategorization frame for the R-construction, given in (44a), where V-DE subcategorizes for a [-t-V] category. This is because in the frame, V-DE is demonstrably lexical and the following complement is phrasal (XP), and its presence is obligatory, as in (44b). Such a subcategorization is m ade explicit in C-R. Huang (1985a;139), and elaborated

in Dai (1992a;96), who also points out that the subcategorized XP must be a [+Vj.

(44) a. VP[+R] - > V(VFORM:DE] XP[-PV]

b. Ta tiao-DE kuai. (*Ta tiao-DE.)

he jump DE fast

‘He jumps fast.’

Ta chang-DE hao ting. (*Ta chang-DE.)

he sing-DE good hear

‘He sings pleasantly.‘

To Summarize, DE affixation is inflectional because of its clear syntactic relevance - its

presence or absence is determined by syntax.

^^hilethe (un)grammaticaiity in (40) can be accounted for on since grounds, one may argue that (41 a) is syntactically irrelevant, because it can be ruled out by the incompatibility of the lexical semantics between tiao-DE and its complement. My position, however, is that any syntactic construction must be associated with its semantic interpretation and pragmatic value. As mentioned earlier, in the R-construction, VP2 must be understood as the result or the description of VI. In addition, the ungrammaticality of (41b) cannot arise from semantic and pragmatic factors alone, since there is no reason why PP zai ma shang cannot denote the result of VI (cf. the string in the parenthesis in (41b)). Therefore (41b) is indeed ruled out by syntax: either the presence of DE or the PP after VI, or both. 188

The requirement of both DE suffixation and the following [+V] category by the R- construction entails the inflectional status of DE even in the traditional sense. An inflectional morpheme often indicates an apparent syntactic relationship between its host and some other constituent in the same local domain. Here, DE indicates a syntactic relationship between its host

V and the following phrase in that the former apparently requires the latter to be [+V], and for the latter to be understood resultatively, it apparently requires that the former be in the DE-form.

5.4. The Aspect Markers as Inflectional Morphemes

The inflectional status of TA and DE will help to identify the syntactic relevance of the three aspect markers, since they will tje shown to occupy the same verbal morpheme slot, just as [PAST] interacts with the other verbal inflections in English. But to begin with, the aspect markers are required by certain highly marked structures, as in the case with [PAST] in certain

English subjunctives.

The two structures to be examined are the double object construction (Chao 1968:317,

Li & Thompson 1981:375) in (45a), where the verb is followed by an indirect object and a direct object, and the pivota! construction (Chao 1968:124, Li & Thompson 1981:607) in (46a), where the verb is followed by an object and a VP with the object as its subject. Here, one may assume that both matrix VPs are ternary-branching - specifically, the verb in the double object construction subcategorizes for two consecutive object NPs, and the verb in the pivotal

construction for an NP followed by an VP complement, as in the b-forms below. Alternatively, one

may follow Y-H. Li (1990) in assuming a two-level binary-branching structure, as in the c-forms.

(45)a. Wo gei Zhangsan yiben shu.

I give Zhangsan one book

‘I give Zhangsan a book.’ 189

b. Wo [vp [v gei] [^p Zhangsan] (^p yiben shu]].

c. Wo [yp [v [v gei] [np Zhangsan]] [^p yiben shu]].

d. ‘Zhangsan yiben shu, wo gei.

(46)a Wo-men weituo ta zuo daibiao.

I-PL entrust he do representative

'We entrust him to be the representative.’

b. Wo-men [^p [y weituo] [f^p ta] [yp zuo daibiao]].

c. Wo-men [^p [y weituo] [^p ta]] [yp zuo daibiao]].

d. *Ta zuo daibiao, wo men weituo.

One common property for the two sets of b- an c-forms is that the two consecutive complements following the matrix V do not form a single constituent, as seemingly confirmed by the d-forms, in which the two complements cannot be moved together. But this may not be a correct analysis, since grammaticality returns in the topicalization pattern when some "modal" (x/ang, yao 'want') or tense" {mingtian tomorrow’) word is present, as in (47) and (48).

(47)a. Wo mingtian gei Zhangsan yiben shu.

I tomorrow give Zhangsan one book

'I will give Zhangsan a book tomorrow.’

b. Zhangsan yiben shu, wo mingtian gei.

(48) a. Wo-men xiang weituo ta zuo daibiao.

I-PL want entrust he do representative 190

‘We want to entrust him to t>e the representative.’

b. Ta zuo daibiao, wo-men xiang weituo.

Of interest here is the fact that when a past tense word (^uotian ‘yesterday’) is present, the verb in the reversed pattern must be suffixed by aspectual -/e or -guo to be grammatical, as in the c- and d-forms below, whereas such suffixes are entirely optional in the default pattern, as represented by the a- and b-forms. Note that in tiie d-forms, zuotian ‘yesterday’ becomes optional.

(49)a. Wo zuotian gei-le/-guo Zhangsan yiben shu.

I yesterday give perf./exp. Zhangsan one book

‘I gave Zhangsan a book yesterday.’

b. Wo zuotian gei Zhangsan yiben shu.

c. ‘Zhangsan yiben shu, wo zuotian gei.

d. Zhangsan yiben shu, wo (zuotian) gei-le/-guo.

(50)a. Wo men zuotian weituo-le/-guo ta zuo daibiao.

I-PL yesterday entrust perf./exp. he do representative

‘We entrusted him to be the representative yesterday.’

b. Wo men zuotian weituo ta zuo daibiao.

c. *Ta zuo daibiao, wo men zuotian weituo.

d. Ta zuo daibiao, wo men (zuotian) weituo-le/-guo. 191

It is beyond the scope of this chapter to discuss the constituency of these verbal complements.

But the data in the topicalized double-object and pivotal constructions provide direct evidence for taking the syntactic relevance of the aspectual suffixes as indicating inflections, since the attachment of these suffixes marks the a-sentences as distinct constructions from the b-sentences in (49) and (50).

The other reason for regarding the aspect markers as inflectional is that they compete with DE or TA for the slot immediately following a verbal stem.

(51) a. xue-xi-DE hen keku

study DE very hard

b. xue-xi-TA san tian.

study TA three day

c. xue-xi(-le/-zhe/-guo) wenjian

study (perf./dur./exp.) document

In (51), the derivational morpheme -xi ‘review’ can be followed not only by the inflectional DE or

TA, as in (51 a/b), but also by any one of the three aspect markers, as in (51 c). More importantly, no two of the five suffixes may co-occur.^® These facts might give rise to an inflectional

interpretation for the aspect markers in Chinese, since the inflectional DE or TA also occupy that slot. An analogue to this scenario can be found in other languages. In English, for instance, the

relevant morpheme slot in a verb seems to be reserved for only one inflectional morpheme, as

in (4). The aspect morphemes are inflectional, if the following hypothesis (HI) concerning the

^®Guo ‘finish’ in xue-xi guo-le zhe ben shu ‘finish studying this book’ is a complement to xue-xi, homophonous to, but distinct from, the experiential aspect marker -guo (Chao 1968:196). 192 final morpheme slot reservation holds in Chinese.

(52) Hi : In Chinese the verb-final morpheme slot may hold at most one inflection, and only an

inflection may occupy that slot.

Up to this point, the only pair of inflections which can be used to test HI are DE and TA.

It is observed that although the R- and M-constructions are not semantically and pragmatically incompatible, as seen in the discourse in (53a), they cannot co-occur syntactically, as in (53b-h).

(53)a. Ni yao pao-TA san quan, erqie yao pao-DE hen kuai.

you must run TA three circle and must run DE very fast

‘You should run three circles, and run very fast.’

b. *Ni yao pao-TA-DE (san quan) hen kuai.

*Ni yao pao-TA-DE hen kuai (san quan).

*Ni yao pao-DE-TA san quan (hen kuai).

*Ni yao pao-DE-TA (hen kuai) san quan.

c. San quan, ni yao pao-DE hen kuai.

d. *San quan, ni yao pao-TA-DE hen kuai.

e. *San quan, ni yao pao-DE-TA hen kuai.’^

f. Hen kuai, ni yao pao-TA san quan.

g. *Hen kuai, ni yao pao-DE-TA san quan.

h. *Hen kuai, ni yao pao-TA-DE san quan.

’^Note that a homophonie San quan, ni yao pao de ta hen kuai ‘As for these three circles, you need to run fast’ is grammatical, but involves a different structure where ta can only be understood as referential to san quan, and de is of an adjectival type. 193

However, there are other explanations available. One reason for the ungrammaticality could be that V-DE and V-TA have different subcategorization frames, since the former requires a resultative

VP immediately following, and the latter a quantified NP. In the absence of the parenthesized material in the b-set, the subcategorization requirement of either the DE- or TA- construction fails to be met. In the presence of the parenthesized material, either DE or TA is not adjacent to its complement, violating the subcategorization requirements. This would explain the b-set above.

The contrast between (53c) and (53d/e) is covered by an independent constraint on the re­ construction, i.e., the quantified NP object cannot be dispiaced, as shown in (17g). The ungrammaticality of (53h) may be caused by an intervening DE between TA and the quantified

NP. Thus so far, there has been no need to appeal to HI. The only telling case for H1 is the

ungrammaticality of (53g). Given that VP2 in the R-construction may be fronted, as demonstrated

by (32b), the unacceptability seem s to be due merely to the violation of Hi, i.e., pao has two

inflections in the morpheme final slot. Note that without DE, (53f) is good, even though Hen kuai

is not necessarily interpreted as resultative, can mean that 'Hurry up (and run three circles)’.

Now turn to the aspect markers and TA. Experiential -guo is not compatible with TA, and

thus the two cannot generally co-occur. This is because -guo signals an event which has been

experienced at least once in an unspecified time (Li & Thompson 1981:232), but TA defaults to

a future imperative reading. One may also want to attribute the non-occurrence of durative -zhe

with TA to mutual incompatibility, as -zhe defaults to a present continuous tense. But this will not

work, since the two can occur together in a discourse such as (54a), even though they are

syntactically incompatible, as seen in (54b).

(54)a. Wo (zai) yun-zhe xigua, yun-TA san che jiu xiuxi.

I (at) transport dur. water-melon transport TA three carts then rest

I am transporting water melons by cart. I’ll take a rest after three loads.' 194

b. *Wo (zai) yun-zhe-TA (-TA-zhe) san che xigua...

Similarly, there seem s to be no other reason why 4e and TA cannot co-occur in the sam e word or construction in (55b), as the suffixes are semantically and pragmatically compatible in (55a), where -le denotes the future perfective.

(55)a. Ni yao zhengzheng pao-TA san quan, pao-le san quan hou, jiu chu han le.

you should total run TA three circle run pen. three circle after therefore out sweat pit.

‘You need to run a total of three circles before sweating.'

b. *Ni pao-le-TA (-TA-le) san quan jiu chu han le.

The non-co-occurrence of DE with an aspect marker cannot be satisfactorily explained, either, for the two categories may occur in discourse, as in (56a/b/c), in contrast with (56d), where the two are not allowed to occur in the same morpho-syntactic structure,

(56)a. Ta chi-le (fan), chi-DE hen bao.

he eat perf. (rice) eat DE very full

‘He had the meal, and he is full.’

b. Ta yong-guo (kuaizi), yong-DE hen hao.

he use exp. (chopsticks) use DE very well

‘He once used chopsticks, and handled them very well.’

c. Ta chi-zhe (fan), chi-DE hen kuai.

he eat dur. (rice) eat DE very fast 195

‘He is eating very fast.’

d. *Ta chi-ie/guo/zhe-DE hen bao (fan).

Such a non-co-occurrence restriction on the aspect markers with respect to TA or DE is consistent with HI : the one-to-one correspondence tœtween an inflection and the final morpheme slot.^® Between DE or TA and the aspect markers, DE or TA gets to fill that slot. But they are merely structural winners. When TA or DE is present, as in the second clause in (55a) or (56a-c), the relevant aspectual reading is still present, indicating only the suppression of a corresponding aspect marker. More crucially, in the absence of contexts, e.g., without the first clause in (54a) or (56a-c), any of the compatible aspectual interpretations is available in the R- or M-construction.

This suggests that aspect markers are incompatible with TA and DE in morphosyntax, but not

^®An apparent counter-example would be the -le- infixed in the verbal delimitative reduplicates (cf. 4.1.1 & 6.0), where DE and -le- do not compete for the final slot, and yet the result is still ungrammatical, as in (i) but cf. (ii). i. *Men, ta qiao-le-qiao-DE hen qing, door he knock-perf.-knock-DE very light ii. Ta qiao-le-qiao men, qiao-DE hen qing. he knock-perf.-knock door knock very light ‘He had a knock at the door a little, and did so very lightly.’

However, the slot competition may not be the issue here, since the delimitatives in general (without -le-) are also ungrammatical with DE. iii. *Men, ta qiao-qiao-DE hen qing. iv. *Men, ta qiao-yi-qiao-DE hen qing.

The ungrammaticality may be due to the functional incompatibilty between the delimitative and the resultative. The former focuses on the delimitative meaning of doing something for a little while, wheras the latter functions to signal that a given event leads to a certain result. This point is made clear in Li & Thompson (1981:235), although they use it to explain why resultative compounds (da-kai [hit-open] ‘open ) cannot be delimitatively reduplicated. Thus in separate sentences, the delimitiative and the resultative are compatible, as in (ii). 196 necessarily in semantics and pragmatics.’® Such slot competition is parallel to the case in

Georgian found by Anderson (1986:8): the formal markers v- for 1st subject and g- for 2nd object are mutually exclusive by virtue of their competition for the same position; the v- is the winner here, and although the g- is suppressed, 2nd objects are still interpreted for verbs with 1st subjects.

Although HI is true, it remains stipulative in morphology. However, HI can be made to follow from further assumptions on other principles of grammar. I make a hypothesis (H2) for universal grammar, as below.

(57) H2: A verb (or any word) heads only one construction.

H2 can be narrowly interpreted as licensing individual verbal subcategorization frames in grammar. Thus verb believe in English can independently head (at least) two constructions, subcategorizing for either an NP {believe the man) or an S {believe that the earth is flat), but not both {*believe the man that earth is flat). By further assuming that an inflectional morpheme marks only one head in Chinese, the co-occurrence of any two of the five verbal inflections is expectedly ungrammatical, since the verb would illicitly head more than one unique construction. Therefore, the non-co-occurrence of two inflectional morphemes in a verb, i.e., the constraint on the final morpheme slot (HI), follows as a corollary from H2.

Recapitulating, the aspect markers are morpho-syntactically incompatible with DE and TA.

Their syntactic relevance resides in their marking a family of syntactic constructions, distinct from the relatively marked R- and M-constructions. In the syntax of Chinese, a verb suffixed with an aspect marker is either intransitive or subcategorizes for a complement other than the sort of complements in the R- or M-constructions. In addition, the aspect morphemes have a place in

’®Thus although HI is reminiscent of the DE affixation rule proposed in Huang and Manjione (1985:85), the two are different in nature. The former deals with the final inflectional slot appealing to no semantics, but the latter is a derivational restriction appealing to semantic coherence. 197 some reversed double-object and the pivotal constructions, although they usually do not overtly mark a syntactic relationship between the host verb and some other constituent in a VP, like

[PAST] -ed in English.

5.5. The Deverbal Nominal -tou as an Infectional Morpheme

Pronounced segmentally the sam e as the nominal formative suffix -tou in mu-tou ‘wood’, lao-tou ‘old man’, or shi-tuo ‘stone’, as well as the word tou ‘head’, the deservative suffix -tou investigated here has oniy rareiy attracted linguists’ attention, and even fails to be included in the lexical entries of standard dictionaries.^ Chao (1968:243) is apparently the first one to note this category. The category is rediscovered by Li & Thompson (1981:45-6). According Chao, the pattern in (58) is productive in the context of verb you ‘have/exist’ or mei(you) ‘not have/exist’^^ followed by an action v erb ^ with the suffix -tou.

(58) a. Zhe dianying you kan-tou.

this movie have see des.

This movie is worth seeing.’

b. Na-ge difang meiyou fangwen-tou.

that M place not-have visit des.

That place is not worth visiting.’

^E .g., Xiandai Hanyu Cidian [A Dictionary of Modern Chinese] (1978): Zhongguo Shehuikexueyuan Yuyanyanjiushuo [Linguistics Institute of the Chinese Academy of Social Sciences], Beijing: Zhangwuyinshuguan [Commercial Press].

^’Henceforth, I use (mePjyou to represent the three possibilities you, meiyou and mei.

^1 tentatively follow Chao in assuming the lexical category to be a verb for convenience, though later on I will argue that it is deverbalized to a nominal. 198

Since Ihe construction at>ove containing -tou denotes the meaning 'It is (not) worth V-ing* or

T here is no point in V-ing', I call the morphological category the deservative (des.) -tou, and the construction the deservative (D-) construction.^^

5.5.1. The Deservative -tou as a Suffix

My contention here is that -tou is a word suffix like the homophonous nominal suffix, as assumed in Chao (1968). -Tou does not behave like a word since nothing may intervene between the verb and this suffix, as in (59a), where a resumptive pronoun intervenes illegitimately. Though the D-construction in (59c) is semantically and pragmatically compatible with the double object construction in (59b), the logical indirect object Zhangsan is not allowed to occur after the verb, further showing that -tou must directly attach to the preceding verb.

(59)a. *Zhe dianying you kan ta -tou.

this movie have see it des.

b. Women meiyou song Zhangsan najian liwu

we didn’t give Zhangsan that present

‘We didn't give that present to Zhangsan.’

c. Najian liwu meiyou song (*Zhangsan) -tou.

that present not-have give (Zhangsan) des.

That present is not worth sending (to Zhangsan).’

^^It should be further noted that you kan-tou in (58a) is not a compound word, since a phrase may intervene between you and kan, as illustrated below. i. Zhe dianying you sheme kan-tou? this movie have what see -tou ‘What’s the point in seeing in this movie?’ 199

The deservative -tuo is usually weak prosodically (i.e., unstressed and toneless), which would give rise to an enclitic analysis. By this clitic account, -tou would be a syntactic word, which phonologically attaches to its preceding word, and thus is inseparable from it, like the classical clitic cases of the contracted English auxiliary verb -’s for is in He's coming, and in

Chinese sentential (final) particles, e.g., the question word ma in Ta hao=ma [he good prt.] Is he all right?’ (cf. ch.7). However, the morpheme -tou cannot be a clitic according to the cross- linguistic evidence which separates clitics and inflectional affixes/morphemes suggested in Zwicky

& Pullum (1983:502).

First, -tou is highly selective of its host. Unlike a clitic which may be phonologically dependent on any category, -tou must attach to a [+V] category. It further selects its host by specifying particular [+V] categories. Stative verbs, traditionally called adjectives, are excluded, as are the resultative compound verbs, as exemplified in (60a) and (60b) respectively.

(60)a. *Zhe dianying you yisi-tou.

this movie have interesting des.

‘?This movie is worth arousing interest from.'

b. *Zhege ren mei kan-jian-tou.

this person not-have look-see des.

‘This person is worth seeing."

While one may attribute the ungramaticality of (60a) to semantics, as the English translation suggests, (60b) cannot be so explained. This is because if one of the components of the compound, kan ‘look’ or jian ‘see’, is missing, the grammaticality would be restored. Neither the bisyllabicity nor the morphological composition of kan-jian seems to play a role here, since the 200 bisyllabic verb-object word dan-xin [carry-heart] ‘worry’ can act as the host to the suffix.

(61) Zhejian shi meiyou dan-xin-tou.

this thing not have worry des.

‘There is no point in worrying about it.’

Thus morphosyntactic selection, one of the properties found for affixes and not for clitics, would be the best term for the attachment of -tou.

Moreover, if -tou is a verbal suffix, arbitrary gaps should not come as a surprise, since, unlike the exceptionless attachments expected of a clitic, the combinability of stem-affix is largely morphological and/or lexical in nature. Consider below the awkwardness of attaching -tou toxuexi

learn’ in contrast to its synonym, xue.

(62) Yingyu meiyuo xue/?xuexi-tou.

English not-have learn/learn des.

’English is not worth learning.’

Similar arbitrary gaps are found with the lexical items goumai ‘purchase’, yilai ‘rely on’, and so on.

One would expect the meaning of a clitic group to be semantically compositional. By

contrast, a stem-affix combination may yield semantic idiosyncrasies. This is borne out in the V-

tou combination. In (63a), the verb shou can only mean ‘talk about’, tell’ or ‘explain’. In the

corresponding D-construction in (63b), however, an additional meaning ‘argue about’ may be

idiosyncratically introduced.

(63)a. Ni he ta shuo zhejian shi.

you with he talk this thing 201

‘You talk with him about this.’

b. Zhejian shi he ta mei shou-tou.

this thing with he not-have argue des.

‘There is no point in arguing about this with him.’

Finally, based on the fact that a stem-affix combination is a (minimal) syntactic constituent

(i.e., a word), but a clitic group may not be, one would expect syntactic rules to refer to ttie V-rou combination. This expectation is fulfilled. The constituency of the D-construction in (64a) may be rearranged for emphatic effect, as in (64b) where -tou still follows the verb. In (64c), the V-fou can even occur alone as the subject for a liberal speaker.

(64)a. Zhe dianying yidian meiyou kan-tou.

this movie at-all not-have see des.

‘This movie is not worth seeing at all.’

b. Zhe dianying yidian kan-tou ye meiyou.

this movie at-all see des. even not-have

‘This movie is not worth seeing at all.’

c. Zhe dianying, kan-tou shi yidian ye meiyou.

this movie see des. be at-all even not-have

‘As for this movie, it is not worth seeing at all.’

To summarize, failure to pass the expansion test shows that -tou is a left-bound

morpheme. It does not behave like an enclitic because it exhibits at least four affixal behaviors 202 uncharacterisitic of clitics. The high degree of selection for hosts, arbitrary lexical gaps, semantic idiosyncracies, and syntactic operations on combinations all indicate that the deservative -tou is a suffix in Chinese. This conclusion is not surprising, given that the deservative -tou is etymologically and historically related to the nominal suffix -tou, which is in turn related to the word tou ‘head’.

5.5.2. Morphosyntactic Government: -tou as an Inflectional Morpheme

As stated earlier, inflectional morphemes are those that are syntactically relevant.

Accordingly, -tou should be a. alyzed as an inflectional rather than derivational morpheme, since its presence or absence is determined by syntactic rules. In the D-construction, the head of the complement must be in the morphological fou-form, since, if the head is monomorphemic in the absence of -tou, either the result is ungrammatical in the presence of you, as in (65a), or there is no deservative reading in the presence of meiyou, as in (65b), though the result is still grammatical.

(65)a. *Zhe dianying you kan.

this movie have see

b. Zhe dianying meiyou kan.

this movie not-have see

'This movie hasn’t been seen yet.’

*‘This movie is not worth seeing.'

The syntactic relevance of -tou thus qualifies it as an inflectional morpheme, marking the complement of the D-construction. The strongest evidence for the inflectional nature of -tou comes from its relevance to the external syntax. Since {mei)you takes D-complements (in 203 coordination) in (66a) as well as nominal complements (in coordination) in (66b), and since the two complement constructions are apparently compatible In a discourse, as in (66c), one might predict the acceptability of the coordinations in (66d) or (66e). The fact that neither is acceptable shows that the language has at least two (mei)you's, one with a possessive reading illustrated in

(66b), the other introducing a deservative reading as in (66a). The deservative -tou co-occurs only with the latter and Is syntactically Incompatible with the former, resulting In the ungrammaticality In (66d/e).

(66)a. Zhe dianying meiyou xinshang he jleshao-tou.

this movie not-have enjoy and Introduce des.

‘This movie is not worth enjoying and Introducing.’

b. Zhe dianying meiyou shengyin he yanse.

this movie haven’t sound and color

‘This movie has no sound and color.’

c. Zhe dianying meiyou shengyin, meiyou xinshang-tou.

this movie haven’t sound not-have enjoy -tou

‘This movie Is soundless and therefore not worth enjoying.’

d. *Zhe dianying meiyou shengyin he xInshang-tou.

e. *Zhe dianying meiyou xinshang-tou he shengyin.

Thus the syntactic relevance of -tou and Its Inflectional nature are borne out even In a traditional sense: the presence of -tou marks a syntactic relationship between its host (the V) and some other constituent ((mei)you). The syntax requires that -tou co-occur with the deservative (mefjyou. 204

Such co-occurrence entails that the upper verb (mei)you is also obligatory in the D-construction, as shown in (67a), where the ungrammaticality remains even in the presence of a synonymous verb zhide ‘worth’. In the absence of -tou, there is no requirement for (mei)you, or more precisely, it must be absent, a s illustrated in (67b).

(67)a. *Zhe dianying (zhide) kan-tou.

this movie (worth) have see des.

b. Zhe dianying zhide kan (*-tou).

this movie worth see (des.)

'This movie is worth seeing.’

I propose for the time being that in the D-construction the main verb (mei)you subcategorizes for a D-ccmplement in which feature D percolates to the head of the complement and is morphologically realized as -tou, as shown in the structure below.

(68) [^p ... (v (mei)you [yp ... [y kan-tou]]]]

In (68), a relationship in morphosyntactic government holds in a traditional sense. The verb

(mei)you governs the morphological (D- or tou-) form of the head of the complement. This is on a par with standard instances of morphosyntactic government in other languages. For instance, a verb or preposition in German governs (or determines) the morphological case of its object NP.

For similar examples in other European languages, see Lyons (1968:239-41). A clear case of government in English is seen between an auxiliary verb and the morphological form ([PSP],

[PRP], [PAS] or [INF]) of the head V of its complement VP (Pullum & Wilson 1977:741). I have thus shown that Chinese possesses morphosyntactic government like other languages, as 205

(mei)you governs the morphological form of the head of its complement in the D-construction.^"*

5.5.3. -Tou as a Deverbal Nominal Inflection

So far I have assumed that the D-complement is a VP. However, this is not the case.

Chan (1984:133) suggests that the slot after you ‘have’ (non-deservative) shows promise for distinguishing stative verbs from other verbs. Because you is used primarily for denoting possession, and the possessed is usually a nominal, a verb in the slot after you is often referred to as static, subject to nominaiization. Perhaps since the deservative (mei)you Is etymologically related to possessive you, kan in (69a) is deverbalized to a nominal in the structure, since it can be modified by the adjectival question word sheme ‘what’, but not by adverbial zenme ‘how’.^^

Sheme is demonstrably an adjective in the language, which modifies only nominals, whereas zenme is an adverb, modifying non-nominals, as illustrated in (69b/c).

(69)a. Zhe dianying you sheme/*zenme kan-tou?

this movie have what/how see des.

‘What is the point in seeing this movie?’

b. Sheme xuexi? c. Zenme xuexi?

What kind of/*How to study? How to study/*What kind of study?

Supporting evidence for the nominaiization comes from (70a), in which an unambiguous AP {hen

da de ‘very big') can modify kan ‘seeing’ in the D-complement (also see (64)), but not in other

structures containing a verbal kan ‘see’, as in (70b).

^"*Note that like the auxiliary VP in English, the D-construction involves no traditional feature of morphological case.

^^Nor can it be nmodified by any VP-adverbs like mashang ‘immediately’, jiu ‘soon’, or jingchang ‘often’. But it can be modified by an adjectival yidian ‘a little, at all’, as in (64). 206

(70) a Zhe dianying you hen da de kan-tou.

this movie have very big de see des.

‘This movie is very worth seeing.’

b. Wo-men (*hen da de) kan zhe dianying.

we (very big de) see this movie

‘We are watching this movie (*very big).’

Given the nominaiization analysis, one might argue that the deservative could be a derivational suffix, since it deverbalizes the original verb into a noun, and the ability to change part of speech is not characteristic of inflection but of derivation. However, there is reason to believe that an independent nominaiization via the process of conversion or zero derivation (Dai

1990a:30) takes place before infectional suffixation, as xue-xi ‘study’ in (69b/c). Therefore the D- complement must be an NP. Prototypical NPs like qiang ‘wall’ oryizi chair’ are ruled out in this position, not because they are not NPs, but because they do not have a verbal meaning, and are thus incompatible with the semantics and pragmatics of the D-construction. English has an analoguous case. The [PRP] -mg form in the verb writing as in stop writing on the blackboard

is determined or governed by the verb stop, and hence inflectional. But the possible co­ occurrence of V-/ng with the as in stop the writing on the blackboard may give rise to a derivational appearance of -mg, since writing is a noun now. However, this is an illusion, since stop has multiple-subcategorization frames. It takes an NP complement (the lecture) or a [PRP]

phrase (writing). The two complements are syntactically incompatible, as evidenced by

coordinations data: *stop the lecture and writing on the blackboard vs. stop the lecture and the

writing on the blackboard. The latter is grammatical, since there is an independent derivation to

form a gerund noun by attaching -mg to the verb in the language. Therefore the inflectional

[PRP]'3 ability to change parts of speech is not a consideration here. The Chinese counterpart 207 is still ungrammatical, as in (66), since the language has two distinct words (mei)you. They are incompatible in syntax.

Recapitulating, if my nominaiization analysis is correct, then Chinese is for the first time shown to have nominal inflections.^ Accordingly, the lower VP in (68) should be changed to

NP, with the resulting rule VP[D] > HW[28] NP[NFORM:-fou] (mei)you (not) worth’.

5.5.4. Grammatical Relations and the Bare Deverbal

The D-construction has one more syntactic requirement; no complements, though semantically and pragmatically compatible, may follow -tou, in addition to the non-intervention constraint between the deverbal nominal and -tou, as in (59). Example (71a) is unacceptable since a resumptive pronoun follows the deverbal, and so is (71 c/d), in which the logical (in)direct object in (71b) follows the deverbal.

(71)a. *Zhe dianying you kan-tou ta.

this movie have see des. it

b. Women meiyou song Zhangsan najian liwu

we didn’t give Zhangsan that present

'We didn’t give that present to Zhangsan.’

c. Najian liwu meiyou song-tou ("Zhangsan).

d. Zhangsan meiyou song-tou ("najian liwu).

Thus from the examples up to this point, it appears that the deverbal is transitive but "bare", with

^®As a reminder, the plural suffix -men is derivational rather than inflectional in the language, because its presence is not required by the syntax. 208 the obligatory absence of object complements, and the subject of the sentence is interpreted as the object of the deverbal. This observation is nevertheless not quite correct. The verb in the a- forms below has a postverbal cognate object (yi yan ‘one eye (look)’, yi jiao ‘a foot (kick)’, cf.

4.1.1) followed by the object; in the corresponding D-construction in the b-forms, the cognate object is promoted to subject, with the direct object the topic.

(72)a. Ta kan yi yan zhezhang hua.

he look one eye this picture

‘He took a look at this picture."

b. Zhezhang hua, lian yi yan dou mei kan-tou.

this picture even one eye all not-have look des.

‘There is no point in having a look at this picture at all.’

(73)a. Ta ti yi jiao zhege zhuqiu.

he kick one foot this soccer ball

‘He gave a kick at this soccer ball.’

b. Zhege zhuqiu, lian yi jiao dou mei ti-tou.

this soccer even one foot all not-have kick des.

'There is no point in giving a kick at this soccer at all.’

As cognate objects function as adverbials, postverbal frequency or duration NP adverbials are

expected to occur in the subject position of the D-construction as well. This is borne out in the

following examples. 209

(74) a. Ta da le yixia

he hit perf. once

‘He hit once.’

b. Yixia mei da-tou.

once not-have hit des.

It is not worth hitting once.’

(75) a. Ta zhu-le san tian.

he reside perf. three day

‘He stayed three days.’

b. San tian mei zhu-tou.

three day not-have reside des.

‘It is not worth staying three days.’

Still, it is not general enough to state that all and only postverbal NPs can be the subject of the

D-construction, since NP adverbials may occur preverbally in non-D-constructions. Further, it is not clear whether the subject PP in (76b) begins as ore- or postverbal In (76a).

(76) a. Ta zhu zai xuexiao/zai xuexiao zhu.

he live at school/at school live

'He lived at the school.’

b. Zai jia meiyou zhu-tou.

at home not-have live des. 210

T here is no point in living at home.’

Finally, (77) and (78) demonstrate that a preverbal PP can be the subject of the related D- construction, since some PPs never occur after the verb.

(77)a. Wo dui ta bu guan-xin.

I to he not concern-heart

I did not show concern for him.’

b. *Wo bu guan-xin dui ta.

c. Dui ta meiyou guan-xin-tou.

to he not-have concern-heart des.

‘There is no point in showing concern for him.’

(78) a. Ta chao nan zou.

he toward south walk

‘He walked toward the south.’

b. *Ta zou chao nan.

c. Chao nan mei zou-tou.

toward south not-have walk des.

‘There is no point in walking toward the south.’ 211

Thus from (72)-(78), I conclude that a dependent (argument or modifier) of the deverbal in a default structure, whether an NP or PP, pre- or postverbal, may be promoted to subject in its corresponding D-construction.

5.6. Summary

This chapter has shown that TA, DE, -tou, and the aspectual markers -/e, -zhe and -guo are best treated as inflections, thereby demonstrating the need to recognize inflectional morphology in Chinese. This result is surprising, given the lack of attention given to TA and 4ou in the literature and the traditional view that DE is an acategorial particle word or clitic. It is further suggested that inflectional morphology might be more widespread than is usually supposed. I have eliminated one example of a language without inflection, which should encourage interested linguists to examine other cases of supposedly non-inflecting languages to see whether they exhibit the same behavior.

The purpose in studying inflectional morphology in Chinese, or in any language, is two­ fold: first, to understand morphology proper, i.e., the word internal structure in the language; second, to see how syntactic relationships are realized through morphology. 1 have focused on the latter. The six inflections mark distinct constructions in Chinese, and their respective syntactic relevance is at least as theoretically interesting and revealing as that of English and other languages. CHAPTER VI

INFLECTIONAL MORPHOLOGY; CONSEQUENCES

6.0. Introduction

This chapter explores the empirical, methodologial and theoretical consequences of the inflectional analysis of DE, TA, -le, -zhe, -guo, and -tou presented in chapter 5, Empirically, their inflectional characteristics noted in the traditional grammar (Bloomfield 1933:222-4) fall out naturally. First, they can be suffixed to virtually every lexical item of the verbal or nominal subclasses they select for; thus the processes are morphologically general and productive, much more productive than derivational affixation, though less general than the word-clitic combinations.

Second, there is no change in part of speech after suffixation. The suffxed word remains a verb or a noun in each construction. Third, DE, TA, and -tou signify the syntactic relationship between their hosts and the fallowing VP, quantified NP, and deverbal NP respectively, and the aspect suffixes mark constructions different from the R- and M-constructions. Fourth, they are located at a word-marginal position, and thus occur after derivational morphemes and close the word.

Now I reinforce the distinction between inflectional and derivational morphology by comparing the inflectional deservative -tou with the derivational nominal formative -tou in mu-tou

'wood' or lao-tou ‘old man’. The most telling characteristic of the deservative -tou is that it serves to indicate the syntactic relationship between its host and the preceding verb (mei)you. More to the point, the existence of the deservative form is required by a particular syntactic construction, the D-construction. By contrast, there is no syntactic rule in Chinese which explicitly requires the presence of the nominal formative -tou. Imagine that the noun lao-tou ‘old man’ occupies subject position, (in)direct object position, or prepositional object position in a sentence. The syntax of

212 213

Chinese only requires that the position be taken by an NP, but never requires that the head of the NP be in a particular -tou form in morphology. Expectedly, nouns without the formative -tou, such as lao-ren ‘old man', lao-hu ‘tiger’ or qi-che car', may also occur in those same NP positions. In the absence of the derivational formative -tou, the original syntactic construction remains unchanged, hence the syntactic irrelevance of the formative -tou. The inflectional deservative -tou, by contrast, figures prominently in syntax. Its absence makes the D-construction ungrammatical, as shown in 5.6. Furthermore, by assuming that the syntactic rule referring to the inflectional -tou is structurally general, it is not surprising that the deverbals to which the deservative -tou can be suffixed constitute an open-ended list in the lexicon, meaning that the process is productive. This contrasts sharply with the unproductive nominal formative -tou, which can attach to only a dozen stems listed in the lexicon. Any theory which aims to capture these distinctions has to assume the distinction between inflectional and derivational morphology.

The inflection’s peripheral position in a word in turn sheds light on the methodology of determining wordhood, a perennial issue in Chinese (Dai 1990a). To date, there has been no serious morphological attempt to determine word boundaries in the literature. As a result of the analysis in this thesis, a minimal morphological criterion can be established to identify word boundaries. One can at least use the five verbal inflectional morphemes to locate verb boundaries and demonstrate whether a verb is monomorphemic, derived or a compound. For example, each of the inflections can only follow -xi in the verb xue-xi ‘study’, but never intervene between xue and -xi, indicating that the verb boundary is after -xi, even though xue learn’ as a free verb can occur with these suffixes elsewhere in the syntax. Therefore, xue-xi is a bimorphemic derived word. In addition, morphological approach to part of speech is conceivable.

It is safe to assume an item with a verbal suffix to be a [+V) category.

There is an apparent exception to the edge position of inflections in Chinese. The perfective -le in kan-le-kan (ta) [look-perf.-look (he)] ‘have had a look at him’ is infixed in the delimitative verb (kan -> kan-kan, cf. 4.1.1) after the derivational reduplication. Since I assume 214 the same syntactic function and relevance of the affix le in the delimitative a s in the corresponding non-deiimitative {kan-le ((a) ‘have looked at him"), this is a case of inflections within derivations.

The current framework allows for such cases, since a morphological rule, inflection or derivation, may choose (or stipulate) either infixation, prefixation or suffixation as the operation type on the target stem having a particular internal morphological structure, even though the same rule selects one of the other affixation types elsewhere. For instance, Nevis & Joseph’s (1992) argue that the Lithuanian -sr (reflexive) is an infix if the stem has one or more prefixes; elsewhere it is a suffix (cf. 5.0). The authors also cite some similar cases in ether languages. In Afar (Fulmer

1991), the plural marker -n- is prefixed to vowel initial verbs (n-ookom-e [PL-won-perf.] 'we won’)

but suffixed to consonant-initial verbs (athn-a [do-PL-imperf.| ‘we do ). The possessive marker -m-

in Miskito (Ultan 1975:170) occurs as an infix or as a suffix, according to the alienability of the

nominal host (na-m-pa ‘your tooth' vs. nina-m ‘name’). In Atayal (Moravcsik 1977:76), stem subcategorizations determine the affixation type of the actor/focus marker -m- (m-citiq ’fight’vs. k-m-uqus ‘scrape’). The Estonian indefinite -ge- (Nevis 1984) can be an infix or suffix (kes-ki

‘whoever’ [nominative] vs. kelle-gi-le ‘to whomever’ [allative]).

Inflectional infixation should be distinguished from a derivational rule invoking a stipulated

form as the target stem, such as related in relatedness in English, which is another case of

inflection (-ed) within derivation (-ness). But the stipulated "past participle" form related is no

longer inflectional in that the form participates in no individual syntactic constructions like the

perfective and passive in English.

These two cases, the inflectional infixation (or other processual operations, see 6.5.3) and

the stipulated form for derivation, are two types of apparent counter-examples to the traditional

observation on inflection outside derivation (cf. 5.0), which a morphological theory must be able

to account for.

The rest of this chaper considers some theoretical consequences of positing inflectional

morphology for Chinese. I am to show how some long disputed issues in Chinese syntax can 215 be better understood by considering the inflectional morphology of the language. They are the identification of the head in the resultative construction (6.1), the artifact of the Phrase Structure

Condition (6.2), and the rethinking of Case theory for Chinese word order (6.3 & 6.4). The reduced A-not-A question (6.5) will be discussed by reference to processual morphology. Finally,

I will demonstrate that the derivational-inflectional distinction in Chinese provides difficulty for the theory of Lexical Morphology and Phonology which resorts to levels in word formation process to capture such a distinction (6.6).

6.1. The Head of the Resultative Construction

One of the hotly debated issues in Chinese is whether VI or V(P)2 is the head or matrix verb in the R-construction (e.g., C-T. Huang 1988 (for VI as the head), Tai 1989a (with some reservations for VI), C-R. Huang 1990 (for V2 as the head)). A clinching arguemnt cannot be made effectively on semantic grounds. The translation for Ta pao-DE kuai [he run fast], for instance, can be either ‘He runs fast’, showing that VI pao-DE is the head, or ‘He is fast in running’, where VP2 kuai is treated as the main verb. On the basis of syntax and morphology, the debate centers on six matters concerning VI and V2 in the R-construction, involving the formation of A-not-A questions, the attachment of aspect markers, the scope of negation, the coreferential relation of pronominals, the morphological complexity of verbs, and the formation of yes-no questions. But so far neither side seems to have offerred a convincing argument, and the issue is still formally debatable (see the summary in C-R. Huang (1990:327-30) and Dai (1992a;84-

7)).

Now that DE is demonstrably inflectional, the head of the matrix VP must be VI, rather than VP2. Following Zwicky (1985b), the notion of head in syntax is crucial for the percolation of

syntactic features relevant to the external syntax of the construction, and the head is where the

morphosyntactic locus is found in the construction. For example, the following are the heads to which the inflection attaches in their respective constructions in English: N in Det + N {the 216 students), V in V + NP {builds the house), AUX in AUX + VP (was going to school), etc/ Back to the case of the R-construction, VI is the head (or the matrix verb), since it is VI. not V2, that is obligatorily inflected with DE, i.e., VI is where the morphosyntactic locus of the R-construction is realized, marking the construction in contrast to other non-R constructions in the language.

If one adopts the Head Feature Convention along the lines of Gazdar et al. (1985:94) and instantiates the feature [-t-R] on the matrix node of the R-construction, then [+R] percolates from the matrix VP to VI, morphosyntactically represented by DE. Of course, one does not have to assume the feature percolation analysis; but suffice it to say that since the inflection DE marking the matrix R-construction is located in VI rather than VP2, VI is the head of the R-construction.

it is noted that aspect markers can attach to the head of VP2 but not to V1 in the R- construction. However, the location of the aspect markers is irrelevant to determining the syntax

(and semantics) of the whole R-construction, in contrast to the crucial role of DE. As argued earlier, VI cannot be aspectually inflected because of the final morpheme slot restriction (HI, cf.

5.4). Aspectual inflection only marks the matrix verb of a default VP construction, which VI-DE in the R-construction may, but does not have to, subcategorize for as its resultative complement.

This seems to the strongest claim one can make from the fact that only V2 can be aspectually marked, which never entails that V(P)2 is the head verb in the R-construction. Thus, technically, the [+R] percolates from the matrix of the R-construction, and an aspectual feature percolates from the matrix of the resultative complement.

Four other pieces of cross-linguistic evidence are provided in Dai (1992a:95-108) to further support taking VI as the head of the R-construction. When more data are consulted, it becomes evident that the R-construction is the hyponym of VI, and VP2 is the semantic argument of VI.

In addition, VI subcategorizes for the VP2, has the same distribution as the R-construction, and is obligatory in the construction. Thus besides being the morphosyntactic locus, VI, but not VP2,

^ Note that AUX in AUX + VP (was going to school) is the head because was is where the morphosyntactic locus is found for the matrix construction: tense, finiteness and agreement features for the external syntax (so -ing in going is irrelevant in this regard). 217 has the universal head properties of being the semantic functor, subcategorizand, distributional equivalent, and obligatory element in the construction (see Bauer (1990:2) for a review of these

properties). Therefore, the account turning VI-DE to an adverbial adjunct to VP2, as implicit in

Huang & Manjione (1985:85), is untenable, since it fails to capture the head properties VI possesses in the R-construction.

6.2. The Chinese Phrase Structure Condition as an Artifact

■Another theoretical controversy in the literature which my analysis of inflections helps to illuminate is the well-known issue of the Chinese Phrase Structure Condition (PSC) proposed by

C-T. Huang (1982 & 1984), stipulating that the head (the verb or VP) branches to the left only once, and only on the lowest level of expansion. One putative motivation for the PSC is to rule out the co-occurrence of a resultative complement (la) and an NP object (1b) following a verb, as in (1 c) and (1 d), presumably because this would require that the verbal category left-branching twice.

(1)a. Ta tiao-DE hen kuai.

he jump DE very fast

'He jumps very fast.’

b. Ta tiao sheng.

he jump rope

‘He plays rope-skipping.’

c. *Ta tiao-DE hen kuai sheng.

he jump DE very fast rope 218

d. *Ta tiao-DE sheng hen kuai.

he jump DE rope very fast

The PSC has been Disputed, though, opinions on it ranging from criticisms (e.g., Ernst (1988),

Tai (1989a & b), Y-H. Li (1990)) to refutations (e.g., C-R. Huang (1990), Jin (1991)) for its descriptive inadequacy and lack of real motivations (also cf. 3.3). In the very least, the PSC is parochial and does not offer an explanation for the distinctions in (1).

If the PSC does not exist in Chinese (C-R. Huang 1990:324), the ungrammaticality of

(1 c/d) must be accounted for independently. In the current framework, the account is from the hypothesis that a verb heads only one construction (H2, cf. 5.4). My analysis of inflections presupposes a subcategorization frame at the V® level. For instance, the head V-DE subcategorizes for a [+V] phrase in the R-construction, as in (la). Independently, the head V

(without suffix DE) subcategorizes for an NP object in the VO-construction, as in (1b). Since DE marks one unique head in a subcategorization frame, H2 also effectively rules out (1 c/d), because in (1 c/d) the verb subcategorizes for two distinct complements (i.e., a resultative complement and an NP object) and hence illegally heads two distinct constructions (i.e., both the R- and VO- constructions). Thus the current framework makes the parochial PSC an artifact by appealing to the universal assumption of headedness in a subcategorization frame and inflectional morphology marking the heads.

6.3. Rethinking Case Theory for Chinese Word Order

There are powerful generalizations of word order by reference to head.^ For instance, the linear ordering of sisters in Japanese is heads following non-heads. Unfortunately, the

^These generalizations are seriously challenged by Dryer (1992:81), based on a study of sample of 625 languages. This is because first adjectives, demostratives, intensifiera (modifying adjectives), negative particles and tense/aspect particles do not precede their heads more often in OV languages than in VO languages. Second, for various pairs of elements (such as article and noun), there is a lack of consensus as to which is the head. 219 generalization does not work reliably for Chinese, and accounting for its word order is a long and hotly debated issue. This is because some heads precede non-heads and some follow. In particular, Chinese is basically head-final, except in the verbal (and prepositional) subcategorization frames, in which the word order is head-initial. This generalization is apparently first noted by C-T. Huang (1982), who resorts to his PSC as a solution in the GB framework. As a follow-up, Y-H. Li’s (1990) work appealing to C ase theory marks another thrust toward a GB solution. She aims to show how Case theory predicts the constituency and word order facts by her centra! proposal on the Chinese word order constraint (p.11) below.

(2) The Chinese Word Order Constraint

a. Chinese is head-final except under the requirements of Case assignment.

b. Case is assigned from left to right in Chinese.

c. A Case assigner assigns at most one Case.

Y-H. Li (1990:10) claims that (2) captures the fact that at most one NP may follow a verb or a preposition if it receives Case from this V or P. Moreover, the constraint not only captures the same facts as the PSC, but also accommodates the counterexamples to it. According to Li, (2) derives the word order facts without stipulating a long list of phrase structure rules. Thus the

Case module servos as a principled account for the deviation of Chinese word order from what

X’-theory predicts.

Aiming at descriptive and explanatory adequacy, Li deviates from the classical GB

(Chomsky 1982) in at least the following respects: a) intransitive verbs assign Case; b) subcategorized clauses need Case; and c) adjunct bare NP adverbials need Case. However, Li’s

Case account of word order still seem s far from the mark. For a critical review of Y-H. Li (1990) on empirical and theoretical grounds, I refer readers to Dai (1992b). Here I concentrate on a methodological or metatheoretical controversy about Li’s assumptions on theory-neutral grounds. 220

Outside GB circles, skepticism often arises over the underlying assumption of Case theory, as well as concerns over its potentially excessive power. Since Chinese has little or no morphological case, a resort to abstract Case in accounting for syntactic facts of the language would amount to an unwelcome move toward teleology. It is true that Case theory is not a morphological case theory; it attempts to state the distribution of NPs. As pointed out by Nichols

(1986:114-6), however, the assumption that grammatical relations are normally dependent-marked appears to underlie the conception the of Case and government approach presented by Chomsky

(1982:48). In English and other European languages, the government relation is marked on the dependent (NP) by morphological cases rather than on the head (V or P). It is conceivable that syntactic C ase theory might co-operate empirically with morphology in German, which overtly case-marks its NPs according to their different syntactic positions. With a careful demonstration, one might also be able to provide evidence for Case in English, given its impoverished case system (in the pronouns and problematic genitive case in NPs). To the surprise of linguists, dependent-marked patterns are not universal; they are possibly not even preferred type, according to the typological study of sixty lesser-known languages by Nichols. Chinese and many other languages are not so dependent-marked. Hence, accepting Case as one of the fundamental analytic notions of formal and theoretical syntax is strongly biased toward the dependent/case-marked patterns that happen to predominate in Indo-European and some well- studied non-lndo-European languages like Japanese, Korean, Finnish, Malayalam and Australian languages (;b/d:115).

As recalled, the deservative -tou is the only instance of dependent-marking patterns in

Chinese I know of analogous to morphological case marking in European languages. -Tou is an inflectional morpheme suffixing to the deverbal noun kan ‘see’ in (3a), and the morphological tou- form of kan is governed (or determined) by the head verb (mei)you ‘(not) have’ in the subcategorization frame. By an Case account, (3b) would be the d(eep)-structure for its s(urface)- structure (3a), consistent with (2a). 221

(3)a. Zhe dianying meiyou yidian kan-tou. (s-structure)

this movie not-have a-bit see -des.

This movie is not worth seeing at all.'

b. *Zhe dianying yidian kan meiyou. (d-structure)

I assume that the deverbal NP yidian kan-tou is morphologically unmarked/uninflected in d- structurs. It moves to the pcshverbal position to get Case, as forced by (2b) to escape the Case

Filter, which stipulates that an NP must have Case. As reflected in the morphology, the postverbal NP is assigned the fou-form, as in (3a).

Kan-tou (together the adjectival modifier yidian) can move from its Case position, giving

(4a). This must be a wh-movement, since the NP keeps its C ase/case {-tou), assigned postverbally, in a Case chain. Thus kan-tou is now in a preverbal non-argument position and zhe dianying ‘this movie" is still the subject. But zhe dianying can also be analyzed as a topic in different theoretical frameworks. To Li & Thompson (1976:457 & 1981:15), it is the topic because it occurs in sentence-initial position and may be followed by a phonological pause or a mood word a, as in (4b).

(4)a. Zhe dianying yidian kan-tou ye meiyou.

b. Zhe dianying (a [mood word]), yidian kan-tou ye meiyou.

By this account, yidian kan-tou can well be analyzed as the subject, forming a comment sentence with the following VP meiyou han-tou, as in (5).

(5) [s [t o p ic NP Zhe dianying] [3 [gygj ^p Y'dian kan-tou][yp ye meiyou]]]. In studing the syntactic aspects of topic and subject in Chinese, Jiang (1991 :ch.2) defines topics as preverbal NPs that have S's as sisters, and subjects as preveitial NPs that have VPs as sisters. Accordingly, zhe dianying is the topic in (5), as it has an 8-sister yidian kan-tou ye meiyou. The latter is an 8 since it can be conjoined with an overt 8, sharing the same topic, as illustrated in (6).

(6) Zhe dianying, bujin fangyingji jingchang chu maobin erqie yidian kan-tou ye meiyou.

this movie not-only film-projector often give problems but-also a-bit see -des. even not-have

‘As for this movie, not only there was often something wrong with the projector but also it was not worth seeing at all.’

What makes a C ase account for (3) suspect is that the NP subject {kan-tou) in (5) has no morphological alternations in different positions, and unexpectedly assumes the same tou-form as in the object position in (3a), a situation which is unlike the Case-to-case correspondence in

European languages (cf. the contrast in I like him vs. *M e like he in English).

One plausible solution would be to argue that the subject NP in (5) is instead a topic NP that have a sister 8, as in (7), where kan-tou as topic keeps its tou-form in a non-argument position via wh-movement.

(7) Is [t o p ic NP Zhe dianying][g [t o p ic n p Y'dian kan-tou][g ye meiyou]]].

However, the subjecthood of kan-tou in (5) cannot possibly be ruled out, since, among other points of Li & Thompson’s account, the NP posesses the most salient subject properties: it is grammatically determined by the main verb meiyou and carries selectional restrictions with it in syntactic and semantic category as well as morphological form. Jiang’s framework would also specify kan-tou as subject, as in (5), because it has ye meiyou as a VP- instead of an 8-sister. 223

According to Jiang, only S but not VP can co-occur with a preceding adverbial modifier at sentence level, such as xingkui fortunately' or dagai 'probably'. Since ye meiyou can be preceded by neither modifier, as in (8c), in contrast to the rest of the examples in (8), ye meiyou can be anaylzed as a VP,^ and thus its sister NP kan-tou is naturally subject, as in (5).

(8)a. Xingkui/Dagai, zhe dianying yidian kan-tou ye meiyou.

fortunately/probably this movie a-bit see -des. not-have

‘Fortunately/Probably, this movie is not worth seeing at all.'

b. Zhe dianying, xingkui/dagai yidian kan-tou ye meiyou.

c. *Zhe dianying, yidian kan-tou xingkui/dagai ye meiyou.

The unaltered morphological tou-form for both subject and object NPs shows that there is a neither morphological support for, nor any constraint on, a Case account of Chinese word order. Not only there are no case alternations in unmarked NPs (with no case-endings) according to their distributions, as in Wo da-le ta I hit him.' vs. Ta, wo da-le 'Him, I hit.', but it also is true for a marked NPs (with -tou), as in (3)-(8).

One last resort would be to deny any connection between syntactic Case and morphological case at all. The fact that Case in Chinese is totally abstract by lacking corresponding morphological case alternations makes a Case theory account for word order in

Chinese less constrained than in other languages. This is because the abstract notion of Case in languages with rich case morphology can be formulated in terms of the principles which are obviously necessary in any event for determining morphologcial cases, as pointed out by

Anderson (1988:175).

^Note that it would be still possible to stipulate ye meiyou as an S containing a subject gap in a theory-dependent account. My contention here, however, is that the VP account cannot be discarded, and may even be preferable for its less complexity. 224

It appears that the more abstract the Case module is in a language, the more excessive power one’s Case theory would have. Thus when Case syntax applies to Chinese, any problems encountered are easily surmountable, thanks to the lack of constraint from case morpfiology. By simply taking advantage of the absence of case morphology in Chinese, one could modify Case theory originally designed for European languages in an ad hoc way to explain away any apparent counterexamples in the language. Such a theory is powerful, but nevertheless uninteresting, since such Case theory for Chinese seems to be unfalsifiable, and full of logical circularities. This may be one of the reasons why non-GB Chinese syntacticians are more interested in discussing C-T. Huang’s PSC with resprect to word order than abstract Case, since the former is empirically falsifiable. In a neutral stand, a statement like "an object follows a verb because the object needs (abstract) Case assigned rightward from the verb which is Case assigner* is no more explanatory than "within a verbal subcategorization frame, the head/subcategorizand precedes the non-head/the subcategorized", both involving only the mechanism of description. To me, the head-final generalization and its exceptions in verbal subcategorization frames have to be stipulated and licensed in the grammar of Chinese. I conjecture that the reasons and explanations would lie in the learnability, markedness, parsing or function of language, and language change - psycholinguistics and historical linguistics, the details of which are beyond the scope of this thesis. Below I will show that the disclosure of the inflectional morphology of the language may help in seeking such explanations.

6.4. Head-Marking vs. Dependent-Marking Grammar: A Correlating Factor to Chinese Word Order

The fact that Chinese is largely head-final, except being head-initial in verbal (or prepositional) subcategorization frames is related to the previously ignored fact that Chinese is basically a morphologically head-marking, but not dependent-marking, language in the sense of

Nichols (1986:56). 225

The claim that Chinese is predominantly head-marking owes mucii to the result of the investigation of inflectional morphology in the language. Of six inflectional morphemes unearthed so far, five are suffixed to verbs and only one (-tou) to nouns - more precisely, the nouns which are derived from verbs. The five verbal inflections are the resultative DE, the measurative TA, and three aspect markers: the perfective -le, the experiential -guo and the durative -zhe. The crucial obervation is that these inflections MUST be attached to the head V rather than to the following dependent (or complement) in verbal subcategorization frames, and hence the term head- marking.

The leading idea proposed here is that such head-markings correlate with head-initial (or

V-initiai) order in the frames. For instance, the word order in chi-le putao ate grapes’ is a VO- rather than OV-order, correlating to the fact that the morphological aspect marker attaches to the

V {phi ‘eat’) but not to the O {putao ’grape ),

One may wonder whether the putative correlation is a significant generalization in grammar or simply an accidental property of Chinese. Specifically, does the head-marking pattern within subcategorizations constitute an explanation for the concurrent head-initial exceptions in Chinese? The answer is probably yes. Two important conclusions drawn from the cross-linguistic study by Nichols (1986) shed light here. First, the head-marking patterns are preferred, thus are default and unmarked, in contrast to the dependent-markings which are disfavored and marked in human languages (/b/d:116). Second, the head-marking morphology favors verb-initial word order, while dependent-marking morphology disfavors it. This appears to have functional motivation: if the verb comes first in a head-marking language, then the grammatical relations which are marked on verbs are established at the beginning, and establishing grammatical relations at the outset must be communicatively efficacious, because it streamlines the hearer’s processing {ibid:8^-2). Given that Chinese word order is head-final in general, it ought not be strange that the language employs the cross-linguistically frequent and unmarked head-marking and verb-initial patterns only in subcategorization frames as an 226

■exception'. Functionally motivated, the V-DE signals the grammatical relation for the coming resultative or descriptive complement, the V-TA for the coming quantified NP. and an aspect marker for a relatively unmarked NP or without any complements, in addition, all these suffixes clearly indicate to the listener that the syntactic category of their hosts is V or A.

Next, the burden is to characterize the verbal subcategorization frame. This is not an easy task, though here are some main points. V-DE and V-TA require appropriate complements, and therefore they are unquestionably subcategorizands. Other verbs are not so clear-cut, because most action verbs in Chinese can be transitive or intransitive. Hopefully, posb/erba! phrases and clauses all turn out to be the subcategorized constituents. Extraction or movement rules are responsible for their preverbal occurrences. Some verbs are allowed to subcategorize for more than one complement, either in ternary branching or binary branching hierarchical structures, such as the double object and the pivotal (Equi-) constructions.

Given that head-marking functionally correlates to the head-initial order in verbal subcategorization frames, does the head-final order elsewhere in the language anticipate predominant concomitant dependent-marking instances? Yes, but the claim needs qualification.

As far as I know there is no such dependent-marking by morphology per se outside the domain of verbal subcategorization. However, it is well-known that while inflecting languages use inflectional morphemes to mark grammatical relations, isolating languages employ function words

(or grammatical words, empty words, or particles) for the same effect. If the notion of head- and dependent-markings is extended from morphology to syntax, as assumed in Nichols (1986:59)

and Zwicky (1992b), Chinese clearly exhibits a predominant dependent-marking tendency outside

of verbal subcategorizations.

Perhaps the most commonly used function words are the adjectival and adverbial de’s.

These form a constituent with the preceding phrase and are cliticized to the last word of that

phrase, as in (9). This is a case of dependent-marking, since de’s attach to the modifier but not

to the following head. 227

(9)a [np (ap meili de][N huayuan]]

beautiful de garden

‘beautiful gardens’

b, Ivp [AdvP manman de] [yp du shujj

slow de read book

‘read a book slowly'

Although the much-discussed ba- or he/-NPs, i.e., PPs, are the semantic arguments of verbs, i.e., generally a ba-NP is taken as the object of the verb, and a be/-NP as the subject, they are syntactically outside the subcategorization frame VP2, since overt adjunct phrases can intervene before the verb.

(10)a. Wo Ivpi [pp [p ba](Np ta]] (henhen de) [y p g [^ da-le][ ^ p yixia]]].

I ba he (hard de) beat perf. once

I beat him once (very hard).’

b. Wo [ypi (pp [p bei][fgp ta]] (henhen de) [ypg [y da-le][^p yixia]]].

I bei he (hard de) beat perf. once

I was beaten once (very hard) by him.’

Since the grammatical relation markers ba- and bei- form a constituent with the dependent, this is another instance of dependent-marking. Preposition-marked dependents abound in Chinese.

They usually occur preverbally. Consider cong ‘from’ below. 228

(11) Wo [ypi [pp [p congK^p Beijing]] (hao-rongyi) [^pg [y qing-lai-le][Np Y'sheng]]].

1 from Beijing (not-easy) send-for perf. doctor

‘I have gone through all the trouble in sending for a doctor from Beijing.’

Like the occasional dependent marking in verbal subcategorization frames (such as -tou), sporadic head-markings are found above the subcategorization level. A case in point is the emphatic shi ‘be’, which marks the head predicate VP but not the dependent subject phrase (Ta

[SHI bu bao] ‘He is really bad’). And there are also double-markings (on both head and dependent) in the language: suiran... keshi... ‘although ... b u t...’, ruguo... jiu ... 'if... then...’, etc.

Recapitulating, the strong tendency towards morphological head-marking at the verbal subcategorization levels corresponds to the exceptional head-initial order within those domains in Chinese. By contrast, the dominant syntactic dependent-marking elsewhere accompanies the general head-final order in the language. Deviating though it appears to be in Chinese, such a

V-initial order is consistent with the unmarked cross-linguistic tendency of that type of order. The functional motivation to build grammatical relations at the outset of a verbal subcategorization frame thus accounts for the V-initial order in that frame. Note that here I re-interpret Nichol s correlations between marking patterns and word order across languages as plausibly applying across syntactic constructions within a single language. Whether such plausibility holds outside

Chinese remains an empirical question.

While synchronically the morphological head-markings and the V-initial order are mutual correlates, diachronically, the V-initial order seems to be the causation. This is because it is possible to reconstruct a verbal status for DE and aspect markers in the history of Chinese and to show that they were once the head of the complements of the matrix VP. A diachronic morphologization called headward migration (Nichols 1986:84) is responsible here, whereby some adposition-like verbs change from the head of a subordinate phrase to the morphological marker of the matrix head, via cliticization (cf. to in want to go becom es wanna go in English). The sam e 229 point can be made for the historical nominal status of TA. The headward migration process can still be observed in Modem Chinese. The postverbal gei ‘give, to ’, dao ‘arrive, to’ and zai ‘at’ have been claimed to become part of the preceding matrix verb by m eans of cliticization, compounding or suffixation by different authors (e.g., Chao 1968, Tai 1989b, Y-H. Li 1990, Huang & Mo 1992).

(12)a Ta [song gei] wo yi ben shu.

he send give me a M book

‘He gave a book to me.

b Ta [lai dao] wo jia.

he come to my home

‘He came to my home.’

c Ta [tiao zai] dishang,

he jump at ground-on

‘He jumped to the ground.’

The main evidence provided by these authors for lexicalization is that no element, including aspect markers, can intervene in [V + V/P], and that aspect markers can only attach to gei, dao or zai. In addition, as I point out in Dai (1992b), there seems to be no question that [V + V/P] forms a prosodic unit, and thus at least som e sort of phonological cliticization of V/P to V is involved (especially when the result is bisyllabic).

However, while these pieces of evidence are consistent with the headward migration leading to head-marking patterns, there is evidence that the morphologization is not completed yet as DE is in Modern Chinese. If the bracketed material in (12) were syntactic words, then they would be the minimal constituents that syntactic rules may refer to. While no convincing evidence 230 can be found that syntax does refer to these "words', there exists some evidence against the constituency of the brackets; the second element in the brackets and its following NP can move around like a single constituent.

(12)'a. Ta song yi ben shu gei wo.

he send a M twok give me

‘He gave a book to me.

b. Ta dao wo jia lai.

he to my home come

‘He came to my home.’

c. Ta zai dishang tiao.

he at ground-on jump

‘He is jumping on the ground.’

Note that (12a) and (12’a) share the same truth-conditions. The examples (12b) and (12’b) are sightly different in semantics, with (12b) but not (12’b) entailing the completion of the action.

Thus the two sentences in each pair are related syntactically by, say, transformational movement, and the bracketed elements in (12a) and (12b) cannot be a single constituent and a word. Only

(12c) and (12’c) are different in meaning and can be claimed to be unrelated in syntax. The examples in (12a), (12’a), (12b) and (12’b) are strong counter-evidence to the various synchronic lexicalization accounts, although the non-intervention constraint, the impossible attachment of aspect markers to the first element,^ and the prosodic encliticization signify a historical

■^Neither of these symptoms is absolute, though. As pointed out by Tai (1989b), lu ‘road’ may intervene between [zou dao] (zher) [walk to], giving zou lu dao zher 'walk to here', and so may zhe ben shu ‘this book’ be inserted between [fang zai] l^huo-shang), as in fang zhe ben shu zai 231 morphologization in progress going toward the head-marking morphology. If these processes are completed in the future, then the results would be clear cases of V-initial order plus morphological head-marking in verbal subcategorizations.

Similar phenom ena are observed in -p ‘not’ in reduced A-not-A questions (x/ bu xihuan

-> xip xihuan ‘like or not?') and in the second verb in the serial verb construction {qu pu chuang

-> qup chuang ‘go make bed’, cf. 2.3.2), wheœ the internal sandhi FE optionally applies across two syntactic words which form no constituent, and thus makes them one phonological word.

The results would easily give rise to a reanalysis of head (,’r/ and qu) and head-marking (-p) in these constructions.

In conclusion, the correlation between Chinese word order and head-marking vs. dependent-marking grammar with its functional motivation opens up a whole new channel of explanation for the head-initial word order exception in Chinese. Such a program stands as a promising competitor to the account appealing to abstract Case, a standard theoretical term stemming from what Nichols (1986:116) calls Eurocentric distortion.

6.5. An Inflectional Analysis of the Reduced A-not-A Questions^

This section addresses the relationship between the Chinese A-not-A question in (13) below and its corresponding "reduced" forms in (14) and (15). As seen in (13), the main VP is composed of a VP followed by its negative counterpart, usually with the negator bu.®

(13) Ta xihuan zhe ben shu bu xihuan zhe ben shu?

he like this M(easure) book not like this M book

zhuo-shang ‘put this book on the table’. In addition, no aspect markers can attach to zai in (12c).

®This section is largely adapted from Dai (1991c).

®lf a disjunction hai(shi) ‘or’ occurs before bu, the question form is called the "alternative question" or "disjunctive question". Its semantics are basically the same as its A-not-A counterpart, if the two disjuncts are in polarity of affirmative vs. negative. 232

‘Does he like this book or not?’

(14) Ta xihuan bu xihuan zhe ben shu?

(15) Ta xi bu xihuan zhe ben shu?

In particular, I intend to provide an inflectional analysis for (13) that improves upon previously proposed analyses.

In the classical approach, (14) and (15) are related to (13) by a transformational deletion

(e.g., Wang 1967), which, however, violates the LIH, since in (13) the first instance of -huan is part of a word xihuan and yet is syntactically deleted, as observed by C-T. Huang (1989:147). To preserve the UH, he adopts a phonological approach, submitting that (13)-(15) are all derived from a simplex d-structure by the insertion of bu after a phonological reduplication of some elements following (-HO). A morphological reduplication is employed by Dai (1989) and Libucha

(1990) to derive an A-not-A word xi-bu-xihuan in (11) from xihuan. Below, I first critically review the previous reduplication analyses and then propose a processual morphological operation of subtraction to derive xi from xihuan in inflectional morphology and explore its theoretical consequences.

6.5.1. Against Phonological Reduplication

C-T. Huang (1989:153) tries to solve the LIH violation problem in the transformational approach by proposing that the A-not-A questions in (13)-(15) are derived by a phonological process triggered by [-t-Q] dominated by an INFL node, and that the A-not-A part is derived from a simplex d-structure with an interrogative INFL constituent as in (16). 233

(16) [,n f l ’[in fl I+Q]][vp xihuan zhe ben shu]]]

The abstract INFL with [+Q] presumably copies a sequence immediately following INFL and inserts the morpheme bu not' between the original and its copy. Owing to the flexibility of the size of the reduplicated materials, the results may be xihuan zhe ben shu bu xihuan zhe ben shu as in (13), xihuan bu xihuan zhe ben shu as in (14), or xi bu xihuan zhe ben shu a s in (15).

Example (15) is generated, since phonological rules, according to Huang, are immune from morphological (the LIH) and syntactic structures.

However, Huang’s proposal first suffers from descriptive inadequacy (Dai 1 S90b:305). For instance, it predicts the ungrammatical utterance *Ta xihuan zhe bu xihuan zhe ben shu to be grammatical, since nothing can prevent the phonological reduplication from ending with the second word after the INFL. Libucha (1990) lists additional problems of Huang’s analysis. In particular, there are underlying VPs which cannot undergo the A-not-A phonological reduplication

(e.g., VP headed with bidei must’ in *bi bu bidei qu ‘must go or not?’), nor can the VP with an aspect marker (cf. piping-le ‘criticized’ in *pi bu piping-le ni criticize you or not?’); there are elements "reduplicated" which do not seem to have d-structures (e.g., le in Ta piping-le ni mei piping ni Did he criticize you or not?’ has no copy in the second conjunct, and the d-structure

*Ta you piping ni for Ta you meiyou piping ni ‘Did he criticize you or not?' is unacceptable in many dialects, including the standard dialect.); and for the potential resultative compounds like zou-DE- dao ‘able to walk to arrive’, an ungrammatical A-not-A *zuo-DE-dao bu zuo-DE-dao able to walk to arrive (somewhere) or not?’ will be generated.

Huang’s analysis also suffers on metatheoretical grounds, since he employs the same rule to generate information questions (i.e., wh-questions) in the language, which makes his theory have potentially excessive power, According to Huang, the question word weisheme ‘why’ in the information question in (17) is also phonologically realized by the [-f Q] in the same way as the

A-not-A question in (13). 234

(17) Ta weisheme xihuan zhe ben shu?

he why like this M book

‘Why does he like this book?’

The theory is now in fact powerful enough to generate all types of Chinese interrogatives. One might expect [+Q] to induce haishi ‘or’ for an alternative question, if it occurs with a disjunction; to induce the question marker ma with a proper intonation for a yes-no question, if it occurs with a sentential particle. This proves not to be a virtue of Huang’s analysis, though, for Huang himself maintains typological dissimilarity among various types of Chinese interrogatives. In addition, it will be later shown that A-not-A questions and alternative questions, for example, have different distributions.

Summarizing, Huang’s analysis is untenable for descriptive and metatheortical considerations.

6.5.2. Against Morphological Reduplication

Approaches in morphology to A-not-A questions have been proposed (e.g., Dai 1989,

Libucha 1990). Dai (1989) formalizes an A-not-A Word Formation Rule (WFR) in derivational morphology, i.e., the A-not-A part in a question like xihuan bu xihuan and xi bu xihuan in (14) and

(15) is derived by the A-not-A WFR. There are two morphological operations involved in the WFR: the préfixai reduplication of the first syllable or the whole word and the infixation of bu ‘not’. For example, the first syllable of the word xihuan is reduplicated to give xi-xihuan, followed by the infixation of bu to form xi-bu-xihuan; in xihuan-bu-xihuan, it appears to be the whole word that is reduplicated, as below. 235

(18) A-not-A Word Formation Rule (WFR)

Reduplication Pu-infixation

a xihuan > xi xihuan -> xi bu xihuan (zhe ben shu)

b xihuan -> xihuan xihuan -> xihuan bu xihuan (zhe ben shu)

By this approach, the LIH is maintained, since the rule applies presumably in the lexicon. But it seems that saving the LIH is the sole motivation for the WFR, and there are reasons to doubt that such a WFR exists in the language.

First, reduplication elsewhere in Chinese is constrained at the stem-level, not at the sub­ stem level (cf. 4.1), i.e., the whole word composed of XY, rather than part of its components X or

Y, must be reduplicated, no matter whether the result is XXYY or XYXY, e.g., piaoliang ‘pretty’ -> piaopiaoliangliang, cf. *piaopiaoliang, *piaoliangliang; xuexi ‘learn’ -> xuexixuexi, of. *xuexuexi,

*xuexixi. The constraint is a rule of thumb to test wordhood of the VO- and R-structures.

However, the reduplication in (18) may operate at the sub-stem level, i.e., in (15) and (18a), only the first syllable of xihuan is copied. This violation of the reduplication constraint unfortunately implies the presence of a non-existent *xi-xihuan (at least at certain level of derivation) in the language.

The reduplication is not only unconstrained for (15) and (18a), but also redundant for (14) and (18b), where the whole word xihuan is presumably copied. This is because the only thing that needs explaining in the syntax of (14) as opposed to (13) is the missing object NP after the first x;7)uan in (14), which does not violate the LIH. Since missing objects in appropriate contexts are well motivated in the syntax of Chinese, e.g., the second object NP zhe ben shu in (13) can also be missing, I conclude that (14) and (18b) can be licensed by an independent syntactic rule, and need no morphological reduplication. 236

Besides, it is possible to add in more syntactic words (W) to the result of the WFR. For many speakers, the disjunctive hai or emphatic shi can occur between the original and its copy, in violation of the LIH.

(19)a. Ni xi hai bu xihuan zhe ben shu?

you like or not like this M book

‘Do you like this book or not?'

b. Ni xi shi bu xihuan zhe ben shu?

you like be not like this M book

‘Tell me, do you like this book or not?’

Thus by recognizing xi bu xihuan as a word, one would have to recognize xi hai/shi bu xihuan as a word as well. Now only an extended WFR could cover (19) and avoid the violation of the UN, which strikes me as a most implausible move. The examples in (19) also suggest that xi might be a syntactic word, free" only in certain marked syntactic constructions, for it passes the expansion test (to be discussed).

Finally, the loss of a significant generalization seem s unavoidable with the WFR, since (15) via the WFR would have a very different syntactic structure from that of (13), indicated in (20a) and (20b) respectively. Example (14) is assum ed to correspond to (13), with a missing object NP of the first verb. 237

(20)a VP

/ \

VP VP [+neg]

/ \ / \

V NP Adv VP

I I I / \

I I I V NP

I I I I !

xihuan zhe-ben-shu bu xihuan zhe-ben-shu (13)

xihuan bu xihuan zhe-ben-shu (14)

b. VP

/ \

V NP

I I xi-bu-xihuan zhe-ben-shu (15)

The examples in (20a) contain a coordinate structure. In addition to the obvious parallel structures of xihuan zhe ben shu and bu xihuan zhe ben shu, (20a) is subject to the Coordinate

Structure Constraint (Ross 1967) as in (21a-c). As expected, across-the-board extraction is allowed as in (21 d) (Each conjunct is put in a brace and the gap is marked with e.).

(21)a. Ta [{xihuan zhe ben shu} {bu xihuan zhe ben shu}]?

he like this measure book not like his measure book

Does he like this book or not?' 238

b. *Zhe ben shu, ta [{xihuan e} {bu xihuan zhe ben shu}]?

c. *Zhe ben shu, ta [{xihuan zhe ben shu} {bu xihuan e}]?

d. Zhe ben shu, ta [{xihuan e} {bu xihuan e}]?

Example (21 d) can also be derived from (14) in (20a), presumably because the object of the first xihuan is also missing, consistent with the across-the-board phenomenon. Nevertheless, no direct evidence shows that (20b) is also a coordination. Instead, it looks like a frame in which the verb

(via the WFR) subcategorizes for an NP. Such a big structural difference implies that similar syntactic, semantic and pragmatic behaviors exhibited by the members of A-not-A questions would be treated as accidental. This is very unlikely, since the similarities are too striking to be accidental, for the shared properties serve as features for distinguishing A-not-A questions from the other types of questions in Chinese.

Semantically, no one would deny that (13)-(15) are questions about a choice between an affirmative and its negative counterpart. Pragmatically, an A-not-A questioner performs a speech act of "hurrying tone" (Tone here is not a technical term in phonetics.), i.e., the question is perceived as if the questioner is in a hurry to urge the listener to give a response. The alternative questions (with haishi 'or') do not have such a hurrying tone (Lü 1985). This pragmatic property is shared by all of the A-not-A questions.

The most interesting behavior common to the A-not-A questions is their syntactic distribution. C-T. Huang (1989:149) claims that there is a systematic distributional difference between the A-not-A and alternative questions in terms of island constraints (Ross 1967).

Specifically, an A-not-A question cannot occur as a sentential subject, as in (22a-c), nor in a relative clause, as in (22d-f), with the islands marked with square brackets. But an alternative question can occur in these islands - simply by inserting a disjunction haishi before bu, the grammaticality comes back. 239

(22)a *[Wo xihuan zhe ben shu bu xihuan zhe ben shu] bijiao hao?

I like this M book not like this M book comparatively good

‘Is it better for me to like this book or not?’

b. *[Wo xihuan bu xihuan zhe ben shu] bijiao hao?

c. *[Wo xi bu xihuan zhe ben shu] bijiao hao?

d. *Ni renshi [xihuan zhe ben shu bu xihuan zhe ben shu] de ren?

you know like this M book not like this M book DE person

'Do you know the person who likes this book or the person who doesn’t?’

e. *Ni renshi [xihuan bu xihuan zhe ben shu] de ren?

f. *Ni renshi [xi bu xihuan zhe ben shu] de ren?

It remains unexplained why none of the versions of the A-not-A questions, long or short, can occur in the islands. However, Dai (1990b:302) points out that A-not-A questions, contra Huang’s claim, may be embedded in certain sentential subjects, as in (23a-c), and in certain relative clauses, as in (23d-f) {ne is a marker for non-yes-no questions.). Here, it is also unexplained why all of the A-not-A questions, not just some of them, are allowed in these islands.

(23)a. [Wo xihuan zhe ben shu bu xihuan zhe ben shu] hao ne?

I like this M book not like this M book good Q-ne

‘Is it better for me to like this book or not?’

b. [Wo xihuan bu xihuan zhe ben shu] hao ne?

c. [Wo xi bu xihuan zhe ben shu] hao ne? 240

d. Ni renshi [xihuan zhe ben shu bu xihuan zhe ben shu] de ren ne?

you know like this M book not like this M book DE person Q

‘Do you know the person who likes this book or the person who doesn’t?’

e. Ni renshi (xihuan bu xihuan zhe ben shu] de ren ne?

f. Ni renshi [xi bu xihuan zhe ben shu] de ren ne?

Another distributional difference between A-not-A and alternative questions is seen in the following examples;

(24)a. Ni ying(gai) bu yinggai xuexi shuxue?

you should not should study math

'Should you study math or not?’

b. *Ni yinggai xuexi shuxue bu xuexi shuxue?

*Ni yinggai xuexi bu xuexi shuxue?

*Ni yinggai xue bu xuexi shuxue?

you should study math not study math

Roughly speaking, the A-not-A part must be located at the left edge of certain phrases (/b/d:309).

Thus in (24a), the A-not-A may involve yinggai, the left node immediately dominated by the predicate VP, but must not involve the right node xuexi shuxue, as in (24b). However, the positions of haishi phrases in alternative questions are not subject to such a left edge condition, as demonstrated in (25). They can occur on either side of the VPs. Compare their grammaticality status with that in (24). Still, the question to be answered is why all of the A-not-A questions, and 241 not just some of them, are subject to the left edge condition.

(25)a. Ni yinggai haishi bu yinggai xuexi shuxue?

b. Ni yinggai xuexi shuxue haishi bu xuexi shuxue?

c. Ni yinggai xuexi haishi bu xuexi shuxue?

‘Should you study math or not?’

Summing up, unlike other reduplications in the language, the reduplication in the WFR in (19) copies sub-stem material. Further, it is possible to insert Ws into the “A-not-A word", a violation of the LIH. More crucially, the treatment entails a large structural distinction in the family of A-not-A questions, and hence falsely implies that the syntactic, semantic, and pragmatic properties strikingly shared among the members of the family are totally accidental. Therefore, the A-not-A WFR which generates the question word in (19) should be abandoned, and a solution which has better motivation and reduces the structural difference between (20a) and (20b) is desirable.

6.5.3. An Inflectional Analysis

Any adequate account of A-not-A questions should keep the LIH in mind while capturing

a generalization among (13)-(14). I assum e that they involve a family of three constructions. The

differences among them should be kept to a minimum, or at least not so great as the one

between (20a) and (20b), as it will be shown that some of the syntactic requirements in the

grammar of Chinese are set for all three constructions.

As the "reduplication" approaches do not work well, I will adopt a "deletion" approach.

The deletion however does not operate in syntax as assumed by the transformationalists but in

inflectional morphology. 242

There is no a priori reason to expect that affixation is the only type of operation in morphology, though it is predominant. Since the morphology of natural languages does employ, in addition to affixation, operations like alternation of vowels or consonants, interdigitation, reduplication, metathesis and subtraction, I adopt a processual morphology rather than a strict concatenation, following the lead of Janda (1983), Hoeksema & Janda (1988), Anderson (1988) and Zwicky (1988).

Xi is one more inflected form of the stem of the lexeme xihuan, which I name AA-form.

The formal operation of subtraction derives xi by deleting the second syllable -huan in xihuan in inflectional morphology. More precisely, the lexeme in question is truncated to its first syllable.^

Subtraction as a type of morphological operation has been proposed to analyze other languages, including English Latinate adjectives (Aronoff 1976), the German diminutive (Lessen Kloeke 1985), the Danish imperative (Anderson 1975), Dutch toponymie adjectives (Hoeksema & Janda 1988), and Icelandic deverbal action nouns (Oresnik & Petursson 1977, Kiparsky 1984).

I start with the Tull" form in (13), and assum e that it takes the coordinate structure in (21 a).

In syntax, the matrix VP requires that a) the second VP must be in its negative form; b) the lexemes of the head verbs in both conjuncts are identical; and c) the (object) complements of both verbs are identical.

I assum e that (14) has basically the sam e structure as (13) and (20a), the only difference being a missing object in the first conjunct in the former but not in the latter. The object can be missing in the second conjunct also, yielding (26a). Moreover, the second conjunct can be

^This is because a monomorphemic polysyllabic (borrowing) word is subtracted to its first syllable. i. Ta xiesidili bu xiesidili? *Ta xiesidi bu xiesidili? *Ta xiesi bu xiesidili? Ta xie bu xiesidili?

he hysteria not hysteria ‘Is he hysterical?' 243 missing provided that bu remains, as in (26b). If the context is clear, the first object may also be null, as shown in (26c).

(26)a Ta xihuan zhe ben shu bu xihuan?

b. Ta xihuan zhe ben shu bu?

c. Ta xihuan bu?

This shows that ellipsis in conjuncts is independently motivated in the language, accounting for

(14). An interpretation rule is presumably responsible for the meaning of the missing constituents in (14), (20a) and (26).

As argued before, (15) cannot have the structure in (20b). Nor can it have the same structure in (20a) as (13), (14) and (26). Otherwise it would predict that the ellipsis works within word in violation of the LIH. Worse still, the structure in (20a) cannot explain why ellipsis cannot delete huan in the second conjunct, as in (27), in contrast to (15), where huan is missing in the first conjunct.

(27) *Ta xihuan zhe ben shu bu xi?

I propose instead a distinct structure for (13) as follows. 244

(28) VP{a}

/ \

VP[AA] VP[+neg]

I / \

V[#,AA] Adv VP

I 1 / \

I I V[#] NP

I I I I

xi bu xihuan zhe-ben-shu (13)

like not like this M book

My proposal supposes that x/ is a W/w in (15) and (28). The evidence for the inflectional status of x/ is its syntactic relevance, as the AA-form is required by the whole construction. The matrix VP sets the following requirements, represented as {a} in (28), on its two conjunct VPs; a) the head verb of the first VP must take the AA-form; b) the AA-form takes no complements, i.e., the object of a transitive verb is obligatorily missing;® c) the second VP must be in its negative form and its head verb must be in the full form and not in the AA-form; and d) the lexeme of the two head verbs is identical, marked as the lexeme number [#] in (28).®

In general, the existence of the construction in (28) with its AA-form is predictable from the corresponding construction in (14) with the full form of the stem of the verb. Any bi- or poly-

®Note that there is no NP trace or gap proposed after xi in (28), for the AA-form of an intransitive verb (adjective) may also occur in this position, giving

i. Ta piao(liang) bu piaoliang? she pretty not pretty ‘Is she pretty or not?’

®The VP{a} expansion rule in (28), together with the one for the matrix VP in (13), (14) and (20a), needs to mention the requirements which are two or three layers’ down from the matrix VP{a} (e.g., the identical lexeme number [#]), which supports No’s (1990) proposal on Depth-r? Grammars. 245 syllabic verb ocurring in the first conjunct with a missing complement in (14) - not necessarily an object, as Chinese has intransitive verbs and non-object complements - is expected to have the corresponding "reduced" construction in (15) and (28) with the AA-form of the verb in the first conjunct.

One might point out that since the full-fledged xihuan can apparently replace xi in first VP conjunct in (28), the proposed inflectional formx/ is not syntactically relevant. However, there are several differences between xi and xihuan in that position, indicating that two distinct constructions are involved. First, when xihuan is in place, the absence of the second conjunct in (28) still gives a grammatical sentence. In contrast, when xi is in that position, the second VP is obligatory, as in (29). Second, xihuan may also subcategorize for an appropriate object, but xi may not, as shown in (30). Its semantic object is obligatorily missing. Third, a bare bu can follow x/Vtuan but not xi, as in (31). Finally, xihuan does not have to occur with its negative counterpart, whereas xi does have to, as shown by (32).

(29)a, Ta xihuan. b. *Ta xi.

(30)a. Ta xihuan zhe ben shu bu xihuan zhe ben shu?

b. *Ta xi zhe ben shu bu xihuan zhe ben shu?

(31 )a. Ta xihuan bu? b *Ta xi bu?

(32)a Ta xihuan (haishi) taoxian Zhangsan?

b. *Ta xi (hai) taoxian Zhangsan?

he like (or) hate Zhangsan

‘Does he like or hate Zhangsan?’ 246

I conclude that only the AA-form can occur in the structure of (26), and the AA-form is taken as an inflection because of its syntactic relevance.

6.5.4. Consequences

My inflectional analysis entails that the distribution of the AA-form is restricted to certain syntactic construction(s), resulting in the apparent boundedness of the AA-form. This is due to the assumption concerning the syntactic relevance of inflectional morphology. The b-sentences in (29)-(32) are bad because Chinese syntax does not allow the .A.A-form in those constructions, just as standard English syntax rules out the [PRP] /ngr-form in *He going, but licenses He is going.

This analysis accounts for the relative productivity of the AA-form. Inflections are morphologically productive because the syntactic rules they have access to are structurally general, i.e., the syntax (associated with compositional semantics) may enforce the generality of certain features to be realized as morphological inflectional forms (rather than the generality of particular morphological rules). For reason of economy (one form, one meaning), most inflections ought to be productive, attaching to the whole word class with few exceptions. Thus in addition to the [+V] categories, certain prepositions and adverbs also have the AA-form, for the A-not-A is a productive syntactic process in Chinese. By contrast, derivational morphology may not be

productive, since the syntax has no access to enforcement of generality of derivations, and

moreover, the lexical function of derivation allows more lexical exceptions than inflection does.

Thus the productivity of the AA-form in the derivational WFR in (18) would be just as accidental

as English derivational noun suffix -ness for adjectives in happiness. Note that the AA-form only

applies to all those stems with more than one syllable. So the lexical exceptions of the inflection

are predictable from the number of syllables of the stem, just like the English inflectional

comparative -er and superlative -est, which are basically restricted to monosyllabic and some

bisyilabic words. 247

The reduplication approach in Libucha (1990) assum es that (18a) is an inflectional rule.

However, (18a) as an inflectional rule lacks syntactic motivation, for nowhere is the 'inflected A- not-A word' from (18a) exclusively required or prohibited by the syntax of Chinese, since, as mentioned before, its syntactic distribution is the sam e as those of other A-not-A phrases in (13) and (14). Such an inflectional stipulation ignores the theoretical consideration of the syntactic markedness of inflections, relative to derivations, because, other things being equal, inflections must be syntactically relevant.

The constructional similarities between (15) and (13)/(14) do not come as a surprise as they do if (15) assum es the structure in (20b), for the structural difference between (20a) and (28) is much smaller than between (20a) and (20b). The extra syntactic requirements for (28) relative to (20a) dem and the AA-form for the first VP and its obligatory missing complement. However, the difference is minimized as the inflectional morphology relates the forms between xi and xihuan of the same lexeme and the interpretation rule takes responsibility for the meaning of the missing complement.

The previous reduplication accounts are claimed by their proponents as motivated in part

by the unexpected preposition stranding in the first but not in the second conjunct if some

monosyllabic prepositions are A-not-A questioned (e.g., cong bu congzher 'from not from here ),

by assuming that Chinese does not tolerate any preposition stranding. Thus cong bu cong (zher)

from not from (here)' would be a word, reduplicated from preposition cong ‘from’ and involving

no preposition stranding at all. Before dealing with the stranded prepositions, however, research

should be done on the definition of preposition vs. verb, and on the extent to which stranded

prepositions are allowed elsewhere in the language. On the one hand, there is controversy in the

literature concerning the definition of preposition in Chinese (Li & Thompson 1981:ch.9 and

references therein); on the other hand, the ban on preposition stranding is not absolute. At least

one preposition, the passive marker bei ‘by’, is exempted in certain cases, as below. 248

(33) a Wo bei Lis! da le.

I by Lisi beat perf.

‘I was hit by Lisi."

b. *Lisi, Wo bei da le.

c. Wo bei da le.

I by beat perf.

‘I was hit.’

In a transformational account, one might want to say that preposition stranding is strictly banned in movement rules, as in the topicalization in (33b), but allowed in deletions, as in (33c).

Assuming the ban on preposition stranding is not absolute in the language, the current account allows for a missing object for the first cong in cong bu cong (zher). The missing object is licensed by the same rule as the rule licensing (14), in which the object is missing after the first verb. Thus the category for the AA-form is generalized to [-N], including verbs, prepositions and adjectives, if one assumes the adjectives are stative verbs in Chinese.

Libucha (1990) claims that (34) and (35) are very strong evidence for a reduplication account, since *you xihuan ni^° underlying the first conjunct in (34) and *bu zou-de-dao Beiji underlying the second conjunct in (35) do not exist.

(34) Ta you meiyou piping ni?

he have haven't criticize you

‘Has he criticized you or not?’

^°lt is ungrammatical in certain dialects, including the standard dialect. 249

(35) Ta zou bu 2 0 u-de-dao Beiji?^^

he walk cannot-walk-arrive North Pole

‘Can he walk to reach North Pole?’

However, the putative reduplication is not necessarily the only account.

Î take meiyou (and its variant mei) in (34) first, as they are negators occurring in sam e position as bu in A-not-A questions. Note that meiyou has no affirmative *you and is the negation of the aspectually perfective -!e, and that both bu and meiyou are incompatible with -le in a local

VP (Wang 1965, also see a summary in C-T. Huang 1988:282), e.g., *bu/meiyou piping-le ni not criticize you’ vs. bu/mei piping ni ’not criticize you’ vs. piping-le ni ‘criticized you’. The present account predicts the AA-form and the related forms with or without -/e.’^

(36)a. Ta pi mei piping ni?

he criticize haven’t criticize you

‘Has he criticized you or not?’

b. Ta piping-le mei piping ni?

c. Ta piping-le ni mei piping ni?

d. Ta piping meiyou piping ni?

e. Ta piping ni meiyou piping ni?

’^The acceptability of this pattern varies across speakers and dialects. Lü (1985) considers it grammatical, but Chao (1968:159) and Packard (1990:30) do not. Here I assume its acceptance and proceed to develop an account shortly.

^^The choice between mei and meiyou is determined by the bisyilabic rhythm rule (Dai 1990b:297-300), which is not the concern here. 250

The existence of (34) as an alternative expression of (36) can be attributed to the properties of the aspectual lexeme meiyou. There are certainly diachronic explanations, as the etymologically related homophorous possessive me/you and deservative me/you (cf. 5.5) may form

A-not-A questions in the way (34) does.

(37)a. Ta you meiyou shu?

he have not-have book

'Dees he have books or not?'

b. Zhe dianying you meiyou kan-tou?

this movie worth not-worth see des.

'Is this movie worth seeing or not?'

The synchronic explanations, on the other hand, lie in the lexical idiosyncrasy of the perfective meiyou, which lacks its affirmative *you but additionally allows for you meiyou in an A-not-A question, as in (34). This seems not to be an ad hoc assumption, if the perfective meiyou is treated as an auxiliary verb (Wang 1965). As pointed out by Libucha (1990), at least one (other) auxiliary bidei 'must' idiosyncratically lacks any form of (negation and) A-not-A question, and so does the other auxiliary de 'must' for some speakers. If auxiliaries can have such idiosyncratic distributions with respect to negation and A-not-A questions, it ought not be strange that meiyou does not have an affirmative form *you but tolerates you meiyou in an A-not-A question, which has strong etymological and analogical motivations.

With the you meiyou case accounted for, I turn to the potential resultative compounds in

(35). Recall that such compounds are formed by "infixation" of DE or its negation bu i^ou-DE-dao

'able to walk to arrive at’, zou-bu-dao 'unable to walk to arrive at’, cf. 3.4) from their corresponding

regular resultative compounds (zou-dao 'walk to arrive at ). The current framework apparently 251 cannot predict the ungrammaticaiity of the AA-form in (38c), though it can account for the full forms in (38a/b). Worse yet, it cannot license the existence of (38d/35), since *bu zou-de-dao

Beiji is presumably not well-formed elsewhere.

(38)a. Ta zou-DE-dao Beiji zou-bu-dao Beiji?

he walk-able-walk-arrive North Pole walk-unable-arrive North Pole

Can he walk to reach the North Pole?’

b. Ta zou-DE-dao zou-bu-dao Beiji?

c. *Ta zou zou-bu-dao Beiji?

d. Ta zou bu zou-DE-dao Beiji?

However, there may be principled explanations. Example (38c) can be ruled out by stipulating that the negator of the negative form of the A-not-A question containing the AA-form must be at

phrasal level, as illustrated in (28). This enforces phrase initial position for an adverbial negator

in the second conjunct. Such an initial position seems to have phonetic motivation for avoiding

two consecutive identical syllables or words (zuo zuo) disfavored in Chinese. The language

accommodates a repetition by haplology (Chao 1968:247), as in (39a), although I note that the

haplology is not obligatory, as in (39b/c).

(39)a. *Ta yijing zou-le le. -> Ta yijing zou(-)le.

he already walk perf. prt.

‘He has already gone.’

b. waijiao-bu-buzhang -> waijiao-buzhang

foreign ministry minister 252

‘Foreign Minister’

c. Wo han ta ta ye bu ting. -> Wo ban ta ye bu ting.

I shout he he still not hear

‘Even though I shouted to him loudly, he didn’t listen."

In the other reduced A-not-A questions, bu or mei(you) separates the two otherwise consecutive identical syllables. But this is not true for the potential resultatives, which requires that bu be suffixed to the first components, let alone the fact that the potential -bu- is not inserted at phrasal level as required, but instead at the morphological compounding level, hence the ungrammaticaiity of (38c). Since the AA-form (zou for zou-DE-dao) is the most economical way of expressing A-not-A questions, the simplest way of saving (38c) would be the readjustment of the positions of the second zou and bu there, giving ?*Ta zou bu zou-dao Beiji?. DE is inserted, presumbly to avoid either the ambiguity between the potential and regular resultative readings or to restore grammaticality, as in (38d). Finally, whether or not *bu zou-DE-dao Beiji exists in the

language remains an empirical question. It can be embedded as a sentential subject, as in (40a),

on a par with the negated V1 in the resultative construction (C-T. Huang 1988:289, due to Y-H.

Li), although speakers may prefer (40b/c) to convey the same meaning. In this way, (38d) can

even be licensed in same way as other reduced AA-forms.

(40)a. Bu zou-DE-dao Beiji bu suan haohan.

not walk-able-arrive North Pole not be hero

‘One is not a hero if unable to walk to the North Pole.’

b. Bu zou-dao Beiji bu suan haohan.

c. Zou-bu-dao Beiji bu suan haohan. 253

Although the explanations above do not seem as neat as the reduplication accounts, counter-balancing to the apparent complication of my account is the fact that the AA-form exists in a few related constructions. These cases would be mostly difficult to motivate in a reduplication account, because the full-fledged syntactic words dao, shi, dou, hai, etc. would have to be

‘infixed*.

(41 )a xi dao/shi xihuan, jiu shi tai xiao le. (concession)

like be like but be too small LE

‘(I) like it, but it’s too small.’

b. xi dou/hai xihuan bu-guo-lai, hai da ta? (tag)

like all like not-come how beat her

‘(I) love her very much, how could I possibly beat her?’

By the present account, the AA-form as an operation is simply invoked by two constructions other than the A-not-A question requiring the inflection.

In conclusion, the morphosyntactic analysis resorting to a subtraction operation in inflectional morphology has been shown to be advantageous when compared to those employing reduplications plus bu-infixation. In addition to the cross-linguistic and language-specific motivations of this proposal, it yields a satifying explanation for the constructional similarities among the A-not-A questions. More importantly, since the deletion operates in the morphological component. It is therefore possible to construct a viable account of this much-disputed construction in Chinese within a highly constrained theory of grammar in which the UH is maintained. 254

6.6. The Derivation-1 nflection Distinction as Obscured by Lexical Morphology^ ^

The theoretical framework of Lexical Morphology (and Phonology) is often referred to as level-ordered morphology (and phonology) (Kiparsky 1982, Mohanan 1986). It claims that there exists a direct link between the morphology and the phonology, and that certain properties of morphological configurations follow from the assignment of levels (also called cycles, layers or strata), i.e., how closely affixes and their stems are related in linear distance, phonological interaction, and to some extent semantic interaction. For instance, one of the morphology- phonology covariations observed is that the farther an affix is away from the most embedded stem, the less active the phonological interactions are between the affix and the stem in terms of changes in segmental and suprasegmental features. Thus -ity and -s in English regular-iti-es are assigned at different levels with respect to the stem regular, because -ity is closer to the stem and causes the vowel and stress alternations in the stem, whereas the relatively distant ^ does not.

While readers are referred to Gussm ann’s (1988:232-9) review and other literature on the theory of Lexical Morphology, a major issue is whether such interactions and covariations should be accounted for by putative language-specific assignments of morphological levels, or made to follow from some fundamental theoretical assumptions concerning the organization of universal grammar (Zwicky 1992a:359). In this section, I will show, using Chinese, that a certain amount of covariation is the result from the distinction between derivational and inflectional morphology and that such a distinction cannot be adequately captured by the last level stipulated in Lexical

Morphology. After introducing Packard’s (1990) 4-level model and locating relevant issues in

6.6.1, I demonstrate that Packard’s levels will yield an ordering paradox if the derivational- inflectional distinction is maintained. The sam e problem arises when the issue of productivity is considered in section 6.6.2. Theoretical implications are briefly discussed in 6.6.3.

^^his section is a condensed version of Dai (1992c). 255

6.6.1. The Issues

Packard (1990:21-37) adopts the Lexical Morphology approach to Chinese word formation. Incorporated in the approach are the leading ideas of the theory, namely, that morphological levels are crucially ordered, there is no ordering within levels, and structure formed at one level is invisible at subsequent levels.

At the outset, Packard lays out the following assumptions for Chinese. First of all,

Chinese words have heads, and the head of a word is canonically determined by position, according to the class of the word, rather than by the relation berjveen categories in syntax.^'*

Then he proposes that a) complex nouns are right-headed; b) complex verbs are left-headed; and c) all affixation (including inflections) applies to the head of a word except at the last level (level

4), where default application is to the word. Based on these assumptions, he stipulates four levels for Chinese, as in (42).

^''Again, the notion of head in morphology is suspect, cf. Zwicky (1985:1-29), Bauer (1990:1- 51) iui iiic aiyuiiieiiis ayaiiisi fiead ed n ess in morphology, in Cninese, since pacKaro oetermines head by marginal positions in a word, this is extensionally equivalent to the first or last syllable/morpheme of the word, or to the morphological operation types of the prefixation, suffixation or compounding in word formation. These notions are needed anyway in morphology, which would render the notion head redundant. In support of the notion of head in morphology, Packard’s primary evidence is from speakers who decompose the words through reanalysis. They treat the canonically defined head as though it were the category of the matrix word, even if the head is transparently a member of a different category. There are at least two uncertainties involved here. First, it is doubtful that equating a theoretical construct (head) to putative psycholinguistic evidence is an accepted methodology in theoretical linguistics. Second, it is hard to see how heads in syntax and in morphology are connected in Packard’s sense. For instance, if word dan-xin [carry-heart] ’worry’ is reanalyzed as a verb-object (VO) phrase dan xin load mind’, in which dan is the head verb (cf. 3.3), it does not logically follow that the notion of head in syntax is automatically carried into morphology, entailing that dan- is the morphological head in word dan-xin. 256

(42) Packard’s 4-level model of Chinese morphology (adapted from Packard 1990:34)

Level 1 (compounding^® only): restrictive resultatives (gai-shan [correct-proper]

‘improve’, exocentric compounds (zuo-you [left-right] ‘influence’)

Level 2 (compounding and affixation): all other endocentric nominal compounds (^uo-qiu

[table-ball] ‘table tennis’), regular resultatives {tan-wan [talk-finish] finish discussing’),

derivational affixation operating on heads (-zi, -r, -tow. zhou-zi [table-zi] ‘table’), potential

affixation (-DE^®/bu-: tan-de-wan [talk-DE-finish] ‘able to finish discussing )

Level 3 (compounding and affixation): endocentric verbal compounds (tao-lun [discuss-

speak] ‘discuss’), VO-verbs {chu-ban [emit-edition] ‘publish’), reduced A-not-A questions

operating on head (^i-bu-xitiuan [like-not-like] ‘like or not? )

Level 4 (inflectional affixation only): aspect markers -le, -zhe, -guo {kan-zhe [look-duration]

‘looking’), plural forms -men {ren-men [person-plural] ‘people’)

With these levels, Packard appears to be able to explain some broad generalizations in morpheme ordering. For instance, *ping-zi-gai [bottle-ZI-cover] is ungrammatical (cf. ping-gai

‘bottle cover’), since -zi attaches at level 2, the same level at which nominal compounding operates, but -zi is forced to attach to the right-hand nominal head. Moreover, because of invisibility of the left-hand heads at earlier levels, the A-not-A questions at level 3 cannot apply to

^®Some of his compounds are only morphologically, not syntactically composite, since their components are bound or there may not be general syntactic rules to license their internal structures. Thus these compounds are equivalent to my bimophemic expressions (cf. 3.2).

^®DE is used here in contrast to the homophonie adjectival de {wo de shu [I de book] ‘my book ) and the adverbial de {manman de pao [slow de run] run slowly ). 257 the level 1 restrictive resultatives {*gai-bu-gai-shan ‘improve or not?') and exocentric compounds

(*zuo-bu-zou-you ‘influence or not?'), nor to the level 2 regular resultatives (*u-fan(-DE)-wan

‘finish talking or not?'). In addition, the so called potential infixes -DEJbu- can attach to the head of the regular resultatives {tan-DE-wan can finish talking') since both are formed at the sam e level

2, but cannot attach to the level 1 restrictive resultatives {*gai-bu-shan ‘cannot improve').

Similarly, regular verbal compounds at level 3 can undergo A-not-A questioning (^i-bu-xihuan ‘like or not?'), since both operate at the sam e level, but cannot undergo the potential infixation {*chu-

DE-ban [emit-DE-editicnj ‘publish’) at the earlier level. Finally, inflections at the last level apply to words by default just in case the head is invisible (jie-jue-le [resolve-decide-perf.] ‘have resolved’, *jie-le-jue).

Compelling counter-arguments to the Lexical Morphology approach are launched by

Sproat & Shih (1992). They point out that Packard’s evidence does not support morphological levels in the language, because some of his data can be covered by independent explanations, whereas other claims are simply false. For instance, Packard’s claim that -zi is restricted to the

right-hand member of compounds is generally false (cf. ju-zi pi ‘orange skin ). Furthermore,

unreduced A-not-A questions demonstrably cannot apply to the restrictive or regular resultatives

and exocentric verbs either {*zui-bu-zui-ying vs. *zui-ying-bu-zui-ying ‘recalcitrant or not?’), which

cannot resort to morphology for explanation, as (in)visibility of the head cannot be at issue. The

authors’ criticism is quite to the point, since it is well-known that, parallel to the resultative

compounds, the resultative phrases are equally immune to the A-not-A questions operating on

the left-hand heads (*pao bu pao-DE kuai vs. pao-DE kuai-bu-kuai ‘run fast or not?', see a

summary in Dai (1992a:109)). Finally, they argue that the potentials cannot apply to regular

compound verbs (e.g., VO-verbs), simply trecause the second morpheme (O) does not denote

the result/completion of the first (V), and hence the putative morphological level ordering is

irrelevant. As Sproat & Shih can offer independent syntactic, semantic or pragmatic accounts for

Packard’s data at level 1 through 3, they conclude that Chinese morphology is not level-ordered. 258

However, to Sproat & Shih (1992), Packard’s stipulation of level 4 for inflections "seems to be correct; it is true that inflectional morphemes in Mandarin attach to the whole word, which implies in turn that inflections apply after all forms of word-formation". But the authors still try to make level 4 otiose by deriving an aspectually inflected V through raising an uninflected V to the

INFL node, and adjoining it to an aspect inflection which originates there.

There are two problems with their proposal. The first one is methodological. Although the inflectional level is now a consequence of their account, the question is whether such a theory-dependent derivation can be accepted by linguists who do net practice GB theory. It seem s that Sproat & Shih’s alternative is preferable. According to them, saying that inflectional aspect markers are word-level affixes rather than these morphemes being attached at level 4 at

least has the advantage of placing them in a cross-linguistically well-instantiated morphological class (like English -ness or -iy, which attach to a full-fledged word, giving happiness or happily).

The second problem is empirical and theoretical. The question is whether Lexical

Morphology, when applied to Chinese, can capture the division between derivational and

inflectional morphology. In the remainder of this section, I will demonstrate, contra Packard’s

(1990) wish and Sproat & Shih’s (1992) belief, that levels do not reproduce the derivational-

inflectional distinction in that lexical morphology neither excludes nor enforces the possibility that

inflection might be at the last level, the upshot of this being that inflectional rules might appear

at any level.

6.6.2. Derivations Occurring at the Last Level and Inflections at the Previous Levels

The putative four levels in (42) strike one as ad hoc, since it is hardly possible to form a

natural class in terms of morphological categories within each level, except for the last level, which

is stipulated for inflections. Unfortunately, Packard’s reservation of level 4 merely for inflections

cannot bear further scrutiny. Since the plural -men {ren-men [person-plural] ‘people’) is

demonstrably a derivational affix, as it is not of syntactic relevance in Chinese (Dai 1991a:76, cf. 259

5.3.2), categories at level 4 are not exclusively inflectional, and hence not a natural class, either.

Worse still, true inflectional operations are found at the previous levels. They are the potential DE affixation for the resultative compounds (cf. 3.4, tan-DE-wan [talk-DE-finish] ‘able to finish discussing’) at level 2 and the reduced A-not-A questions (cf. 6.5, xi bu xihuan [like-not-like] ‘like or not? ) at level 3.

In order to maintain the derivational-inflectional distinction, which is the only merit Sproat

& Shih (1992) recognize for the level-ordered approach, Packard would have to move -men from level 4 to level 3. It cannot be moved to level 2, because there would be no way of prohibiting

*hai-men-zi [child-PL-nominal] ‘children’ (cf. hai-zi-men), on the assumption of no ordering within levels. However, the re assignment of -men to level 3 would reduce the elegance of the level ordered theory. Note that Packard’s (1990:23) stipulation is that all affixation (including inflection) applies to the head of a word except at the last level, where the default application is to the word, just in case the head is not visible. Now Packard would have to add some affixations other than level 4 morphology to th e default application to the word", which would be otherwise unique for the level 4 inflections. While this does no substantial harm to the theory, since the nominals which -men attaches to occur at the earlier levels 1 and 2, and at level 3 the heads are indeed rendered invisible, one is nontheless struck by the potentially excessive power of the theory: both head operations and default applications could possibly occur at any level. Of course, one could let -men stay at the last level by maintaining the unprincipled assumption that the plural morpheme is inflectional in the language.

The second hole Packard would need to fix for his levels is to move the DE/bu-affixation from level 2 and the A-not-A operation from level 3 to level 4 respectively, in order to keep a homogeneous level for inflections. Unfortunately, a famous ordering paradox occurring in other

languages would arise here, which would cause damage to the theory. This is because, after the suggested re-assignments, the A-not-A operation at level 4 would not be allowed to apply to the verbal compounds at level 3, and -DEIbu- could not be affixed to the resultatives at level 2, 260 because of the invisibility of the heads caused by bracket erasure at the previous levels.

Unfortunately, neither prediction is fulfilled. The A-not-A reduction can apply to the left-hand head of verbal compounds: tao-bu-tao-lun "discuss or not?', chu-bu-chu-ban "published or not?’, and so can the potentials to the regular resultatives: tan-DE-wan "able to finish discussing’, xie-bu- qingchu "unable to write clearly’. As far as I can see, there is no principled remedy, except to assum e that inflections may occur at any level (i.e., maintaining the potentials at level 2 and the

A-not-A question at level 3). Therefore I come to the conclusion that Lexical Morphology cannot capture the theoretical distinction between derivational and inflectional morphology.

6.6.3. The Issue of Productivity

An issue related to the derivation-inflection division conoerns the relative productivity of rules, which Packard (1990) takes as THE independent evidence for motivating the levels.

According to the theory, structures built at later levels are more productive than those at earlier levels. As pointed out by Sproat & Shih (1992), however, while inflectional morphology at level

4 is surely more productive than much of the rest of Mandarin morphology, some of Packard’s claims about relative productivity are completely unsubstantiated.

Worse still, Packard’s model is demonstrably unable to predict relative productivity amongst the relevant levels. First, nominal compounding at level 2 is fully productive, licensing an open-ended list of nouns in the language (cf. 3.5). Nominal compounding is far more productive than the nominal affixation (-z/, -r and -tuo) at the same level 2. For instance, suffixation of -tou as in shi-tou "stone’, mu-tou ‘wood’ is an unproductive process (Chao 1968:242) and must be learned on an item-specific basis (Li & Thompson 1981:43, also cf. 6.0). If level ordering could capture relative productivity, then one wonders why the nominal compoundings and nominal suffixations occur at the same level.

Moreover, Thompson (1973:369) demonstrates that the resultative compoundings (with and without the potential mode) at level 2 are completely productive (cf. 3.4). They are certainly 261 more productive than the endocentric verbal compounds and VO-verbs at the next level. In addition, the so-called VO-compounds at level 3 {dan xin [carry heart] -> dan-xin ‘worry'), which can take an object {dan-xin ta ‘worry about him’), are so isolated that Chinese only has about two dozen such combinations (of. 3.3).

Lastly, the level 3 A-not-A question is a general process (cf. 6.5.4), more productive than verbal compounding at the sam e level.

All of these, the regular nominal compoundings, resultative compoundings and reduced

A-not-A questions, contradict what the level order predicts regarding relative productivity in morphology. More problematic to the level-ordered model is that the three processes are as productive as the verbal aspectual inflections at the last level. A re-arrangement of levels according to relative productivity could not possibly keep the distinction between derivation and inflection, since the nominal and resultative compounds would have to occur at the same level as the verbal inflections. More damaging to the theory, the re-assigment would cause an ordering paradox, since on the one hand, nothing would prevent the A-not-A rule from applying to the resultatives: on the other hand, the rule would be wrongly prohibited from operating on the level

3 endocentric verbal compoundings and VO-verbs. Thus any effort to correlate relative productivity with level ordering while keeping the derivation-inflection division seem s doomed to failure. Since productivity a s the only putative independent motivation for levels induces more problems than it solves, Packard’s (1990) levels would simply, to put it in his own words, reduce to "nothing but fancy descriptive or notational device" (p.35).

6.6.4. Theoretical Implications

Having seen the failure of the level-ordered morphology to reproduce the derivational- inflectional distinction, one may wonder whether this distinction is so important that an adequate theory ought to capture it. The answer is yes. As summarized in 5.0 and 6.0, the theoretical assumption that inflectional morphology is syntactically relevant accounts for the behavioral 262 characteristics distinguishing the two grammatical subcomponents observed in traditional grammar (e.g., Bloomfield 1933:222-4).

First, inflectional morphemes do not change parts of speech but derivational morphemes are not similarly restricted. Second, inflectional morphemes apparently indicate the syntactic relationship between the host and another word in a phrase or a sentence. Derivational morphemes are powerless in this regard. Third, inflectional morphemes are morphologically general, or productive. Fourth, inflectional morphemes occur at the margins of words, outside of derivational morphemes.

In Chinese, the potential DE and the reduced A-not-A question, together with the resultative DE, the measurative TA, the aspectual -le, -zhe, -guo, and the deservative -tou, have all the characteristics that inflections of other languages have. They can apply to virtually every lexical item of the verbal subclasses they select for; thus the processes are morphologically general and productive, in contrast to derivations, which may have many exceptions. There is no change in part of speech after inflectional operations. The potential suffixed form in a resultative compound and the truncated form in an A-not-A question remain a verb In each construction. They signify the syntactic relationship between their hosts and the following VP, i.e., an unreduced VP and a resultative complement VP respectively. Like the other inflections, DE is located at the marginal position of the first word in a compound, and thus occurs after derivational morphemes and closes the word.^^

As put forward by Anderson (1988:171 ), the assumption that both inflection and derivation are in the lexicon is based on the observation that both deploy the same operations; there is apparently no formal operation which is unique to one domain. However, if the classes of productive inflections constitute the last level in the sense of Lexical Morphology, this ought to follow from the organization of grammar, rather than stipulation.

^^Forthe A-not-A question, I assume that after all the word-formation processes are completed l^i-huan 'like'), the truncation takes place (x;). 263

To reinforce the positional marginality and morphological productivity of inflectional

morphology, I should first point out that there is one important reason why an inflection often closes a word (or attaches at the last level in terms of Lexical Morphology) in addition to the assumption that inflectional rules apply after derivations: in realizing proper morphosyntactic

representations in natural languages, the most common morphological operations an inflectional

rule would choose are prefixation and suffixation. Provided that syntactic generality leads to an

associated morphological productivity, it is therefore the derivational-inflectional distinction and the syntactic reievance of inflections, but not at which level the operation occurs, which predict the productivity and marginal postions of an inflection in default. However, inflectional

morphology allows for exceptions, since a morphological rule may have lexical gaps. Moreover,

an inflection may call on a processual operation like infixation, reduplication or truncation, and

there is no guarantee that the effect must take place at the rim position of a word. Thus it would

be perfectly all right in the current framework for traditional grammarians to view DE as an infix

in the resultative potential, an inflection which appears within a compound, or at an earlier level.

To push the point of syntactic generality further, given that syntactic rules associated with

compositional semantics are productive, the current framework predicts nominal compounding

(though at level 2) to be productive, because its internal structure is licensed by a general

syntactic rule N+N > N, with the first N as the modifier to the second (cf. examples in 3.5). In

addition, the N+N compounding is licensed by a morphological rule, which imposes lexical

integrity on the matrix compound and allows for lexical gaps and semantic/phonological

idiosyncrasies.

It is true that cross-linguistically there exists certain covariation, following from the

assignment of levels, in both linear distance and phonological interaction between affixes and

stems. But such covariation is lacking in Chinese. For example, the main productive

phonological rule in the language. Third Tone Sandhi, shows sensitivity to morphological

structure, and must be able to apply in the lexicon, but it shows no sensitivity to morphological 264

levels (Shih 1986, Sproat & Shih 1992). The sam e is true of the rule of Final Elision (Dai

1990d:332 and references therein). This rule operates across the syntactic word boundary to

resyllabify a following bilabial as the coda of the first syllable on the level 3 A-not-A questions (x/ bu xihuan -> xip xihuan ‘like or not’), as well as within a morphological word on the plural -men

at level 4 {ta-men > tarn ‘they’). To say the least, the fact that no phonological rules in Chinese

crucially depend on morphological levels makes a Lexical Morphology approach to this language

much less interesting than for some other languages^® (Sproat & Shih 1992).

There is now a considerabie iiterature contra Lexical Morphology, which argues, by

reference to various paradoxes in level assignments in particular languages, that the hypothesis

of level assignments brings in its train more problems than it solves. See Gussmann (1988:237-9)

for a summary: Chinese morphology is another case in point.

Zwicky (19923:359-60) notes that the scheme of levels doesn’t follow (as does the

distinction between inflection and derivation) from foundational considerations in grammatical

theory (assumptions about universal grammar or the functions of rules). Levels are built on

observations about linguistic facts. He says "... This lack of grounding in an existing theory makes

layering [i.e., levels] metatheoretically suspicious, on the one hand, but exciting to theoreticians,

on the other, since we have the sense of discovering an unexpected fact about language, of

being surprised by a hidden orderliness in our world. Neither the suspicion nor the excitement

makes an argument, of course...".

In contusion, Packard’s lexical morphology approach fails to adequately account not only

for the morpheme ordering at the earlier levels (Sproat & Shih 1992), but also the derivational-

inflectional distinction at the last level. On my account, it is such a theoretical distinction rather

than the level stipulations which explains certain facts with respect to "level 4 and earlier levels"

in Chinese Morphology. Therefore levels should not be treated as grammatical constructs.

^®Even in other languages, as pointed out by Zwicky (1992a:359), such a covariation is a natural reflection of historical change, and a certain amount of the covariation follows from the distinction between derivational and inflectional morphology. 265

6.7. Summary

The analysis of inflectional morphology in Chinese not only has satisfactory empirical and methodological consequences, but also can solve some long-debated issues in Chinese syntax.

Specifically, since DE is an inflectional morpheme and attached to the first verb, it is the first verb, but not the second, which serves as the head of the resultative construction. Moreover, the universal assumption of headedness in a subcategorization frame and inflectional morphology marking the heads make the parochial PSC an artifact. Further, the result of this study shows that a syntactic Cass theory accounting for Chinese word order does not have the morphological support which is found for other languages. Instead, the predominant morphological head- markings in verbal subcategorization frames and syntactic dependent-markings elsewhere correlate to, and provide the motivation for, the head-initial word order within verbal subcategorizations, in contrast to the head-final word order elsewhere, in the language. The inflectional analysis appealing to subtraction improves on the previous analyses of the reduced

A-not-A question. Finally, the derivational-inflectional distinction along with the productivity issue

in Chinese reveals the inadequacy of the theory of Lexical Morphology which resorts to levels in word formation processes to capture such a necessary distinction. CHAPTER Vil

CLITICS

7.0. Introduction

Many languages have bound elements commonly referred to as clitics, which possess mixed properties of words and affixes. However, the term clitic is used in the literature to label a collection of related phenomena, but not necessarily the same. For instance, a clitic is broadly taken as having no independent accent and forming an accentual unit with the adjacent word, such as the 'em in spoken English: Give'em the money. Another case is the contracted auxiliary

-’s for is in He's going. These are called simple clitics by Zwicky (1977), which have the same syntactic distributions as their corresponding full syntactic words (them and is). By contrast, the

English possessive marker -’s in the boy's cat and the Latin -que in Senatus populusque Romanus

‘the senate and people of Rome' are called special clitics, which do not have corresponding full syntactic words. In addition to prosodic dependence, one characteristic common to clitics is that their distributions are syntactically determined: they occur in construction with some phrasal constituent in a syntactic domain. Thus Klavans (1982) takes the position that clitic is a theoretic construct, not a merely a useful pretheoretical cover term, and must be distinguished from words and affixes. This position gives rise to two lines of study. One is the formal syntactic and phonological description of clitics (Klavans 1985, Kaisse 1985, etc.); the other is the methodology of differentiating clitics from independent words (Zwicky 1985a), as well as from affixes (Zwicky

& Pullum 1983). But it is still not clear from these authors what clitics are, or at least there is no consensus on how the term clitic is used.

266 267

Appealing to the different notions of word, Nevis (1988) implicitly takes a clitic as a syntactic word attaching to its neighbor word to form a phonological word. Syntactically distributed affix-like elements (English possessive -s), called phrasal affixes, or syntactic words that are prosodically weak, and depend on and interact with their neighbors, called bound words

(the contracted auxiliaries in English), are classified as clitics. Zwicky (1990:212) explicitly points out that a clitic group thus formed is in general not a syntactic constituent (hence not a syntactic word). He also suggests that the combination is also a morphological word.’ Note that to Nevis and Zwicky, the notion clitic is derived from various notions of word. Therefore, cliticization constitutes another object of interest because the internal structure of a clitic group lies in the interface of syntax and phonology (perhaps also morphology).

In the following investigation of clitic-like elements in Chinese, I take a clitic to be a syntactic word prosodically attaching to and phonologically interacting with its adjacent syntactic word to form a phonological word (and, with some hesitation, a morphological word as well).

Note that my choice of definition is purely terminological, as different theoretical frameworks may

characterize and label clitics in different ways. The grammatical significance of such clitics, though, lies in the mismatch between syntactic words and phonological/morphological words,

rather than in the label clitic per se, analogous to the notion of compound, which shows

mismatches between syntactic words and morphological words (cf. 3.0).

The grammatical status of clitics contrasts sharply with that of particles, another related

and widely used term in the literature. Zwicky (1985a:290) maintains that particles constitute no

grammatically significant category. Most particles fall into either words or clitics by the tests which

distinguish between words and clitics (/b/d:286-9). Thus given the term clitic, together with well-

motivated structural levels of affixes, words, phrases and clauses (in different grammatical

’The following may indicate the morphological status of a clitic group (Zwicky, p.c.). There are arbitrary gaps for word-clitic combinations. For instance, the auxiliaries was and were cannot be so contracted. Further, particular slots in a clitic group may not be categorially definable in syntax, and there are suppletive forms for the clitic, as in the Tagalog data by Schatcher (1974). 268 components), particles have no place in grammar. Much for this reason, C-R. Huang (1985b) proposes a clitic analysis of Chinese sentential particles, inspired by Chao’s (1968:149) long ignored definition of those particles as enclitics (unstressed final syllables) in construction with a preceding phrase or sentence, but phonetically attached to the syllable immediately before it.

In line with Huang’s effort to suppress the atheoretical use of the term of particle in

Chinese grammar, I will show that the present framework will classify those sentential particles as words (syntactically and phonologically) instead of clitics (7.1). Their non-clitichood bears on some previously proposed universals with respect to the markedness and distributions of clitics.

If the analysis is extended to other prosodically weak words, the grammar of Chinese demonstrably does not make much reference to the notion of clitic (7.2).

7.1. Sentential Particles as Words but Not Clitics

The sentential particles in question include, but are not restricted to, ma (for yes-no questions), ne (for contrastive/expecting addressee’s expectation), be (for conjectural and advisative remarks), de (presuppositional) and le (inchoative/new state). They are located at the sentence-final position. Here I adopt Huang’s notation "=" for the putative clitic attachments.

(1)a. Zhangsan xihuan Lisi=ma?

Zhangsan like Lisi=ma

’Does Zhangsan like Lisi?’

b. Zhangsan xihuan Lisi=ne.

Zhangsan like Lisi=ne

‘(You might know, but) Zhangsan (really) likes Lisi.’ 269

c. Zhangsan xihuan Lisi=ba?

Zhangsan like Lisi=ba

‘(I suppose that) Zhangsan likes Lisi, (does he?)’

d. Zhangsan xihuan Lisi=de.

Zhangsan like Lisi=de

‘Surely, Zhangsan likes Lisi.’

e. Zhangsan xihuan Lisi= le.

Zhangsan like Lisi=!e

‘Nowadays, Zhangsan has begun to like Lisi.’

Huang puts sentential particles on the tests suggested in Zwicky & Pullum (1983:502) and

Zwicky (1985a:286) for differentiating clitics from affixes on the one hand and from words on the other hand. Since the results show that sentential particles behave neither like affixes nor words,

Huang conludes that they are clitics.

Huang gives good reasons for not taking sentential particles as affixes. First, they do not select the grammatical categories of their hosts as affixes do. Second, unlike affixation, the attachment of the particles has no arbitrary gaps and semantic idiosyncrasies. Last, they do not form part of the host words to which syntactic rules may refer. While readers are referred to

Huang for relevant examples, these behaviors contrast sharply with those exhibited by words suffixed by DE and -tou, as given in 6.3.1 and 6.5.1. Since sentential particles can pass the expansion test, as in (2), suffice it to say that they are not parts of their preceding words, and hence cannot be suffixes. 270

(2) Ta lai= ma? vs. Ta lai xuexiao=ma?

he come ma he came school ma

‘Is he coming?’ vs. ‘Is he coming to the school?’

The current framework offers explanation for the behaviors of the particles. They do not select their hosts because they attach at the sentential level, i.e., they form a constituent with the preceding sentence. But since they prosodically attach to the last words of their preceding constituents (S), there is no reason for the host words to belong to one category. Their attachment to the host shows no arbitrary gaps and results in no semantic idiosyncrasies, since the syntactic rules which refer to them are structurally general (syntactically productive and semantically compositional). Finally, no syntactic rules apply to the combination of the particles and their hosts, because the two do not usually form a syntactic constituent. Some of these explanations are consistent with or implicit in C-R. Huang’s (1985b) assumptions.

Having ruled out the possibility of being affixes, the sentential particles are either words or clitics. Because the particles can pass the expansion test, the current framework will categorize them as words by default, if there are no other reasons for classifying them as marked clitics (cf. Zwicky’s (1985a:289) metaconsideration, i.e., clitics are more marked than inflectional affxes and syntactic words).

Huang lists several pieces of evidence against the wordhood of these particles, but none of them strikes me as convinving. He apparently ignores Zwicky’s (1985a;284) remarks on the nature of lests" in linguistics. The point is whether these tests are necessary and sufficient conditions for defining the term clitic in the language. Zwicky cautions that the tests are more in the nature of characteristic symptoms, or tendencies, of linguistic state of affairs, than of exclusive

invariant concomitants of it. Independent and interfering factors can prevent even clear cases from exhibiting a certain symptom, and a particular symptom might result from some condition which is not at issue. This is what happens to Huang’s tests for sentential particles. Below, I 271 comment on Huang’s tests one by one, leading to a conclusion that Chinese sentential particles cannot be clitics but must be words in my framework.

Huang’s morpho-syntactic tests include boundness, closure and distribution. For Huang, words can typically occur in isolation, but the sentential particles cannot. However, atypical bound words exist. Function or grammatical words generally cannot occur alone cross- llnguistically, nor can the Chinese demonstrative zhe, monosyllabic numeral san and classifier ge in the specifier position of an NP occur in isolation as a short answer to a question, as seen in

(3).

(3) [zhe san ge] ren

this three classifer persons

’these three persons’

The sam e is true of prepositions ba, bei and cong ‘from’. More importantly, sentential particles are always bound, presumably since they make no sense at all to listeners when in isolation.

Therefore binding has little bearing on the putative clitic status of the particles. They could be words in the language as well.

Words, provided they are of the correct categories, allow affixes to attach to them.

Huang argues that sentential particles close off affixation, as in (4), where ma closes off the verb to block further suffixation of -zhe to the host. Therefore the particles cannot be words.

(4)a. Ta hai ku-zhe=ma?

he still cry dur.= ma

Is he still crying?’ 272

b. *Ta hai ku=ma-zhe

There are two independent explanations for such a blocking, even assuming ma is a word. First, since ma is not a verb, it is not the right category for suffixation of an aspect marker. Second, sentential particles seemingly close off affixation, because the syntax requires them to be at the absolute sentence-final position. Therefore they always linearly follow the suffixes, if any, of the preceding words.

Such a nonnegotiable sentence-final position for the particles would in turn render

Huang's distributional evidence moot. According to him, sentential particles are not like words because they lack the freedom of typical words in ordering relative to other constituents.

However, syntactic movement, or any syntactic rule, is subject to language- and structure-specific conditions. For instance, the ordering of demonstrative-numeral-classifer within a specifier phrase is as strict as the absolute final position for sentential particles. The assumption of specific conditions on rules also sheds light on the deletion test Huang gives. He points out that unlike words, sentential particles do not allow deletion under identity, as in (5a/b).

(5)a. Ta muqin lai=ma? Ta lai= ma?

he mother come=ma he come=ma

‘Is his mother coming? Is he coming?'

b. *Ta muqin lai? Ta lai=ma?

c. *Ta muqin= ma? Ta lai= ma?

d *fyiuqin lai=ma? Ta lai= ma? 273

Ironically, the words fa and lai are equally immune to the same deletion in the same structure, as shown in (5c/d).^ Therefore, resistance to deletion is not a sufficient condition on non-wordhood and clitichood.

The above exhausts Huang’s morpho-syntactic tests. It should be noted that there are clearly syntactic words in the language which exhibit all of the behaviors above. Preposition cong

‘from’ in the PR in (6) is a case in point.

(6) cong di-yi tiao jie

from ordinal-one M street

‘from the first street’

Cong is always bound; it closes off the prefixation of the following word it attaches to {*di-cong yi tiao jie)', it obeys a strict ordering relative to a PP {*di-yi tiao jie cong); and it resists deletion under identity. Therefore, the morpho-syntactic arguments made by Huang do not directly support the non-wordhood or clitichood of sentential particles, though the evidence is consistent with typical clitic behaviors. However, the counter-arguments launched here admittedly do not necessarily nullify the clitic analysis, and nor do they provide direct evidence for the wordhood of the particles. Below I seek direct phonological evidence for their wordhood, thereby refuting their supposed clitic status.

Huang’s phonological argument against sentential particles as words is straightforward: since sentential particles are always prosodically weak in having neutral tone and not bearing stress, they cannot be words. Underlying this contention is that each word has one of the four

^One might want to argue that (5c/d) are not really bad syntactically, since the problem seems to stem from the difficulty in interpreting them as deletion under identity. However, the sam e point can also be made for the grammaticality of (5b). 274 lexical tones in Chinese.^

However, such an accentual test is most unreliable when administered in various languages, as pointed out by Zwicky (1985a:287), although it is probabiy the most popular rule of thumb for distinguishing clitics from words. By assuming that a ciitic group is one phonological word corresponding to two syntactic words, there are two major problems here. The first is whether the prosodic features of tone and stress are sensitive to phonological words or to phonological phrases in Chinese. The second is whether there are clearly independent syntactic words which normally occur without tone and stress in the language.

There are good reasons to believe that domains for tone and stresse in Chinese count as phonological phrases, not necessarily phonological words, as suggested by Kratochvil

(1968:36-47) in a study of interactions between tone, stress and phrasal intonation. The directional complements of resultative compounds are a case in point for prosodically weak words. According to Chao (1968:438), directional complements {chu-lai [out-here]) are normally unstressed and occur in neutral tone (na-chu-lai ‘take out here’). They are clearly independent syntactic words, since they have freedom of ordering with respect to adjacent constituents with the same cognate meaning, as below. Chao (1968:478) Indicates that the alternative orderings do not affect prosodic weakness of the complement.

(7)a. na chu-lai yi zhi e

take out here one M goose

‘take out here a goose’

b. na yi zhi e chu-lai

c. na chu yi zhi e lai

^ h e current framework must assume that words in Huang’s sense here are phonological words, defined in purely phonological terms. 275

The complement bears a tone (Tone 1 and 2), only if the potential form is present in an R- compound {na-DE/bu-chu-lai ‘be (un)able to take out here'), while the potential -DE/bu- takes the neutral tone. If no one would want to recognize that one phonological word can instantiate a combination of verb-potential-complement, then stress and tone operate and alternate at the level of phonological phrase in the language. Thus the prosodic test is demonstrably as unreliable for differentiating clitics from words in Chinese as in other languages.

While prosodic weakness does not entail the non-wordhood or clitichood of sentential particles, neither does it provide evidence for their wordhood, either. In fact, one might maintain that in Chinese, prosodic weakness defines clitics in phonological words rather than weak phonological words in phrases, since which rules are used for defining the primitive prosodic domains can be determined on an arbitrary basis. For instance, one might use stress to define the phonological word in English. Thus monosyllabic prepostions, auxiliary verbs and determiners without a phrasal accent would all be clitics, prosodically leaning on its neighbor to form a phonological word. Of course, sentential particles and many other unstressed and neutral toned

items would be clitics in Chinese by this account. There seems to be no principled way of

defining clitics except when one encounters prosodically weak words in a language, some of which do and some of which do not phonologically interact with their hosts. In Finnish, for

instance, there are quasi- or semi-clitics (juuri ‘just’, sità (a modal adverbial)) which are not subject to word-internal vowel harmony, as well as true clitics {-hAn, -pA (sentential particles)) which

undergo the harmony (Nevis 1988).

To make this distinction, I maintain, with Nevis (1988), that a clitic is a syntactic word

which phonologically, not merely prosodically, interacts with its host within a phonological word,

evidenced by applicability of word internal sandhi rules. In addition to vowel harmonies in Finnish,

typical word internal phonological rules involve processes like the aspiration of initial stops and

devoicing of final consonants in English. Elsewhere, weak items default to “leaners", those 276 syntactic words which are prosodically dependent on their hosts, characterized by application of prosodic rules involving suprasegm ental features of stress, tone, or length. On this assumption, the English function words and Chinese sentential particles are not clitics, but prosodically weak words.'*

It should be emphasized that sentential particles do not merely default to words but are also by no means clitics under phonological considerations. This is because there is no phonological interaction between the particles and hosts, even though the condition for the word internal sandhi Final Elision (FE) tor ma and ba are met. FE cannot optionally delete the rime of a second syllable and resyllabify its bilabial stop onset as the coda of the first syllable as it can in wo men -> wom 'we' and ba-ba > bap ‘dad’ (cf. 2.3.2). The inapplicability of word internal sandhi FE to sentential particles in (8b/c) clearly demonstrates that they cannot be clitics in

Chinese.

(8)a. Zhangsan xihuan wo=m a/ba.

Zhangsan like l=ma/ba

‘Does Zhangsan like m e?’ (ma)

’(I suppose that) Zhangsan likes me, (does he?) (ba)

b. ‘Zhangsan xihuan Lisim?

c. ‘Zhangsan xihuan Lisip?

By Huang’s clitic account, clitics which lack such word-internal phonological interaction in a domain for application of internal sandhi rules must be distinguished from those not lacking (to

*Since clitics are taken to be more marked than words and affixes, if phonological evidence is lacking for one or the other, a non-affixal item defaults to word rather than being assum ed as a marked clitic by metaconsideration (Zwicky 1985a:289). In this regard, the prosodic approach has to recognize more marked clitics; by contrast, the phonological approach decreases the markedness of grammar of human languages. 277 be mentioned).

Because sentential particles are neither affixes nor clitics in the current framework, I conclude that they are (syntactic and phonological) words, more precisely, prosodically weak and bound, function words in Chinese.

7.2. Discussions

In a separate work, C-R. Huang (1987:49) argues that the adjectival de is a clitic for the sam e reasons above, in particuiar, he adduces prosodic weakness, ciosing off affixation, boundness, and strict ordering with other words to prove that de is not word. Here my account applies equally well, namely, these tests are not sufficient conditions on the non-wordhood of de.

Since there is no phonological evidence (as I find with FE) for its clitic or word status, de defaults to word.

Not being clitics, sentential particles and de have no place in C-R. Huang’s (1985b) evaluation of Zwicky’s (1985a) metaconsideration of markedness among words, affixes and clitics, and of Zwicky’s (1977) and Kaisse’s (1982) proposals on the universal distributional positions of sentential clitics. According to Huang, Zwicky’s hierarchy of markedness in which clitics are more marked than inflectional affixes is called into question, because Huang believes, from his study, that Chinese is relatively rich in clitics and relatively poor in inflections (he only assum es -fe and

-zhe are the “most probable candidates' for inflections). Thus inflections in Chinese seem to him to be at least as marked as, if not more marked than, clitics. Huang also casts doubts on

Zwicky’s observation that sentence-final position is the least common position for sentential clitics and Kaisse’s universal that sentential clitics may only occur in second position. But Huang’s questioning of the proposed language universals is rendered irrelevant if one assumes the strict phonological (and syntactic) notion of clitics, since in the current framework, the items in question are not clitics in Chinese. 278

Although I am not necessarily committed to these particular supposed language universals atx)ut the distributional positions of clitics, the result of this thesis shows that, at least in Chinese, inflectional affixes are less marked than clitics, thus supporting Zwicky’s metacriterion of markedness. In such a typical isolating language, six inflectional affixes have t>een documented, in contrast to only two marginal cases of clitics. They are -p ‘not’ in reduced A-not-

A questions (x/ bu xihuan > xip xihuan ‘like or not?’) and in the second verb in the serial verb construction {qu pu chuang > qup chuang ‘go make t>ed‘, cf. 2.3.2), where the internal sandhi

FE optionally applies across t^vo syntactic words which form no constituent, and thus makes them one phonological word. The reason for the marginality of these two cases is that they are simple clitics in Zwicky’s (1977) sense, which serve as the reduced forms at the sam e positions as the corresponding full forms {bu ‘not’ and pu ‘make (bed)').

Summarizing, the label clitic in the current framework is reserved for weak syntactic words exhibiting word-internal behaviors in phonology and perhaps affixal behaviors in morphology.

Given the well motivated notions of word in syntax, phonology and morphology, clitics have no place in syntactic rank. Cliticization constitutes yet another set of mismatches among words in different grammatical components, calling for an adequate grammatical description, In Chinese, prosodically weak function words seems to abound, while clitics are apparently rare, therefore there is not much need for its grammar to refer to the notion of clitic at all. CHAPTER VI!!

CONCLUSIONS

8.0. introduction

This thesis has clarified a number of universal concepts of morphology and syntax as applied to Chinese by investigating the compounding process, derivational morphology and inflectional morphology of the language. In this concluding chapter, 1 will try to draw together the most important assumptions, positions, results, and conclusions that have been discussed in this work, and suggest directions for future research.

8.1. Word, Affix and Clitic

There are at least three notions of word in individual grammatical components, and the mapping between words in syntax, phonology and morphology is not always one-to-one. A syntactic word is a minimal constituent to which syntactic rules may refer; a phonological word is a certain prosodic domain in which word-internal phonological rules may apply; and a morphological word is a maximal domain in which morphological rules may apply. The main object of morphology is the morphological word and its relation with other components of grammar. Chinese grammar needs to refer to each of these three notions of word.

Affixes are characterized by five tendencies they exhibit: boundness, prcxJuctivity, generalized meaning, grammatical function, and attachment to a free form. A net parochial result of my study is the enlargement of the affix inventory of Chinese, as well as an expanded role for affixes in the grammar. Since affix has not been strictly defined in the literature, these five tendencies could serve as a starting point toward a universal notion of affix in human languages.

279 280

A clitic is taken as a syntactic word that phonologically interacts with, not merely leans prosodically on, its host syntactic word to form a phonological word, as evidenced by the applicability of word internal sandhi rules. It follows that Chinese sentential particles are not clitics but ordinary words, because there is no phonological interaction between the particles and the hosts, even though the condition for the application the word internal sandhi rule of Final Elision is met. Furthermore, given well motivated notions of word in syntax, phonology and morphology, there does not seem to be much need for the grammar of Chinese to refer to the notion of clitic at ail.

8.2. The Methodology of Identifying Morpheme Boundaries within a Word

Some methods have been developed in order to differentiate word boundaries in a phrase from morpheme boundaries in a word, based on the assumption that phrasal syntactic rules do not refer to proper parts of a morphological word. In particular, a constituent cannot be part of a word if it is pronominalizable, able to occur in isolation, deletable in discourse, modifiable at the phrasal level, or expandable by inserting phases on both its left and right side.

8.3. The Distinction between Syntax and Morphology

A division between syntax and morphology can be drawn in Chinese on the assumption that principles governing the two components of grammar are distinct with regard to allowing alternative ordering of constituents, the ability to change parts of speech, reference to phonology, acceptance of optional dependents, possible suppression of a constituent, and the extent to which exceptions to rules are tolerated.

Without reference to morphology, Chinese syntax (plus phonology) alone is not sufficient to describe the structure of the language. It has been shown that, among other things, the following puzzles call for morphological solutions as well: many items which apparently belong

to a certain major syntactic category cannot freely occupy a proper slot in a construction where 281 the corresponding syntactic words occur; some putative syntactic deletion rules (such as those needed to derive A-not-A questions) seem to work into words, apparently violating the UH; topicalization and expansion are unexpectedly blocked in certain semantically VO structures; prosodic domain formations for reduced A-not-A questions and certain A-t-N concatenations are sensitive to word structures; and some "phrasal" nominal categories (A+N and N+N) exhibit lexical, semantic and/or phonological idiosyncrasies, and also observe the UH.

8.4. Compounding and Syntax-Morphology Interface

The term compound/compounding is reserved for a syntactic word (W) composed of two syntactic words, and correspondingly, a morphological word (w) composed of two morphological words. Put another way, the immediate constituents of the matrix w are licensed by a morphological rule concatenating nonaffixal items, and the internal structure of the matrix W is licensed by a syntactic rule. Such a joint syntactic and morphological compounding has a special status in a grammar, because a compound is licensed by the interface between syntax and morphology, but not by either grammatical component alone. It follows that a) syntactic operations referring to W may refer to the internal Ws of a compound; b) compounding rules may be as productive and recursive as phrasal rules; c) there may be grammatical relations between the component W/w’s of a compound; d) compounding may have arbitrary lexical gaps and semantic and/or phonological idiosyncrasies; and e) W's in a compound observe the lexical integrity of the matrix compound w.

Some VO-structures have been examined with respect to the (in)separability of their constituents. The proposed analysis not only abandons the parochial PSC and maintains the universal LIH but also yields a corollary hypothesis that lexicalization of a sequence of W/w’s is determined by both syntax and morphology. This hypothesis is further substantiated by the study of the R-compound and the R-construction. In particular, a sequence of Ws is a compound if the syntax is able to categorially describe its internal structure and to place the sequence in a 282 position calling for rank W, and if morphology imposes lexical integrity on it. While a compound can be described by the interface between syntax and morphology, the formal similarities and differences between the compound and its putative corresponding phrase are captured by rule relationships, i.e., the degree to which the formal requirements of the syntactic rule licensing the compound overlaps with and deviates from the requirements of the rule licensing the phrase.

The synchronic quadrisyllabic N+N and A+N combinations, traditionally taken as phrases, have been shown to be compounds. They have lexical gaps and specialized meanings, and their components are inseparable and unable to receive phrasal modifiers. This investigation has also shown that coordination as a syntactic operation cannot apply in morphology except for compounding. Coordination in morphology creates a compound within a compound, and conjunction in compounds is semantically coordination plus equation making each conjunct parallel.

Little evidence is available for Chinese compounding as simply the reflection of the organization of its phrasal syntax. The view of morphology as the syntax of words is misleading even when it applies to compounding, since it downplays two crucial differences between compounds and phrases, i.e., syntactic words in a compound behave differently from those in a phrase in that the former must observe the lexical integrity of the compound, and a compound as a morphological word may exhibit lexical, semantic and/or phonological idiosyncrasies.

8.5. The Distinction between Derivational and Inflectional Morphology

Based on the theoretical assumption that inflectional morphology is the part of morphology which is relevant to syntax and that derivational morphology is of largely lexical function, arguments are put forward for the claim that the measurative TA, the resultative DE, the deservative -tou, and the three aspect markers -le, -zhe and -guo are inflectional morphemes in

Chinese. Of particular interest is the fact that a morphosyntactic government relationship is observed in the deservative construction: you subcategorizes for its complement and governs the 283 morphological tou-form of its head, on a par with the well-known case government in European languages. The conclusion of the study not only refutes the predominant view that Chinese has little or no inflectional morphology, but also yields some important empirical, methodological and theoretical consequences.

Empirically, the behavioral characteristics of these suffixes are consistent with those of inflectional morphemes in other languages - namely, the productivity of their attachment, the lack of change in part of speech of their hosts after suffixation, the marking of syntactic relationships between their hosts and some other constituents in construction, and the marginal positions they occupy in a word.

As for methodology, the inflection's rim position in a word in turn can be used as a morphological criterion for identifying word boundaries, as a verbal inflectional morpheme closes a verb.

8.6. Chinese Inflectional Morphology and Its Theoretical Consequences

Some long-debated theoretical issues in Chinese syntax can be better understood by considering the inflectional morphology of the language. In particular, since the first verb inflected with DE is the morphosyntactic locus of the R-construction, this verb is the head of the matrix VP.

Moreover, the fact that the inflected head V-DE subcategorizes exclusively for a resultative complement independently rules out the co-occurrence of a resultative complement and an NP object following the V-DE; as a result, the PSC, formulated as a filter just .or that purpose, is unnecessary.

In addition, inflectional morphology in Chinese does not lend empirical support to the

Case account of word order in Chinese as it does in European languages. Not only are there no case alternations in morphologically uninflected NPs in different syntactic distributions, an inspection of the "case" marking -tou demonstrates that NPs "case'-marked by -tou also have no morphological alternations, whether they occur preverbally or postverbally, or in an argument 284 position or a non-argument position.

A correlate to Chinese word order is observed. Most inflections in Chinese attach to the head V rather than to the following dependent (or complement) in verbal subcategorization frames, which is to say that Chinese is largely a morphologically head-marking language. Cross- linguistically, a preference for head-marking correlates with a default head-initial (i.e., V-initial) order in the frames. Given that Chinese word order is head-final in general, it is not surprising that the language employs the head-marking and verb-initial patterns just in subcategorization frames, as an ’exception", if the notion of head- and dependent-markings is extended from morphology to syntax, Chinese clearly exhibits a dependent-marking tendency (via function words) outside of verbal subcategorizations, correlating with its basic head-final order.

The controversy over the analysis of the reduced A-not-A question can be settled by proposing a verbal inflection formed by morphological truncation to the first syllable of the stem.

The analysis not only captures what those employing reduplication plus bu-infixation can, but also better accounts for the constructional similarities among the family of A-not-A questions.

The derivational-inflectional distinction in Chinese reveals the inadequacy of the theory of Lexical Morphology and Phonology, which resorts to levels in the word formation process to capture such a distinction. Specifically, levels do not reproduce the derivational-inflectional distinction, in that lexical morphology neither excludes the possibility that inflection might occur at the last level, nor enforces this, so that inflectional rules could in fact appear at any level. The fact that a reassignment of levels to maintain the derivational-inflectional distinction would unavoidably cause an ordering paradox, among other things, leads to the conclusion that levels should not be treated as a construct of grammar.

8.7. Concluding Remarks

As mentioned at the beginning of this work, it has been hotly disputed whether or not distinctions should be drawn in a theory of grammar between syntax and morphology, between 285 compounds and phrases, and between derivational and inflectional morphology. Chinese is the object language of this study because the language has long been cited as one of the most isolating and analytical languages of the world, having little or no morphology. In this respect, its syntax (plus phonology) should be necessary and sufficient for describing its structure, including word structure. However, this work convincingly argues for maintaining the syntax- morphology, compound-phrase, and derivation-inflection distinctions even in such a highly analytical language. I hope that this study substantially contributes to the understanding of these three universal distinctions in grammar, which i believe have been insufficiently appreciated in most modern frameworks of linguistic theory.

For the first time, Chinese has treen systematically demonstrated as possessing most or all of the varieties of morphology in universal grammar -- derivation and inflection in addition to compounding, like European and other languages. The debunking of the misconception that the language is morphology-free may encourage interested linguists to re-examine the structures of other analytical languages. For example, Vietnamese and Classical Chinese have long been presumed to be more isolating and analytical than Modern Chinese. However, from a future investigation of these two languages, 1 would hope to see a further claim made about human languages: that languages without morphology do not exist.^

1tThis i claim is made implicitly in Joseph and Janda (1988). BIBLIOGRAPHY

Anderson, Stephen R. 1977. On the Formal Description of Inflection. Chicago Linguistic Society 13, 15-44.

Anderson, Stephen R. 1982. Where Is Morphology?. Linguistic Inquiry 13, 571-612.

Anderson, Stephen R. 1984. Rules as 'Morphemes' in a Theory of Inflection. 1983 Mid-America Conference Papers, 3-21. Boulder: University of Colorado.

Anderson, Stephen R. 1985. Inflectional Morphology. Language Typology and Syntactic Description Vol.lll, ed. by T. Shopen, 150-201. Cambridge; Cambridge University Press.

Anderson, Stephen R. 1986. Disjunctive Ordering in Inflectional Morphology. Natural Language and Linguistic Theory 4:1, 1-31.

Anderson, Stephen R. 1988. Morphological Theory. Linguistics: The Cambridge Survey Wo\.\, ed. by F.J. Newmeyer, 146-191. Cambridge: Cambridge University Press.

Aronoff, Mark. 1976. Word Formation in Generative Grammar. Cambridge, Massachusetts: MIT Press.

Awbery, Gwen. 1975. Welsh Mutation: Syntax or Phonology? Archivum Linguisticum 6,14-25.

Bach, Emmon. 1984. Some Generalizations of Categorial Grammar. Varieties of Formal Semantics, eds. by F. Landman & F. Veltman, 1 -23. Dordrecht: Foris.

Baker, C. L & Michael K. Brame. 1972. 'Global Rules’: A Rejoinder. Language 48:1, 51-75.

Baker, M irk C. 1987. Incorporation: A Theory of Grammatical Function Changing. Chicago: Chicago University Press.

Bates, Dawn. 1988. Prominence Relations and Structures in English Compound Morphology. Ph.D dissertation. University of Washiongton.

Bauer, Laurie. 1988. Introducing Linguistic Morphology. Edinburgh: Edinburgh University Press.

Bauer, Laurie. 1990. Be-Heading theVJord. Journal of Linguistics 26, 1-31.

Bloomfield, Leonard. 1933. Language. New York: Henry Holt & Co.

Callaghan, Catherine A. 1987a An 'Indo-European' Type of Copula in Plains Miwok. In Honor of Use Lehiste, eds. by R. Cannnon & L Shockey, 331-338. Dordrecht: Foris.

286 287

Callaghan, Catherine A. 1987b. Northern Sierra Miwok Dictionary. Berkeley: University of California Press.

Chan, Ning-ping. 1984. The NounyAiternative: Crossing the Verb/Noun Borderline in Chinese. Ph.D dissertation. University of California, Berkeley.

Chao, Yuen-Ren. 1968. A Grammar of Spoken Chinese. Berkeley: University of California Press.

Chen, Matthew V. & William S-Y. Wang. 1975. Sound Change: Actuation and implementation. Language 51:2,255-281.

Chen, Ping. 1987. Shi Hanyu Zhong Yu Minci Chengfeng Youguande Sizhu Gainian [On Referentiality vs. Non-Referentiality, Identifiability vs. Non-ldentifiability, Specificity vs. Non- Sepecificity, and Genericity vs. Non-Genericity in Chinese]. Zhongguo Yuwen [Chinese Language and Writing] 1987:2, 81-92.

Cheng, Chin-chuan. 1973. A Synchronic Phonology of Mandarin Chinese. The Hague/Paris: Mouton.

Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge MA: MIT Press.

Chomsky, Noam. 1970. Remarks on Nominalization. Readings in English Transformational Grammar, eds. by R. Jacobs & P. Rosenbaum, 184-221, Waltham, Massachasetts: Ginn and Company.

Chomsky, Noam. 1982. Lectures on Government and Binding: The Pisa Lectures. 2nd Edition. Dordrecht: Foris.

Dai, John Xiang-ling. 1989. The Alternative and A-not-A Interrogatives in Chinese. Ms., The Ohio State University.

Dai, John Xiang-ling. 1990a. Historical Morphologization of Syntactic Structures: Evidence from Derived Verbs in Chinese. Diachronica Vil:1, 9-46.

Dai, John Xiang-ling. 1990b. Some Issues on A-not-A Questions in Chinese. Journal of Chinese Linguistics 18:2, 285-316.

Dpi .Inhn viann-lînn 1QQ0r» nofotinn I InHor A.nrvf.A »r^ Revisited. Paper presented at the Second Northeast Conference on Chinese Linguistics, University of Phiiadephia, May 4-6, 1990.

Dai, John Xiang-ling. 1990d. Syntactic Constructions in Serial Verb Expressions in Chinese. The Ohio State University Working Papers in Linguistics 39, 316-339.

Dai, John Xiang-ling. 1991a. The Resuitative DE as an Inflectional Morpheme in Chinese. Eastern States Conference on Linguistics 7, 67-78.

Dai, John Xiang-ling. 1991b. Inflectional Morphology in Chinese. Paper presented at The Third North American Conference on Chinese Linguistics, Cornell University, May 3-5,1991.

Dai, John Xiang-ling. 1991c. The Negator BU and a Morphosyntactic Analysis of A-not-A Questions in Chinese. Chicago Linguistic Society 27.2. 288

Dai, John Xiang-ling. 1992a The Head in Wo Pao De Kuai. Journal of Chinese Linguistics 20:1, 84-119.

Dai, John Xiang-ling. 1992b. Rethinking Case Theory for Constituency and Word Order in Chinese. Journal of the Chinese Language Teachers Association 27:1.

Dai, John Xiang-ling. 1992c. Derivational-Inflectional Disitinction and Levels in Chinese Morphology. Chicago Linguistic Society 28.2.

Defrancis, John. 1984. The Chinese Language: Fact and Fantasy. Honolulu: University of Hawaii Press.

Di Sciulio, Anna M. & Edwin Williams. 1987. On the Definition of Word. Cambridge: MIT Press.

Dcwty, David. 1978. Applying Montague’s Viev/s on Linguistic Metathscry to the Structure of Lexicon. Chicago Linguistic Society 15.2, 97-137.

Dowty, David. 1979. Word Meaning and Montague Grammar. Dordrecht: D. Reidel.

Dryer, Matthew S. 1992. The Greenbergian Word Order Correlates. Language 68:1, 81-138.

Ernst, Thomas. 1988. Structure vs. Function in the Chinese Verb Phrase. Bloomington: Indiana University Linguistics Club.

Fabb, Nigel. 1984. Syntactic Affixation. Ph.D dissertation, MIT.

Fulmer, S. Lee. 1991. Dual-Position Affixes in Afar: An Argument for Phonologicaliy-Driven Morphology. WCCFL 9: Proceedings of the Ninth West Coast Conference on Formal Linguistics, 198 203.

Gazdar, Gerold, Ewan Klein, Geoffrey Pullum, & Ivan Sag. 1985. Generalized Phrase Structure Grammar. Cambridge: Harvard University Press.

Gussmann, Edmund. 1988. Review of Mohanan 1986. Journal of Linguistics 24:1, 232-239.

Hall, R. M. R. & Beatrice Hall. 1970. A Problem in Rule Ordering. Studies in Honor of J. Alexander Kerns, eds. by R. Lugton & M. Saltzer, 49-58. The Hague: Mouton.

Hockett, Charles F. 1954. Two Models of Grammatical Description. Word 10, 210-231.

Hockett, Charles F. 1958. A Course in Modern Linguistics. New York: Macmillan.

Hoeksema, Jacob. 1984. Categorial Morphology. Ph.D dissertation, Rijksuniversiteit, Groningen.

Hoeksema, Jacob & Richard Janda. 1988. Implication of Processual-Morphology for Category Grammar. Categorial Grammar and Natural Language Structures, eds. by R. Oehrle, E. Bach & D. Wheeler, 199-247. Dordrecht: D. Reidel.

Huang, Chu-Ren. 1985a. On Pseudo-Possesive NPs in Mandarin Chinese. Cornell Working Papers in Linguistics 7,129-148. 289

Huang, Chu-Ren. 1985b. Chinese Sentential Particles: A Study of Cliticization. Paper presented at the LSA Annual Meeting, Seattle, WA.

Huang, Chu-Ren. 1987. Mandarin Chinese NP de: A Comparative Study of Current Grammatical Theories. Ph.D Dissertation, Cornell University.

Huang, Chu-Ren. 1989. Cliticization and Type-Lifting: A Unified Account for Mandarin NP de. Bloomington: Indiana University Linguistics Club.

Huang, Chu-Ren. 1990. Review of Functionalism and Chinese Grammar. Journal of Chinese Linguistics 18:2, 318-334.

Huang, Chu-Ren & Louis Mangione. 1985. A Reanalysis of de: Adjunct and Subordinate Clauses. West Coast Conference on Formal Linguistics 4, 80-91.

Huang, Chu-Ren & Ruo-ping Mo. 1992. Mandarin Ditransitive Constructions and the Category of gel. Berkeley Linguistic Society 13.

Huang, C-T. Jam es 1982, Logical Relations in Chinese and the Theory of Grammar. Ph.D dissertation, MIT.

Huang, C-T. Jam es 1984. Phrase Structure, Lexical Integrity, and Chinese Compounds. Journal of Chinese Language Teachers Association 19, 53-78.

Huang, C-T. Jam es 1988. Wo Pao De Kuai and Chinese Phrase Structure. Language 64:2, 274- 311.

Huang, C-T. Jam es 1989. Modularity and Explanation: The C ase of Chinese A-not-A Questions. Proceedings of the Third Ohio State University Conference on Chinese Linguistics, eds. by M. Chan & T. Ernst, 141-169. Bloomington: Indiana University Linguistics Club.

Hudson, Richard A. 1976. Regularities in the Lexicon. Lingua 40:2/3, 115-130.

Hudson, Richard A. 1987. Zwicky on Heads. Journal of Linguistics 23, 109-132.

Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge, Massachusetts: MIT Press.

Jackendoff, Ray. 1975. Morphological and Semantic Regularities in the Lexicon. Language 51, 639-671.

Janda, Richard D. 1983. Morphemes Aren’t Something That Grows on Trees: Morphology as More the Phonology than the Syntax of Words. Chicago Linguistic Society 19.2, 79-95.

Jensen, John T. 1990. Morphology: Word Structure in Generative Grammar. Amsterdam/Philadephia: John Benjamins Publishing Company.

Joseph, Brian D. & Richard D. Janda. 1988. The How and Why of Diachronic Morphologization and Demorphologization. Theoretical Morphology, eds. by M. Mammond & M. Noonan, 193-210. Orlando, FL: Academic Press. 290

Jiang, Zixin. 1991. Some Aspects of the Syntax of Topic and Subject in fAandarin Chinese. Ph.D dissertation, The University of Chicago.

Jin, Xiaochun. 1991. Modularity and Chinese Compounds. Journal of the Chinese Language Teachers Association 26:1, 33-48.

Kaisse, Ellen. 1982. Sentential Clitics and Wackernagel's Law. kVesf Coast Conference on Formal Linguistics 1,1-14.

Kaisse, Ellen. 1985. Connected Speech: The Interaction between Syntax and Phonology. Orlando: Academic Press.

Kiparsky, Paul. 1982. Lexical Morphology and Phonology, Lingustics in the tdorning Calm, ed. by Linguistic Society of Korea, 3-91. Seoul: Hanshin.

Kiparsky, Paul. 1984. On the Lexical Phonology of Icelandic. Nordic Prosody III: Papers from a Symposium, 135-164. Stockholm: Almqvis & Wiksell.

Klavans, Judith. 1982. Some Problems in a Theory of Clitics. Bloomington: Indiana University University Linguistics Club. 1980 Ph.D dissertation. University College London.

Klavans, Judith. 1985. The Independence of Syntax and Phonology in Cliticization. Language 61:1, 95-120.

Kratochvil, Paul. 1968. The Chinese Language Today: Features of an Emerging Standard. London: Hutchinson University Liberay.

Lakoff, Robin T. 1974. Linguistic Theory and the Real World. Studies in Syntax and Semantics 18. 1-49.

Lapointe, Steven. 1980. A Theory of Grammatical Agreement. Ph.D dissertation. University of Massachusetts.

Lessen Kloeke, W. U. S. van. 1985. Truncation in German - Morphotactic versus Phonotactic Constraints. Linguistics in the Netherlands 1985, 121-128. Dordrecht: Reidel.

Li, Charles N. & Sandra A. Thompson. 1976. Subject and Topic: A New Typology of Language. anrt Trxn'm o H c h i ; C\ |_î 4 ^ 7 - 4 0 0 M o \a / Vrtrl^* Anarfom ir* P r o c c

Li, Charles N. & Sandra A. Thompson. 1981. Mandarin Chinese: A Functional Grammar. Berkeley: University of California Press.

Li, Ya-Fei. 1990. On V-V Compoundings in Chinese. Natural Language and Linguistic Theory 8:2, 117-207.

Li Yen-Hui Audery. 1990. Order and Constituency in Mandarin Chinese. Dordrecht: Kluwer Academic Publishers.

Lieber, Rochelle. 1983. On the Organization of the Lexicon. Ph.D dissertation, MIT. 291

Lieber, Rochelle. 1992. Deconstructing Morphology. Chicago: University of Chicago Press.

Libucha, Mark. 1990. Arguing against Mandarin A-not-A as Phonological Reduplication in PF (Early version: Beijing Mandarin A-not-A in the Morphology and in the Syntax). Ms., The Ohio State University.

Light, Timothy. 1979. Word Order and Word Order Change in Manderin Chinese. Journal of Chinese Linguistics 7 :2 ,149-180.

Lu, Zhi-wei. (eds.) 1964. Hanyu Goucifa [A Grammar of Chinese Word-formation]. Beijing: Kexue Chubanshe [Science Publication Co.].

LQ, Shu-xiang. 1963. Xiandai Hanyu Danshuangyinjie Wenti Chutan [A Preliminary Investigation on the Bisyllabification of Chinese]. Zhongguo Yuwen [Chinese Language and Writing] 1963:1,10- 22.

LÛ, Shu-xiang. 1985. Yiwen Pouding Kending [Interrogation, Negation and Affirmation]. Zhongguo Yuwen [Chinese Language and writing] 1985:4, 241-250.

Lyons, John. 1968. Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press.

Maling, Joan. 1983. Transitivity Adjectives: A Case of Categorial Reanalysis. Linguistic Categories: Auxiliaries and Related Puzzles Vol.l, eds. by F. Heny & B. Richards, 253-289. Dordrecht: D. Reidel.

Marantz, Alec P. 1984. On îLie Nature of Grammatical Relations. Cambridge: MIT Press.

Matthews, Peter H. 1965. The Inflectional Component of Word-and-Paradigm Grammar. Journal of Linguistics 1 , 139-171.

Matthews, Peter H. 1972. Inflectional Morphology. A Theory Study Based on Aspects of Latin Verb Conjugation, Cambridge: Cambridge Unversity Press.

Matthews, Peter H. 1974. Morphology: An Introduction to the Theory of Word-Structure. Cambridge: Cambridge Unversity Press.

Mohanan, K.P. 1986. The Theory of Lexical Phonology. Dordrecht: D. Reidel.

Moravcsik, Edith A. 1977. On Rules of Infixing. Bloomington: Indiana University Linguistics Club.

Nespor, Marina & Irene Vogel. 1986. Prosodic Theory. Dordrecht: Foris.

Nevis, Joel. 1984. A Non-Endoclitic In Estonian. Lingua 64:2/3, 209-224.

Nevis, Joel. 1988. Finnish Particle Clitics and General Clitic Theory. New York: Garland Publishing Co. 1985 Ph.D dissertation. The Ohio State University. An Early Version in The Ohio State University Working Papers in Linguistics 33 (1986).

Nevis, Joel & Brian D. Joesph. 1992. Wackernagel Affixes: Evidence from Balto-Slavic. Yearbook of Morphology 5. 292

Nichols, Johanna 1986. Head-Marking and Dependent-Marking Grammar. Language 56:1,56-119.

No, Yongkyoon. 1991. Case Alternation on Verb Phrase Internal Arguments. Ph.D dissertation. The Ohio State University.

Norman, Jerry. 1988. Chinese. Cambridge: Cambridge University Press.

Cresnik, Janez & Magnus Petursson. 1977. Quantity in Modern Icelandic. Arkiv for Norjisk Filologi 92, 155-171.

Packard, Jerome L 1990. A Lexical Morphology Approach to Word Formation in Mandarin. Yearbook of Morphology 3, 21-37.

Pullum, Geoffrey & Deirdre Wilson. 1977. Autonomous Syntax and the Analysis of Auxiliaries. Language 53:4, 741-788.

Revero, Maria-Luisa & Douglas C. Walker. 1976. Surface Structure and the Centrality of Syntax. Theoretical Linguistics 3, 99-124.

Ross, John R. 1967. Constraints on Variables in Syntax. Ph.D dissertation, MIT.

Ross, Claudia. 1984. Adverbial Modification in Mandarin. Journal of Chinese Linguistics 12, 207- 234.

Sadock, Jerrold M. 1985. Autolexical Syntax: A Theory of Noun Incorporation and Similar Phenomenon. Natural Language & Linguistic Theory 3:4. 379-439.

Sag, Ivan A., Thomas Wasow, Gerald Gazdar & Steven Weisler. 1985. Coordination and How to Distinguish Categories. Natural Language & Linguistic Theory 3,117-171.

Sapir, Edward. 1921. Language: Introduction to the Study of Speech. New York: Harcourt, Brace, and World.

Scalise, Sergio. 1984. Generative Morphology. Dordrecht: Foris.

Scalise, Sergio. 1988. Inflection and Derivation. Linguistics 26. 561 -581 ycnacnier, raui. lem . Oonsrrainrs on Oiitic ùrder in Tagaiog. UCLA Papers in Syntax a, y b -nb.

Selkirk, Elisaberth. 1982. The Synax of Words. Cambridge: MIT Press.

Sheu, Ying-yu. 1990. Topics on a Categorial Theory of Chinese Syntax. Ph.D dissertation. The Chio State University.

Shih, Chilin. 1986. The Prosodic Domain of the Tone Sandhi in Chinese. Ph.D dissertation. University of California, San Diego.

Sproat, Richard. 1985. On Deriving the Lexicon. Ph.D dissertation, MIT.

Sproat, Richard & Chilin Shih. 1992. Mandarin Morphology Is Not Stratum-Crdered. Paper presented at the LSA Annual Meeting, Phiiadephia, PA. 293

Sun, U-May & Peter Cole. 1991. The Effect of Morphology on Long-Distance reflexives. Journal of Chinese Linguistics 19:1, 42-62.

Tai, Jam es H-Y. 1972. Conjunction Reduction. Ph.D dissertation, Indiana University.

Tai, Jam es H-Y. 1989a. Toward a Cognition-Based Functional Grammar of Chinese. Functionalism and Chinese Grammar, eds. by J. Tai & F. Hsueh, 187-226.

Tai, Jam es H-Y. 1989b. Verb-Copying in Chinese: Syntactic and Semantic Typicality Conditions in Interaction. Paper presented at the LSA Annual Meeting, Washington, DC.

Tegey, Habibullah. 1975. A Study of Pashto Clitics and Implications for Linguistic Theory. University of Illinois Studies in Linguistic Sciences 5:1, 154-196.

Thompson, Sandra A. 1973. Resuitative Verb Compounds in Mandarin Chinese: A Case of Lexical Rules. Language 49:2, 361-379.

Traugott, Elizabeth C. 1977. Pidginization, Creolization, and Language Change. Pidgin and Creole Linguistics, ed. by A. Valdman, 70-98. Bloomington: Indiana University Press.

Ultan, Russell. 1975. Infixes and Their Origins. Linguistic Workshop III, 157-205. Munich: Fink.

Wang, Williams S-Y. 1965. Two Aspect Markers in Mandarin. Language 41:3, 457-470.

Wang, Wiliiams S-Y. 1967. Conjoining and deletion in Mandarin Chinese. Monuments Serica 26, 224-236.

Zhang, Shou-kang. 1957. Lue Tan Hanyu Goucifa [A Brief Consideration of Chinese Word- formation]. Xiandai Hanyu Cankao Ziliao [Reference for Modem Chinese], ed. by Y-S Hu (1981), 241-256. Shanghai: Shanghai Jiaoyu Chubanshe [Shanghai Education Publication. Co.].

Zhang, Zheng-sheng. 1987. Reduplication as a Type of Operation, Chicago Linguistic Society 23:1, 376-388.

Zhu De-xi. 1956. Xiandai Hanyu Xingrongci de Yanjiu [A Study of Adjectives in Modern Chinese]. Yuyan Yanjiu [Studies in Linguistics] 1, 83-112.

^ # A A • A # "4 f a t ^ ^ # i ^ ^ i i wm ^ * I i• • ^«viofvyi mi I i v i u ivi. vyii o i u t o o . w iv w iiiii l y i v i i. i i i u i a i i o w iiivv^ ioity L.11 l y u i o i i u a w iu w .

Zwicky, Arnold M. 1984. Autonomous Components and Limited Interfacing: Phonology-Free Sytax, the Hallean Syllogism, and Their Kin. Chicago Linguistic Society 20, 365-386.

Zwicky, Arnold M. 1985a. Clitics and Particles. Language 61:2, 283-305.

Zwicky, Arnold M. 1985b. Heads. Journal of Linguistics 21, 1-29.

Zwicky, Arnold M. 1987a. Slashes in the Passive. Linguistics 25, 639-669.

Zwicky, Arnold M. 1987b. Suppressing the Z’s. Journal of Linguistics 23:1,133-144.

Zwicky, Arnold, M. 1988. Morphological Rule, Operations and Operation Types. Eastern States Conference on Linguistics 4, 318-334. 294

Zwicky, Arnold M. 1989a. Quicker, more quickly, *quicklier. Yearbook of Morphology 2,139-173.

Zwicky, Arnold, M. 1989b. Idioms and Constructions. Eastern States Conference on Linguistics 5, 547-558.

Zwicky, Arnold M. 1990. Syntactic Words and Morphological Words, Simple and Composite. Yearbook of Morphology 3, 201-216.

Zwicky, Arnold M. 1992a Some Choices in the Theory of Morphology. Formal Grammar: Theory and Implementation, ed. by R. Levine, 327-371. Oxford: Oxford University Press.

Zwicky, Arnold. 1992b. Jottings on Adpositions, Case Inflections, Government, and Agreement. The Joy of Grammar, eds. by D. Brentari, G. Larson & L Macleod, 369-383. Amsterdam/Philadephia: John Benjamins Publications Co.

Zwicky, Arnold M. & Geoffrey Pullum. 1983. Cliticization vs. Inflection: English NT. Language 59:3, 502-513.

Zwicky, Arnold M. & Geoffrey Pullum. 1986. The Principle of Phonology-Free Syntax: Introduction Remarks. The Ohio State University Working Papers in Linguistics 32, 63-91.