The Chinese Resultative Construction A Cognitive Constructional Analysis Ronald Fong LINCOM Studies in Chinese Linguistics

LINCOM Studies in Chinese Linguistics

E 0 0 C: F

The Chinese Resultative Construction A _ Cognitive Constructional Analysis Ronald Fong 2020 LINCOM GmbH

Published by LINCOM GmbH 2020. LINCOMGmbH Hansjakobstr.127 a D-81825 Muenchen contact@lincom.eu www.lincom.eu webshop: www.lincom-shop.eu All rights reserved, including the rights of translation into any foreign language. No part of this book may be reproduced in any way without the permission of the publisher. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de jJ FSC ,..,..,.I,uq MIX Paper from respon1lble 1ourcea FSC' C083411 Cover painting: Distant Secret, courtesy Sumiko Ito. Links to third-party websites are provided only as information and help for the reader. LINCOM disclaims any liability or responsability for the contents of links provided by the authors and editors of this volume.

Acknowledgements In preparation of this monograph, many people helped me in various ways. Particularly, I am grateful to KK Luke and L-H Tan for their encouragement and help at different stages. Andrew Spencer commented on parts of the draft, for which I am thankful. I would also like to express my gratitude to Hannah Gibson.

Some Common Abbreviations asp cl COMP RVC aspect classifier complement resulative verb compound 11

Tabel of Contents Aclmowledgements Some Common Abbreviations Chapter One .. ...... .... ... . . .... . ...... ... . . . ....... ... ...... ... . .. .. ..... ..... ...... ... 1 Introduction Chapter Two ... . . . ......... . . . . .. ....... . ........ .. ................ . .. ..... ..... ..... . ... 7 Major Research on the Chinese Resultative Construction Chapter Three . ..................... ..... . .. . . . . ..... .. .. .. .... .... ... . ...... ..... ... .. . 66 Verb Classes and Constructions Chapter Four . . . .. . ... ... .. .. ... . .. .... . ...... ...... . ..... ...... ... .. ... . .. ... .. . ... . ... 82 Alternative Analyses of the Resultative Construction Chapter Five . .. . ....................................... . ...... . .... . ... . ... . . .. .. . . .... 132 The ha-Construction Chapter Six .. .... ... .. . .. . ...... ....... . ... ... ... . . ... . ... . .... . ..... . ... ............ . 162 Conclusion "' Bibliography iii

Chapter 1 Introduction 1.0 Brief Overview An important kind of secondary predication or sometimes known as complex predicates in Chinese is realised by the resultative verb complement construction. This phenomenon is certainly not unique in Chinese; in fact, the construction of complex predicates is observed crosslinguistically (eg Butt 1995, Washio 1997, 1999, Rappaport Hovav and Levin 2001). In recent years, the interface between syntax and lexical semantics has intrigued researchers in linguistics and related fields more than ever. Argument structure has been postulated to capture the relationship between syntax and semantics. Resultative verb compounds in Chinese have been shown to provide insights into the theory of argument structure as the syntax-semantics interface. In recent studies (eg Cheng 1989; Li 1990, 1993; Cheng and Huang 1994), attempts to account for the argument structure of resultative compounds have been made by analysing the argument structure of the verbs within the compounds. Indeed, Chinese provides a rich field for the study of compounds. Several classes of compounds are identified and they are structurally similar to the syntactically related phrase (Chao 1968, Thompson 1973, Li and Thompson 1981, Anderson 1985), which are subject-predicate verb compounds ([N-V]: tou-teng 'head-pain: headache'), coordinate verb compounds ( [V -VJ: sheng-chan 'bear-bear: produce'), verb-complement verb compounds ([V-V/A]: da-po 'hit-break: break'), verb-object verb compounds ([V-N]: kai-dao 'open-knife: operate on') and finally modifier-head verb compounds ([Adv-V]: re-xin 'hot-heart: enthusiastic'). Among the five types of verbal compounds the most widely known type is the verb-complement verb compound, which includes the widely studied resultative verb compound (Chao 1968, Li and Thompson 1981). The resultative verb compound in Chinese consists of two elements, or structurally [Vl-V2]v ( eg xie-cuo 'write-wrong: write incorrectly'; zuo-wan 'do-complete: finish') and the second element, most of which involves virtually all intransitive verbs/stative verbs, indicates some kind of result

of the action designated by the first element. The results can be of different types, for example: (1) Chabei da-po-le teacup hit-broken-asp The teacup broke. (2) Yingwen yao xue-hao English want learn-good Learn English well. (3) Ta kai-guo-qu-le he drive-cross-go-asp He drove across. The examples consist of resultative compounds. (1) signals the cause and effect, while (2) signals the achievement and (3) indicates directions. (3) is also referred to as an instance of directional resultative verb compounds, for it contains the directional verbs guo 'cross' and qu 'go'. This book is organised in the following way. Chapter One is a brief introduction to the study of the resultative construction in Chinese. Chapter Two surveys the major studies of the Chinese resultative construction done in the last few decade's and Chapter Three will examine the verb classes that can enter into the relation with the resultative construction. In Chapter Four we will show how the resultative can be regarded as a construction and how it exhibits its structural and functional properties. In particular, we will put forward several innovative or 'radical' proposals such as the catenative analysis and the conceptual ( co )subordination anaylsis. In Chapter Five the ba-construction is discussed and argued to be a kind of resultative construction. Chapter Six is the conclusion. I.I The Resultative Construction Resultative constructions can be of two types: lexical and phariphrastic. The former exists in Chinese lil

cry. In terms of meaning, they express the event type cause and effect. Structurally, there 1 are two predicates. In English, for example, the first predicate expresses activity related to CAUSE and the second predicate expresses result related to a change of state ie BECOME. A noun phrase intervene between the two predicates in English and in Chinese, they merge like a single item. Carrier and Randall (1992) identify the English resultative as transitive and intransitive: The gardener watered the tulips flat (transitive) and The joggers ran their Nikes threadbare (intransitive). The first type subcategorises for an NP and the second does not. Thus, the tulips is an argument of water and their Nikes is not an argument of run. In English, as Simpson (1983) notes, the postverbal NP is predicated of by the secondary predicate, whether it is a fake object or underlying object. This is referred to as the Direct Object Restriction (DOR). Since the resultative secondary predicate occurs before the NP, the NP argument may not be there as in John chi-bao-le 'John ate full' . The subject NP is predicated ofby both chi and bao. The resultative construction can be analysed as binary branching or ternary branching. It can be considered to be a. 'small clause' (Hoekstra 1988) [SC [the tulips] flat] ] or to be ternary [VP [VNP AP]]. Some have recently adopted the VP shell analysis. In this study, we will see some advantages of taking the resultative as a construction, having syntactic and semantic features. The resultative construction in Chinese also interacts with the ba-construction, which typically expresses cause and effect. w·~ will also note that the construction has some bearing on transitivity, as a syntactic or semantic concept. The resultative and the ba-construction are highly transitive structures, and provide human beings with a means of understanding the cause and effect in the world. The two constructions can be seen as a constructional family having related senses. Typologically, the Chinese resultative construction is similar to cosubordination, which exhibits some coordination and subordination features. Simultaneously, the Chinese resultative construction is similar to the English counterpart in that they can be recognised as belonging to the same family of constructions. The current approach can be translated into other frameworks fairly easily and can be taken as a guide to subsequent research on resultative constructions in various languages and dialects. 3

One of the interesting problems that has been the centre for research is exemplified by the following sentence: (4) John qi-lei-le ma john ride-tired-asp horse a. John rode the horse and he got tired as a result. b. John rode the horse and it got tired as a result. The fact that this sentence is ambiguous is puzzling, and while we notice similar examples that are not ambiguous, for instance: (5) John qi-lei-le che john ride-tired-asp bicycle a. John rode the bicycle and he got tired. b. *John rode the bicycle and it got tired. A formalist approach has to deal with the theta-role assignment and head percolation features, etc. Still, there are exceptions to the formal generalisations. For instance, the principle of theta-criterion is violated in example ( 4) if John is assigned both the Agent and the Experiencer/Theme roles. As a result, only one interpretation must be chosen according to the principle, although the principle is observed in example (5), as the sentence is not ambiguous. If we examine this phenomenon from another angle, as will be argued for in this study, perhaps it can yield a different result. Take the previous sentence ( 4) as an example, again: cause-result John---~ T I ma ~ --r ----------------------------------------------------------------- ------------------------ --4

The interpretations of the sentence require different types of information such as syntactic, semantic and pragmatic-discourse, which will be explored in this book. The dotted-lined box above indicates that the canonical or prototypical interpretation of the resultative construction applies Causer and Causee to John and ma, respectively. John being both Agent (Causer) and Experience (Causee) is not canonical, while the horse being the Patient (Causee) and Experiencer (Causee) is compatible with the roles given by the construction - Causee/Theme/Experiencer. The cause-result conceptual structure lurks behind the entire resultative construction and it is supposed to be the Patient argument that is affected and caused to produce a result. In other words, it is the horse that is supposed to be tired under the canonical structure and meaning of the resultative construction. This is also supported by our world knowledge that an animate or living thing can be tired. This knowledge allows us to interpret John as in John qi-lei-le che as the one that gets tired, because a bicycle is not animate and thus cannot be tired. 1.2 Frameworks This work attempts to be eclectic but adopts different frameworks where appropriate. In particular, we will consider the resultative in Chinese as a construction in terms of syntax and semantics. Further, we also attempt to integrate the formal and functional endeavours while we suggest that a functional or semantic perspective may be preferable in the study of the Chinese resultative construction. Being eclectic does not imply that we do not make use of any theoretical devices. In fact, we will apply certain major concepts and principles of--Cognitive Construction Grammar (eg Goldberg 1995, 2006) and Cognitive Linguistics (eg Langacker 1991, 2008) to the analysis of the resultative construction when we see fit. 1.3 Terminology The notions of subject and object are used conveniently to avoid repeating terms such as 'preverbal element' or 'postverbal element'. As argued by LaPolla (1993), quite convincingly, the concepts of subject and object may not be appropriately applied to Chinese. Furthermore, when we discuss semantic roles, labels like Agent and Actor, and Patient and Undergoer will be used interchangeably. They are used according to the discussion of the relevant literature. The resultative verb compound (RVC) is called differently in the literature, for example, the verb-resultative compound and the resultative verb. In this work we will 5

use such names where appropriate, particularly according to the literature. However, we will consider it as a syntactic construction, and so we will use 'resultative construction' for our purposes. 1.4 Chinese Orthography Since the resultative verbs have been traditionally regarded as compound verbs, they are hyphenated in orthography. We follow this tradition especially when we cite examples in the sources that consider them as compounds. However, in our later catenative analysis ( Chapter 4) we will provide evidence that there should be a syntactic boundary between the activity verb and the result verb and thus they should not be written together with a hyphen. The traditional way to represent the verbs is merely for convenience in this study. 6

Chapter2 Major Research on the Chinese Resultative Construction 2.0 Introduction The resultative verb complement construction or commonly, the resultative verb compound (RVC) has been one of the most interesting areas of Chinese linguistics. The classic work on Chinese grammar available in English is Chao ( 1968), which follows the tradition of structuralist linguistics particularly influential in America before the 'Cognitive revolution' in the fifties. Chao distinguishes several types of compounds in Chinese, and the resultative verb compound is a subtype of the verb-complement compound. In the early days of transformational grammar, Hashimoto (1966) and Thompson ( 1973) were among the first few works which applied the transformational analysis to Chinese resultative verb compounds. Hashimoto favoured the underlyingly bi-clausal analysis of resultative verb · compounds while Thompson proposed that resultative verb compounds are derived from lexical rules. Although Thompson made use of lexical operations, her work still had the transformational grammar spirits. Hence, Thompson (1973) can still be regarded as a transformational approach. In response to Thompson, Lu ( 1977) argued that the underlying representation of RV Cs must come from a bi-clausal source because some RVCs can undergo the BA and BEi constructions, while others cannot. Transformational-generative grammar, however, was not the sole attention of linguists working on Chinese. The late seventies and early eighties witnessed the development of the functional approach to Chinese linguistics. The representative example was Li and Thompson ( 1981). This grammar revived the interest of linguists in the functional perspective on linguistics while transformation-generative grammar occupied so much of linguists' attention then. Along the functional-cognitive analysis, L Huang (1980), following Chu (1978), analysed the resultative verb compound by the concept 'conceptual closeness', contrasting with the structural and transformational approaches. Since Chomsky (1981), much linguistic work related to Chinese has been done .in the Government-Binding (GB) framework. J Huang (1982) was a very influential 7

work on both linguistic theory and Chinese linguistics. Harlow (1989) remarked that J Huang's work was the 'catalyst' for the subsequently large numbers of works done on Chinese within GB or other theoretical frameworks. One such work, for our purposes, was Sybesma (1992) on the ha-construction. While J Huang (1982) was primarily concerned with establishing Chinese phrasal structures, especially the X-bar theory of Chinese, his later works ( eg 1988, 1994, 2010) have contributed to the syntax and semantics of verb compounds. The more elaborated theory on resultative verb compounds seems to be his work with L L-S Cheng (1994), which addresses the verb classes and argument structure of Chinese verb compounds. Still within the GB framework, Y Li (1991) was the first articulated theory on Chinese V -V compounds. Li developed a theory of argument structure of V -V compounds out of Grimshaw' s ( 1990) argument structure theory. He revised his theory later in his papers (Li 1993, 1995). Li (1995) even proposed a Cause role in the argument structure of resultative verb compounds. Other recent research is the reaction to Li's proposals. Cheng and Huang (1994) is an example at hand. Within the framework of Lexical-functional grammar, Huang and Lin (1992) proposed a theory of argument templates based upon Dowty (1991) in resultative verb compounds. Since Cheng and Huang (1994) is rather influential in this area, we will discuss their proposal in more detail. Huang (2010) is the most recent representative work in this area. Research on resultative verb compounds has not ceased here, and many, if not most, researchers as above are still working and revising their theories of Chinese verb compounds. Moreover, since there has been a recent interest in the investigation of complex predicates and secondary predications across languages, linguists-,working on Chinese will consider verb compounds in terms of more recent proposals and theories. This chapter aims at providing the necessary background of the previous approaches and simultaneously, reviews the more influential and representative work pertaining to our study. This survey tends to be a little long but the backgrounds are necessary. Many examples come from the published sources. 2.1 Traditional Study As mentioned in the Introduction, Chao (1968) is the first comprehensive treatment of Chinese grammar in English and represents the tradition of structural linguistics prevalent in America. However, the structural tradition of Chinese linguistics had begun long time ago in China. This was witnessed by the earliest contribution to Chinese grammar, which is generally regarded as the first Chinese grammar, Ma's 8

Grammar ( 1898), named after the author Ma Jianzhong. Ma borrowed the grammatical model of European languages then current in Europe. It is often criticised to commit the mistake of squeezing Chinese into European languages, for Chinese is a markedly different from European languages in syntax, semantics and morphology, etc. Later Chinese grammarians realised the problems with Ma's grammar, and started to find another grammatical model. When American structuralism was pervasive after Bloomfield's classic work Language (1933) had been published, Chao produced the classic mammoth work on Chinese (1968). Among the many topics treated in Chao ( 1968) is the one of verbal compounds. Verbal compounds in Chinese occupy most of the vocabulary of the language. Chao distinguishes several types of compounds, for example, verb-object compounds, verb-complement compounds, and so on. The verb-complement construction in Chinese is the most interesting type, and received quite a large degree of attention in Chao ( 1968). Chao distinguishes verb-complement (V-R) constructions into five types of complements, which are resultative complements, phase complements and aspect suffixes, intensifying complements, potential complements, and directional complements. Resultative verb complements are further classified into four types, which are cause, achievement, direction and phase. The five types of complements can also be grouped into two major categories, one of which is free predicative complement, and the other of which is bound verb-complement compounds. ( 1) John ZOU de hen man john walk de very slow John walks very slowly. (2) John chi-bao le john eat-full asp John has eaten full. Examples (1 and 2) exhibit the free predicative complement and bound verb-complement, respectively. Apart from the distinction between being free and bound, the first example is different from the second in that the second example has a canonical resultative compound, whereas what the first example has can be a phrase or a compound, depending on one's theory. Moreover, Chao's classification of free and bound complements is not entirely sufficient to capture the difference between these 9

two constructions. Resultative verb compounds as in example (2) are not always bound, for instance: (3) John chi le john eat asp John has eaten. (4) John bao le john full asp John is full. The separability of the RVC chi-bao 'eat-full' explains that RVCs are not necessarily bound, as Chao originally suggested. Furthermore, the RVC chi-bao 'eat-full' can be expanded as a verb-predicative complement construction, as in (5): (5) John chi john eat de de hen bao very full John is very full as a result of eating a lot. As a consequence, the free-bound distinction is not a satisfactory pr_inciple accounting for the difference between these two kinds of constructions. Most of verb-complement verb compounds are also marked by phonological features, expandability and the ability to occur with the aspect marker -le. Chao only mentioned these in passing but did not intend to uncover the underlying structure of resultative verb complements. Nevertheless, the different meanings of some resultative verb complements were observed by Chao, for example: (6) John ba fan zhu-lan le john ba nee cook-soft asp John cooked the rice soft. (7) John ba fan chi-bao le john ba nee eat-full asp John is full as a result of eating rice. Chao noted that the result state of the resultative verb complement can refer to the object and occasionally refer to the subject of the entire verb compound. In example ( 6), Zan 'soft' refers to the object fan 'rice', while in example (7), bao 'full' refers to the 10

subject of the whole compound John. This is exactly the issue that subsequent linguists (egLi 1991, 1993; Cheng and Huang 1994) take up. We shall consider this in more detail later. Chao's application of structural linguistics to the Chinese language was fully displayed in his (1968) grammar, which prepared an access to Chinese linguistics for the subsequent generations of linguists working on Chinese. This is inter alia reflected in his treatment in Chinese resultative verb complements. Chao divided verb-complements into free/bound complements and the classifying principle is not adequate to capture the entire classes of verb complement constructions. As has been pointed out, some resultative verb compounds can be separated as chi-bao 'eat-full' can allow the component verbs to occur individually and separately, as in examples (3 and 4) above. Although Chao provided the basic classification of resultative verb complement constructions, there was no discussion about the derivations of the structures, while Chomsky's transformational-generative grammar was being developed in full swing then. The issue of derivations and deep structure clouded on much of the linguistic work and debate, and this was also taken up by linguists researching on Chinese. The next section considers the resultative verb complement constructions within the transformational-generative grammar framework. 2.2 Transformational Approaches In the heyday of transformational-generative grammar, especially after Chomsky's Aspect was published, debates about deep structure surrounded almost the whole linguistic community ( cf Harris 1994) . ...English was not the sole focus of inquiry though much work then was done on English. Hashimoto (1966) should be the first attempt to analyse Mandarin Chinese resultative verb complement constructions systematically. Hashimoto argued that the underlying structure of resultative verb complements was derived from two clauses, an underlyingly bi-clausal analysis. However, . some resultative verb compounds cannot be analysed as derived by two clauses because of semantic constraints. This led Thompson (1973) to propose lexical rules that those resultative verb-complement compounds that cannot be transformationally derived must list in the lexicon. Like Hashimoto, Lu (1977) argued for a bi-clausal analysis of resultative verb compounds. Lu drew evidence from the BA and BEI constructions, for some RVCs can allow these two constructions, but some cannot. As it happens, linguistic research seems to go around the circle at times. 11

2.2.1 Hashimoto (1966) The first attempt to apply transformational grammar to the analysis of resultative verb complement construction was made by Hashimoto (1966). Hashimoto's treatment ofRVCs was based upon a two-sentence analysis. The concept of bound forms favoured by traditional grammarians like Chao was rejected by Hashimoto, and hence she was not so interested in the RVCs as they occurred in the potential form, for example: (8) John ting-bu-jian Mary john hear-not-see mary John didn't hear Mary. (9) John kan-bu-qi Mary john look-not-up mary John looked down on Mary. Hashimoto was, however, concerned with the differences between resultative complements and descriptive/degree complements, and demonstrated that these two types of constructions shared similar surface structures. (10) John pao de lei le john run de tired asp John ran himself tired. (11) John pao de Imai john run de quick John runs quickly. Example (10) has a resultative reading, while (11) has a descriptive/degree reading. Although resultative and descriptive/degree complements are similar in their surface structures, their underlying structures are not identical. As Hashimoto argued, the resultative verb complement can be derived from an underlying sentence. Thus, example (10) was said to be derived from two sentences as the following: (12) John pao (13) John lei le However, the descriptive/degree complements are not generated from two sentences, and so (11) cannot be derived as (10). 12

(14) John pao (15) * John kuai Another piece of evidence that resultative complements differs from descriptive/degree complements in their underlying structures comes from the fact that descriptive/degree complements can be realised by a whole clause, rather than a verb compound. This is demonstrated by the example below: (16) John Im john cry de de Mary dou bu mary also not John cried so that Mary couldn't sleep. neng shui can sleep As also noticed by traditional grammarians, resultative complements can contain a whole sentence or clause but descriptive complements never can. In the case of ( 16), the resultative complement is the clause Mary dou bu neng shui 'Mary can't sleep'. This fact led Hashimoto to postulate two different underlying structures for resultative and descriptive complements, and supported her analysis that resultative complements were derived transformationally from two sentences. Let us consider how the bi-clausal analysis works in a transformational fashion. One of the examples Hashimoto treated transformationally can be observed in the following sentence: (17) John la-kai-le men john pull-open-asp door John pulled the door open. The sentence ( 17) was treated as being derived from the matrix sentence containing an embedded clause, as shown below: (18) John la men john pull door John pulled the door. (19) Men kai-le door open-asp The door was open. The surface structure of John la-kai-le men 'John pulled the door open' is derived from (18 and 19) by transformational rules, as illustrated in the phrase-marker (20). (20) 13

shows that the second component verb kai 'open' is analysed as a predicate of the underlying resultative complement sentence or clause, and that the transformational rule is applied so that it deletes the identical noun men 'door' of the lower clause before the predicate kai 'open' is moved to the position preceding la 'pull'. As a consequence, the surface structure of example ( 17) is derived, John la-kai-le men 'John pulled the door open'. The syntactic irregularities of some resultative verb complements further confirm the transformational analysis that Hashimoto proposes. Hashimoto observes that there is syntactic idiosyncrasy between the resultative verb compounds chi-baa 'eat-full' and chi-wan 'eat-finish'. Although these two compounds are resultative compounds, they are considered to have two different deep structures (L Huang 1980). (21) (22) (20) S ~ NP VP //\ V NP S A NP VP I \ V ASP I I john la men men open le john pull door door kai asp John chi-bao-le fan john eat-full-asp nee John was full as a result of eating rice. John chi-wan-le fan john eat-finish-asp nee John ate up the rice. 14 (Deep structure of ( 17))

The sentences (21 and 22) were observed to have the deep structures, as shown in phrase-markers (23 and 24), respectively. The underlying arguments predicated by the second component verbs are different in these two structures. In (21), it is the subject of the whole sentence that is predicated of by the second component verb baa 'full', that is, John was full. However, the second component verb of the resultative compound in (22) wan 'finish' is predicated of the object of the entire sentence, so rice was finished. This syntactic and semantic asymmetries exhibited by resultative verb compounds also persuaded Hashimoto to rely on transformational analysis. (23) S ~ NP VP //\ V NP S /\ NP VP I \ V ASP I I jahn chi fan jahn baa le john eat nee john full asp (Deep structure of (21)) 15

(24) S ~ NP VP //\ V NP S A NP VP I \ V ASP I I john chi fan fan wan le john eat nee nee finish asp (Deep structure of (22)) The differences between these two resultative compounds are reflected not only in the argument that can be predicated of, but also in their occurrence in the BA and BEI constructions. The resultative compound chi-baa 'eat-full' cannot occur in the BA and BEI constructions while the one of chi-wan 'eat-finish' can. (25) a. b. (26) a. b. * Fan nee bei bei John chi-bao-le john eat-full-asp * The rice was eaten full by John. * John ba fan chi-bao-le john ba nee eat-full-asp * John ate the rice full. Fan bei John chi-wan-le nee bei john eat-finish-asp The rice was eaten up by John John ba john ba fan nee John ate up the rice chi-wan-le eat-finish-asp 16

Hashimoto explained that in the underlying structure the passive BEI and BA transformations are not allowed when both the matrix subject and the subject of the embedded clause are identical. This explanation only makes sense if we believe in the transformational analysis of resultative verb compounds that Hashimoto proposed. Although this analysis seems to work well in the resultative verb compounds in question, the heavy syntactic flavour of Hashimoto's approach provided insufficient semantic properties of lexical items in resultative compounds. Many resultative verb compounds cannot really be derived via transformation, and this issue was taken up by Thompson (1973), who adopted a lexical view on this matter. 2.2.2 Thompson (1973) Thompson (1973) argues for a lexically derivative approach to resultative verb compounds. First of all, Thompson defines a resultative verb as a compound verb which consists of two parts, the first part denoting an action and the second part the result of the action. For example, those RVCs like la-kai 'pull-open' and chi-bao 'eat-full', as we have seen earlier, fall under Thompson's category. Apart from that, Thompson entertains that the ability for an RVC to occur in the potential form plays a crucial role in characterising an RVC. In other words, whether a compound is a resultative verb compound relies on its ability to occur in the potential mode. There are two potential forms, one positive, the other negative. The positive is formed with de meaning 'able', and the negative with bu meaning 'not'. If we apply the above examples la-kai and chi-bao to it, we will have the positive: la-de-kai, chi-de-bao, and the negative: la-bu-kai, chi-bu-bao. In response to the transformational approach to the derivation of resultative verb complement constructions, Thompson proposes that resultative verb compounds are derived by lexical operations rather than transformational rules. The evidence comes from those many resultative compounds that cannot be derived transformationally. Resultative verbs, in Thompson's view, are morphologically complex verbs derived via some lexical rules. Thompson admits that if all resultative verb compounds were like the one of la-kai 'pull-open' in John la-kai-le men, which can be derived transformationally from two underlying sentences, (as in (17) proposed by Hashimoto), there would not be any alternatives to analysing RVCs. However, some RVCs will produce nonsense under the bi-clausal analysis, for instance: 17

(26) John pao-kai-le john run-open-asp John ran away. Under Hashimoto' s analysis, the deep structure of the sentence should be like the phrase marker of ( 17) having two underlying sentences, ie John pao-le 'John run-asp' and* John kai-le 'John open-asp'. The embedded sentence *John kai-le is unacceptable if not ungrammatical because it is semantically anomalous and cannot make any sense in (26). 1 Further evidence that Thompson draws to support her lexical analysis originates from the fact that some endings that often attach to resultative verbs cannot be derived by transformational rules. These endings include dao and zhao, etc. (27) John mai-dao john buy-arrive John bought the book. shu le book asp Thompson maintains that examples Wee (27) cannot be derived transformationally from (28) for the same reason as (26). (28) a. John mai shu john buy book John bought the book. b. * Shu dao le book arrive asp * The book arrived. Note that (286) alone could make sense in another context but is not acceptable on this resultative reading as part of the resultative compound of (27). A further type of RVC that cannot be generated by transformational rules observed by Thompson is the type of RVC with directional complements with a metaphorical but not literal interpretation. These directional complements also behave 1 It seems that to regard kai in pao-kai as meaning 'open' is a little odd. It is more natural to simply interpret it as meaning 'away', and then the bi-clausal analysis as Hashimoto proposes can still be applied. However, other RVCs may not be so interpreted as to avoid the problem. 18

like the above suffixes such as dao 'arrive' in this manner. Example (29) is not able to be generated from example (30). (29) (30) a. b. John xie-shang-lai-le john write-up-come-asp neige z1 that character John recalled how to write that character. John xie ne1ge z1 john write that character John wrote that character. * neige zi that character shang-lai-le came-up-asp * That character came up. On the basis of the evidence of these observations, Thompson argues that resultative verb compounds must be derived by lexical operations, as opposed to the derivation by transformational processes. Thompson proposes the RV-formation rule such as (31). (31) V + V • [V-V] RV action intransitive action la + kai • la-kai This lexical rule can generate the RV compound by combining the action verb with an intransitive verb, for example, la-kai 'pull-open'. Not only does Thompson propose a rule for RV-formation, a lexical rule for the derivation of the potential forms is also formulated to avoid the ad-hoe nature of creating transformational derivation for potential forms of RVCs. According to Thompson, no well-motivated underlying structure can represent the same meaning of a resultative verb with a potential form without an ad-hoe rule applying to it. The potential forms have the following lexical derivation: (32) [V-V] RV la-kai RV • • [V-de-V] Rv [V-bu-V] Rv la-de-kai la-bu-kai 19

This rule simply inserts the positive or negative infix into a resultative compound to produce a potential form of resultative compound. The rules for potential forms and the RV-formation rules together make up a small set oflexical rules for resultative verb compounds. However, those that cannot be explained by these rules, Thompson maintains, must be listed in the lexicon. On the one hand, there are resultative complements whose semantics can be predicted and they are derived by lexical rules. Those resultative complements whose semantics cannot be predicted compositionally from the constituent verbs are not generated by transformations but are listed in the lexicon. In the classical transformational grammar as outlined in Aspect, there are three components of the grammar, ie categorial, transformational and lexical components. Thompson's lexical theory could simplify the first two components but it could complicate the lexical component. This is not desirable particularly when viewed from a contemporary linguistic angle, which requires the grammar to be as simple as governed by a few principles and is simultaneously rich enough to account for the variations among languages. The narrowly defined notion of resultative compound also sets a drawback to Thompson's lexical approach. Thompson's definition of a resultative verb compound would exclude many others which are typically regarded as resultative compounds by grammarians. For example, resultative compounds like e-bing 'hungry-ill' would not be accepted as such because it cannot occur in the potential mode, * e-de/bu-bing (recall that the ability to do so is the major characteristic of resultative verb compounds in Thompson's view). Another example is ku-pao 'cry-run', *ku-de/bu-pao. On the other hand, the definition would also include those which some linguists like Lu ( 1977) may not considered to be resultative compounds, for instance, directional complements, eg tiao-shang-lai 'jump-up-come: jump up'. Thompson ( 1973) was able to observe the weaknesses of the transformational approach to Chinese resultative compounds and to make some improvement on the derivational problems. However, due to Thompson's narrowly defined notion of resultative compounds and the heavy dependence on lexical operations, Lu (1977) suggests that a modified definition of resultative compounds with a functional perspective should be adopted and that a transformational analysis would be more plausible based upon this definition. It is Lu's approach to which we shall turn our attention. 20

--2.2.3 Lu (1977) Lu ( 1977) discusses two types of verb compounds, resultative and directional, both of which have been regarded as belonging to the same category ' resultative' by linguists like Thompson (1973) and Li and Thompson (1981). Lu's distinction between resultative and directional verb compounds mainly relies on the definition he gives to resultative verb compounds. Contrasting to the traditional notion of RVC, Lu (1977: 28) defines a resultative verb compound in terms of its function in a sentence as in (33). (33) an RV is a V-V construction in a sentence where the second verb indicates a result caused by the action or the process represented by the first verb At a glimpse of this definition, it does not seem to be much different from Thompson's. However, this definition differs from Thompson's in that it does not rely on the ability to enter the potential mode to determine the status of a resultative verb compound. Recall the ability to appear in the potential form is almost a decisive characteristic of RVC in Thompson's terms. However, Lu does not resort to this characterisation. Another major difference between Lu and Thompson is that Lu first separates directional verb compounds from resultative verb compounds. Lu also differs from Thompson in the derivation of resultative verb compounds. Although Thompson offers a plausible lexical account on the derivation of resultative verb compounds, Lu argues for a transformational account similar to that of Hashimoto (1966). The following represents Lu's analysis (1977: 284): (34) S ~ NP VP /A V (NP) S [+action/process] /\ NP VP I V I [+intransitive] 21

As we can see, this analysis, similar to Hashimoto (1966), also assumes a bi-clausal underlying representation of resultative verb compounds. Lu suggests a movement rule called Verb Raising to raise the verb in the embedded clause to the matrix sentence to form an RV with the matrix verb. Unlilce the previous transformational approach, Lu observes that the selectional restrictions between the accusative NP and the second component verb of the RV construction are consistent with those between the nominative NP and the correspondent simple verb. (35) John la-kai-le men john pull-open-asp door John pulled the door open. (36) Men kai-le door open-asp The door opened. (37) * John kai-le john open-asp * John opened. Examples (35 and 36) show these relations between the second component verb kai 'open' and the argument NP men 'door'. In other words, kai 'open' predicates of men 'door' in both sentences, despite the fact that men 'door' in (35) is the 2bject of the sentence, but men 'door' in (36) the subject of the sentence. Put in modern terms, kai 'open' exhibits unaccusativity (cf eg Perlmutter 1979, Levin and Rappaport Hovav 1995), predicating of the internal argument of these two sentences. Although Lu does not describe these syntactic and semantic properties of kai 'open' explicitly, his observation is still significant pertaining to both his own analysis of resultative verb compounds and the understanding of the nature of resultative verb compounds in general. In fact, later works discuss how the component verbs in a resultative verb compound predicates of the arguments. Furthermore, in spite of the grammaticality of (35 and 36), (37) is not grammatical because kai 'open' does not hold the kind of relation with the argument NP John as it does with the argument NP men 'door' in (35 and 36). That is, kai 'open' does not predicate of the agent John. 22

Further syntactic constraints on sentences with a resultative verb compound are also observed. Some sentences with RVCs are allowed to undergo the BA and BEi constructions, but some cannot be transformed in these constructions. (38) a. b. c. (39) a. b. c. John la-kai-le men john pull-open-asp door John pulled the door open. John ba john ba men la-kai-le door pull-open-asp John pulled the door open. Men bei door bei John la-kai-le john pull-open-asp The door was pulled open by John. John he-zui-le jiu john drink-drunk-asp wine John was drunk (by drinking wine). * John ba jiu he-zui-le john ba wine drink-drunk-asp * John drank the wine drunk. * Jiu bei John he-zui-Ie wine bei john drink-drunk-asp * The wine was drunk by John drinking it. On the surface, the syntactic structures of both (38 and 39) are identical but only (38), but not (39) can undergo the BA and BEi constructions. Suppose the surface structures and deep structures of the RVCs in these sentences are basically identical, we cannot, Lu argues, explain why some RVCs with an object can undergo BA and BEi constructions, while others with a more or less identical object cannot. Thus, Lu concludes from this evidence that these RVCs must have a bi-clausal origin. This conclusion is like Hashimoto's. As we have seen, Lu is concerned not merely with the syntax of resultative verb compound, but also with its semantics. Another semantically significant and interesting 23

issue is also uncovered by Lu. Based upon the above examples, Lu further observes that sentences like (38 and 39) have the same presuppositions. Consider the following sentences: ( 40) a. John la-le men, men ca1 kai de john pull-asp door, door until-then open par The door did not open until John pulled it. b. Men kai-le shi yinwei John la de door open-asp be becausejohn pull par The door opened because John pulled it. ( 41) a. John he-le jiu (ta) ca1 ZUI de john drink-asp wme (he) until-then drunk par John was not drunk until he drank wine. b. John zui-le shi yinwei (ta) he-le jiu john drunk-asp be because(he) drink-asp wine John got drunk because he drank wine. As we can see from the English interpretations of the sentences, we will notice that a particular kind ofrelations hold between the pair of the sentences (a and b ). Lu observes this cause and result relation exhibited in the presuppositions of the sentence, which is realised through the resultative verb compounds. There are two clauses in each of these sentences and one clause indicates a cause and the other a result. Also, the non-existence of the result prior to the cause is presupposed in the clauses. Likewise, the negation of the sentences will still share the same presuppositions. This is exemplified in (42 and 43). (42) (43) John mei-you john neg la-kai pull-open John didn't pull the door open. he-zui men door jiu John mei-you john neg drink-drunk wme John didn't get drunk. 24

These two sentences still presuppose that in ( 42) the door was still closed and John was not drunk. Since John didn't pull the door, the door didn't open; since John didn't drink, he didn't get drunk. The cause and result relation is still assumed. This observation of cause-result relation of resultative verb compounds has an enormous influence on the subsequent work on this aspect of syntax. For example, Li (1991, 1993) discuss this semantic relation between the resultative verb compound and its arguments. Li (1995) takes the issue even further to assume a Causer role in the verb compound, following work such as Grimshaw (1990). We shall defer our discussion of this aspect to a later section. Lu's analysis of RVCs differs not only from the previous transformational approach, but also from the lexical approach like Thompson (1973) in assuming that a rewriting rule in the categorial component derives the potential forms of RVCs. Lu observes that the two potential forms are mutually exclusive, that is, they are in complementary distribution. ( 44) a. b. * John la-de-bu-kai men john pull-pot-open door * John can and cannot pull the door open. * John la-de-bu-kai-le men john pull-pot-open-asp door * John can and cannot pull the door open. As Lu notices from the examples, the aspectual marker le is not compatible with the potential form. When they appear together, the sentence is not grammatical. Based upon the characteristics exhibited by the potential forms occurring in RV Cs, Lu devises a rewriting rule to account for the derivation of ASP and POT. This is illustrated in ( 45) below: ( 45) a. AUX • b. POT • ASP POT pot positive pot negative 25

Lu suggests that the rule can better capture the nature of potential forms and is superior to Thompson's lexical potentialisation rule, which fails in Lu's view to provide an accurate characterisation of potential forms occurring in RV Cs. In summary, Lu's analysis shares the transformational derivation with Hashimoto's but differs in capturing the semantic features exhibited by the RV construction, and differs from Thompson (1973) in assuming an underlying bi-clausal source ofRVCs. Lu's definition of RV is also different from Hashimoto, Thompson and others, as it pertains to a functional perspective. 2.2.4 Summary Thus far, we have considered the important analyses of the Chinese resultative verb complement construction within the framework of transformational grammar prevalent in the 60s and 70s. Hashimoto (1966) could be considered to be the catalyst for the emergence of subsequent studies on RVCs, and she was among the first to apply the transformational grammar to study of Chinese syntax. As syntacticians were concentrating upon the derivation of deep structure of sentence, Hashimoto attempted to analyse Chinese resultative verb complement constructions in terms of transformational rules. Transformations can capture that fact that the resultative complement is derived from two sentences, as (17) can be derived from (18 and 19), repeated here as ( 46): ( 46) a. b. John la-kai-le men john pull-open-asp door John pulled the door open. John la men john pull door John pulled the door. c. Men kai-le door open-asp The door was open. However, as we have already seen, Thompson (1973) demonstrates that Hashimoto's transformational account assuming a bi-sentensial analysis is not tenable because some RVCs, for example, pao-kai 'run-open: run away' in John pao-kai-le, cannot be derived from an underlying two-sentence representation. Instead, Thompson proposes that 26

---RVCs are derived by lexical operations and a set oflexical rules. In Thompson's model the lexicon is responsible for the resultative verb compounds whose semantics cannot be predicted. The advantage of adopting Thompson's account is that it could reduce the load of transformational and categorial components of the grammar, but at the same time make the lexical component more complicated. Furthermore, Thompson's account could not provide us with an adequate analysis of the resultative complement construction involving a clause like our example (16), repeated as ( 47) below: (47) John ku john cry de de Mary dou bu mary also not John cried so that Mary couldn't sleep. neng shui can sleep Example (47) contains two subjects, John and Mary, and two predicates ku 'cry' and bu neng shui 'cannot sleep' . According to Thompson, we would expect the predicate to be listed in the lexicon since the semantics of the predicate cannot be derived by lexical rules. However, it is obvious that the sentence consists of a matrix sentence and an embedded sentence. ( 48) a. b. John ku john cry John cried. Mary dou bu mary also not Mary couldn't sleep. neng shui can sleep It seems that it is not possible to account for this in Thompson's terms. Thus, Thompson's approach only applies to some canonical RVCs, but not resultative complement constructions like the above example. As mentioned earlier, Thompson's lexical approach cannot even account for the syntactic irregularity why some RV Cs can undergo BA and BEI constructions while others cannot. This was exemplified in examples (38 and 39), repeated here as ( 49and50): ( 49) a. John la-kai-le men john pull-open-asp door John pulled the door open. 27

b. c. (50) a. b. c. John ba john ba men la-kai-le door pull-open-asp John pulled the door open. Men bei door bei John la-kai-le . john pull-open-asp The door was pulled open by John. John he-zui-le jiu john drink-drunk-asp wme John was drunk (by drinking wine). * John ba jiu he-zui-le john ba wme drink-drunk-asp * John drank the wine drunk. * Jiu bei John he-zui-le wine bei john drink-drunk-asp * The wine was drunk by John drinking it. However, the transformational analysis is not entirely satisfactory. As L Huang ( 1980) points out, some RVCs like ning-gu 'congeal-solid: solidified' cannot be derived transformationally by two underlying sentences. (51) a. b. c. Heliu ning-gu-le nver congeal-solid-asp The river solidified. Heliu ning-le river congeal-asp * Heliu gu-le river solid-asp While ning-gu 'congeal-solid: solidified' fulfils the requirement of a resultative verb compound since the first component verb indicates the action and the second result, the RVC cannot be derived transformationally. The fact that (51c) is not grammatical suggests that the transformational analysis cannot fully capture the underlying sources 28

of resultative verb compounds, except that there are more constraints imposed on RVCs. A more obscure example of RVC kan-jian 'look-see' provides further evidence against the transformational analysis. (52) a. b. c. John kan-jian-le Mary John look-see-asp mary John saw Mary. ? John kan-le John look-asp Mary, John ca1 mary John until-then ? John didn't see Mary until John saw Mary. jian see Mary de mary par ? John jian John see Mary shi mary be yinwei John kan-le Mary becausejohn look-asp mary ?John saw Mary because John saw Mary. According to Lu (1977), the (band c) sentences should be the underlying sentences for (a), but they cannot be because they are anomalous. Despite the fact that kan 'look' indicates the action and jian 'see' the result, they should not be considered as indicating two separate events. In fact, speakers will consider kan-Jian to denote one single event only. There is a close relationship between these two component verbs, and it is so close that speakers will only recognise one single event, just the event of seeing. It is argued that kan-jian has been lexicalised in the lang1Jage, and is always used together to mean see. If these two verbs are used separately, they will have some other meanings. This kind of close relationship is referred to conceptual closeness by linguists like Chu (1978) and L Huang (1980). We shall discuss this concept in the next section of functional approaches. 2.3 Functional Studies (Chu 1978 and L Huang 1980) In the last section, we considered the strengths and weaknesses of the treatments of RVCs within the transformational-generative grammar framework. This section discusses the treatments ofRVCs done in the functional perspective. The first functional study of resultative verb compounds is L Huang (1980), which I refer to as functional because it attempts to explain the nature of resultative verb complement construction in cognitive functional terms. The second major functional linguistic study of Mandarin 29

Chinese RVCs is Li and Thompson (1981). Li and Thompson contribute not only to the study ofRVCs, but also to Chinese linguistics in general. In the above discussion, we considered the problems with the previous approaches to resultative verb complement constructions. L Huang (1980) also notices the inadequacy of the structuralist and formalist analyses, and following Chu (1978), proposes an analysis of RV Cs base upon the concept of conceptual closeness as follows (L Huang 1980). The postulation of conceptual closeness was first conceived by Chu (1978) to attempt to solve some syntactic problems in Chinese. If an event entails another, these two events are considered to be conceptually close to each other in the mind of the speaker. Such a relation can be said to be conceptual closeness. In terms of syntactic realisation, this concept can be amalgamation of some elements. If there is a close conceptual relation between the elements, there will be fewer grammatical markers to indicate the relationship. Compounds are usually the realisation of the closest relation. Resultative verb complements are observed to exhibit the kind of variations that can be captured by conceptual closeness. The examples in Table 1 illustrate the variations shown in resultative verb complement constructions, which can be explained by means of the concept of conceptual closeness (L Huang 1980). VERB COMPLEMENT V-VRC POTENTIAL V-DE COMPLEMENT ning gu ning-gu *ning-de-gu *ning-de (hen) gu congeal solid solidified congeal-pot-solid congeal-de (very) solid tui fan tui-fan *tui-de-fan *tui-de (hen) fan push turn over overthrow push-pot-turn over push-de (very) turn-over kan jian kan-jian kan-de-jian *kan-de (hen) jian look perceive see look-pot-perceive look-de (very) perceive la kai la-kai la-de-kai *la-de (hen) kai pull open pull-open pull-pot-open pull-de (very) open tui dao tui-dao tui-de-dao tui-de (*hen) dao push fall push-fall push-pot-fall push-de (very) fall chi bao chi-pao chi-de-pao chi-de hen pao eat full eat-full eat-pot-full eat-de very full 30

pao lei run tired e bing hungry ill e shou hungry thin pao-lei run-tired e-bing hungry-ill ?e-shou hungry-thin *chi-xiao pao-de-lei run-pot-tired *e-de-bing hungry-pot-ill ?e-de-shou hungry-pot-thin *chi-de-xiao pao-de (hen) lei run-de (very) tired e-de (*hen) bing hungry-de (very) ill e-de (hen) shou hungry-de (very) thin chi-de (*hen) xiao eat laugh eat-de (very) laugh eat-laugh eat-pot-laugh ku shui-bu-zhao *ku-shui-bu-zhao *ku-de-shui-bu-zhao ku-de(hen)shui-bu-zhao cry sleep-not-par cry-sleep-not-par cry-pot-sleep-not-par cry-de(hen)sleep-not-par ku dajia shui-bu-zhao shui-bu-zhao *ku-dajia-shui-bu-zhao *ku-de-dajia-shui-bu-zhao (John)ku-de-cry everyone sleep-not-par cry-everyone-sleep-not-par cry-par-everyone-sleep-not-par (John)cry-de-sleep-not-par Table 1: a continuum of'resultativeness' (adopted from L Huang 1980: 24) The examples in Table 1 from L Huang (1980) show the degrees of grammaticality of the various resultative complement constructions. Except the last four examples, all others can occur as resultative verb compounds. However, the first two examples, ning-gu and tui-fan are not able to occur in other complement constructions. The fact that they must occur together and they cannot be inserted any words in between suggests that they share a mutual dependence for the semantics of the entire construction. Besides, they seem to have become lexicalised and can be treated as a single lexical item, ninggu and tuifan. Nevertheless, since ninggu and tuifan are so semantically close to each other, they do not allow the potential de!bu insertion, which Thompson regards as the main characteristic of resultative verb compounds. Nor do they can appear in the de-complement clause. Kan-jian 'look-perceiver' and la-kai 'pull-open' cannot be used in the de-complement clause but it is fine with other constructions. The examples that can occur in all different types of constructions are chi-baa 'eat-full' and pao-lei 'run-tired'. RVCs like these two can be considered to be more canonical that others. The last four verbs plus complements show their inability to form resultative verb compounds. They only occur in the V -de-complement construction. The compound e-shou 'hungry-thin' is arguable to be an RVC, but the further down the 31

table the compound is the less acceptable it is. Again, according to Thompson's characterisation of RV Cs, many of them cannot be regarded as RV Cs. The examples, more importantly, show a continuum of'resultativeness' ranging from RVCs unable to occur in the potential and de-complement forms at one extreme, to those able to be used in all the constructions, at the other. Thus, the items on the left hand need no conceptual closeness, and this is where V-de-complement appears. On the contrary, those items on the right need absolute closeness and this is where lexicalised V-V patterns can occur. In between are those that require various degrees of mutual dependence or conceptual closeness. If we draw a line along which the above examples can be seen at different points, we will have something like the following (L Huang 1980: 29): V-DE-COMPLEMENT Im-de dajia shui-bu-zhao ku-de shui-bu-zhao e-de shou pao-lei LEXICALISED V-V kan-jian ning-gu tui-fan This line of continuum comparable to the table containing all these examples we have seen expresses the same idea that there are degrees of grammaticality of the combination of verb and complement occurring in different types of complement constructions. This analysis is not only sufficient for its own sake, but it also tries to explain and complement with the transformational approaches. It can be stated that the constructions appearing at the left extreme of the continuum, that is, V -de-.?omplement end, can be generated by transformational rules as proposed by Hashimoto and Lu, while those appearing at the right extreme, that is, lexicalised V-V end, can be derived by lexical operations as proposed by Thompson. As for those items in between, it may depend upon the individual's conceptual closeness. If an individual assigns an item to a closer position to the V-de-complement extreme, the assigned item may be considered to be generated by transformational rules. Lilcewise, if an item is assigned to a closer position to the lexicalised V-V end, it may be said to be derived by lexical operations. It should be noted, however, that speakers may assign the same item to a different position along the continuum according to some slightly different intuitions since individuals may have different conceptual structure. As mentioned earlier, Li and Thompson (1981) (henceforth LT) is a major study of Chinese grammar from a functional perspective. Among the many topics treated in LT is the time-honoured topic of resultative verb compound. LT shares many of the 32

--views on RV Cs held by Thompson ( 1973), and one of which is the entertainment that the ability to occur in the potential form characterised most of the RV Cs, and that the class of RV Cs subsumes a subclass of verb compounds, directional verb compounds. In the following, we shall describe the major features of RVCs and directional verb compounds as presented in LT. Some characteristics are associated with the RVC. First, RVCs allow the adverbs de (derived originally from dei meaning 'get/obtain') and bu 'not' to be inserted within the two verbs to indicate positive and negative possibilities, for instance: (53) (54) ta tiao-de-guo-qu he jump-get-cross-go He could jump across. ta tiao-bu-guo-qu he jump-not-cross-go He couldn't jump across. This construction is referred to as potential form (Li and Thompson 1981). As Anderson (1985) notes, the structures and the interpretations ofRVCs are comparable to those of phrasal constructions. However, RVCs have a special formation process as just illustrated, namely, [V-de-V]v and [V-bu-V]v. These have no parallel phrasal patterns. It is pointed out that although de and bu can be translated into the English can and cannot, it is more appropriate to translate them as achievable and unachievable. (53) means that he jumped and achieved getting across, and (54) means that he jumped but failed to achieve getting across. The semantic properties of de and bu can be explained in terms of their functions as those of adverbials. That is the matter of scope. Like adverbs that can have their scope over the following elements, de and bu work in the same manner, where the meanings of 'achievable' and 'unachievable' are included as part of the RVC that follows de and bu, but not the preceding part of the RVC. Thus, (53) conveys the meaning of achieving getting across by jumping and (54) the meaning of not achieving getting across by jumping. 3 2 It should be distinguished from the nominative particle de, as in zhong tian de 'farmer', though they share the historical origin. They are pronounced the same in modern standard Chinese, however. 3 There are, however, cases ofRVC, where de and bu insertion would yield something unnatural if not ungrammatical, eg ?ta e de bing le he-hungary-de-ill-asp 'He got ill from hunger', and ?ta zuo bu lei he-walk-not-tired 'He can't get tired from walking'. Li and Thompson (1981) note that these are not cases of exception to the general rule, but their strangeness seems to come from the fact that the meanings of de and bu implies some effort on the subject, and people don't normally make an effort to achieve such undesirable results as to getting sick and tired. 33

The semantic functions of de and bu naturally remind us of the auxiliary verb neng 'can'. Although their meanings are similar, they are not always interchangeable in a resultative construction. We have noted that in examples (53 and 54) the subject performs the action of jumping. The difference between them is that in ( 53) the subject is able to get across, but in (54) the subject is not able to get across. Suppose the subject's feet are injured and not able to jump, (54) cannot express such meaning of inability. However, neng'can' is used with bu 'not' to indicate inability, eg: (55) ta bu neng tiao-guo-qu he not can jump-across-go He cannot jump across (he cannot jump). As a result, in an RVC, the agent initiates the primary action referred to by the compound, but the use of neng 'can' only signals the possibility of the initiation of the action. While they have potential form, RVCs cannot be reduplicated as most action verbs can to signal delimitative aspect. For instance, we have kan 'to look' and kan-kan 'have a look', but we don't have *tiao-guo-qu-tiao-guo-qu 'jump-across- jump-across' or *la-kai-la-kai 'pull-open-pull-open'. From the functional point of view, the ill-formedness of the reduplication of RV Cs thanks to the communicative function of an RVC to tall< about whether the result of an action can/cannot happen. By contrast, the delimitative aspect means doing the action 'a little', and does not concentrate upon the comments on the result. A further characteristics of RVCs is the resistance against other elements intervening between the two constituents, except for the potential infix -de- and -bu-, and directional verbs. This property separates RVCs from the verb-object compound. The structure of a directional RVC consists of a displacement Vl and a directional V2. Consider the following example: (56) ta zuo-chu-lai-le he walk-out-come-asp He came wall

--The most common type of displacement verbs (Vl) of the RVC include those expressing the meaning of motion, for instance, zuo 'walk', pao 'run', liu 'flow', fei 'fly', etc. The second type of displacement verbs involves those action verbs that intrinsically imply the direct object to undergo change of location, for example, ban 'remove', reng 'throw', song 'send', etc. There is another type that involves the action verbs that cause the direct object to undergo displacement, for instance, da 'hit'. The verb da 'hit' does not imply the location change of the direct object. Rather, it may cause the change of location to take place, as in da-jin-qu 'hit-in-go: hit something so that it goes in'. Just as the displacement verbs (Vl) of the RVC have different types, so do the directional verbs (V2) of the RVC. Generally, directional verbs can be classified into three types. The first type of directional verbs includes the verbs of lai 'come' and qu 'go'. The verb lai 'come' indicates the direction towards the speaker, while qu 'go' means 'away from the speaker'. Consider the following examples: (57) (58) ta song-lai-le yi he send-come-asp one ge cl dianshi TV He sent over a TV (towards the speaker). ta he na-qu-le take-go-asp san ban shu three cl book He took three books away ( away from the speaker). In (57) the directional RVC is song-lai. Song 'send' is an action verb that indicates the displacement of the direct object, while the second verb lai 'come' signals the direction towards the speaker. The same holds for (58). The directional RVC of (58) is na-qu, where the first verb na 'take/bring' indicates the displacement of the direct object, and the second verb qu 'go' signals the direction away from the speaker. Another class of directional verbs includes the following eight verbs (Li and Thompson 1981: 59): (59) a. shang ascend/up b. xia descend/down c. jin enter/in d. chu exit/out e. qi rise/up f. hui return/back 35

g. h. guo kai cross/over open/away, apart Apart from their lexical meanings and uses as independent full verbs, these verbs are used to indicate directions in the RVC. We can illustrate this with some examples: (60) (61) (62) (63) (64) (65) (66) (67) WO I dai-shang-le wear-ascend-asp I put on a hat. WO fang-xia shubao I put-descend bag I put down my bag. WO zuo-jin keshi maoz1 hat I walk-enter classroom I wall

--From the above examples, it is not difficult to see how the directional verbs are used in conjunction with the action verbs to constitute directional resultative verb compounds. The third type of directional verbs in RVCs are comprised by adding the directional verb of the first type to that of the second type. Hence, for example, we can combine the second type directional verb shang 'ascend' with the first type directional verb lai 'come', and produce a new double directional verb shang-lai 'come up (towards the speaker)'. Since there are two directional verbs in the first type and eight in the second type, we then have sixteen combinations of double directional verbs for this third type. This type specifies whether the motion is towards or away from the speaker, and thus differs from the second type, for instance: (68) John zuo-jin-lai le John walk-enter-come asp John walked in. The directional verb lai 'come' signifies that the speaker is inside the same building as John was entering. Since in (69) qu 'go' indicates that the subject is moving away from the speaker, the speaker is outside the building that John was entering. (69) John zuo-chu-qu-le John walk-exit-go-asp John walked out. There is a member of the class which may be used with a nondisplacement verb, that is, xia-qu 'descend-go: come down (away"from the speaker)'. The nondisplacement verbs that occur with it are durative verbs, which together with xia-qu produce the meaning 'continue'. This can be understood as a metaphoric extension from direction to time, for instance: (70) chi-xia-qu zuo-xia-qu eat-descend-go: keep eating work-descend-go: keep working Structurally, the third type, which involves lai 'come' and qu 'go', can be separated and allow the direct object to intervene between the directional verb and the nondisplacement verb, for example: (71) yi ban shu lai ta he dai-le bring-asp one cl book come He brought a book (towards the speaker). 37

(72) ta she Sal fill na ban that cl shu gin-qu book enter-go She squeezed that book in (away from the speaker). In (71) the resultative verb compound is dai-lai 'bring-come: bring', but it is split by the direct object yi-ban-shu 'one-cl-book: a book'. The same is true to the resultative verb compound sai-gin-qu in (72), where the direct object is positioned between sai 'fill/squeeze' and gin-qu 'enter-go: go in'. This phenomenon finds its parallel to the phrasal verbs in English, for example, John put on the hat vs John put the hat on, where the hat can intervene between the verb put and the particle on. For an RVC that contains lai 'come', the syntactic pattern can alter in several ways. Besides the kind of split that we have just observed above, where the displacement verb (eg sai 'fill') is separated by the direct object from the whole directional verb (eg gin-qu 'enter-go: go in'), we can put the direct object between the two directional verbs if the last directional verb is lai 'come', for example:4 (73) a. ta lin-shang-lai-le yi fen liwu he carry-ascend-come-asp one cl gift b. ta Jin-le yi fen liwu shang-lai he carry-asp one cl gift ascend-come c. ta lin-shang yi fen liwu lai-le he carry-ascend one cl gift come-asp He carried a gift up (towards the speaker). Although all three alternations in (73) differ in their syntactic structures, they are 4 This is Li and Thompson's (1981) observation. However, not all directional verbs of the second type can have these alternations. For instance, guo 'cross' may not be used in this alternation meaning the same: (a) ta song guo lai le yi ben shu he send cross come asp one cl book (b) ta song yi ben shu guo lai le he send one cl book cross come asp (c) * ta song guo yi ben shu lai le he send cross one cl book come asp He sent a book over (towards the speaker) It seems to me that (c) is not possible on the reading that guo 'cross' signifies direction, but it is grammatical when guo 'cross' expresses experience of the subject meaning something like 'ever/already'. 38

semantically synonymous. (73a and b) demonstrate the different separations of the two constituents of the compound lin-shang-lai 'carry-ascend-come: bring up'. While only the RVC with lai 'come' can alternate as such, the RVC that contains qu 'go' do not have all the alternations as those of the RVC containing lai 'come' above. Let us look at an example: (74) a. ta ti xiao gou chu-qu -le he kick little dog exit-go-asp b. * ta ti-chu xiao gou qu-le he kick-exit little dog go-asp He kicked the little dog out (away from the speaker). As in (74b ), the RVC ti-chu-qu 'kick-exit-go: kick out' cannot allow the direct object to intervene between the chu 'exit' and qu 'go', while the counterpart of lai 'come' can. We can observe that only an RVC split as in (73c), we can have the potential form with the infix -de- or -bu- between the displacement verb and the whole directional verb phrase, as in (75) below: (75) ta lin-de/bu-shang . yi fen liwu lai-le he carry-can/not-ascend one cl gift come-asp He can/cannot carry a gift up (towards the speaker). The potential form is not available for the split structure of (736 ), where the whole directional verb phrase is separated from th~" displacement verb by the direct object, thus: (76) * ta lin-de/bu-le he carry-can/not-asp yi one fen cl liwu shang-lai gift ascend-come He can/cannot carry a gift up (towards the speaker). We cannot insert -de- or -bu- after Zin' carry', neither can they be inserted between shang 'ascend' and lai 'come'. A further observation (Chao 1968, Li and Thompson 1981) about the first type of directional verbs, ie lai 'come' and qu 'go' puts another constraint on the split of RVCs. If the RVC containing either of these two verbs has a direct object realised by a place noun, the RVC must split, for instance: 39

(77) a. WO xia shan qu I descend hill go b. * WO xia-qu shan I descend-go hill I went down hill. (78) a. ta zuo-jin fangjian qu-le he walk-enter room go-asp b. * ta zuo fangjian jin-qu-le he walk room enter-go-asp c. * ta zuo-jin-qu fangjian le he walk-enter-go room asp He walked into the room (away from the speaker). The nouns shan 'hill' in (77) and fangjian 'room' in (78) are place nouns and thus following Chao and Li and Thompson, the two place nouns must be used between the first type of the directional verb lai 'come' or qu 'go' and the second type of the directional verbs in the compound. Otherwise, ungrammaticality may be yielded.5 There are, moreover, some other types of verb compounds subsumed under the heading ofRVCs by LT, but we shall not discuss them here. One major difference of LT from others is that LT do not undertake the effort to distinguish the surface and deep structures of RV Cs, but only consider how communicative functions are -~onveyed by RVCs. We have so far surveyed two pieces of research conducted in a functional framework. L Huang's work has the advantage of reconciling the opposite views on the derivation of the underlying structure of RVCs. The theory of conceptual closeness is said to capture the nature of RVCs, and simultaneously works without failing to recognise the strengths of the transformational and lexical approaches. Furthermore, the recognition of continuum of grammatical constructions is important. In fact, the concept has been observed by linguists, for example, Quirk at al ( 1985), which refers to it as gradience. L Huang provides us with a descriptive framework and attempt to 5 The judgement of the well-formedness of the (c) type sentence is arguable. According to Chao (1968) and Li and Thompson (1981) this is not possible, but some may find it quite all right to say that without difficulty. 40

explain the different degrees of grammaticality of resultative verb complement constructions in terms of the different cognitive organisation of speakers. While they do not concentrate upon the underlying representation ofRVCs, LT provide us with a functional framework within which the functions of RVCs are described and explained. For example, the semantic functions of the potential forms de!bu are discussed in terms of scope of auxiliary, which were ignored in previous studies. However, one major drawback of LT is that they fail to account for the semantic relations between the RVC and its arguments, which has become a major issue in more recent studies. 2.4 Some Recent Representative Studies The recent years have witnessed the increasing interest in the study of the syntactic-semantic interface. This interest has led researchers to investigate the argument structure of particular constructions among languages. One of these interesting topics is that of complex predicates. Recently, there has been a revival of interest in Chinese complex predicates, in particular, the resultative verb compound, which has been recently studied by linguists within different theoretical frameworks and approaches, for instance, GB (eg Li 1990, 1993, 1995; Huang 1991, Cheng and Huang 1994), LFG (eg C-R Huang and Lin 1992), syntactic (eg Huang 1991, 2010) and lexical ( eg Li 1990) and mixed ( Cheng 1993) approaches. We shall review some of these recent major studies of resultative predicates. 2.4.1 Li (1990) In the previous sections, we saw the linguists' research on resultative verb compounds conducted more or less in a traditional line. However, Li (1990) is the first articulate theory ofRVCs done in the Government-Binding framework. 6 Li (1990) is a significant work in that it first attempts to account for the syntactic and semantic properties exhibited in RV Cs by a theory of argument structure, which has become one of the main foci of inquiry in linguistics. Li proposes a lexical analysis of compounding within a theory of argument structure assumed by Grimshaw (1992). Apart from the Case theory, Li makes use of three assumptions in relevance to theta-roles. 6 Li (1993, 1995) also deal with the argument structure of RVCs and they are more or less the development ofLi's theory. We note the influence of his later work in places. 41

a. Following Grimshaw ( 1992), Li assumes that the theta-roles of a verb has a structured argument structure and they have the hierarchy based upon relevant prominence b. Li also adopts from Higginbotham (1985) to assume that theta-roles can be identified and assigned to a single NP c. Borrowing from inter alia Lieber (1983), head-feature percolation is used to assume that the theta-grid of a compound preserves the theta-role prominence of the head of the compound The identification of a compound consisting of two component verbs VI and V2 is said to take place when the theta-roles of VI are 'identified' with those of V2 and then produce a theta-structure with combined theta-roles. For instance, consider the compound zhui-lei 'chase-tired' below. John zhui-lei-le Mary john chase-tired-asp mary John chased Mary as a result John/Mary got tired. The transitive VF has two theta-roles, < l, 2>, and the V2 has one theta-role < l'> or <2'> because it is ambiguous in that V2 can assign the theta-role to either the subject argument or object argument in this case. Thus, the compound zhui-lei 'chase-tired' has the theta-role either < 1-1', 2'> or < l, 2-1'> ('-' indicates that the two theta-roles are identified). Li's use of theta-identification is to maintain that the theta-criterion is not violated by the ability of assignment of multi-theta-roles to arguments. The prominence of the theta-roles rule of the other combinations. Theta-identification and theta-role prominence contribute to the explanation of a large number of resultative compounds. Examples include the following: (80) John qi-ku-le Mary john annoy-cry-asp mary John made Mary so angry that she cried. 7 Li (1990) uses 'Vl' and 'V2' to represent the two component verbs in a resultative compound, but regards them as having a semantic CAUSE. Li (1993) makes use of the notation 'Vcaus' and 'Vres' to represent the two verbs in the compound, and '01, 02' and '0a, 0b' to represent the theta-roles of Vcaus and Vres, respectively. We adopt the notation of Li (1990) here for convenience. 42

(81) John wan-wang-le zijide zhize john play-forget-asp own duty John played so much that he forgot his duty. (82) John xiao-feng-le john laugh-crazy-asp John laughed as a result of being crazy. In (80) Vl has two theta-roles and V2 only has one, but in (81) Vl has one theta-role and V2 has two. Vl and V2 in (82) have one theta-role, respectively. Furthermore, the relative prominence of theta-role does not exist in (82) because both verbs merely have one single theta-role. There is no need for theta-identification because of the Case theory. There are two structural Cases, namely, nominative and accusative Cases. If the verb has only one theta-role, we will have a transitive or intransitive (with(out) theta-identification) compound. Since there are two Cases, there are only two arguments available for a compound. Thus, in (82) there should be a transitive and an intransitive reading available, but due to some pragmatic influence, Li suggests that the transitive reading is not available in this case. Apart from the theta-identification, Li regards the complex predicates as involving headed structures (ie Vl as the head of the compound). The hierarchical lexical argument structures adopted from Grimshaw is controlled by the head feature percolation principles. This is also important for Li to maintain that the theta-prominence of the head is inherited by the co:11pound. Although Li (1990) provides an attractive theory of resultative verb compounds based upon a theory of argument structure, there are none the less some problems associated with this approach, which have been noted by linguists (eg C-R Huang and Lin 1992, Cheng 1993, Cheng and Huang 1994, etc). First of all, we note that some RV Cs show ambiguity, as illustrated below. (83) John qi-lei-le ma john ride-tired-asp horse a. John rode the horse and he got tired as a result. b. John rode the horse and it got tired as a result. 43

Thus, according to Li, there are two theta-identifications, namely, < 1-1', 2> and < l , 2-1'>. As Cheng (1993: 3) points out, the ambiguity can be resolved 'when the object is specified or definite'. The sentence below therefore is not ambiguous. (84) John qi-lei-le nei-pi ma john ride-tired-asp that-cl horse a. John rode the horse and he got tired as a result. b. *John rode the horse and it got tired as a result. No ambiguity arises here because the object is definite or specified, and V2 cannot predicate of the object argument. In other words, there is only one theta-identification for qi-lei 'ride-tired' in this case, <1-1', 2>. This phenomenon is not captured by Li's theory. Li's analysis will predict that the above example can have the same ambiguous readings as the previous one because the constraints of definite or specified NP arguments or referential NP arguments are not part of the mechanism of his analysis, and they play no role. Accordingly, Li's account will overgenerate. Moreover, qi-lei ' ride-tired' is expected to have ambiguous interpretations in (85) ifit is ambiguous in (83) (Cheng 1993). However, there is no ambiguity here but only one acceptable reading, that is, only the subject NP argument; but not the object NP argument (as expected), is predicated ofby V2. (85) John qi-lei-le che john ride-tired-asp bicycle a. John rode the bicycle and he got tired. b. *John rode the bicycle and it got tired. On the other hand, Li's account also undergenerates the data (cf Cheng 1993, C-R Huang and Lin 1993). (86) John ti-pou-le qiuxie john kick-break-asp sneaker John kicked (eg played soccer) so that his sneakers are broken. For example, C-R Huang and Lin (1993) show the inadequacy of Li's account by presenting this example. The object argument of ti-pou 'kick-break' is qiuxie 'sneak.er' in (86), but there is a patient argument absent from the syntax, namely, the argument qiu 'ball'. The second argument of Vl is not identified with the argument of V2. 44

However, the sentence is grammatical and has only this reading. Thus, the theta-prominence is not maintained and Li's analysis undergenerates. There is no doubt that Li relies on the abstract concept of theta-prominence, but we do not know how it is represented and 'percolated' with on thematic structure and roles involved. Also, it is generally assumed in GB that head features involves Case-assignment, and Case-assignment should be inherited from the head. However, as shown in the next section of LPG analysis, the transitivity of the whole resultative compound does not follow from either Vl or V2. This in turn means that Li's analysis cannot predict that the number of arguments of the compound sometimes contradicts the number of arguments that the purported head takes. In summary, Li (1990) (and in his subsequent work) has offered an attractive analysis of resultative verb compounds, which can account for a wide range of compounds. However, there are problems with overgenerating and undergenerating data, as we have already seen. More recent work also builds on Li's analysis and attempts to solve the problems by different means. We shall consider an LPG account of resultative compounds in the next section. 2.4.2 Huang and Lin (1993) C-R Huang and Lin (HL 1993) provide an alternative analysis ofRVCs within the framework of lexical-functional grammar (LPG). As we have noted in the last section, they show that Li (1990) is not adequate. They propose a template-based approach to argument selection in resultative complex predicates. This approach partly relies on the argument selection proposed by bowty's ( 1990) Proto-roles. First and foremost, HL show that Mandarin resultative verb compounds possess complex argument selection and argument sharing phenomena. For example, it is not possible to predict the transitivity of the resultative verb compound by the transitivity of its component verbs, Vl and V2. (87) a. [Vl(intr) - V2(intr)] Vr(intr) John ku-lei-le john cry-tired-asp John cried such that he's tired. 45

b. [Vl(intr) - V2(intr)] Vr(tran) John ku-shi-le zhentou john cry-wet-asp pillow John cried such that the pillow was wet. c. [Vl(tran) - V2(intr)] Vr(tran) John da-si-le louhu john hit-die-asp tiger John hit the tiger to death. d. [Vl(intr) - V2(tran)] Vr(tran) John zou-jin-le jiaoshi john walk-enter-asp classroom John walked into the classroom. e. [Vl(tran) - V2(tran)] Vr(tran) John ting-dong john listen-understand zhongwen Chinese John listened and understood Chinese. All these examples show that the transitivity of the resultative compound is not predicted from the transitivity of its component verbs. Just as HL claim that there is no way to predict the transitivity of resultative compound, they also deny the existence of the headed structure of the compound as other researchers (eg Li 1990, Cheng and Huang 1994) assume. Instead, HL argue that the resultative compound involves composite structures rather than headed structures like causative constructions. This assumption is based upon semantic considerations of Mandarin resultative complex predicates. Consider the example (86) again, repeated as (88) (HL 1992: 92). (88) Ta he (tiantian (day-day ti qiu) ti-puo-le ta-de qmXIe kick ball) kick-break-asp his sneaker (lit) He kick-break his sneaker (because he plays soccer everyday). Under the circumstance described in the sentence, the sneaker gets worn and broken as a result of the actor playing soccer, but not as an intended result. That is to say, the 46

result expressed by V2 is not controlled by the action of the whole predicate, and the actor is not the causer like that of the abstract predicate CAUSE. Furthermore, the object theme argument qiu 'ball' is not present. According to HL, Li's (1990) argument structure analysis and Baker's incorporation analysis (1988) cannot provide an adequate explanation for what happens in this example. Thanks to the problems of previous research, m particular, Li (1990), HL propose that RVCs undergo the same mapping between thematic and argument structures (Bresnan and Kanerva 1990). They also rely on Dowty (1990) to postulate two argument templates for argument selection. We shall briefly comment on their treatment ofRVCs below. HL treat resultative compounds as composite event structures, which are considered to fit into one of the 'templates': the Accusative Template is connected to the arguments bearing the roles of Proto-Agent and Proto-Patient, and the Unaccusative Template (following Dowty 1990) is connected to the argument bearing the Proto-Patient role. Basing on these two templates, HL claim that the mapping between the selected base argument and the resultative argument is one-to-one, hence 'one-argument-per-argument-structure'. Moreover, HL do not entirely do without consideration of transitivity, since they stipulate that the component of the RVC contributes one argument to the compound. This proposal is said to predict the possible and impossible readings of RVCs. However, some widely used RVCs are not accounted for in such an approach. (89) John lei-si-le Mary john tired-die-asp mary John made Mary tired to death. Note that the subject argument is not an argument of Vl or V2. Rather, it is the argument of the whole RVC. There is however a causative interpretation where the Causer argument is not projected from the argument structure ofVl or V2. The causative RVC also poses problems for the one-to-one principle. (90) Nei-pi ma qi-lei-le wu-ge ren that-cl horse ride-tired-asp five-cl person That horse made five people tired (from riding it). 47

According to the one-to-one principle, HL would assume that the subject argument of the RVC is the contribution ofVl, and the object argument is that ofV2. In this case, the object argument is contributed by V 1. In other words, it is five people who rode the horse. Normally, Vl contributes to the external argument in some 'canonical' resultative compounds. Furthermore, the fact that the subject argument in this example is a Causer rather than a Patient is not captured by HL's theory. 2.4.3 Tang (2010) Tang (2010) follows Huang (1988) to assume a VP-Shell analysis of the Chinese resultative construction and consider the potential form of the construction as a complex predicate - the resultative complement of the first verb eg chi-de ('eat') ie hen kai-xin ('very happy') being the 'secondary predicate' (Tang 2010: 123) . ... vP D~' v~VP l Tb Won en B chi-de e hen kai-xin We eat very happy This phrase marker can be interpreted as women in some state, which is 'we ate happily'. The state is expressed by BE (BECOME) to which chi-de moves from the V position . ... vP D~' v~VP Tb Won en chi-de t e hen kai-xin /1\ __ I We eat very happy The question here is why it moves. Tang (2010: 123) says 'it is relevant to the nature of de'. The movement merges BE and de so that they express a stative event. However, how relevant it is to the nature of de we are not told. If there is no movement, is there still be a state as implied by BE? 48

-2.4.4 Cheng and Huang (1994) Apart from Tang (2010), Cheng and Huang (1994) is a representative work in the formal paradigm. They argue for an analysis based upon the classes of argument structure and transformation. They also examine the transitivity alternations of the classes. Thus, according to Cheng and Huang (1994), a predicate can be distinguished between accusative and unaccusative along with transitive and intransitive. (91) a. John chang-le yi-ge xiaoshi john sing-asp one-cl hour John sang for an hour. b. John chang-le liang-shou ge john sing-asp two-cl song John sang two songs. c. Mary xia-le yi tiao mary shock-asp one jump Mary was shocked. d. John xia-le Mary yi tiao john shock-asp Mary onejump John surprised Mary. ( unergative) (transitive) (ergative) ( causative) (91 a-d) illustrate the patterns of predicates thi t are generated from the two dimensions. (91a and c) are different from (916 and d) in transitivity, where the former is intransitive and the latter is transitive. (91a and b) also differs from (91c and d) in their aspectuality, where the former denotes an activity and the latter denotes a state and/or a change of state. While they denote an activity, (9 la and b) show in their transitivity alternations the presence or absence of an internal argument as a Theme or Patient. (91c and d) denoting a state or change of state show in their transitivity alternations the prescience or absence of a Causer as an external argument if it is present. In fact, Cheng and Huang ( 1994) show that resultative verb compounds can also be seen to exhibit such kinds of transitivity alternations and aspectual properties in their argument structure. 49

(92) a. John qi-lei-le ( unergative) john ride-tired-asp John rode himself tired. b. John qi-lei-le yi-pi ma (transitive) john ride-tired-asp one-cl horse John rode a horse tired. c. John qi-si-le (ergative) john anger-dead-asp John was very angry. d. Nei-jian shi qi-si John le (causative) that-cl matter anger-dead john asp That matter angered John. As shown in (92), RVCs can have the same structures as those of simple verbs. (a) is unergative denoting an activity of the Vl of the RVC and can alternate with the accusative construction in (b ). There is a question of control in (b) as used transitively. In (a) the subject John is the subject of both Vl and V2, but in (b) the subject John is the subject ofVl and the subject ofV2 is now understood to be the object of the whole RVC a horse. To use a relevant jargon, a horse is the controller of V2, whereas John is the controller ofVl in this sentence (cfHuang 1991; Li 1990, 1993, 1995). (92a and b) has only the unergative-transitive relationship since the subject ofVl is nat involved in this interpretation. Some examples of the unergative pattern include chi-bao 'eat-full', he-zui 'drink-drunk', zhui-lei 'chase-tired', etc. As for the transitive RVC, (926) is the pattern that most transitive RVCs have whose V2 predicates on the object (ie the controller of V2 is the object), such as ku-shi 'cry-wet', ti-kai 'kick-open', da-si 'hit-dead', etc. (92c and d) exhibit the ergative-causative alternation because the Theme object is the same in them, but they differ in whether a Causer exists. (92c) has no Causer, and the Theme John occupies the subject position. If there is a Causer, it will occupy the subject position and force the Theme to occur in the object position instead. Examples that can enter into this alternation are lei-si 'tired-dead', zui-dao 'drunl<-arrive', xia-pao 'scare-run', mang-lei 'busy-tired', and so on. 50

For the event type of the RVC, Vl in the unergative-transitive alternation usually denotes an activity and so its subject is an Agent as shown in the above example. In the ergative-causative alternation, Vl denotes a state or a change of state and so its subject is a Theme or an Experiencer. Vl of a few ergative-causative RVCs can denote a state, however. (93) a. John xie-lei-le john write-tired-asp John wrote himself tired. b. Zhe-ben shu :xie-lei-le John this-cl book write-tired-asp john This book got John to write himself tired. Example (93) shows the ergative-causative alternation and Vl designates a state rather than an activity now. Some transitive RVCs like (926) are sometimes found to exhibit the causative construction in some context. Hence, an RVC denoting an activity may well be used causatively as well as transitively, as shown in (94) below: (94) a. John ku-:xing-le Mary john cry-awake-asp mary John cried and made Mary awake. b. Zhe-jian shi ku-xing-le Mary this-cl matter cry-awake-asp mary This matter (in the dream?) made Mary cry herself awake. One may be tempted to consider the transitive and causative constructions to be derived from the unergative pattern such as John ku-xing-le 'John cried himself awake' in (94). There is a difference between the causative use and transitive use, however. In the causative construction, the subject of Vl is a Causee, but the subject of Vl in the transitive counterpart is an Agent. Thus, it would be more appropriate to characterise them as ergative-causative alternation relationship. As noted in Huang (1988, 1991), these patterns can be extended to larger syntactic units than verbs and compounds. One example is the complex predicate as resultative complement constructions with the so-called 'complementiser' de, which is regarded as a projection V' and falls within the four types of predicates as outlined above. 51

(95) a. John Im-de hen shangxin john cry-de very sad John cried very sadly. b. John ba Mary Im-de hen shangxin john ba Mary cry-de very sad John cried so much that Mary got very sad. c. John jidong-de liu-chu-le yanlei john excite-de drip-out-asp tear John was so excited that he came to tears. d. Zhe-jian shi jidong-de John liu-chu-le yanlei this-cl matter excite-de John drip-out-asp tear This matter got John so excited as to come to tears. As predicted, (95c and d) show ergative alternation and have an Experiencer ofVl. If the subject of Vl is a Causee or an Experiencer, causativisation may take place with a Vl denoting an activity like ku 'cry', as in ku-de hen shangxin 'cry very sadly'. (96) Zhe-jian shi ba John ku-de hen shangxin this-cl matter ba john cry-de very sad This matter caused John to cry very sadly. Noted in Cheng and Huang (1994), all resultatives are treated as causatives by nature. So, it is not accurate to say 'unergative resultatives' and other 'non-causative resultatives' in a strict manner. Causatives are divided into 'event causatives' and 'factive causatives' in S Huang (1982), though they roughly correspond to the 'transitives' and 'causatives' in what we have discussed here. So far, we have seen the RVC and simple verb as being grouped into different classes on the basis of their argument structure. Verb compounds in Mandarin Chinese generally have their syntactic counterparts (Anderson 1985, Chao 1968, Li and Thompson 1981). This leads linguists to conclude that verb compounds are basically syntactic in nature. As Chao remarks: As for the formation of compounds ... the relations involved are mostly analogous to these in syntactic construction . . . (Chao 1968: 194) 52

In Chinese most compounds are syntactic ... (Chao 1968: 366) This seems to be consistent with the projection principle proposed by Chomsky in the 80s: .. .lexical structure must be represented categorially at every syntactic level. (Chomsky 1986: 84) This syntactic property seems even to be reflected in compounds, that is, the syntactic nature of the whole compound would be subject to the syntactic nature of the head of the compound. The question of headedness is what we shall try to tackle in this section. Thus, what exactly is the head of the resultative verb compound? There have been debates about the head of an RVC in the literature recently (eg Huang 1988, 1992; Li 1990, 1993, 1995; Cheng and Huang 1994). The study of the headedness of RVCs being of theoretical interest, it is actually about the syntax of compounds, since the notion of a head is a structural notion rather than a conceptual one. Be that as it may, the problem can be approached from a semantic point of view, which, in fact, has been the traditional focus on the head of a resultative compound. Semantically, Vl of a resultative verb compound seems to function as an adverbial in the meaning of the whole compound. V2, on the other hand, seems to be the so -called 'core of predication' which Vl seems to modify. For instance, a compound such as la-kai 'pull-open' may preferably be understood as 'to open by pulling' rather than 'to pull with the result of opening'. Other examples having the same kind of interpretation include chui-kai 'blow-open', tui-open 'push-open', ti-kai 'kick-open', etc. This reason alone has led some to propose that V2 shol})d be considered to be the head of the compound. However, it is argued in Cheng and Huang (1994) that this view should be rejected. To begin with, if the semantic explanation is adopted, that is, Vl is the adverbial of the compound because of some examples interpreted as 'different ways of opening', then one would be led to consider them (eg ti-kai 'kick-open', ti-bian 'kick-flat', ti-shang 'kick-up', etc) to have 'different results' that arise from the action designated by Vl. These would not provide us with a correct basis for choosing either Vl or V2 as the head of the compound, however. The meaning of the compound depends upon the meanings of both components rather than that of any one of them. We can draw some parallel to the English phrasal verbs or verb-particle constructions here, eg push up, pull up, etc in which the emphasis is on different manners of upward movement; and push up, push down, etc in which the emphasis is on different results gained from the action of pushing. In these examples, we would not want to argue that the particle rather than the verb is 53

the head of the whole phrase, and so the head is the first predicative elements as the Vl in RV Cs. Thus, it is argued that the issue of whether Vl or V2 is the head of compounds should be better decided on by syntactic considerations rather than semantic ones. Attempting to solve this problem, Li (1990) argues that Vl is the head of a resultative verb compound by resorting to syntax. By doing so, Li tries to formulate a theory which can predict the argument structure of a resultative verb compound from the argument structure ofVl and V2. Transitivity relations of the verbs are of particular concern in his treatment ofRVCs. Li maintains that the external argument ofVl must be expressed as the external argument of the whole compound. However, the argument ofV2 is not restricted by the same requirement. All these are indication ofVl being the head, because some related features of the head are compulsory, but not the peripheral features as the head in X-bar theory is obligatory, whether overtly or covertly. Furthermore, the argument structure of the head Vl, which is considered to be a feature by Li, is said to be 'obligatorily percolated' to the whole compound, and thus, the external argument of Vl is realised as the external argument of the resultative verb compound. One of the interesting issues raised by Li ( 1990) is concerned about the paradigm of causativity, where the subject ofVl is a Theme/Experiencer, and the subject ofVl is a Causee, as in the following examples: (97) Nei-jian shi qi-si John le that-cl matter anger-dead john asp That matter angered John. (98) Zhe-ben shu kan-lei-le John this-cl book look-tired-asp john This book got John to read himself tired. However, the external argument of the compound in (97 and 98) is not an argument of Vl or ofV2. Rather it is a Causer from outside the argument structures ofVl and V2. In (98) the external argument of the compound is not the external argument ofVl, but is a Causer that is the same as the internal argument ofVl. Accordingly, it seems that causativity might override the hierarchy for non-causative accusative and unergative constructions. Some modification of Li' s theory would be necessary. Li's main focus being on thematic structure and transitivity relations, the very idea of Li's that transitivity relations do not really help decide the transitivity of the 54

whole resultative verb compound is shared by Cheng and Huang (1994) despite their disagreement that 'it is the aspectual property of an argument structure that plays a crucial role in the making of a resultative compound' (Cheng and Huang 1994: 194). Let us consider the following sentences: (99) a. Ta qi-lei-le zhe-pi ma he ride-tired-asp this-cl horse He rode and made this horse tired. b. Ta ku-shi-le maujin he cry-wet-asp towel He cried the towel wet. c. Wo ti-po-le yi-shuang xie kick-break-asp one-cl shoes I kicked and got a pair of shoes broken d. Wo song-suan-le send-sore-asp tui leg le asp I sent (presents) and my legs got tired. (100) a. Ta kan-fan-le he look-annoyed-asp He looked and got annoyed. b. Ta xiao-feng-le he laugh-silly-asp He laughed himself silly. c. Ta song-lei-le he send-tired-asp He sent and got tired. Example (99) is the transitive pattern [a. transitive Vl; b. intransitive Vl; c. transitive Vl, object not realised in compound; d. ditransitive Vl, neither object expressed] and (100) is the intransitive [a. transitive Vl, object not realised; b. intransitive Vl; c. ditransitive Vl, neither object expressed]. Both (99 and 100) show that the transitivity 55

of an RVC does not solely depend on the transitivity ofVl. The RVC can, in fact, be either transitive or intransitive regardless of whether Vl is transitive or intransitive. In this connection, perhaps it is worth mentioning the parallel linguistic phenomenon exhibited by English resultative constructions. Consider the following sentences (Cheng and Huang 1994: 195, with modifications): (101) a. John laughed himself silly. b. Mary wall

indicates a state or a change of state. It is generally true that those compounds with V2 being an inherently active verb, eg xia-tiao9 'frighten-jump', xia-han 'frighten-cry', etc cannot occur as resultative verb compounds, but those like xia-pao 'frighten-away' and xia-xing 'frighten-awake' are fine, for pao10 'away' and xing 'awake' are not active. On the contrary, Vl seems to be more responsible for the aspectuality and event types of the RVC. For example, the whole compound is considered to be unergative or transitive contingent upon whether it has an object, ifVl is active. If, however, Vl is stative/non-active, the whole compound will be ergative or causative contingent upon whether there is an external Causer argument. Thus, in this connection, it seems reasonable to regard Vl as the head of the compound, if the event type of the entire compound is dependent upon the event type ofVl but not the event type ofV2. It is, nevertheless, not merely those semantic considerations such as aspectuality of an argument structure but also syntactic concerns that support the postulation that Vl should be the head of a resultative verb compound. To substantiate this argument, Cheng and Huang (1994) draw further evidence from the structure of resultative phrases larger than compounds. Let us consider some examples (we will take the following as examples of resultative constructions, despite the uses of de-structure): (102) a. John Im-de john cry-de hen shangxin very sad b. c. John cried so that he got very sad. John ba Mary ku-de hen shangxin john ba mary cry-de very sad John cried so much to make Mary very sad. John jidong-de liu-chu-le yanlei john excite-de flow-our-asp tear John was so excited that he came to tears. d. Nei-jian shi jidong-de John liu-chu-le yanlei that-cl matter excite-de john flow-out-asp tear 9 It is however possible to have xia-tiao 'frighten-jump' and xia-han ' fr ighten-cry' in a larger construction like a sentence, eg John xia-de tiao le qi lai 'John was so frightened that he jumped up', and John xia-de jiao jiuming'John was so frightened that he screamed for help'. According to Cheng and Huang (1994), this difference is due to the result expressed by the aspectual features of a phrase or clause. 10 However, pao can be an activity verb meaning ' run' in other contexts. 57

That matter got John so excited as to come to tears. As has been pointed out earlier, the standard assumption is that a phrase must have a head (as stated in the projection principle). Hence, in the sentences above the main verb must be Vl or V2. It is argued in Huang (1988) that V2 cannot be possibly the main verb or head of the verb phrase, and that ( 102d) can be understood of its causative sense by assuming the underlying structure like ( 103) below with a covert causative predicate taking (30d) as its complement: (103) nei-jian shi that-cl matter [CAUSE] John jidong-de liu-chu-le yanlei john excite-de flow-out-asp tear The verb jidong-de 'excite-de' is said to have undergone movement to the position of the causative predicate to yield the surface structure. Vl is then affected by the movement but V2 is not. Since V -movement can only move a verb from the head of a complement (not from an adjunct, though) into the position of the higher selecting verb, this can be predicted under the assumption that Vl is the head. Moreover, it is generally accepted that Chinese resultative compounds reflect the word order of resultative phrases, and thus, it is reasonable to assume that resultative verb compounds have the same headed structure of phrasal resultative constructions. If this is correct, then the assumptions that V2 is the head and that RVCs are headless would be unfounded. Thus, attempting to answer the question concerning the head of an RVC in the beginning of this section, we are tempted to suggest that Vl is the head of a resultative verb compound. We have considered the role that aspectual properties of verbs., play in the determination of the head of a resultative compound unique to the Chinese language. Aspect is certainly also one of the determining factors in the classification of event types. Events are telic (indicates a successive stage without natural endpoints, eg pushing the trolley), atelic (indicates a successive stage with natural endpoints or outcome, eg writing a letter) or change of state (indicates only change of state, eg finding). In the many languages in the world that are found to grammatically distinguish event types, which suggests that the event type are based upon the grammaticalised cognitive universal in the languages of the world. However, it is interesting to see whether or not this hypothesis also holds for Mandarin Chinese, as it may have some impact on the generalisation of the distinction of event types. One interesting way to study event types in Mandarin involves their interaction with resultative verb compounds. 58

-As discussed before, the directional resultative verb compound indicates the direction of action and could locate an event from the viewpoint of the speaker, for example, chu 'out', shang 'up', guo 'cross', etc. Also, two directional verbs can occur together, for instance, jin-qu 'enter into', shang-lai 'come up', etc. We have also seen that resultative compounds can express not only result, but also accomplishment and achievement, and so on, for example: (104) a. result state: chi-bao 'eat enough', xie-qingchu 'write clearly' b. accomplishment: qu-dao 'arrive', kan-jian 'see' The combinations of the component verbs of a resultative compound are often predictable. 11 Result-state RVCs function in event aspect, while accomplishment (completive) RVCs function in viewpoint aspect, in which they contribute to the perfective viewpoint, and in event aspect, in which they change event type to change-of-state. We can illustrate this point by looking at some examples in the following, and they will show the event types may be changed by RVCs of different classes. (105) a. Niao fei-le ( atelic) bird fly-asp The bird flew. b. Niao fei-shang-le (atelic) bird fly-ascend-asp The bird flew up. c. Ta peng-lewo de meimei ( change of state) he bump-asp me de sister He bumped my sister. d. Ta peng-jian-le WO de me1me1 ( change of state) he bump-see-asp me de sister He encountered my sister. (105 a-d) are two pairs of sentences which show no change of event type by the RVC, while (a and b) differ from (c and d) in the class ofRVC (the former directional, the latter completive) and hence in event type. 11 Some unpredictable combinations, however, include xia-qu 'descend-go: keep on', xiang-dao ' think-arrive: know', etc. 59

However, RV Cs may be able to change the event type of a sentence. For instance, the next examples show a change from atelic or stative to change-of-state. (106) a. Ta zhao-le nei-ben he look for-asp that-cl He looked for that book. shu book b. Ta zhao-dao-le nei-ben shu c. d. he look for-arrive-asp He found that book. John a1 Mary john love mary John loved Mary. John ai-dao-le john love-arrive-asp that-cl Mary mary John fell in love with Mary. book ( atelic) ( change of state) (stative) (change of state) ( 106a and b) show a change from atelic to change-of-state, and ( c and d) show a change from stative to change-of-state. This pattern is important to the Mandarin language. The change of state may be viewed as the outcome of the process (not a necessary outcome, though). Examples of such pairs include: (107) a. atelic: kan 'look' , ting 'listen' , zhao 'look for' b. change of state: kan-dao 'see', ting-dao 'hear', zhao-dao 'find' Although the semantics between Mandarin and the English equivalents are similar, they differ in morphology. As Smith (1990) observes, the lexicalisation pattern for the related verbs differ between English and Chinese as there are in English few pairs with the same morpheme, for instance, cai 'guess' and cai-dao 'guess'. Thus far, our discussion about the interaction between event types and resultative compounds has prepared us for the consideration of the event structure of resultative compounds. Recall, towards the end of last section, we have seen that the event type ofVl can affect the event type of a resultative compound. In order to capture such a generalisation within a theory of argument structure, Cheng and Huang (1994) propose a complex underlying event structure for resultative compounds. In the event 60

structure, the event denoted by Vl has its complement the event denoted by V2. (108) is the proposed event structure of the unergative and transitive constructions. (108) [1\V Vl Act ;vc [V2 Statc/changcof stateJl Given its semantics and the component denoting an activity, this structure can be called the Active RV Cs, which obligatorily select an Agent as their external argument. Some of these RVCs must, and some may take a Theme as their internal argument. If a Theme as an internal argument is selected, the RVC is transitive, and it is unergative if there is no internal argument selected. Accordingly, three argument structures are possible: < 1 > , < l, 2>, < l (2)>, and 1 is linked to an Agent and 2 a Theme. This leads to the thematic linking assumed by Cheng and Huang (1994) to be controlled by Universal Grammar principles which arise from the compositional semantics. Thus, as implied in (108), the argument structures of unergative and transitive RV Cs are as in ( 109). (109) a. b. c. ( unergative) ( transitive) (mixed) Examples of these types of RV Cs include: pao-lei 'run-tired', tiao-fan 'jump-annoyed', ku-lei 'cry-tired' (unergative); tui-kai 'pull-open', sha-si 'stab-dead', la-ping 'pull-flat' (transitive); he-zui 'drink-drunk', chi-bao 'eat-full', qi-lei 'ride-tired' (mixed). Contrasting active RVCs are non-active RVCs, which are ergative and causative, and have the event structure as in (110) below: (110) [1lV Vl Non-active [V2 Statc/changcof statet] Since the semantics of RVC is non-active and denotes a state or a passive action, the non-active RVC selects a Theme or an Experiencer as an internal argument. It, however, selects a Causer as an external argument optionally. Assumed to observe the Unaccusative Hypothesis, the non-active RVC has an argument structure such as <2> or , and again, 1 is linked to a Causer and 2 a Theme, an Experiencer or a Causee by principles ofUG. (111) b. (ergative) c. (causative) An ergative sentence appears as if it was a purely intransitive sentence because it has an object but has no thematic subject. The structure of (112) shows that in an ergative 61

sentence, the object is moved up to the surface subject position in the derivation, and so it looks like an intransitive sentence (Cheng and Huang 1994: 199-200). (112) IP ----- ----------NP I' I / "yp / " [e] V NP 1 lei-si WO I The subject and object positions of a causative sentence, on the other hand, are occupied, as shown in (113). (113) IP ----- ---------NP I' I / ---.....__ VP / " [ta] V NP lei-si wo Hence, from (112) and (113) (Cheng and Huang 1994: 199-200), it can be deduced that the unergative-transitive alternation arises from the addition of an internal argument, and the ergative-causative alternation arises from the deletion of an external argument from the argument structure of a verb which has the identical event structure. (114 and 115) demonstrate the alternation between ergative and causative of a non-active RVC. (114) a. John john mang-lei-le busy-tired-asp John was busy and got tired. b. mary mang-lei-le John mary busy-tired-asp John Mary made John so busy that he got tired. 62

(llS) a. John john zui-dao-le drunk-fall-asp John got so drunk that he fell. zui-dao-le John b. Zhe-bei-jiu this-glass-wine drunk-fall-asp John This glass of wine caused John to be so drunk that he fell. There are some RVCs which can show not only the ergative-causative alternation but also the unergative-transitive alternation. One example is the RVC qi-lei 'ride-tired', as shown in (ll6): (ll6) a. Ta qi-lei-le he ride-tired-asp He rode and tired. b. Ta qi-lei-le yi-pi ma he ride-tired-asp one-cl horse He caused a horse to be tired by riding it. c. Zhe-pi this-cl ma qi-lei-le ta horse ride-tired-asp him This horse got him tired (oy his riding on it). As observed, even Vl of a resultative verb compound denoting an activity can be used in a non-active manner as long as the logical subject ofVl is not the initiator of the event at issue. In other words, the subject ofVl will be the Causee or Experiencer of the event if an external Causer exists. This can be seen from ( 116a ), which is the ergative counterpart with a Causee subject of the causative sentence. Nevertheless, such an interpretation is obtained only when the external Causer is understood in the universe of discourse. On the other hand, if there is no understood Causer, the subject ofVl is assigned an Agent role by default, and thus the RVC is unergative. For some intransitive RV Cs like xiao-feng 'laugh-mad', pao-lei ' run-tired', etc, theres seem to be ambiguity in the interpretation of the subject of the RV Cs as a Causee or an Agent. 63

(117) a. Ta xiao-feng -le he laugh-mad-asp He laughed so much as if he'd got mad. 12 He was caused to laugh so much that he got mad. b. Ta pao-lei-le he run-tired-asp He ran so that he got tired. He was caused to run so much that he got tired. The English translations of the sentences should hopefully do justice to the interpretations of the Chinese counterparts in ( 117). The subjects of the sentences can be interpreted as a Causee if in the universe of discourse an external Causer is understood. Or they can be interpreted as an Agent if there is no external Causer understood in the discourse. In contrast, for some other intransitive RVCs like lei-si 'tired-dead', le-Jeng 'happy-mad', etc, since the first verb denotes a (psychological) state, the subject is interpreted not as an Agent, but an Experiencer only. (118) a. John lei-si-le john tired-dead-asp John was tired to death. b. Mary le-feng-le mary happy-mad-asp Mary was so happy as if she'd got mad. John and Mary are the experiencers of the events denoted by Vls of the sentences. Accordingly, such RVCs are ergative and show only ergative-causative alternation. Thus far, we have considered the unergative-transitive and ergative-causative alternations shown by resultative verb compounds. They are defined by two event structures which are headed by the event denoted by Vl, ie the active and non-active patterns. The active pattern shows the unergative-transitive alternation, and the non-active the ergative-causative alternation. Furthermore, the event types denoted by a verb contribute to the formation of the active and non -active patterns. The transitivity of a 12 This is a preferred reading, however. Some speakers may even find this the only acceptable interpretation, and so, for them, the sentences in the example are not ambiguous. 64

-verb, however, plays no role in determining these patterns. As has been argued, the transitivity of an RVC is, in fact, determined by the composite meaning of the entire RVC in question, but not by the transitivity that its component verb has. For instance, ku-si 'cry-dead' meaning literally 'to cry oneself till one is dead' is an event affecting nobody else except the subject, and so it is intransitive. On the contrary, ku-shi 'cry-wet' meaning 'to cry till something gets wet' does has an effect on something but the subject, and so it id transitive. For some amphibians, both transitive and intransitive uses can be found. Like the intransitive RVC k~-lei 'cry-tired', the cry may happen with or without an external cause. If there is one, the RVC could be used ergatively ( the cause is not from within the sentence) or causatively (the cause is a syntactic argument); if there is not, the RVC could be used unergatively. The above discussion of the previous studies on the Chinese resultative construction is not meant to be exhaustive but attempts to show the complexity of the nature of the construction and the great deal of efforts linguists have put. Those general studies of the construction will be discussed in the appropriate places later. 65

Chapter3 Verb Classes and Constructions 3.0 Introduction Over the last several decades, and with the development of theories, more and more analyses of argument structure, which stands for the complicated information of a lexical item, have emerged. This lexical information also provides the information of syntactic behaviours of lexical items. There has been some recent work adopting argument structure theory to explain phenomena in adjectival and verbal passives, middles, light verb constructions, verbal compounds, etc (eg Grimshaw 1990). In contemporary theoretical linguistics, the study of lexicon - lexical semantics has emerged within various current linguistic frameworks such as Transformational Grammar, Lexical-Functional Grammar, Head-driven Phrase Structure Grammar, and so on (Horrocks 1987; Sells 1985; Sag I et al 2003 ). Linguists have switched their focus on syntactic rules to the properties of the lexical items, which means that linguistic phenomena can be explained in terms of argument structure, that is, the representation of properties that allow lexical items to take arguments. The study of these properties of lexical items inevitably yields the detailed analysis of the meanings of their predicates. Despite the fact that the study of lexical semantics separated linguists from the interpretive semantics and generative semantics camps in the 1970s (Huck and Goldsmith 1995; Newmeyer 1981, 1983), it has a unifying implication today allying linguist of various current theories. The results of their investigations have been incorporated into most linguistic theories in a similar way. The underlying assumption of most of the current research in linguistics that meanings of words are represented in the syntactic properties of phrases makes way for the new method for the study of word meanings. A set of words sharing similar meanings may require similar syntactic representations, and thus provides us with the semantic structures of word meanings. In the Chinese linguistic tradition, argument structure has been studied under the name of verbal construction. The last decades saw the change of Chinese linguistics against the background of general modern linguistic theory. Since Huang (1982) 66

appeared, more linguistic enquiries into the syntax of Chinese have been conducted, for example, Y HA Li (1990) etc. While Huang and others concentrated upon the study of LF, subjacency and postverbal constraint, others have worked on topics such as argument structure. One of the Chomsky's doctrines about the nature of linguistics is that syntax is autonomous, and should be studied on its own. This attitude has been dominant since Chomsky' s first book Syntactic Structures was published in the 1950s. It is, however, fair to say that Chomsky did not entirely neglect the importance of semantics - the study of meaning in linguistic theorising, as his closing remarks in Syntactic Structures (1957: 108) expresses his concern about the relationship between syntax and semantics: ... one result of the formal study of grammatical structure is that a syntactic framework is brought to light which can support semantic analysis. Description of meaning can profitably refer to this underlying syntactic framework, although systematic semantic considerations are apparently not helpful in determining it in the first place ... Nevertheless, we do find many important correlations, quite naturally, between syntactic structure and meaning; or, to put it differently, we find that the grammatical devices are used quite systematically. These correlations could form part of the subject matter for a more general theory of language concerned with syntax and semantics and their points of connection. Chomsky maintains that semantics has to be studied through the interaction with the theory of syntax, and he admits that there is a·correlative relationship between sentence structure and meaning. Whether one accepts Chomsky's conception of syntax and semantics or not, one would agree that verbs play a crucial part in determining the meaning of sentences. The crux of the propositional content of a sentence is generally expressed by the concept of the verb. The study of meaning in formal terms will embrace the study of meaning in a single word - word-internal semantics and the study of meaning relations a word has with other parts of the sentence - word-external semantics 3.1 Verb Classes and the Resultative Construction As has been discussed, Chinese compounds are of different types ( eg Chao 1968, Li and Thompson 1981, Anderson 1985, Tang 1989), for example, subject-predicate, predicate-object, predicate-complement, modifier-head. The RVC belongs to the class 67

of predicate-complement compounds. Recall that the RVC consists of a verbal morpheme and another verbal morpheme or an adjectival morpheme. The relation between the first verbal morpheme and the second verbal morpheme or adjectival morpheme is identical to the syntactic relation between a verb and a complement in a sentence. According to their internal structure, predicate-complement verbal compounds can be analysed as follows, and examples are included (from Tang 1989): (a) free verbal morpheme+ free verbal morpheme transitive: xue-jian ' reduce', shuo-fu 'persuade', ya-sui 'fragment', da-po 'break', tui-fan 'overthrow', jiu-huo 'save', yao-dong 'waver', jia-xing 'wake', kan-po 'see through', kan-jian 'see', kan-dao 'see', xue-hui 'learn', ting-jian 'hear', tui-dao 'push down' intransitive: zuo-kai 'walk away', san-kai 'scatter', pao-kai 'run away' (b) bound verbal morpheme + free verbal morpheme transitive: ji-po 'attack successfully', ji-lou 'shot down' (c) bound verbal morpheme+ bound verbal morpheme transitive: ju-jue 'refuse' (d) free verbal morpheme+ free adjectival morpheme transitive: tui-guang 'promote', ti-kao 'increase', xue-ruo 'weaken', xiao-hung 'burn red' intransitive: zhan-wen 'stand firmly', zuo-hung 'getting popular', kan-hao 'look upon' (e) free verbal morpheme+ bound adjectival morpheme transitive: shou-ming 'explain', fa-ming ' invent', kai-shen 'improve' intransitive: zuo-pi 'wallc tired' (f) bound verbal morpheme + free adjectival morpheme transitive: kang-da 'enlarge', ge-xin 'innovate' (g) bound verbal morpheme + bound adjectival morpheme transitive: ji-zhong 'concentrate', zheng-ming 'prove' The RVC is arguably left-headed (see Cheng and Huang 1994; Fong 2015, 2016), that is, the head of the compound is the first verbal morpheme. Although the components of the compound are referred to as morphemes, they can be categorised as 68

lexical verbs or adjectives in certain context. For example, hong'red' in zuo-hong'getting popular' is a bound morpheme, but is itself a verb/adjective in Ta hen hung 'He is very popular'. Apart from the classification based on its internal morphological structure, the RVC can also be categorised into different classes based on different principles and analyses. For instance, the RVC can be viewed in terms of its syntactic and semantic properties. Levin (1993) is a representative of the semantic approach to the analysis of verb classes in English, whose thesis is that the syntactic behaviour of words can largely be determined by their semantic properties. In other words, meaning plays a role in the determination of how a word appears in a sentence. It has long been held by many, if not all, linguists that the opposite is true since Bloomfield's (1933) neglect of meaning in linguistic theory. In recent years, the interest in the study of meaning has been revived and linguists would believe that meaning is still attainable even though all attempts may not have been successful. On the other hand, the syntactic approach to verb classes is still widely adopted by linguists (eg Keyser & Roeper 1984 [in the case of identifying middles and ergatives in English]; Li 1990, 1993, 1995; Cheng and Huang 1994 [in the case of identifying verb classes in RVCs in Chinese]; Levin and Rappaport Hovav 1995). The lexical representation may be represented in an argument structure or in a more semantic construct as the lexical semantic representation. In what follows we shall consider the verb classes that can undergo the resultative verb complement construction in Chinese in the approach exemplified by Cheng and Huang (1994). In Cheng and Huang ( 1994: 188-189) four classes of resultative compounds are identified by virtue of their syntactic behav.iours and transitivity alternations. They argue that the verb class that can occur in the resultative compound shows a range of transitivity alternations that is also shared by the verb used in isolation. The classes of verb include transitive, causative, unergative and unaccusative/ergative. The RVC is also observed to have the same types of alternations exemplified as follows: (1) a. John qi-lei-le john ride-tired-asp John rode himself tired. b. John qi-lei-le yi-pi ma john ride-tired-asp one-cl horse John rode a horse tired. 69 ( unergative) (transitive)

c. John qi-si-le john anger-dead-asp. John was very angry d. Nei-jian shi qi-si (ergative) John le (causative) that-cl matter anger-deadfohn par That matter angered John. The following are examples of verbs that can undergo all four of these alternations (from Cheng and Huang I 994: 198) a. Unergative: zhui-lei 'chase and get tired', tiao-fan 'jump and get annoyed', pao-dao 'run and arrive', shui-bao 'sleep enough', xiao-lei 'smile-tired', pao-lei 'run-tired', ku-lei 'cry-tired', shui-zhao 'fall-asleep', zou-diao 'walk-away', pao-kai 'run-away' b. Transitive: da-si 'hit-dead', ku-shi 'cry and make wet', ma-ku 'scold and make someone cry', ti-kai 'kick away', la-ping 'pull-flat', tui-dao 'push-fall', liu-chu 'drop out, ti-po 'kick-broken', tui-kai 'push-open', sha-si 'stab-dead', song-sua'n 'send-sore', si-po 'tear-broken', sao-ganjing 'sweep-clean' c. Unergative and transitive: chi-baa 'eat and become full', he-zui 'drunk', ku-xing 'cry-awake', da-sheng 'hit-win', qi-lei 'ride-tired', ku-fan 'cry-annoyed', da-bai 'hit-lose', han-ya 'shout oneself hoarse', ku-xing 'cry-awake' d. Ergative and causative: lei-si 'tired to death', zui-dao 'drunk-fall', xia-pao 'scare-away', le-huai 'happy to death', mang-lei 'busy-tired', kan-hua 'look and get blurred', xie-lei 'write-tired', chi-ni 'eat enough', ti-po 'kick-broken' Most RVCs are transitive; while (le and d) are ergative alternations and both have the semantic object/Theme. (2c) is a mixed class, whose members can occur in both unergative and transitive alternations. However, as Cheng and Huang (1994) point out, there is a class of verb which does not seem to fall into these four categories. This is a class of verb referred to as the pseudo-passive, which is similar to the middle in English. Examples of the pseudo-70

passive class include ku-shi 'cry-wet', chui-po 'blow-broken', zui-dao 'drunk-fall', lei-si 'tired-dead', yao-si 'bite-dead' etc. This class of verb differs from all the others in that the event structure and transitivity expressed by this class are not so predicted as the others. They show the following alternations: (2) a. Maojinku-shi-le towel cry-wet-asp The towel was wet because of somebody's crying. b. Qiqiu chui-po-le balloon blow-break-asp The balloon was blown broken. Following the analysis of Cheng and Huang (1994), the Vl in these sentences are activity verbs, and the compounds should be unergative or transitive plus an Agent subject. However, the subject of the Vl is Theme or Patient, in fact. These sentences can also show the ergative-causative alternation as follows: (3) a. John ku-shi-le maojin john cry-wet-asp towel John cried the towel wet. b. John chui-po-le 'qiqiu john blow-break-asp ' balloon John popped the balloon (blew the balloon broken). (3) cannot be understood as ergatives while they have no Causee subjects. Also, it is transitive rather than causative. Thus, it exhibits the unexpected pattern - transitive-ergative alternation, by the characterisation of Cheng and Huang (1994). Here, we are presented with some RVCs showing a transitivity alternation different from that of the other classes. In addition, this class of verb is also observed to behave as the English middle. English middles alternate with transitive verbs but not causative verbs, for instance: ( 4) This book reads well. Keyser and Roeper (1984) distinguish between surface ergatives, ie middles, from deep ergatives that alternate with causative verbs. (5) is an example of this Chinese middle, 71

which only differs from the passive construction in that the passive marker is not present: (5) Maojin bei (John) ku-shi-le towel bei john cry-wet-asp The towel was cried wet (by John). 3.2 Causative Verbs and Causative Constructions We have considered some examples from the causative verb. (eg Comrie 1993a, 19936; Goddard 1998, 2011; Goldberg 1995; Levin 1993; Matsumoto 1991; Shibatani 1976; Wierzbica 1988; Yuan 2005). Most of them are concerned with the syntax of causatives, but the semantic counterpart is very limited, as Wierzbica (1988) notes. The conception of causation is of particular interest to philosophers and sociologists since they are interested in the relations between different types of causation. The relations between them, in daily events, are not easy to observe, the syntax of a language, however, provides us evidence of causation if the evidence can be formally analysed. In so doing, an appropriate and consistent set of terminology should be employed in order that the phenomena are investigated in a systematic way. As Shibatani ( 1976) comments, it is not easy to define causative construction. He attempts to define the conditions for causative construction, though. Roughly, if a speaker believes that an event causes another event, there is a 'causative situation'. Also, the speaker believes the event caused will not happen without the causing event. Under this formulation, a sentence like John knows that Mary is married will not__be classified as causative. Rather, a sentence such as John made Mary marry is a causative sentence because the speaker commits himself/herself to the belief that the event of Mary's marrying will not happen without John. Causative verbs in English constitute a large stock of the vocabulary of the language. Examples are break, drop, flatten, etc. All of these causative verbs can undergo transitive/intransitive alternations. The transitive use of the verb can be paraphrased as something like 'cause to V-intransitive'. Thus, Mary broke the vase can be translated into Mary caused the vase to break. Levin ( 1990) divides the set of verbs referred to as causative verbs into two subtypes, namely, causative/inchoative and the induced action alternations. However, we shall not discuss them further, but we shall consider the English causative 72

construction described in traditional grammars. In English, the verbs have, make and let are usually considered to be causative verbs. The verb have has two syntactic patterns. ( 6) a. John had his shirt washed. b. John had Mary wash his shirt. In (36a) had forces the verb wash to be a past participle to mean that John's shirt was washed by somebody. But in ( 6b) since Mary - the one who did the washing appears, had requires the verb wash to be in the bare infinitive form. Like have, make and let also require the bare infinitive, as in (7) a. John made her wash his shirt. b. John let her do it. Note they require the to-infinitival when they are in passive construction. Languages seem to possess several constructions that can convey the causative situations, often with some subtle difference in both syntax and semantics. Apart from the resultative verb construction, Chinese also exhibits several patterns involving causation. The analytic structure with de as in Ta chi M hen baa "He ate a lot and was full' expresses the degree of the result. Another construction with dao is also seen as in Ta chi dao baa le. 'He ate a lot and became full'. The resultative can also interact with the ba-construction as in Ta ba mianbao chi-wan le 'He finished eating the bread', as will be discussed in Chapter Five. All these patterns referred to above involves transitivity alternations to some extent. 3.3 Lexical Information and Grammatical Blending While language can be approached from a typological perspective, it can also be viewed from the internal organisation and structure. The examination of the predicates above shows that the predicates interact with various constructions closely such as the middle construction and causative-resultative constructions. A certain class of verbs appears in a certain kind of constructions. To recap, a clause like (8) can have the middle meaning through the verb sell, while (9) does not work. This is on a par with the resultative construction, where ( 10) is unremarkable, but ( 11) is not all right. (8) The book sells well. (9) *The TV watches well. (10) John hammered the metal flat. ( 11) *John thought the metal flat. 73

These two phenomena suggest a strong correlation between the predicates and the constructions. While linguists of different persuasions are concerned with the relation between the lexicon and the construction, Construction Grammar (Boas 2003, Fillmore et al 1988, Goldberg 1995, 2006; Kay 1997; Lakoff 1987 etc) and other similar models such as Radical Construction Grammar (Croft 1991, 2001; Croft and Cruse 2006) seem to me to highlight the importance of this most. Take, for example, Boas's (2003) study of resultative constructions, which argues that while the discourse contextual factors are important to the interpretation and licensing of the resultatives, the event-frame semantics of verb further restricts the co-occurrence of the verb and the resultative phrase. The following examples (Boas 2003: 183) contain an intransitive verb run (12 - 16) and a transitive verb paint (17 - 19) in English, which licences different complements: (12) Jim ran. (13) Jim ran himself exhausted (14) Jim ran his feet sore. (15) Jim ran his shoes threadbare. (16) Jim ran the pavement thin. (17) Carol painted the house. (18) Carol painted herself to exhaustion. (19) Carol painted the brush to pieces. The verb run in (12) does not require any complements, while the ~'erb paint, a transitive verb, only requires the object noun phrase but does not require a resultative phrase in ( 17). It looks as if whether a verb is transitive or not does not solely determine the occurrence of the resultative (as noted by Huang and Lin 1993 earlier). The resultative phrases must have come from the argument roles from the construction (Goldberg 1995). However, without any concern with transitivity, this does not work perfectly either. ( 17), for instance, cannot do with the object, as shown in the following examples: (20) *Carol painted _ . (21) *Carol painted _ to exhaustion. (22) *Carol painted_ to pieces. 74

These examples show that while the intransitive predicate can become transitive, the transitive predicate cannot become intransitive - this is also observed in Chinese, for instance, John ku-shi le zhentou 'John cried and made the pillow wet'. The supplement of the object and other complements depends upon the general world knowledge we have - 'each event-frame includes information about the prototypical end result states or locations for each of its prototypical and non-prototypical event-participants' (Boas 2003: 183). This kind of restriction comes from our view of the world; in other words, it is semantic or conceptual information that contributes to the requirement of the result complements. Boas (2003: 183) attributes this to the 'conventionalised collocational patterns' that certain meaning occurs with certain structure. The representations of the event-frame of the prototypical sense of run and that of paint are as follows: (23) run paint ~ IGOA~ ~ Ag (W) ) Pt W stands for 'world knowledge', associated with the event-frame, which lexically licenses a particular type of result states and whether there are participants. In other senses of run, including theme participant and result complements, are supplied by world knowledge connected with them. The world knowledge can also provide some clues as to how the resulative construction in Chinese such as John ku-shi le zhentou can be interpreted. The simplified representations of the event-frame of the verbs ku and shi are as follows: (24) ku shi The Agent of ku is to a general cover for experiencer and causer etc. It is our world knowledge that a pillow cannot be crying but it can become wet. Of course, the Agent John can become wet but it is not the meaning of the construction and it is also not a 75

canonical interpretation of the sentence. The difference between shi and paint is that the latter subcategorises for a complement, without which ungrammaticality will result. (25) Zhentou shi le pillow wet asp The pillow was wet. (26) *Zhentou shi-le John pillow wet-asp john The pillow made John wet. (27) John ku-shi-le zhentou john cry-wet-asp pillow John cried and made the pillow become wet. The grammaticality of (26) suggests that there is a causative-resultative mechanism when the two predicates merge, and make ku a causative predicate and shi the result. This can also be seen as the interaction between lexicon and construction, both of which influence each other at different levels. The independent predicate only contains the information about its own canonical or prototypical behaviour and extra information has to be supplied by other means such as a construction or an appropriate context or universe of discourse. If we follow Fauconnier (1994, 1997), which recognise mantel spaces as important constructs for human beings' thought, through which linguistic expressions are used to mean, we may regard the two predicates about ku and shi as two different mental spaces: Mental Space Ag Pt Frame As observed in (25), there is one participant involved in both events, respectively, and create two mental spaces, which are structure by the Idealised Cognitive Model (Lakoff 1987: Ch4), which provides the frame of an event. The frame for crying, for example, has one participant - the crier; and the frame for being wet, again, involves one participant. The connection between these two events or frames lies in the cause and 76

effect relationship between crying and being wet. The participants and the events will create mappings between them and they fill in appropriate slots (Fauconnier 1997: 12). Given the insights of mappings, Construction grammarians (eg Goldberg 1995) suggest the fusion of participant roles given by the predicates and argument roles determined by the constructions, under two principles - the Semantic Coherence Principle and the Correspondence Principle. The former Principle asserts that 'only roles which are semantically compatible can be fused'; the latter Principle demands that a profiled participant role (such as Agent) 'must be fused with a profiled argument role of the construction' (Goldberg 1995: 50). Thus, ku, for example, can be represented in the following manner between syntax and semantics. Thus, the participant role 'crier' and the argument role 'agt' are fused together because they are semantically compatible and profiled. Sem EMOTION < agt > I I CRY < crier > l l Syn V SUBJ The second Principle also unites the construction and the lexical predicate in another way. The construction will provide the appropriate role such as the non-profiled argument role when there are more than two profile participant roles. In other words, the construction can assign roles so that ku-shi, if entering into the causative-resultative construction will be given the Causer argument role to the Agent -Agent often involves volition and, hence, causation. (28) John Zhentou c~ <>---.- ~ --.- This diagram shows the mappings between the lexical conceptual structure and the predicate argument structure, plus the constructional meaning in the background (The 77

three dimensional box aims to emphasise the constructional meaning lurking behind). First, ku assigns the Agent 'crier' role to John and then shi assigns the Patient (Theme) role to zhentou. Then when ku-shi merges as a causative-resultative structure, the new meaning coerces the Causer role and Causee role into John and zhentou, respectively. Thus, the conceptual structure (meaning) of the resultative construction finds its way to macro-manage the predicate argument structure of the lexical predicates. In our present constructional model, after the participant roles are fused with the argument roles, that is, crier > agent, the construction provides a further conceptual level of meaning, assigning the Causer and the Causee roles. In this line of reasoning, we find the concept of grammatical blending (Fauconnier 1994, 1997) relevant to the discussion of the Chinese resultative construction. Following construction grammarians such as Goldberg (1995), a resultative can be seen as a metaphorical extension of the Caused Motion Construction - 'A causes B to move to C by doing D'. However, to adopt it, we will need to modify the schema. The following schema is our modified version for the Chinese resultative (Fauconnier 1997: 173, modified): Cause Motion Construction and its schematic meaning NP AO V DO VP/AP BO NP co Input 1 Input 2 Here, we replace Fauconnier's 'Move' to 'Change to' to denote the change of state in the resultative structure. Since the order of the result phrase and the Patient argument is different, as is reflected in the above diagram - C changes to B. Further, since the result can also be predicated of the Agent argument, there is another modified kind of mapping: 78

-Schematic meaning of the Chinese resultative construction NP AO V DO VP/AP BO (NP) (CO) Input 1 Input 2 The above mappings can accommodate the two kinds of Chinese resultatives below: (29) John da-po-le huaping. john hit-break-asp vase John broke the vase. (30) John chi-bao le (fan). john eat-full asp (rice) John ate and became full. This diagram allows the post-verbal NP to be optional as in John chi-bao le. Instead of having one kind of grammatical blending, we suggest that there are two for the resultative. This echoes Croft's idea that there is 'mapping of conceptual function onto grammatical form'. The blending model can be translated into the Construction Grammar model, which we only use minimally (e.g. Goldberg 1995, 2006). We propose the following constructions: The ku-construction is an intransitive construction Sero BECOME < agt > R t R : instance, PRED < > Means J t Syn SUBJ 79

The shi-construction is an intransitive construction Sem BECOME < agt > I R l R : instance, PRED < > means l l Syn V SUBJ The ku-shi resultative Sem CAUSE-BECOME < agt pat > I R l l R : instance, PRED < > means l l l Syn V SUBJ OBJ It is the over-arching semantic and conceptual structure of the resultative construction that produces the ku-shi construction while both the verbs do not license a Patient argument when they are used in isolation. The following diagram shows this in a three-dimensional model, which attempts to emphasize the power of the construction: John zhentou cause-result r-ru7 ~ ~ ~ ~ ~ The entire construction implies that causation and result are more powerful than lexical licensing, assigning the Causer and Causee roles to the participants in the structure. Once the verbs enter into the construction, they are at the mercy of the structure. The construction comes close to human experience and our world knowledge, which helps us interpret the ambiguity of such a sentence as John qi lei le ma - where either John or the horse may be tired in a riding event. 80

This is controversial, however. Shi C (2008: 285), for example, conducted a survey in Beijing in 2002 and found that more than 30% of the informants did not accept the sentence Wo qi lei le ma, and for the sentence Xiaozhang kan lei le xiaoshuo ('Xiaozhang read a novel and got tired'), more than 50% of the informants did not think it was acceptable. Instead, most informants prefer to use the serial-verb construction Wo qi ma qi lei le ('I rode the horse and got tired') to obtain the reading that the preverbal argument wo got tired. Note even if the Agent Causer may become a Causee in a certain context, the Patient Causee cannot become a Causer 13, which observes the universal principle of iconicity. As has been noted, the Chinese resultative is causative in nature and thus is transitive. There are intransitive verbs becoming transitive in the Chinese resultative construction, but it does not allow two transitive predicates to become intransitive. This is naturally constrained by the entire resultative construction, whose nature, again, is causative and resultative. Only through the interaction between lexical predicates and constructions will we understand the nature of the resultative construction. The understanding of the nature of the resultative is further reinforced by the interaction between syntax, semantics, pragmatics and discourse and they are not to be considered as straight discrete linguistic levels. The question of the relations between syntax and semantics, however, has triggered many studies (eg Dixon 1991; Levin and Rappaport 1995) and one way to get it is through the study of argument structure. We have not argued for a specific position held by linguists like Perlmutter, who maintain that unaccusative construction is syntactically represented but semantically determined. Nevertheless, this hypothesis may be extended to any other constructions of argument structure since the behaviour oflexical items contributes to a large extent determining the structure of syntax. In echo to Chomsky's remark (1957), 'the general theory' has been developing and will remain the centre of researchers' interest in the foreseeable future. 13 Exceptions occur. Li ( 1995) takes a sentence such as John zhui-lei le Mary to have one of the meanings that Mary chased John and she got tired. 81

Chapter4 Alternative Analyses of the Resultative Construction 4.0 Introduction This chapter discusses several alternative analyses of the resultative constructions. We draw on English examples and then Chinese ones. We will argue that the Chinese resultative construction do not observe the Direct Object Restriction, as it applies to English. Following Goldberg (1995), we will consider how the Chinese resultative works within the constructional framework. Meanwhile, we draw our attention to a case in Chinese: Wangmian si-le fuqi 'Wangmian's father died', whose purpose is to discuss the two principles related to argument realization, and thus related to the resultative construction. The resultative construction is then argued to be a case of catenative, then a phenomenon of ( co )subordination and finally mini-constructions (Fong 2019). 4.1 DOR and Unaccusativity One of the problems raised by the study of resultative constructions is to explain the strong transitivity of the construction - the Direct Object Restriction (DOR) and the nature of unaccusativity found in the constructions. Levin and Rappaport Hovav (henceforth L and RH) (1995) investigate the unaccusative and unergative verbs in English by virtue of the resultative construction. Quite different from Goldberg ( 1995), LandRH attempt to bridge the gap between a syntactic approach and a semantic approach within the GB framework. L and RH start with a review of studies of the distribution of resultative phrases, where the DOR is shown to hold for resultative constructions based upon transitive, unergative, unspecified object, and passive and unaccusative verbs. The following discussion attempts to apply their analysis based on the DOR in Chinese resultative complex predicates, simultaneously pointing out similarities and differences between English and Chinese in this respect. In English several classes of verbs can occur in the resultative constructions. Resultative phrases can occur with transitive verbs, of whose objects may be predicated, for example: 82

(1) Woolite safely soaks all your fine washables clean. 14 In example ( 1) the resultative phrase clean predicates of the object washables. Thus, it is the washables, not W oolite, that are clean. This phenomenon is said to be generally true in English, where the postverbal NP is predicated ofby the resultative phrase, and hence, the Direct Object Restriction proposed by Land RH. However, Land RH observes that a seemingly counterexample of the DOR occurs when the resultative phrase is predicated of the subject. According to L and RH, English does not allow such a resultative construction. However, when there seems to be a case like this such as the sentence Mary burned the cookies dirty, there does not exist a resultative reading. Instead, the sentence has a depictive interpretation. Thus, the sentence cannot mean that Mary got dirty as a result of burning the cookies, but it can only be interpreted as Mary was dirty while she was burning the cookies. As a consequence, Land RH state that they are not aware of any counterexamples to the DOR that involve transitive verbs (1995: 35). Their claim may well be valid for English, but I believe that seemingly counterexamples to their claim could be found in other languages. In fact, Mandarin Chinese is one of such languages, where the unique resultative complex predicates exhibit some opposite properties to the DOR that L and RH propose. The following examples can illustrate this point: (2) a. John zhui-lei-le Mary le john chase-tired-asp mary par John chased Mary so that John/Mary got tired. b. John chi-bao .fan le john eat-full rice par John ate the rice and became full. c. John ting-fan-le yinyue john listen-annoy-asp music John listened too much of the music and he became annoyed. d. John he-zue-le jiu john drink-drunk-asp wme John drank wine and got drunk. 14 English examples are taken quite freely from published work such as Land RH (1995) and Goldberg (1995), where appropriate. 83

e. John xue-hui-le john study-know-asp John has learned Chinese. Zhongwen Chinese All of the resultative verb compounds in example (2) show that the DOR wrongly predicts that the postverbal NP objects are predicated ofby the resultative V2s. Contrary to the DOR, all V2s are predicated of the subjects John, except that (2a) is ambiguous in that the resultative V2 can be predicated of either the subject John or object Mary. Thus, the resultative construction based on transitive verbs in Chinese resultative complex predicates can be predicated of the subject, or occasionally of the object. English resultative constructions are transitive m nature. In the case of unergative verbs occurring in resultative constructions, a fake reflexive object (Simpson 1983) or a postverbal NP is added to yield a grammatical resultative. Let us consider example (3): (3) a. Dora shouted herself hoarse. b. *Dora shouted hoarse. c. The dog barked him awake. d. *The dog barked awake . Example (3b) is ungrammatical for a resultative reading because English does not allow the resultative construction based upon unergative verbs without any qualifications. The reflexive pronoun herself is such a qualification yielding the resultative construction grammatical. The subject and the reflexive have the same referent in the discourse of universe. This phenomenon is referred to as the 'fake object' case. Example (3d) is ungrammatical for the same reason but this time a postverbal NP is not a reflexive pronoun but a personal pronoun denoting a different refe_rent of the subject. Based on the above evidence from fake object cases and postverbal NPs, L and RH argue that the Direct Object Restriction can be extended to resultative constructions based upon unergative verbs. However, the Chinese resultative construction with unergative verbs does not conform to this restriction. For example, the Chinese counterpart of the English example (3a) can do without the fake reflexive object. 84

( 4) Dora jiao-ya-le dora shout-hoarse-asp Dora shouted herself hoarse. It does not equal to saying that Chinese resultative constructions based on unergative verbs cannot allow a postverbal NP. In the case of ( 4), a reflexive object can occur in the resultative constructions in Chinese. (5) Dora jiao-ya-le dora shout-hoarse-asp Dora shouted herself hoarse. (ta) ziji (her) self This means that the reflexive object ( ta) ziji in Chinese in cases like ( 4) is optional, without which does not render the resultative construction ungrammatical. A similar case of the personal pronoun referent is also observed in Chinese resultative constructions, where a postverbal NP is allowed, as in the case of (3c). The Chinese equivalent of (3c), however, does not give us any help regarding the issue here, because the verb bark in Chinese can be transitive, and so it can take a postverbal NP, as illustrated in ( 6). (6) Zhe-ji gou fei ta this-cl dog bark 3sg The dog barked at him/*The dog barked him. When the verb fei 'bark' combines with xirig 'awake' to form a resultative complex predicate fei-xing 'bark-awake', the whole predicate is still transitive 15• (7) a. Zhe-ji gou fei-xing-le ta this-cl dog bark-awake-asp 3sg b. Zhe-ji gou fei-xing-le ta le this-cl dog bark-awake-asp 3sg par The dog barked him awake. Both sentences (7a) and (76) are acceptable equivalents of (3c); while the presence or absence of the sentential particle le does not concern us here. The important point to note is that the entire resultative verb compound fei-xing 'bark-awake' is 15 The transitivity of a resultative verb compound does not rely on a particular component verb in the compound, as argued in Huang and Lin (1992), Li (1991), etc. 85

transitive and thus takes the postverbal NP ta '3sg', equivalent to the corresponding case in English. Like the English example in (3d), the Chinese resultative complex predicate of fei-xing 'bark-awake' requires the postverbal NP, without which ungrammaticality of the sentence is produced. (8) *Zhe-jigou this-cl dog fei-xing-le bark-awake-asp *The dog barked awake. Now things are more complicated than we thought. We cannot seem to simply reject Land RH's DOR. Chinese resultative verb compounds seem to conform to the DOR in example (7), where a postverbal NP must be inserted to produce a grammatical resultative sentence. In example (5), however, the DOR is not observed by the Chinese resultative verb compound jiao-ya 'shout-hoarse'. Thus, what we can provisionally conclude is that the DOR applied in the case of Chinese should need modifications. We shall leave this issue aside for the time being and continue our examination of the DOR in the resultative constructions based upon other types of verbs. Alongside the resultative construction based on unergative verbs that require postverbal NPs in English, a related construction is that require postverbal nonsucategorised NPs (generally denoting a body part) which are possessors coreferential with the subject of the verb, for example: (9) a. Sylvester cried his eyes out. b. *Sylvester cried his eyes. c. Sleep your wrinkles away. d. *Sleep your wrinkles. Although the postverbal NPs in the example are subcategorised and reflexive, they involve possessive pronouns coreferential with the subject. L and RH thus argue that this case is similar to the fake object case and accordingly, the DOR still holds. In Chinese, We shall show that the V2 of a resultative verb compound functions like the prepositions out and away in example ( 9): (10) a. Sylvester sylvester ku-qu cry-out Sylvester cried his eyes out. 86 yanjing lai le eyes come par

b. Shui-zou sleep-go Ill you de de Sleep away your wrinkles. zhouwen wrinles Thus, qu 'out' is a directional verb forming a resultative compound with ku 'cry', and zuo 'go' forms a resultative compound with shui 'sleep'. It seems reasonable to compare qu and zou with the English particles, for instance, in this case. In fact, most directional verbs function like many English particles to yield a resultative reading as in (9 and 10). Besides, some intransitive verbs, though sometimes can be used as transitives, can occur in resultative phrases with an unspecified object interpretation. Consider (11 ). ( 11) a. Slyvia ate. b. Sylvia ate the grapes. Within this class, some resultative constructions based on unergative verbs and unspecified object verbs have 'arguments' as the object of the preposition heading an oblique PP complement (eg Jackendoff 1990), for example: (12) The dog barked at them. This approach is rejected by L and RH in that 'it is difficult to imagine such a source for many of the fake reflexives or nonsubcategorised NPs' ( 1995: 38). We follow Land RH's argument here and assume that Jackendoffs approach is not adequate to cover most types of resultative constructions. In terms of example (12), it is not natural to apply the same analysis of Jackendoff and others in the Chinese resultative compound, as we have seen in -example (6) above. Moreover, being a transitive verb, fei 'bark' does not have an 'argument' as the object of the preposition heading an oblique PP complement. However, fei 'bark' can be used as an unergative verb: (13) Gou fei dog bark The dog barked. In this case, fei cannot take any complement, let alone oblique PP complement. The case of verbs with unspecified objects also finds its parallel in Chinese. Example (14) illustrates this similarity between English and Chinese resultatives: 87

(14) a. Sylvia ate. Sylvia chi-le b. Sylvia ate these grapes. Sylvia chi-le zhe-xie putao Chi 'eat' can take unspecified objects, or take nothing at all. However, the fact is more complicated than this if we have a subject like qi 'chicken '. (14') Qi chi-le Chicken eat-asp a. Chicken ate. b. Chicken was eaten. In terms of passive and unaccusative verbs, the DOR seems to be violated in that they predicate of their surface subjects, for instance: ( 15) The river froze solid But L and RH recognise the surface subject as the underlying object of the passive and unaccusative verbs; thus, the DOR still holds. Unaccusative verbs cannot occur with resultative phrases predicated of fake objects or nonsubcategorised NPs, though, eg (16) *The snow melted the road slushy. This property of unaccusative verbs can be explained by Burzio's Generalisation, which states that unaccusative verbs cannot assign Case. Unaccusative verbs cannot even have cognate objects, as unergative verbs can. The movement analysis of passives and unaccusatives in resultative constructions in English appears to be applied in the case of Chinese resultative constructions. ( 17) Heliu jie-bing le river form-ice par The river froze solid. In terms of (16), the Chinese counterpart is also not grammatical unless the ba-construction is used. (18) Xue ba lu rong cheng xueni snow ba road melt become slush 88

Finally, resultative phrases cannot be predicated of VP-internal NPs which aren't direct objects like obliques, eg (19) a. John loaded the wagon full with hay. b. *John loaded the hay into the wagon full. All of these phenomena are captured by the DOR, assuming a class of unaccusative verbs in English. In Chinese, (19) will be like the following: (20) a. John feng-man yiche de cau b. *John feng cau zai che sheng man In this section we have shown the similarities and differences between English and Chinese resultative constructions, and at the same time put some challenges on the Direct Object Restriction proposed by Land RH. Furthermore, we agree with Land RH (1995: 54) that 'the interpretation of the resultative construction is compositionally derived: specifically, it is derived from the meaning of the verb plus that of the resultative XP'. L and RH show unaccusativity exhibited in English by means of resultative constructions. Although they present both the syntactic and semantic approaches to the issue, we lean towards the semantic end. 4.2 Constructional Analysis The theory of Construction Grammar (eg Goldberg 1995, 2006; has been popular since the late 1980s after the publication of Fillmore et al (1988), and a very important work is that by Goldberg ( 1995), which provides one of the early treatments of resultatives in this framework. This section will discuss the main arguments of the resultative, the different kinds of resultatives, the interaction between middles and resultatives, and finally different accounts of anlaysis. Along with Goldberg we also hold that 'the necessary constraint on the appearance of resultatives can be stated in semantic terms: the resultative can only apply to arguments that potentially (although not necessarily) undergo a change of state as a result of the action denoted by the verb' (Goldberg 1995: 180). However, exceptions of 'fake object' cases to the semantic constraints of patienthood led linguists to state the constraint in syntactic terms, but Goldberg argues against it. The phenomena Goldberg tries to account for include: 89

(i) resultatives apply to direct objects of some transitive verbs, but not others: I had brushed my hair very smoothly. *He watched the TV broken (ii) resultatives apply to subjects of passives corresponding to acceptable actives: The tools were wipe clean. (iii) resultatives often apply to subjects of unaccusative verbs, but not those of unergatives: The river froze solid. *He talked hoarse. (iv) resultatives sometimes occur with postverbal NPs without argument relation with the matrix verb: She laughed herself crooked. Having the facts laid out, Goldberg considers the status of the postverbal NP 'fake object'. Some processes such as middle formation, adjectival passive formation, and process nominalisation are observed to apply to direct internal arguments but not 'fake object' resultatives, for instance: (i) Transitive resultative: He hammered the metal (flat) middle formation: This metal hammers flat easily. -adjectival passive: the hammered-flat metal nominalisation: the hammering of the metal flat (ii) Fake object resultative: He drove his tires *(bald). middle formation: *Those tires drive bald easily. adjectival passive: *the driven-bald tires nominalisation: *the driving of the tires bald The middles in (I and ii) can be ungrammatical in the same pattern. Most adjectival passives and nominalisations based upon the transitive resultative are also ungrammatical. An implication is yielded if 'X occurs in the middle construction and 90

adjectival passive and nominalisation construction, then X is an argument, the converse is clearly false' (Goldberg 1995: 183). Middles also provide Goldberg with the probe to the examination of this statement. One of the properties of middles is the inherent quality of the patient subject argument responsible for the property expressed in the predicate. Resultatives are also observed to exhibit the same kind of property, for example: (7) The metal hammers flat easily. (ie People can hammer the metal flat easily because of an inherent quality of the metal) However, the fake object case is not entirely compatible with middles. Fake objects occur with objects coreferential with the subject: (8) a. He cried himself asleep. (fake object) b. *He cries asleep easily. (middle) Fake object cases do not have a subject patient argument with inherent quality that makes the predicate true. (9) a. The joggers ran the J)avement thin. (#People can run the pavement thin easily because of an inherent quality of the pavement) b. #The pavement runs thin easily. The fact that middles typically occur with un_expressed indefinite volitional agent also rules out the fake object cases. Fake object cases often express a negative consequence, and yield anomaly together with volitionality. It is, however, noted that given the right context, middles with fake object cases are acceptable. Given the semantics of middles, Goldberg explains why fake object resultatives cannot normally be acceptable as middles. Other proposals exist (eg Jackendoff 1990, Bresnan and Zaenen 1990, Van Valin 1990). The idea that the postverbal NP is an argument is rejected by Jackendoff (1990a), who proposes that it is an adjunct. However, Goldberg shows that the postverbal NP fails the test for adjuncthood; for example, it could be the subject of a passive construction. 91

(10) a. The baby was barked awake every morning by the neighbour's noisy dog. b. *The dog barked ferociously the baby awake. ( 1 0b) is not grammatical because the fake object and the predicate cannot be intervened. Goldberg thus argues that Jackendoffs adjunct rule is not warranted. Furthermore, the Lexical-functional grammar approach is postulated by Bresnan and Zaenen (1990), where a feature [-r] ('unrestricted') is supposed to mark the argument predicated of by the resultative. This feature is shared by subjects and objects but not prepositional objects and secondary objects. The argument then bears a thematic role of patient/theme or it does not. In the latter case, the fake object is not assigned any semantic role, and is assumed to be unsubcategorised for by the verb. Thus, it is assigned the 'unrestricted' feature. However, the problem with this proposal, as Goldberg points out, is how 'the internal object makes its way into the argument structure of the verb in order to receive its critical [-r] marking' ( 1995: 187). Goldberg postulates a lexical rule that accounts for the problem: 'resultatives can only be predicated of patient argument' (1995: 187). Alongside Goldberg's lexical rule, Van Valin (1990) proposes a similar rule for resultatives. He argues that a resultative should be predicated of an 'undergoer', which is defined as not necessarily undergoing a change of state, and in English the argument that can be passivised is an undergoer. This proposal wrongly predicts that the sentence that has an undergoer but non-patient argument is acceptable. Van Valin also claims that resultatives only occur with accomplishments and achievements, and argues that unergative verbs do not allow resultatives. Goldberg argues that activity verbs lil

(11) a. wipe (wiper wiped) He wiped the table clean. b. talk (talker) He talked himself blue in the face. In the case of wipe, the construction adds the result-goal argument if the verb has a participant role fusing with the patient role of the construction. However, in the case of talk, the construction adds the patient and result-goal roles. However, the intransitive resultative construction may be a little different, and requires a further apparatus in the construction. For example, became will need two resultative construction grids. The constructional approach can also capture the idiosyncrasy of some resultatives, such as eat with sick in He ate himself sick. In addition, Goldberg notes the constraints on the resultative construction. The following diagram is the resultative-construction template (Goldberg 1995: 189): Sem CAUSE-BECOME < agt pat result-goal > I R R : instance, means Syn PRED l < l OBL AP/ PP > The argument structure construction provides the argument roles for the resultative while the verb supplies the participant roles, which 'may be semantically fused with argument roles of the argument structure construction' (1995: 50 original emphasis) . In this way, the construction can help contribute to the argument that the verb does not subcategorise for. For instance, the verb sneeze does not subcategorise for an NP object, and yet it can appear in a construction with an object: (12) John sneezed the napkin off the table. As in the canonical case, the verb sneeze does not require an object and the construction does not have to supply the oblique role. 93

Furthermore, there are (animate) instigator, aspectual, end-of-scale constraints and constraint against deverbal adjectives. In the two-place resultative construction, the subject argument must be an animate instigator though not necessarily an agent, eg She coughed herself sick. But lexical causatives do not observe this constraint, eg Water filled the tub half full. Diverging from others, Goldberg postulates the aspectual constraint that '[t]he change of state must occur simultaneously with the endpoint of the action denoted by the verb' (1995: 194), eg Harry shot Sam dead. The class of adjectives that can occur in the resultative construction is limited. These adjectives normally code a state with a delimited lower bound and hence nongradable, eg He cut himself free, ? a little free (cf Wechsler 2005). However, used with fake objects, the adjectives are understood to have a clearly delimited boundary, eg He ate himself sick. This implies that the patient argument cannot bear more than the normal limit. Exceptions include, again, the lexical causatives, eg He made her a queen. Finally, deverbal adjectives derived from participles cannot occur in the resultative construction, eg She painted the house red vs She painted the house reddened/reddening. As Goldberg argues the resultative construction can apply to the patient argument, (even in fake object cases). This can only be done in semantic terms that resultatives only apply to arguments potentially undergoing a change of state as a result of the action denoted by the verb. If we examine the definition of a construction in Goldberg's (1995: 4 original emphasis) type of Construction Grammar, we will regard the resultative in Chinese is also a construction: C is a CONSTRUCTION iffd,f C is a form-meaning pair such that some aspect of Fi or some aspect of Si is not strictly predictable from C's component parts or from other previously established constructions. Although Goldberg may not have meant this to apply to different languages, this definition captures the idiosyncrasies of the resultative constructions typologically. We have seen examples like He wiped the table clean, whose patient argument 1s a propotypical argument and the result of being clean is the expected result. The unpredictable results are not difficult to find, and we have also seen the example John sneezed the napkin off the table, whose oblique argument is not expected from the verb sneeze. 94

4.3 A Digression - Wangmian si-le fuqin At this juncture, the classical example in Chinese may be appropriate. (13) Wangmian si-le fuqin Wangmian die-asp father Wangmian's father died. Starting from the lexical semantics of the verb si 'die', one cannot expect an object theme role after si. The argument structure of si is <~ > and it subcategorises for nothing si [Theme, 0]. The problem here is that father occurs postverbally and Wangmian occurs preverbally and they have a relationship -father and son. Instead of Wangmian's father died, where father occurs before the verb (that is what one will say in English), Chinese has this 'unusual' word order for the verb si. This is unusual in that it violates several principles or constraints for argument structure, one of which is the Argument Realisation Principle (Goldberg 2005: 18, Rappaport Hovav and Levin 1998): There must be one argument XP in the syntax to identify each subevent in the event structure template. Following this principle, there should be an argument for a subevent. If the verb si is a predicate expressing the event (there ·is no subevent here), it is predicated of an argument, in this case, father. However, there is one more argument Wangmian, and it occupies the preverbal position - a typical subject position, which should be barred by a relevant condition - Subevent Identification Condition (Goldberg 2005: 18, Rappaport Hovav and Levin 1998: 112): Each subevent in an event structure template must be identified by a lexical predicate. The appearance of an extra argument is puzzling, according to this principle, for there are two arguments and one predicate. The event of dying is identified by si. Nevertheless, the argument Wangmian seems to be redundant. We would expect another lexical predicate to link with Wangmian. If we follow the principles and consider this sentence: (14) Fuqin si-le Father die-asp (My) father died. 95

Under a natural interpretation, if this is acceptable, that will mean 'my father died' and my is implicit. Without any qualification, the noun father here will imply the speaker's father. The function of the first argument is to identify the status of the father ie Wangmian's father. The genitive noun phrase Wangmian's provides the identity of the head noun father. The Idealised Cognitive Model (ICM) (Lakoff 1987: 68-76) suggests that father is a category and a radial category (Lakoff 1987: 91-114). There are central and less central subcategories eg a birth father and a foster father. I believe that Wangmian's father has some bearing on this categorisation that in the world, to understand the sentence Fuqin si-le, we will need to know whose father died. In the Chinese literature (eg Liu 2010; Zhang 2002), most attention has been paid to the meaning of si 'die', but fail to notice this aspect of meaning of father. It seems that father can 'assign' a semantic role to Wangmian, which is the possessor, similar to the function of the English genitive, to determine the nature of the following noun. Furthermore, the ICM of death is that somebody undergoes a change of state from being alive to being dead. There is no inherent meaning that more than one person dies, so that Wangmian fuqin si-le cannot mean that both Wangmian and his father died. Thus, Wangmian is not an argument of si, and its role is perhaps given by father and supported by our encyclopaedic knowledge of the world. This example Wangmian si-le fuqin poses theoretical problems to the principles above. In the resultative construction the above principles receive further challenge. If we take the definition of states of affairs in Cristofaro (2003) and that of (sub )event in Goldberg (1998) (more elaboration follows), the resultative seems to violate them for discourse-pragmatic factors. In the eating event, an eater and the food are expected, and thus a natural sentence with eat is as follows: (15) John ate dinner with Mary yesterday. Where the arguments of eat, ie John and dinner appear. And yet, often the Patient argument is omitted as in (16) John has already eaten. as an reply to the question: Has John eaten dinner yet? It makes sense to suggest that there is a null complement/argument here after eaten, which is recoverable from the discourse. In a Chinese resultative counterpart, for example, Ta chi-baa le fan meiyou?, a similar reply can occur: Ta chi-baa le, dropping the patient argument fan because this is 96

understood in the context. Two subevents occur here: chi 'eating' and baa 'full', which are supposed to be identified with their arguments, and yet chi can only be predicated of the Agent, although the Agent is also the Theme for baa. A more serious problem arises in this example: ( 17) Ta ku-shi le zhentou. he cry-wet asp pillow He cried and the pillow got wet. While the two subevents are identified by the lexical predicates ku and shi, respectively, there is another subevent (Goldberg 2005: 19) - the causing subevent (CAUSE), which is not identified by any single predicates but connects the two subevents. This can be seen as the meaning of the sentence is decomposed: (18) Ta BECOME [zhentou ] The causing subevent does not seem to come from either of the predicates, and leads us to consider the constructional and pragmatic factors. Before we attempt to explore the pragmatic factors, it may be appropriate for us to consider the two positions which are prominent in a clause and discourse ie the subject and object positions. The question to ask is how the roles are linked to these positions. One of the proposals that attempts to answer this question is worked out by Dowty (1991) in his Proto-Agent and Proto-Patient hypothesis. This theory is based upon the idea tliat role types are not 'discrete categories' but 'cluster concepts'. Arguments of a role type are considered to have various 'degrees of membership'. Explaining these two roles, Dowty presents preliminary lists of entailments that characterise them, which include: (i) contributing properties for the Agent Proto-role, i.e. (a. Volitional involvement in the event or state, b. Sentience (and/or perception), c. Causing an event or change of state in another participant, d. Movement (relative to the position of another participant), and e. Exists independently of the event named by the verb); (ii) contributing properties for the Patient Proto-role, i.e. (a. Undergoes change of state, b. Incremental theme, c. Causally affected by another participant, d. Stationary relative to movement of another participant, and e. Does not exist independently of the event, or not at all). To illustrate these entailments, Dowty gives the following examples: 97

(i) Proto-Agent: a. Volition: John is being polite to Bill. b. Sentience/ perception: John knows/ believes/ is disappointed at the statement. c. Causation: His loneliness causes his unhappiness. d. Movement: The rolling tumbleweed passed the rock. e. Independent existence: John needs a new car. (ii) Proto-Patient: a. Change of state: John made a mistake. b. Incremental theme: John crossed the driveway. c. Causally affected: Smoking causes cancer. d. Stationary relative to another participant: The bullet entered the target. The sets of entailments are like distinctive features to distinguish Proto-Agent from Proto-Patient. The boundaries, however, are not clear-cut. Examples are lilce a computer doing some actions or having some sentience, e.g. The machine switched itself off. Causation is another similar problem. The point Dowty tries to make here is that the unclear boundaries of the entailments just match the events found in the real world. To further support his ideas of these entailments, Dowty remarks that people do concern themselves with causations in everyday life, with whether an action is volitional, etc. If Dowty (1991) is right, examining the cause and effect as expressed in language 1s an important that a language should express this phenomenon. Resultative constructions serve to provide us with such a testing ground. In the previous example Ta ku-shi-le zhentou, the subject is a human Agent, so it is volitional, perceptional, able to cause something to happen, etc. And the object is a Patient, which involves a change of state, is causally affected, and stationary. In other words, the more proto-agent properties an argument has the more lilcely that it is an Agent, potentially aligned with the subject. The two positions are filled and they express prominence of the roles ( Goldberg 20056: 185). The pillow zhentou undergoes a change from being dry to being wet, and the reason for this is that the Agent cried. In other words, Agent and Patient (Undergoer) are both prominent. If the sentence is paraphrased as 'That the man cried caused the pillow to become wet', the cause is overtly expressed. In the original sentence 98

the causing event is implicit inside the construction. This may also be seen as a causative sentence. Another aspect of this construction is that the referents ( the arguments of the predicates) are overtly expressed. The sentence * Ta ku-shi-le does not work if the patient argument is missing, because according to the Argument Realisation Principle, there must be an argument for shi to identify the subevent in the event structure. In other words, the Argument Realisation Principle further reinforces the transitivity nature of the resultative construction, which typically requires an affected patient. Nevertheless, if we adopt this Principle crosslinguistically, as in Chinese, it will have to be reformulated as the following: There is at least one argument XP for the whole event (including subevents) in the event structure template. This will cover cases where there is only argument which is concerned with two subevents - the whole resultative event such as chi-baa 'cry-wet' that only requires on overt argument. The verbs chi and baa are only predicated of the same argument as in John chi-baa le ('John ate and became full ') . The relevant Subevent Identificaiton Condition, however, can be observed in both English and Chinese. To recap, the Condition states (Goldberg 2005: 18, Rappaport Hovav and Levin 1998: 112) that ' [ e] ach subevent in an event structure template must be identified by a lexical predicate'. In the chi-baa case, the subevents are identified by two lexical predicates. This principle seems to work well but it will carry more force if a further factor can be identified. For a patient argument to !;>e omitted, there is another factor ie the 'low discourse prominence' (Goldberg 2005, 2006). If the patient argument is not emphasised or in the focal position, it can be omitted (Goldberg 2006: 196). (19) He was always opposed to the idea of murder, but in the middle of the battle field, he had no trouble killing. There is a 'discourse topic' here and killing does not take an overt patient argument. Returning to our example ku-shi, the patient argument must be expressed because there is no discourse relevant information from which the patient argument of shi will be retrieved. Apart from discourse factor, the Subevent Identification Condition also suggests that the predicates expressing the subevents have to be very close. The Chinese resultative construction contains together the two verbs expressing the subevents, which are inside the resultative event. The first subevent is a cause and the second is a result or 99

effect, which reflects the natural order of the events in the real world. This order expresses the nature of iconicity that items that are close together in terms of function or cognition are arranged closely together (Giv6n 1991). The order of ku-shi demonstrates this: first the cause, then the result; besides, the two verbs are so tight that they do not allow any intervention ( except bu 'not'), for instance,* ku-feichang-shi ('cry-very much-wet') . They form a complex predicate syntactically and semantically integrated. There is, in other words, iconic motivation in the Chinese resultative construction. The iconicity principle is also connected to the conceptual space or semantic map model as 'The universals of language are found in conceptual structure and in the mapping of conceptual function onto grammatical form' (Croft 2001: 105). 4.4 Catenative Analysis This section attempts to argue that the resulative verb is a legitimate verb and not just a morpheme in a compound, and thus to establish our argumetns for the catenative analysis that follows. As has been said earlier, the broad sense of resultative structure includes directionals, and since resultatives are traditionally classified as verb compounds, directionals are also considered to be compounds (eg Chao 1968, Henne et al 1977, Hu et al 1978, Li and Thompson 1981, Packard 2000). The majority of grammarians working on Chinese take this view, partly because I think it is intuitively acceptable and partly because it is a convenient approach to the structure of Chinese, particularly when it is regarded as an SVO language, where the V has as its constituents the verb compounds: (7) John (S) zou-jin-le (V) wuzi (O). John walk-in-asp house John walked into the house. This model has the advantage that the predicator or verb phrase consists of several verbs similar to the verb phrase in English such as will have found him. If, however, the verb phrase is analysed further, it is not difficult to imagine the different layers inside it, as a phrase structure tree can show, for instance: 100

(8) IP VPl //r I Vl V2 NP will have found him In line with current thinking, all the verbs above are heads of their own phrases. That is, found is the head of found him, which is complement of have found him, of which have is head. The entire phrase is headed by will, which projects into an IP. If there is universal grammar, which is not very controversial now, a similar template of verb phrase structure is expected of Chinese. The two verbs in the Chinese 'compound' zou-jin 'walk in' and zou-jin wuzi 'walk-into house' can be argued to belong to different layers, rather than having a flat structure ( cf Culicover and J ackendoff 2005). (9) VPl //1P2 Vl V2 NP ZOU ;m WUZl (10) V V NP ZOU ;m WUZl Analysis (9) shows hierarchical structure of the VP, while ( 10) proposes a flat structure. There is a third possibility: [VP [ V zou-jin] wuzi] where zou and jin form a constituent - the verb, which many scholars endorse, for instance, Huang et al (2009: 45 (14) and Tang 2010: 123 (75), although their examples may not be the same. While the flat structure has the advantage that one does not need to choose the head, it fails to indicate the internal relations of the items inside the phrase. Acknowledging this, however, we will present evidence that the first analysis ie (9) is preferred. It shows that instead of having zou-jin as one single constituent, jin forms a constituent with the following noun phrase wuzi before forming a verb phrase with zou. 101

However, Huang et al (2009: 45) show that a similar item chi-guo 'eat-over' belong to the same constituent and they actually form a verb. They consider the three verbs sheng-chi-guo 'raw-eat-over' (eaten something raw) 'form a compound with the argument structure <8a <86-81 >>. These theta-roles are assigned to the subject and object of the compound in syntax, yielding the reading that the object of the verb chi also refers to the material which is raw'. This is done, according to Huang et al, by assuming that chi is the head of the compound. Similar to Huang et al (2009), Butt (2012: 61 -62) analyses the Chinese directional items as 'etymons' and light verbs, constituting a complex predicate and she notes that the item guo can have four uses - as a main verb, as a directional, as a perfect marker and as a light verb. The following are the examples with guo: (11) Guo malu. cross road To cross the road. (12) Pao guo qu. run cross go To run across. (13) Wo shuai duan guo tui. I fall sever cross leg I broke my leg once (but it has healed since) . Apart from these uses, Butt notices another use - guo as a light verb. (14) Wo chi guo le. I eat cross asp I have eaten. This use is different from the previous ones, Butt argues, for there are phonological and selectional restrictions. This light use of guo does not receive tone and cannot be separated from the main verb. And it does not allow a locative argument (2010: 62). As a consequence, Butt considers this as a distinct category, given its distinct semantics. A further consequence from this allows Butt to claim that the directional in Chinese forms part of the complex predicate - 'a construction that involves two or more predicational elements ... which predicate as a single unit, ie their arguments map onto a monoclausal 102

-syntactic structure' (Butt 2010: 49). All of the examples above are meant to be monoclausal, as Huang et al (2009) would claim. Although Butt suggests several factors that distinguish the experiential use (13) and the light use (14) of guo, it is, however, very subtle and I doubt that speakers of Chinese in fact realise the difference. For example, if guo in (14) appears as an answer to the question Ni chi guo mei? ('Have you eaten yet?), there will be a tone on guo. And if the question Ni qu guo nali? ('Where have you been?) is asked, a possible answer is Wo qu guo xuexiao ('I have been to the school'), whose guo takes a locative argument xuexiao 'school'. These uses of guo do not seem to me to constitute the evidence that there are four verbs of guo. Rather, this issue, in my opinion, is related to polysemy. One can keep this question in mind: Should there be one verb with several related senses or different verbs with different senses? Butt seems to suggest that there are several verbs, although she is careful to speak of'uses' and 'usages'. This polysemous phenomenon is quite pervasive in grammar and in Chinese grammar as well. To deal with this, I believe that a plausible analysis is to consider the phenomenon as demonstrating a scale of gradience, following Quirk et al ( 1985) and Aarts (2007), for instance. The following continuum can represent the various uses of the item guo: lexicalised grammaticalised The first guo1 can be treated as a lexical main verb, the second guo2 a directional, the third guo3 aspectual marker and the fourth.. guo4 a light verb, showing a dine of grammaticalisation. The fact that the directional guo2 is adjacent to the main verb guo1 and farther away from the light verb suggests that the directional shows more properties of a main verb than grammatical functions. This continuum illustrates whatAarts (2007) calls 'Subsective Gradience', taking the various senses as belonging to the same lexeme, and hence polysemous. If the motion verb merges with the directional verb, they form a construction, producing a special type of gradience - 'Subsective Constructional Gradience' (Aarts 2007: 171). The guo directional has not been fully grammaticalised and hence shows more lexical properties, which seems to be the case for Chinese directionals in general. Implicit here in Butt's argument is the claim that the verbs form a flat structure - a complex predicate inside the verb phrase, similar to the structure in (10) above. However, it may sound counter-intuitive to claim that jin wuzi is a constituent, as in 103

(9) . The convincing evidence, I think, lies in the adjunct-insertion phenomenon in Chinese directional. Let us illustrate this: (15) John zou bu jin wuzi. John walk not into house John could not walk into the house. (16) John pao bu guo qu. John run not cross go John could not run over there. The 'potential' form using bu 'not' separates the two verbs, for example, zou and jin, demonstrating that they do not seem to be a unit. Rather jin wuzi will form a constituent. Furthermore, the potential form cannot occur in between the second directional and the third directional verbs or between the second directional and its complement. (17) *John pao guo bu qu. John run cross not go John could not run over there. (18) *John zou-jin-le bu wuzi. John walk-into-asp not house John could not walk into the house. The potential form is not the only item that can separate the motion a~d directional verbs; the aspectual marker le can sometimes behave in the same way. It can separate the first and the second verbs but not the second and the third verbs. (19) John zou-le jin qu. John walk-asp into go John walked into it. (20) *John zou-jin le qu. John walk-into asp go John walked into it. All this suggests that the directional verb has a closer relationship with its complement and they constitute subordination, and thus the hierarchical structure in (9) is preferred. Thus, the item jin qu ' into go' forms a constituent, and then forms a larger constituent 104

-with zou 'walk'. In this manner, jin qu will constitute a complex event, with qu indicating the further direction, that is, away from the speaker. Thus, this study views the Chinese resultative construction as a bi-clausal unit but it differs from Hashimoto (1966) and Lu (1977) in that the construction is not derived from underlying structures via any transformations. Following a similar line of enquiry, we now offer a catenative analysis based upon Huddleston and Pullum's (2002) model on English. In fact, the controversies about this analysis in English started some time earlier among linguists such as Huddleston (1974, 1976) and Palmer (1973, 1990) . A catenative is a verb that takes another verb as its complement, and there is no limit on the number of verbs that a catenative can take. For example, in John wanted to avoid seeing Mary there are three verbs ( wanted, avoid and seeing) but there are two in John wanted to go (wanted and go). Palmer (1973) refers to this construction as a 'complex phrase', which involves subordination. The catenative analysis seems to have some relevance to the verbal construction of Chinese. The above examples can find their counterparts in Chinese too. (20) John xiang bi-kai bu jian Mary john want avoid not see mary John wanted to avoid seeing Mary. (21) John xiang zou john want go John wanted to go. The negative 'not' in the Chinese example malces the sentence more idiomatic, and the sentence still has three verbs xiang, bi-kai and jian. It seems that catenation is a common strategy that languages employ to link several events together. The question to ask here is whether the resultative construction in Chinese can be taken to be a catenative construction. If we can show that the verb that occurs after the catenative is not merely a verb but a clause, perhaps we can argue for the catenative analysis. At the same time, the related issue of finiteness arises as the catenative is usually defined as a verb that takes a non-finite clause (Huddleston and Pullum 2002, Palmer 1987). As Chinese is an isolated language, there are no tense markers on the verb and thus morphologically there is no tense in Chinese. The concept of finiteness is not very useful in Chinese. There are of course studies of the issue in Chinese such as Li ( 1985) and Tang (2000), both of which argue that a distinction between finite and non-finite 105

clauses can be made in Chinese. For instance, Tang (2000) uses the A-not-A question as one of the diagnostics for finiteness. If, he claims, a structure can appear with the A-not-A question, it is a finite; otherwise, it is non-finite. However, Tang (2000: 210-211) also admits that there are problems with such an analysis. Some structure may form the A-not-A question with different predicates in the same sentence. It is not quite clear how this should be analysed in terms of finiteness. (22) Ta pao de lmai bu lmai? he run deg fast not fast Did he run fast? (23) Ta pao bu pao de kuai? he run not run deg fast Did he run fast? If we only examine the morphosyntactic structure of Chinese verbs, it does not make much sense to distinguish between finite and non-finite clauses in Chinese. And I will make no effort to do so. Crosslinguistically, Cristofaro (2003) shows that relying solely on the morphosyntactic properties of languages will not be fruitful _in the examination of subordination. Whether finite or not, a Chinese sentence can consist of (subordinate) clauses. Given the lack of contrast of tenses in Chinese verbs, it would not be unreasonable to take them as 'non-finite' ie not showing tenses or agreement. The problem with this is that there is no finite counterpart. Thus, catenatives are simply verbs in Chinese and the term may be unnecessary, even for English (Aarts 2004). What it suggests is that the catenative complement is a clausal complement, rather than simply a verb or verb phrase. The different resultative constructions in Chinese might be analysed in the same fashion. (24) Tada-po-le huaping he hit-break-asp vase He broke the vase. (25) Tada dao huaping po le he hit arrive vase break asp He broke the vase. 106

The items po le huaping and dao huaping po le are the clausal complements of da. The plausibility of this analysis comes from the cognitive-functional definition of subordination. According to Cristofaro (2003: 2), subordination can be seen as 'a particular way to construe the cognitive relation between two events, such that one of them (which will be called the dependent event) lacks an autonomous profile, and is construed in the perspective of the other event ( which will be called the main event)' . This way of viewing subordination can allow us to capture the relation between the verb and its complement in the Chinese resultative construction. In the above examples, it makes sense to say that breaking depends on hitting; without hitting there will be no breaking of the vase. It is, however, not always clear what constitutes an event or subevent. Goldberg (1998: 42) defines events as such: Two events el and e2 are distinct subevents of an event E designated by a verb V, iffE ~ eland e2 is not completely within the temporal extent of e2. Goldberg gives the example of saute, which designates one event 'since the two aspects of heating and stirring overlap temporally such that the stirring is completely within the temporal duration of the heating' ( 1988: 42). Structurally, examples (24 and 25) show similarities too and the separation of da and po by dao huaping in (25) ie Tada [dao heaping] po le suggests further that da-po has a flexible structure; their combination is macro-managed by the resultative construction. Functionally, the event da 'is not completely within the temporal extent' of the event po. However, some resultative constructions in Chinese may not show the same temporal control. The temporal scheme 'tor English resultatives involves the action and then the change of state (Goldberg 1995). Although this works for the general cases in Chinese, Chinese resultatives are not restricted in this use, for example, da-si 'hit-die' as in John da-si le Peter. It is arguable that John can keep hitting Peter after he's dead. In order to interpret this, one would have to rely on the interpretation of the whole event and that would mean that we will assume other semantic and pragmatic factors. English resultatives differ from Chinese RVCs in its aspectual properties. According to Goldberg (1995), the time of the action denoted by the verb is the same as that of the result in the resultative construction in English. The following constraint of English resultatives is construed by Goldberg (1995: 194): The change of state must occur simultaneously with the endpoint of the action denoted by the verb. 107

Possible: Not possible: action action change of state change of state time time The temporal schema for English as demonstrated above is different from that for Chinese in that Chinese RVCs are not subject to this constraint, which requires that the ending point of the first verb occur simultaneously as the initial point of the state caused by the first verb. For example, in Harry shot Sam dead, it cannot be interpreted as Harry's shooting happened last night and Sam's death in the morning. Instead, this sentence must be understood as Sam died immediately when Harry fired the shot at him. However, in the Chinese counterpart, this restriction does not necessarily hold. Following Chang (1994: 22), the temporal scheme for Chinese RVCs may be captured as below: Vl: cause 111--1 EH- I E'l~ )) V2: result 121 -----Time Whether we take the Chinese resultative to include two subevents or two events does not prevent us from taking the 'catenative' analysis of resultatives that they contain two clauses. If the hypothesis is correct, it may counter Goldberg' s claim that resultatives are single clauses and single clauses represent basic human experience, since the resultative constructions seem to be basic to the experience of the Chinese. In this case, we may make use of Rothstein's 'Time-Participant Connected' hypothesis, 'which holds between two events and an individual, such that TPCONNECT(el, e2, y) holds for the following circumstances (Rothstein 2004: 70-71): TPCONNECT(el, e2, y) iff: (i) , (el)= , (e2) (ie the run time of el is the same as the run time of e2); (ii) el and e2 share a participant y.' There is a difference between depictives (cf Halliday 1967) and resultatives, and for Rothstein the idea behind this is that there is 'a relation between eventualities which reflects some intrinsic connection between events stronger than the relation between 108

elements in an equivalence class' (2004: 71). The conditions above work for the 1 depictives while the resultative has different conditions that 'the TPCONNECT relation holds between the culmination of the event introduced by the matrix verb and the event of the adjectival predicate', (2004: 75 original emphasis) as this: ... A TPCONNECT(Cul(el),(e2). Depictives are represented as: . .. A TPCONNECT(el, e2). Thus, the main difference between the two kinds of construction lies in 'culmination of the event' denoted by the main verb. In the Harry-shot-Sam-dead case above, the death of Sam can be seen as 'culmination of the event' that his death may have occurred any time during the spanning of the time ofVl. Typologically, Dixon (2010: 129) defines a complement clause by the following three main criteria: (I) It has the internal structure of a clause, at least as far as core arguments are concerned. Whether peripheral arguments may be included - and if so, which ones - varies from language to language. (II) A complement clause functions as a core argument of another clause. If a complement clause is in O function, for instance, it should show at least some of the syntactic properties of O in that language. For example, in English a THAT co~plement clause in O function may generally be passivized, as in [ That John A did [ the burglary] 0 ] c0 c1,5 is believed by everyone. (III) A complement clause will always describe a proposition; this can be a fact, an activity, or a state (it ruay be a place or a time). The criteria above may seem at the first sight incapable of describing the structure of Chinese; yet by applying the above criteria to the resultative construction, we will find that they contribute to our understanding of the construction. The following sentence serves as an example: (26) Chi-bao fan shi rensheng yi da lequ eat-full rice is life one big interest Life is enjoyable when one eats and becomes full. The subject of this sentence is chi-bao fan, having a predicate-object structure, which is realised by a resultative construction. This is a kind of nominalization, quite common in Chinese. The fact that there is not much morphology in Chinese does not distinguish 109

between the verbal form and the nominal form. Both of them occur in the same shape, which is quite similar to the gerund-participle or ing-form in English. (27) John was eating his breakfast this morning when Mary got up. (28) Eating breakfast is what John will never do. In the first case, eating is a no doubt a verb, albeit non-finite, but whether eating is the second example is more similar to a noun is controversial. In fact, it shows both verbal and nominal properties. For example, both cases of eating take an object breakfast, whereas eating (breakfast) occupies the subject position in (b) and eating (his breakfast ... ) in inside a verb phrase and eating is a main verb. The Chinese resultative has the internal structure of a clause with arguments, for instance, the Agent eater, albeit covert, and the Patient or Undergoer fan. And it can be argued that chi-baa fan functions as the subject of the main clause. This observes the first of the criteria. The subject, argued in Huddleston and Pullum (2002), serves as a complement or argument of the verb. For instance, in the sentence Whether John will move to the Philippines depends upon his wife's decision the subject begins with a complementiser or subordinator, depending upon one's view. And if the verb is replaced by another verb, the 'whether' clause may not work here, for example, *Whether John will move to the Philippines hurts Mary's feelings. It seems, thus, the Subject can be seen as a kind of complement. It makes sense to say that the Chinese example contains the Subject (ie complement) chi-baa fan, and it satisfies the second criterion that a complement clause serves a grammatical function. In terms of meaning, chi-baa fan produces a proposition that somebody ate rice with the arguments and the predicate, and it expresses an activity or a state. Hence, this satisfies the third criterion. Moreover, the Chinese resultative resembles the English counterpart in that the first verb can sometimes appear with an aspectual marker. The following example can illustrate this (Huang et al 2009:45): (29) Hendou shucai, many vegetable ta he sheng-zhe chi-guo raw-asp eat-asp He has eaten many vegetables raw. This example contains topicalisation of hendou shucai, which makes the sentence more acceptable than ?Ta sheng-zhe chi-guo hendou shucai. It would not be so far-fetched if we claim that there is a boundary (maybe covert) between the two verbs in the resultative. llO

In a similar vein, Wechsler (1997, 2001, 2005) analyse the English resultative as 'subordination ' although he does not explicitly claim that. This can be seen by his distinction two classes of resultatives, namely, the Control resultative and the Exceptional Case-Marking resultative (ECM), similar to the non-finite clausal structures John persuaded Mary to eat and John expected Mary to eat, respectively. In the former type, the subject of the resultative phrase is an argument of the matrix verb, while in the latter, the subject of the resultative phrase is not an argument of the matrix verb. The following examples can illustrate the basic difference here, with (a) and (b) control construction and (c) and (d) ECM construction. (30) a. b. c. John hammered the metal flat. The water froze solid. The dog barked itself hoarse. (transitive) ( unaccusative) ('fake' reflexive) d. Mary ran the soles off her shoes. This analysis seems to implicitly support our own here. The catenative analysis in fact reflects the iconicity principle that the subordinate verb comes close to the first verb as they show cause and effect. Thus, the resultative shows a complex event and complex structure since 'the concepts which are always, or frequently, expressed by complex structures are cognitively complex' (Croft 2002: 204). On the other hand, the 'catenative' analysis ofresultatives differs from the 'small-clause' analysis (eg Hoekstra 1988) in mainstream generative syntax that treats the secondary predication (ie the result) as a small clause in that no movement is assumed in the current model. If Wechsler is right about the structure of th~ English resultative, it will be equally important that the Chinese resultative be examined in a similar way. And also if he is accurate, it may pose, again, a challenge to Goldberg's (1995) claim that a single clause reflects basic human experience. This will be seen in this chapter that if a Chinese resultative is composed of two 'clauses', then this will provides cross-linguistic evidence that basic human experiences do not have to be reflected through simple single clauses. Here is the mismatch between syntax and semantics. SEMANTICS SYNTAX Resultative constructions English result~nese resultatives Thus, the syntactic representations may be manifested in different forms and patterns to provide a similar semantic dynamic or static scene. And these manifestations are often not uniform between languages, but they recognise a similar, if not universal, semantic basis of the structure and in this case resultative. 111

4.5 The Resultative as Cosubordination The catenative analysis argues that there is subordination in the resultative construction in Chinese. Following the argument, this section further suggests that the Chinese resultative construction sometimes shows the phenomenon of cosubordination that exists in some other languages such as Tamil (Croft 2001: 323), for example: ( 31) [ avaru kavide erudiitt u] [naaval moripeyarttaaru] He.NOM poetry.ACC write.PRF.GER novel.ACC translate. PST.3SG He wrote poetry and then translated a novel. Many languages show this phenomenon and when translated into English, sometimes coordination is used; other times, subordination is used. We begin by taking the functional definition of states of affairs and then see how the resultative construction expresses them through cosubordination. The typological work suggests that defining states of affairs in semantic functional terms is more preferable to the definition from a formal point of view (Cristofaro 2003). Cristofaro (2003: 25) takes the definition of states of affairs from Siewierska (1991) and Dil< (1997) alil

Cristofaro argues that although Langacker's approach was not meant to be used as a typological tool, it can actually contribute to the study of subordination in diverse languages. This is the asymmetrical assumption, which puts unequal weightings to the main clause and the subordinate clause. This asymmetrical assumption is further supplemented by another dimension - figure/ground and complex-figure, along with the dimension of foregrounded and backgrounded information. This is particular fitting within the typological framework as the features are not only specific to English but are also general enough to be used for other languages. Figure-ground relations consist of asymmetrical conceptual distinctions. The state of affairs is the ground if it provides the 'cause, preconditions or reference point' for another state of affairs, which is considered to be the figure (Cristofaro 2003: 27). This can be shown by an English sentence such as The phone rang when John was sleeping. The temporal adverbial clause provides the ground for the event that the phone rang, as the phone ringing depends upon John's sleeping. In this regard, the resultative construction can be considered to be a figure-ground manifestation, as the first verb typically provides the cause for the result indicated by the adjective phrase or another verb as in Chinese. For example, John hammered the metal flat or John chi-baa le fan 'John ate rice and as a result he was full' show the causes (ground) and results (figure). The grounds are John's hammering the metal and John's eating rice, and the figures are the metal being flat and John being full. This phenomenon can be referred to as 'semantic integration' (Givon 1990). If we are on the right track, we should be able to use this approach to the resultative as a typological study. And if this is convincing, we may further argue that the resultative construction is some kind of subordination or cosubordination. However, there are problems with these arguments. Cristofaro (2003) shows that morpho-syntactic criteria are not universal but the asymmetrical assumption is typologically valid. To test the dependent and independent state of affairs suggested by the assumption, one can use, for instance, negation and sentence questioning. Thus, for example, the negative counterpart of John thought that Mary was in Hong Kong is John didn't think that Mary was in Hong Kong. This means that the main clause is negated but not the subordinate clause, which is supposed to be insensitive to the negation. The dependent state of affairs is not subject to such operations as negation, while the independent state of affairs is. The Chinese counterpart of the previous example is the same as that in English John bu renwei Mary zai Xiangkang. What would happen to the resultative constructions 113

in the two languages if they are considered to be a type of subordination? In John didn't hammered the metal flat and John meiyou chi-baa fan 'John didn't eat rice to the extent that he was full', they seem to be equal and show that the dependent states of affairs are not subject to sentence negation. However, if one asks the question Ni chi-baa le meiyou?, the negative answers will be meiyou (chi-baa), not meiyou ( chi). It seems that the negation scope falls on the resultative verb. The English counterpart suggests the same: Did John hammer the metal flat? No, he didn't (hammer the metal flat), but not he didn't (hammer). This shows in both English and Chinese resultatives the dependent states of affairs fall into the scope of negation, which is not predictable, if the resultative is seen as subordination. The negation diagnostic works well for the canonical subordinate clause that designates a dependent state of affairs. Rather than suggesting that the resultative construction is not subordination, we believe that it is the diagnostic that is problematic. In English the scope of negation is not always on the main clause or independent state of affairs, for instance, the sentence John did not come back because he wanted to see Mary may be ambiguous precisely because of the scope of negation. It can mean 'John did not come back and he wanted to see Mary' or 'John came back but it was not because he wanted to see Mary'. Another example at hand will be All the students did not come, where not can modify all (meaning 'not all') or come. This test may suggest that the negated verb or element should be head of the construction, which is also used by Song (2008) who argues that the head of the Chinese resultative construction is the first verb. We can take the first verb of the Chinese resultative to be a syntactic head and the second verb of the construction a_s a semantic head (Payne 2011), when usually a syntactic head is also a semantic head. A mismatch occurs here. As argued in Song (2007: 37), the behavior with respect to the negation in the Chinese resultative differs from that with an adjunct. There are two ways that can negate the Chinese resultative. One is bu as in the 'potential' form and the other is meiyou or just mei before the first verb. For example, in John mei da-sui boli 'John didn't break the glass', John did hit the glass but it did not break into pieces ('sui'). However, in the structure John mei qiaoqiao-de jin-qu 'John did not enter quietly', the adjunct qiaoqiao-de 'quietly', according to Song, is modified by mei 'not' and the sentence indicates that John entered, but not quietly. The item jin-qu is a resultative construction, in the broad sense of it, and neither jin 'enter' or qu 'go' falls into the scope of negation here. 114

If this is tenable, can we apply the test to the resultative construction and obtain the same results? I believe that we can. It makes sense when the sentence occurs in isolation. Suppose a reply to John mei da-sui boli 'John didn't break the glass' is John meiyou da boli; shi boli ziji po le 'John did not break the glass; it just broke', where the speaker attempts to correct the utterance by negating the first verb rather than the second verb complement. Song further argues that the resultative behaves in the same manner as the verb-object construction. In John mei chi pingguo 'John didn't eat an apple', the negation scope falls onto pingguo 'apple' but not chi 'eat'. It means that John ate (something) but it was not an apple. Nevertheless, if this appears in a piece of discourse, the meaning may change. A reply can make it clear what the utterance means: Bushi, John shenme dou mei chi 'No, John did not eat anything'. This shows that the negative marker may not always modify the second element of the construction. The complexity of the structure can reflect the fact that the Chinese resultative construction consists of a complex predicate, 'which are multi-headed' (Alsina et al 1997: 1). This results from a semantic assumption that a complex predicate can show that ' [ t] he argument structure is complex (two or more semantic heads contribute arguments)' (Butt 1995: 2). If the Chinese resultative construction is multi-headed rather single-headed, whether being the first verb or the second, this construction shares some features of coordinate structure (coordination). As Van Valin and Foley (1984) and Croft (2001: 322-328) note, there is a continuum of complex sentence types: Coordination Cosubordination Adverbial Clauses Coordination and adverbial clauses are the two poles and extremes, in the middle of which lies Cosubordination, whose clauses are deranked and balanced, respectively. Cosubordination hence contains the syntactic structure of adverbial clauses and coordinate clauses. In many European languages cosubordination can be translated into other languages in coordinate clauses or subordinate clauses. For instance, a sentence that consists of a serial verb is ambiguous between the coordinate and the subordinate interpretations: 115

(32) Wo jide guan men I remember close door I remembered to close the door. Or, I remembered; I closed the door. An example of the resultative construction such as Wo chi-bao fan would also be arguably ambiguous between the two interpretations. The coordinate structure interpretation for this example is 'I ate and I was full', while the subordinate structure interpretation is 'I ate so that I was full' . If this makes sense, it will be possible to claim that the Chinese resultative construction exemplifies the general strategy of cosubordination. To further illustrate the phenomenon, one can rely on the assertion test from Cristofaro (2003). The serial verb example Wo jide guan men can be questioned by Ni jide guan men ma? And the replies can be Wo jide 'I remembered' or Wo guan le 'I closed ( the door)'. We can apply this strategy to the Chinese resultative and question Wo chi-bao le by Ni chi-bao le ma?, which can be replied by Wo bao le 'I am full' or Wo chi-guo le 'I have eaten'. Further, we can reply, with respect to Ni chi-bao fan le ma?, Wo chi le, dan mei bao 'I have eaten, but I am not full' or Wo bao le, dan bushi chi fan ' I have eaten, but not rice'. Since the second verbs of bao and guo in the replies carry aspectual meanings such as perfective, I believe that they are both acceptable, given the appropriate context. Thus, in Wo jide guan men, either verb (clause) can be asserted. The same, I believe, applies to the resultative. If one argues that only the second verb of the resultative is asserted, that ignores the pragmatic context that allows for the interpretation with the assertion of the first verb. The negation test that we previously used can also be applied to the serial verb example and the resultative one. In Wo bu jide guan men, the negative marker bu can modify either jide or guan; if the aspectual marker le is added, the sentence is not ambiguous. The natural interpretation of this sentence Wo bu jide guan le men for me will be 'I don't remember the fact that I closed the door'. We have already seen the examples with negation above such as John mei da-sui boli 'John didn't break the glass' with the reply John meiyou da boli; shi boli ziji po le 'John did not break the glass; it just broke'. According to the continuum of complex sentence types above, the Chinese resultative construction seems to combine the two characteristics of cosubordination and yet it is not a typical example where on the one hand, the construction behaves as a whole, much lilce a lexical item - a single word; on the other, it shows syntactic 116

structure such as the bu insertion, which suggests that the two verbs do not seem to be a constituent at a lower level. This idea seems to be shared by Shi (2008: 83), whose Boundary Principle states as follows: As constrained by the semantics of resultative construction, there seems to be a covert syntactic boundary between the two verbs in the resultative construction. This limits, in the process of integration, the projection of the underlying arguments onto the surface level. However, Shi does not recognise a syntactic position between the two verbs, and in fact, except from few items such as bu ('not'), it is not possible to insert any element between them - their tie is so strong that they are difficult to be separated in present-day Chinese. Similarly, Van Valin and LaPolla (1997: 456-457) take the two verbs to form a nucleus at one level and a core at a higher level. They treat the Chinese resultative as having nucleus junctures, as, for example, the aspectual perfective marker le attaches to the entire construction and has scope over both verbs: (33) Ta qiao-po-le yi ge fanwan he hit-break-asp one cl bowl He broke a ricebow. *Ta qiao-le-po yi ge fanwan The fact that the two verbs share a single aspectual operator suggests that the nexus type is cosubordination. A simplified diagram can illustrate this (Van Valin and LaPolla 1997: 456): CLAUSE I CORE NLc

Recognising that the Chinese resultative is a subordinate structure has further typological consequences and advantages. For example, we have already seen the structure as representing two states of affairs in terms of event structure. The phenomenon of deranking can also be expressed in the resultative construction, which has been found in many different languages. Cristofaro (2003) examines over eighty languages about such a phenomenon and discovered an implication hierarchy about the relationship between complement clauses and adverbial clauses: The Hierarchy of Subordinate Clause Deranking modal, phasal < purpose < manipulative, desiderative < perception < before < after, when < reason, reality conditional < knowledge, propositional attitude < utterance This scale measures the degree of semantic integration, which shows connections between two states of affairs. Two states of affairs are seen as one single tight event by modals and phasals, while manipulatives and desideratives can denote events that are dependent upon another (Croft 2002: 218). The ranking above should also include the category of 'resultatives' - perhaps before manipulative - many of which express purposes that are parallel to emphasising the outcome of an event (Boas 2003). Thus, resulatives are high on the scale and Chinese resultatives appear to be more semantically integrated than the English ones. Otherwise, adopting the hierarchy, we may suggest that resultatives are similar to purposes, between modal/phasal and manipulative. While most linguists of Chinese consider the Chinese resultative as a compound and the English resultative as a phrasal structure, we maintain that they are both structural constructions. Apart from the reasons given above, the following arguments attempt to make the same point. 4.6 The Resultative as a (Mini-)Construction The resultative has been analysed as a pairing of form and meaning (eg Croft 2001, Fillmore and Kay 1988, Goldberg 1996). Decomposing the resultative construction as mini-constructions is postulated by Boas (2003), where he analyses English and German. As has been noted, the Chinese resultative analysed as a verb compound has been supported by most linguists (Cheng and Huang 1994; Huang et al 2009; Packard 2000; Li 1990, 1993, 1995; Lu 1977, Thompson 1973 etc). In particular, Li (1990, 1993) analyse the resultative as two verbal morphemes merging into a compound and the 118

argument structure is shown through theta-identification (ie semantic role-indentification) (Higginbothem 1985). The head features ofVl will be percolated to the entire compound. Take qiao-sui 'hit-break' as an example: (34) qiao 'hit' > sui 'break into pieces' qiao-sui Theme> Along the line, Huang et al (2009: 43-46) argue that the resultative is a compound and make use of the phrase structure ofX-Bar Theory to illustrate this: Their argument is that the resultative is a compound because the aspectual marker eg guo attaches to V2, not Vl such as (35) Ta qiao-sui-guo boli he hit-break-asp glass He has broken a glass. The compound qiao-sui will merge into a compound word under the X node in the phrase marker above. In their example Ta sheng-chi-guo henduo shucai 'He has eaten many vegetables raw', they claim that chi 'eat''is the head of the compound, and sheng is attached to a higher node WP, for the reason that sheng can also be attached with an aspectual marker zhe, as in Ta sheng-zhe chi-guo henduo shucai with the same reading. However in the phrase marker XP Y~' ~, shenizhe ~p I eat-guo 119

there is nothing that claims that sheng-zhe is part of the compound word, as an adjunct can occupy the same position, as in Ta gaaxing de chi-zhe shucai 'He is eating vegetables happily'. XP Y~' W~' gaaxiAg-de X~P I eat-zhe Furthermore, the argument structure of sheng-chi 'raw-eat' is <01> and <0a <06>>, as in Cai sheng'vegetables are raw' and Ta chi cai 'He eats vegetables'. The view is in fact consistent with a constructional one in that there is one construction inside another construction. If the resultative construction can be decomposed into the two clauses as follows: (36) Ta chi-le fan he eat-asp rice 'He eats rice' . (37) Ta bao-le he full-asp He was full. we should be able to see the internal event structure more clearly. The two clauses express two states of affairs or propositions independently: E(x, y) and F(x), and can be represented as two (mini-)constructions (Boas 2003). Perhaps we could replace the above phrase marker with the following: chi-baa resultative construction N~struction )a V~struction I ~lchi bag fan le Instead of combining chi and baa as a single compound, as indicated by the box, the item baa 'projects' into a mini-construction and chi another one before they fuse into the resultative construction. (The two constructions can be seen as instances of the 120

intransitive construction.) Following the catenative analysis above, the construction may be considered to be a clausal construction. The fact that chi and bao merge into one single whole echoes the observation by Huang and Cheng (1994: 191) that 'all resultatives are causative by nature' - the causative-resultative pattern. This will lead us to the discussion of event-frame semantics model, which in fact is based upon the meaning of a construction. Ifwe adopt the event-frame semantics model, we can see the event denoted by the each verb producing its own frame and scene. The event-frame semantics developed by linguists such as Fillmore (1982) and Fillmore and Atkins (1992) provides us with a convenient way to view the meanings of individual verbs and their connection with the world. For instance, the transaction example consists of buying and selling, and the verbs buy and sell supplies the frame that involves the buyer and seller, money and charge and goods. There is more such as 'property ownership' and 'money economy' ( 1992: 78) etc. Thus, knowing a verb implies knowing its connection with the world and that is in our encyclopedic knowledge. Every verb has its own internal semantic structure and contributes to the mini-construction in which it can occur. In the resultative construction the mini-construction denoted by the first verb associated with activities become relevant 'once the speaker decides to point out a special perspective of the event's outcome' (Boas 2003: 170). In other words, when the verb denotes the activity alone, it does not contribute to the meaning of the resultative construction but it does when it enters the construction with the secondary predicate as a whole. As in John painted the wall, while painted denotes the activity but there is no emphasis on the outcome although one can imi!gine that the wall has changed. However, in John painted the wall red, the construction produces a result reading, which emphasises the outcome that the wall has now become red. In the Chinese resultative example above, chi denotes the activity of eating with the eater and the food, and bao denotes the change of state, without which we do not know the state of the speaker, which was full. The outcome here is emphasised by the baa-construction. This is consistent with Boas's (2003) observations of the communicative functions of the resultative: (1) emphasizing the endpoint of an event, (2) being less vague about the endpoint of an event, (3) perspectivising a participant different from the prototypical patient participant, ( 4) perspectivsing a participant outside of the conventionalised event-scene. All these functions suggest that the secondary predicate in the Chinese resultative plays a very important role. And it is a semantic centre ( eg Chao 1968, Huang 121

et al 2009). The fourth communicative function can in fact help us understand those sentences with verbs that do not subcategorise for the patient argument, for example: (38) John sneezed the napkin off the table. Ta ku-shi-le zhentou. he cry-wet-asp pillow He cried and made the pillow become wet. Traditionally, the verbs sneeze, ku and shi are all intransitive verbs, and yet in the above sentence there are arguments after them. The verbs are used unconventionally (Boas 2003). Rather than producing two distinct clauses such as John sneezed, and the napkin flew off the table, we can package them into one construction to 'highlight the immediate cause and effect relation between the two events' (Boas 2003: 156). In the Chinese example, instead of having Ta ku, ranhou zhentou shi le 'He cried and then the pillow got wet', ku and shi merge into a resultative construction which behaves like a lexical item. The merge still sticks to the universal principle of semantic roles (Dowty 1991) that the Actor (Agent) or Cause occupies the subject position, and the Patient is aligned with the postverbal object position. This analysis can be extended to other cases such as Ta chi-bao-le fan 'He ate and became full', with Ta as the Agent s_ubject and fan the Patient object. This phenomenon also echoes Haiman's (1985) Principle of Maximised Expressive Power that the inventory of constructions is maximised for communicative purposes. This is in consistency with the Constructional models that different semantics may drive different structures: Chinese: Resultative verb ku-shi, qi-lei cry-wet, ride-tired English: John shouted himself hoarse Causative-resultative shi zhentou shi le cause pillow wet asp Other result constructions ba zhentou xi le ba pillow wash asp John caused the pillow to become wet In Chinese the resultative ( compound), the 'shi' causative-resultative construction and the ba-construction can all express the idea of cause and result/effect; and on this continuum English exhibits a similar kind of phenomenon having the resultative 122

construction at one end and the transitive construction - using cause- at the other end. If there are more constructions in Chinese that can show similar meanings, it may, from Haiman's (1985) point of view that Chinese is more maximally expressive in terms of communicative functions. Compared with the resultative in English, the Chinese resultative construction only permits two arguments - Agent/Actor/Causer and Patient/Theme/Undergoer. There is no oblique. In the case of chi-baa, chi assigns the Actor and Patient roles to the subject and object and baa assigns the Actor (for some linguists, Theme) to the subject. The verb chi is assigned the Agent and Patient roles, and the verb baa is assigned the Agent role, which has to be in one of the two slots. It goes to the subject position. The chi-construction is a transitive construction: Sem CAUSE-BECOME < agt pat > R R : instance, PRED < eater food > means l l l Syn V SUBJ OBJ l 123

The baa-construction is an intransitive construction: Sem BECOME < agt > R R : instance, PRED < > means l l Syn V SUBJ l Sem CAUSE-BECOME < agt pat > R R : instance, PRED < > means l l l Syn V SUBJ OBJ This analysis can be extended to other cases such as Ta ku-shi-le zhentou 'He cried and made the pillow wet'. Imagine that we have two intransitive constructions: 124

The ku-construction is an intransitive construction: Sem BECOME < agt > R R : instance, PRED < > means l l Syn V SUBJ l The shi-construction is an intransitive construction: Sem BECOME < agt > R R : instance, PRED < > means l l Syn V SUBJ l 125

Sem CAUSE-BECOME < agt pat > R R : instance, PRED < > means l l l Syn V SUBJ OBJ The resultative construction has the intrinsic semantics: CAUSE-BECOME, which coerces the two intransitive constructions ku and shi to become causative. Croft ( 1991) states that 'the meaning designated by the verb and the meaning designated by the construction must be integrated via a (temporally contiguous) causal relationship' (Causal Relation Hypothesis). Since the Chinese resultative construction allows two argument roles, there is no oblique-goal argument. The goal function in English finds its counterpart in Chinese as another construction - the 'dao-construction'. For example, Ta ku dao zhentou dou shi le 'He cried so that the pillow got wet'. This construction can be considered to be an extended resultative construction. This dao-construction provides a further extent of the result, that is, the extent of wetness is sufficient to make the pillow wet. Embedded within this construction is another construction projected by shi 'wet'. Ta ku dao zhentou dou shi le (Clausal construction) N~truction )a V~onstruction I ----------_ lku dmizhentoudoushf Construction Grammar or other related models have been developed to account for the syntax, semantics and pragmatics of language and it deals with the integration of them. Using it as a model for resultatives can allow for the 'integration of event-frame information in terms of (mini-) constructions' (Boas 2003: 318). This integration in fact shows strong properties of transitivity, and so strong that even single intransitive verbs entering into this construction can have arguments caused by the construction such as To ku-shi-le zhetou ('He cried and made the pillow wet'), in which both 126

-predicates are intransitive. Similar to the resultative construction m English, the Chinese resultative is also highly transitive, whether in syntax or semantics. This approach can also explain why there are certain intransitive resultatives, ie not taking an object argument in the Chinese resultative structure, which exists as a case of 'grammatical blending' (Fauconnier and Turner 1996, Fauconnier 1997). There are two inputs in grammatical blends or blending - the predicates' (sub )event structures and the contribution of the entire blend - which is consonant with the spirits of the constructional models. The observation made by Fauconnier (1997: 173) that 'grammatical constructions are blends, which are entrenched but evolve diachronically' is particular fitting in the discussion of the Chinese equivalents. The historical link of the predicates in the resultative construction suggests that they were once separate and some still are. However, the construction is itself contributes to a further meaning -cause and result. Over time, the structure changed (Shi 2002, Liang 2006), which is followed by the 'linguistic pressure to represent complex integrations of events by making maximum use of existing grammatical constructions' (Fauconnier 1997: 173) . Thus, grammatical blending has a higher pressure that even intransitive predicates entering into the resultative construction become more transitive. 4.6 The Motion-directional Construction as a Resulative Construction As has been mentioned, a special kind of construction that many linguists ( eg Chao 1968, Li and Thompson 1981) will recognise as the resultative, where there are so-called displacement verbs such as lai 'come' and qu 'go' , is the motion-directional construction as we call it here. Our proposed _fatenative anlysis implies that the head of the resultative and motion-directional constructions precedes its complement, and thus leans support to Talmy's arguments that Chinese is satellite-framed. In this regard, we have to discuss Talmy's ideas about the phrase structure of Chiense briefly. Talmy (eg 2000, 2008 rev) has proposed the typology oflanguages and argue that there are two broad types - verb-framed and satellite-framed language, in which Chinese is classified as the latter type. Languages that characteristically map the core schema into the verb will be said to have a framing verb and to be verb-framed languages . .. 'languages that characteristically map the core schema onto the satellite will be said to have a framing satellite and to be satellite-framed languages (Talmy 2000: 222). 127

However, Slobin (2004) and Chen and Guo (2009) and subsequent work argue that there is a third type of language - "equipollently-framed", and that Chinese is equipollently-framed. Their arguments hinge on the main verb status, which we will examine from a (Cognitive) Construction Grammar perspective (Goldberg 1995, 2006; Langacker 2008, 2013), and argue for Talmy's position that Chinese is satellite-framed, with the insights from Cognitive Construction Grammar and Cognitive Grammar. According to Langacker (2008: 167), verbal complexes can realize constructional schemas, which 'provide the basis for semantic and grammatical composition'. One such schema is the motion-directional construction. The following examples can illustrate this: (39) La botella sali6 flotando de la cueva. Talmy (2000: 224, Example 4b) "The bottle exited floating from the cave." (40) Pingzi piao-chu-le 16 dongxue. bottle float-exit-asp cave "The bottle floated exiting from the cave." The two examples are similar in their internal structures. The verbs for 'float' and 'exit' occur in a different word-order in Spanish and Chinese but the two sentences mean approximately the same thing. The Spanish example contains the non-finite form flotanda but it is not relevant in the case of Chinese, as it does not distinguish between a finite and a non-finite verb form. Whatever form chu is, it's a verb. One might argue that chu could be a non-finite form in this example as it shares the subject of the matrix clause (Pingzi 'bottle') (Fong 2016). However, whether the verb chu is a finite or non-finite form of the verb in Mandarin is not relevant here. The following diagram represents the constructional schema for the motion-directional construction and it can instantiate this use of three verbs together. ( 41) Constructional schema ' Directionals " ~Path Deixis 16 I still hyphenate the verbs here just to be consistent with the traditional practice. In fact, to support our catenative anlaysis properly, we should separate the verbs such as piao chu 'float out'. 128

( 42) Pingzi piao-chu-qu-le. bottle float-exit-go-asp "The bottled floated out." Although V2 indicates the path along which the motion moves and V 3means 'come' or 'go', I group V2 and V3 as one class and refer to them collectively as directional verbs for two reasons. First, their meanings are similar i.e. indicating orientations, but both differ from the meaning of the motion verb. Second, either of them can independently merge with the motion verb to create the motion-directional construction, which has undergone grammaticalization - the process referred to as 'constructionalization' (Traugott 2014, Traugott and Trousdale 2013: 22). Constructionalization is the creation of form""' - meaning"'" (combinations of) signs . ..... New micro-constructions may likewise be created gradually, but they may also be instantaneous. Thus, the motion-directional construction has been a new form with meaning of its own, and the verbs constitute the micro-constructions. Our catenative analysis already implies that the head of the construction is the main verb and in Chinese the first verb (see Fong 2016 for more detailed arguments). Thus, our analysis support Talrny's claim that Chinese is satellite-framed. Let us look at some evidence. Although the three verbs can occur freely as single predicates of clauses, they conform to strict word order in syntax. They show strong selectional restriction or subcategorization. (Perhaps we can dubbed the motion-directional construction with two directional verbs as "double-satellite" construction.) ( 43) John pao-jin-lai-le. John run-enter-come-asp "John came in by running." ( 44) *John jin-pao-lai-le. ( 45) * John lai-jin-pao-le. (46) *John pao-lai-jin-le. 129

While (43) is impeccable, (44-46) are ungrammatical. Ifwe take a lexical view, we will suggest that the motion verb pao 'run' selects or licenses the directional verbs and the first directional verb licenses the second directional verb. From a constructional perspective, the motion-directional construction provide the schematic template for the verbs to fill the slots. Since there is a continuum between lexicon and syntax, both lexical and constructional evidence should be considered. If we examine the interrogative sentence in Chinese, we will find a similar phenomenon. Pragmatically, the interrogative sentence in Chinese also questions the directional verb but not the motion verb. Observe the following piece of discourse: (47) A: John zou-pn WUZI le ma? John wall<-enter house asp par "Did John entered the house by wall

between a head and its complement in that in generative syntax the complement is said to depend on the head, which selects its complement.) In the Mandarin example the verb jin specifies (i.e. elaborate) the path along which John walked, which is a salient substructure within zou as opposed to the opposite with the verb chu ('exit'). (48) zou-jin-construction John ZOU Jin wuzi Jin-construction zou-construction Motion-Directional Construction 4.7 Summary This Chapter examines several .analyses of the resultative construction, which can be taken as a clausal construction, cosubordination and mini-constructions etc. In particular, the arguments behind all these are essentially that a verb alone cannot produce the syntax and semantics of the resultative in both language-general and language-particular levels. Examining both Chinese and English, we find that there exists an independent and highly transitive resultative construction with syntax, semantics and pragmatics, and this can also be considered a grammatical blend. In particular, showing certain clausal or constructional characteristics, the Chinese resultative is viewed as a construction rather than merely as a compound, which many linguists hold. The special kind of the resultative construction involving the displacement verbs (motion-directional construction) can also be analysed in the same framework here. The transitivity hypothesis postulated by Hopper and Thompson (I 980, 1982) will be discussed in the next Chapter. This transitive nature is further argued to exist in a special kind of causative-resultative construction - the ba-Construction. 131

Chapter 5 The ha-Construction and the Resultative 5.0 Introducing the ha-Construction and ha as a Verb Among the widely discussed grammatical constructions in Mandarin Chinese is the so-called the ha-construction. This construction is named as such because the grammatical item ba occurs in the sentence, as shown in (1). (1) John ba john ba Mary da-si mary hit-die John hit Mary to death. Since the ha-construction has been widely studied by linguists, it is given various names. The ha-construction is called the 'disposal' construction by Wang (1954), Chao (1968) etc. Some regard it as the ergative construction (eg Frei 1956; Li and Yip 1979). Hashimoto (1971) called it the 'executive' construction, while Teng (1977) analyses it as the accusative construction. Other linguists simply consider it to be the ba-construction, for example, Cheung (1973), Li and Thompson (1981), Mangione (1982), Wang (1985) and Sybesma (1990). Furthermore, the status of ba has not been fully settled. Huang et al (2009: 165) conveniently summarise the previous analyses of the categorical status of baas, for example, a preposition (eg Chao 1968; Li 1985, 1990) or a Case assigner (eg Huang 1982; Goodall 1987). Seldom are there analyses that consider baas a verb, exceptions include Hashimoto (1971) and Ding (2007). The present study, following Fong ( 2015), provides further evidence and argues that ba is a legitimate verb or complement-taking predicate in Chinese, although it lacks some of the common verbal properties shown in many other verbs. There are some standard tests for verbhood in Chinese such as the occurrence with an aspectual marker like le or guo to show the perfective meaning. 132

(2) Wo chi-le. I eat-asp I have eaten. (3) *Wo ba-le fan chi le. I ba-asp rice eat asp I have eaten rice. It is true that ba cannot take an aspectual marker such as le, while many other verbs in general can. However, ba is certainly not the only verb that behaves as such. While the verb ai 'love' can occur with guo 'cross', it cannot take le. (4) John ai-guo Mary. John love-asp Mary John has loved Mary. (5) *John ai-le Mary. John love-asp Mary John has loved Mary. Since Chinese verbs do not inflect for tenses, we cannot depend on this to distinguish between verbs and adjectives, for instance. (6) Wo zuotian/mingtian chi huoguo. I yesterday/tomorrow eat hot-pot I had hotpot yesterday. I will have hotpot tomorrow. Yet, many linguists (eg Huang et al 2009, Li and Thompson 1981, Liu et al 1996) will take example (3) as evidence that ba is not a verb because it cannot occur with the aspectual marker le. However, I believe that is not the whole picture, and the model of gradience (Quirk et al 1985, Aarts 2007) seems to do ba more justice. According to Quirk et al (1985: 90), A gradient is a scale which relates two categories of description (for example two word classes) in terms of degrees of similarity and contrast. At the ends of the scale are items which belong clearly to one category or 133

to another; intermediate positions on the scale are taken by 'in-between' cases - items which fail, in different degrees, to satisfy the criteria for one or the other category. They give the example of coordinating and subordinating conjunctions: Coordinating Subordinating and---· for --- if With this principle in mind, we can extend it to other categories. For example, verbs are said to exhibit a continuum of verb hood, in which some verbs are more verb-like than others, ie possessing more verbal properties. A more recent study on gradience (Aarts 2007) provides us with a more detailed and scientific method for word categorisation. We may be able to solve the problem of categorisation of the Chinese baby applying what Aarts calls the 'Subsective Gradience' model, which 'is the phenomenon whereby a particular set of elements displays a categorical shading in propotypicality from a central core to a more peripheral boundary' (2007: 97) . Following Aarts, I will argue that the Chinese ba shows Subsective Gradience in that it belongs to the class of verb, but does not lie in the central core of the class. As has been shown, ba does not license the suffixes le and guo immediately, and thus for many Chinese scholars, it is not a verb. However, apart from the verb suffixes, there are other diagnostics available, such as the ability to take a complement, the possibility of being modified by adjuncts and by 'not'. If we take a prototypical verb chi 'eat' and compare it with ba, we will obtain the following findings. (7) chi ba (i) Verbal suffixes yes no (ii) Taking complement yes yes (iii) Being a verb complement yes yes (iv) Standing alone yes no (v) Being modified by adjuncts yes yes (vi) Being modified by meiyou 'not' yes yes (vii) Being in a predicator position yes yes (following the first argument) 134

-(viii) Occurring in A-not-A questions yes yes (8) (i) a. Wo chi-le fan. b. *wo ba-le fan . (ii) a. Wo chi-le fan. b. Wo ba chi-fan de shi wangle. I ba eat-rice de matter forget asp I forgot about the eating-rice business. (iii) a. W o xiang chi fan. I want eat rice I want to eat rice. b. W o xiang ba fan chi-le. (iv) a. Chi! b. *Ba! (v) a. Wo kuaikuaide chi-wan fan. I quickly eat-finish rice I fmished eating rice quickly. b. Wo kuaikuaide ba fan chi-wan-le. (vi) a. Wo meiyou chi-wan fan. b. Wo meiyou ba fan chi-wan. (vii) a. Wo chi-le fan. b. Wo ba fan chi-wan-le. ( viii) a. Ni chi bu chi fan? You eat not eat rice Do you ( want to) eat rice? b. Ni ba bu ba ta dang pengyou? ni ba not ba him regard friend Do you consider him as a friend? 135

In the above eight diagnostics or behaviour constructions (cfCroft 2012: 20), chi has eight while ha possesses six - more than half, which makes me believe that ha is a verb lacking some of the verbal properties. The last piece of evidence (7 and 8 viii) suggests that there is a close connection between ha and auxiliary verbs. Perhaps ha is moving towards to the auxiliary group, since it has been grammaticalised. All in all, the conclusion we should draw here is that ha is a verb. If we establish a continuum here, we will put them in different places: chi ha shu'book' V Non-verb Historically, there is also another item similar to ha, jiang, which occurred in the position of ha, and it means the same as ha, and so can be considered as a variant of ha. (9) John jiang Mary da-si le john ha mary hit-die par John hit Mary to death. However, jiang is not widely used in modern Mandarin Chinese, though in some 'dialects' of Chinese it is generally used like the Mandarin ha. In Cantonese, for instance, ha is never used but jeung (10) John jeung Mary da-sei jeoh john ha mary hit-die asp John hit Mary to death. Generally speaking, the ha-construction often has its counterpart without the item ha, having more or less the same meaning. (11) John da-si Mary john hit-die mary John hit Mary to death. Note that Cantonese also exhibits the same kind of ha-alternation as Mandarin. Much said about Mandarin could also be said to Cantonese or other 'dialects' of Chinese, albeit with some subtle difference. 136

(12) John da-sei-jeoh Mary john hit-die-asp mary John hit Mary to death. Next, we will discuss the general characteristics of the ha-construction and the constraints observed by linguists. The ha-construction, though extensively studied by linguists, is still as interesting as ever, partly because it apparently differs from the canonical SVO word order of Chinese, and partly because the ha-construction presents challenge to different theoretical frameworks. 5.1 The Functions of the ha-Construction First of all, let us consider the sentences (1 and 11). They are interpreted as almost semantically equivalent. And if two sentences share the same meaning with or without ba, why does there exist the ha-construction? However, the sentences ( 1 and 11) exhibit some nuance between them, which would be considered by some linguists to be characteristic of the ha-construction. In these cases, the direct object in the ba sentence is said to be different semantically from that in the non-ba sentence in that it is more focused, empathetic or definite (eg Teng 1.971 , Thompson 1977, Li and Thompson 1981). As we shall argue, there is motivation for the existence of the ha-construction in Mandarin. Many, if not all, linguists believe different linguistic forms produce different meanings. Wang (1954), for example, maintained that there must be difference in meaning between the ha-construction and its counterpart without ba. However, there are cases which demand the presence of ba in a sentence, as shown by an example of Song (1991: 2) below. (13) Dajia shuyao ba wenti bai-dao All need ba problem put-arrive [ (lit) We need to put the problem on the table] We need to discuss the problem frankly. zuo-min-shang-lai table-face-up-come There is no counterpart of this sentence without ba and fronting of the object. (14) ''Dajia shuyao bai wenti dao zuo-min-shang-lai 137

Thus, the presence of ba is obligatory for this sentence to be grammatical. Not only is ba required to occur, but the word order has to change. In sentences like ( 14), the object of ba cannot occur after the verb, and ba, behaving like a preposition, must take the object before the verb. In other words, one of the characteristics of the ha-construction is the fronting of the postverbal NP argument, as we have observed in (1 and 11). This is one of the constraints of the ha-construction. Although we have seen some of the reasons for the use of the ha-construction, the difference between the ba sentence and its non-ba counterpart has been much debated among Chinese grammarians. For example, Hu and Wen (1957) noted that if two constructions (ba and non-ba) having the same meaning can be expressed in different linguistic organisations, the listener understands without ambiguity, and the change of linguistic forms does not entail any extra meaning, then there is optionality for the use either of the two constructions (in this case, the ba and non-ba constructions) (Song 1991). Wang (1954), however, pointed out that if the meaning of the disposal construction and its 'plain' counterpart (non-ba, SVO) are identical, why does Chinese have these two linguistic forms? We maintain, following Wang and others, that these two constructions possess some semantically subtle differences. One of the central concept of the ha-construction observed by linguists is that it expresses the meaning of disposal, and we shall discuss it in the next section. We already noted that the ha-construction has been given different names. Perhaps the most widely known name for the ha-construction is the disposal construction, as first coined by Wang (1954), and adopted later by Chao (1968), Y-C Li (1974), and Li and Thompson (1981), etc. Let us consider the following examples first: (15) John ba cha-beida-po-le john ba tea-cup hit-break-asp John broke the teacup. (16) *John ba Mary at john ba mary love John loves Mary. The contrast in grammaticality between (15 and 16) can be explained by the idea of disposal ( eg Li and Thompson 1981). The disposal construction in Wang's words ( 194 7) states 'how a person is handled, manipulated, or dealt with, how something is disposed 138

of, or how an affair is conducted' (translated in Y-C Li 1974; cited Li and Thompson 1981: 468). In other words, disposal concerns what happens to the direct object. This concept as applied to the case of (16) is what (16) does not have. The verb ai 'love' in ( 16) is a verb of state/ cognition, which cannot affect the object in the sense of disposal. The concept of disposal can also be understood metaphorically or implicitly. While sentences like (16) cannot be disposed of, sentences like (17) is acceptable. (17) John ba Mary ai de yao si john ba mary love de want die John loves Mary so much that he wants to die. The phrase yao si 'want to die' helps indicate that the loves is so intense that it has some effect on Mary. The concept of disposal is not overtly expressed by the verb ai 'love' but by the attached phrase covertly. Thus, linguists ( eg Wang 1956, Li and Thompson 1981) conclude that the idea of disposal is the warrant for the ha-construction. Put it in modern terminology, the disposal idea can be said to have the same constraints as the idea of affectedness (e g Tenny 1987). M Wang (1987: 72), for example, treats 'disposal' as 'a high degree of the affectedness of the verb upon the direct object'. 5 .2 Variations of the ha-Construction Despite the fact that linguists and grammarians working on Chinese have considered the ha-construction as a particular construction, the ha-construction is by no means homogeneous. There are several ways to distinguish different types of the ba-construction, one of which is to observe the syntactic and semantic relations between the verb and its arguments, and between the preverbal and postverbal NP arguments in the ha-construction. Then, four types of the ha-construction are identified (eg Lu and Ma 1985, Zou 1993): objective, causative, locative and instrumental. Zou (1993) investigates two typical types of the ha-construction in some detail, namely, objective and causative, and provides a morphosyntactic analysis of these two types of the ba-construction within Government and Binding framework. Based upon Zou and others, we shall also consider these two frequently studied types of the ha-construction in the following sections. The most typical kind of ha-construction is the so-called objective ba-construction and has been observed by grammarians lilce Lu and Ma (1985) and Zou (1993). Examples of this kind are shown in the following. 139

(18) John ba pmgguo bo-le john ha apple peel-asp John peeled the apple. (19) Mary ba zhe-jian yifu xi-le mary ha this-cl clothes wash-asp Mary washed the clothes. (20) John ba fan chi-wan-le john ha nee eat-finish-asp John ate his rice. Structurally, this kind of ha-construction comprises a subject, ha, a preverbal NP argument, a transitive verb, and an aspect marker or a resultative/directional verb (particle). Moreover, from these examples, we can see that the preverbal NP argument is the logical object of the verb and the surface object of ha. Within this objective ha-construction, another type is also observed. This type is an alternation of the previous one, for instance: (21) John bo-le pinnguo-de pi john peel-asp apple's skin (22) John ba pinnguo bo-le pi john ha apple peel-asp skin John peeled the skin of an apple. (23) John bang-le Mary-de liang-zhi jiao john tie-asp mary's two-cl foot (24) John ba Mary bang-le liang-zhi JlaO john ha mary tie-asp two-cl foot John tied up Mary's two feet. 140

This type differs from the previous one in that this type comprises not only all the elements of the previous one, but also a postverbal NP after the aspect marker or resultative/directional verb (particle). In addition, the preverbal NP argument and the postverbal NP are the logical objects of the verb, but the preverbal NP argument is affected by the action of the verb and the postverbal NP is the target of the action denoted by the verb. On the surface, however, the preverbal NP is the object ofba, and the postverbal NP is the object of the verb. Furthermore, there is an inherent possessive or partitive relation between the preverbal and the postverbal NPs; the preverbal NP is a possessor and the postverbal NP possessee, or the preverbal NP indicates a whole and the postverbal NP indicates its part. As the name of this type of ha-construction suggests, this ha-construction denotes a causer and causee relation between the subject of the sentence and the preverbal NP of the verb phrase. (25) John ba yi-ge ze1 pao-le john ba one-cl thief run-asp John let a thief run away. (26) Mary ba John xiao-si-le mary ba john laugh-die-asp Mary made John laugh madly. This causative ha-construction is formed by a subject, ba, a preverbal NP, an intransitive verb, and an aspect marker or a resultative/directional verb (particle). The difference between the objective ha-construction and the causative ha-construction is the transitivity of the verb; the former being transitive and the latter intransitive. Just as there are two subtypes of the objective ha-construction, so can two subtypes of the causative ha-construction be identified. Another type of the causative ha-construction involves an embedded clause. (20) John ba Mary qi-de fa-feng-le john ba mary anger-till become-mad-asp John made Mary so angry that she went mad. 141

(21) Zhe-ping jiu ba John zui-de ZOU bu dong this-bottle wine ha john drunk-till walk not move This wine made John so drunk that he couldn't walk. The preverbal NP argument is the logical subject of the matrix verb and the embedded clause. In other words, in (20) Mary was angry and mad, and in (21) John was drunk and couldn't walk. In addition, the subject of the sentence and the preverbal NP argument has a causer and causee relation. Hence, the causative type can be considered to be the canonical type of the ha-construction. 5.3 The ha-Construction and Transitivity Now we consider the issue of transitivity and its relevance to the ha-construction. The notion of transitivity has long been a topic in linguistic scholarship ( eg Lyons 1968, Robins 1989; Hopper and Thompson 1980; Tsunado 1985, 1994). Almost all grammars will touch upon the topic (eg Quirk et al 1985; Huddleston 1984, 1988). Although the term seems very established, some linguists have some reservations about its actual application in language, and attempt to modify and extend it ( eg Tsunoda 1994, Hopper and Thompson 1980). We shall first consider the adequacy of the term and then the proposals of a new look at transitivity. We shall pay particular attention to Hopper and Thompson's proposal for transitivity, and its relevance to the Chinese ha-construction. The term transitivity is derived from Latin: trans 'across' and ire 'go'. The traditional sense of transitivity generally includes transitive clauses and intransitive clauses. Linguists often use such terms as transitivity, transitive and intransitive clauses, etc without bothering to define what they really mean. This is, understandably, not always easy to do. And it seems that these terms are clear enough, though it is not so if one takes a closer look at them. Let us first consider some statements of traditional concept of transitivity. They are usually like the following (Tsunoda 1994: 4670). • A transitive clause contains an object. An activity 'goes across' from the subject to the object. The object is sometimes affected by the activity, and/or a transitive clause can be turned into passive • An intransitive clause has no object, and no activity is transferred The examples The man hit her and She cried we have given already fall into these definitions of transitivity. However, this traditional sense of transitivity is far from being 142

adequate for explaining the wide range of linguistics phenomena regarding transitivity. Let us consider some examples: (22) John likes Mary. (23) Mary has two houses. Although these two sentences are said to be transitive, (22) cannot be regarded as having something to 'go across' from John to Mary. Also, if (22) has anything that passes across, that is the stimulus or perception from Mary. Thus, it does so in the opposite direction from the object to the subject. (23) is a transitive clause but it cannot have the passive counterpart, ie *Two houses are had by Mary. Example (24) is an intransitive clause, but it has a passive counterpart (25). This violates one of the traditional definitions of transitivity that an intransitive clause cannot be passivied. (24) John bumped into Mary. (25) Mary was bumped into by John. Furthermore, although (24) is regarded as an intransitive clause, the action seems to 'go across'. The traditional idea of transitivity cannot capture these phenomena displayed in these examples. Thus, we need to rriake up another term or we have to extend the traditional sense of transitivity in order to accommodate these linguistic phenomena. As linguists are already familiar with the traditional sense of transitivity, it seems to be sensible to extend or modify what transitivity can mean. In addition, the problem with the traditional definition of transitivity is that it does not make any clear difference between syntax and semantics of the concept ( Tsunoda 1994). The notional definition of transitivity has been criticised for its lack of precision and inappropriateness (Lyons 1968, Robins 1989). Another problem is that the traditional concept of transitivity assumes a clear distinction between transitive clauses and intransitive clauses, and so it cannot accommodate examples lil

hit her the verb hit is an transitive verb, while in She cried, the verb cried is intransitive. However, these terms are not, arguably, applicable to those languages very different from English. It seems, thus, reasonable not to refer to 'subject' and 'object' if a universally acceptable concept of transitivity is attempted. Another objection to the traditional concept of transitivity is that many languages, such as German and Turkish can have intransitive clauses passivied, even though passivizability is considered to be a feature of transitive clauses (Lyons 1968: 379). Moreover, many languages do not exhibit passives. Passivisability is then not a universal feature of transitive clauses. Some other features of transitive clauses should be examined in a full range and the concept should be extended or modified in order to achieve universality. According to Tsunoda (1994), transitivity should have such a characterisation that should distinguish between the semantic and morphosyntactic features of transitivity. Furthermore, transitivity is a matter of degree or a continuum/scale (Hopper and Thompson 1980), rather than exhibiting a clear separation. Based on this assumption, a prototypical transitive clause, which could serve as a basis of measuring the degree of transitivity of other clauses, can be postulated. In other words, the more transitivity features a clause has, the more transitive a clause is. A clause with more transitivity features will be more transitive than a clause with fewer transitive features. These features are said to be systematically co-vary. Hopper and Thompson (1980: 254 original emphasis) observe: In the grammars of all the languages we have examined, we find that these component features of Transitivity CO-VARY ext~nsively and systematically whenever an obligatory pairing of two Transitivity features occurs in the morphosyntax or semantics of a clause, THE PAIRED FEATURES ARE ALWAYS ON THE SAME SIDE OF THE HIGH-LOW TRANSITMTY SACLE Upon this observation, they further (1980: 255: 9) propose the transitivity hypothesis, which states: If two clauses [a] and [b] in a language differ in that [a] is higher in Transitivity according to any of the features lA-J, then, if a concomitant grammatical or semantic difference appears elsewhere in the clause, that difference will also show [a] to be higher in Transitivity. 144

id is intransitive. es very different md 'object' if a objection to the nan and Turkish msidered to be a .nguages do not ve clauses. Some and the concept characterisation ctic features of :ontinuum/scale m . Based on this .sis of measuring more transitive a ive than a clause 1atically co-vary. 1ed, we find that extensively and 1res occurs in the FEATURES ARE TRANSITMTY tivity hypothesis, t [a] is higher in if a concomitant m the clause, that y. The transitivity hypothesis makes a very strong claim about transitivity feature correlates. However, the hypothesis is shown to be too strong in that some languages do not exhibit such kind of correlates stated by the transitivity hypothesis (Tsunoda 1985). In Hopper and Thompson (1980), ten semantic parameters are proposed to be subsumed by transitivity, particularly, the prototypical transitive clauses. They are participants, kinesis, aspect, punctuality, volitionality, affirmation, mode, agency, affectedness of O(bject) and individuation of O(bject). The prototypical transitive verbs can be defined as those involving an activity that causes a change in the patient. Together with these ten parameters, a sentence is said to exhibit a continuum of transitivity. In other words, some sentences have a high degree of transitivity and others low. In English, examples of such verbs are kill, break, clean, stop, and give, etc. However, there are transitive verbs in English that would not be qualified as prototypical transitive verbs in terms of the definition given above. These verbs would include, for example, hit and kick, because they do not necessarily imply a change in the patient. Be that as it may, these two types of verbs are not immediately obvious in English. In this connection, prototypical transitive clauses can be considered to consist of two participants and a predicate that is a prototypical transitive verb. Apart from the semantic properties, prototypical transitive clauses are also characterised by possessing the mrophosyntactic properties such as linguistic expressions of the two participants, namely, agent and patient; case frame; verbal morphology; and the possibility of parallel constructions like passive and antipassive, or reflexives and reciprocals, and so on. These properties are also used traditionally to distinguish between transitive clauses and intransitive clauses. Let us consider the linguistic expressions of two participants. A clause is traditionally classified into transitive and intransitive, that is, in terms of the number of participants. Transitive clauses are at least two-place -with two participants {or arguments), while some clauses are three-place. For instance, in English, John gave a book to Mary is regarded as three-place. Some grammarians, for example, Huddleston {1988) will consider the sentence, which contains two 'objects', to be 'ditransitive', while the sentence which only has one 'object' are said to be 'monotransitive'. Intransitive clauses are usually one-place, but there are two-place intransitive clauses. In some languages, zero-place intransitive clauses are allowed, but it is not permitted in English, and a 'dummy subject', eg it is employed, according to the Theta-criterion (Chomsky 1986). The 'dummy subject', 145 , . .,. , , I

however, does not refer to any participant. In a language like Chinese, this is not entirely clear if we should consider the word Tianqi 'weather' is a 'dummy subject'. (26) Tianqi Ieng weather cold It's cold. Another relationship between the semantics and syntax oflinguistic expressions is that it is not one to one. (27) He drinks between meals. Semantically, drink certainly requires a drinker and a drink. However, ( 6) does not contain the drink. We have noted that transitivity is considered as a continuum or scale (Hopper and Thompson 1980; Tsunoda 1994). It can be visualised in terms of the table below with illustration of English examples. (28) John killed Mary. (29) He hit her. (30) John likes Mary. (31) John bumped into Mary. (32) Maryleft. (33) It is cold today. 146

Now let us look at the table (Tsunoda 1994: 4675). Semantic aspect Morphosyntactic aspect Participants impingement change expressions case frame passive reciprocal (34) + (35) + (36) + (37) + (38) (39) + + + + + + + + + + + + + + + reflexive + + + + + + + The clauses are tested against the semantic and morphosyntactic aspects and we can see different degrees of transitivity exhibited by the clauses. (34) means that the clause has all the properties of transitivity, while (39) indicates the absence of the property. It is argued that to treat transitivity as a continuum achieves universal validity and better describes the prototypical transitive clauses of the world's languages. The connection between the transitivity hypothesis and the ba-construction is that the ba-construction shows very strong transitivity according to the scale (Thompson 1973). Thus, the sentence John ba Mary da-si-le 'John killed Mary by hitting her' is fine, but John ba Mary xihuan-le is bad. The resultative da-si shows a change of state, impingement on the patient, and there are the participants killer and killee. If we can organise the different kinds of resultative constructions along such a continuum, it may look lil

high transitivity low transitivity Chinese: resultative verbs: Ta ku-shi-le zhentou ba-cosntruction: Ta ba zhentou ku-shi-le English: resultative John shouted himself hoarse John cried so that the pillow got wet There is overlapping between the resultative and the ha-construction because they are highly transitive and compatible such as the examples above. Further, the minimum requirement of the number of arguments for both constructions is one, which looks like intransitive, for example, Ta chi-baa le ('He ate and became full) and Ba wo xia-si ('I was scared'). If there are any differences between them, one of them is that there must be an argument after ba, even though the agent may not be there. For the English resultative, a high transitivity seems to be the nature of the construction, hence the requirement of Direct Object Restriction. 5.4 The 'Disposal' Construction and ba as a Radial Category The transitivity nature of the ha-construction has its origin as a 'disposal' structure. Consider L Wang's original formulation: The disposal form states how a person is handled, manipulated, or dealt with, how something is disposed of, how an affair is conducted. Since it is specifically designed for disposing, the disposal form cannot be used unless the action possesses the quality of disposal. For instance, wo ai ta 'I love him' cannot be restated as wo ba ta ai'I "take" him and love [him]'. Again, tao shu kai hua 'Peach trees bloomed' cannot be restated as tao shu ba hua kai 'Peach trees "take" the blooms and bloomed' (L Wang 1947: 160-161; translated by Y-C Li 1974: 200-201) 148

That the term disposal is so widely used though many linguists, for example, Lii 1948, H Wang 1959, and Chao 1968 have pointed out the inadequacy of Wang's original theory of disposal before the idea of affectedness was introduced. In fact, more recent studies like Q Wang (1985) understand the concept as basically the same as that of affectedness but just use the traditional term of disposal. Although the notion of affectedness has been one of the issues among linguists recently (eg Tenny 1987), we maintain that disposal is a useful concept, and that the disposal type is the prototypical ha-construction. The problems with the disposal explanation of the ha-construction can be observed from the example below. ( 40) Zhe-ci bing ba this-cl illness ba John zhemo de john torture de The illness made John suffer so much. ku le bitter part It is difficult to imagine that the postverbal NP arguments can be said to be 'handled, manipulated, or dealt with'. For this reason alone, various linguists (eg Hu and Wen 1957) are led to reject the disposal theory. In a nutshell, the literature against the meaning of disposal as the defining property of the ha-construction includes the following reasons: (i) The verbs in some of the ha-sentences express emotion, thought, perception have no disposal meanings (ii) In some ha-sentences, the element introduced by ba does not represent the patient of the action, but the agent of the action Seemingly, the disposal approach is defective. However, despite this criticism, the idea of disposal is still widely adopted among Chinese linguists. As Song (1991) points out, what is noteworthy of the disposal idea depends upon how one interprets the term. For many of the linguists, for example, Hu and Wen, 'disposal' means that one intentionally disposes of something. According to this literal interpretation, many ha-sentences cannot be accounted for. If we understand the term in a broader sense, as noted by Li and Thompson's (1981) metaphorical use of the term, the idea of disposal can be accepted more easily. Song ( 1991: 5) gives a revised version of the interpretation of 'disposal': 149

The action represented by the predicate of a sentence has some positive effect on the element introduced by ba, making the element undergo some kind of change, produce some kind of result, or be in some kind of state. Accordingly, 'disposal' indicates the relationship between verbal structure and the element introduced by ba, b_ut not the intentional behaviour of the agent or theme represented by the subject. There are advantages of recognising the disposal function or meaning of the ba-construction. As we noted earlier, there are cases where the ha-construction is optional. For instance, we can say this in the normal SVO construction: (41) Dai ta shang-lai bring him up-come or in a ha-construction: (42) Ba ta dai shang-lai ba him bring up-come Bring him up. to mean basically the same thing. However, these two sentences may have some pragmatic differences. One of the differences between them is that in an appropriate situation, the ha-construction is used for emphasis and has a stronger tone tban the non-ba sentence. Very often, whether to use a particular construction depends upon not merely its wellformedness in isolation, but also its connection in the discourse and context. (43) Yisheng ge-le lanwei doctor cut-asp appendix (44) Yisheng ba lanwei ge-le doctor ba appendix cut-asp The doctor cut the appendix. 150

Although these two sentences are semantically the same, the ba-sentence is preferred in the following piece of discourse. (45) Yi-ge ren fa-le lanweiyan, yisheng ba lanwei one-cl man suffer-asp appendicitis doctor ba appendix ge-le, zhe-ge ren jiu jiu-qu-lai-le cut-asp that-cl man at once save-out-come-asp A man suffered appendicitis. The doctor cut his appendix. The man was then saved. The ba-construction is preferred in this discourse because it emphasises or focuses the action and its result, ie the disposal function, but not the object. Let us consider an opposite case, where the non -ba-sentence is preferred. If we replace the second coordinating clause with a ha-clause (ie erjie yao ba lunwen xie-wan 'also need ba thesis write-finish'), the clause will be rhetorically inappropriate, even though the two clauses are basically semantically equivalent. (46) Women we xie-wan write-finish bu-dan not-only lunwen thesis yao nian-hao want read-good shu, erjie yao book also want We not only need to study well, but also need to write up the thesis. Since the first coordinating clause is the normal SVO pattern emphasising the direct object, and the whole sentence expresses a gradual strength of tone, so the second coordinating clause follows the same tone of voice using the SVO non-ba-construction, and emphasising the direct object. If the ba-construction is used to replace the non-ba-construction to emphasise the action and its result, this will not be compatible with the first coordinating clause. In a word, the evidence above shows that there are certainly differences between the ba-construction and the canonical SVO non-ba counterpart in grammatical functions. The ba-construction produces another means of expression in Chinese. This structure, like the resultative construction, carries out a particular communicative function that 'represent[s] complex integrations of events by making 151

maximum use of existing grammatical constructions' (Fauconnier 1997: 173). What is different is that the pragmatic function of the ha-construction typically subsumes that of the resultative construction, ie making further prominence of the patient that undergoes some kind of change. Apart from the discourse-pragmatic functions, taking the ha-construction as a disposal structure has some bearing on Hopper and Thompson's transitivity hypothesis that the ha-construction is on a high degree of the scale, producing change of state and result. Since the ha-construction causes to have result and change of state, it is highly compatible with the resultative construction. We have already seen the examples of their co-occurrence abound, for instance: ( 47) John da-si-le Mary ( 48) John ba Mary da-si-le John killed Mary. In some cases, they have to co-occur, such as John ha huaping da-po-le 'John broke the vase', as opposed to *John ha huaping da. Hopper and Thompson (1981) and Tsunoda (1994) provide us with a useful way of capturing the difference between the ba-construction and the resultative construction as they both exhibit a high degree of transitivity. !John ba Ma1y da-si-lq ~ The ha-construction is a larger construction than the resultative construction and therefore the resultative construction is contained within it. The shaded area is the resultative construction as we have already seen and the ha-construction contains it. Here we also assume that the ha-construction also consists of the subject argument. Another way to look at this is that ha is considered to be head of the entire construction of the clause, although there can be adjunct insertion between the subject argument and ha. liohn henhende ba Mary da-si-lq The adjunct henhende 'fiercely' shows the manner modifying ha Mary da-si-le. While different from the resultative construction, the ba-construction takes three arguments: 152

the Agent/Actor/Causer, the Patient/Undergoer and the Result-Goal, which is realised by a resultative construction. Sem CAUSE-BECOME R R : instance, PRED means Syn 1 < agt pat result-goal > < > l COMP(lement) The reason for not taking Mary da-si-le as a constituent is that we can ask: Ta ba Mary · zeme le? 'What did you do to Mary?' In the above diagram result and goal is simply used as a convenient label and should be taken metaphorically as change of state. Moreover, we recognize the causative nature of the construction. Instead of taking the resultative as an oblique, we consider it simply to be a complement subcategorised by ba. We can name this the 'disposal' construction, which exists independently of other constructions, given its syntactic, semantic and pragmatic functions. The analysis has the implicit claim that the ba is similar to a catenative verb, much the same as the first verb of the resultative. The verb ba is a catenative verb taking a complement clause, which is realised by a resultative construction. Thus, ba is like a verb persuade in English taking an NP argument and a clausal complement. ( 49) John persuaded Peter to kill Mary. (50) John ba Mary da-si-le. The ha-structure is then an embedding structure having the resultative as subordination. This is also found in English such as John intended Mary to expect to find a better husband. The fact that after ba no aspectual marker can occur may act against it as cosubordination. (51) *John ba-le Mary da-si-le. 153

Thus, on the continuum of the complex sentence types (Croft 2001), the ba-construction leans towards the complement end of the continuum between adverbial clauses and complements. The complement NP, for example, Mary in the above case, must be adjacent to ba. This 'adjacency requirement' is derived from the general iconicity principle that conceptually closely related items come close in the language. A subtype of iconicity is isomorphism (Haiman 1985) between meaning and form that exists in language. This has already been seen in the discussion of the resultative construction. In the ba-construction the meaning of the construction ie 'disposal' requires that the NP argument be next to it, since the argument is an Undergoer for some kind of change or for some effects. The structure matches the meaning in the way that things happen. (52) John ba Mary da-si-le This diagram means that John made Mary to become in a particular state. In this connection, it will be appropriate to adopt the polysemy links use_d in Construction Grammar (eg Goldberg 1995). These links 'capture the nature of the semantic relations between a particular sense of a construction and any extensions from this sense' (Goldberg 1995: 75). For instance, the caused-motion construction has several related senses (Goldberg 1995: 76): • X causes Y to move Z • Conditions of satisfaction imply 'X causes Y to move Z' • X enables Y to move Z • X causes Y not to move from Z • X helps Y to move Z The first one is a central sense and others are derived from it. The derived senses have to do with force-dynamics (Talmy 1985), which is related to causation. Thus, one way oflooking at the ba-construction is through this polysemy link with the caused-motion construction: 154

Sem CAUSE-MOVE < cause theme goal > R R : instance, Pr < > means 1 1 l Syn V COMP(lement) This is similar to the previous diagram but this one is explicitly about caused-motion. In John ba Mary da-si-le, John makes Mary have a different state (death) as ifhe moved her to a different place. This is a metaphorical extension. Another way to see this is to treat baas a radial category (Fong 2015). The following shows a variety of meaning denoted by the ba-construction through the interaction of the complement verbs: !change of ,ta~ :eu)ationl lposses~,"-.~ \intensiv~ Some examples include: (53) John ba yingyinben fang-dai-le. John ba photocopy open-big-asp John enlarged the book. (54) John ba John ba pingguo bo-le pi apple peel-asp skin John peeled the skin of an apple. (55) John ba shuben ban-dao waimian qu. John ba book move-arrive outside go John moved the books to the outside of the building. 155 (change of state) (possession) (movement)

(56) John bu ba wo clang pengyou. John not ha me as friend John does not regard me as a friend. (57) John ba fan chi-wan-le. John ha rice eat-finish-asp John finished eating his rice. (intensive) (result) As a consequence, a better way to characterise the ha-construction in terms of meaning, in my view, is to consider them in a scale of gradience from the cental ha-construction meanings to the non-ha-construction meanings. Furthermore, in the construction grammar framework (eg Goldberg 1995), multiple inheritance is allowed so that the exact interpretation of the ha-construction depends upon the input from the predicate, the ha-construction and the construction associated with a particular meaning. If we take John ha Mary da-si-le as an example, the following diagram may represent how the sentence obtains its meaning. Instead of two inputs, ha also contributes to a more abstract meaning associated with result, manipulation, movement, etc. In other words, ha can be considered to be a radial category (Lakoff 1987) in the sense that it is defined by conventions but may not be predicted by rules. As a result, if the -historical meaning of ha as 'take' or 'hold' is regarded as the basic model, from which all other semantic models are derived, the following models will be produced: (58) a. disposal b. result c. change of state d. manipulation e. possession f. motion g. intensive This list is not exhaustive, and as has been said earlier, the interpretation of the ha-construction depends on its interaction with the complement. When the different conditions meet collectively, they may form a prototypical case of a category, which may instantiate the Idealised Cognitive Model (ICM, Lakoff 1987) of the category. In the case of ha, speakers may consider result or disposal to be central properties of the function of the ha-construction. 156

This metaphorical extension is, in my view, involves transitivity in different senses. Under the transitivity hypothesis, John acts on Mary in the example above, which involves participants, change and impingement. If ha is considered to be a case marker, it involves the morphosyntactic aspect of the hypothesis. In terms of structure, the postverbal NP argument must be present and that fits the traditional concept of transitivity. Another sense is that the relationship between the caused-motion construction and the ha-construction is transitive ie extension of senses. The strategy of the extension of senses has further consequences that connections between the canonical resultative, non-canonical resultative and the ha-construction can be established. (59) John~daoMary~le. ( 60) John~ Mary lda-sil le. (61) Johnlda-sij-leMary. This connection echoes what Fauconnier (1997: 137) observes - that complex events are integrated into a structure by 'making maximum use of existing grammatical constructions'. Example (59) contains. two separate predicates showing two separate (sub )events. The patient Mary followed da 'hit' and and si 'die' is predicated of Mary, which reflects the natural sequence of the happenings - iconicity. Example ( 60) contains the two predicates merged together but with Mary before the verb da 'hit', and Mary is still understood as the patient. This time, the structure, however, relies on the use ofba. Since Mary precedes the two predicates da-si,••the use ofba, which manipulates Mary in a particular way, is appropriate. Without ha, the structure of (60) contains only the two predicates that are adjacent to each other making a minimum effort for the action and result, and the structure requires that the patient Mary should follow the two verbs, like an object of them, if they are considered to be a single predicate, because this is the only position it can take. The predicates become tighter and closer, but si 'die' is still predicated of the patient Mary. The phenomenon of ha also has typological implications. The ha-construction manifests the typological type of manipulative meaning as that argued for by Cristofaro (2003), which is high on the subordinate clause deranking hierarchy. That shows more support to the idea that the ha-construction must take complement (clauses), and that 157

ha also expresses a kind of purpose and result. Thus, ha should occur between purpose and manipulative on the scale, after the resultative construction. They are close together because they denote similar concepts, and another way to interpret the ha-construction is to see it as a 'double-resultative' structure ( or 'disposal-resultative'), whose resultative function is taken to be primary in this study. ha-resultative (62) John ba Mary da-si le. lresultativ~ If the canonical resultative emphasises the outcome of the event, the ha-construction emphasises both the patient argument (sometimes regarded as a secondary topic by linguists such as Tsao 1990, Xu and Liu 2007) and the outcome of the resultative event, by making the patient argument close to ba, so that the hearer or addressee must wait until the end to find out the activity and outcome of the event. Typologically, creating emphasis and focus of attention requires different structures, and very often non-canonical ones. The example of John kicked Mary to death in English specifies the result path to death after the patient in the canonical clause while the cleft sentence, which is non-canonical, can put the patient in a focusing perspective It's Mary that John kicked to death. The Chinese ha-construction seems to perform a similar function in John ha Mary da-si le in that the patient Mary is focused and emphasised by ha. In other word, the patient Mary is 'manipulated' by ha and by the whole construction. Let us look some more examples. The first example (63) below is grammatical but the second one ( 64) is not. (63) John ba ma qi-lei le john ha horse ride-tired asp John rode the horse and the horse became tired. (64) *John ba yifu xi-lei le john ha clothes wash-tired asp *John washed the clothes and they became tired. 158

The second example (64) is bad because it is semantically anomalous - clothes cannot feel tired because they are not animate. What is syntactically odd is the use of the ha-construction here. If we understand the structure correctly, we would expect the clothes to be manipulated in a certain way by ha, and more particularly, the clothes are manipulated in a way that they are tired. On the other hand, a horse can be manipulated and it can feel tired. This can only be done by considering the ha-sentence as a special construction in Chinese, which has its own form, meaning and function. 5.5 Blending In Chapter 3 we established that the resulative construction is a kind of grammatical blending. In fact, we can argue that the ha-construction is a form of blednign. Thus, the verb ha can be regarded as a space builder that anticipates the following event, as shown below: (65) ha-event In this view, the main function of ha is to present the stage for the following event to unfold in the discourse, in relation to the Agent. It does not only highlight the Patient but the entire event associated with the arguments. Since the word ha itself does not have much lexical content, it interacts with its complement as a whole to achieve the construction's meaning such as manipulation and change of state. Thus, we know what kind of events is involved in John ha Mary da-si-le, but not* John ha. The exact meaning of ha is derived from the entire construction. As a space builder, ha can contribute to various mental spaces (Fauconnier 1994, 1997), depending upon how it is actually used in the discourse. It creates different spaces matching the individual meaning of the complement. Although we recognise that perhaps the more central meaning of the ha-construction is 'disposal' or 'cause-result', 159

the less prototypical, unpredictable meaning provides further support to the idea that the ba-construction is a construction of its own. (66) Ta ba yi-ge da hao jihui cuoguo le. he ba one-cl very good opportunity m1SS asp He missed a very good opportunity. The exact interpretation of the ba-construction depends on several factors. Structurally, there are the verbal elements, and the construction is transitive in nature. Semantically, the meaning of the verbal elements must be compatible with the meaning of the ba-construction, for example, producing results or change of state. In other words, the construction's argument structure roles must be fused with the corresponding participant roles provided by the verbal element (Goldberg 1995). There is, last but not least, context in which the construction is used. On the other hand, the mental spaces that ba builds provides us with a convenient way to understand the discourse and the relationship between the participants and the spaces. (67) ba-event In this example cuoguo is more like a single unit as the bu 'not' insertion is not allowed here: cf. * cuo bu guo 'not miss', which perhaps constitutes a single event. So, on one interpretation the Agent may or may not have intentionally missed the opportunity, and the Patient or Theme here does not seem to undergo any transformation and change. The function of ba, again, is to establish a stage for the event to begin in the discourse. The diagram above is meant to illustrate the flexibility of the blending model to fit the various senses of the ba-construction. The current proposal does not merely revises Goldberg's (1995: 5) claim that 'simple clause constructions are directly associated with semantic structures which reflect scenes basic to human experience' but also admits that 'scenes basic to human 160

experience' like cause and effect or result are reflected by more integrated constructions, and while the cause-and-effect scene is one of the basic scenes of human experience, its manifestations may take various forms across languages. 161

Chapter6 Conclusion This work has examined the Chinese resultative construction and its relevance to the ha-construction. We have considered both the formal and functional approaches to the resultative construction and how the constructions differ from each other in English and Chinese. For instance, the Direct Object Restriction is quite well observed in English but it does not apply to Chinese very well. The resultative construction as a research focus lies in its interface between lexical semantics and syntax. Different frameworks have emerged. For the last fifteen years a constructional approach has been active and fruitful (eg Goldberg 1995, Croft 2001). Linguists have switched their focus on syntactic rules to the properties of the lexical items and constructions, which means that linguistic phenomena can be explained in terms of argument structure, that is, the representation of properties that allow lexical items to take arguments. The study of these properties of lexical items inevitably yields the detailed analysis of the meanings of their predicates. Despite the fact that the study of lexical semantics separated in the 1970s linguists from the interpretive semantics and generative semantics camps (Huck and Goldsmith 1995; Newmeyer 1983, 1986), it has a unifying implication today allying linguist of various current theories. The results of their investigations have been incorporated into most linguistic theories in a similar way. The underlying assumption of most of the current research in linguistics that meanings of words are represented in the syntactic properties of phrases makes way for the new method for the study of word meanings. A set of words sharing similar meanings may require similar syntactic representations, and thus provides us with the semantic structures of word meanings. This study argues for an integrated approach to the resultative and by 'integrated' is meant examining the syntax and semantics of the structure in English and Chinese from both functional and formal angles, with special reference to Chinese. Chapter Two reviews the essential and important issues raised by the resultative construction investigated by different theoretical frameworks. Although the problems associated 162

with the issues have not been entirely solved, they received a fair treatment and discussion. Chapter Three discussed the interaction between lexical verbs and the resultative construction and, particularly, the classes that can enter into the structure. Meanwhile, the verbs' transitivity alternations showed a strong connection between transitivity of the verbs and the resultative construction. Chapter Four deals with several different analyses of the resultative construction, namely, catenative analysis, cosubordination analysis and constructional anlaysis, etc. These analyses share one property in common that the resultative construction in Chinese exhibits clausal features in the complement. Instead of taking it as a compound, as many linguists do, we argue that the resultative in Chinese is similar to that in English being a legitimate syntactic-semantic construction in language. Interacting with the ha-construction, the resultative construction is seen as an extended caused-motion construction in general in Chapter Five. In particular, we argue that the resultative is a syntactic structure as manifested as a special predicate-argument (mini-) construction and subordinate structures. We argue that ha is a legitimate verb - a catenative verb and that it is a radial category. The ha-construction requires to a large extent the resultative complement because of its intrinsic nature of being causation or 'disposal'. This area is a theoretically challenging field and theories are tested against the background of the study of the resultative. Descriptively, the issues are complex and there is much more to understand. At the general linguistic level, the resultative construction is 'associated directly with semantic structures which reflect scene§_ basic to human experience' (Goldberg 1995: 5). It seems that in different cultures the basic human experience involves cause and effect. The happenings in the world are often ascribed to human or spiritual controls; and thus having a cause for almost everything is common among societies and this is reflected in different linguistic structures, which may show surface discrepancies but a shared experience and a common conceptual structure. One of the strategies to show these common properties in languages is to blend or integrate grammatical constructions together into a single structure such as a causative, resultative or serial verb construction. This strategy has the advantages of packaging more information, over time, into a highly tight single whole than information shown by different unconnected structures demonstrating strong transitivity. And even for clauses there are ways of information packaging that requires minimum effort from the language structure, for instance, concatenation so that the predicates can lump together 163

economically to reflect the sequence of happenings in the world. This demonstrates some typological significance. At the language-particular level, we have seen that the resultative in Chinese differs from its English counterpart to the degree of using pragmatic-discourse factors, on which Chinese is more heavily dependent than English. The Direct Object Restriction in English is an example, while Chinese does not observe this principle well, and thus allows a null complement easily such as John qi-lei le as opposed to John rode the horse and became tired. The missing argument in Chinese is covert and understood by the application of the world lmowledge and pragmatics that qi requires a thing to be ridden on. And lei is predicated of John as it is the only argument in the sentence but it is not in the universe of discourse. The approach taken here should be transferred to other models fairly smoothly. It should be useful to examine the structure of language through a more developed cognitive model of language whose metal spaces model has been proven valuable in the study of counterfactuals, conditionals and blending. The model works well with construction grammars that all levels of language are studied together and they represent different levels of constructions. The resultative construction can be considered as an extension of grammatical blending and integration where the Chinese predicates integrate more closely that those found in English. This can also be seen as a continuum where on the one hand, there are completely analytic structures lil

Bibliography Aarts B 2004 Grammatici Certant. Journal of Linguistics 40: 365-382 Alsina A et al (eds) 1997 Complex Predicates. CSL!: Stanford Anderson S 1985 Typological Distinctions in Word Formation. In Shopen (ed) Language typology and syntactic description (vol 3). CUP: Cambridge Boas H 2003 A Constructional Approach to Resultatives. CSL!: Stanford Bresnan J and A Zaenen 1990 Deep Unaccusativity in LPG. In K Dziwirek et al eds Grammatical Relations P45-57 CSL!: Stanford Butt M 1995 Complex Predicates in Urdu. CSL!: Stanford Butt Mand W Geuder (eds) 1998 The Projection of Arguments. CSL!: Stanford Carrier J and J Randall 1992 The Argument Structure and Syntactic Structure of Resultatives. Linguistic Inquiry 23: 173-234 Chang S-M 1994 V-qi-lai Constructions in Mandarin Chinese. MA diss National Chenghua University Chao Y-R 1968 A Grammar of Spoken Chinese. UC Berkeley: California Cheng L LS 1989 Transitivity Alternations in Mandarin Chinese. In M Chan and T Ernst (eds) 1989 Proceedings of the third Ohio State University Conference on Chinese Linguistics. Indiana University Linguistics Club, P81 - 94. Cheng Land C-T J Huang 1994 On the Argument Structure of Resultative Compounds. In In honour of William S-Y Wang: Interdisciplinary Studies on Language and Language Change. Pyramid: Taipei Chomsky N 1957 Syntactic Structures. Hague. Mouton Chomsky N 1965 Aspects of Syntax. MIT: Cambridge Chomsky N 1986 Knowledge of Language. Praeger: London Comrie B (ed) 1993 Causatives and Transitivity. New York: Benjamins Comrie Band M Polinsky (eds) 1993 Causatives and Transitivity. Amsterdam: John Benjamins Cristofaro S 2003 Subordination. OUP: Oxford Croft W 1998 Event Structure in Argument Linking. In M butt and W Geuder (eds) The Projection of Arguments (p21-64) . CSU: Stanford Croft W 2001 Radical Construction Grammar. OUP: Oxford Croft W 2002 Typology and Universals (2nd ed). CUP: Cambridge 165

Croft W and Cruse D A 2006 Cognitive Linguistics. CUP: Cambridge Ding P S-Z 2007 Studies on Ba Resultative Construction: A comprehensive approach to Mandarin Ba sentences. Lincom: Miinchen Dixon RM W 1991 A New Approach to English Grammar, On Semantic Principles. OUP: Oxford Dixon RM W 2010 Basic Linguistic TheoryVol 1. OUP: Oxford Dowty 1979 Word Meaning and Montague Grammar. Dordrecht: Reidel Dowty D 1991 Thematic Proto-roles and Argument Selection. Language 97 (3): 54 7-619 Fauconnier G 1994 Mental Space. CUP: Cambridge Fauconnier G 1997 Mappings in Thoughts and language. CUP: Cambridge Fauconnier G and M Turner 1996 Blending as a Central Process of Grammar. In A Goldberg ed Conceptual Structure, Discourse and Language. CUP: Cambridge Fillmore C and B Aitkens 1992 Toward a Frame-based Lexicon. In Frames, Fileds and Contrasts. eds A Lehrer and E Kittay 75-102 Erlbaum: Hillsdale Fillmore C et al 1988 Regularity and Idiomaticity in Grammatical Constructions: The Case of Let Alone. Language 64: 501 -538 Foley W and R Van Valin 1984 Functional Syntax and Universal Grammar. CUP: Cambridge Fong R 2015 The Chinese Ba as a Verb - a Cognitive Constructional Cpproach. Open Journal of Modern Linguistics 5, 48-64 Fong R 2016 Chinese as Satellite-framed - a Constructional-cognitive Intepretation. Cognitive Linguistic Studies 3:2, 234-259 -Fong R 2019 Resultative Constructions in English and Chinese - a Functional-cognitive Analysis. Linguistics and the Human Sciences 13(3):241-263 Frei H 1957 The Ergative Construction in Chinese Part II. Genga Kenkyu 32: 83-115 Givon T 1990 Syntax: A Functional-Typological Introduction Vol2. John Benjamins: Amsterdam Goddard 1998 Semantic Analysis. OUP: Oxford Goddard 2011 Semantic Analysis Rev ed. OUP: Oxford Goldberg A E 1991 A Semantic Account of Resultatives. Linguistic Analysis 21: 66-96 Goldberg A E 1992 The Inherent Semantics of Argument Structure: The Case of the English Ditransitive Construction. Cognitive Linguistics 3: 37-74 166

Goldberg A E 1995 Constructions: A Construction Grammar Approach to Argument Structure. University of Chicago Press: Chicago Goldberg A E 2005 Argument Realization. In Construction Grammars. Ostman J and Mirjam Fried (eds). John Benjamins: Amsterdam Goldberg A E 2006 Constructions at Work. OUP: Oxford Goldberg A E and Jackendoff R 2004 The English Resultative as a Family of Constructions. Language 80 (3): 532-568 Grimshaw J 1991 Argument Structure. MIT: Massachussattes Guo R 1995 Valent Structure of Verb-resultative Complement and Integration. In Y Shen and D Zheng (eds) Studies on Valent Grammar in Modern Chinese. Peking Univ Press: Beijing Guo R 1995 Valent Structure ofVerb-resultative Complement and Integration. In Shen and Zheng (eds) Studies on Valent Grammar in Modern Chinese. Peking University Press: Beijing Gu Y 2003 On the Syntactic Projections of Causatives. In Xu Jet al (eds) Chinese Syntax and Semantics (p27-62). Pearson: Singapore Haiman J 1985 Natural Syntax. CUP: Cambridge Hale K and S J Keyser 1993 On argument structure and the lexical expression of syntactic relations. In Hale and Keyser (eds) The View from Building 20. MIT: Massachusetts Halliday MAK 1967 Notes on Transitivity and Theme: Part I. Journal of Linguistics 3: 37-81 Hashimoto A 1966 Embedding Structures in Mandarin. Ohio State Univ diss. He Y 2003 On the Syntax of Causatives in Mandarin Chinese. In Xu J etal (eds) Chinese Syntax and Semantics (p63-78). Pearson: Singapore Higginbothem J 1985 On Semantics. Linguistic Inquiry 16: 547-593 Hoekstra T 1988 Small Clause Results. Lingua 74: 101-139 Hopper P and S Thompson 1980 Transitivity in Grammar and Discourse. Language 56: 251-295 Horrocks 1987 Generative Grammar. Longman: London Huang J 1988 Wo Pao De Kuai and Chinese phrase structure. Language 64: 274-311 Huang C-T J et al 2009 The Syntax of Chinese. CUP: Cambridge 167

Huang C-T J 1992 Complex predicates in control. In Lahiri, Iatridou, Larson and Higginbotham (eds) Control and grammar. Kluwer Academic Press: Dordrecht Huang L 1980 Resultative Verb Compounds. MA diss University of Florida Huang S-F 1982 Papers in Chinese syntax. Crane: Taipei Huck G and J Goldsmith 1995 Ideology and Linguistic Theory. Routledge: London Huddleston R 1974 Further Remarks on the Analysis of Auxiliaries as Main Verbs. Foundations of Language 6: 322-361 Huddleston R 1976 Some Theoretical Issues in the Description of the English Verb. Lingua 40: 331-383 Huddleston R 1984 An Introduction to English Grammar. CUP: Cambridge Huddleston R 1988 English Grammar. CUP: Cambridge Huddleston R and Pullum G 2002 The Cambridge Grammar of the English Language. CUP: Cambridge Jackendoff R 1990 Semantic Structures. MIT: Cambridge Kay P 1997 Words and the Grammar of Context. CSU: Stanford Keyser S J and T Roeper 1984 On the middle and ergative constructions in English. Linguistic Inquiry 19: 335-391 Lakoff G 1987 Women, Fire and Dangerous Things. Chicago University Press: Chicago LaPolla R 1993 Arguments against the Subject and Object in Chinese. Bulletin of the Institute of History and Language. Academia Sinica: Taipei LaPolla R 2009 Chinese as a Topic-Comment (Not Topic-Prominent and Not SVO) Language. In JZ Xing (ed) Studies of Chinese Linguistics. HK University: Hong Kong Langacker R 1987 Foundations of Cognitive Grammar Vol J. Stanford University Press: Stanford Langacker R 1991 Foundations of Cognitive Grammar Vol 2. Stanford University Press: Stanford Levin B 1993 English Verb Classes and Alternations. Chicago Univ Press: Chicago Levin Band M Rappaport Hovav 1995 Unaccusativity. MIT: Massachussetts Li Y HA 1990 Order and Constituency in Mandarin. Kluwer: Dordrecht Li C and S Thompson 1981 Mandarin Chinese: a Functional Reference Grammar. UC Berkeley Press: California Li Y 1990 On V-V compounds in Chinese. Natural Language and Linguistic Theory 8: 168

177-207 Li Y 1993 Structural heads and aspectuality. Language 69: 480-504 Li Y 1995 The Thematic Hierarchy and Causativity. Natural language and linguistic theory: 255-282 Liang Y 2006 Hanyu Dongbu Jiegou de Changsheng yu Yanbian (The Origins and Changes of the Chinese Verb Complement Structure). Xuelin: Shanghai Lu J H-T 1977 Resultative Verb Compounds vs Directional Verb Compounds in Mandarin. Journal of Chinese Linguistics 5: 276-313 Newmeyer F 1983 Grammatical Theory. Chicago University Press: Chicago Newmeyer F 1986 Linguistic Theory in America (2nd ed). Academic Press: New York Newmeyer F 1991 Language Form and Language Function. MIT: Massachusetts Newnham R 1971 About Chinese. Penguin: Hardmondsworth Mangione, L 1982 The Syntax, Semantics and Pragmatics of Causative, Passive and Ba Constructions in Mandarin. Ph.D. dissertation, Cornell University. Packard J 2000 The Morphology of Chinese. CUP: Cambridge Palmer F 1974 The English Verb. Longman: London Palmer F 1987 The English Verb (2nd ed). Longman: London Palmer F 1990 Modality and the English Modals (2nd ed). Longman: London Payne T E 2011 Understanding English Grammar. CUP: Cambridge Perlmutter D M 1978 Impersonal Passives and the Unaccusative Hypothesis. BLS 4: 157-189 Quirk Rand S Greenbaum 1991 A Student's Grammar of the English Language. Longman: London Quirk R et al 1985 A Comprehensive Grammar of the English Language. Longman: London Rappaport Hovav Mand B Levin 1998. Building Verb Meanings. In M butt and W Geuder (eds) The Projection of Arguments (p97-135). Stanford: CSLI Rappaport Hovav Mand B Levin 2005 Chang-of-State Verbs. In N Erteschik-Shir and T Rapoport (eds) The Syntax of Aspect (p274-288). OUP: Oxford Rothstein S 2004 Structuring Events. Blackwell: Oxford Sag I et al 2003 Syntactic Theory 2nd ed. CSLI: Stanford Sells P 1985 Lectures on Contemporary Syntax. CSLI: Stanford 169

Shi C 2009 Hanyu Dongjieshi de Qufa Yuyi Yan jiu (The Study of Syntax and Semantics of the Chinese Verb-resultative Constructions). Beijing Language and Culture University Press: Beijing Shi Y 2001 Xiandai Hanyu de Dongbu Jiegou de Neixingxue Tezheng (The Typological Characteristics of the Resultative Construction in Modern Chinese). Jiangxi Jieyu Press: Nanchang Shi Y 2002 The Establishment of Modern Chinese Grammar. John Benjamins: Amsterdam Shi Y and C Li 2001 A History of Grammaticalization in Chinese. Peking University Press: Beijing Simpson J 1983 Resultative. In L Levin, M Rappaport and A Zaenan (eds) Papers in lexical-functional grammar. Indiana University Linguistics Club: Bloomington Smith C 1990 Event types in Mandarin. Linguistics 28: 309-336 Song W 2007 On the Syntax of Modern Chinese V-R Compounds. Peking University: Beijing Song Y 1991 Xiandai Hanyu Yufa Jiben Zhishi (Basic Knowledge of Modern Chinese Grmmar). Yuwen Press: Shanghai Sybesma R 1992 Causatives and accomplishments: the case of Chinese BA. Holland Institute of Linguistics: Dordrecht Tai J 1984 Verbs and time in Chinese. CLS 20: 289-296 Talmy L 1985 Lexicalisation Patterns. Language Typology and Syntactic Description ( ed) T Shopen 57-147 CUP: Cambridge Talmy L 1998 Force Dynamics in Language and Cognition. Cognitive Sciences 2: 49-100 Tang S-W 2010 Formal Chinese Syntax. Peking University Press: Beijing Tang T-C 1977 Studies in transformational grammar of Chinese (vol I): movement transformations. Student Book Co: Taipei Tang T-C 1979 Chinese relative clauses. In Studies in Chinese Syntax. Student Book Co: Taipei Tang T-C 1979 Studies in transformational grammar of Chinese (vol 1): movement transformations. Student Boole Taipei Tang T-C 1989 Studies on Chinese Morphology and Syntax: 2. Student Book: Taipei Teng S-H 1985 Readings in Chinese Transformational Syntax. The Crane Publishing Co: Taipei 170

Thompson S 1973 Resultative Verb Compounds in Mandarin Chinese: a case for lexical rules. Language 49: 361-379 Tsao F-F 1990 Sentence and Clause Structure in Chinese. Student Book Co: Taipei Tsunoda T 1993 Transitivity. In Asher R and JMY Simpson eds The Encyclopedia of Language and Linguistics. Pergamon: Oxford Van Valin R 1990 Semantic Parameters of Split Intransitivity. Language66 (2): 221-260 Van Valin R 1993 A Synopsis of Role and Reference Grammar. In Advances in Role and Reference Grammar (pl-164) . John Benjamins: Amsterdam Van Valin R and R LaPolla 1997 Syntax: Structure, Meaning and Function. CUP: Cambridge Vendler Z 1967 Linguistics and Philosophy. NY: Cornell University Press Wang H 1984 Baziqu he Beiziqu (Ba-sentences and Bei-sentences). Shanghai Education: Shanghai Wang H 1995 Valency of the Resultative Verb Compounds. In Y Shen and D Zheng (eds) Studies on valent grammar in modern Chinese. Peking U niv Press: Beijing Wang L 194 7 Zhongguo Xiandai Yufa. ( Chinese Modern Grammar). Chinese Bookshop: Shanghai Washio R 1997 Resultatives, Composidonality and Language Variation. Journal of East Asian Linguistics 6: 1-49 Wechsler S 1997 Resultative Predicates and Control. In Blight Rand Michelle Moosally (eds) Texas Linguistic Forum 38 (p307-321) Wechsler Sand B Noh 2001 On Resultative Predicates and Clauses. Language Sciences 23:391-432 ~ Williams E 1981 Argument Structure and Morphology. The Linguistic Review 1: 81-114 Xu L and D Liu 2007 Topic: Structural and Functional Analysis. Rev ed. Shanghai Educational Publishing: Shanghai Yuan X 2005 Xiandai Hanyu Zhishiju Yanjiu. (The Study of Modern Chinese Causative Sentences) Zhejiang University: Zhejiang 171

LINCOM Studies in Chinese Linguistics In this series 01 Weiyuan Xu 02 Liwei Gao 03 Liang Chen 04 Haiyong Liu 05 Ruixi Ressy Ai 06 Chao Li 07 Daniel z. Kadar 08 Christopher Button 09 Markus Greif 10 Yuntong Liu 11 Patrizia Pacioni 12 Yishan Huang 13 Yishan Huang 14 Ronald Fong A Study of Chinese Colour Terminology Chinese Internet Language A Study of Identity Contructions The Acquisition and Use of Motion Event Expressions in Chinese Bu-Yu, the Complex-Predicate Structures in Mandarin Chinese Elliptical Predicate Constructions in Mandarin Mandarin Resultative Verb Compounds: Where Syntax, Semantics, and Pragmatics Meet Model Letters in Late Imperial China -60 Selected Epistles from Letters from Snow Swan Retreat Phonetic Ambiguity in the Chinese Script A Palaeographical & Phonological Analysis Corrective Focus in Mandarin Chinese -A Question of Belief? A Dictionary of Slang among Chinese Youth Noun Classifiers: a View from Cantonese Zhangzhou Southern Min: Rhyme Tables, Homonyms, Heteronyms Zhangzhou Southern Min: Syllables and Phonotactics The Chinese Resultative Construction A Cognitive Constructional Analysis

This study examines the structure and function of the Chinese resultative construction. Specifically, it makes an attempt to integrate the form and function , syntax and semantics, Chinese and English, and formal and functional perspectives. Since the interface between syntax and lexical semantics has intrigued researchers in linguistics and related fields more than ever, the argument structure has been postulated to capture the relationship betw~en syntax and semantics. Resultative verb compounds in Chinese have been shown to provide insights into the theory of argument structure as the syntax-semantics interface. In particular, this work examines the verb classes that enter into the resultative construction in Chinese and proposes an innovative analysis of the construction based upon the insights from contemporary studies such as Cognitive Construction Grammar (Goldberg 1995, 2006) and Huddleston and Pullum (2002). Further, this study incorporates typological evidence and argues that the resultative construction signals the concept of subordination. The analysis is then extended to the language-particular phenomenon in Chinese - the ba-construction, which shows further usefulness of the proposed analysis. In fact, the current framework demonstrates the interesting results of the interaction between the ba-construction and the resultative construction in Chinese. LSCHL 14 ISBN 978 3 86290 219 4 LINCOM EUROPA academic publications JLL lLlJ