ABSTRACT 

The term serial verh construcrion(SV C) refers to a construction in whi ch mo re than one 

verb are not connected by any lex ical device such as conjunction and punctuation, etc . . This 

construction is quite popular in Chinese, Cari bberu1 creoles, West African lang uages , ru1d Dravidiru1 

languages. Structurally, the SVCs may be compounds, c lauses or phrases. The clausal SVCs have 

been attested in all the seriali zed languages. The compound SVCs are reported in Chinese and Edo . 

Only a few instances of phrasal SVCs have been found in Dravidian languages . 

The compound SVCs in Chinese can consi st of two verbs or a verb plus an adj ective. The 

productive compound can only have the following structures: V, ram. + Vin,r,u,u V,ra,.,. + A and 

Yin,r.ms. + A. Among the three compounds, the head of the compound must link to both the ex ternal 

and the internal arguments when it is transitive. The nonhead adjective can link to either the ex ternal 

argument or the internal argument, while the nonhead verb cru1 only link to the internal argument , 

and in most cases, thi s intransitive verb must be unaccusative. Feature Percolation Convention aJld 

the Case requirements generate all the grammatical compound SVCs and rule out a ll the 

ungrammatical ones. These rules also can well account for the differences between the 

VV compound and the VA compound . A three-place verb cannot be used to form verb compounds 

except for the redupli cated construction . This conforms to what the Feature Percolation Convention 

and the Case requ irement predict. 

It is suggested that the syntactic SVCs under di scussion are clausal because of facts 

concerning their binding properti es . This analysis is first proposed by Bickerton & latridou( 1989) in 

their study of the Caribbean creoles. With this analysi s, if the first verb in an SVC is in the matrix 

clause, the second verb is in an adjunct clause attached to the V' position of the matrix clause. Thi s 

analysis obtains support from Chinese. 1n this respect, the only difference between Chinese ru1d the 

Caribbean creoles is that the adjunct clause has two adjoining positions. Furthermore it is proposed 

in this thesis that the adjunct clause is a CP rather than an [P. 



ON SERIAL VERB CONSTRUCTIONS 

by Daoping Wu 

Dissertation submitted to the Faculty of the Graduate School 
of The University of Maryland in partial fulfillment 

of the requirements for the degree of 
Doctor of Philosophy 

1992 

(I 

Advisory Committee: 

Professor Norbert Hornstein 
Professor David Lightfoot 
Associate Professor Stuarts. Sargent 
Assistant Professor Amy Weinberg 
Assistant Professor Juan Uriagereka 



DEDICATION 

To those who are struggling for human rights in my motherland. 

II 



ACKNOWLEDGEMENT 

I am fortunate to have studied in this Depar tment, which not only has a faculty of high 

academic reputation, but also provides me with atmosphere of a happy family. In fact, all faculty 

members of this Department have helped complete this thesis. Therefore I owe so much to so many 

people that this short acknowledgement can hardly be a complete one. 

I should first of all like to thank my thesis supervisor Prof. Norbert Hornstein for his 

considerateness, warmheartedness and encouragement. His exceptional talent to inspire students 

to find and solve problems bas proved invaluable. In fact, our endless discussions in the past two 

years became the basis for this thesis. Without them, it is unimaginable that I would have 

successfully completed this study. 

Special thanks should also be given to Prof. David Lightfoot. Although he shoulders the 

heavy task of heading the Department, be has also helped me in countless ways with this work: 

discussing linguistic issues with me, raising questions, offering suggestions, providing materials, 

and more. 

I am grateful to Prof. Stuart S. Sargent who as a sinologist has often given me fresh ideas 

in our discussions. I am also grateful to Prof. A. Weinberg and Prof. J . Uriagereka for their 

patience in reading this thesis and their enlightening suggestions. I am also indebted to Prof. David 

Lebeaux whose comprehensive knowledge of lexical structure, argument structure and 0 theory 

helped me a great deal in writing this thesis. 

Several professors from Fudan University have also provided me help. Prof. Liejiong Xu's 

study on the binding theory in Chinese is crucial in solving a key problem of this study. Prof. Yum in 

Cheng, in bis visit to Washington, D.C. in 1989 gave me several inspiring ideas about Chinese lexical 

structure. Prof. Xiao Fan provided me with bis detailed descriptive study of the verb compound in 

Chinese, which is one of the bases of this study. 

iii 



I can never forget that it was Prof. J.-R. Vergnaud and Prof. M.-L. Zubizarreta who aroused 

in me an interest to study Chinese morphology. Prof. C.-T. J. Huang kindly offered me materials 

I needed for this study. Dr. S.Iatridou provided me the information about her paper. My deep 

gratitude to all of them. 

I learned a lot from my good friends and classmates in discussing Chinese issues and testing 

linguistic intuitions, they are: Mr. Qimin Chen, Ms. Lilin Zhuang, Mr.Jie Xu, Dr. Yafei Li, Dr. San 

Duanmu, and Dr. Zhiming Bao. 

Deep thanks to Mr. Shuigen Xiao, who used his knowledge and talent to help me prepare 

a readable format of the thesis. He also printed out this thesis for me several times. 

I should not forget to give my deep gratitude to Ms. Kathi Faulkingham, the secretary of 

the Department. She is always ready to help us with everything. This dissertation would have been 

many times more difficult without the help that she provided. 

iv 



List of Symbols and Abbreviations 

A Adjective, Agent 

Acc. Accusative 

adnp Adnominal Participle 

Adj. Adjective 

Arg. Agreement, argument 

Adv. Adverb 

Asp. Aspect 

C Complementizer 

CL. Classifier 

Conj. Conjunction 

CP Complementizer phrase 

CV Compound verb 

DE Chinese particle 

EC Empty category 

ext!. External 

f Feminine 

fut. Future tense 

iff if and only if 

INFL Inflection 

IP Inflection Phrase 

int Interrogative 

inti. Internal 

Loe. Locative 

m. Masculine 

V 



MA Chinese question particle 

MSVC Morphological serial verb construction 

n Neutral 

N Noun 

NCV Non-Productive verb compound 

Neg. Negative 

Norn. Nominative 

NP Noun phrase 

Obj. Object 

p. Page, Plural 

P Patient 

PCV Productive compound verb 

pl Plural(noun) 

PP Prepositional phrase 

Pred. Predicate 

RC resu ltative clause 

s Singular 

SVC Serial verb construction 

SSVC Syntactic serial verb construction 

Subj. Subject 

Top. Topic 

Ten. Tense 

Tpe. Truncated personal ending 

V Verb 

VP Verb phrase 

1 First person 

vi 



2 Second person 

3 Third person 

,# Not equal 

E Member 

i Not Member 

vi i 



The Tao that can be told of is not the Absolute Tao; 

The Names that can be given are not Absolute Names. 

Laotse 

V111 



TABLE OF CONTENTS 

Section 
Abstract 
Dedication 
Acknowledgement 
List of symbols 

Chapter I . In troduction 
1.0. Introduction 
I . I . Introduction to the Framework 

1.1 . I. Bounding T heory 
1. 1.2 . Government Theory 
1. 1.3 . 0-Theory 
1. 1.4. Binding Theory 
1.1 .5 . Case Theory 
1. 1.6. Contro l Theory 

1.2. Introduction to the Chapters 
1.2. 1. Introduction to Chapter 2 
1.2.2. Introduction to Chapter 3 
1.2.3. Introduction to Chapter 4 

Notes 

Chapter 2 . The Distribution and the Properties of Seri al Verb 
Constructions 

2.0 . Introduction 
2 . 1. Distribution of Seri al Verb Construction 

2 . 1. I . Chinese 
2. I . I . I . On PP + V and V + PP 

2 . 1.2. West African Languages 
2. l. 3 . Caribbean Creoles 
2. I .4. Dravidian Languages 
2. 1.5 . SYCs and English 

2.2. Classifications of Serial Verb Construction 
2.2 . I . Classification of SV C Structures 

2 .2. 1. I . SVC with clausal Structure 
2.2 . l. 2. SVC as Verb Compound 
2. 2. 1.3. SVC with Phrasal Structure 

2.2.2. Finite vs. Non-Finite clause 
2.2.3. VY Compound and VA Compound 
2.2 .4. Transitive Verb vs. Intransitive Verb 
2.2. 5. Fixed vs. Free Combinations 
2.2.6. Some Typical SVCs 

2.2.6. I . Take Construction 
2.2.6 .2 . Go Construction 
2.2. 6 .3 . Double Go Construction 
2.2. 6.4. Give Construction 
2.2 .6 .5. Comparative Construction 

2.3 . Properti es of Serial Verb Constructions 
2 .3. 1. Multiple Verbs 
2.3 .2. Lexical Connection 

2.3 .2. 1. Complementi zer-Like Forms in Saramaccan 

I X 

Page 

II 

11 1 

V 

I 

2 
7 
7 
10 
12 
13 
15 
17 
17 
20 
22 
24 

26 

26 
28 
28 
29 
3 1 
32 
34 
36 
37 
37 
38 
40 
41 
42 
42 
44 
45 
49 
49 
50 
52 
53 
56 
59 
59 
60 
64 



2.3.3. Tense or Aspect Markedness 65 
2.3.4. Control in SVC 69 
2.3.5. Extraction of Object of V 2 72 
2.3.6. Reduplication of Verb 72 
2.3.7. SVC and Argument Position 73 

2.4. Defin ition of Serial Verb Construction 76 
2.4.1. Does SVC Exist in English? 78 

2.5. Previous Literature on SVCs: A General Survey 79 
2.5.1. Christaller (1985):A Grammar of the Asante and Fante Language 

Called Tshi (Twi) 79 
2.5.2. Steward(1963):"Some Restrictions on Objects in Twi" 79 
2.5.3. Li & Thompson (1973):"Serial Verb Construction in Mandarin 80 

Chinese:Subordination or Coordination?" 
Mandarin Chinese"; (1974a):"Co-Verb in Mandr in Chinese" 

2.5.4. Bamgbose (1974):"On Serial Verbs and Verbal Status" 82 
2.5.5. Williams, W. (1976)Linguistic Change in the Syntax and 

Sem antics of Serra Leone Krio 
83 

2.5.6. Jansen, Koopman & Muysken (1978):" Serial Verbs in 
the Creole Languages" 84 

2.5.7. Sebba (1978):771e Syntax of Serial Verbs: A n In vestigation 85 
into Serialization in Sranan and Other Languages 

2.5.8. Byrne (1987) :Grammatical Relations in a Radical Creoles:Verb 85 
Complementation in Saramaccan 

2.5.9. Steever (1988): The Serial Verb Formation in Dravidian Languages 86 
2.5.10.Baker (1989):"Object sharing and Projection 86 

in Serial Verb Construction" 
2.5.11.Bickerton & Iatridou (1990):"Serial Verb Constructions 87 

and Empty Categories" 
2.5.12.Li(1990a): "On V-V Compound in Chinese"; 88 

2.6. Conclusion 
Notes 

Chapter 3. Serial Verb Construction as Compound 
3.0. Introduction 
3.1. Compounding: the Major Device of Word-formation 

in Chinese 
3.1.1. Illustration of Compound in Chinese (I) 
3.1.2. Illustration of Compound in Chinese (II) 

3.2. Classification and Properties of Compound Verbs in 
Chinese 
3.2.1. Productive Compound Verb (PCV) vs. Non-Productive 

Compound Verb (NCV) 
3.2.2. Scope of Modification 
3.2.3. Head of CV 

3.2.3.1. Will iams (1978, 1981); Di Sciullo 
& Williams (1987) 

3.2.3.2. Baker (1989) 
3.2.4. Compound or Phrase? 

3.2.4.1. Insertability 

X 

89 
90 

92 
92 
95 

95 
97 
97 

98 

102 
105 
105 

106 
110 
110 



3.2.4.2. Questionability 
3.2.5. Reduplication of V1 
3.2.6. Ba Construction 
3.2.7. Argument-Linking 

3.2.7.1. Cases of Argument-Linking of PCV 
3.2.8. Two Issues Related to Argument-Linking 

3.3. Formation of CV in Chinese 
3.4. Argument-Linking of CVs 

3.4.1. Feature Percolation 
3.4.2. 8-position Licensing 
3.4.3. Case Requirement and Argument-Linking 
3.4.4. 8-Critcrion: A Revisit 
3.4.5. Analyses of Grammatical and Ungrammatical 

PCVs 
3.5. Reduplicated Construction 
3.6. Alternative Solutions 

3.6.1. Bao (1986) 
3.6.2. Li (1990a) 

3.7. Conclusion 
Notes 

Chapter 4. Syntactic Serial Verb Constructions 
4.0. Introduction 

4.0.0. Syntactic Serial Verb Construction (SSVC) 
4.0.1. Some Examples of SSVC 
4.0.2. Some Issues of Clausal SVCs 

4.1. Structure of SSVC 
4.1.1. Phrasal Structure Analyses 

4.1.1.1. Explanation of Schahter(1974) 
4.1.1.2. Analysis of Williams(1976) 
4.1.1.3. Jansen, Koopman & Muysken (1978) 
4.1.1.4. Sebba(1978)'s Explanation 
4.1.1.5. Object-Sharing Approach:Baker(1989) 

4.1.2. Clausal Structure Approach 
4.1.2.1. From Stahlke(1970) to Lu(1977) 
4.1.2.2. Complex Predicate Analysis:Huang(1989) 
4.1.2.3. SVC and the Binding Theory: the Analysis of 

Bickerton & Iatridou(1989) 
4.2. Structure of SSVCs in Chinese 

4.2.1. Binding in Chinese 
4.2.2. Contrastive ziji and Structure of SSVCs in Chinese 
4.2.3. Some Complexity 

4.3. Generation of Clausal SVCs 
4.3.1. Lexical Approach 
4.3.2. Adjunction and Formation of SSVCs in Chinese 

4.4. Conclusion 
Notes 

Chapter 5. Conclusion 

Bibliography 

xi 

112 
116 
118 
120 
122 
125 
129 
131 
131 
134 
138 
143 
146 

155 
160 
160 
161 
163 
165 

168 
168 
168 
170 
171 
178 
179 
179 
181 
183 
186 
191 
198 
198 
199 
209 

212 
212 
216 
219 
222 
223 
224 
229 
229 

231 

235 



CHAPTER 1 

Introduction 

1.0. In troduction 

The Serial verb constmction (SVC) is a linguistic phenomenon which has drawn the attention 

of li nguists for more than a century. In the last two decades, dozens of works on this topic have 

appea red in literature. T hey contain discussions of the distribution, formation, and structure of SVCs 

in several languages. Over tim e these studies have shifted from pure description lo explanation. 

While Christall cr (1875) just wanted to report that in some Afri can languages, there existed SYCs 

which were not attested in Engli sh or French, from Steward (1963) to Li (1990b ), almost all the 

studies of SY Cs pursue explanations of their structure and form ation. Another important change is 

re fl ected in the attitudes to linguisti c universa ls. While trad itional descriptions were only concerned 

wi th the phenomena of one parti cular language, the explanatory studies have tri ed to generalize their 

results, and pursue theories which arc universal to all the human languages. This trend is clearly due 

to the emergence of linguistic theory since 1950, especially the establishment and development of the 

Generative-Transform ational Gramm ar first sketched in Chomsky (1957) . 

The ultim ate goal of Generative-Transform ational G rammar is to discover the psychological 

and the biological basis fo r the language faculty of hum an beings. Along this line, a linguistic theory 

ultimate ly is expected to answer the foll owing question: H ow is it that human beings, whose contacts 

wi th the linguistic data arc so brief, personal and lim ited, a rc nevertheless able to know so much as 

they do know about language? This question is analogical to the question first raised by Bertrand 

Russell about knowledge in general, and cited by Chomsky (1986a). Chomsky termed it as "Plato's 

Problem". A lthough we still have long way to go before we can provide a full account of Plato's 

Problem, explanatory study of a specific natu ra l language can se rve to expand our understanding. 

1 



I hope that this study will make a contribution to the ente rprise, although neither psychological nor 

biological issues a rc discussed directly in this thesis. 

This thesis studies SYCs in several languages, but concentrates mainly on Chinese. The 

SYCs in other languages will also be discussed in pursuing ling uistic universa ls. The discussions 

involve the distribution, properties, origin, structures, formation and constraints, argument-linking, 

reference of empty categories of SYCs, etc., which arc supposed to be controlled by linguistic 

principles. I also want to discuss briefly why such a construction does not appear in languages like 

E nglish. Of course, to discuss so many issues in a short thesis like this risks the danger of a shallow 

and sim plificd account. T o avoid that, J shall focus on some more interesting issues which are 

expected to be more concerned with linguistic universa ls and will leave other issues open. 

In this chapter, a brief introduction of the framework and the contents of the chapters which 

will follows will be given. This study assumes the fr amework of a recent version of Generative­

Transformational Grammar, the Government and Binding Theory proposed by Chomsky (1981b) . 

Although this thesis only discusses one type of construction, the serial verb constmction, the 

explanations of the facts exposed by this construction involve almost all the princi pies of this theory, 

which signifies how complex those linguistic phenomena are. In order to lay the base for the 

substantive discussion, the Government and Binding Theory is introduced in 1.1.. Other proposals 

such as the theory of argument-structure proposed by Williams (1981a), and feature percolation 

suggested by Lieber (1983) will also be mentioned as they are important in what fo!Jows. In section 

2, a short introduction of the contents of the following chapters will be presented. 

1.1. Introduction to the Framework 

Generative - Transformational Grammar is a lingui stic theory of universa l grammar (UG) 

which aims to reveal the principles underlying natural languages. The most rece nt version 

Government and Binding (GB) Theory, is organized as (1): 

2 



(1) D-Structurc (DS) 

I 
J move a 

I 
S-Structure (SS) 

I \ 
PF LF 

DS is the level al which the a ll the lexical items are arranged according to the specifications indicated 

in the lexicon. Applying move-a to DS results in SS. PF, the phonological form, gives a 

phonological interpretation of SS; LF, the logical form, gives a semantic interpretation of SS. 

D S has two sub-components: the lexicon and the categorial component. In the lexicon, all 

the lexical items arc li sted and specified. In this thes is, I claim that punctuation is also listed in the 

lexicon . For deta ils, sec chapter 2. The specification of a lexical item includes its morphological, 

phonological, and syntacti c fea tures. The categorial compone nt consists of the X-bar system, which 

is shown in (2): 

(2)a. X' = X Y "* 

b. X" = Z"* X' 

The Y" in (2a) refers to complement of X, and Z" in (26) is the specifier of X. Y" * and Z"* stand 

for zero or more occurrences of some maximal projection (Chomsky 19866). The re lative orders of 

X and Y" in (2a), X'and Z" in (26) are determined by the paramete rs of a particular language. For 

example, while (2a) in English and Chinese has the form like (2a), in Japanese it is like (3) 

(3) X' = Y" X 

In Chomsky (198Gb), the X-bar system is extended to nonlcxical categories. In 

Chomsky (1981b), a cla use is labe lled as S, and a clause with its complcmcntizer is labelled as S'. 

With the new trea lm c nl, INFL is assumed as the head of a clause, a nd complcmcntizer is conside red 

as the head of S' . T herefore we have schema ( 4) : 

(4)a. S = l" = [NP f1, [yp Y ... ]]] 

b. S'=C" = [ ... [c I"]] 

3 

[Chomsky (19866, p.3)] 



Accord ingly, a language li ke E nglish has the X -bar structure as (5): 

(5) C" 

I \ 
C' 
I \ 

C I" 

I \ 
NP1 I' 

I \ 
I VP 

I \ 
V' 

I \ 
V NP 

TL wi ll be seen in Chapte r 4 that the X-bar structure o f Chinese in substance is the same as (5). 

T here wi ll be a discussio n of the X-bar system in this thesis. Chomsky (1986b) p ro poses 

a mechanism of adjunction. Adjunction, as Chomsky says, is e ithe r the result of the operation of 

move- a , or can be base generated. The form er is illustrated as (6) : 

(6) X 

I \ 
a X 

I \ 
I t 
L_J 

Chomsky (198Gb) stipulates that such an adjunction is only possible to a maximal projection that is 

a nonargume nt. T his requireme nt has been re laxed by F ie ngo & Higginbotham (1981) and Chomsky 

himself. H owever, adjunction as an operation of m ovement has to be co nstra ined by the principles 

and rules governing movement. Fo llowing Lightfoot & Weinberg (1988), I disa llow with VP­

adjunction. For deta ils, sec the late r part of this chapte r and the discussion in chapter 4. 

Base-generated adjunction is important in what fo llows. Fo llowing Bickerton & 

Iatri<lou (1989), it is ass umed that some serial verb constructions conta in adjoined clauses. Such a 

treatment has its adva ntages: It simplifies X-bar syntax and distinguishes the complement and the 

adjunct. We wi ll discuss this in greate r detail be low. 

DS is inte rpreted as a pure re presentation of them atic re lations. D S, especially the lexicon, 

4 



is the base of a grammar in the following sense: the specification of the lexical items must be 

represented at other levels of the syntax. This principle is stated in (7): 

(7) Projection Principle 

R epresentations at each syntactic level (i.e., LF & D- & S- structure) are projected from 

the lexicon, in that they observe the subcategorization properties of lexical items. 

[Chomsky (1981b, p.29)J 

For example, the verb hit is specified in the lexicon as assigning a 0-role AGENT(A) to its subject, 

and REClPIENT(R) to its object. According to (7), such a relationship must be realized in each 

of the levels of the syntax. In the sentence (8) 

(8) I d idn't want [ to hit him] 

the verb hit does not have an overt subject. If (7) is correct, it is expected that it must have a non­

overt subject, which is termed as empty categ01y indicated by e in (9): 

(9) I didn 't want [ e to hit him] 

Suppose we construct a passive sentence with verb hit: 

(10) I was hit by him. 

According to the Projection Principle and the specification of the verb, the DS of (10) is 

(11) [ was hit I by him l 

In the mapping from the DS to SS, the transformation "move a " applies. Move-a leaves traces (one 

type of empty category). For example, to apply move-a to (11) resu lts in SS (12): 

(12) f I was hi t t by him] 

PF is the phonological interpretation of a str ing, and LF is its semantic interpretation. Some 

phonological rules, such as deletion rules, and stylistic transformation ru les, are applied at the PF, 

while semantic ru les such as Wh-raising, Quanti fier Raisi ng in the sense of May (1977) are applied 

al LF. One of the generally accepted explanations of Chinese says that in Chinese, there is no 

syntacti c Wh-movement, and Wh-movement applies at LF(Huang 1982; Chomsky 19866). For 

example, sentence (13) has different representations as shown in (14): 

5 



(13) Z hangsan 1 xuan le shenme ke? 

Zhangsan select Asp. what course 

"What courses have Z hangsan selected?" 

(14)a. DS 

fer [1p Z hangsan [yp xuan le shenm e ke ]J] 

b. SS 

[0 , lIP Z hangsan [y p xuan le shenme ke]ll 

c. LF 

fer shenme ke [1p Zhangsan [ vr xuan le t ll] 

O ne of the recent developments of GB proposes some constraints on PF and LF. Aoun et 

al (1987) requires an empty category to be lexically governed at PF, while bound at LF in the sense 

of generalized binding proposed by Aoun (1986) . With the theory of Chomsky (1981b), PF and LF 

do not interact with each other. This development is not directly re lated to the discussion of this 

thesis, but the principles under Aoun ct al's proposal --- the definition of governm ent and the 

theori es of ECP and Bounding, are adopted here, and are the basis of some of the discussions. For 

details, see Aoun et al (1987), Lightfoot & Weinberg (1988) and C hapter 4 of this thesis. 

Chomsky (1981b) proposes that Universal G ramm ar consists of several interacting 

subsystems, which arc both the va rious subcomponents of the rule system of gramm ar as stated in 

(1) and subsystem of principles. The subsystems of principles include the foll owing: 

(15) a. bounding theory 

b. government theory 

c. 0-theory 

d. binding theory 

c. Case theory 

f. cont rol theory 

In the following sections, a brief introduction to these theories will be given. Focus will be made on 

6 



the theories which are most relevant to the investiga tion of SVCs such as governme nt theory, Case 

theory, and bindi ng theory. 

1.1.1 . Bounding Theory 

Bounding theory poses locality conditions on cert ain processes and related items, especially 

the locality condition on movement. Such locality conditions may differ at different levels of a 

gramm ar and in different languages. Chomsky (19816) proposes that NP and S are two bounding 

nodes fo r syntactic movement in E nglish, and any moveme nt in the syntax cannot cross more than 

one bounding node . In Chomsky (19866), bounding theory and government theory are generalized 

into one principle: barrier theory,2 which stipulates that movement is optimal when it crosses no 

barriers, and it degrades sharply with crossing of more than one barrier; a barrier will block the 

governm ent relationship among constituents. 

As slated in the later part of this thesis, the theory of ba rriers still has many difficulties in 

explaining the behaviors of movement and governm ent ( cf. Lightfoot & Weinberg (1988)), the theory 

suggested in Chomsky (19816) is adopted which sta tes that moveme nt must observe the Subj acency 

Condition and thus cannot cross more than one bounding node. 

1.1.2. Government Theory 

Government is the relation between the head of a construction and the categories dependent 

on it. Other theories, such as Case theory and 8-theory exploit this relation. There are several 

differe nt defin itions of governm ent. In this paper, three of them will be discussed: the definitions 

given by Chomsky (19816), (19866) and by Be lletti & Rizzi (1982). 

Chomsky (19816) defines government in the foll owing way: 

(16) a: governs P if and only if 

(i) a: = XO 

(ii) a: c- commands P and if y c-comm ands p then y either c-commands a: or is c­

comm anded by p. 

[Chomsky (19816, p.163)] 

7 



a is a L±.N, ...±. V] category. C-command is a relation between constituents defined by (17): 

(17) C- command 

A c-commands B if and on ly if the first branching node dominating A also dominates 

B, a nd A docs not itself dominate B. 

[van Ricmsdijk & Williams (1986, p.142)] 

Accordingly, we have fo llowing examples of government: 

(18)a. [yp V NP (NP')] 

b. [pp P NP] 

c. (ci, for f1p NP1 t0 [yp NP2 ]]] 

d. [x p X [1p NP to VPI] 

e. [yp V [c p COMP [II' NP INFL YPJ]] 

f. [yp V INPI NP2'S N]] 

g. [y p V [J\p A (of) NP]] 

h. lvr V [pp P NP ]J3 

[Chomsky (1981b, p.162)] 

These are the basic cases of government. Chomsky explains that in a ., V governs NP and NP'. In 

b., P governs NP. In c., the prepositional complementizer for governs NP1 but not NP2 which is 

governed by V. In cl ., X governs NP. In e., V does not govern NP, but CP; if INFL is tensed, it 

governs NP. In f., V gove rns NP1 but not NP2. In g., V does not govern NP; NP is governed by A. 

In h ., NP is governed by P but not by V; PP is governed by V (Chomsky (1981b, p.162). 

Bcllctti & Rizzi (1982) suggest another definition of government, which enables a head not 

only to govern its complement, but al so the head of it comple ment: 

(19) Governm e nt 

a (X0) governs P iff all maximal projections dom inate a also dominate P, and a is also 

dominated either by all maximal projections dom inating p or by all maximal projections 

8 



dominating the maximal projection of p. 

With this definition, in configuration (20), a governs P: 

(20) X" 

I \ 
a P" 

\ 
p 

We will see in Chapter 3 that definition (19) is crucial in explaining the properties of verb 

compounds in Chinese. Therefore this definition has been adopted throughout this thesis. 

Along the line suggested in the Government and Binding Theory, movement is not 

constrained by any specific ru le. Instead, it is constrained by principles such as the Adjacency 

Condition. One of th e wcll-formedncss condition is the E mpty Category Principle (ECP), which 

constrains the distribution of empty categories(ECs): 

(21) Empty Category Principle 

[e] m ust be properly governed. 

ECP constrains the ecs left by movement, thus restricting the app lication of movement. Proper 

government is defined in (22) 

(22) a properly governs p if and only if a. governs p and a. I' AGR. 

[Chomsky (1981b, p.250)1 

However, this condition is changed in Chomsky (19866 ). In the framework of barriers, the definition 

of proper government is divided into two parts: 

(23) a properly governs P iff a 0-governs or antecedent-governs p. 

[Chomsky (1986b, p.17)] 

0-governmcnt is defi ned as follows: 

(24) a 0-governs P iff a. is a ze ro-level category that 0-marks p, and a., p arc sisters. 

[Chomsky (19866, p.15)] 

Antecedent government is the relation between clements in a link of a chain, in which the govcrner 

9 



is conindcxcd with lhc govcrncc. Jn other words, there is no barrier between the governer and the 

govcrnee which blocks the relation between the two clements. 

Chomsky claims that the lwo parls of proper governme nt work disjunctively, meaning 

fulfillment of any parl of the condition yields a well-formed string. Obviously, the first part of 

defin ition (23) is en tailed by definition (19) . The second parl of definition (23) accounts for the 

adjunct movement: 

(25)a. the meal is IAP cooked [how wellJJ 

b. (how well] is the meat cooked t 

[Chomsky (198Gb, p.79)] 

(25a) is the DS of (25b). In (25b), the adjunct how well is moved into CO MP position. The trace 

of the moved conslitucnl is not governed by a x0 element since no constitue nt 0-marks il. It is 

governed by ils antecedent how well through coindexing. 

The revision of the definition of proper government brings extensive consequences. Based 

on this notion, Chomsky (1986b) suggests a kind of new operatio n of move- a: VP adjunction. 

Baker(1988) develops a theo ry of incorporation which is a type of head moveme nt constrained by 

antecedent government. 

However, it has been found lhat the antecedent-government is not sufficient to guarantee 

the well-formedness o f movement. As pointed oul in Lightfoot & Weinberg (1988), in the fo llowing 

sentence, the trace of the moved element is governed by ils antecedent, bul the sentence proper is 

still ungrammatica l: 

(26) *Whati did John ta lk yesterday [pp about e i ]? 

It means that the theory of barrie rs still has serious problems. Therefore, in this thesis, I will not 

employ the new definition of proper government. For the de tails of discuss ion, see Chapter 4. 

It is suggested that the ECP is one of the locality conditions which constrain the well­

formedness of a sentence. It is no longer a condition applied al SS or LF as suggested by Chomsky 

(19816), bul at PF. For lhe deta ils, see Aoun ct al (1987). 

10 



1.1.3. 0-Theory 

0-theory, as stated by Chomsky(1981b), is concerned wi th the assignment of thematic roles 

such as agent-of-action, goal-of-action, etc., lo an NP in the argument position. The central notion 

of the 0-theory is the 0-Crilerion: 

(27) 0-Criterion 

Each argument is assigned one and only one 0-role and each 0-role is assigned lo one 

and only one argument. 

[Chomsky (19816, p.36) 

0-assignment is constrained by government. A position to which 0-role is assigned is a 8-position: 

(28)a. [they] persuade [John][that [he] should lcaveJ 

b. jwe] hold jthat [ these truths] are [self-evident]! 

c. [we] hold [[these truths] are [self-evident]] 

d. it is held (that (these truths ] are (self-evident]] 

e. these truths are held [[t] to be [self-evident]] 

f. (these truths] are [self-evident] 

g. jwe] put !the books] [on [the table]J 

h. the books were put [t] [on [the table]! 

i. advantage was taken t of [Bill] 

[Chomsky (1981 b, p.36)] 

All the bracketed positions are 0-positions. 

In regard to 0-theory, there are several issues discussed in this thesis. First, some other 

studies and this study have found, in some circumstances, one 0-position may receive more than one 

0-role. Several revisions of the 0-Criterion have been suggested, and they are all the same in 

substance. This paper adopts a more comprehensive well-formedness condition proposed by 

Chomsky(1986a), the Full Interpretation Principle, which allows one 0-position to receive more than 

one 0-role. Detailed discussion of this theory will be given in Chapter 3. 

11 



This paper also proposes that the Full Interpretation Principle is not on ly a well-formedness 

condition of phrases, but also constrains the formations of compounds. In chapter 3, it will be 

argued that some combinations of verbs are not grammatical because they cannot meet the 

requirement of the Full Interpretation Principle in the syntax. 

1.1.4. Binding Theory 

Binding theory is concerned with relations of NPs to poss ible anteceden ts. In Chomsky 

(1.981b), it is basica lly the relation that holds for elements in argument positions (A-positions). In 

Chomsky (1981b), binding theory is stated as follows: 

(29) Binding Theory 

a. An anaphor is bound in its governing category. 

b. A pronominal is free in its governing category. 

c. An R-cxpression is free. 

(30) Governing Category 

a is the govern ing category for 13 if and only if a is the minimal category containing and 

a governor of 13, where a = NP or S. 

[Chomsky (1981.b, p.188)] 

There arc several important modifications of the binding theory since Chomsky (1981b) . Aoun(l981 ) 

extends the binding relation between A-positions to A'-position (non-a rgument positions), thus 

proposes his theory of general ized binding: 

(31) Genera lized Binding Theory 

a. An anaphor must be X-bound in its governing category. 

b. A pronominal must be X-free in its governing category. 

c. An R -express ion must be A-free. 

X = A or A'. 

The governing category for Chinese is defined as follows: 

(32) Governing Category for Chinese 

12 



a. a is a governing category fo r an anaphor iff a is the minimal ca tegory containing 

a SU BJ ECT accessible fo r this anaphor. 

b. a is a govern ing category fo r a pronom inal iff a is the minimal category containing 

this pronominal. 

SUBJ ECT is the most salient nominal constituent in a maxim al category. It can be the subj ect or 

!NFL of a clause, the specifier of the head of an NP. T he notion accessible is defined in (33): 

(33) a is accessible to p if and only if P is in the c-comm anded domain of a and assignment 

to p of the index of a would not violate (34) 

(34) * [y ... o ... ], where y and o bear the same index. 

[Chomsky (19816, p.211)] 

In Aoun et al, generalized binding is one of the locality conditions exercised at LF. 

This paper docs not discuss the binding theory directly, and the difference between different 

versions of binding theory docs not affect the discussion and conclusion in the re levant chapters. 

Therefore I will not go into details of the binding theory in this thesis. H owever, we will see that 

the binding theory docs play an important role in the analyses of the structure of SVCs. For the 

details, see Chapter 4. 

1.1.5. Case Theory 

Case theory deals with the ass ignment of abstract Case to an NP and its morphological 

rea lization. Government and Binding theory claims that any NP in a phrase must have an abstract 

Case, no matter whether the Case is morphologically rea lized or not. 

As regards Case theory, although there are several controversies which we will not go into, 

generally linguists agree on the fo llowing points: 

First, any lexical argum ent NP must have a Case. This principle can be slated as follows: 

(35) *f NP a] if a has no Case and a contains a phonetic matrix or is a variable. 

For example, the foll owing sentence is ungrammatical: 

(36) *who docs it seem [ t to see Mary] 

13 



fo r who in (36) cannot inherit Case from its t race, since tis not assigned a Case from the verb seem. 

Second, Case assignment can only be do ne under govern ment. In othe r words, only if a 

Case assigner governs a Case ass ignee can the fo rmer assign a Case to the latter. 

T hi rd, there is an adj acency condition for Case assignm ent in some languages , e.g. E nglish 

and Chinese. In other words, the Case assignee must be adjacent to the Case ass igner. We have a 

great deal of evidence from English and Chinese to support this condition. For example, in both 

English and Chinese, there must not be anything intervening between the ve rb and its complement, 

because the intervening clement will block Case assignm ent. 

On the issue of adjacency for Case ass ignment, there is a seeming "exception" --- th ree-place 

verbs which assign Case Lo each of its complements. Assignment of Case to the second complement 

seems to vio late the adj acency condition since one NP has to be crossed. To pursue a unified 

solu tion to this problem, Stowcll (l 981) proposes that after a three-place verb assigns a Case lo the 

first complement, it incorporated with the NP to form a new V, and then the new V assigns Case 

to the second NP under adj acency. We wi ll see in the Chapter 3 that there is sup porting evidence 

fo r this so lution in Chinese. Fourth, it has been accept ed that in a chain, only one Case is ass igned, 

and the unaccusative verb such as /ai(come), qu(go), zou (walk) , pao(run) in C hinese, is the optional 

Case assigner. If an unaccusative verb does not assign a Case to its complement, the 

complement NP lacks Case, and therefore NP movement applies so that it can meet the Case 

requirement. 

This thesis docs not study Case theory. H owever, from the investiga tion of SYCs, positive 

evidence has been fo und Lo support this theory. Chinese, and many other languages mentioned in 

this thesis, such as Sranan, Yoruba, etc., do not have concrete cases. But it has been attested that 

in fo rm ing an SVC, the Case requirement must be met, otherwise the SVC wi ll be ungrammatica l. 

We will sec in Chapter 3 that this is why only an unaccusative verb can act as Y2 in form ing a VY 

compound, and why three-place verbs in Chinese cannot form verb compounds. ln fact, the Case 

req uirement is one of the two morpho-syntactic conditions of verb compound fo rmation in Chinese. 

14 



Koopman(1984), Travis(J984) and Li(1985) propose that the direction of Case 

ass ignment and 0-assignment in some languages such as Chinese may be different. Huang(1989) 

genera lly accepts this assumption . However, up until now, we have found no evidence to support 

this hypothesis. Because this is not only indirectly related to the topic of this thesis, only a few points 

wi ll be discussed mention here: Chinese is not an SO Y language as some linguists cla im, and the 

function of ba construction is merely to raise the object prcverbally. This will be discussed in 

Chapter 4. lf these views arc correct, we need not assume different directions for Case and 0-

assignment in C hinese. 

1.1.6.Control Theory 

The last subsystem which will be discussed is control theory. Control determines the 

potential for reference of the clement PRO. The Government and Binding theory divides empty 

ca tegories into four types: 

1. NP trace: the trace left by NP movement, e.g.: 

(37) [1p this booki was published ti by the Academic Press] 

NP this book is moved from V complement position, leaving the trace t. 

2. Wh-tracc, also called variable because it is always bound by an Wh-clement in A'­

position : 

(38) f Cl' what1 I have you seen ti] 

3. PRO: a pronominal acting as the subject of a non-finite clause: 

(39) [1p he used a knife fir PRO to cut the bread]] 

4. pro: a pronominal whose contents is determined by INFL: 

( 40) liP pro arriva Giovanni] --- Ita lian 

arrive Giovanni 

"Giovanni is coming." 

In (40), the inflection of the verb arriva (arrive) determines the number, person, etc. of the 

15 



subject. Control theory is concerned with the reference of PRO. 

O ne type of SVC is analyzed as a type of clause, and empty categories arc therefore 

involved. In Bickerton & Iatridou (1989), the reference of the empty categories, incl uding those in 

the finite and the non-finite clauses arc treated under control theory. It has been reported that in 

most cases, the pronom inal in a clausal SVC is controlled by the subject of the matrix clause, which 

is termed Subject Cont rol in cont rol theory. 

However, in determining the nature of the pronominal in SVCs in Chinese, there is a 

difficulty. Since Chinese sen tences does not have tense and subj ect-predicate agreement, it is hard 

Lo determine whether a clause is finite or not. Consequently, we do not know if the pronominal in 

the subject position of a clause is PRO or pro. There have been several studies on the issue of 

fin iteness of clauses in Chinese. Chinese docs not have tense. But it docs have aspect, and aspect 

plays an important role in the well-formedness of clauses (that is, some constructions and operations 

are limited for non-finite clause. For the details, see the relevant sections of Chapter 4). 

H owcvcr,the properties and features of subjects are never determined by the aspect of the verb. 

Several proposals have been advanced to aviod the embarrassment. Since this paper does not focus 

on this issue, this problem is left open for further investigation. Following Bickerton & 

[atridou(1990), a compromise stance is taken in this thesis: whether a clause has aspect or not, its 

empty subject is studied along the line of control theory, since its properties and features cannot be 

dete rmined by the aspect. 

Before concluding the discussion of the framework of the theory, two other theories should 

also be mentioned. One of them is the theory of argument structure proposed by Williams(1981), 

which is defined as a labelled list of the arguments that a lexical item can have. H erc labelling 

means thematic relations. For example, the verb chase has the following argument structure in the 

lexicon: 

(41) chase (Agent, Theme) 

which means that the verb chase can take two arguments, and its external a rgument (subject) will 

16 



receive a 0 -ro le Agent, whi le the internal argument will receive a 0-role Theme. In this thesis, 

a rgument structure is considered to be one of the specifications of verbs, or in other words, the 

features of verbs. It is not ass umed that there is an independent syntactic structure to specify this 

this property of verbs. 

T he o ther theory is the Feature Percolation suggested by Lieber(1983). She claims that in 

form ing compounds in English, stems and affixes percola te their fea tures upward according to some 

rules: 

(42) V 

I \ 
A \ 

I \ 
[modern] v izeJv 

in ( 42), it is sti pulated that a ll the features of stem morpheme percolate Lo the first non branching 

node dominating that morpheme, and that all the features of an affix morpheme percolate to the first 

branching node domina ting that morpheme. Therefore node A receives all the features from 

m odem , while V rece ives a ll the feature from -ize . This expla ins why the word should be labelled 

[ + V, -NJ in the lexicon. This has proved a useful theory in explai ning some inte resting phenomena 

in verb compound formation in Chinese: In Chinese, to construct a VA compound is much easier 

than Lo construct a VY compound, and the rules of argume nt-linking of the two compound a re quite 

different. These arc accounted for by different feature percolation rules of the two categories. For 

the details, see Chapter 3. 

1.2.lntroduction to the Chapters 

In addition to this introduction, this thesis consists of four o ther chapte rs. In the following 

sections, I wi ll survey each of them. 

1.2.1. Introduction to Chapter 2 

Chapte r 2 dea ls with the distribution, classification and properties of the serial verb 

constmctions (SYCs). SVCs have been attested in Chinese, West African languages such as Yoruba, 

and Edo. Caribbean creoles as Sranan, Seselwa, etc., and Dravidian languages. A ft e r giving a short 

17 



introd uction lo each of the languages or language groups, the SYCs in these languages are presented 

with a variety of types and structures. F urthermore, SYCs in these languages are classi fi ed by 

several cri te ri a as follows: It has been foun d that SVCs in differe nt language may have different 

structures. We have fo und three types of structures: clausal structure, phrasa l structure and 

compound. Examples of all th ree types of SVCs will be given in Chapter 2. As regards SYCs of 

clausal structures, the non-matrix clause in an SVC may be either finit e or non-finit e; if the SVCs 

arc of compound structure, they may consist of two verbs or one verb plus an adjective. As regards 

the transitivity of verbs which form SYCs, they can be either transitive or intransitive. The more 

important criterion to class ify SVCs is the fixity of the combination. In fact, all the SYCs can be 

divided into two types: the fixed combination and the free combination. The fi xed combination 

means that if an SVC is of clausal or phrasal structure, certain verbs appear aga in and again in the 

same position; if it is a compound, it is not productive. 

M any examples wi ll be shown lo illustrate the stability of the combination. While the 

structu ra l analyses and the propert ies of clausal and phrasa l SVCs do not depend on the stability 

of the combination, the properties of the productive compound and the non-productive compound 

have fundamental differences. For the productive compound, change of its c lements is always 

possible if no linguisti c princip les or rules arc violated, while for the non-productive form , such 

change is impossible. The meaning of the non-productive form is always compositional, while that 

of the non-productive form is not always compositional. There arc more possible ways of a rgument­

linking for the non-productive form s, wh ile those for the productive form s are more restricted. It is 

stated that because the non-productive compounds are more strictly constrained, thus more 

interesting, this study therefo re wi ll concentrate on this type of verb com pound . It has been found 

that there exist some typical SYCs in the sense that the fr equency of some verbs appearing is much 

higher than that of other verbs in SYCs, and the high frequency of use of such verbs is found in most 

serialized languages. In section 2.2.6 of Chapter 2, four typical SVCs are presented. They are the 

18 



take, go, double go, and give constructions. In this section, another construction, the comparative 

construction is also discussed. Evidence will be given that such constructions cannot be analyzed as 

SVCs in Chinese. Contrary to the genuine SVCs, the object of the comparative construction is not 

extractable, which shows it is a PP construction rather than a verb construction. 

Section 3 of this chapter deals with the properties of SVCs. The first one mentioned in this 

section is that any SVC must consist of more than one verb, which seems straightforward, but still 

worth discussing. Some of the studies define other marginal categories, such as verb-like, coverb, but 

fail to define by what standard a marginal category can act as the constituent of SVC. In this section, 

I claim that a verb- like can become a part of SVC only if it has all the functions a verb has in that 

language (not necessarily vice versa). This thesis also proposes the notion lexical connection which 

means two or more constituents are connected by any lexical device, including conjunctions, 

punctuation, etc .. Two verbs or one verb and one verb-like clement is considered as serialized if and 

only if they are not lexically connected. By this definition, we can exclude some sentences like 

(43): 

( 43) John picked up his book, read a few lines and threw it away. 

Several other properties are revealed in this section. It has been found that in most cases the object 

of the V2 is always extractable, which indicates that two parts of an SVC can hardly be in coordinate 

relation. Neither V 1 nor V 2 can be in the argument position, which not only excludes V + object 

clause construction from the discussion, but also means that for the clausal SVCs, its non-matrix 

clause must not be subcategorized by the matrix verb, thus it must be in adjunct position. This fact 

is helpful fo r the structural analyses of SVCs. 

Based on the properties discussed in section 3, a formal dclinition of SVCs is worked out. 

Here three subdcfinitions arc involved: the definition of SYCs, the definition of verb-like and the 

definition of lexical connection. With these definitions, the scope of the study of this paper is 

clarified. 

There is a short discuss ion on whether English has SVCs. It is claimed that although some 

19 



constructions (such as the make constmctio11) is somewhat simi lar lo SVCs, they do not have all the 

diagnostics of SVCs in other languages. Therefore by any standard English is not a serialized 

language. 

Section 5 of Chapter 2 is a general survey of previous literature, which includes studies of 

SVCs from Christallcr (1875) lo Li(1990a). More than a dozen books or papers arc mentioned. 

However, a lthough r give short comments to each of them in this section, no detail is gone into until 

the relevant sections of later chapters. 

l.2.2. Introduction to Chapter 3 

Chapter 3 discusses one type of the SVC---productive compound vcrbs(PCV) in Chinese. 

To show the importance of compounding in Chinese, I present some examples in section 1, 

illustrating that in Chinese, words of all the categories can be formed by compounding, and the 

structures of compounds arc parallel to the structures of phrases. 

Section 2 deals with classification and properties of compound verbs in Chinese. In this 

section, the fundam ental distinction between the productive compound verbs and the non-productive 

compound vcrbs(NCV) is discussed, claiming that some structu ra l "exceptions"of the compounds 

mentioned in other studies actually arc non-productive in nature. In discussing the scope of 

modification of PCVs, one properly of the PCV is mentioned: if the compound has a modifier, it 

either modi1ies the whole compou nd, or the head of it. The non-head of the compound is never 

solely modified. To explain this, it is assumed that the scope of modification is constrained by 

government: a constituent is modified by another if it is governed by the mod ifier. The de1inition 

of governm ent given by Bcllclli & Rizzi (1982) is adopted which enable an x0 governs not only the 

maximal projection, but the head of it as well. 

This raises question of the head condition: what is the head of a compound? On this issue, 

Williams (1978, 1981), Di Sciullo & Williams (1987) and Baker(1989) make different proposals. 

After showing the inadequacy of their proposals, I give new dc1initions of the H ead Licensing 

Condition(HLC) and Projection Licensing Condition(PLC), which results in the following points: 

20 



1. A maxim al projection may have the same bar level as its head, which accounts for the fact 

of verb compound, in which both the maximal projection and the head a re of x0-bar. 

2. T he category with minimal bar and identical fea tures with the maxim al projection is the 

head. 

3. If there is no category which has identical features with the maximal projection, the 

category of minimal bar and with identical categorial fea ture is the head. It expla ins why 

in Chinese, an in transitive verb and an adjective can constitute a transitive verb compound. 

On the status of the phenomena discussed in th is chapter, one may argue that they are in 

fact verb phrases rather than verb compounds. Some tests arc used lo show they cannot be phrases: 

In Ch inese, a phrase can be inserted by other constituents, the constructions under discussion cannot; 

if two phrases are coordinate, each of them can be questioned, but no part of the construction under 

discussion can; if two verb phrases are coordinate, we can reduplicate a verb in one of the phrases, 

but fo r the VY construction under discussion, no verb can be reduplicated. In conclusion, the 

construction under discuss ion docs not have the typical properties of verb ph rases. 

In discussing argument-linking of PCVs, some examples are given to show that if a PCV 

consists of two verbs, V 2 can only link to the internal argum ent; if it consists of one verb and one 

adjective, the adjective can link to either internal or external argument. Furthermore, three-place 

verbs cannot be part of compounds, and a compound cannot consist of more than two verbs. These 

phenomena raise severa l questions about well-form edness conditions and 0-theory. 

Section 3 deals with fo rmation of CVs in Chinese, and discusses the conditions fo r 

constructing verb compounds with respect to selectional fea tures and subcatcgoriza tions. Section 

4 d iscusses further conditions of well-formedness of PCVs. It is claimed that a PCV is constrained 

not only in the lexicon, but in syntax as well. A PCV must link all the 0-roles of its argument 

structure to the argument in a proper way, otherwise the form must be ruled out. 

To explain the syntacti c constra ints of PCVs, I divide the process of argument-linking into 

two parts: argument position li censing and argument-linking. A rgument position li censing is done 

21 



by the Feature Percolation Convention in the sense of Lieber(1983), which stipulates in a PCV, the 

non-head V can only percolate its fea tures upward to v 0, thus can only license an internal argument 

posit ion, while a non-head adj ective can percolate its fea tures up to VP, thus can li cense ei ther the 

internal or the external argum ent position. It explains why in Chinese, Vi nt rans·+ V, Vt ra ns. + Vt ra ns. 

combinations are ungramm atical. 

It is also claimed in this section that a compound must meet the Case requirement in the 

syntax ---- the arguments it links to must have proper Case assigned by it --- ,otherwise it is 

ungrammatical. The evidence to support it is the fact that generally only unaccusative verbs can act 

as V 2 in a com pound, because using an accusative assigning verb will cause Case conflict. This 

explains why three-place verbs cannot be used to construct PCVs: in such cases, because the second 

constituent of the com pound blocks the incorporation of the fi rs t constituent of the com pound and 

the fir st internal argument, the second internal argument cannot receive a Case since the Case 

assignment must observe an adjacency condition. To explain that in some cases, a compound assigns 

more than one 0-role lo an argument, the F ull ln le rpretalion Principle, a more general princip le 

than the 0-cri lerion, is adopted by this study. This principle allows one 0-position receives more 

than one 0-role, or in other words, a compound 0-role. 

In analyses of all the grammatical and the ungrammatica l compounds in section 3.4.6, I 

ill ustra te that all the gramm atical combinations can be generated by the Feature Percolation 

Convention, while all the ungrammatical combinations can be excluded by the linking convention and 

Case theory. 

Section 5 discusses alternative so lutions, especia lly that proposed by Li(199Oa). By 

comparing the so lution proposed in this thes is and Li's, it is shown that Li's solution cannot exclude 

all the ungramm atica l combinations, e.g., vinl rans. + vlrans. and vl ra ns. + vl ra ns: Therefore the 

theory of compound formation worked out by this study is a more constrained than Li(199Oa). 

1.3. Introduction to Chapter 4 

Chapter 4 discusses syntactic serial verb construction(SSVC), mainly clausal SVCs. 

22 



Clausal SYCs have been altcstcd in all the fo ur languages and language groups mentioned 

in th is thesis. There arc several issues concerning the study of this type of SYCs. Section 1 of this 

chapter dea ls with the structure of SSYCs. Generally, there a re two types of analyses of this 

construction. Some linguists prefer all the SSYCs to have clausal structure. Some of the ir analyses 

arc discussed in this section. Schachter (1974) assumes that SSVCs are of coordinate verb phrase 

structure. H e provides evidence to argue against the transformational approach, but fa ils lo 

demonstrate that SSVCs musl have coordinate phrase structure. Williams(1976) employes the 

notion of thematic relations lo explain the relations between verbs in SSYCs. In his detailed 

discussion on Sranan, Scbba docs not agree with the theory that all the SSVCs must have one 

possible structure. He prefers some of them to be of coordinate ph rase structure, while others are 

of subordinate structure. In discussing the object sharing approach given by Bakcr(1989), it is pointed 

out that the approach needs radical revision of the X-bar theory proposed by Chomsky (1981b ) . The 

trouble is that his revision cannot be supported by conslruclions other than a few sentences he cites, 

and object sharing is not always necessary for SSVCs. 

Some other linguists suggest clausal structure for SS VCs. Among them, Stahlkc(1970), 

Bamgbo§e (1974) adopt the Generative Semantics and Standard Theory of the G enerative -

T ransform ational Grammar lo explain SS VCs, the fr amework of which arc generally conside red 

inappropriate. Li & Thompson (1974) give lwo possib le clausal slruclurcs for SSYCs, but the two 

structures do not conform to either the linguistic principles or lhc general rules of Chinese gramm ar. 

Bickerton & lalridou (1989) give very convincing evidence thal SSYCs under discussion must 

be of clausal structure with rcspccl to binding theory, which is based on the following observation: 

y
2 

of SSVCs in some languages can have an anaphor as object which cannot be coindexed with the 

object of V 
1

. They explain that this phenomenon means there must be a governing category which 

docs not include V 1 and its objccl. Such a governing category cannot be VP, but IP or CP. 

It also has been found lhat in Chinese, only an anaphor can be inserted before v2, which 

implies that there is a prcvcrbal empty subjecl. It indicates that V2 musl be in a non-maximal 

23 



clause. However I do not agree with Bickerton & lalridou (1989) that the clause is an IP. 

Considering the extractability of the object of V 2, and the invalidity of VP-ad junction outlined by 

Chomsky (198Gb), it is suggested that the clause is a CP. I also claim that unlike SSVCs in the 

Caribbean creoles, SSYCs in Chinese have three subtypes of structures. 

Section 3 of this chapter deals with generation of clausal SVCs. Several proposals have been 

advanced in the previous literature. For example, Schach ter (1974) , Williams (1976), and Baker 

(1989) suggest that SSYCs arc generated by phrase structure rules, Stewart (1963) suggests a 

transformational approach, Dechaine(1986) and Lefcbvre(1988) agree that SSVCs are formed in the 

lexicon a discontinuous complex predicates. In this section, it is suggested that all these proposals 

wi ll have difficulties in some respects. For instance, the approach that SSYCs a re generated by 

phrase structure rules always needs specific rules for SSVCs, which not only loses the unification of 

the theory, but oft en creates ungrammatical sentences as we ll ; the lexical approach will complicate 

the lexicon in a way which can hardly be accepted by linguists, and we do not have any motivation 

to assume such a complex theory; the transformational approach will not be considered as a 

appropriate model beca use recent theory no longer contains any specific transformation rules. 

On this issue, the approach proposed by Bickerton & Iatridou (1989) is partially adopted. 

SSVCs are generated by basc-adjunction of a clause to V', which is based on the theory that a 

controller must m-command its controllce. While in other languages, the non-matrix clause can only 

adjoin to a poslverbal position, in Chinese, both prevcrbal and poslverbal positions arc the possible 

landing sites for adjunction. 

24 



Notes : 

1 . For unknown reason , in the di scussion of Chinese, names Zhangsan and Lisi are often 

employed in literature. However, these two names, literally mean "Zhangsan the third" "Li the 

fo urth" are seldom found in modern Chinese society. Jn this thesis, I, following others, use these 

names just for releasing the burden of learning new Chinese names, which are neither easy to 

remember nor easy to pronounce, since the two names have been familiar to many linguists. 

2 . Because this thesis is not a study under the framework of barriers theory , I will not introduce 

the barrier theory proposed by Chomsky (1986b). The related aspect of that theory is di scussed 

in relevant sections of the thesis. 

3 . For consistency, I changed a S' into CP, S into IP. 

25 



CHAPTER 2 

The D istribution and the Properties of Serial Verb Construction 

2.0. Introduction 

The scientific study of the serial verb construction (SVC) began in 1875.1 Since thcn,dozens 

of studies have appeared. Although, before Sebba (1987) and Bickerton & Iatridou (1989), the 

definiti on of the notion was seldom discussed in the literature, the term serial verb constmction has 

been generally used to refer to a surface string of verbs, or verb plus verb- like, or verb phrase- like 

items which are not linked by any functional words and occur within a single matrix clause. 

Serial verb constructions arc widely used in the following languages: 

l. Chinese 

2. West Afri can Languages 

3. Caribbean creoles 

4. Dravidian languages 

Among these languages or language groups, Chinese is a relatively well-studied language in the sense 

that it is rich in historical written records and its grammar has been studied descriptively for several 

decades. There has also been detailed discussion of the SVC in the Dravidian languages2
. As 

regards other languages and language groups, there have been several studies, but historical records 

are not sufficient. Considering the availability of the linguistic data and the reliability of linguistic 

intuition, this study mainly focuses on data from Chinese. 

In this chapter, a general survey of SVCs will be made so as to lay a base for further 

discussion. In section 2.1., SVCs from several languages are surveyed with brief int roductions and 

explanations of the languages. From the discussion in this section, we will end up with a sim ple 

26 



picture of the distribution of SVCs. Section 2.2. will discuss the classification of SVCs. In this 

section, SVCs arc classified by the fo Uowing properties and criteria: 

1. Structures of SVC 

SVCs usually have three types of structure: clausal structure, phrasa l structure and 

compound structu re. 

2. F inite vs. non-finit e clause in SVCs 

If an SVC has clausal structure, the clause in it can be e ither finit e or no-finite. 

3. Different inner structures of compound SVCs 

A compound SVC in Chinese may consist of two verbs, or a verb plus an adj ective which 

is analyzed functionally as a verb-like category. We also discuss why this classifica tion is 

crucial in this study. 

4. Transitivity of the verbs which constitute SVCs 

5. Fixed and free combinations 

Some verbs always occur in SVCs in fixed positions, and have the same fun ction and 

meaning in different environments. They are called fixed combination SVCs. This type of 

SVC also includes the non-productive verb compounds. A ll other SVCs arc termed free 

combination S VCs. 

Section 2.3. deals with the properties of SVCs. In addition to the properties mentioned in 

the previous literature, especially those in Bickerton and fatridou (1989), I will introduce one 

important property of the languages which may have SVC: In all these languages, many of the linking 

devices, for example conjunctions, lexical complementizer or infinitival clause introducer the to in 

E nglish arc unnecessary for connecting two clauses or phrases. This property provides evidence for 

the structural analysis of this study. A form al definition of SVCs is given in 2.4 .. 

Section 2.5. discusses the previous literature on SVCs, from Christallcr (1875) to Li (l 990b) 

with brief comments. We will provide a full er discussion of Jansen ct al(1978), Sebba(1987), Steever 

27 



(1987), Baker (1989), Bickerton & Iatridou (1989), Li (1990a) as they arc of special importance due 

to their elaborate analyses. 

2.1. The Distribution of Serial Verb Construction 

Serial verb constructions have been attested in Chinese, West African languagcs(thc Kwa 

languages group and other languages), Caribbean creoles and Dravidian languages . In the fo llowing 

subsections, we will give a brief introduction to these languages and the SVCs in them. 

2.1 .1. Chinese 

Chinese is now spoken in mainland China, Taiwan, Hong kong and Singapore. Chinese has 

eight major dialects and hundreds of minor dialects. Although phonologically they arc so different that 

some dia lects e.g. Cantonese, arc hardly comprehensible to people from other dialect areas, 

syntactica lly, Chinese manifests a high homogeneity. The syntactic rules of standard Chinese, usually 

termed Mandarin or Common Speech (Putonghua), are applicable in all these dialects3
. So in this 

thesis, only Mandarin is discussed and it is assumed that the syntactic rules discussed in this thes is 

apply to all the Chinese dialects. 

Serial verb constructions in Chinese have been investigated in Chao (1968), Thompson 

(19736), Li & Thompson(1973,1974a, 19746), Bao(1985), Wu(1989) and Li(1990a), although the 

proper term was not employed in all these works. They provide many examples of SYCs: 

(1) Gou gan-pao le mao
4

. 

dog chase-run Asp. cat 

"The dog chased the cat and the cat has run away." 

(2) Zhangsan chi-bao le fan. 

Zhangsan cat-full Asp. meal 

"Zhangsan has eaten his meal and he has been full." 

(3) Zhangsan chi-wan le fan. 

Zhangsan cat-complete Asp. meal 

"Zhangsan has eaten his meal and nothing left." 

28 



( 4) Zhangsan qi-lei le ma. 

Zhangsan ride-tired Asp. horse 

"Zhangsan rode his horse (loo much) lhal he was tired." or 

"Zhangsan rode his horse (loo much) that lhc horse was tired." 

(5) Zhangsan qi ma qi-lei le. 

Zhangsan ride horse ride-tired Asp. 

"Zhangsan rode his horse (loo much) lhal he was tired." 

(6) Zhangsan zou-lei le tui. 

Zhangsan walk-tired Asp. leg 

"Zhangsan walked (too much) so that his legs were tired." 

(7) Zhangsan xi-ganjing le yifu. 

Zhangsan wash-clean Asp. clothes 

"Zhangsan washed his clothes clean." 

(8) Zhangsan kai men yingjie keren. 

Zhangsan open door welcome guest 

"Zhangsan opened the door lo welcome his guests." 

(9) Zhangsan yong dao qie rou. 

Zhangsan use knife cut meal 

"Zhangsan used knife lo cut meat." 

(10) Zhangsan jie luxiangdai kan. 

Zhangsan borrow video tape watch 

"Zhangsan borrowed video tapes lo watch." 

2.1.1.1. On PP + V and V + PP 

Scbba(1987) included (11) in SYCs by showing lhe contrast between (11) and (12): 

29 



(11) Wo zhu zai Shanghai. 

I live in Shanghai 

"I live in Shanghai." 

(12) Wo zai Shanghai zhu. 

I in Shanghai live 

"I live in Shanghai." 

He argues that sentence (12) is not an SVC, while (11) is, since zai(in) in (11) is preceded by a verb 

and becomes a verb. Consequently, sentence (13) is a lso included in this class: 

(13) Ta jie yi kuai qian gei wo. 

s/he lend one CL money give me 

"S/he lent me a dollar." 

[Sebba (1987, pp.31,180)] 

Along this line, he also classifies comparative sentences as instances of SYCs: 

(14) Zhang bi ni xi/wan juzi. 

Zhang compare you like orange 

"Zhang likes oranges more than you." 

(15) Zhang bi ni gao. 

Zhang compare you tall 

"Zhang is taller than you are." 

[Sebba (1987, p.207)] 

These sentences show that Sebba(l987) analyses some cases of Prep. + V or V + Prep. structures 

as SVCs. However we have evidence to argue that zai(in) and bi (compare), etc. arc not verbs, but 

prepositions. In Chinese, prepositions and verbs are historically closely related , and they do share 

certain common properties. However, considering the form al definition which will be given in 2.4., 

these sentences are not classified with SY Cs. We will go into details of this issue in sections 2.2.6.4. 

30 



and 2.2.6.5 .. 

2.1.2. West African Languages 

The Serial verb co11structio11 is very popular in the Kwa languages and olher west African 

languages. Moreover, it was lhe SVC in the Kwa languages that first allracled lhe allenlion of 

linguisls in lhe 19th century (Christaller, 1875). Kwa is a group of Niger-Congo languages spoken by 

lhc inhabitants in West Africa, exlcnding along lhe Allanlic coast from Weslern Liberia inlo Nigeria, 

Ghana, Togo and Benin. The most important languages in this group are Yoruba which is spoken 

in Western and Southweslern Nigeria, Igbo (or Ibo) spoken in Southern Nigeria, Akan subgroup, 

including Twi, Anyi, Beale, Guang, Metyibo which are spoken in Ghana, Ewe, Bini, and !gala 

(Lightfoot, 1979). Besides lhe inilial study by Christaller (1875), lhe SYCs in these languages have 

been investigated in the lileralure by Westermann (1930), Slewart (1963), Awobuluyi (1973), 

Bamgbose (1974), Williams (1976), Jansen ct al (1978), Lighlfoot (1979), and Baker (1989). The 

following arc examples of SYCs from these languages: 

(16) 0 mu 1we wa. ---- Yoruba 

he take book come 

"He brought the book." 

(17) Olu sare wa ilc. 

Olu ran came home 

[Bamgbo~e (1974) I 

---- Yoruba 

"Olu ran and came home." or "Olu came home quickly." 

(18) U la ebi ba naka. 

he took knife cut meal 

[Sebba (1987, p.12)] 

---- Nupe 

"He cut lhe meat wilh knife." 

[Lightfoot (1979, p.217)] 

(19) Dada gbe akoti lo ile ni ana. ---- Akan 

31 



Dada took box went home on yesterday 

"Dada took the box home yesterday." 

(20) Ozo rele le cva,e re. 

Ozo ra n cook food cat 

[Schachtcr(1974)] 

----Edo 

"Ozo quickly cooked and ate." 

(21) Lagos tye Ozo re. 

Lagos called Ozo come 

"Lagos summoned Ozo." 

(22) Lagos tye-re Ozo. 

Lagos ca lled-come Ozo. 

"Lagos summoned O zo." 

[Agheyisi(1987)J 

T he SVC in E do is very interesting in that it may be a compound verb as in (22), or a clause as in 

(21) . 

In addition to the languages mentioned here, SVC is a lso fo und in Elik Jukun Fanti , , , 

Yatye, Kru, Anyi, Ewe, Ga and Fon. 

2.1.3. Caribbean Creoles 

The Serial verb constmction has been attested in several Caribbean creoles· Saramaccan . , 

Sranan, A ucaans (Djuka), Sesclwa and M orisycn. Saramaccan, Sranan and A ucaans are E nglish­

related creoles which are spoken in Surinam. Scselwa and M orisyen are F rench-related and spoken 

in islands of the Indian Ocean. These languages are originally re lated to the West A fri can languages. 

T he studies of SVCs in Caribbean creoles include Bendix (1972), Dechaine & Lefebvre (1986) and 

Sebba (1987) . They contain rich linguistic data from Sranan; Byrne (1986, 1987), B icke rton (1989) and 

Bickerton & Jatridou (1989) . The following are examples from Sranan: 

(23) Lon go teki a buku tyari go gi a lcriman. 

32 



ran go take the book carry go give the teacher 

"Run and fetch the book and take it to the teacher." 

(24) De Jon owrukuku kiri. 

they beat owl kill 

"They beat owl to death. " 

[Sebba(1987,p.40)] 

Tn Aucaans: 

(25) A kai a dataa kon luku a pikin. 

he call the doctor come look the child 

"He called the doctor to come to look at the child." 

[Bickerton & Iatridou (1989)] 

In Saramaccan: 

(26) A lei goni suti di pingo. 

he take gun shoot the pig 

"He shot the pig with a gun." 

(27) Tei a lei goni suti di pmgo. 

take he take gun shoot the pig 

"He shot the pig with a gu." 

[Byrne (1987,pp. 166-167)1 

In Seselwa: 

(28) Mon ti pran balye koko ti bat Ii. 

J Ten. take broom coconut Ten. hit him 

"I hit him wi th a coconut broom." 

(29) Zot ti amas Ii ti al amene11 lopital. 

33 



they Ten.gather him Ten. go bring hospital 

"They took him to hospital." 

[Bickerton & Tatridou(1990)] 

2.1.4. Dravidian Languages 

One of the sa lient features of the Dravidian languages is that the serial verb constmclion 

exists or has existed in all the languages of this language group. Moreover, the possible syn tactic types 

of SVCs are sim ilar in all these languages. 

Dravidian is a language group of non-Indo-European languages spoken in Southern India and 

Southern Ceylon. It is generally divided into four language sub-groups: South Dravidian, South-central 

Dravidian, Central Dravidian and North Dravidian. The SVC has been reported in all these sub-

groups. 

South Dravidian includes Tamil, Kola, Tada, Kannada and Kodagu. A mong these languages, 

evidence has been found to confirm the presence of SVC. H owever, in this language group, while 

languages such as Old Tamil shows both clausal or compound SVCs, the modern languages in this 

group only display compound form, mainly in form of ki-echo compound. The ki-echo compound is 

a structure in which a noun or a verb is duplicated lo form a distr ibutive compo und. Consider the 

following examples from modern Tamil: 

(30)a. Malu Varum. 

oxen-Norn. come-Ten.-3sn 

"Some oxen will come." 

b. Matu-Kitu varum. 

ox-ki-echo-Nom. come-ful-3rn 

"Some oxen or stuff like it will come." 

Steever (1988) explains that when ki-echo formation applies to the noun m a Lu ( ox) in (30), it crea tes 

the echo compound noun matu-kitu ( oxen and the like) in (306). In the resulting compound noun kilu 

34 



partia lly reduplicates the form of the echoed word matu ( ox). The first syllabic of the echoed word 

is replaced in the echo word by a standard syllabic ki, which is short or long depending on whether 

the initial syllabic of the echoed word is short or long (Stecver,1988, p.48). The verb echo-compound 

is formed in a similar way as will be shown later. First let us observe some examples from Old 

Tamil: 

(31)a. kan-an-cn va,y-v-al. 

sec-Ten.Asp.-] s come-Ten.-ls 

"I will see and come." 

b. Tank-in-ir cen-mo. 

stay-Ten.-2p go-Asp.-(2p) 

"Stay, (then) go." 

[Purananwu 23.17; 233.7; Steever (1988, p.45)] 

The SVC in (31) has disappeared from the language by the early modern period. In modern Tamil, 

instead of such SVC, the Id-echo Formation is used to create compound verbs: 

(32) Contrast: 

a. Avan-ukkup paci-kkir-atu. 

he-Dat. hunger-Ten.3sn 

"He is hungry." 

b. Avan-ukkup paci-kkir-atu kici-kkir-atu. 

he-Dat. hunger-Ten.-3sn ki-Ten.-3sn 

"He is always hungry, or some damn thing!" 

[Steever (1988, p.49)] 

Sentence (32b) is analyzed with containing an SVC in the sense that the matrix verb and its 

reduplicated form (echo-compound) are not connected by any conjunction. SVCs in Toda also 

presents a form of the ki-eclzo compound: 

(33) Al Oid-xid-fi-c-a? 

35 



lhere eat-ki- Adv.-Tcn.-2s-Int. 

"Did you eat lhcrc or do any such (polluting) lhing?" 

lSlccver (1988, p.54)) 

Structurally, these languages are of most interest for their varieties of forms, which include balance, 

ilerative and echo compound verbs and co-ordinate verb constructions.(Steevcr,1988). Herc arc some 

examples: 

(34) Cus-t-amu cus-t-amu . 

sec-Ten.-ls sce-t-ls 

---- Modern Telugu 

"I shall continue seeing for sometime." 

[Sleever (1988, p.63)) 

(35) Conlrast 

a. Va-t-icler e sur-l-idcr. ----Kovi 

come-Ten.-2p and see-Ten.-2p 

"You came and saw." 

b. Va-t-ti sur-icler. 

come-Ten.-Tpe. see-Ten.-2p 

"You came and saw." 

(36) Baral-in en-mote-cl 

[Steever (1988, p.73)) 

cantk-mote-cl. --- Pariji 

bullock-Ace. speak-Ten.-3sm drive-Ten.-3sm 

"He drove lhe bullocks (by) speaking to lhem." 

[Steevcr(1988, p.89) 

2.1.5. SVCs and English 

The senlences shown above provide almost all the possible types of SYCs in modern 

languages. The survey shows thal SVCs exists exlcnsivcly in several major language families or groups 

in Asia, Africa and America. The languages having SYCs include uninfleclcd language such as 

36 



Chinese, or inllcclional languages such as the D ravidian languages. T he dislribulion of SVCs 

in<l icalcs lhal lhe presence of SVC is not relaled lo language lypc. T his may raise a question: D ocs 

lhe SVC exisl in an Inda-Europea n language like E nglish? Can E ng lish sente nces (38), (39) and ( 40) 

be analyzed as SVCs? In discussing lhe properlics and dcfinili on of SVC in 2.4., lhis queslio n will 

be answered: 

(37) John painled lhc house red. 

(38) John wcnl to bed hungry. 

(39) John made Bill painl lhc house. 

2.2 . Class ifications of Serial Verb Construction 

Serial Verb Constmctions are not onJy popular in many languages and language groups, they 

lhemselves are synlactica lly and semanlically varied as well. Theore tically, the notion serial verb 

constmction , as poinled o ul in Sebba(1987), is not used consislenlly in the lilcralure, a llhough the 

constructions thal lhe a uthors refer to share common properties. The deficiency of consistency can 

be parli ally a ttribuled lo the lack of a definiti on, partia lly to the facl lhat o ne author generally can 

only study the SVCs in a few languages or language groups, thus fa iling lo recognize the va rie ty of 

lhe phenomena. Furlhermore, some aulhors assume thal lhe rules of SVCs lhal they discovered in 

one language, which may have been supported by strong evidence, a re potenlia lly applicable lo other 

languages. It somclimes over-generalizes and simplifies the phenomena under discuss ion. A clually, 

as will be shown in fo llowing chapters, a lthough the evidence from SVCs provides strong suppo rt for 

the linguislic universals assumed by the present linguistic lheorics, the phenome na themselves a rc by 

no m eans syntaclically and semantically as simple as some works ass umed. 

In 2.1., the varielies of SVCs have been prcsenled. H owever, lhe proper picture is fa r from 

being clear and is not sufficient for a theoretical investigation. In this seclion, I will classify the 

phenomena shown in lhe last seclion from differenl angles and wilh differenl crite ria, so that lhe 

phenomena become bclle r organized. 

2.2.1. Class ification of SVC Structures 

37 



Most studies of SVC, except Steever (1988), have assumed that all the SVCs in languages 

have only one possible structure. For example, Bickerton & Iatridou (1990) proposes an IP structure. 

Baker (1989) proposes a V + V' structure, etc. . H owever, if we expand the investiga tion into a ll the 

languages which have reported SVCs, we find that at least three structures arc possible, and each of 

them can be analyzed into several subtypes: There are clausal structures, compound structures and 

phrasal structures. 

2.2.1.1 . SVC with Clausal Structure 

SVCs with clausal structu re are common lo many languages. In Chinese, there is evidence 

lhal the sentences (40) and (41) have structure (42) : 

( 40) Zhangsan kai che song xing. 

Zhangsan drive car deliver letter 

"Zhangsan drives a car to deliver letters." 

( 41) Zhangsan yong kuaizi chi fan. 

Zhangsan use chopstick cat meal 

"Zhangsan uses chopsti cks lo cal." 

38 



(42)a. IP 

b. 

I \ 
NP I' 
I I \ 

Zhangsan I VP 

I \ 
V' CP 

I \ I \ 
V NP IP 
/ I / \ 

kai che NP I' 
(drive) (car) / / \ 

IP 
I \ 

NP I' 

e I VP 

I \ 
V NP 
I I 

song xing 
( deliver )(letter) 

I I \ 
Zhangsan I VP 

I \ 
V' 

I \ 
CP V' 

/ I / \ 
IP V NP 

I \ I \ 
NP I' chi fan 

/ \ (eal) (meal) 
l VP 

I \ 
VP NP 
I I 

yong kuaizi 
(use) (chopstick) 

Evidence and arguments lo support lhe slruclure (42) is given in Chapter 4. Bickerton & Iatridou 

(1989) argues convincingly that lhc SVC in the Aucaans sentence (43) is a clause: 

( 43) A pikin gi mi a sopu wasi cn/miseli. 

the child give me the soap wash him/myself 

39 



"The child gave me the soap to wash him/myself with. 

This type of SVC is also popular in other languages: 

(44) In Seselwa 

Zot ti anvoy mwai al ape! doker vm get tifi 

they Ten.send me go call doctor come look girl 

"They send me lo call a doctor to come and look al the girl." 

[Bickerton & Jatridou (1989)] 

2.2.1.2. SVCs as Verb Compounds 

Another type of SVC is the compound verb: Two verbs or one verb and one verb­

like combine together to form a new word. This lype of SVC is allesled in Chinese and Edo: 

(45) VP 
I \ 

V' 

I \ 
V 

I 
I \ 

a 13 

a = verb, 13 = verb or verb-like 

( 46) ln Chinese: 

Z hangsan die-duan le tui. 

Zhangsan fa ll-broken Asp. leg 

"Zhangsan fe ll down and his legs were broken." 

( 47) Zhangsan kan-lei le dianshi. 

Zhangsan walch-lired Asp.TV 

"Zhangsan watched the TV (too much), so he was tired." 

(48) In Edo: 

Lagos tye-re ozo. 

40 



he called-come Ozo 

"He summoned Ozo." 

This example is from Agheyisi (1987). Although it only gives ( 48) as example of the sen tence of 

which the predicate consists of a compound verb, it also lists other compound verbs as follows: 

( 49)a. mu-hee carry-begin "start; begin" 

b. fya- 1: cut-eat "bite" 

2.2.1.3. SVCs with Phrasal Structure 

l 11 addition to clausal structure and compounding structure, another possible structure for 

SVCs is phrasal structure. Much previous literature, such as Schachter (1974), Williams (1976), Jansen 

ct al (1978) and Baker(l989) have proposed phrasal structures for all SVCs. The varieties of 

structures of SVCs surveyed in the previous section show that this is probably not correct. In Chapter 

4, we give evidence show that their claim is not correct. 

However, we do find that there is at least one type of SVC which can only be analyzed as 

having phrasal structure. Observe the Koda sentences (50). These two sen tences have exactly the 

same interpretation. In (50a), two clauses are coordinated by a conjunction e (and); in (50b), the 

conjunction disappears, but two parts of tbe sentence arc still in a coordinate relation: 

(50)a. Va-t-idcr e sur-t-idcr. 

come-Tcn.-2p and see-Ten.-2p 

"You came and saw." 

b. Va-t-i sur-idcr. 

come-Ten.-Tpc. sec-Tcn.-2p 

"You came and saw." 

In (50b), the first verb only has a truncated personal ending. If here there arc two coordinate 

clauses, we do not expect V 1 (Va-t-i) lo have a trunca led ending. This would follow that both verbs 

must share one JNFL, thus being in one clause rather than in two coordinate clauses.If the approach 

to (50) is correct, (50b) shou ld have a D-structure (51): 

41 



(51) IP 
I \ 

NP I' 

I I \ 
pro I VP 

I I \ 
VP Conj. VP 

I I \ 
va (e) sur 

From the discussion above, it appears that SVCs can have three poss ible inner structures: 

clause, phrase and compound. Furthermore, we are going to see in the following sections that the 

clausal structure and the compounding structure should be analyzed into several sub-classes. 

2.2 .2. Finite vs. Non-finite Clause 

Bickerton and Iatridou (1989) divides the structures of a ll the SVC into two classes: linitc 

clause and non-finite clause. The fo llowing are the examples: 

(52) In Scsclwa ---- Finite clause: 

Nui ti sorti proi ti al !apes. 

we Ten. leave Ten. go fishing 

"We set out to fish." 

(53) In Saramaccan ----Non-finite clause: 

Mi tei faka koti beee. 

I take knife cut bread 

"I cut bread with a knife." 

In Chapter 4., we will see that the distinction between the finite clause and the non-finite clause pl ays 

a role in determining the structures of some SVCs in Chinese. 

2.2.3. VY Compounds and VA Compounds 

fn Chinese, if an SVC is a verb compound, it can consist of two verbs, or a verb and an 

adjective which shares many important properties with the verb (we will return to this topic in 

Chapter 3): 

42 



(54) In Chinese ---VY compound 

Zhangsan song-zou le Lisi. 

Zhangsan send-go Asp.Lisi 

"Zhangsan scnl Lisi off." 

(55) In Chinese ---YA compound 

Zhangsan qi-lei le ma. 

Zhangsan ridc- lircd Asp. horse 

"Zhangsan rode lhe horse (too mush) so he was tired." 

"Zhangsan rode lhe horse (too much) so lhc horse was lired. " 

(56) Zhangsan die-duan le Lui. 

Zhangsan fall-broken Asp. leg 

"Zhangsan fell down and broke his leg." 

From examples (54)-(56), we have found lhat lhe VA slruclure is synlaclially more complex than lhal 

of lhe yy structure. In VY slruclures, lhe first Y musl be lransilivc, whi le in VA slruclurc, il may 

cilher lransilive (as in (55)) or inlransilive(as in (56)); in VY slruclurcs, the argumenl of lhe second 

V must be lhe internal argumcnl of lhc VP ((54)) ,--- lhe person who has gone can only be Lisi, not 

Zhangsan in (55) . In VA struclurcs, lhe second elemenl can be linked eilhcr lo Lhe internal or 

cxlcrnal argument of Lhe VP ( as shown in (56)). Conscqucnlly only YA slruclu rcs can be ambiguous. 

One aim of the work is to explain lhcse differences bclwccn VY slruclure and YA slruclures. 

To sum up, SYCs can have lhc following poss ible slruclurcs: 

(57) Classification of SYCs 

Clause: Finile clause 

Non-Finilc Clause 

Phrase 

Compound: 

V +Y 

43 



V+A 

2.2.4. Transitive Verb vs. Intransitive Verb 

As regards transitivity of verbs, verbs lhal form a serial verb constmction can be either 

transitive or intransitive: 

(58) In Yoruba 

Bola se eran ta. 

Bola cook meat sell 

"Bola cooked some meat and sold it." 

[Lord(1974),; Baker (1989)] 

(59) In Chinese 

Zhangsan changchang kao JI chi. 

Zhangsan often roast chicken cat 

"Zhangsan often roasts chicken to eat." 

It is not unusual to find an SVC consisting of a transitive verb and an intransitive verb as in (60) and 

(61): 

(60) SVC with one intransitive verb ---lo Yoruba 

Olu Li omo naa subu. 

Olu push child the fall 

"Olu pushed the child down." 

(61) In Sranan 

Kofi lyari a Amba gwe. 

Kofi throw the book fall 

"Kofi threw the book down." 

[Bamgbo§e (1974); Baker (1989)] 

[Baker (1989)J 

44 



An SVC can consist of two intransitive verbs: 

(62) In Yoruba 

Titi rin lo ni, ko sare lo. 

Tili walk go is not run go 

"Titi left walking, not running." 

[Baker (1989)] 

(63) In Sranan 

A watra dropu fadon. 

the water drop fall 

"The water drips down." 

[Sebba (1987, p.71)] 

A most interesting case is the VA structure of the Chinese SVC. In this structure, the verb can 

be either transitive or intransitive. Moreover, even if it is intransitive, the whole structure will still 

be transitive and thus take an object: 

(64) Zhangsan zou-lei le tui. 

Zhangsan walk-tired Asp. leg 

"Zhangsan walked (too much) so he was tired." 

(65) Zhangsan die-slzang le yao. 

Zhangsan fall-wounded Asp. waist 

"Zhangsan fell clown and hurt his waist." 

However, what is the mechanism that makes the combination of two intransitive elements transitive? 

It is proposed that the mechanism involves feature percolation and argument-linking in the sense of 

Lieber (1983). We will return to the issue in Chapter 3. 

2.2.5. Fixed vs. Free Combinations 

Considering structural stability, serial verb constructions can be class ified into two types: the 

fixed combination and the free combination. Herc fixed means 

45 



1. in clausal or phrasal structu res, certain verbs appear again and aga in in the same 

position, or 

2. in a compounding structure, certain VY or VA combinations arc 11011 -productive. 

In other words, we cannot replace one of the clements to fo rm a new compound 

freely. 

If an SVC is not in the cases 1. or 2., it is free. Some Sranan examples from Scbba (1987) will be 

presented to illustrate the classification. 

ln Sranan either the V 1 or the V 2 may be [LXed: 

( 66) Fixed V 2 

You dyompo tanapu lcki wan bubrompki. 

you jump stand like one bushflower 

"You spring up like a windflower." 

Tanapu (stand up) in Sranan always follows a verb of motion or change of position, to indicate that 

the object whjch changes position ends up in a vertical plane. O ther possible combinations include 

opo ... tanapu (get up .. . stand,"stand up"), dyompo lanapu (iump ... stand up, "stand up")(Sebba 1987, 

p.42). 

(67) Den fon owrukuru kiri . 

they beat owl kill 

"They beat owl to death." 

This is a typical example of the fixed combination, in which the V2 kiri (kill) docs not have its own 

object, and follows V1 which yields to an action which may be used as a means of killing or 

extinguishing. O ther combinations in this language as sutu kiri (shoot...kill, "shoot and kill"), anga kiri 

( hang ... kill, "hang and kill"), bro a kangera kiri (blow the candle kill, "blow out the candle"), (Scbba, 

l 987, p.43). 

v
1 

in Sranan SY Cs also can be fixed: 

46 



( 68) Fixed V 1 

Gorogoro sa bari si ngi. 

throat Ten./ Asp. shout sing 

"Throats will sing out loud." 

In this sentence, the intransitive verb bari may be followed by a V which refers to a vocal activity, 

such as krey ( cry), lafu ( laugh) , kari ( call) lo add information that the action of V 2 is particularly loud 

or intense (Sebba, 1987, p.56). 

(69) No tel<i baskita tyari watra. 

Neg. take basket carry water 

"Don't carry water with a basket." 

In (69), both verbs have their own objects respectively, teki (take) + NP serves as the instrument of 

the action of V
2

. Other examples of this type include teki. .. go (take ... go, "take ... to ... "), teki ... kosi 

(take ... curse, "cursc ... by ... "). Teki can also be used in other structures as a fixed verb: 

(70) A teki a fish seri. 

S/he take the fi sh sell 

"S/he sold the fish." 

In (70), teki (take) has an object, but the second verb docs not have overt complement. T o explain 

(70), Scbba states that the semantic contribution of teki is negligible, and the verb in this position 

seems to serve mainly the syntactic purpose of moving the direct object of V 2 further to the left 

(Scbba, 1987,pp.59-60). 

The distinction between the fixed and free combination is even more crucial lo compounding 

structures. In Chinese, if a VY verb compound is productive, the first verb must not be intransitive, 

and the second verb must not be transitive: 

(71) Zhangsan ma-zou le Lisi. 

Z hangsan scold-go Asp. Lisi 

"Zhangsan scolded Lisi so that he left." 

47 



In con trast, the following combinations are not grammatical: 

(72) vt rans. + vt rans. 

a.* gan-m a : chase- scold 

b. * da-gong: light-attack 

(73) vintrans. + vint rans. 

a.* pao-duo: run-hide 

b. * zuo-shui:sit-sleep 

(74) vint ra ns. + v tra ns. 

a.* zuo-ka11 : sit - watch 

b. * zhan-mai: stand-buy 

However, with the non-productive combinations, Vt,·a,
1
s. + V V + y a d y 

trans., in trans. intra ns. n int rans. 

+ Vtrans_are possible: 

(75) v int rans. + vint ra ns. 

Tamen jingchang lai-wang. 

they often come-go 

"They often visit each other." or 

"They often get in touch to each other." 

(76) vtrans. + trans. 

Tamen zhan-bai le diren. 

they fight-defeat Asp. enemy 

"They defeated the enemy." 

(77) vintrans. + vt ra ns· 

Tamen zuo-shi le liangji. 

they sit-lose Asp. golden-opportunity 

"They lost the golden opportunity." 

Fixed combination compounds (non-productive verb compound) and the free combination 

48 



compounds (productive verb compound) have different rules of formation, constraints, feature 

percolation and argument linking. Chapter 3 discusses the grammar of productive compounds. 

2.2.6. Some Typical SVCs 

Although it is not yet clear that how many verbs arc frequently used to form SVCs in 

languages, we do have find that certain verbs are widely used in different languages in forming 

SVCs. Here two facts are worth noting: the frequency of appearance of these verbs is much higher 

than that of other verbs in SVCs, and the high frequency of use of such verbs is unexceptionably 

found in most languages which can have SVCs. 

2.2.6.1. The Take Constructions 

As Sebba (1987) points out, the take construction is one of the commonest type of SVCs. 

Take can be used as a V1 or V2 denoting instrument. Sometimes, this verb lacks semantic specificity 

in a given construction: 

(78) In Yoruba 

Mo ft ada ge iqi . 

I take machete cut tree 

"I cut the tree with a machete." 

[Stahlke (1970); (Sebba (1987, p.168)] 

(79) In Ijo 

Eri ogidi aki-ni indi pe1-1111. 

he machete take-L. fi sh cut-up-SP. 

"He cut up a fish with a machete." 

[Sebba (1987, p.169)] 

(80) Eri, edein-bi aku bo-mi. 

he knife-the take come-SP. 

"He brought the knife." 

[Williamson (1963); Sebba (1987, p.169)] 

49 



Williamson notes an interesting fact about aku in (80) : When aki (lake) imm ediately precedes bo 

( come), its final vowel becomes unrounded. Such a change lakes place in a few other common set 

phrases. Therefore he assumes that aku-bo (take-come) is a compound. 

(81) In Saramaccan 

H en lei di fa ka koli di bee. 

he take the knife cul the bread 

"He took the knife (to) cut the bread, but didn't cut it."5 

(82) In Chinese 

Z hangsan na bi xie zi. 

Zhangsan take pen write character 

"Zhangsan look a pen lo write." 

Benn ell (1981) argues that in Chinese Lhe ba ( and ils variant jiang) structure is para llel lo the take 

construction in the sense that historically ba was a verb meaning "lake, hold use", for example: 

(83) Wo ba juzi bo le pi . 

I BA orange peel Asp. skin 

"I peeled the orange." 

[Li & Thompson (1981, p.471] 

We will argue in later sections lhal this classification is nol correct, even though in O ld Chinese ba 

or jiang were both verbs. 

2.2.6.2. The Go Constructions 

Another common construction is the go construction. We include not only go, but also come 

and other motion verbs in this construction. One of the salient features of this construction is that 

these verbs work almost in the same way in different languages. They always occur aft er V 1 to 

distinguish the motion of V1 towards or away from the speaker: 

(84) In Akan 

Miguaree mebaa mpono. 

50 



I-swim-Ten. I-come-Ten. shore 

"I swam to the shore." 

(85) In Yoruba 

Olu gbe aga wa. 

Olu take chair come 

"Olu brought a chair." 

(Ch ristallcr (1875, p.131); Sebba (1987,P.184)] 

[Scbba (1987, p.186)j 

(86) In Ijo --- three verb combination 

Eri bani mu weni-mi. 

he run go walk-Sp. 

"He went running and walked."6 

[Williamson (1963, p.153 fn .); Sebba (1987, p.187)] 

(87) In Saramaccan 

Di womi subi go a liba. 

the man go up go LOC river 

"The man went up to the river." 

(88) In Chinese---contrast 

a. Zhangsan na lai le diannao. 

Zhangsan take come Asp. computer 

"Zhangsan brought the computer." 

b. Zhangsan na zou le diannao. 

Zhangsan take go Asp. computer 

"Zhangsan took the computer away." 

The two examples above show that lai (come) and zou (go), like their equivalents u1 other languages, 

indicate the direction of the motion done by the first verb. 

51 



2.2.6.3. Double Go Cons tructions 

As reported by Sebba (1987), double - go construction exists in some languages. The 

form of it is [go PP go VPJ. fn this construction, the first go is usua lly interpreted as a motion 

verb, fo llowed by a locative directional complement; the second go is the so called pwpose go (Sebba, 

1987, p.192): 

(89) In Sranan 

Mi go na wowoyo go bay krosi. 

I go Loe. market go buy clo thes 

"I went lo the marke t Lo buy clo th ." 

[Sebba (1987, p.194) J 

Tn Twi the exact counterpart has been found: 

(90) Mek Niere k t nlama. 

I-go-Ten. Accra go buy cloth 

"I went to Accra and bought cloth." 

[Sebba (1987, p.194] 

Scbba (1987) proposes that such constructions also exist in Chinese: 

(91) Tamen dao Beijing qu nian shu qu le. 

they to Beijing go read book go Asp. 

"They went lo Beijing lo study." 

[Sebba (1987, p.l96)J 

H owever, there is an important difference between the case of Chinese and tha t of Sranan or Twi: 

In Chinese, either the first or the second qu (go) can be dropped: 

(92) Tamen dao Beijing qu nian shu le. 

they to Beijing go read book Asp. 

"They went lo Beijing to study." 

52 



(93) Tamen dao Beijing nian shu qu le. 

lhcy lo Beijing read book go Asp. 

"They went lo Beijing to sludy." 

F urlhcrmore, stylislically, (92) is standard Chinese, while (91) and (93) arc a Iilllc colloquial. It seems 

lhat scmanlically qu (go) contribulcs litlle to lhe sentence lhat it is in. As in the single go 

construclion, lhe second qu (go) only indicates the direction of lhc mot ion performed by the first qu 

(go). Since the meaning of firsl qu (go) has conlained an interprelalion "leaving away", the second 

one consequenlly will be oplional. This raises anolher qucslion: What is the funclion of qu (go) in 

(93)? In this sentence, there is no verb denoling a motion. The fact is lhat the PP dao Beijing, which 

means to Beijing or to wards Beijing, also implies a molion. Therefore lhe fun ction of qu (go) will nol 

change. Consider the following sentence, if we replace qu (go) wilh lei (come), lhc meaning of lhe 

scnlcnce changes as we expect: 

(94) Tamen ciao Beijing nian shu lai le. 

they to Beijing read book come Asp. 

"They have come to Beijing lo study." 

For (93), the speaker must not be in Beijing, while for (94), lhe speaker must be in Beijing. The 

change of meaning is triggered by replacing qu (go) with lai (come). Therefore in all these sentence, 

the fun clion of qu (go) as V2 remains the same. This is different from the funclion of second go in 

Sranan or Twi, but as same as thal in the single go conslruction. 

2.2.6.4. Give Constructions 

Another popular construclion is the give construction. This construction is very interesting 

for the verb give is a three-place verb, which takes lwo objects in a sentence. Evidence has shown that 

such constructions do not exist in Chinese. Sebba (1987) has found its existence in other languages: 

(95) In Sranan 

Mi tyari en gi Kofi. 

J carry it give Kofi 

53 



"I brought it to Kofi (as a gift) ." or "I carried it for Kofi." 

[Sebba (1987, p.174)] 

The ambiguity of the sentence comes from the different interpretations of the verb gi (give): it may 

be interpreted as dative, or benefactive. As pointed out by Sebba (1987), such ambiguity is not 

unusual in some languages: 

(96) In Akan 

Ogyaw ne sika mfil! me. 

he-take knife the give-Ten. me 

"He left me his money." 

(97) In Yoruba 

0 Ii owo naa Jun me. 

H e put money the give me 

"He gave me the money." 

[Christaller (1895, p.118); Sebba (1987, p.174)] 

[Scbba (1987, P.176)] 

Sebba (1987) also considers that Chinese has an extensive range of such construction using give to 

mark either an indirect object or a bencfactive. The fo llowing arc the examples: 

(98) Ta jie yi kuai gian gei wo. 

s/he borrow one CL. money give me 

"S/he lend me a dollar." 

(99) Ta gei wo xie le xm. 

s/he give me write Asp. letter 

"He wrote letter to me." 

[Scbba . (1987, p.180)] 

However, sentences (98) and (99) are not well accounted for in Scbba (1987) . Gei in Chinese belongs 

54 



lo two categories and has two different functions. Let me label the two different geis as gei
1 

and geiz 

respectively. Gei1 is a verb which lakes two objects, meaning "give": 

(100) Zhangsan gei le Lisi yi ben shu. 

Zhangsan give Asp. Lisi one CL. book 

"Zhangsan gave Lisi one book." 

This gei is categorially and functionally very similar to the English word give. Gei
7

, on the other hand 
- ' 

is a preposition, meaning "lo, for": 

(101) Zhangsan mai le yi ben sh u gei Lisi. 

Zhangsan buy Asp. one CL. book for Lisi 

"Zhangsan bought a book for Lisi." 

These two geis are not difficult to distinguish: In Chinese, there is an important grammatical rule: 

the object of a verb can be moved, while the object of a preposition cannot. For example, from (100) 

we can form (102)a.,b. by virtue of movement: 

(102)a. Yi ben shu, Zhangsan gei le Lisi. 

one CL. book Zhangsan give Asp. Lisi 

b. Lisi, Zhangsan gei le yi ben sh u. 

Lisi Zhangsan give Asp. one CL. book 

"Zhangsan gave Lisi one book." 

For (101), the object of gei must not be moved: 

(103) *Lisi, Zhangsan mai le yi ben shu gei. 

Lisi Zhangsan buy Asp. one book for 

So gei Lisi in (101) is a prepositional phrase rather than a VP. We have furth er evidence to support 

the analysis: I