* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download CHAPTER 5 THE LIGHT VERB SYNTAX IN CHINESE
Germanic strong verb wikipedia , lookup
Polish grammar wikipedia , lookup
Japanese grammar wikipedia , lookup
Old Irish grammar wikipedia , lookup
Scottish Gaelic grammar wikipedia , lookup
Macedonian grammar wikipedia , lookup
Udmurt grammar wikipedia , lookup
Swedish grammar wikipedia , lookup
Modern Hebrew grammar wikipedia , lookup
English clause syntax wikipedia , lookup
Old English grammar wikipedia , lookup
Ancient Greek grammar wikipedia , lookup
Kannada grammar wikipedia , lookup
Navajo grammar wikipedia , lookup
Portuguese grammar wikipedia , lookup
Sotho verbs wikipedia , lookup
Hungarian verbs wikipedia , lookup
Latin syntax wikipedia , lookup
Yiddish grammar wikipedia , lookup
Icelandic grammar wikipedia , lookup
Chinese grammar wikipedia , lookup
Spanish grammar wikipedia , lookup
Kagoshima verb conjugations wikipedia , lookup
Georgian grammar wikipedia , lookup
Serbo-Croatian grammar wikipedia , lookup
CHAPTER 5 THE LIGHT VERB SYNTAX IN CHINESE 1. Introduction In this chapter we look at the light verb syntax from several different perspectives. In section 2, we look at other some other light verbs in Mandarin Chinese sentences, in particular those with higher functional contents. In section 3 we summarize all the light verbs we have seen in Mandarin Chinese, and demonstrate the "Davidsonian" character of Mandarin Chinese. In section 4, we extend our perspective to cover other Chinese dialects. It is shown that, though some Chinese dialects do not exhibit the full range of unselectiveness of subject and object as Mandarin Chinese, some regularities can be discerned. Our focus will be specially placed on Taiwanese, which makes extensive uses of lexical realization, rather than head movement, as the means to support the light verbs in the phrase structure. In section 5, we turn to the historical perspective and examine the case of Archaic Chinese. It is shown that Archaic Chinese, surprisingly, was more like the presentday English in allowing productive deadjectivals and denominals. We provide an explication for the relevant phenomena, and propose an account for the historical changes of the light verb syntax in Chinese. At the end of this chapter is a concluding remark. 2. Other light verbs in Mandarin Chinese In Chapter 3 and 4, we examined the phenomena of unselectiveness of subject and object in Mandarin Chinese. We proposed that all arguments in Mandarin Chinese sentences are licensed by light verbs. All of the light verbs that we saw introduce an argument into the sentence and help to shape the eventuality by contributing a specific aspect to it. But there can be light verbs that simply contribute to the shaping of the eventuality without 252 introducing arguments. In the following, we will look at two kinds of elements that fall into this class, the aspectual markers -le (perfective), -zhe (durative), and -guo (experiential), and the extension marker -de. Besides, there are higher functional categories in a Mandarin Chinese sentence that take the whole light verb structure as scope. We will look at two of them in this section, the raising modals and the tense category T. 2.1 The aspectual markers In Chapter 2 and 3 we encountered some questions related to the aspectual markers in Mandarin Chinese. Here we will look at more details. Some researchers in Chinese linguistics notice that the post-verbal aspectual markers in Mandarin Chinese, the perfective -le, the durative -zhe, and the experiential -guo, play a substantial role in defining the type of eventuality for a given predicate. Smith (1994) proposes that these aspectual markers denote different temporal viewpoints to be superimposed on types of event. Smith (1994) gives the following characterization for these aspectual markers: (1) a. -le: perfective viewpoint that closes non-stative situations b. -zhe: imperfective viewpoint on continuous, stable situations c. -guo: perfective temporal quantifier ranging over a set of situations To illustrate the functions of these aspectual markers, let's look at the following example. Suppose we have a predicate chi niu-rou mian 'eat beef noodle', as in (2a). This predicate represents an eventuality that is temporally neutral. As Mandarin Chinese requires all predicates to be aspectually interpreted, (2a) by itself is an incomplete predicate. To make it complete, a temporal viewpoint must be superimposed upon it. For this requirement, the verb chi 'eat' can be suffixed with one of the aspectual markers, and thereby receive a temporal interpretation, as in (2b-d).1 1 (2a) has a generic reading according to which "eat beef noodle" is a habit or acceptable thing to the subject (a "generalizing" predicate in the sense of Krifka et al. (1995)). We ignore this reading and the 253 (2) a. chi niu-rou mian eat beef noodle 'eat beef noodle' b. chi-le niu-rou mian eat-Perf beef noodle 'ate beef noodle' c. chi-zhe niu-rou mian eat-Dur beef noodle 'is eating beef noodle' d. chi-guo niu-rou mian eat-Exp beef noodle 'have the experience of eating beef noodle' relevant questions, but simply assume that a generic operator Gen is responsible for this reading (cf. Carlson 1977, 1989, Krifka et al. 1995). There is an aspectual marker in Mandarin Chinese that many Chinese linguists focus on but we don't discuss in the text, the progressive zai 'at'. According to Smith (1994), the progressive marker zai 'at' "presents an internal interval of a durative event and has the connotations of activity and the temporary which is associated with non-statives" (p.120). The use of zai 'at' can be illustrated by the following example: (i) Laowang zai chi niu-rou mian. p.n. at eat beef noodle 'Laowang is/was eating beef noodle.' We have pointed out in section 3.1 of Chapter 4 that zai can be used as the main verb of a Mandarin Chinese sentence, as the following example shows: (ii) Laowang zai xuexiao. p.n. at school 'Laowang is/was in the school.' We will simply take the progressive marker zai 'at' as an aspectual verb. It can take a VP, as in (i), or an NP, as in (ii), as complement. This is actually in line what we will propose for all the other aspectual markers. The special thing about zai is that, while verb movement applies in the cases of all other aspectual markers, for zai, the main verb doesn't incorporate to it. This explains why other aspectual markers are post-verbal and suffixal, but zai is not. We will ignore questions related to zai in the following discussion. 254 Though the aspectual markers -le, -zhe, and -guo are suffixes to the verb, according to the viewpoint theory of Smith (1994), they actually take the whole VP as scope -- the viewpoints represented by these aspectual markers in (2b-d) are not just superimposed on the verb chi 'eat' alone; they are superimposed upon the whole predicate chi niu-rou mian 'eat beef noodle'. Syntactically, these aspectual markers should occur in a structural position higher than the VP. Indeed, this is the assumption underlying Cheng's (1991) postulation of an independent syntactic category for these aspectual markers, Asp. In Cheng's (1991) proposal, AspP is a projection in Mandarin Chinese that corresponds to IP in English. The head Asp contains aspectual features to be checked with the verb. However, Cheng (1991) notices a problem if the verb is assumed to move to Asp to check the aspectual features. If the verb in Mandarin Chinese moves to Asp overtly, then VP adverbials, which presumably are adjuncts of VP, should occur post-verbally. But this is not correct. The following examples show this point: (3) a. [AspP Subj [Asp' Asp0 [VP AdvP [VP ...V... ] ] ] ] b. *Laowang manman-di qie-zhe cai. *p.n. slowly cut-Dur vegetable *'Laowang cuts the vegetable slowly.' c. *Laowang qie-zhe manman-di cai. *p.n. cut-Dur slowly vegetable *'Laowang cuts the vegetable slowly.' If the aspectual markers are hosted in Asp0 in (3a), then, since the aspectual markers, such as the durative -zhe in (3b-c), end up suffixed to the verb in the surface structure, the verb V must move overtly to Asp0 in Syntax. This head movement, however, leaves the Adv, manman-di 'slowly' in (3b-c), in post-verbal position, ungrammatical in Mandarin Chinese. 255 Therefore, it must be the case that, firstly, the V-to-Asp0 movement in Mandarin Chinese takes place in LF, and, secondly, the aspectual markers in fact don't occur under Asp0 -Asp0 only contains the relevant aspectual features, and the verb V has already had an aspectual marker suffixed to it at the beginning stage of the derivation. The V-Suffix complex moves up to Asp0 in LF to check the relevant features. While all this appears to be plausible, there are still other important points about the aspectual markers that we have to notice. Firstly, the superimposition of a temporal viewpoint via the suffixation of an aspectual marker doesn't seem to have anything to do with the thematic status of the subject. For instance, (2b-d) are all agentive, regardless of which aspectual marker is suffixed to the verb. Thus there seems to be no harm if we assume that the scope of the aspectual markers doesn't include the subject-selecting light verb. Secondly, all the three aspectual markers, the perfective -le, the durative -zhe, and the experiential -guo, exhibit strong verbal characters. Historically, all these three aspectual markers were verbs; they underwent grammaticalization and became functional elements (cf. T. Mei 1981, 1991 for relevant discussion). What is more, these aspectual markers nowadays still retain verbal uses. Below are some examples:2 (4) a. Zhe-ge wenti bu liao-jie, women dou hui you mafan. this-Cl problem not end-terminate we all will have trouble 'If this problem doesn't come to an end, we will all have trouble' 2 In (4a-b), -le and -zhe are phonetically realized as liao and zhao. The formers are generally believed to result from phonetic reduction from the latters. See Sun (1996) for relevant discussion. Also in (4a) liao 'to end' occurs as a bound morpheme. This seems to be typical for the use of this morpheme, in verbal compounds and idioms such as zhong-liao 'end-end (= finish)' and mei-wan-mei-liao 'no-finish-no-end (= endless)'. But liao 'to end' can be used as a free morpheme in some limited cases, as below: (i) Laowang liao-le yi-chuang xin-shi. p.n. end-Perf one-Cl heart-matter 'Laowang had a bothering matter resolved.' An interesting thing in (i) is that both liao and -le occurs, both originating from the same source. But in an example like (i), they have very different functions -- liao is the main verb with the specific meaning 'to end', but -le is a pure functional element denoting perfectivity. 256 b. Laozhang yi dian huochai, zhe-dui baozhi jiu zhao-le huo p.n. as-soon-as lit match this-pile newspaper then get-Perf fire 'As soon as Laozhang lit up the match, this pile of newspaper got fire' c. zhi zou, guo san-ge hong-lü-deng, jiu dao Laowang jia le straight go pass 3-Cl traffic-light then arrive p.n. home Prt 'Go straight, pass three traffic lights, then you will arrive at Laowang's home' In their verbal uses, these three elements have the meanings "to end," "to attach," and 'to pass." The aspectual uses of these three elements, as verbal suffixes, are clearly related to their original meanings. In fact, it has been claimed that the aspectual meanings of these three elements resulted from grammaticalization of their original verbal meanings. (See Peyraube 1996 and the references cited there.) As a matter of fact, Sybesma (1997) proposes to treat the perfective -le and the durative -zhe as verbs, specifically resultative and resultative-stativizing predicates respectively. According to Sybesma's (1997) analysis, (5a) have the structure in (5b) ((9ab), Sybesma 1997: 222): (5) a. Ta mai-le ta-de nei ji-tou zhu. he sell-Perf his that several-Cl pig 'He sold those few pigs of his.' b. Ta mai [ [ ta-de nei ji-tou zhu ] [ le ]] he sell his that several-Cl pig Perf In (5b), the perfective marker -le stands in a very low structural position. It has the function to mark the resultative state of the main predicate. The durative -zhe can be analyzed in the same way, according to Sybesma (1997). 257 We do not plan to provide a complete argumentation against Sybesma's (1997) structural analysis of -le and -zhe, but it is clear that, in the present context, they are better considered as occurring in a structural position higher than the predicate, as we have pointed out above. At any rate, it seems quite plausible to assume that, even in their suffixal uses, the perfective -le, durative -zhe, and experiential -guo still retain the verbal status in the syntactic structure. In other words, they are light verbs that license no arguments. Though these aspectual markers were verbs in Archaic Chinese and are still being used as verbal elements, they have been grammaticalized to an extent that they lose the ability to license NP arguments, and only participate in the shaping of an eventuality. If this line of thinking is correct, (2b-d) can be analyzed in the following way: .... (6) VP V' Subj VP V DO V VP NP −le − zhe − guo niu-rou mian 'beef noodle' V' V chi 'eat' In (6), the main predicate chi niu-rou mian 'eat beef noodle' (order irrelevant) is the complement of the Asp-VP. This Asp-VP, in turn, is the complement of the DO-VP. The main verb chi 'eat' moves up, firstly to the aspect marker, and then to DO, resulting in the surface structure in (2b-d). The advantages of this analysis include the following: in this structure, the aspectual markers take the main predicate as the scope (excluding the subjectselecting light verb), and, assuming that the VP adverbials such as manman-di 'slowly' are adjoined to the subject-selecting VP, higher than the Asp-VP, syntactically overt movement 258 of the main verb to the higher light verbs will not result in the kind of ungrammaticality as in (3c), where the adverbials occur post-verbally. To conclude, it is likely that the aspectual markers in Mandarin Chinese are actually light verbs. They are eventuality predicates and contribute to the shaping of the eventuality. They used to be full-fledged verbs in Archaic Chinese, and still retain verbal uses in Mandarin Chinese, but they have lost the ability to license an argument due to grammaticalization. They are suffixal, hence they attract the main verb to move to them. They have specific functions; that is, they help to determine the temporal viewpoints of eventualities. Thus they are no different from other light verbs in Mandarin Chinese, in the sense that they are all predicated of eventualities. 2.2 The extension marker In Chapter 2, we saw some phenomena related to the extension marker -de. Here, again, we will look at more details about this morpheme. The questions about -de are important in several respects. Though we will not go into all of them, we will suggest that the extension marker -de, like the aspectual markers, is a light verb that contributes to the shaping of the eventuality but licenses no NP argument. The extension marker -de introduces a clausal complement to the verb it is suffixed to. The clausal complement can be either descriptive or resultative (cf. section 3.2.4 of Chapter 3). In the former case, the -de complement describes the manner with which an eventuality is happening, and, in the latter, it denotes the resultant state of the eventuality. Some examples are given below: (7) Descriptive a. Laowang pao de hen kuai. p.n. run Ext very fast 'Laowang runs fast.' 259 b. Xiaoli ku de hen da-sheng. p.n. cry Ext very loudly 'Xiaoli cries very loudly.' (8) Resultative a. Laowang pao de hen lei. p.n. run Ext very tired 'Laowang runs to the extent that he gets tired.' b. Xiaoli ku de dao zai di-shang. p.n. cry Ext fall at ground-on 'Xiaoli cries to the extent that he fells on the ground.' While the examples in (7-8) contain intransitive verbs, transitive verbs can take the -de complement, too. Below are some examples: (9) a. Xiaoli (chi niu-pai) chi de hen kuai. (Descriptive) p.n. (eat steak) eat Ext very fast 'Xiaoli eats (steaks) fast.' b. Xiaoli (chi niu-pai) chi de hen ni. (Resultative) p.n. (eat steak) eat Ext very sick 'Xiaoli eats (steaks) to the extent that he gets sick of it.' But there are interesting questions with the intransitive verbs with the -de complement. Firstly, when a transitive verb takes a -de complement, the "logical" object of the transitive verb cannot occur in its canonical post-verbal position -- the transitive verb is just followed by the -de complement, as though the -de complement were an object-like expression. Secondly, though the object of the transitive verb cannot occur in the canonical post-verbal position, it may occur pre-verbally, sometimes with a repetition of the main verb, a 260 phenomenon known as "verb copying" (cf. K. Mei 1972, 1978, Huang 1982, C-R, Huang and Mangione 1985, Huang 1988, Li 1990, among others). Both points manifest in (9a-b). In (9a-b), the main verb chi 'eat' takes a -de complement post-verbally, without a post-verbal object; on the other hand, the object niu-pai 'steak' occurs pre-verbally, with a repetition of the main verb chi 'eat'. Though we will not go into detailed discussion on the "verb copying" phenomenon, it seems that this phenomenon has to do with some semantic/pragmatic requirements for the identification of the eventuality, rather than with syntactic factors in the strict sense.3 As to the absence of the post-verbal object when the de complement is present, here we adopt Huang's (1988) analysis, discussed in section 3.2.1 of Chapter 3, that the main verb and the -de complement conjointly form a complex predicate. A schematic analysis for (9a-b) is given in the following (cf. Sybesma 1992):4 3 The crucial point that motivates such an assertion is that, the copied verb and the pre-verbal object seem to be adverbial in nature. The VO expression that the two elements constitute can occur in an adverbial clause, as in the following examples: (i) [ Na tian chi niu-pai de-shihou, ] Laowang chi-de hen kuai. that day eat steak when p.n. eat-Ext very fast 'That day, when [he] at steaks, Laowang ate [them] very fast.' In (i), the copied V and the object occur in a 'when'-clause. Presumably, elements in such an adverbial clause don't enter into any selectional relationship with the verb in the main clause. This is the major reason that we propose that the copied verb and the (pre-verbal) object are probably required due to some semantic/pragmatic factors. (Examples of (i) kind were first pointed out to me by Jim Huang in a lecture given at Academia Sinica, Taipei, 1989.) 4 Sybesma (1992) proposes a structure for the -de complement very similar to the one in (10), reproduced below (adapted from (38), Sybesma 1992: 77) (i) VP V ExtP Ext XP In (i), the head Ext is lexically realized as de, which introduces a resultative complement XP. Though (i) looks almost exactly the same as (10) in the text, there are several crucial differences. First, Sybesma (1992) does not discuss the question how the "logical" object of the verb V can be absent. Second, Sybesma (1992) considers the resultative expression ExtP as directly selected by the verb V. But on these two points we are in quite different positions. We assume that V doesn't select an internal argument, hence the internal argument doesn't need to (and sometimes cannot) occur (cf. section 3.3 of Chapter 4). Also, we assume that ExtP is not a selected item of V; it is a secondary predicate (cf. section 3.2.3 of Chapter 3, and Huang 1988). 261 .... (10) VP Subj V' V VP DO V' ⇒ Complex predicate V VP chi 'eat' V' V -de CP description {Manner } Resultative state In (10), the -de complement, as a VP (see below), functions as a secondary predicate to the main verb chi 'eat'. The element -de, as a bound morpheme, incorporates to the main verb chi 'eat'. The result is a V' complex predicate. This V' doesn't take an external object and projects a VP directly. The light verb DO then is merged to the structure, introducing an agent subject to the sentence. There are several questions to be answered regarding the analysis in (10). For example, in (10), the complex predicate V' projects directly to a VP without taking a subject first. But is a subject possible in this structure? Another question is, in (10), we identify the extension marker -de as a verb that introduces a CP complement. Is there reason that it should be so identified? In the following, we will take a closer look at these questions. 262 2.2.1 The internal subject Firstly, let's look at the question on the subject of the complex predicate V' in (10). In this structure, V' doesn't take a subject, and, indeed, an (internal) subject in this structure is usually unacceptable. Consider the following examples: (11) a. *Laowang (chi niu-pai) chi de hen kuai. *p.n. (eat steak) eat Ext very fast *'Laowang eats steaks very fast.' b. *Laowang chi de niu-pai hen kuai. *p.n. eat Ext steak very fast *'Laowang eats steaks very fast.' (12) a. *Laowang (chi niu-pai) chi de hen ni. *p.n. (eat steak) eat Ext very sick *'Laowang eats steaks to an extent that he gets sick of them.' b. *Laowang chi de niu-pai hen ni. *p.n. eat Ext steak very sick *'Laowang eats steaks to an extent that he gets sick of them.' The ungrammatical examples (11b) and (12b) have a structure similar to (10), except that the V' complex predicate takes an internal subject, the underlined expression niu-pai 'steak'. On the other hand, however, we do observe some grammatical examples where an internal subject occurs, such as the following: (13) a. Laozhang xi de zhe-liang che gan-gan-jing-jing. p.n. wash Ext this-Cl car completely-clean 'Laozhang washed this car to the extent that it is completely clean.' 263 b. Laozhang zhuang de zhe-zuo men yao-yao-yu-zhui. p.n. strike Ext this-Cl door shaky-fallen 'Laozhang struck this door to the extent that it became shaky.' Notice that two factors distinguish the grammatical examples in (13a-b) from the ungrammatical ones in (11b, 12b). The first factor is that, both the grammatical examples in (13a) and (13b) are resultative. It seems that, with a descriptive -de complement, the structure can not take an internal subject. The second factor is, the resultant state which the -de complement denotes must be one which the internal subject, not the main subject of the sentence, ends up in. In both (13a) and (13b), the resultant states, gan-gan-jing-jing 'completely clean' and yao-yao-yu-zhui 'shaky'', are situations that the internal subjects, zheliang che 'this car' and zhe-zuo men 'this door' end up in. It seems that if these two factors are satisfied, the internal subject becomes acceptable. On the other hand, (11b) involves a descriptive -de complement, hence is ungrammatical; (12b) involves a resultant state for the main subject, therefore the sentence is ungrammatical. We believe that these two factors -- resultative, and internal subject oriented -- have to do with the semantics of the V' complex predicate. A predicate takes a subject that undergoes the eventuality denoted by the predicate. If there is an internal subject for the complex predicate V', the internal subject should denote the entity that undergoes the eventuality denoted by the predicate. In the ungrammatical examples in (11b) and (12b), the internal subject of the V' complex predicate, niu-pai 'steak', doesn't really undergo the complex eventuality chi de hen kuai 'eat very fast' and chi de hen ni 'eat to the extent so as to get sick'. It is the subject of the sentence, Laowang, that does. Therefore, the internal subjects in (11b) and (12b) are not semantically appropriate, probably due to failure of predication.5 On the other hand, the internal subjects in (13a-b), zhe-liang che 'this car' and 5 That is, it is not the subject of any predicate. See the condition of Formal Licensing in section 3.3.2 of Chapter 4, which requires all NP arguments in Mandarin Chinese sentences to be the subject of a predicate. 264 zhe-zuo men 'this door', do undergo the complex eventualities xi de gan-gan-jing-jing 'wash to be completely clean' and zhuang de yao-yao-yu-zhui 'strike to become shaky'. Thus they are acceptable internal subjects. 2.2.2 The disposal construction A syntactic construction in Mandarin Chinese deserves special mentioning here, the ba construction (also known as the disposal construction; cf. Wang 1980, K. Mei 1978, Huang 1988, Li 1985, 1990, F. Liu 1997, among many others). Since many intriguing questions are involved in the ba construction, we cannot provide a complete and thorough discussion for this construction. But we would like to address some questions relevant for the discussion here. A cursory look on the function of the disposal marker ba suggests that it seems to license a preposed object: (14) a. Xiaowang da-le Xiaoli. p.n. hit-Perf p.n. 'Xiaowang hit Xiaoli.' b. Xiaowang ba Xiaoli da-le. p.n. Disp p.n. hit-Perf 'Xiaowang hit Xiaoli.' (15) a. Wo wang-le yaoshi le. I forget-Perf key Prt 'I forgot [to bring up] the key.' b. Wo ba yaoshi wang le. I Disp key forget Prt 'I forgot [to bring up] the key.' (Of course, for this explanation to work, we must assume that multiple subjects for a single predicate are not permitted.) 265 An important function of the disposal marker ba, as noticed by Li (1985, 1990) and Huang (1988), is that it provides a syntactic position for object-like elements when they are not able to occur post-verbally. A typical case is the situation in which a -de complement occurs. Consider the following examples: (16) a. Laozhang ba zhe-liang che xi de gan-gan-jing-jing. p.n. Disp this-Cl car wash Ext completely-clean 'Laowang washed this car and made it completely clean.' b. Laozhang ba zhe-zuo men zhuang de yao-yao-yu-zhui. p.n. Disp this-Cl door strike Ext shaky-fallen 'Laozhang struck this door and made it become shaky.' (16a-b) are identical to (13a-b) except that, in the present case, the internal subjects zheliang che 'this car' and zhe-zuo men 'this door' are associated with the disposal marker ba. According to a popular view (e.g. Li 1985, 1990, Huang 1988), the occurrence of ba here has to do with Case-theoretic reason. A verb in Mandarin Chinese can assign only one accusative Case (cf. section 3.3 of Chapter 4). Since in (16a-b) the post-verbal position, which gets Case-marked, is occupied by the -de complement, the objects of the transitive verbs xi 'wash' and zhuang 'strike' must be realized at pre-verbal position, where ba can assign an independent accusative Case to them. According to this view, therefore, ba is a accusative Case marker that assigns Case to an NP that cannot receive one elsewhere. This view, however, is difficult to maintain in our analysis, if the post-ba NP is considered the internal subject of a complex predicate, V' in (10). Here we would like to suggest a different view on ba. Following Sybesma (1992), we propose that the disposal marker ba actually is a verb, just like the passive marker bei in Mandarin Chinese (Huang 1999). It takes an NP as specifier (which is the subject of the sentence), and a VP as complement. There seems to be 266 evidence for this proposal. Firstly, the ba-NP string cannot move as a constituent, evidenced by (17); secondly, the conjunction test shows that ba takes the main VP as complement, as in (18): (17) a. *Xiaozhang ba Laowang henhen-di da-le yi-dun. *p.n. Disp p.n. violently hit-Perf one-time *'Xiaozhang did a violent beating on Laowang.' b. *Ba Laowang, Xiaozhang henhen-di da-le yi-dun *Disp p.n. p.n. violently hit-Perf one-time *'Xiaozhang did a violent beating on Laowang.' (18) Laozhang ba Shengjing nian-le san-bian, Kelanjing nian-le wu-bian. p.n. Disp Bible read-Perf three-time Koran read-Perf five-time 'Laozhang read the Bible three times, the Koran five times.' The structure containing ba, in our view, is like the following: (19) a. Zhangsan ba Lisi da le. p.n. Disp p.n. hit Prt 'Zhangsan hit Lisi.' .... VP b. NP Zhangsan V' VP V ba NP V' V Lisi da 'hit' 267 (19a) is a typical ba sentence. (19b) is a partial structure for (19a). In (19b), the disposal marker ba occurs as a verb licensing a specifier, Zhangsan. It also takes the VP that the main verb projects as complement. The object of the main verb, Lisi, is the internal subject of the lower VP. Thus, though in the surface structure Lisi seems to be introduced by ba, actually it is not. It is merged to the VP as a specifier of the main VP. This structural analysis accounts for the observation that the disposal marker ba and the NP that follows it do not make a constituent. Just like the aspectual markers in Mandarin, ba used to be a full verb in Archaic Chinese, meaning "to hold, to grasp." It underwent grammaticalization and became a highly functional word. (See Peyraube 1996 for an overview on the grammaticalization of ba and references.) As many researchers have noticed, ba has the function to attribute complete transitivity to the predicate (Teng 1975, Sybesma 1992, F. Liu 1997, among others). Let's assume that ba is an eventuality predicate that denotes complete transitivity. In this regard ba is fairly like a subject-selecting light verb, and its thematic function is very much like DO or CAUSE. This view on the grammatical status of ba, then, associates it with the subjectselecting light verbs and dissociates it from the NP that follows it. Now we come back to the -de complement. In the light of the discussion on ba, (16a-b) can be analyzed as involving a complex predicate structure with an extra ba-VP layer. Take (16a) as an example: (20) a. Laozhang ba zhe-liang che xi de gan-gan-jing-jing. p.n. Disp this-Cl car wash Ext completely-clean 'Laowang washed this car and made it completely clean.' 268 .... VP b. NP Laozhang V' VP V ba NP V' zhe-liang che 'this car' V V xi 'wash' -de VP CP Pro gan-gan-jing-jing 'Pro completely clean' In (20b), the main verb xi 'wash' and the -de complement gan-gan-jing-jing 'completely clean' constitute a complex predicate. This complex predicate takes the NP zhe-liang che 'this car' as specifier. The resulting VP, furthermore, is embedded under the ba-VP, which introduces Laozhang as the subject of the sentence. The point here is that the post-ba NP, zhe-liang che 'this car', is not introduced by ba, but is licensed by the complex predicate V'. With this understanding on the disposal marker ba, we can now go back to the general question about the internal subject of the complex predicate V' with a -de complement. We have seen that the complex predicate V' sometimes can still take an internal subject, as shown in (13a-b), repeated below: (13) a. Laozhang xi de zhe-liang che gan-gan-jing-jing. p.n. wash Ext this-Cl car completely-clean 'Laowang washed this car to the extent that it is completely clean.' b. Laozhang zhuang de zhe-zuo men yao-yao-yu-zhui. p.n. strike Ext this-Cl door shaky-fallen 'Laozhang struck this door to the extent that it became shaky.' 269 We suggest that (13a-b) are no more than variants of (16a-b), where the disposal marker ba is utilized. We repeat the examples: (16) a. Laozhang ba zhe-liang che xi de gan-gan-jing-jing. p.n. Disp this-Cl car wash Ext completely-clean 'Laowang washed this car and made it completely clean.' b. Laozhang ba zhe-zuo men zhuang de yao-yao-yu-zhui. p.n. Disp this-Cl door strike Ext shaky-fallen 'Laozhang struck this door and made it become shaky.' The structural analysis of (13a-b) in fact is identical to the ba structure in (20b), except that (13a-b) involve head movement of the main verb to the higher subject-selecting light verb. In short, the higher V in (20b) can be either ba or DO/CAUSE. In the former case, the main verb stays in-situ, but in the latter case, the main verb moves up and leaves the internal subject behind.6 In both cases, however, the licensing of the internal subject is the function of the complex predicate V'. It has no bearing on ba or the subject-selecting light verbs. Further evidence can be elicited to show the parallelism between the ba structure and the (13) type of sentences. We have seen that, for the complex predicate V' in the (13) type of sentences to take an internal subject, the -de complement must be resultative, not descriptive, and the resultant state must be for the internal object, not for the subject of the sentence. We observe the same effects in the ba construction: (21) a. *Laowang (chi niu-pai) chi de hen kuai. *p.n. (eat steak) eat Ext very fast *'Laowang eats steaks very fast.' 6 Sybesma (1992) proposes an analysis that essentially makes the same claim. That is, the ba construction and the -de structure are transformationally related. 270 b. *Laowang chi de niu-pai hen kuai. (=(11b)) *p.n. eat Ext steak very fast *'Laowang eats steaks very fast.' c. *Laowang ba niu-pai chi de hen kuai. *p.n. eat Disp steak eat Ext very fast *'Laowang eats steaks very fast.' (22) a. *Laowang (chi niu-pai) chi de hen ni. *p.n. (eat steak) eat Ext very sick *'Laowang eats steaks to an extent that * he is sick of them.' b. *Laowang chi de niu-pai hen ni. (=(12b)) *p.n. eat Ext steak very sick *'Laowang eats steaks to an extent that * he is sick of them.' c. *Laowang ba niu-pai chi de hen ni. *p.n. Disp steak eat Ext very sick *'Laowang eats steaks to an extent that he is sick of them.' The ungrammaticality of (21c) and (22c) is expected if the post-ba NP is licensed by the V' complex predicate to which it is a sister. In other words, they are ungrammatical for exactly the same reason as the ungrammatical examples (21b) (=(11b)) and (22b) (=(12b)). 271 2.2.3 The extension marker -de as a verb In (10), we identify the extension marker -de as a verb taking a CP complement. This is a view adopted from Huang (1988) and Sybesma (1992). But is there evidence that this identification of -de is the right one? For one thing, just like the aspectual markers and the disposal marker ba, -de used to be a verb in Archaic Chinese, and it still retains verbal uses in Mandarin Chinese. Its use as an extension marker is a result of grammaticalization in history. (See Sun 1996 for a detailed discussion on the grammaticalization of the morpheme de in the history of Chinese.) When it is used as a verbal element, de means "to acquire," as the following examples show: (23) a. Laowang de-le di-yi ming. p.n. acquire-Perf first number 'Laowang becomes the top one.' b. Juren-dui qu-de guanjun. Giant-team grasp-acquire champion 'The Giants won the championship.' Thus there is a ground to assume that the extension marker -de still retain the status of a verb when it introduces descriptive and resultative complements. There are other possibilities, of course. First of all, -de can be just a verbal suffix that attaches to a verb via morphological operations in the lexicon. Alternatively, it can be a preposition heading a PP following the main verb. Still another possibility is that it is the complementizer of the descriptive/resultative CP. But these possibilities are unsatisfactory in one way or another. If -de is a verbal suffix that attaches to a verb in the lexicon via some morphological process, then the verb will enter Syntax with the form V-de, taking a CP complement. But 272 just like the discussion of the Zhe operation in Chapter 3, such a morphological process would be very isolated in the grammar of Mandarin Chinese, to say the least. So it is not a very plausible hypothesis to assume that the morpheme -de gets suffixed tom the verb in the lexicon. Taking -de as a preposition that heads a PP with a CP complement is even more untenable. First of all, -de doesn't have independent prepositional use in Mandarin Chinese. Second, except for some limited cases, PPs in Mandarin Chinese typically occur in preverbal position.7 So -de cannot be the head of a PP. Notice that in the structural analysis in (10), -de is regarded as a verb that takes a CP complement. Is it possible to consider the descriptive or resultative clause an IP, and -de the complementizer of a CP? This possibility is equally unlikely. If we take this possibility serious, we have to claim that all verbs in Mandarin Chinese are proposition-taking verbs such as think and say. There is no reason that this claim could be correct. So -de cannot be a complementizer of a CP. In conclusion, we think that the best hypothesis for the morpheme -de is to consider it an independent head in the syntactic structure. It contributes a resultant or descriptive state to the eventuality. In this regard, it is no different from other light verbs, since they all help to shape the eventuality. 2.3 Higher functional categories So far we have seen many light verbs in Mandarin Chinese. They all project a VP, and contribute information to the eventuality. But conceivably there can be even higher and more functional categories that take the whole predicates as complement and assume propositional scope. These categories should be raising predicates. In a sense, the raising predicates are even more grammaticalized than such light verbs as the aspectual markers and the extension marker -de. They do not license NP argument. Furthermore, they even 7 The post-verbal PPs in Mandarin Chinese typically involve the element zai 'at' denoting the goal of some action. But we analyzed zai 'at' as a verb heading a secondary predicate in section 2.2.3 of Chapter 3. Thus it is very likely that, in Mandarin Chinese, there is no post-verbal PP at all. 273 permit a thematically unrelated NP to be their subject, via movement. Here we look at two cases, the raising modals and the tense projection TP. 2.3.1 Raising modals Huang (1989) notices that modals in Mandarin Chinese can be divided into two types, the control type and the raising type. The two types of modals are exemplified below: (24) Control modals a. Laozhang gan zai laohu tou-shang pai cangying. p.n. dare at tiger head-on flap fly 'Laozhang dares to hit a fly on a tiger's head.' b. Xiaoli neng yi-ci chi wu-wan niu-rou mian. p.n. can one-time eat five-bowl beef noodle 'Xiaoli is able to eat five bowls of beef noodle at a time.' (25) Raising modals a. Shu yinggai zhang-jia. book should raise-price 'The prices of the books should be raised.' b. Zhe-ge dianshi jiemu keyi rang xiaohai kan. this-Cl TV program can let children watch 'Children are permitted to watch this TV program.' Typically, the control type of modals have to do with the ability or intention of the subject, and the raising type of modals have to do with the possibility or obligation of/for the 274 proposition.8 Sometimes a particular modal in Mandarin Chinese can assume either of the two uses, a typical example being hui 'will': (26) a. Laowang hui kai che. p.n. will drive car 'Laowang can drive.' b. Shu hui zhang-jia. book will raise-price 'The prices of the books will be raised.' It seems that the crucial difference between the control type and the raising type of modals is the presence vs. absence of a selectional relationship between the subject and the modal verb. Huang (1989) proposes the following analysis for the two types of modals: (27) a. [IP Laowangi [VP hui [IP Proi kai che ]]] p.n. b. will drive car [IP shuj [VP hui [IP tj zhang-jia ]]] book will (Control) (Raising) raise-price It appears that the control type of modals are no different from the regular kinds of control verbs. They take an IP complement, and their subject controls an empty pronominal in the IP complement. What interests us is the raising type of modals. According to Huang (1989), the raising type of modals take an IP as complement, and the subject of the IP raises to the specifier position of the modal projection. But since we have proposed that the subject of a sentence is introduced by a light verb, there is no need for an embedded IP. 8 Thus, the raising type of modals are epistemic and deontic in the conventional sense. The control type of modals can be classified as intensional verbs, on a par with English verbs like want and hope. 275 The projection embedded under the modal can be just a light VP. Thus, we have the following structural analysis for a sentence with a raising modal: (28) a. Laowang hui kai che lai xuexiao. p.n. will drive car come school b. 'Laowang will drive to the shool.' .... VP V' NP Laowang V hui 'will' VP V' t V VP DO NP che 'car' V' V VP kai 'drive' lai xuexiao 'come school' In (28b), the subject Laowang is the specifier of the DO-VP in the underlying representation. From there it raises to the specifier position of hui 'will' VP. (In this structure, the main verb kai 'drive' also moves to DO. We omit this part.) The modal hui 'will' takes the propositional DO-VP as its scope.9 Other kinds of subject, of course, are possible. Consider the following examples: (29) a. Gaosu-gonglu-shang yinggai (zheng) kai-zhe yi-pai tanke-che. Expressway-on should (right-now) drive-Dur one-line tank 'There should be one line of tanks on the expressway (right now).' 9 This analysis is compatible with the VP-Internal Subject Hypothesis (Kitagawa 1986, Kuroda 1988, Koopman and Sportiche 1991, among others), according to which the subject of a sentence is not directly licensed by I', but raises to the specifier position of IP from the specifier position of VP. DO-VP in (28b) corresponds to the VP in the traditional sense. 276 b. Zhe-liang po-che yinggai xiu-le ni bu-shao qian ba! this-Cl broken-car should repair you not-little money Prt 'Repairing this car should have cost you a lot of money.' In (29a), the subject is a locative; in (29b), it is a causer. As we have proposed in Chapter 3, these different subjects are licensed by different light verbs. It is clear that, in (29a-c), the subject raises from the specifier of the subject-selecting VP to the specifier of the modal yinggai 'should', as shown below: (30) Locative ...[VP Causer [ yinggai [ EXIST i V' VP ti CAUSE [VP ... kai ... ]]]] In (30), dependency holds between the subjects and the subject-selecting light verbs. The modal yinggai 'should' has nothing to do with this dependency. It just hosts the raised NP at its specifier position. Thus, the raising type of modals provide a clear example for raising to subject in Mandarin Chinese. 2.3.2 Tense projection There is another instance of raising to subject in Mandarin Chinese. In Lin (2000a) I propose that the tense projection, TP, exists in Mandarin Chinese, and its head is an empty operator, Op, which receives a value from the intra- or extra-sentential context. Mandarin Chinese have been assumed to be a language without morphological tense, and the temporal interpretation of a sentence is largely assumed to be highly discourse-oriented. The following example illustrates this point: 277 (31) Shang-ge xingqi Meiguo zongtong Kelindun fangwen Eguo, last-Mod week U.S. president Clinton visit Russia huijian Ye'erqin zongtong. Liang-wei zongtong zhendui meet Yeltsin president two-Cl president targeting Eguo yu Ou-zhou jingji qingshi jinxing taolun. Russia and Europe economy situation proceed discussion 'Last week President Clinton visited Russia and met with President Yeltsin. The two presidents had a discussion focusing on the economic situations in Russia and Europe.' (31) is a piece of discourse consisting of two sentences. It has the following schematic structure: (32) {discourse .... [S1 last week .... ] [S2 .... ] .... } There is a temporal adverb, shang-ge xingqi 'last week', in sentence S1. Thus S1 is understood as denoting a proposition that holds true in some past time last week. Sentence S2 doesn't have any temporal expression in it. Nonetheless, just like S1, S2 is understood as denoting a proposition that holds true at about the same time as S1, i.e., some past time last week. In other words, the temporal interpretation of S1 is carried over to S2. Thus, from examples like (31), it seems that the temporal interpretation of Chinese sentences is much discourse-oriented. This kind of observation, however, leaves an important question unanswered. That is, in terms of compositional semantics, how can S2 in (2) be interpreted? Lin (2000) assumes with Fukui's (1995) hypothesis on the universality of the TP category in human languages and proposes that the temporal interpretation for sentences of the S2 type sentences proceeds by the aid of an empty tense operator Op. The empty tense operator Op 278 has the following properties: it is the head of the tense projection TP, it has its value fixed via binding, and, once its value is fixed, it denotes the event time of the proposition (cf. Reichenbach 1947, Hornstein 1990). The evidence elicited to support this proposal is the grammatical behaviors of the temporal expression yiqian 'before, in the past' in Mandarin Chinese. The interesting thing about yiqian is that, though it denotes the indefinite past time, its past-time denotation can be conferred to the sentence as the event time only when it occurs in a predicate-external position of the sentence. Consider the following examples: (33) a. Yiqian de Laozhang xihuan gou. (Hold in the past) yiqian Mod p.n. like dog 'Laozhang in the past liked dogs.' b. Laozhang yiqian xihuan gou. (Hold in the past) p.n. yiqian like dog 'In the past Laozhang liked dogs.' c. Laozhang xihuan yiqian de gou. (Hold now) p.n. like yiqian Mod dog 'Laozhang like dogs in the past.' In (33a), yiqian occurs as a modifier of the subject NP, and in (33b), it is a sentential adverbial. In both cases, the proposition is meant to hold true in some past time. In (33c), yiqian occurs as a modifier of the object NP, but here in this case the proposition is supposed to hold true in the present time (that is, the speech time). Thus it appears that only when yiqian occurs in a sentential-level position can it affect the temporal interpretation of the sentence. To account for this effect, I propose that an empty operator Op is at work here. The relevant points are represented in the diagram below: 279 (34) TP Subj [past] yiqian... T' AdvP[past] yiqian T' T Pred Op Obj[past] ...yiqian... Binding under c-command No c-command, no binding From the subject position or the sentential adverbial position, the temporal expression yiqian c-commands Op, and hence binds it, conferring its past-time value to Op. But when yiqian occurs low in the phrase structure, such as in the objet position, it doesn't c-command Op. As such binding doesn't obtain, and there is no way for yiqian to confer its past-time value to Op. Thus, the value-setting of the tense operator Op is in fact structurally determined -- it receives a value only from a c-commanding binder. Further support for this proposal comes from the topicalization of the object. If the value-setting of Op is structurally determined by a c-commanding binder, then, though the occurrence of yiqian in an object NP cannot affect the temporal value of Op, it should be able to do so if the object NP undergoes topicalization to the initial position of the sentence. This prediction is borne out, as the following examples show: 280 (35) a. Wo nainai huo-zhe de-shihou*i/j, my grandma alive-Dur time hen xihuan yiqiani de xiangcun shenghuo. very like yiqian Mod country life 'When my grandma was alive, she liked the country life in the past very much.' b. Yiqiani de xiangcun shenghuo, yiqian Mod country life wo nainai huo-zhe de-shihou i/j hen xihuan t. my grandma alive-Dur time very like 'Country life in the past, when my grandma was alive, she likes it very much.' In (35a), the yiqian-object remains in object position, and the sentence has a temporal interpretation such that yiqian and huo-zhe de-shihou 'the time when she was alive' must be disjoint in time. That is, yiqian doesn't have any effect on the temporal interpretation of the sentence, which is restricted by the 'when'-clause huo-zhe de-shihou 'the time when she was alive'. But in (35b), where the yiqian-object moves up and becomes the topic of the sentence, we find that the past-time denotation of yiqian and the time denotation of the 'when'-clause huo-zhe de-shihou 'the time when she was alive' can now be identical. The discrepancy in interpretation between (35a) and (35b) supports the proposal on Op in a straightforward way. In (35a), since yiqian occurs in object position, it doesn't bind Op and hence cannot affect the temporal interpretation of the sentence. In (35b), yiqian occurs in topic position, from where it c-commands Op, and as a result it binds Op and passes its past-time denotation to Op, giving rise to the identical reading. There are still other interesting questions related to yiqian and Op, which we omit. For detailed discussions, see Lin (2000a). Here we just want to point out an important 281 observation on the predicate-external vs. predicate-internal contrast with respect to yiqian's ability to affect the event time of the proposition. In fact, the predicate-external vs. predicate-internal contrast is not exclusively seen with yiqian. Universal quantification in Mandarin Chinese, in particular quantification involving the universal quantifier dou 'all', exhibits this pattern as well. Let's take the quantificational determiner renhe 'any' as an example. The determiner renhe 'any' belongs to the class of elements in Mandarin Chinese whose occurrence require dou 'all' quantification. Dou 'all' typically occurs between the subject and the predicate, and elements like renhe 'any' must occur at a predicate-external position to be licensed by dou 'all' quantification. The following paradigm shows this point: (36) a. *Renhe reni dou i xihuan chi Yidali cai. *any person all like eat Italian food *'All people like to eat Italian food.' b. *Laowang renhe shihoui dou i xihuan chi Yidali cai. *p.n. any time all like eat Italian food *'Laowang likes to eat Italian food all the time.' c. *Laowang dou i xihuan chi renhe caii. *p.n. all like eat any food *'Laowang likes to eat all kinds of food.' (36a-c) are parallel to (33a-c) precisely: in the former, renhe 'any' must be predicate-external to be grammatical; in the latter, yiqian must be predicate-external to affect the event time of the proposition. Furthermore, if we topicalize the renhe-object, the resulting sentence becomes grammatical, just like (35a-b), where topicalization of an yiqian-object makes it possible for yiqian to affect the event time of the proposition: 282 (37) Renhe caii, Laowang dou i xihuan chi ti. any food p.n. all like eat 'All kinds of food, Laowang likes to eat.' The parallelism between renhe 'any' and yiqian, therefore, is very suggestive: the predicateexternal vs. predicate-internal contrast in the case of renhe 'any' is due to the binding of dou, so the same contrast in the case of yiqian must be due to Op, the corresponding element that enters into a binding relationship with yiqian. In this way, we infer the existence of Op, supported by very strong evidence. If all these are correct, then we have established the existence of the TP projection in Mandarin Chinese. Just like the raising type of modal VPs, TP is a raising category. Since T (or Op) takes propositional scope, it doesn't license an independent specifier. Its specifier is an NP that raises from below in the structure. The relevant points are shown in the following example: (38) a. Laozhang yiqian chou henduo yan. p.n. yiqian smoke many cigarette 'Laozhang smoked a lot of cigarettes in the past.' 283 b. TP NP T' Adv Laozhang yiqiani T' T VP t Opi V' V DO VP NP henduo yan 'many cigarettes' V' V chou 'smoke' In (38b), the light verb projection DO-VP is embedded under TP. The subject of the sentence, Laozhang, is the specifier of the DO-VP underlyingly, and raises to the specifier position of TP. The situation here is exactly the same as that of the raising modals. Raising to subject in Mandarin Chinese can be cyclic. Consider the following example: (39) a. (Ruguo yudao mafan,) (if encounter trouble) Laozhang yiqian hui chou henduo yan. p.n. yiqian will smoke many cigarette '(When there was trouble,) Laozhang would smoke a lot of cigarettes.' 284 b. TP NP T' T' Adv Laozhang T yiqiani Opi VP V' t V hui 'will' VP t V' V DO VP NP henduo yan 'many cigarettes' V' V chou 'smoke' In (39b) an epistemic modal hui 'will' occurs. The subject Laozhang, licensed by DO, raises to the modal VP and then to TP in a cyclic way. To conclude, we have seen two types of raising category in Mandarin Chinese, the (raising type of) modal VP, and TP. They take propositional scope and are more functional than the light verbs. They don't have independent specifiers of their own, but take a specifier raised from below in the structure. Since these raising predicates do not seem to contribute to the shaping of the eventuality in a direct way, they are not regarded as light verbs. Their existence in Mandarin Chinese sentences, however, indicates that functional categories higher than the light verbs are needed in the construction of Mandarin Chinese sentences. Right now, for Mandarin Chinese, the available evidence only points to the existence of the modal VP and TP; there is no evidence that other functional categories that have been frequently referred to, such as the Agr projections, are factually existent in 285 Mandarin Chinese. Thus we will assume that the modal VP and TP are the only higher functional categories in Mandarin Chinese.10 2.4 Summary In this section we saw more light verbs in Mandarin Chinese, in particular those that don't introduce an NP argument into the sentence, including the aspectual markers (the perfective -le, durative -zhe, and experiential -guo) and the extension marker -de. We also had some discussion of the higher functional categories, specifically the raising modals and the tense projection TP, and showed that they are all raising predicates. In the next section we will see how all these elements, light verb and functional categories, participate in the construction of Mandarin Chinese sentences. In particular, The existence of all these elements in the phrase structure leads us to the claim that Mandarin Chinese is a "Davidsonian" language. 3. Mandarin Chinese as a "Davidsonian" language In this section we illustrate the "Davidsonian" character of Mandarin Chinese sentences. Firstly, we have a brief discussion on the nature of the light verbs postulated in this thesis. Then we take an general overview on the construction of phrase structures in Mandarin Chinese. It is shown that a Mandarin Chinese sentence is a constellation of verbs, full or light. Since each verb in the syntactic structure corresponds to a semantic predicate in the logical representation, Mandarin Chinese can be said to be a "Davidsonian" language. 3.1 Why are there light verbs? So far we have postulated a number of light verbs in Mandarin Chinese. They are listed below: 10 We omit discussion on CP, which have been assumed to exist in Mandarin Chinese by syntacticians. See, for example, Tsai's (1994) discussion on C in Mandarin Chinese as an unselective binder. 286 (40) Light verbs in Mandarin Chinese a. b. c. Subject-selecting light verbs DO agentive EXIST existential CAUSE causative PROCEED progressive INCLUDE inclusive OCCUR occurring etc. Object-selecting light verbs USE instrumental AT locational/time FOR reason etc. Light verbs without specifier Aspectual markers aspectual (-le, -zhe, -guo) The extension marker -de descriptive/resultative These light verbs compose sentences in Mandarin Chinese, contribute to the shaping of eventualities, and introduce arguments into the structure. In other words, they are the "building blocks" of Mandarin Chinese sentences. At this point, the reader may wonder about an important question: Why are there light verbs? Here we would like to provide some speculations. First of all, we don't think there is reason to claim that all of these light verbs are universal to all languages. That is, they don't seem to be elements in the universal grammar (UG). Look at the object-selecting light verbs, for instance. They have very specific thematic functions. It is hard to propose that UG would contain things like these as part of 287 human linguistic faculty. But on the other hand, they have a clear stance in the grammar of Mandarin Chinese. We have assumed that these object-selecting light verbs are directly correlated with the serial verb construction and originate from grammaticalization of the corresponding verbs (cf. section 3.2 of Chapter 4). Thus, they are no different from other lexical items. The only particular thing for them is that they lack a phonetic form, and as a result they need some other lexical item to conflate with them. So we assume that they are lexical items in the grammar of Mandarin Chinese, learned by speakers via exposure to the linguistic data in the Mandarin Chinese community. Since their correlation with the serial verb construction is regular, we assume that it is not difficult for the Mandarin Chinese speakers to infer their existence as lexical items in Mandarin Chinese. Thus Mandarin Chinese speakers learn the meanings and functions of the light verbs USE, AT, etc. in the same way that they learn the meanings and functions of the verbs yong 'to use', zai 'at', and so on. But the possibility still remains that some light verbs are universal to human languages and constitute an important portion in the universal lexicon in UG. This possibility seems to be particularly plausible for such frequently referred light verbs as DO, CAUSE, BECOME, etc. But, again, some of the subject-selecting light verbs, e.g. PROCEED and INCLUDE, are very specific in thematic function. We assume that they enter the lexicon of Mandarin Chinese in the same way as the object-selecting light verbs, via inference from the linguistic data available to the speakers. There are still important comparative issues related to the light verbs; for example, the licensing of arguments in a sentence. Throughout the discussion on the light verb syntax in Mandarin Chinese, we have been trying to show that, in Mandarin Chinese, arguments are introduced into the sentence via the functions of light verbs. An immediate consequence from this is that verbs in Mandarin Chinese don't have arguments of their own (cf. section 3.3 of Chapter 4). On the other hand, however, there don't seem to exist such a rich array of light verbs in English. An inference will be that, in English, the introduction of 288 arguments into a sentence crucially depends on the inherent argument structure of the verb. Thus light verbs and argument structures are actually complementarily distributed, the two faces of a coin. From this observation, a parametric theory on the licensing of arguments and the construction of phrase structure can be foreseen. All this can be accounted for by the Lexicalization Parameter proposed in section 5.2 of Chapter 2. We will have more discussions on the related questions in Chapter 7. 3.2 The "Davidsonian" character of Mandarin Chinese With all these in mind, let's go on to examine the construction of the Mandarin Chinese sentence. Essentially, a Mandarin Chinese sentence is constructed via complementation of verbs, full or light, topped with raising functional categories. Below is a diagram on the different verbs in the syntactic structure in Mandarin Chinese: (41) .... VP NP Functional V V' V VP NP Subejct-selecting V V' V VP Aspectual V V' V VP NP Object-selecting V V' V Main V VP (NP) V' V VP Secondary predicate In the structure in (41), from bottom up, we have a secondary predicate, the main verb, the object-selecting light verb, the aspectual verb, the subject-selecting light verb, and the functional verb. The hierarchical organization of the verbs reflects the closeness of the aspects to the core of the eventuality, the main verb. The secondary predicate, e.g. the 289 locative zai 'at' phrase for verbs of placement (cf. section 2.2.3 of Chapter 3) or the descriptive/resultative -de (cf. section 3.2.2 of Chapter and section 2.2 earlier in this chapter), typically denotes the "downstream" (e.g. goal or resultant state) of an eventuality, hence is the complement of the main verb. The theme/patient object of the main verb (the bracketed NP in (41)), being the direct undergoer/experiencer of the action/state denoted by the main verb, is the thematically closest element to the eventuality, hence it occurs as the specifier of the main verb. Other NPs, such as the adverbial objects, are indirect participants in the eventuality, so they are farther and occur above the main VP. The aspectual markers take the predicate in its scope, so it is even higher. The subject-selecting light verb, which introduces the subject that initiates or holds the eventuality (the "upstream"), occurs in the highest position in the light verb structure. Even higher is the raising functional category, the modal VP and TP. In summary, in (41), there appears to be an iconic correspondence between the light verb structure and the eventuality structure -- the lower/higher an element (either an N or V) is in the phrase structure, the closer/farther the relationship it bears with the core of the eventuality.11 Thus, the phrase structure in Mandarin Chinese exhibits a highly thematic character -- to wit, its composing elements map in the composing elements of an eventuality. In this sense, the phrase structure in Mandarin Chinese can be said to be Davidsonian. There is a one-to-one correspondence between the syntactic predicates that compose the phrase structure and the eventuality predicates that compose the eventuality. Let's look at some examples. It's not hard to find sentences in Mandarin Chinese in which most or all the classes of light verbs in (41) occur.12 In those sentences, the formation of 11 It is likely that this is the true source of the thematic hierarchy proposed in the linguistic literature (Gruber 1965, Fillmore 1968, Jackendoff 1972, Grimshaw 1990, among many others). If the proposals in this thesis are correct, the thematic hierarchy can be understood as a reflection of the complementation of different light verbs. Thanks to Mamoru Saito (personal communication) for relevant discussion. 12 There are co-occurrence restrictions among different light verbs, though. For instance, the -de complement and the aspectual marker cannot occur together in the same sentence, as (ia) shows. But this doesn't mean that the secondary predicate and the aspectual markers never meet each other. The post-verbal locative expression headed by zai 'at' is fine with the aspectual markers, as in (ib): 290 the phrase structure involves stacking of VPs, full or light (plus the higher functional categories). Consider the following example: (42) a. Zhe-zhi bi hui xie de wo shouzhi suan-tong. this-Cl pen will write Ext I finger sore-painful 'Using this pen to write made my fingers sore and hurt.' (i) a. b. *Laowang pao--le-de hen lei . *p.n. run-Perf-Ext very tired *'Laowang has run to the extent that he feels tired.' *Laowang fang-le yi-ben shu zai zhuo-shang. *p.n. put-Perf one-Cl book at table-on *'Laowang put a book on the table.' Another instance is the observation that we noticed in section 2 of Chapter 4; that is, the object-selecting light verbs USE and AT seem to require high agentivity of the verb, and hence the occurrence of the light verb DO or CAUSE is mandatory. As a result, the subject-selecting light verb EXIST is not compatible with these object-selecting light verbs, as exemplified in (ii): (ii) ?*Chufang-li qie-le san-ba dao. ?*kitchen-in cut-Perf three-Cl knife ?*'In the kitchen three knives are used to cut [something].' We attribute these co-occurrence restrictions to idiosyncratic semantic properties of specific light verbs, since we haven't found any evidence that shows that a particular class of light verbs cannot co-occur with another class of light verbs. 291 b. CP C' C TP T' NP T VP Op t zhe-zhi bi 'this pen' V' V hui t 'will' VP V' V VP V' CAUSE NP V wo 'I' VP V' DO NP VP V pro V' USE NP V shouzhi 'finger' VP V xie -de 'write' c. CP pro suan-tong 'pro sore-hurt' ∀ <w,t> ∃x [ CAUSE (x,this pen) ∧ DO(x,I) ∧ USE(x,this pen) ∧ Write(x) ∧ RESULT(x,[(My) fingers sore and hurt] ] (42a) is a causative sentence with an instrument causer and a resultative -de complement. (42b) is the complete syntactic structure for (42a). In (42b), we see that the phrase structure is built up via stacking of VPs -- the -de VP, the main VP, USE-VP, DO-VP, CAUSE-VP, and the modal VP. The main verb xie 'write' (suffixed with the extension marker -de) moves up cyclically, firstly to USE, then to DO, and lastly to CAUSE. All these semantic ingredients are conflated into the verb xie 'write' at the output level of the derivation. (42c) is a (simplified) logical representation for (42a), constructed in the event-semantic framework of Parsons (1990). We see that all the relevant eventuality predicates in (42c) have a 292 corresponding V in the syntactic structure in (42b). Thus (42a-c) provide a very clear demonstration of the one-to-one correspondence between the syntactic predicates and the semantic predicates in Mandarin Chinese. Examples like (42a) are abundant in Mandarin Chinese. The light verb structure can be very complex, as in (42a), or fairly simple, where only one or two light verbs are present. The resulting eventualities are correspondingly complex or simple. The reason for this, of course, is that Mandarin Chinese, as a Davidsonian language, uses syntactic means to realize the eventuality structure. All this provides further support to what we proposed in Chapter 2: light verbs are eventuality predicates. They are rich in semantic function, and they play a substantial role in the construction of phrase structure. 3.3 Conclusion Summing up the discussion on the light verb syntax in Mandarin Chinese, an important conclusion recommends itself: Mandarin Chinese is a Davidsonian language. In this language, the syntactic structure is composed of elements that map directly to eventuality predicates in the logical representation. This makes the phrase structure in Mandarin Chinese highly thematic. An important claim ensues: Mandarin Chinese syntax is light verb syntax, and the light verb syntax is motivated by eventuality semantics. This is a remarkable typological feature in the grammar of Mandarin Chinese. 4. Unselectiveness of subject and object in other Chinese dialects So far we have been focusing on the light verb syntax in Mandarin Chinese. However, as we have mentioned in Chapter 1, the grammatical judgments provided in this thesis on Mandarin Chinese are largely drawn from my own intuition as a native speaker of Taiwanese Mandarin, Mandarin dialect. China is well known for the multiplicity of dialects spoken within its territory. Even for Mandarin, there are at least four major dialect groups 293 (see below). These Mandarin dialects exhibit different phonological and syntactic characteristics. There are even more non-Mandarin Chinese dialects in the southern area of China. Each of the southern Chinese dialect is itself a large dialect group. In view of this multiplicity of dialects in China, it is important to examine the syntactic behaviors of other Chinese dialects. Though a complete comparative syntax among Chinese dialects is far beyond the scope of this thesis (in fact such work may require tens or even more theses), in this section, we will discuss some questions related to the light verb syntax in some dialects. We examine some Chinese dialects and examine if they exhibit unselectiveness of subject and object as Taiwanese Mandarin does. It is shown that Chinese dialects differ from each other in the degree that the array of phenomena of unselectiveness of subject and object are exhibited. This diversity, however, is not arbitrary. Some regularities can be discerned. We then go on to a special case in this diversity, Taiwanese, and compare it with Mandarin. It is shown that the differences between the two Chinese dialects in light verb syntax can be characterized in terms of a few principled factors, in particular the choice between verb movement and lexical spell out as the means to realize the light verb. 4.1 Subject and object in Chinese dialects In China, there are thousands of accents that can be labeled as "dialects." In the following we provide a rough portrait for the Chinese dialects. After that we report a survey conducted to examine the syntactic behaviors in several Chinese dialects with respect to the phenomena of unselectiveness of subject and object. 4.1.1 Mandarin and non-Mandarin dialects Chinese is a branch of the Sino-Tibetan family of languages, and is itself a large language sub-family. As Norman (1988) comments, there are few language names in the world that are so "all-encompassing" as Chinese (Norman 1988:1). Innumerable dialects and accents are discerned that impress observers as totally different languages. However, classification 294 of dialect groups is possible. Basically, the Chinese languages can be classified into two major groups: the Mandarin dialects, and the non-Mandarin dialects. Compared to other Chinese dialects, Mandarin is a relatively new language. It is believed that the birth of Mandarin has a strong bearing on the invasion of the Mongolian from the north in the 13th century A.D. Since then, Mandarin has served as the official language of the nation and spread to the whole country from the northern China. According to Norman (1988), there are four sub-groups for Mandarin, each of which constitutes an independent Mandarin dialect: (43) Mandarin dialects (i) Northern Mandarin Spoken in the northern areas of China, including the provinces of Hebei, Henan, Shandong, etc., and the city of Beijing. It is also spoken in Manchuria. (ii) Northwestern Mandarin Spoken in the western and northwestern areas of the central China, including the provinces of Shanxi, Gansu. (iii) Southwestern Mandarin Spoken in the southwestern areas of China, including the provinces of Sichuan and Yunan. (iv) Eastern Mandarin Spoken in the eastern area of China, including parts of the provinces of Jiangsu and Anhui. Lexical items aside, the most salient differences among all these Mandarin dialects are the tones. Differences in other aspects are not very significant. So the Mandarin dialects listed in (43) more or less form a coherent dialect group with only minor intra-dialectal variations. 295 The situation of the non-Mandarin dialects, on the other hand, is dramatically different. First of all, the non-Mandarin dialects do not form a coherent dialect group. As a matter of fact, the non-Mandarin dialects differ from one another with the same (or even more) magnitude as they differ from Mandarin. Furthermore, each non-Mandarin dialect is itself a major dialect group, encompassing a large number of dialects. Some non-Mandarin dialects have a history much longer than Mandarin. For instance, Hakka and the Wu dialects can be dated back to the 4th century A.D. (cf. Chang 1996). So, contrary to the Mandarin dialects, the non-Mandarin dialects exhibit a great deal of diversity in all linguistic aspects, including phonetics, phonology, and syntax. But there is no doubt that all the non-Mandarin dialects are descendants of an early common language. Many factors contributed to the diversity we observe among the nonMandarin dialects -- influences from other Chinese dialects or non-Chinese language, dialect-internal changes, conformity to the literary norms, and so on and so forth (cf. Xu 1991). Whatever the factors are that cause such diversity, all the non-Mandarin dialects are Chinese dialects. Yuan (1961) classifies the non-Mandarin dialects into six major groups: (44) Non-Mandarin Chinese dialect groups (i) Wu dialects Spoken in the provinces of Jiangsu and Zhejiang, including the city of Shanghai. (ii) Min dialects Spoken in the province of Fujian and some areas in the province of Guangdong; also spoken in Taiwan. (iii) Hakka Spoken in the provinces of Guangdong and some inland provinces; also spoken in Taiwan. (iv) Cantonese Spoken in the provinces of Canton and Hainan, and in Hong Kong. 296 (v) Xiang dialects Mainly spoken in the province of Hunan. (vi) Gan dialects Mainly spoken in the province of Jiangxi. All the non-Mandarin dialects listed in (44) are spoken in the southeastern areas of China. Wu, Min, and Cantonese are spoken in the sea-adjacent areas, from north to south. Xiang and Gan are spoken in more inland areas. Hakka is spoken in Taiwan, Guangdong, and forms many islands in some inland provinces. For detailed discussions on these dialects, see Yuan (1961) and Norman (1988). 4.1.2 A survey Our focus in earlier discussions has been on the Mandarin dialect spoken in Taiwan, the Taiwanese Mandarin. However, in other regions of China, there are very different Mandarin and non-Mandarin dialects. In view of this linguistic diversity, it is of great interest to see if other Chinese dialects exhibit the same range of unselectiveness of subject and object as Taiwanese Mandarin does. This concern motivated a survey on the responses from speakers of other Chinese dialects for those sentences exhibiting unselectiveness of subject and object. Below is a report on the survey results. There is a caution before we move on to the details. Due to various limitations on the resources available, this survey is small in scale and only serves a demonstration purpose. It is not meant to be exhaustive and conclusive. As we have mentioned, each major Chinese dialect group encompasses a large number of dialects, so the intent for this survey is just to show the diversity of the responses from speakers of different Chinese dialects. It is tentative and suggestive in nature. A real serious, large-scaled, and welldesigned survey is worth pursuing, but we will leave it for future research. In this survey, several Mandarin and non-Mandarin dialects was chosen as representatives, and a speaker of a particular Chinese dialect was asked to translate a list of sentences into that dialect and judge if the corresponding sentence in that dialect is 297 acceptable or not. This survey involves 11 Mandarin and non-Mandarin Chinese dialects, listed below:13 (45) Group I: Mandarin dialects (i) Bejing Mandarin (Northern Mandarin) (ii) Tianjin Mandarin (Northern Mandarin) (iii) Henan Mandarin (Northern Mandarin) (iv) Shanxi Mandarin (Northwestern Mandarin) (v) Sichuan Mandarin (Southwestern mandarin) Group II: Non-Mandarin dialects (vi) Shanghai dialect (Wu dialect) (vii) Cantonese (Yue dialect) (viii) Fuzhou dialect (Northern Min dialect) (ix) Xiamen dialect (Southern Min dialect) (x) Taiwanese (Southern Min dialect) (xi) Hakka (Hakka) There are 15 examples in the sentence list that the speaker was asked to judge in his/her native dialect. Each of the sentences represents a special feature of unselectiveness of subject or object. The list is given below: 13 This survey was conducted in April-May 2000 in Irvine, California. I am grateful to the following informants for participating in this survey and providing their grammatical judgments: Ge Ren Li Shen Songmei Han Maokai Tsai Xiaolong Yang (Sichuan Mandarin) (Tianjin Mandarin) (Henan Mandarin) (Beijing Mandarin) (Shanxi Mandarin) Bonnie Chen Terence Lam Rong Lin Zhigang Lin Luther Liu (Shanghai dialect) (Contonese) (Fuzhou dialect) (Xiamen dialect) (Taiwanese Hakka) Among the informants, Li Shen and Luther Liu are linguists; all others are not. Most of them are graduate students at UC Irvine, specialized in engineering or social sciences. All the informants only faithfully reported their own judgments for the purpose of this survey. They should not be held responsible for any possible errors and misinterpretations in my analysis. The judgments for Taiwanese are my own. 298 (46) A1. Gaosu-gonglu-shang kai-zhe yi-pai tanke-che. express-way-on drive-Dur one-line tank 'There are one line of tanks on the expressway.' A2. Chufang-li jian-le yi-tiao yu. kitchen-in fry-Perf one-Cl fish 'There is a fish fried in the kitchen.' B1. Zhe-zhi bi xie de wo shouzhi teng. this-Cl pen write Ext I finger hurt 'Writing with this pen made my fingers hurt.' B2. Zhe-tang lu zou-le wo zheng-zheng san tian. this-Cl road walk-Perf I whole three day 'This journey took me three whole days to walk.' C. Zhe-chang sai-che yijing kai-le yi-ban de lucheng le. this-Cl car-racing already drive-Perf one-half Mod distance Prt 'This car racing have proceeded for half of the distance.' D. Zhe-tang renwu yigong kai-le san-liang che, zai-le shi-dun huowu. this-Cl mission altogether drive-Perf three-Cl car carry-Perf ten-ton cargo 'This mission involved ten cars, carrying ten tons of cargo.' E1. qie zhe-ba dao cut this-Cl knife 'use this knife to cut' E2. kan wanyuan-jing watch telescope 'use a telescope to watch' 299 E3. xi leng-shui wash cold water 'use cold water to wash' F1. shui huoche-zhan sleep train-station 'sleep in the train station' F2. chi guanzi eat restaurant 'eat in some restaurant' F3. kai gaosu-gonglu drive express-way 'drive on the expressway' G1. ku nü-penyou cry girlfriend 'cry for the girlfriend' G2. qi xiaohai bu-tinghua angry child not-obedient 'get angry for the child's disobedience' G3. (da pai) da quwei (play card) play fun 'play card games for fun' The specific grammatical features tested in these examples are given below, along with the responsible light verbs: 300 (47) Example Focus Light verb A1-A2: Locative subject EXIST B1-B2: Causative subject CAUSE C: Domain subject PROCEED D: Source subject INCLUDE E1-E3: Instrument object USE F1-F3: Location object AT G1-G3: Reason object FOR These examples are divided into two sub-sets: the 6 examples in A-D exhibit different types of unselectiveness of subject, and the 9 examples in E-G exhibit different types of unselectiveness of object. The result of the survey is shown in the following tables. (In (48a), we list the judgments from Taiwanese Mandarin (TM) as a reference. The judgments are my own.) 301 (48) The survey results a. Mandarin dialects (√: acceptable to the speaker; ×: unacceptable to the speaker) Beijing Tianjin Henan Shanxi Sichuan TM Unselectiveness of subject A1 √ √ √ √ √ √ A2 × × √ √ √ √ B1 √ √ √ √ √ √ B2 √ √ √ √ √ √ C √ × √ √ √ √ D × × √ √ √ √ Unselectiveness of object E1 × × × × × √ E2 √ √ √ × × √ E3 √ √ × √ √ √ F1 √ √ √ √ √ √ F2 √ √ √ √ √ √ F3 √ √ √ √ √ √ G1 √ √ √ √ √ √ G2 × × √ √ √ √ G3 √ √ √ √ √ √ 302 b. Non-Mandarin dialects (√: acceptable to the speaker; ×: unacceptable to the speaker) Shanghai Cantonese Fuzhou Xiamen Taiwanese Hakka Unselectiveness of subject A1 √ × × √ × × A2 √ × × × × × B1 √ √ × √ × × B2 √ × × × × × C √ × × √ √ × D √ × × √ √ × Unselectiveness of object E1 × × × √ √ √ E2 √ √ × √ √ √ E3 √ √ √ √ √ √ F1 √ √ × √ √ √ F2 × × × √ √ √ F3 √ √ × √ √ √ G1 √ × × √ √ √ G2 √ √ × √ √ √ G3 √ √ √ √ √ √ 303 In (48a-b), we have quite diverse responses to the sentences in (46A-G) with respect to their acceptability. For instance, there are dialects that resist the relevant examples to a great extent, such as Cantonese (Yue) and Fuzhou dialect (Min), but there are also dialects that accept most or all the examples, such as Henan Mandarin, Shanxi Mandarin, Sichuan Mandarin, Shanghai dialect (Wu), and Xiamen dialect (Min). The occurrences of ×, which stands for unacceptability, range from 2 to 13. At the first sight, the general situation seems to be random. However, a few interesting distributions emerge as we examine the responses closely, and some of them turn out to be regularities. We will look at two such distributions below. First of all, area resemblance appears to be a significant distribution in the overall responses in (48a-b). On this point, see the resemblance of Beijing Mandarin and Tianjin Mandarin. Both Mandarin dialects belong to the northern Mandarin dialect group, and they are spoken in adjacent areas in the northern area of the province of Hebei, in the city of Beijing and the city of Tianjin. The distributions of the acceptable and unacceptable examples are almost identical -- both have A2, D, E1, and G2 unacceptable. (The Tianjin Mandarin has one more unacceptable example, C.) Thus the northern Mandarin appears to be more restrictive than Taiwanese Mandarin (and the northwestern and southwestern Mandarin dialects; see below) with respect to the unselectiveness of subject and object in sentences. The northern Mandarin has been the dialect that the Chinese governments (the Nationalist government in Nanjing before 1949, and the Communist government in Beijing after 1949) officially posit as the model language for the lingua franca of the nation (putong hua 'common language' in China, and guoyu 'national language' in Taiwan) (cf. Norman 1988). However, comparing Beijing Mandarin and Tianjin Mandarin (both being northern Mandarin dialects) and Taiwanese Mandarin (which has been under great influences from Taiwanese), we find that Mandarin dialects can differ from one another in significant ways. Next let's look at the other three Mandarin dialects in (48a), Henan Mandarin, Shanxi 304 Mandarin, and Sichuan Mandarin. It is interesting to note that these three Mandarin dialects are very similar in the responses -- all the three Mandarin dialects share a × for E1, though Henan Mandarin has another one for E3, and Shanxi Mandarin and Sichuan Mandarin also have one for E2. It seems that, as Mandarin spreads away from its original region (the vicinity surrounding the city of Beijing), its restrictiveness on the unselectiveness of subject and object gets lessened. Shanxi Mandarin and Sichuan Mandarin represent the northwestern Mandarin and southwestern Mandarin respectively. Henan Mandarin, though a dialect in the northern Mandarin group, is located at the far south of the northernMandarin-speaking area, adjacent to the province of Hubei, where the northwestern Mandarin is spoken. All these three Mandarin dialects exhibit much more lenient restrictions on the selectiveness of subject and object. It thus seems that there is a lessening of restrictiveness in a radiational fashion for the Mandarin dialects, from the center to the periphery. Another instance of the area resemblance is observed among Xiamen dialect, Taiwanese, and (Taiwanese) Hakka. An interesting phenomenon to notice is that most Chinese dialects in the survey don't accept instrument objects of "cut this knife" type, namely, E1. But Xiamen dialect, Taiwanese, and (Taiwanese) Hakka accept it. Xiamen dialect and Taiwanese both belong to the southern Min dialect group, opposed to Fuzhou dialect, which is a member of the northern Min dialect group. Since Fuzhou dialect doesn't accept E1 (in fact, it is the most restrictive dialect in the survey), instrument objects of E1 type appear to be a special feature in the southern Min dialects. As to Hakka, the Hakka dialect adopted into the survey is Taiwanese Hakka, which may have been influenced by Taiwanese and as a result comes to adopt this special feature into its syntax. (In fact, it seems reasonable to suppose that the acceptability of E1 in Taiwanese Mandarin arises from the influence from Taiwanese too.) So Xiamen dialect, Taiwanese, and (Taiwanese) Hakka exemplify another instance of area resemblance, with respect to the instrument objects of E1 type. 305 Another significant distribution in the responses in (48a-b) is a north-south discrepancy with respect to the unselectiveness of subject. The general picture we get from (48a-b) is that, the northern Chinese dialects, including the Mandarin dialects and Shanghai dialect,14 exhibit quite extensive unselectiveness of subject (with Beijing Mandarin and Tianjin Mandarin a little more restrictive). On the other hand, the southern Chinese dialects, including Cantonese, Fuzhou dialect, Xiamen dialect, Taiwanese, and Hakka, exhibit very strong resistance to unselectiveness of subject. (But Xiamen dialect seems to be freer than other southern dialects in this respect.) This discrepancy seems to be a typological feature that distinguishes the northern and southern Chinese dialects. While there can be more than one factor that contributes to this discrepancy, in the following we will show that grammatical accounts are possible, at least for some of its aspects. In next subsection, we will focus on the case of Taiwanese, and propose that an important syntactic factor is involved in the difference between the northern dialects, such as Mandarin, and the southern dialects, such as Taiwanese -- how high the main verb may move to in the light verb structure. 4.2 A case study: Taiwanese In this subsection, we look at some syntactic constructions in Taiwanese, and show that, though in appearance Taiwanese differs from Mandarin Chinese with respect to the unselectiveness of subject, this divergence actually can be retrieved to a principled factor: in Taiwanese, the main verb doesn't move to the subject-selecting light verb, but in Mandarin it does. Typically, the subject-selecting light verb in Taiwanese is lexically spelled out, and hence there is no need for the main verb to move to it. Thus one of the major syntactic distinctions between Taiwanese and Mandarin is the way that the subject-selecting light verb manifests itself, lexical spell-out or Attract-Move. 14 Shanghai dialect resembles the Mandarin dialects in many interesting aspects. The unselectiveness of subject is just an instance. See Lin (1998) for another instance, related to the Wh/PQ interaction and its correlation with verb movement in various Chinese dialects. 306 4.2.1 Subject-selecting lights verb in Taiwanese From the survey in (48b), we find that Taiwanese resists unselectiveness of subject to quite an extent. In particular, it doesn't permit the locative and causer subjects.15 The relevant examples are repeated below: (46) A1. Gaosu-gonglu-shang kai-zhe yi-pai tanke-che. express-way-on drive-Dur one-line tank 'There are one line of tanks on the expressway.' A2. Chufang-li jian-le yi-tiao yu. kitchen-in fry-Perf one-Cl fish 'There is a fish fried in the kitchen.' B1. Zhe-zhi bi xie de wo shouzhi teng. this-Cl pen write Ext I finger hurt 'Writing with this pen made my fingers hurt.' B2. Zhe-tang lu zou-le wo zheng-zheng san tian. this-Cl road walk-Perf I whole three day 'This journey took me three whole days to walk.' Though the four examples in (46) are all grammatical in (Taiwanese) Mandarin, they are unacceptable in Taiwanese. Below are the Taiwanese counterparts of (46) (A1-A2), (B1B2):16 15 To my ear, example (C) and (D) in (46) sound more or less acceptable. But in view of the claim that the main verb in Taiwanese typically doesn't move to the subject-selecting light verb, the acceptability of (C) and (D) in (46) will be counterexamples. Right now I don't have an explanation for this problem. 16 All the grammatical judgments on the Taiwanese examples in the following discussion are my own. A clarification is in order here. I have talked with some other native speakers of Taiwanese, and came to realize that different speakers may have different sets of judgments for the examples that I discuss in the text. But such divergence is not surprising. After all, even Taiwanese has its own dialectal accents. Taiwanese itself originates from a mixture of linguistic ingredients from other Chinese dialects, notably from Zhangzhou dialect and Chuanzhou dialect of the southern Min dialect group. Furthermore, rapid and dramatic social changes (e.g. mass immigration from the mainland China after 1949 along with retreat of 307 (49) a. *Kesok-konglo-tingthao le sai tsintsue chia. *express-way-on right-now drive many car *'There are many cars on the expressway.' b. *Tsaokha-laibin tsen tsit-be hi. *kitchen-in fry one-Cl fish *'There is a fish fried in the kitchen.' c. *Jit-ki pit xia ka gua chiuji tïä. *this-Cl pen write Ext I finger hurt *'Writing with this pen made my fingers hurt.' d. *Jit-tsua lo kïä wo jok-jok sä zit. *this-Cl road walk I whole three day *'This journey took me three whole days to walk.' There are imaginably many possible accounts for such discrepancy between Mandarin and Taiwanese, but Cheng et al. (1997) point out an important observation on the light verbs in Taiwanese: in a Mandarin sentence the verb typically moves up to the (subject-selecting) light verb, but in Taiwanese the light verb usually manifests itself overtly (also see L. Liu 1997). Cheng et al. (1997) support this claim with evidence from the possessive agent construction in Mandarin and Taiwanese. Recall from section 4.2.3 of Chapter 2 that Huang (1997) analyzes the possessive agent construction in Mandarin as involving head movement of the main verb to the light verb DO. The following examples demonstrate this analysis: the Nationalist government to Taiwan, which resulted in mutual assimilation between the native southern Min dialect and Mandarin) have imputed a great number of new elements into Taiwanese. So it is natural to expect that the Taiwanese speakers respond to the Taiwanese sentences discussed here in different ways, since Taiwanese is actually a language under rapid change. For relevant discussion on this issue, see Cheng (1985). 308 (50) a. Ni pa ni-de shan ba. you climb your mountain Prt 'You [go ahead and] climb the mountain.' b. [IP you [VP DO [ IP your [VP climb the mountain ] ] ] ] In Taiwanese, however, a word-to-word translation of (50a) will result in an ungrammatical sentence, as shown in (51a). To express the same proposition, the verb tso 'do' must be inserted into the structure, as in (51b). The analysis for (51b), along Huang's (1997) proposal, is given in (51c) (cf. Cheng et al 1997, L. Liu 1997). (51) a. *Li pe li-e süä. *you climb your mountain *'You [go ahead and] climb the mountain.' b. *Li tso li pe süä. *you do you climb mountain *'You [go ahead and] climb the mountain.' c. [IP you [VP [DO tso ] [IP you [VP climb the mountain ] ] ] ] A comparison of the Taiwanese case in (51c) and the Mandarin case in (50b) indicates that, the crucial distinction between these two Chinese dialects in the possessive agent construction is the way that the subject-selecting light verb is realized. In the Mandarin case in (50b), the subject-selecting light verb DO is phonetically empty, and hence requires a "phonetic support" from the main verb. As such DO attracts the main verb pa 'climb' to incorporate to it. In the Taiwanese case in (51c), the subject-selecting light verb DO is 309 simply spelled out as a lexical word tso 'do'.17 Since an independent phonetic form is available, DO in Taiwanese doesn't require any "phonetic support". Thus, general principles of economy on derivation forbid head movement of the main verb pe 'climb' to tso/DO.18 17 Here are some examples for the verb tso 'do' in Taiwanese showing its status as a full-fledged verb: (i) a. b. Gua le tso khangkhue, mai tsa. I Prog do work don't bother 'I am working; don't bother me.' Yi tso daiji jin pintüä. he do things very lazy 'He is lazy in work.' 18 There is a caveat for this point. Though we state that tso 'do' is the lexical realization of the light verb DO, actually it is not the case that all occurrences of DO in Taiwanese are realized as tso 'do'. For one thing, simple active sentences such as John hit Bill cannot be expressed as in (ia). It should be like (ib), on a par with English and Mandarin: (i) a. b. *Ong-e tso pha Tïü-e. *p.n. do hit p.n. *'Ong-e hit Tïü-e.' *Ong-e pha Tïü-e. *p.n. hit p.n. *'Ong-e hit Tïü-e.' (ib) involves head movement of the main verb pha 'hit' to the higher light verb DO. Thus lexical spell-out of the subject-selecting light verb is just a significant pattern in Taiwanese, not a law-like regularity. As a matter of fact, the tso 'do' construction in Taiwanese (also the possessive agent construction in Mandarin) has a special connotation that other DO sentences (such as (ib)) don't -- it has a sense of "at your will," represented in the gloss by the expression "go ahead" in (51b). Thus the use of tso as a lexical realization of DO can be marked. The same point is seen with the light verbs EXIST and CAUSE, to be discussed shortly. Though we will see that these two light verbs typically get lexical spell-out in Taiwanese sentences, there are instances where the main verb moves to them. An example is the verb of placement kheng 'put'. Just like its correspondent in the English verb put and the Mandarin verb fang 'put', the verb kheng 'put' may assume causative/accomplishment use, as in (iia). In (iia), however, CAUSE is not lexically realized. Similar to the case of Mandarin, the verb kheng 'put' in Taiwanese can take a locative subject, as in (iib). But again the light verb EXIST is not lexically realized, which indicates that head movement of the main verb to EXIST has occurred. (ii) a. b. Ong-e kheng jit-pun tse ti to-ting. p.n. put one-Cl book at table-on 'Ong-e put a book on the table.' To-ting kheng jit-pun tse. table-on put one-Cl book 'There is a book on the table.' However, markedness seems to play a role in the cases of CAUSE and EXIST, too. It seems that, in Taiwanese, when a verb is used in a way deviated from its typical use, the subject-selecting light verb must be lexically spelled out. Of course, many questions remain, and right now I don't have explanations for all of them. We will just leave them open. 310 It thus seems that, the distinction between Mandarin and Taiwanese with respect to the unselectiveness of subject can be reduced to the individual factor of the realization of the subject-selecting light verb DO, in the case of the possessive agent construction. In what follows, we will show that this distinction is actually a general pattern in Mandarin and Taiwanese, seen in many syntactic constructions. The opposition between lexical spell-out and Attract-Move appears to be a major typological parameter that distinguishes between Taiwanese and Mandarin. 4.2.2 Locative and causer subjects in Taiwanese Now we look at the locative and causer subjects in Taiwanese. The relevant examples are repeated below: (49) a. *Kesok-konglo-tingthao le sai tsintsue chia. (Locative subject) *express-way-on right-now drive many car *'There are many cars on the expressway.' b. *Tsaokha-laibin tsen tsit-be hi. (Locative subject) *kitchen-in fry one-Cl fish *'There is a fish fried in the kitchen.' c. *Jit-ki pit xia ka gua chiuji tïä. (Causer subject) *this-Cl pen write Ext I finger hurt *'Writing with this pen made my fingers hurt.' d. *Jit-tsua lo kïä wo jok-jok sä zit. (Causer subject) *this-Cl road walk I whole three day *'This journey took me three whole days to walk.' (49a-d) are (near) word-by-word Taiwanese translations of the corresponding Mandarin sentences in (46) (A1-A2), (B1-B2). They are all unacceptable. However, exactly like the 311 possessive agent construction, we can lexically spell out the subject-selecting light verbs, EXIST and CAUSE, and make them acceptable. Look at the following examples: (52) a. Kesok-konglo-tingthao wu le sai tsintsue chia. express-way-on have right-now drive many car 'There are many cars on the expressway.' b. Tsaokha-laibin wu tsen tsit-be hi. kitchen-in have fry one-Cl fish 'There is a fish fried in the kitchen.' c. Jit-ki pit ka gua xia ka chiuji tïä. this-Cl pen Disp I write Ext finger hurt 'Writing with this pen made my fingers hurt.' d. Jit-tsua lo ho gua kïä jok-jok sä zit. this-Cl road cause I walk whole three day 'This journey took me three whole days to walk.' In (52a-b), the existential verb wu 'have' occurs in the structure. The sentences become acceptable, with existential connotation. In (52c-d), we have the disposal marker and causative morphemes ka and ho in the structure, and, again, the sentences become acceptable, with causative connotation. The pattern that we observe here is exactly the same as in the possessive agent construction. Thus, the cases of the locative and causative subjects once again exemplify the distinction between lexical spell-out vs. Attract-Move in these two Chinese dialects. The following diagram shows the relevant points:19 19 The case of Taiwanese, as represented by the examples in (52a-d), actually poses a problem for one of our earlier analyses. Notice that these examples have the surface structure in (i) below: (i) [ Subject { EXIST CAUSE } [ V Obj ]] 312 .... VP (53) NP V' V VP Taiwanese / Mandarin tso DO wu / EXIST ka / ho CAUSE ....V.... (53) shows the two different ways that the subject-selecting light verbs are realized in Mandarin and Taiwanese. In Mandarin, the subject-selecting light verbs DO, EXIST, CAUSE don't have independent phonetic forms of their own, hence they attract the main verb V from below to them. In Taiwanese, the light verbs are lexically spelled out. Lacking an attracting force, the main verb V cannot incorporate to the light verbs. Thus, lexical spellout preempts head movement, due to economy effects (cf. Tsai's 1994 principle of Lexical Courtesy). This account explains the asymmetry that we observe in the survey in (48a-b) on the general resistance to unselectiveness of subject in Taiwanese. The morphological properties (i.e. whether an independent phonetic form is available) of the subject-selecting We claim that in Taiwanese the subject-selecting light verb (EXIST or CAUSE) is lexically spelled out, and the main verb V doesn't move up to it. But remember from the analysis in Chapter 2-3 that the object of a transitive verb is merged to the structure as the specifier of the main VP. In other words, the structure should be as in (ii), which differs from (i) in the order or V and Obj: (ii) [ Subject { EXIST CAUSE } [ Obj V ]] According to the analysis in Chapter 2-3, it is the incorporation of V to the subject-selecting light verb that makes Obj post-verbal. Now the problem is: if no verb movement occurs in the case of Taiwanese in (i), the word order should be [ Obj V ], rather than [ V Obj ]. How can this problem be solved? At this point we don't have a solution for this problem. We could posit an intermediate light verb between the main verb and the subject-selecting light verb, and the main verb moves to it, leaving the object behind. But such an intermediate light verb requires justification. In view of the analytic character of the Taiwanese phrase structure (see section 4.2.4 below), the existence of such intermediate light verb might not be implausible. 313 light verbs in the two Chinese dialects determine the applicability of the head movement of the main verb, leading to the two different patterns in the surface syntax of the two Chinese dialects. If our analysis is correct, lack of the Mandarin type of unselectiveness of subject in Taiwanese doesn't entail absence of the relevant light verbs in the lexical inventory. As shown in (52a-d), subject-selecting light verbs do exist in Taiwanese. They require a different means to get realized. This analysis may carry over to other Chinese dialects similar to Taiwanese, including Fuzhou dialect, Cantonese, Xiamen dialect, and Hakka.20 In other words, the light verb structure remains the same across all Chinese dialects, and their differences arise from the morphological properties of light verbs and the way that the light verbs are realized in the syntactic structure. 4.2.3 The "source" double object structures Another construction that shows the same contrast between Mandarin and Taiwanese is the "source" double object structure. Mandarin Chinese has a special type of double object construction that involves "inward" transaction of the theme object (cf. Tang 1979, Li 1990, Cheng et al 1999). This construction usually takes verbs such as fa 'fine', pian 'cheat', qiang 'rob', etc., as the main verb. Some examples are given below: 20 Relevant facts in Hakka indicate that the same analysis applies to this dialect. In Hakka, unselectiveness of subject of the Mandarin type is not permitted, as shown in (ia). However, if we insert the existential verb ziu 'have' into it, the structure becomes grammatical, as in (ib): (i) a. b. *Thailu-hong tho si thongto chia. *road-on Dur drive many car *'There are many cars on the road.' ?Thailu-hong ziu tho si thongto chia. ?road-on have Dur drive many car ?'There are many cars on the road.' Thanks to Luther Liu for discussion on Hakka. 314 (54) a. Laowang qiang-le Taiwan Yinhang ba-baiwan. p.n. rob-Perf Taiwan Bank eight-million 'Laowang robbed eight million dollars from Taiwan Bank' b. Jingcha fa-le Xiaoli san-qian kuai de fa-jin. police fine-Perf p.n. three-thousand dollar Mod fine 'The police fined Xiaoli three thousand dollars.' The source double object construction, as exemplified in (54a-b), has the structure [NP1-VNP 2-NP 3], in which NP1 is the goal of the transaction, NP2 the source of the transaction, and NP3 the theme transacted. This thematic structure is reversed to the typical "outward" kind of double object construction, in which NP1 and NP2 are the source and goal of the transaction respectively. Sometimes the source double object construction can take a regular activity verb, such as chi 'eat' or du 'read', as the main verb, as in the following examples: (55) a. Laozhang chi-le Xiaoli san-zhi ji. p.n. eat-Perf p.n. three-Cl chicken 'Laozhang ate three chickens from Xiaoli.' b. Laowang du-le Xiaozhang si-ben shu. p.n. read-Perf p.n. four-Cl book 'Laowang read four books from Xiaozhang.' The thematic structure in (55a-b) is the same as that in (54a-b): NP1 is the goal of the transaction that consumes or affects the theme NP3 from the source NP2. An even more interesting example for the source double object construction is the kind of verbs like jie 'lend'. It has been observed that a double object structure with jie 'lend' 315 as the main verb is ambiguous, as it allows both the goal and the source readings (Tang 1979). See the following example: (56) Laowang jie-le Xiaoli san-bai kuai. p.n. lend-Perf p.n. three-hundred dollar 'Laowang loaned three hundred dollars to Xiaoli.' (Goal reading) or 'Laowang borrowed three hundred dollars from Xiaoli.' (Source reading) (56) has two possible readings: in one of them, the subject Laowang is the source of the lending transaction, and in the other, it is the goal. To disambiguate these two readings, the speakers have to resort to the discourse contexts. There have been analyses proposed to account for the source double object construction in Mandarin Chinese. Here we will adopt the one developed by Cheng et al (1999), according to which the source double object structure involves a complex predicate and the head movement of the main verb. Before getting to the analysis, we turn to the case of Taiwanese first, which is very different from the case of Mandarin. We observe that the source double object structure is generally ungrammatical in Taiwanese.21 That is, word-to-word translations of (54a-b) and (55a-b) into Taiwanese result in ungrammatical sentences. Furthermore, the Taiwanese counterpart of (56) only has the goal reading, not the source reading. The relevant examples are given below: (57) a. *Ong-e qïü Taiwan Ginhang pe-paban. *p.n. rob Taiwan Bank eight-million *'Ong-e robbed eight million dollars from Taiwan Bank.' 21 Some researchers, such as Tsao (1988), Cheng and Tsao (1995), and Cheng et al. (1997), take certain sentences that we judge ungrammatical below in the text as grammatical. As I have emphasized at several places, the grammatical judgments in this thesis only represent my own intuitions. The diversity in grammatical judgments must not be conceived as implicating any objection to other researchers' analyses. 316 b. *Kingtsat huat Li-e sä-pa kho e huat-kim. *police fine p.n. three-hundred dollar Mod fine *'The police fined Li-e three hundred dollars.' (58) a. *Tïü-e chia Li-e sä-jia kue. *p.n. eat p.n. three-Cl chicken *'Tïü-e ate three chickens from Li-e.' b. *Ong-e thak Tïü-e si-pun tse. *p.n. read p.n. four-Cl book *'Ong-e read four books from Tïü-e.' (59) Ong-e jio Li-e sä-pa kho. (Goal reading only) p.n. lend p.n. three-hundred dollar (i) *'Ong-e loaned three hundred dollars to Li-e.' (ii) *'Laowang borrowed three hundred dollars from Xiaoli.' To express the same propositions, the disposal marker ka must be used in the Taiwanese sentences, as below: (60) a. Ong-e ka Taiwan Ginhang qïü pe-paban. p.n. Disp Taiwan Bank rob eight-million 'Ong-e robbed eight million dollars from Taiwan Bank.' b. Kingtsat ka Li-e huat sä-qieng kho tso huat-kim.22 police Disp p.n. fine three-thousand dollar do fine 'The police fined Li-e three thousand dollars.' 22 Instead of using the modification marker e (which is equivalent to the Mandarin de; cf. (54b)), Taiwanese uses the verb tso 'do' here. 317 (61) a. Tïü-e ka Li-e chia sä-jia kue (khi).23 p.n. Disp p.n. eat three-Cl chicken (go) 'Tïü-e ate three chickens from Li-e.' b. Ong-e ka Tïü-e thak si-pun tse (khi). p.n. Disp p.n. read four-Cl book (go) 'Ong-e read four books from Tïü-e.' (62) Ong-e ka Li-e jio sä-pa kho. (Source reading) p.n. Disp p.n. lend three-hundred dollar 'Laowang borrowed three hundred dollars from Xiaoli.' The account for this contrast depends on the analysis of the morpheme ka in Taiwanese. We consider the morpheme ka a disposal marker (cf. section 2.2.2 above in this chapter for the discussion on the disposal marker ba in Mandarin). But the semantics of the morpheme ka seems to be more complex than the Mandarin ba. Cheng and Tsao (1995) have shown that the morpheme ka in Taiwanese can represent at least four distinct thematic roles: source, theme/patient, benefactive, and goal. Below are examples illustrating these four uses of ka (adapted form (1), (4), (5), and (7), Cheng and Tsao 1995: 23): (63) a. Kingtsat ka gua huat lak-pa kho. (Source) police Disp I fine six-hundred dollar 'The police fined me six hundred dollars.' b. Yi ka tsu be khi a. (Patient/theme) he Disp house sell go Prt 'He sold out the house.' 23 For the directional verb khi 'go' in Taiwanese, see below. 318 c. Yi ka gua ching sä. (Benefactive) he Disp I wear cloth 'He [help] me put on the cloth.' d. Yi ka gua kïä le. (Goal) he Disp I perform courtesy 'He performed courtesy to me.' As shown in (63a-d), the NP that ka introduces can be thematically interpreted as a source, a patient/theme, a benefactive, or a goal. Thus the thematic functions of ka seem to be allencompassing. But we suggest that further reduction of ka's thematic functions is possible. The common semantic feature that runs across all the examples in (63a-d) is affectedness -the police affected me by putting a fine on me ((63a)), he affected the house by the action of selling ((63b)), he affected me with putting cloth on me ((63c)), and he affected me by performing courtesy ((63d)). We therefore propose that the semantics of the morpheme ka is simply "to affect," and all the different thematic roles of the NPs that ka introduces arise from the semantics/pragmatics of the predicate as a whole. So we propose to take ka in Taiwanese as a light verb standing for affectedness. Furthermore, we assume that there is a corresponding light verb in Mandarin too. Let's call it AFFECT. Now we return to the complex predicate analysis of the source double object construction proposed by Cheng et al (1999). Though Cheng et al (1999) intend this analysis to be for Taiwanese, it can be readily carried over to Mandarin Chinese. The central point in this analysis is that, the source double object structure is actually not a real double object construction. It involves a light verb which takes a complex VP predicate with an internal subject. With the light verb AFFECT, the analysis is as follows: 319 (64) a. Laowang qiang-le Taiwan Yinhang ba-baiwan. p.n. rob-Perf Taiwan Bank eight-million 'Laowang robbed eight million dollars from Taiwan Bank.' .... VP b. NP V' Laowang V AFFECT VP NP V' Taiwan Yinhang 'Taiwan Bank' V NP qiang 'rob' ba-baiwan '8 millions' (64a) has the structure and derivation represented in (64b). In (64b), the subject Laowang is licensed by the light verb AFFECT, which takes a VP as complement. This VP has a V' complex predicate qiang ba-baiwan 'rob eight millions' and an internal subject Taiwan Yinhang 'Taiwan Bank'. Now the same story applies: the main verb qiang 'rob' incorporates to the light verb AFFECT, yielding the surface structure in (64a). Thus, according to this analysis, the source double object structure is just another instance of complex predicate formation and head movement of the main verb in Mandarin Chinese. Taiwanese, on the other hand, exhibits a quite different pattern, as we have seen throughout this section. In a nutshell, Taiwanese lexically spells out the light verb AFFECT as the morpheme ka, rather than moving the main verb to it. An example is given below: 320 (65) a. Ong-e ka Taiwan Ginhang qïü pe-paban. p.n. Disp Taiwan Bank rob eight-million 'Ong-e robbed eight million dollars from Taiwan Bank.' .... VP b. NP V' V VP Ong-e ka AFFECT V' NP V Taiwan Ginhang qïü 'Taiwan Bank' 'rob' NP pe-paban '8 millions' The underlying syntactic representation for the Taiwanese example in (65a) is the same as the Mandarin example in (64a) -- there is a complex predicate that licenses an internal subject, and the light verb AFFECT takes the resulting VP as complement. The crucial difference, as is the case in all other constructions that we have seen in this section, is that, rather than attracting the main verb qïü 'rob' to move to it, AFFECT gets lexically spelled out as ka. So, ka is the lexical version of the light verb AFFECT. Lexical spell-out preempts Attract-Move, therefore head movement of the main verb becomes ungrammatical. This analysis accounts for the Mandarin examples in (54-56) and the Taiwanese examples in (57-59). 4.2.4 Taiwanese is more analytic than Mandarin In view of the contrast between Mandarin and Taiwanese illustrated in the preceding subsection, we believe that the two different patterns for the realization of the subject- 321 selecting light verb, namely, Attract-Move vs. lexical spell-out, are firmly established. This explains the discrepancy between Mandarin Chinese and Taiwanese with respect to unselectiveness of subject. Notice that the discrepancy between Mandarin and Taiwanese does not entail that Taiwanese has a distinct light verb syntax, or even lacks subject-selecting light verbs entirely. On the contrary, the light verb syntax remains essentially the same. All the subject-selecting light verbs that occur in Mandarin occur in Taiwanese too. Since Taiwanese makes use of the strategy of lexical spell-out, Taiwanese appears to be more "analytic" than Mandarin with respect to the syntactic realization of the eventuality predicates. This is so because, in Taiwanese, conflation doesn't apply to fuse the subjectselecting light verb and the main verb into an individual element in surface syntax (Cheng et al 1997). Thus, the discrepancy between Mandarin and Taiwanese in the survey in (48a-b) is accounted for in a principled way. This approach, conceivably, can be carried over to other Chinese dialects. Though there are still other factors to be considered, we believe we have demonstrated a viable approach to certain important aspects in the comparative dialectology of the Chinese languages. The magnitude of the relevant questions prevents us from delving into substantial discussion, though. We will leave them for future research. Before closing this section, we would like to present more evidence for the general analytic character of Taiwanese. There are in fact two distinct facets in the pattern that Taiwanese adheres for the realization the subject-selecting light verb -- lexical spell-out of the subject-selecting light verb, and the prohibition of head movement of the main verb. We have seen evidence in Taiwanese that illustrates the first facet. We will see that the second facet is also independently motivated in Taiwanese. Here we will look at the transitivity alternation in Taiwanese. Remember from section 3.1 of Chapter 2 that transitivity alternation in Mandarin Chinese is syntactically transparent. Here we repeat the examples. First of all, (66a-d) show that the verb po 'break' in Mandarin cannot be readily used as a causative verb: 322 (66) a. *Chuangzi po le. *window break Prt *'The window broke.' b. *Laozhang po chuangzi. p.n. break window 'Laozhang broke the window.' c. *Mutou po chuangzi. wood break window 'The stone broke the window' d. *Taifong po chuangzi. typhoon break window 'The typhoon broke the window.' To make po 'break' causative, an additional VP layer must be merged to the structure specifying the causative-agentive component of the eventuality. The verb, po 'break' can incorporate to the higher verb, as in (67a-c), or stay in the lower VP heading an independent phrasal category, as in (68a-c). (67) a. Laozhang da-po chuangzi. p.n. hit-break window 'Laozhang broke the window.' b. Mutou zhuang-po chuangzi. wood strike-break window 'The wood broke the window.' 323 c. Taifong chui-po chuangzi. typhoon blow-break window 'The typhoon broke the window.' (68) a. Laozhang da-de chuangzi po le. p.n. hit-Ext window break Prt 'Laowang hit [the window such that] the window broke.' b. Mutou zhuang-de chuangzi po le. wood strike-Ext window break Prt 'The wood stroke [the window such that] the window broke.' c. Taifong chui-de chuangzi po le. typhoon blow-Ext window break Prt 'The typhoon blow [on the window such that] the window broke.' The diagram in (69) illustrates the relevant point the analyses of (64) and (65): (69) .... VP V' V hit strike blow ... VP NP the window Causative-agentive V' Inchoative-resultative V break Optional Incorporation Now we turn to the transitivity alternation in Taiwanese. First of all, just like Mandarin, the unaccusative verb in Taiwanese can only assume unaccusative use; it cannot be used as a 324 causative verb directly. Below are examples with the verb phua 'break' in Taiwanese (we will come back to the directional verb khi 'go' in (70a) and other examples to come): (70) a. Thang-a phua-khi a. window break-go Prt 'The window broke.' b. *Ong-e phua thang-a. *p.n. break window *'Ong-e broke the window.' c. *Tsathau phua thang-a. *wood break window *'The wood broke the window.' d. *Hongthai phua thang-a. *typhoon break window *'The typhoon broke the window.' Like Mandarin, an additional VP layer must be merged to the structure of phua 'break' to be causative. However, there is a divergence: in Taiwanese, the merged VP, which specifies the causative-agentive component of the eventuality, cannot be compounded with the unaccusative verb; rather, the two verbs must be kept separate, heading their own VP projections. Thus, (71a-c), all of which are examples of compound verbs, are ungrammatical:24 (71) a. *Ong-e khok-phua thang-a. *p.n. break window *'Ong-e broke the window.' 24 The ungrammaticality of (71a-c) actually conforms to a more general phenomenon in Taiwanese: Taiwanese lacks resultative verbal compounds. See T. Mei (1991) for discussion. 325 b. *Tsathau kong-phua thang-a. *wood break window *'The wood broke the window.' c. *Hongthai tsue-phua thang-a. *typhoon break window *'The typhoon broke the window.' To express the intended meanings, one of two ways must be taken: either the two verbs are kept discontinuous intervened by the affected object (thang-a 'window' in the present case), as in (72a-b), or the disposal marker ka is used and the two verbs go together, as in (73a-c). (72) a. Ong-e khok-ka thang-a phua-khi.25 p.n. knock-Ext window break-go 'Ong-e knocked the window and broke it.' b. Tsathau kong-ka thang-a phua-khi. wood strike-Ext window break-go 'The wood stoke the window and broke it.' c. Hongthai tsue-ka thang-a phua-khi. typhoon blow Ext window break-go 'The typhoon blew on the window and broke it.' (73) a. Ong-e ka thang-a khok phua-khi. p.n. Disp window knock break-go 'Ong-e broke the window by knocking it.' 25 The extension marker -ka here is not the same element as the disposal marker ka. The two morphemes are distinguished by tones. The extension marker -ka in Taiwanese is equivalent to the (resultative) extension marker -de in Mandarin. (In Taiwanese, -ka cannot introduce descriptive complements; a different morpheme is used for descriptive complements, -le. See Li (1990) for discussion.) 326 b. Tsathau ka thang-a kong phua-khi. wood Disp window strike break-go 'The wood broke the window by striking it.' c. Hongthai ka thang-a tsue phua-khi. typhoon Disp window blow break-go 'The typhoon broke the window by blowing on it.' Either of the two ways, however, does not involve verb movement of the lower, unaccusative verb to the higher action verb. Firstly, (72a-c) are identical to the Mandarin (68a-c), where head movement doesn't apply. Secondly, though in (73a-c) the action verb and the unaccusative verb are linearly adjacent to each other, incorporation or compound formation is not really involved here, since we can freely insert a syntactic element between the two, as shown below: (74) a. Ong-e ka thang-a khok-ka phua-khi. p.n. Disp window knock-Ext break-go 'Ong-e knocked the window and broke it.' b. Tsathau ka thang-a gong-ka phua-khi. wood Disp window strike-Ext break-go 'The wood stoke the window and broke it.' c. Hongthai ka thang-a tsue-ka phua-khi. typhoon Disp window blow-Ext break-go 'The typhoon blew on the window and broke it.' (75) a. Ong-e ka thang-a khok jit-le phua-khi. p.n. Disp window knock one-time break-go 'Ong-e broke the window by knocking it.' 327 b. Tsahtau ka thang-a gong jit-le phua-khi. wood Disp window strike one-time break-go 'The wood broke the window by striking it.' c. Hongthai ka thang-a tsue jit-le phua-khi. typhoon Disp window blow one-time break-go 'The typhoon broke the window by blowing on it.' In (74a-c) and (75a-c), a syntactic element is inserted between the action verb and the unaccusative verb phua 'break', the extension marker ka and a measure jit-le 'once'. These facts indicate that the unaccusative verb phua 'break' doesn't forms a unit with the action verb that precedes it. The more appropriate analysis for (73a-c) (as well as (74a-c) and (75a-c)) should be one in which the action verb (khok 'knock', kong 'strike', and tsue 'blow' in the examples) takes a VP complement that contains the expression phua-khi 'break go': (76) khok ' knock' ...[VP Subj kong ' strike' [VP phua-khi ] ] tsue ' blow' Our point here is, in (72a-c) and (73a-c), head movement of the main verb doesn't apply. And since the examples with compounds in (71a-c) are ungrammatical, verbs in Taiwanese don't move up to the verb. This is a major typological difference of Taiwanese in syntax. One last thing. In (70a) and (72-75), we see that a directional verb, khi 'go', is suffixed to the verb phua 'break'. There is good reason for it to occur in these examples. Tai (1984) argues that, in Mandarin Chinese, there are only three event types for verbs: states, activities, and results. The verb po 'break' in Mandarin seems to belong to the class of results, as it can assume inchoative use (cf. (66a) and (67-69)). But the verb phua 'break' in Taiwanese seems to be able to assume stative use. It cannot be used as an inchoative predicate. The following examples illustrate this point: 328 (77) a. *Pingzi po le. (Mandarin) *bottle break Prt *'The bottle broke.' b. *Kanna phua a (Taiwanese) *bottle break Prt *'The bottle broke' c. *Kanna phua-khi a (Taiwanese) *bottle break-go Prt *'The bottle broke.' In (77a), the verb po 'break' in Mandarin can be directly predicated of a theme subject, pingzi 'bottle'. The verb phua 'break' in Taiwanese cannot, however, as shown in (77b). To make the sentence grammatical, the directional verb khi 'go' must be inserted, as in (77c). Thus, the function of the directional khi 'go' is to impute the sense of inchoativity to the predicate. Without it, an unaccusative verb like phua 'break' in Taiwanese cannot assume inchoative use. This phenomenon indicates that verbs in Taiwanese have much more impoverished eventuality properties than verbs in Mandarin. We would like to provide one more example to illustrate this phenomenon, and suggest the possibility that, richness or impoverishment of the eventuality properties in the verb may be the major cause for the presence or absence of head movement in a language. Consider the following examples: 329 (78) a. *Ni mingtian qu Taipei ma? (Mandarin) *you tomorrow go Taipei Q *'Are you going to Taipei tomorrow?' b. *Li mi'atsai khi Taipak bo? (Taiwanese) *you tomorrow go Taipei Q *'Are you going to Taipei tomorrow?' c. *Li mi'atsai be khi Taipak bo? (Taiwanese) *you tomorrow want go Taipei Q *'Are you going to Taipei tomorrow?' (78a) is a question sentence in Mandarin that asks about whether the addressee is going to Taipei the day after the speed time. The eventuality meant in this question is irrealis, as it has not happened yet at the time of speech. We see that the verb qu 'go' along suffices to carry the relevant modalistic information. Surprisingly, however, word-to-word translation of the Mandarin sentence in (78a) into Taiwanese results in ungrammaticality. That is, a verb like khi 'go' in Taiwanese cannot carry the modalistic irrealis information. This point is shown by the ungrammaticality of (78b). To express the same semantic content as in (78a), an intentional verb, be 'want', must be overtly present in the sentence, as in (78c). How can we account for this discrepancy? Along the line we have been pursuing, an account based on head movement recommends itself. We can assume that the element responsible for moods such as the realis/irrealis distinction heads an independent projection in the syntactic structure. Let's simply call it MoodP. Verbs in Mandarin can incorporate to the head of MoodP, but verbs in Taiwanese cannot. This contrast is represented in the following diagrams: 330 (79) a. Mandarin b. Taiwanese .... MoodP .... MoodP Mood' Mood Mood' .... VP Mood ...qu... 'go' be 'want' .... VP ...khi... 'go' Mood, if regarded as an independent projection in the syntactic structure, should stand in a very high position in the phrase structure.26 The account we suggest for the discrepancy between Mandarin and Taiwanese in the realization of mood elements, represented in (79ab), thus accords with the conclusion we have reached about the general distinction between Mandarin and Taiwanese -- in Mandarin, the main verb moves up to the light verb; in Taiwanese, the light verb is lexically spelled out. The conclusion to draw is that verbs in Taiwanese do not move as high as verbs in Mandarin, an important typological difference that distinguishes the two Chinese dialects. 4.3 Summary In this section we extended our discussion to other Chinese dialects. We brought up the issue on the diversity of Chinese dialects, and introduced a survey conducted to test the unselectiveness of subject and object in different Chinese dialects. Several significant distributions were discerned and discussed. We then focused on Taiwanese and showed that some important aspects in the differences between Taiwanese and Mandarin are susceptible to grammatical accounts. We observed that Taiwanese exhibits unselectiveness 26 See Shen (2001) for the proposal that Mood is an independent head in the syntactic structure of Mandarin Chinese. 331 of object, but not unselectiveness of subject. However, we found that this discrepancy between Taiwanese and Mandarin can be reduced to one single grammatical factor: the choice between Attract-Move and lexical spell-out as the means to realize the subjectselecting light verb. Our analyses show that grammatical accounts for the Chinese dialects are not only possible, but also rewarding. All this provides support to the light verbs syntax we have been developing in this thesis. 5. The diachronic perspective So far we have seen that the light verb syntax developed in this thesis provides a good basis for a deeper understanding on the nature of phrase structure in Mandarin Chinese and other Chinese dialects. We have shown that the theory of light verbs is successful in accounting for intriguing syntactic facts in the contemporary Chinese languages. In this section, we turn our attention the historical perspective of the Chinese syntax. Of course, just like the comparative study of Chinese dialects, a complete study of the historical syntax of Chinese goes far beyond the possible scope of a thesis. We will just focus on some important aspects related to the light verb syntax of Chinese. In Chapter 2 we have briefly discussed some historical facts in Archaic Chinese in connection to the denominals and deadjectivals. An example given there is that, in Archaic Chinese, adjectives can freely assume causative use, as shown below (Wang 1980): (80) a. xiao zhi small it 'to make it small' b. guang zhi wide it 'to widen it' 332 c. huo zhi alive it 'to make it alive' d. jie zhi clean it 'to make it clean' In (80a-d), the adjectives xiao 'small', guang 'wide', huo 'alive', and jie 'clean' are used as causative verbs. Such deadjetivals, however, are not permitted in Mandarin Chinese. The following examples are all ungrammatical: (81) a. *xiao zhe-ge dongxi *small this-Cl thing *'to make this thing small' b. *kuan na-ge churu-kou *wide that-Cl entrance *'to widen that entrance' c. *hong zhe-fu hua *red this-Cl painting *'to redden this painting' d. *liang na-zhan dian-deng *bright that-Cl light *'to brighten that light' The discrepancy between (80a-d) and (81a-d) is just an instance of a more general difference between Archaic Chinese and Contemporary Chinese. In Archaic Chinese, nouns and adjectives can freely assume verbal uses, in sharp contrast from Mandarin (cf. 333 sections 3.2.2 to 3.2.4 of Chapter 2). In view of such discrepancy, it is of high interest to explore the relevant grammatical changes in the history of Chinese. In the following subsections, we will first look at various types of derived uses of verbs, nouns, and adjectives in Archaic Chinese. Then we look into the historical changes that result in the discrepancy alluded above. We propose that the major factor responsible for such ancientmodern discrepancy is the weakening of the light verbs, from highly morphological to purely syntactic. 5.1 Liberal uses of verbs, adjectives, and nouns in Archaic Chinese It has been noticed by many researchers of historical Chinese linguistics that, in Archaic Chinese, the uses of verbs, adjectives, and nouns are much more liberal than in Contemporary Chinese (cf. Wang 1980, Tan 1981, among many others). Several major types can be discerned: the causative use of verbs, adjectives, and nouns, and the intentional use of adjectives and nouns. There are still some minor types. We will look at the relevant phenomena below. 5.1.1 Causative use of verbs, adjectives, and nouns In Archaic Chinese, almost all major lexical categories, including verbs, adjectives, and nouns (excluding prepositions), can be used as causative verbs.27 This type of derived 27 Peyraube (1996) gives the following chronological framework of the different stages of Chinese: 1. 2. 3. 4. 5. 6. 7. 8. 9. Pre-Archaic Chinese: Early Archaic Chinese: Late Archaic Chinese: Pre-Medieval Chinese: Early Medieval Chinese: Late Medieval Chinese: Pre-Modern Chinese: Modern Chinese: Contemporary Chinese: language of the oracle bone inscriptions. 14th-11th c. B.C. 10th-6th c. B.C. 5th-2nd c. B.C. 1st c. B.C-1st c. A.D. 2nd-6th c. A.D. 7th-mid-13th c. A.D. 13th-14th c. A.D. 15th-mid-19th c. A.D. mid-19th-20th c. A.D. This chronological framework is given as a reference for the reader. In the text, we do not strictly follow the nomenclature in this or other chronological frameworks. We simply refer to gross terms such "the ancient Chinese." This will not pose significant problem for the exposition of points in the text, since we attach dates with the examples illustrating the relevant phenomena. 334 uses of lexical items is traditionally referred to as shi-dong shi (“causative construction”). The semantic schemes for the causative uses of the lexical items are given below (cf. Wang 1980, Tan 1981): (82) Derived causatives in Archaic Chinese a. Verb V NP → cause NP to V b. Adjective A NP → cause NP to be A c. Noun N NP → cause NP to have/become N Below are some examples: (83) Verbs used as causatives a. zhi Zilu su, sha ji wei su er shi zhi. stop Zilu stay-for-night, kill chicken cook rice and eat him '...invited Zilu to stay for a night, slaughtered a chicken and cooked rice to feed him' (Weizi, Lunyu, 500 BC.) b. gu yuan-ren bu fu, therefore far-people not obedient ze xiu wen-de yi lai zhi. then repair literary-virture so-that come them '...therefore, if people in other states refuse to be obedient, one should improve his own virtues so as to make them come' (Jishi, Lunyu, 500 BC.) 335 (84) Adjectives used as causatives a. Gu tian jiang jiang da-ren yu si-ren ye, therefore Heaven will fall great-responsibility to that-person Prt, bi xian ku qi xin-zhi, lao qi jin-gu definitely first bitter his heart-mind, laborious his muscle-bone 'Therefore, if Heaven is going to assign great responsibility to some person, His Majesty definitely will first set his heart in bitterness and his body in laborious situations...' (Gaozi, Mengzi, 300 BC.) b. Ranyou ue: "Ji shu yi, you he yi jia yen?" p.n. say now prosperous already more what add Q Ue: "Fu zhi." say rich them ‘Ranyou asked: "Now the people have a stable life; what to add to them?” Confucius said: "Make them rich.”’ (Zilu, Lunyu, 500 BC.) (85) Nouns used as causatives a. Wu jian Shenshu, Fuzi suo-wei sheng si er rou gu ye. I see p.n. master so-call live death and flesh bone Prt 'I saw Shenshu, who, as said the master, is like a dead man brought back to life, and a bone caused to regrow flesh.' (22nd year, Xianggong, Zuozhuan, 300 BC.) 336 b. Qi Huangong he zhuhou er guo yi-xing. p.n. ally feudal-ruler and state different-name 'Qi Huangong allied the feudal rulers and made people of different names into independent states.' (Jin shijia, Shiji, 100 BC.) In (83a-b), the transitive verb shi 'eat' and the intransitive verb lai 'come' are used in a causative way. So is the case of the adjectives ku 'bitter', lao 'old', and fu 'rich' in (84a-b), and the nouns rou 'flesh' and guo 'state' in (85a-b). Causative uses of verbs and adjectives are fairly productive in Archaic Chinese. Causative uses of nouns are much rarer (Wang 1980), but they are by no means unusual. All of the examples in (83-85) are not acceptable in Mandarin Chinese.28 We will come back to this point later. 5.1.2 Intentional use of adjectives and nouns A special type of derived uses of lexical items in Archaic Chinese is what is commonly referred to as the "intentional" use (yidong shi, 'intention-action construction'; see Wang 1980). Only adjectives and nouns are eligible for the intentional use, with the following semantic schemes: (86) Intentional uses in Archaic Chinese a. Adjective A NP → consider NP A b. Noun N NP → consider NP an N 28 The causative and intentional uses of lexical items are not permitted in all the Chinese dialects that I know of. So it is likely that these derived uses are not permitted in all the contemporary Chinese languages. 337 The intentional use turns an adjective or a noun into a transitive verb, with the meaning that the subject takes or considers the object as having the property of the adjective or the noun. Below are some examples: (87) Intentional use of adjectives a. xin bei qing-xiang, zhi xiao wan-cheng, heart humiliated official-minister, ambition small emperor ji shou zhi zheng, hun-luan bu-zhi once give him governance dark-chaotic not-ruled '[Those people] despise the officials and ministers, and consider the emperor humiliated; but once they are granted the power to govern, they only make things chaotic and cannot maintain an order.' (Yan-tie lun, 100 BC.) b. Shi Chong nian qi-shi yu, shang lao zhi at-the-time p.n. age 70 more, emperor old him 'At the time Chong's age had been more than 70 years old, and the emperor considered him too old.' (Hanshu, 100 AD.) 338 (88) Intentional use of nouns a. gu ren bu du qin qi qin, bu du zi qi zi so people not only parent their parent not only son their son '...so people not only take their parents as parents, only their sons as sons [they also treat other people just like they treat their own parents and sons.]' (Liyun, Liji, 300 BC.) b. buru wu wen er yao zhi ye rather I inspect and medicine him Prt 'it would be better for me to inspect him and take him as my medicine' (31st year, Xianggong, Zuozhuan, 300 BC.) Again, examples like (87-88) are not permitted in Mandarin Chinese -- as a matter of fact, the intentional use of adjectives and nouns has disappeared entirely in Mandarin. We will come back to the relevant questions later. 5.1.3 Other verbal uses of nouns The causative use of verbs, adjectives, and nouns, and the intentional use of adjectives and nouns, are the two major types of derived uses of lexical items in Archaic Chinese. Compared to verbs and adjectives, instances of nouns in these two uses are rarer. But this statement doesn't cover other derived verbal uses of nouns, which were very productive in Archaic Chinese. Here we present two types: nouns used as instrument and location verbs. Some examples are given below (Wang 1980): 339 (89) Noun used as instrument verb a. Zuoyou yu ren Xiangru guard intend sword p.n. 'The guards tried to kill Xiangru with swords.' (Lian Po Lin Xiangru liezhuan, Shiji, 100 BC.) b. Caozi shou jian er cong zhi p.n. hand sword and follow him 'Caozi held the sword in the hand and followed him' (13rd year, Zhuanggong, Gongyangzhuan, 300 BC.) (90) Noun used as location verb a. Qin shi sui dong Qin troop therefore east 'Qin's troop therefore moved toward the east' (32nd year, Xigong, Zuozhuan, 300 BC.) b. Jin jun Hanling, Qin jun Fannan. Jin army Hanling Qin army Fannan 'The Jin state placed its troop in Hanling, and the Qin state, Fannan.' (30th year, Xigong, Zuozhuan, 300 BC.) In (89a-b), a noun is used as an instrument verb, denoting an action making use of that instrument. In (90a-b) we have two different sub-cases. In (90a), a direction noun is used as a verb denoting the direction that the subject is moving toward; in (90b), a common noun is used as a location verb, followed by a location object. 340 5.2 The evolution To account for the liberal derived uses of lexical items in Archaic Chinese, we propose that there were very productive verb movements in Archaic Chinese. The productive verb movements, we suggest, arose from richer verbal morphology. The ancient-modern discrepancy in Chinese, therefore, results from weakening of the verbal morphology in the Chinese verbs. 5.2.1 Differences between Archaic Chinese and Mandarin Chinese The most salient difference between Archaic Chinese and Mandarin Chinese is the productive deadjectival and denominal formation in the former but not in the latter. This somehow makes Archaic Chinese quite "English-like." The derived uses of verbs, adjectives, and nouns in Archaic Chinese can be readily accounted for with a light verb analysis. The general point is represented in the following diagram: (91) .... VP V' V .... XP ... X ... In Archaic Chinese, X can be either of the three categories, V, A, or N, which incorporates to the higher light verb V. In Mandarin Chinese, however, X can only be V. If X is A or N, incorporation to the higher V cannot apply. This is the major distinction between Archaic Chinese and contemporary Chinese. 341 Throughout the discussion, we have been assuming light verbs with substantial thematic functions. Here, again, we postulate light verbs with specific thematic functions to account for the derived uses of lexical items in Archaic Chinese. First we look at the derived causative. The causative use of verbs, adjectives, and nouns involve complementation of VP, AP, and NP to the light verb CAUSE, as in the examples below. (92) a. zhi Zilu su, sha ji wei su er shi zhi. stop Zilu stay-for-night, kill chicken cook rice and eat him '...invited Zilu to stay for a night, slaughtered a chicken and cooked rice to feed him' b. .... VP V' V CAUSE VP NP V' V Zilu shi 'eat' 342 (93) a. Ranyou ue: "Ji shu yi, you he yi jia yen?" p.n. say now prosperous already more what add Q Yu: "Fu zhi." say rich them 'Ranyou asked: "Now the people have a steady life; what to add to them?" Confucius said: "Make them rich."' b. .... VP V' V CAUSE AP NP A' A zhi 'them' (94) a. fu 'rich' Wu jian Shenshu, Fuzi suo-wei shen si er rou gu ye. I see p.n. master so-call live death and flesh done Prt 'I saw Shenshu, who, as said the master, is like a dead man brought back to life, and a bone caused to regrow flesh.' 343 b. .... VP V' V NP CAUSE NP N' N gu 'bone' rou 'flesh' In the (b) examples in (92-94), the light verb CAUSE takes a VP, an AP, and an NP complement respectively. The V, A, and N incorporates to CAUSE, yielding the surface structures in the (a) examples. This is the case of Archaic Chinese. In Mandarin Chinese, however, only verbs can move to a higher light verb CAUSE (see the discussion on the causer subject in Mandarin in Chapter 3). Adjectives and nouns cannot: (95) a. *Gupiao shichang fu-le henduo ren. *stock market rich many people *'The stock market made many people rich.' b. *Gutou rou le. *bone flesh Prt *'The bone grew flesh.' The intentional use of adjectives and nouns in Archaic Chinese can be analyzed in the same way as (92-94), with the light verb CONSIDER:29 29 Mei (2000) suggests a similar analysis for the intentional use of adjectives and nouns in the ancient Chinese. This analysis can be illustrated by the following example: (i) a. xian qi-ren moral that-person 'consider that person morally superior' 344 (96) a. shi Chong nian qi-shi yu, shang lao zhi at-the-time Chong age 70 more, emperor old him 'At the time Chong had been more than 70 years old, and the emperor considered him to be too old.' b. VP NP V' take VP NP V' as SC NP A 'that person' 'moral' Mei (2000) proposes that an intentional construction like (ia) has a multiple-VP structure, where the light verbs take and as occur in the structure heading two independent projections. The most embedded complement in the structure in (ib) is a small clause SC. The derivation involves two distinct sets of movements: firstly, the subject of SC raises up to the specifier of the as-VP, and then the adjective A incorporates to the light verbs as and take. The analysis we present in the text differs from (ib) in the following aspects. We assume that the eventuality structure of the intentional construction is simple; it just involves one single light verb INTEND and an AP or NP complement. Furthermore, we assume that the AP or NP complement takes an internal subject, as shown in the diagrams in (96b) and (97b). That is, we assume that the AP or NP is itself clausal. 345 b. .... VP V' V AP CONSIDER NP A' A zhi 'him' (97) a. lao 'old' gu ren bu du qin qi qin, bu du zi qi zi so people not only parent their parent not only son their son 'so people not only take their parents as parents, only their sons as sons' [they also treat other people just like they treat their own parents and sons.]' b. .... VP V' V NP NP N' qi qin 'their parents' N CONSIDER qin 'parents' 346 The light verb CONSIDER is not compatible with verbs. It only takes a complement denoting a state or a property, an AP or an NP.30 Since in Mandarin Chinese only verbs can move up to higher light verbs, the intentional use of A and N becomes extinct in Mandarin, as the following examples show: (98) a. *Zhangsan lao Lisi. *p.n. old p.n. *'Zhangsan considers Lisi old.' b. *Xiaoli ma bie-ren de ma. *p.n. mother other-people Mod mother *'Xiaoli treats other people's moms like his own mom.' In the case of verbal uses of nouns, complication arises, since some of the examples require analyses that are not seen in the more familiar cases in English and Mandarin Chinese. First let's look at the followiing example: (99) a. zuoyou yu ren Xiangru guard intend knife p.n. 'The guards tried to kill Xiangru with swords.' 30 Jim Huang (personal communication) points out to me that this is the same for the verb consider in English, since the verb consider only takes an A or an N as complement (in a small cause construction), but not a V: (i) a. b. c. *John considers Bill smart. *John considers Bill a genius. *John consider Bill get an excellent grade. 347 .... VP b. NP zuoyou 'guards' V' VP V CAUSE V' NP V Xiangru BECOME AP A DEAD PP P NP Instr ren 'knife' In (99a), the instrument ren 'sword' is used as the main verb of the sentence, meaning "to kill someone by a knife." (99b) is the structural analysis that we suggest for (99a). There are several special points in (99b). Firstly, an empty preposition, Instr, is postulated in the structure to introduce the instrument ren 'knife'. (For the use of empty prepositions in English, see Kayne (1984) and HK (1997).) This doesn't seem to be an implausible postulation, since Archaic Chinese, unlike Mandarin Chinese, permits post-verbal PPs (cf. Sun 1996, among others). Secondly, there is an empty A, to which we assign the semantics of "dead" (structurally a light-verb-like element or a light adjective), that heads an AP, taking the instrument PP as complement. Rappaport (1999) postulates empty APs in syntactic structures (his Aspectual Structures; cf. section 3.3.1 of Chapter 2) responsible for the sense of resultativeness in accomplishment and achievement events. But for our purposes, it seems necessary to assign more concrete semantics to the empty A in (99b). Semantically, if X ren Y, then necessarily Y is dead, not only wounded or getting some adverse effect via the use of ren 'sword'. In other words, X ren Y in Archaic Chinese can be paraphrased as "X kills/executes Y with a knife".31 Thus the resultant state (the AP in (99b)) of the 31 So the situation is different from the English expressions like to stab Caesar, since the verb stab, though implying an instrument (e.g. "with a dagger"), doesn't necessarily result in the death of the patient. 348 accomplishment eventuality (the CAUSE-BECOME light verb structure in (99b)) must be concrete. With these two special points, the derivation of the surface structure is straightforward: the NP ren 'knife' incorporates to the empty P, the empty A, BECOME, and CAUSE, yielding (99a). Next let's see the case of an NP used as a location verb. Here the case is even more intriguing: (100) a. Jin jun Hanling, Qin jun Fannan. Jin army Hanling Qin army Fannan 'The Jin state deployed its troop in Hanling, and the Qin state, in Fannan.' .... VP b. NP Jin V' VP V CAUSE NP jun 'troop' V' V BECOME PP P NP Loc Hanling In (100a-b), the theme NP, jun 'troop', is used as the main verb of the sentence. In the structure in (100b), we have an empty P, Loc, that introduces the location Hanling, a place name. The theme NP jun 'troop' occurs as the specifier of the BECOME-VP. Here an intriguing problem arises. There is good reason to posit (100b) as the underlying structure For example, in English, one can say Bill stabbed Max and seriously wounded him, without being committed to the belief of Max's death. 349 for (the first conjunct of) (100a), since (the first conjunct of) (100a) can be paraphrased as "The Jin state deployed its troop in Hanling." The underlying structure is identical to what HK posited for predicates such as shelve the books in English, paraphrased as "put the books on the shelf." But curiously, in (100a-b), it is the (internal) theme subject jun 'troop" of the BECOME-VP that undergoes incorporation to the higher light verb CAUSE. This is comparable to the kind of expressions *book the shelf or *church the money in English, which, however, are not permitted (HK 1993). According to HK (1993), these expressions are ungrammatical because incorporation from the specifier position is ungrammatical, since it violates Chomsky's (1986b) Minimality Condition. Some researchers have challenged the validity of such ECP violation (cf. Fodor and Lepore 1999). It seems to us that (100a) really poses a problem for HK's proposal. Right now we don't have an account that accommodates these problems. An insightful analysis such as the HK's should not be dismissed simply in the face of an individual counterexample like (100a). It is likely that further investigation into the related questions would lead to a theory of eventuality structures and their syntactic representations with more explanatory power. We will leave such a feat for future research. At this point, we will simply suggest the analysis represented in (100b) as a possibility. In (100b), the theme NP jun 'troop' incorporates to the light verb CAUSE from the specifier position of the BECOME-VP. There is yet an incorporation that occurs in the structure, the incorporation of the empty P (Loc) up to CAUSE. Notice that this second incorporation is unusual. If P must be conflated to a lexical element due to its phonetic emptiness, the most natural result that we would expect is that P attracts its complement, Hanling. But in fact this doesn't happen. What happens instead is the incorporation of the empty P itself up to the light verbs BECOME and CAUSE. These two applications of incorporation conjointly yield the locational verbal uses of the theme NP jun 'troop' in (100a). 350 One might wonder if the analysis represented in (100b) is empirically supported in any way. While sorting out all relevant facts and present a complete argument in support for it go beyond the capacity of this chapter, we would like to point out a curious historical fact in Chinese that seems to be relevant. Checking the uses of the location noun xia 'case' in the Twenty five Historical Records (Er-shi-wu Shi) in the electronic database built by Academia Sinica (Taipei),32 we find that, in Archaic Chinese documents, xia 'case' can only be used as a noun, as in (101a-b). The use as a location verb occurs very late in Medieval Chinese, as in (102a-b): (101) a. chu qi chuang xia-zhong zhu take he pack case-in zhu 'take out the zhu [a particular kind of musical instrument] which he put in the case' (Cike liezhuan, Shiji, 100 BC.) b. Qi wang zang yong yu-xia its king bury use jade-case 'When the king of [that state] is buried, a jade coffin is used' (Dong-yi liezhuan, Hou-Hanshu, 300 AD.) 32 The website is: http://saturn.ihp.sinica.edu.tw/~hanchi/hanchi/. 351 (102) a. da-jiao fu ming yi-tong, jie juan fan, she shi, big-horn again sound one-Cl all roll flag take-back arrow chi gong, xia ren loosen bow case knife 'When the horn sounds for the second time, all [soldiers] roll the flags, put the arrows back to the bags, loosen the bow lines, and put the knives back to the cases' (Fu-bing zhi, Shin Tang-shu, 1100 AD.) b. wu nian, Di xia zhan-ma-dao yi shi Tsai Ting five year Emperor case chop-horse-knife for show Tsai Ting 'The fifth year, His Majesty put the horse knife in a case and showed it to Tsai Ting' (Qi-jia zhi zhi, Song shi, 1300 AD.) If the contrast between (101) and (102) is a general one between Archaic and Medieval Chinese, then a puzzling question arises. In the history of Chinese, there are two kinds of locational denominals, the one represented by (100a-b), and the one represented by (102ab). The former is comparable to the ungrammatical expression *book the shelf in English (in the sense of "put the book on the shelf"), and the latter to the grammatical expression shelve the books. In the history of Chinese, the former occurs much earlier in the history of Chinese, but the latter, fairly late. If HK's theory is correct, this contrast is not expected, since we would expect the former to represent the unmarked type of locational denominals in Chinese. At this point we don't have an answer for this question. However, in view of the historical contrast between (100a) and (102a-b), it is certain that there are important issues to be explored. We will leave them for future research. 352 At any rate, we would like to point out that these verbal uses of nouns are not allowed in Mandarin Chinese. This is evidenced by the following examples: (103) a. *Zhangsan dao Lisi. *p.n. knife p.n. *'Zhangsan killed Lisi by a knife.' b. *Su-lian jun Luomaniya. *USSR troop Romania *'The USSR deployed troops in Romania.' c. *Xiaoli xiang-le naxie wenjian. *p.n. box-Perf those document *'Xiaoli put those documents in a box.' In summary, from all these discussions, it is clear that in Archaic Chinese liberal trans-category movement of a head to a higher light verb is very productive, a special feature that distinguishes it from Contemporary Chinese languages. But why is there such a distinction? What happens in the history of Chinese that results in such restrictiveness in the contemporary Chinese? In the next subsection we will suggest an account based on a historical change that brings Chinese from a subject-oriented language to a topic-oriented language, and, furthermore, from a morphological language to a syntactic language. 5.2.2 From morphological to syntactic In the realm of Chinese linguistics there is a rich body of literature on Chinese at different historical stages, but works based on generative assumptions aiming at the theoretical aspects of the grammar are still rare. Since there are still many questions in Archaic Chinese that are difficult for generative linguistics to approach, what we will propose below are largely speculative and suggestive in nature, requiring further confirmation. 353 We suggest that the ancient-modern contrast with respect to the light verb syntax discussed above has to do with a hypothesis that T. Mei (1991) proposes. According to this hypothesis, the basic structure of the Chinese sentence has been changed from subjectpredicate to topic-comment. The major cause for this change, according to T. Mei (1991), is the neutralization of morphological transitive/intransitive distinction in the history of Chinese. In the following we look at this hypothesis. 5.2.2.1 From subject-predicate to topic-comment Contemporary Chinese has been known as a topic-prominent language (cf. Tsao 1979). This characteristic of Contemporary Chinese is reflected in the fact that, in a Mandarin sentence, the major break may not be between the NP subject denoting an individual and the VP predicate denoting a property that the subject possesses; it can be between an NP topic and a clausal comment that is semantically related to the topic in a loose way. A famous example given by Tsao (1979) is the following: (104) Zhe-chang huo, xingkui xiaofang-dui lai-de zao. this-Cl fire fortunately fireman-team come-Ext early 'As to this fire, it is fortunate that the firemen came early.' In (104), the topic NP zhe-chang ho 'this fire' doesn't bear any selectional relation with any element in the sentence. It is licensed simply by being a topic. Since the topic only needs to bear a loose semantic relationship with the comment, the thematic status of the subject/topic NP in the Mandarin sentences may be ambiguous. This effect is particularly clear in the case of transitive predicates. T. Mei (1991) cites Chao's (1979) observation and points out that sentences of the following kind are ambiguous: 354 (105) a. Yu chi le. fish eat Prt b. (i) 'The fish ate [something].' or (ii) 'The fish was eaten [by some animate being].' Da-mao ya-si le. big-cat press-dead Prt (i) 'The big cat pressed [some animate being] to death.' or (ii) 'The big cat was pressed to death [by something].' Both (105a) and (105b) are ambiguous: the subject/topic NPs yu 'fish' and da-mao 'big cat' can be understood either as the agent of the action (chi 'eat' and ya-si 'press-dead'), or the patient of the action. Thus, the agent/patient distinction is "neutralized" in the subject/topic position of a sentence in Chinese (cf. T. Mei 1991). But it is not the case that Chinese has been always like this. Many researchers (cf. Ota 1958, Mei 1991, among others) observe that in Archaic Chinese agents and patients in subject/topic position are clearly distinguished via lexical or morphological markings. Take compound verbs such as ya-si 'press-dead' in (105b) as an example, cited in T. Mei (1991). In Mandarin, a compound verb like ya-si 'press-dead' can take either an agent or a patient as subject, as the following examples show: (106) a. Da-mao ya-si xiao-mao le. (Agent) big-cat press-dead small-cat Prt 'The big cat pressed the small cat to death.' b. Xiao-mao ya-si le. (Patient) small-cat press-dead (i) 'The small cat pressed [some animate being] to death.' or (ii) 'The small cat was pressed to death [by something].' 355 However, in Archaic Chinese, the compound verbs of the ya-si 'press-death' type can only assume intransitive use and take a patient subject. For transitive use, a different verb, sha 'kill', must be used replacing si 'dead', yielding the compound verb ya-sha 'press-kill'. To be specific, there is a general contrast in Archaic Chinese on the thematic status of the subject/topic NP that occurs with X-sha 'X-kill' and X-si 'X-dead', X being an action verb. For expressions of X-sha 'X-kill' type, the subject/topic must be an agent, in conformity with the agentivity of the action verb sha 'kill'. For expressions of X-si 'X-dead' type, the subject/topic must be a theme, in conformity with the resultativity of the stative verb si 'dead'. This contrast is illustrated in the following two examples: (107) a. An beng, jin ya-sha wo zhe. slope collapse all press-kill lie person 'The slope collapsed, and [the fallings] pressed to death all those who lied there.' (Waichi shijia, Shiji, 100 BC b. Bai yu ren tan beng jin ya-si. hundred more person coal collapse all press-dead 'More than one hundred people were pressed to death due to the collapse of the coal mine.' (Jiyan, Lunheng, 100 AD.) (107a) has an phonetically null instrument subject (which patterns with an agent), and the verb used is the transitive verb ya-sha 'press-kill'. (107b), on the other hand, has a patient subject, and the verb employed is the intransitive verb ya-si 'press-dead'.33 Thus the agent 33 In both (107a) and (107b), the first verb ya 'press' in the verbal compound seems to be adverbial in nature, without an effect on the thematic status of the arguments in the sentence. Huang (1995) attributes this to the right-headedness of compounds in the ancient Chinese, like the verbal compounds in Japanese (cf. Y. Li 1993 and Nishiyama 1998, among others). 356 subject in (107a) and the patient subject in (107b) are distinguished by the lexical items used. There are still other contrasts between transitive and intransitive predicates in Archaic Chinese,34 and all this indicates that, in Archaic Chinese, the subject and the predicate of a sentence are tied up with a much stronger thematic relationship than in the contemporary Chinese. Starting from the early Medieval Chinese (about 700 AD.), these distinctions began to weaken, a process that eventually leads to complete extinction of all these lexical/morphological distinctions, resulting in the "neutralization" of the thematic status of the subject in a Chinese sentence. According to T. Mei (1991), this neutralization brings Chinese from a subject-oriented language to a topic-prominent language, as the strong selectional bond between the subject NP and the predicate becomes loosened. In Mandarin Chinese (and other Chinese dialects), either an agent or an patient may occur as the subject of an action verb, without explicit lexical or morphological indication on the particular thematic relation involved. This is not possible in Archaic Chinese, where the agents and the patients are clearly distinguished by lexical or morphological markings. 34 Other lexical or morphological markings that distinguish transitive from intransitive predicates in the ancient Chinese include the following (T. Mei 1991): (i) For the modal verb expressing permission or possibility, two distinct forms are used: neng must take an agent subject, whereas ke must take a patient subject. (ii) In relative clauses, if the relativized NP is an agent, the morpheme zhe must be used, yielding the structure V zhe 'the one who does...'. But if the relativized NP is a patient, the morpheme suo must be used, resulting in the structure suo V 'the one who has been...'. (iii) The negation markers used are different as well: for a transitive predicate with an agent subject, the negation marker is fu or wu; for an intransitive predicate, the negation marker is bu or wu. There are still other means to distinguish transitive predicates from intransitive predicates. An important one, which is highly morphological, is ching-zhuo bie yi (literally meaning "voiceless-voiced distinguish meaning"). We will turn to it later in the text. For details and references on all these distinctions, see T. Mei (1991). 357 5.2.2.2 Weakening of morphology in the history of Chinese T. Mei (1991) lists a number of phenomena in Archaic Chinese that show explicit distinctions between agent subjects and patient subjects. Among them, there is one that stands out with high interest for our concern -- the contrast in the voicedness of the onset consonant of a verb as an indicator for the transitivity property of the predicate. This contrast is traditionally known as ching-zhuo bie yi '(lit.) voiceless-voiced distinguish meaning'. In Archaic Chinese, a verb may assume either transitive or intransitive use, and the two uses are marked by the voicing quality of the onset consonant of the verb -- if the onset consonant is voiced, the verb is intransitive, and if the onset consonant is voiceless, it is transitive. Below are some examples that T. Mei (1991) gives: (108) a. b. bai zhe i. 'to defeat' [p-] ii. 'to fail' [b-] i. 'to break (tr.)' [ts-] ii. 'to break (intr.)' [dz-] Thus the transitivity properties of verbs in Archaic Chinese can be identified via their morphological constitutes. Chin-zhuo bie yi starts to decline in the late ancient Chinese and becomes completely extinct in Medieval Chinese. T. Mei (1991) takes the decline of chingzhuo bie yi in Medieval Chinese as a major cause for the transition of Chinese from a subject-oriented language to a topic-oriented language. In other words, the decline of morphology in the history of Chinese is responsible for the neutralization of the thematic status of subject NPs. We will base our analysis on T. Mei's (1991) hypothesis. To be specific, we propose that, in history, Chinese has been changing from a morphological language to a syntactic language. In the following, we will take the light verb CAUSE as an example for a demonstration of this change. The historical changes that occur to the light verb CAUSE in Chinese syntax can be represented in the following diagram: 358 (109) The evolution of CAUSE in the history of Chinese Strong morphological (Prefix *s-) ⇓ Weak morphological Morphological feature only ; ching - zhuo bie yi ⇓ Syntactic (No morphology at all) We start with the very beginning of the history of Chinese. As mentioned in section 4, Chinese is known to be a branch of the Sino-Tibetan language family. But there is an interesting difference between Tibetan and Chinese: though the contemporary Chinese is extremely isolating and lacks morphological alternations in the verbs, Tibetan has a rich verbal morphology and exhibits characteristics of an agglutinating language. Thus, as T. Mei (1991) points out, proto-Chinese should have undergone a dramatic change from a synthetic language to an analytic language. Some reconstructed phenomena in Archaic Chinese support this hypothesis. One important phenomenon in this regard is that the preancient Chinese should have a rich set of morphological elements that contains a prefix *s-, an infix *-r-, and a suffix *-s (see Chmielewski 1949, T. Mei 1980, 1991, Sagart 1999, among many others), each of which contributes a substantial thematic function to the verb, e.g. *s- for causativization, and *-s converting a verb into a noun or a noun into a verb. At this stage, Chinese is no different from verbal agglutinating languages, such as Japanese, where affixation is essential for the alternation of the thematic status of the arguments in the sentence. To wit, Chinese, at this stage, was an agglutinating language. 359 We assume that, at the time of the proto- Chinese and the pre-ancient Chinese, affixes were light verbs. Thus the prefix *s-, which causativizes a verb, is the lexical form of CAUSE, a bound morpheme. Furthermore, the prefixation of *s- to the verb in fact involves head movement of the verb to the higher light verb CAUSE. The following diagram illustrates the relevant points: (110) Head movement in the proto-Chinese and pre-ancient Chinese: (400 BC and beyond; driven by affixation) .... VP X = V, A, N, etc. V' V XP X' CAUSE s- X ω At this stage, head movement was completely morphology-driven. The head X in (110) incorporates to the light verb CAUSE because CAUSE is a prefix, *s-, which requires lexical support from some other lexical item. X does not have to be a verb; it can be an adjective or a noun. The morphological feature in the light verb, in the form of an affix, attracts X, and forces a lower head to incorporate to it. As Chinese evolved in time, however, verbal affixes started to weaken, and eventually disappeared, probably due to simplification of consonant clusters (cf. T. Mei 1980). Chingzhuo bie yi arose and took over the function of the prefix *s-. The function of distinguishing transitive and intransitive predicates was shifted to the phonetic quality of the verb. Thus, at this historical stage, Chinese became an inflectional language -- the verb inflected along the transitive vs. intransitive contrast. In terms of the contemporary syntactic theory (for instance, Chomsky 1995), there must be some morphological feature F in the 360 light verb that attracts the main verb with a particular inflection to check with it. At this stage, the light verb CAUSE was still morphological in nature, though it didn't have an independent phonetic form. The following diagram illustrates the relevant points: (111) Head movement from Archaic Chinese to the early Medieval Chinese: (400 BC - 1000AD; driven by feature checking) .... VP X = V, A, N, etc. V' V XP X' CAUSE [F] X ω [F] From Archaic to the early Medieval times, Chinese changed from (110) to (111). The phonetic form of the light verb was lost. However, the morphological feature still persists, and the head X has to move up to check with it with an appropriate inflectional form, i.e. the correct voice quality of the onset consonant. At this stage, CAUSE remained morphological, since the triggering force for head movement was the checking need of the inflectional feature F, a formal requirement. Chinese, at this stage, kept clear-cut transitive vs. intransitive distinctions, as T. Mei (1991) points out. The verb has to be inflected in the right way to check off the morphological feature in the subject-selecting light verb. In a sense, the transition from (110) to (111) is only quantitative in nature, since both are morphological. The only difference is whether the morphological feature has an independent phonetic form or not. The last transition, however, is a qualitative one, since it involves complete erasure of the morphological feature in the light verb. Before Chinese reached the medieval era, ching-zhuo bie yi had begun to decline, and eventually became 361 completely extinct. Inflection doesn't exist any more. What remains is a light verb as a syntactically free word. The following diagram illustrates the points: (112) Head movement from Medieval Chinese to Contemporary Chinese: (400 BC - 1000AD); driven by semantic requirements .... VP X = V only V' V XP X' CAUSE X ω Let's assume that, as a result of further weakening, the feature F is completely erased, and morphological requirements in CAUSE do not exist any more. Without morphological features, the light verb CAUSE becomes a purely syntactic, morphologically free predicate. The main verb incorporates to CAUSE simply for semantic composition, supposedly the requirement for event identification (Kratzer 1996). The disappearance of deadjectivals and denominals in the contemporary Chinese can be explained in this rubric. Several factors seem to come into play. The most important one, we suppose, is the following. In earlier stages of Chinese, the force that drives an adjective or a noun to move up to a light verb is the morphological requirement of the light verb -- that is, the need for an affix to attach to a lexical item or for a morphological feature to be checked off. These morphological elements attract a lexical head to them, regardless of the syntactic category and the semantics of that lexical item. Now in the contemporary Chinese, the morphological ingredient in the light verb is completely erased, and the light verb does not have morphological requirement any more. Adjectives and nouns are not compatible with the light verb CAUSE in syntactic category. The categorial status doesn't 362 match. In earlier stages of Chinese, the categorial mismatch can be overridden by the morphological requirement. But now, lacking a morphological drive, the categorial mismatch amounts to a cost for the movement of the adjectives and nouns to a higher light verb. Therefore deadjectivals and denominals don't exist in the contemporary Chinese any more. In this way, we explain the extinction of the free derived uses of lexical items in the contemporary Chinese. We attribute it to the erasure of morphological ingredients in the light verbs along the historical change of Chinese. As the nature of light verbs changes from morphological to syntactic, the morphological force that attracts adjectives and nouns diminishes. This is the reason that, in Mandarin Chinese, only verbs undergo head movement to higher light verbs, but adjectives and nouns do not. 5.3 Summary In this section we looked at Archaic Chinese and proposed an account for an important discrepancy between Archaic Chinese and Mandarin Chinese, that is, productive deadjectivals and denominals in the former but lack of them in the latter. We showed that the productive head movements in Archaic Chinese actually have a morphological root -- in Archaic Chinese, the light verbs are morphologically bound, which functions as a driving force for trans-categorial head movement to higher light verbs. Historical changes brought Chinese from a morphological language to an inflectional language, and finally to an syntactic (isolating/analytic) language, as it is now. Weakening of the morphological content in the light verb motivate such changes. Consequently, deadjectivals and denominals are no longer viable, since the morphological force in the light verb vanishes. All these proposals, of course, are suggestive in nature. There are still important questions to be clarified. But we believe that at least some important aspects in these proposals are close to the facts. Further investigations must be carried out, of course. Limited by space, we will leave the relevant questions open. But we believe that fruitful 363 grammatical analyses for the historical syntax of Chinese (and other languages) are not only possible, but also highly explanatory. In this section we tried to lay a preliminary basis for such analyses, and we believe that it is on the right track.. 6. Concluding remarks This chapter is devoted to three different perspectives on the light verb syntax in Chinese: the Davidsonian nature of phrase structure in Mandarin Chinese, the light verb syntax of Chinese dialects, and the historical change of the light verb syntax in Chinese. Some proposals in this chapter are only sketchy, but we believe that we have successfully shown that the light verb syntax we have developed in this thesis can be applied to other aspects of the Chinese languages yielding interesting results. Further research is needed, for sure, but we hope that the analyses proposed in this chapter can serve as a basis for further discussions and proposals on the related issues and lead to fruitful results. 364