V-V Compounds in Chinese: Syntactic Perspective Revisited

This article revisits the formation of the resultative V-V compounds in Chinese. While presenting evidence to show the inadequacies of the lexicalist approach, we instead argue that these compounds are derived via syntactic operations. Meanwhile, the multiple readings of V-V compounds, once claimed to be the strong argumentation of the lexicalist account, are actually the results of such characteristics of Chinese syntax as the object realization, focalization, and topic-prominence. The evidence provided by the lexicalists thus does not constitute an argument against the syntactic account.

tant differences as follows. First, although both V-V compounds and SVCs bear some resemblances, they are not the same in terms of the connectedness between the two verbs. The most obvious manifestation for this lies in the possibility of inserting some elements between the verbs (or verb morpheme). It has been shown in the literature that the first verb in an SVC, if it is transitive, can take a noun phrase (NP) object across various languages (Baker, 1989;Foley, 1997;Givón, 1979). Besides, it is often said that the verbs in SVCs must have the same aspectual value; it remains unclear how universal this is (Haspelmath, 2016). Some scholars believe that the two verbs of an SVC can have different value (Dixon, 2011;Muysken & Veenstra, 1995). This is also true of the SVCs in Chinese because the objects and aspect markers can both appear between the two verbs. For example, Chao (1968)  yesterday cry EXP wet one-CL handkerchief Intended reading: "He cried so much that a handkerchief was wet." However, it has been a quite different scenario for the V-V compounds in Chinese: The two verb morphemes must be linearly adjacent and no elements can ever occur in between, even if the first verb is transitive. For example (3SG = third person singular; IMPF = imperfective; EXP = experiential; CL = classifier), With the above examples, we can easily see the crucial differences between V-V compounds and SVCs. For SVCs, the objects or aspect markers can appear after the first verb in SVCs, as shown in (1). However, the ungrammaticality in (2) indicates that no such thing can ever occur in V-V compounds. The V-V compounds function like the single verb and only take, for example, the aspect marker after the second verb.
Second, Haspelmath (2016, p. 309) puts forward 10 generalizations, some of which can be examined to distinguish a V-V compound and an SVC. The most prominent one is the Generalization 7, which says "in all SVCs, all the verbs share at least one argument." This generalization is also among the most frequently mentioned characteristics of SVCs, at least since Foley and Olson (1985, p. 24). For instance, in the examples of SVC in White Hmong and its Chinese counterparts of (3) below, the agent is shared by the two serial verbs; for V-V compounds in Chinese, things are different because the two verbs (morphemes) may not share any argument at all, as is exemplified in (4) (1SG = first person singular; Perf. = perfective).
(Y. Li, 2014) 1SG take knife cut bread "I took the knife to cut the bread." (4) a. Zhangsan chi-huai-le duzi. Zhangsan eat-bad-Perf. stomach "Zhangsan has eaten (something bad. As a result,) his stomach is upset." b. Zhangsan chang-ku-le Lisi. Zhangsan sing-cry-Perf. Lisi "Zhangsan's singing made Lisi cry." In Example (4a), the first verb chi "eat" has the agent Zhangsan and something implicit as its arguments, while the second verb huai "upset" is intransitive and only has duzi "stomach" as its theme. This is also true of the V-V compound in (4b) because the first verb chang "sing" has Zhangsan as its agent and a song as its implicit patient, while the second verb ku "cry" is intransitive and only has Lisi as its agent. This indicates that the two verbs in V-V compounds are indeed causally related, but they need not necessarily share the argument(s).
Third, while the above two criteria may be used to define SVCs and presumably to distinguish SVCs and V-V compounds across languages, we have also noticed a languageparticular aspect which can illustrate the key difference between the two, that is, the phonological size. As far as Chinese is concerned, all the V-V compounds discussed in Chinese are disyllabic, with the first monosyllabic verb morpheme denoting an activity while the second one indicates a change-of-state or result. This can be evidenced by the V-V compound facts presented in the previous literature, including both the lexical account and the syntactic account (to be discussed in a later section). However, there is no such phonological constraint on the SVCs in Chinese which essentially have nearly all exceeded such a syllabic number.
With all the above discussions, it is believed that the V-V compounds and SVCs in Chinese are linguistic entities of different types and not difficult to be distinguished from each other. Compared with SVCs, the V-V compounds are subject to some restrictive delimitation in some aspects like phonological size and element insertion, but enjoy a relaxation in others such as argument sharing. Next, we will move on to discuss the two competing approaches to such linguistic phenomenon.

Lexicalist Account Versus Syntactic Account
In the literature, the derivation of resultative V-V compounds in Chinese has been a subject of heated debate. Two approaches are most influential: lexicalist account and syntactic account. While the syntactic account still finds favor with many syntacticians (L. L.-S. Cheng, 1997;L. L.-S. Cheng & Huang, 1994;C. Shi, 2008;D. Shi, 1998;Sybesma, 1992Sybesma, , 1999Sybesma & Shen, 2006;Xiong, 2006, among many others), the lexicalist account, with the efforts of Y. Li (1990Li ( , 1993Li ( , 1995Li ( , 1999Li ( , 2005, has gained much success during the last two decades. Advantages and shortcomings of lexicalist account. As a representative of lexicalist account, Y. Li (1990) believes that V-V compounds enter syntax directly from the lexicon and that the multiple readings of V-V compounds are essentially a case of lexical ambiguity. His explanation relies on two theories: one is the theta identification (Higginbotham, 1985) and the other is the Case Theory. Y. Li (1990) states, When a compound verb heads the VP in a clause, only two structural Cases, nominative and accusative Case, can be assigned to its NP arguments. This in turn means that such a compound can assign theta-roles to no more than two arguments, with one of them receiving Case from the subject position and the other from the object position. Therefore, if the total of the theta-roles of the two component morphemes exceeds two, some of them must be identified to guarantee that each theta-role is eventually assigned to some Case-marked argument, as the Theta Criterion demands. (p. 184) Besides, to account for the distribution of legitimate configurations, Y. Li (1990) assumes that the θ-grids of the two Vs are structured according to the prominence hierarchy as "agent > goal > theme . . ." That is to say, the θ-role assigned to the external argument should be higher than the one assigned to the internal argument, and the hierarchy of θroles of the two verbs must be kept in the result θ-grid when they combine to form a V-V compound.
Li's account has shown its power to explain the two readings of Baoyu qi-lei-le ma (Y. Li, 1990) and the three readings of Taotao zhui-lei-le Youyou (Y. Li, 1995), as shown in (5) and (6) Y. Li (1990Li ( , 1993Li ( , 1995Li ( , 1999Li ( , 2005 provides a beautiful explanation for the properties of many resultative V-V compounds and presents a possible explanation for the absence of certain interpretation of V-C constructions. However, although Li's account covers most of the phenomena, exceptions still can be found. Consider the examples in (7-10): (7) Zhangsan chi-huai-le duzi. Zhangsan eat-bad-Perf. stomach "Zhangsan has eaten (something bad. As a result,) his stomach is upset." (8) Zhangsan chang-ku-le Lisi. Zhangsan sing-cry-Perf. Lisi "Zhangsan's singing made Lisi cry." (9) Zhangsan chou-guang-le jiachan. Zhangsan smoke-empty-Perf. estate "Zhangsan smoked (e.g., opium. As a result, he) finished all his money." (10) Zhangsan xi-shi-le xiezi. Zhangsan wash-wet-Perf. shoes "Zhangsan was washing (something. As a result,) his shoe(s) got wet." An approach, whether it is lexical or syntactic, as long as it is carried out within the generative framework, must be ready to be confronted with the challenges from the θcriterion and/or the Case Theory. For example, the θ-grids of chi "eat" and huai "bad" in (7) are shown in (11): (11) chi: <1, 2> huai: <1'> According to Y. Li (1990), because the argument number of the verb component is more than two, these θ-roles must be identified to be assigned to two arguments. Logically, there are two ways: However, neither of them is applicable to chi-huai "eat-bad," which actually has a θ-grid like (13), with <1> assigned to the argument which occupies the subject position and <1'> assigned to another argument which occupies the object position.
(13) <1, 1'> But which argument is the θ-role <2> assigned to? Obviously, there is no argument to bear this θ-role. However, if it is not assigned at all, a violation of the θ-criterion will arise.
Zhangsan cry-wet Perf. handkerchief "Zhangsan cried so much that his handkerchief got wet." Therefore, to the syntactic account, the derivation of (7) seems not to be a problem, because in the D-structure, both chi "eat" and huai "bad" have their own independent projections. This means that they have separate arguments to bear the θ-roles assigned to them.
According to previous studies (e.g., Sybesma, 1992Sybesma, , 1999) the derivation of V-V compounds is through a syntactic incorporation. It can simply be represented by (16). Note that the formation of huai-le can either be accounted for as through aspect-lowering (e.g., L. L.-S. Cheng & Li, 1991;Zhuang & Liu, 2011) or V-raising (e.g., J. Hu, 2008;Sybesma & Shen, 2006). Within the framework of minimalist program, however, there is no controversy anymore, because it is believed that huai-le is full-fledged when it comes out of the lexicon.

(16) VP
Obviously, the syntactic account here is still not fully developed. As far as we are concerned, it is confronted with at least three questions: (a) What the syntactic status small clause (SC) is? (b) Where the object of chi "eat" has gone? and (c) How to account for the two readings of Baoyu qi-leile ma (Y. Li, 1990) and the three readings of Taotao zhui-leile Youyou (Y. Li, 1995)?
In the current study, we will demonstrate that the syntactic account, with certain hypotheses taken into consideration, can overcome these problems. That is to say, it can not only solve the phenomenon that the lexical account cannot explain but also solve that phenomenon that the lexical account can explain.
The remainder of the article is organized as follows. In the "Syntactic Account Revisited" section, the syntactic account is revisited. It is shown that with certain hypothesis taken into consideration, the two readings of Baoyu qi-lei-le ma (Y. Li, 1990) and the three readings of Taotao zhui-lei-le Youyou (Y. Li, 1995) can be explained. It also demonstrates that the evidence lexicalists provide does not constitute an argument against syntactic account. The "Conclusion" section provides a conclusion.

Syntactic Account Revisited
Let us get down to the first question: What does the SC in (16) stand for? Is it a complementizer phrase (CP) or a tense phrase (TP) or neither? According to Zhuang (2014), V-DE constructions share the same D-structure with V-V compounds, and they came from the same "V NP V" construction in Middle Chinese. The reason why both V-DE constructions and V-V compounds can be found in Modern Chinese is that the coexistence of V-DE constructions and V-V compounds satisfies the need to express different emphases. Chinese is an end-focus language, in which the end of a sentence carries the natural focus. Therefore, we can examine the sentences with V-DE constructions first.

SC and V-DE Constraint on Object
Passivization provides a piece of evidence that the SC in (16) is a TP. Consider (17) The contrast between (17b) and (17c) can be explained if the V-DE constructions are assumed to have an exceptional Case marker (ECM) structure. That is to say, when the matrix clause passivized, only the subject in the embedded clause is affected. The object in the embedded clause, however, will be licensed by the verb that governs it and will not rise (to the subject position of the matrix clause). This is perhaps a result of the CP reduction. The TP following V-DE is an infinite one (Sybesma, 1992(Sybesma, , 1999Sybesma & Shen, 2006) and cannot assign a case to its subject; however, it has to grab the accusative case assigned by the V in V-DE construction.
Along the same lines, (7-10) can be understood to contain ECM-style structures. The first question is thus solved. That is, the SC in (16) stands for TP.
Of course, the fact that the SC in (16)  Because toufa "hair" in (18) and liang tiao tui "two legs" in (19) should be in Spec,TP, the only position for ta "she" in (18) or Zhaoliu in (19) should be Spec,CP. That is to say, the clauses after DE can be a CP. (This also indicates that the assumption that DE is a complementizer by some scholars, for example, Huang, 1982, is not unquestionable.) The hypothesis that the SC in (16) is a TP explains also the second question, that is, where the object of chi "eat" has gone. The (accusative) Case that is assigned by the matrix causative verb is grabbed by the subject of the embedded clause. The object of the matrix causative verb-unable to get a Casecannot be realized in situ and has to retreat, although it can be realized through certain roundabout ways, as is shown later. By "retreat," we mean that it is merely not mapped in the derivation of the V-DE construction onto the object position of ku "cry." For convenience, this phenomenon will be referred to as V-DE constraints on Objects (VDCO) henceforth. Note that DE can be a zero form. In many languages, causative morpheme is overtly spelled out, while in many other languages (like Old Chinese and English), it is covert. DE in causative/ resultative sentences of Modern Chinese is considered to be a clitic-like causative morpheme .
It is necessary to note that in derivation of resultative V-DE constructions, the cliticization of DE onto a verb will affect its argument structure in the following way: As shown in (20), the cliticization of the causative morpheme DE affects the θ-role(s) that the base verb assigns. The resultative V-DE construction actually reflects the relationship called causality. As is well known, causality is the relationship between two events: one is the cause and the other the effect. The effect is understood as a consequence of the cause. Therefore, the two θ-roles the resultative V-DE construction assigns should be Cause and Effect. As both of them are concepts of events, they are destined to be assigned to Propositions, that is, clauses, which refer to events or state of affairs in a given world.
In the literature, Causer and Causee (Huang, 1988), Cause and Affectee (Y. Li, 1995), and so on, have been proposed to suggest that these θ-roles are considered to be assigned to particular participants. This, however, incurs some problems. Below are some examples from Huang (1988) According to Huang (1988), Zhangsan in (21) and (22) is an experiencer, while in (23) and (24), it is a causee, with zhe ping jiu "this bottle of alcohol" and zhe jian shi "this matter" being their respective causers. This kind of analysis is faced with severe problems concerning θ-role assignment. As is well known in linguistic studies, θ-roles are encoded in the lexical entry of predicates. This means that the θ-roles that a verb can assign are determined before they enter the syntactic operation. Therefore, zui de "drunk-DE" in (21) and (23) should assign the same θroles (the same is true for jidong de "excite-DE" in (22) and (24)). They should not change in the way as Huang describes.
The problem raised above actually can be avoided in the present analysis. As there exists a reason-result inference between the two verbs of V-V compounds in (21-24), and they are linked by de, we propose that the two θ-roles, Reason and Result, are assigned to clauses, that is, in both (21) and (23), the Reason is assigned to Zhangsan zui "Zhangsan was drunk," and the Result is assigned to the event (Zhangsan) zhan bu qilai "Zhangsan couldn't stand up." While in both (22) and (24), the Reason is assigned to Zhangsan jidong "Zhangsan was excited," the Result is assigned to the event (Zhangsan) shuo bu chu hua lai "Zhangsan couldn't speak a word." However, this solution will be faced with the question as to where zhe ping jiu "this bottle of alcohol" in (23) and zhe jian shi "this matter" in (24) come from. It seems that they are from implicit arguments, which can be illustrated by English verbal passives (Ouhalla, 1999, p. 172): (25) The ball was kicked (by Mary). (26) The room was unoccupied (*by Mary).
According to Ouhalla (1999), the role of the by-phrase in (25) specifies the identity of the individual who performs the act described by the verb, that is, the external argument. (When the by-phrase is missing, the external argument is said to have an arbitrary interpretation, roughly paraphrased as "[some] one or other.") The fact that understanding (25) involves an external argument implies that this external argument is an implicit one (in contrast, understanding the adjectival passive in (26) does not involve one).
In a similar way, in (21) and (23), the Cause argument is (yinwei zhe ping jiu) Zhangsan zui "because of this bottle of alcohol, Zhangsan was drunk," and in (22) and (24), the Cause is assigned to (yinwei zhe jian shi) Zhangsan jidong "because of this matter, Zhangsan was excited." In (21) and (22), these implicit arguments are suppressed (or suspended), while in (23) and (24), they are spelled out. The position they occupy in the structure should be the Specifier of the topic phrase (Spec,TopP).
It seems that the semantic frame of a verb in many languages in the world can be changed by a causative morpheme in such a way as an external cause appears as the subject. For example, in Konso, a language in Southern Ethiopia, unaccusative intransitive verbs can be changed by the causative morpheme -sh into transitive verbs (27)  They retreat from their object positions because of VDCO but are realized before the V-DE constructions.

The Realization of Retreated Objects
The third issue is concerned with the realization of the retreated objects. It cannot be answered, however, until one hypothesis as well as characteristics of Chinese are taken into consideration, namely, the Split CP hypothesis and topic-prominence and focalization of Chinese.
Split CP hypothesis. In recent generative linguistic studies, a renewed interest has arisen in the roles of such notions as topic and focus in the syntax of various languages (Culicover & LMcNally, 1998;Frascarelli, 2000;King, 1995;Reinhart, 1995Reinhart, , 1996Rizzi, 1997Rizzi, , 2001Rizzi, , 2004Soh, 1998;Zubizarreta, 1998). Among them, the most influential one is the Split CP hypothesis (Rizzi, 1997(Rizzi, , 2001(Rizzi, , 2004. For many years before the Split CP hypothesis was put forth, syntactic analyses were carried out under CP hypothesis. It is successful in analyzing left peripheries of most sentences, except those illustrated by (32) (Radford, 2004, p. 328 It is obvious that CP hypothesis can be applied to analyze (32), in which no other colleague stays in Spec,CP, and would occupies the head of CP. However, once CP hypothesis is extended to (32), one problem arises then: if would is positioned under C, then there is no place for that, which, being a complementizer, usually occupies C; if that is assumed to be located under C, then which position will would be in? (Note that it is higher than TP.) To solve this problem, Rizzi (1997Rizzi ( , 2001Rizzi ( , 2004 splits CP into ForceP (Force Phrase), TopP (Topic Phrase), FocP (Focus Phrase), and FinP (Finiteness Phrase), and posits, according to Kayne's (1994) LCA (Linear Correspondence Axiom), that TopP and FocP must be positioned to the left periphery of a sentence. Accordingly, the structure of (32) can be shown as (33) (Radford, 2004, p. 328): (33) ForceP Where should TopP be positioned? Let us look at (34): (34) "He had seen something truly evil-prisoners being ritually raped, tortured and mutilated. He prayed that atrocities like those, never again would he witness." (Radford, 2004, p. 329) According to Radford (2004), that in (30) occupies the head of ForceP; atrocities like those, the proposed object of verb witness, is the topic of the sentence; and the negative constituent never again is focused. As a result of the focalization, auxiliary inversion is triggered, and would is raised to the head of FocP-The difference between focus and topic, according to Radford (2004, p. 329), is that a focus represents new information, while Topic represents given information. The structure of the that-clause in (34) is thus shown as (35): (35) ForceP Note that in (35), only three functional projections are presented, namely, ForceP, TopP, and FocP, while FinP is not included in the tree diagrams. However, this does not mean that it is invalid. The only reason that we do not put it in the tree diagrams is that whether there is Fin in Chinese is still a controversy (see Pan & Hu, 2001;Xu, 1985Xu, -1986. The proposal of a Split CP hypothesis-being a new way to solve the syntactic problems in many languages-has been applied widely in cross-linguistic study, for example, Italian (Frascarelli, 2000), Bellunese (Munaro, 2003), German (Newmeyer, 2005(Newmeyer, , 2009, Japanese (Munakata, 2006), Persian (Darzi & Beyraghdar, 2010), Cantonese (Wakefield, 2011), and so on. In English, for instance, it has been successfully applied to explain the "Directional/Locative Inversion" and the "Preposing around Be" structures, such as (36) (Frascarelli, 2000, p. 130): (36) a.
Into the room walked John. b.
Sitting in front of her was Bill.
According to Frascarelli (2000), "these structures serve to realize the in situ focusing of SUBJ constituents" (p. 131). This can be shown in two steps: Step 1: The verb raises to Foc 0 to check the [+Foc] feature.

(37) [ TopP [ FocP [ Foc' walked v [ AgrSP John i [ AgrS' t' V [ VP t i [ V' t V [ PP into the room ]]]]]]]].
At this stage, [+Foc] is still not visible, because two maximal projections, that is, John at Spec,Agr S P and the PP into the room, can possibly serve as complements of the verb in Foc 0 . Therefore, into the room must be extraposed.

t VP ]]]]]]] [+Foc]
It is important to note that Frascarelli (2000) provides evidence to show that it is not the PP that is extraposed in this construction but the whole VP (pp. 131-132). (According to Cinque, 1990, VP extraposition is allowed in English.) Because topicalization of the PP or the VP means no difference to the discussion here, we will not go into details of her evidence.
The SUBJ NP thus serves as the only complement of the proposed verb "walk" and is thus assigned Focus.
It seems that the "Directional/Locative Inversion" in Chinese can be derived in the same way. For example, the derivation of (39) and (40) is shown in (41) and (42), respectively (Impf. = Imperfective). Similar to that of English, the in situ focusing of SUBJ constituents in Chinese can be realized in two steps: first, the verbs lai le "come-Perf." in (41) and zuo-zhe "sit-Impf." in (42) raise to Foc 0 to check the [+Foc] feature; then jia li "in (my) home" in (41) and tai shang "on the stage" in (42) are extraposed to realize the focusing of SUBJ constituents in situ, that is, keren "guest" and zhuxituan "presidium."

Characteristics of Chinese
Topic-prominence in Chinese. C. N. Li and Thompson (1976, p. 461), according to a number of criteria, establish the following typological table, as in (43)

Subject-Prominent and Topic-Prominent Languages
Neither Subject-Prominent nor Topic-Prominent Languages Japanese Tagalog  Korean  Illocano  : : : : Chinese is treated as topic-prominent because its "topic is always in initial position" (C. N. Li & Thompson, 1976, p. 467).
In fact, before C. N. Li and Thompson (1976), Chao (1968) has already pointed out, "the grammatical meaning of subject and predicate in a Chinese sentence is topic and comment, rather than actor and action" (p. 69). That is to say, "The subject in Chinese is actually the topic" (Shen, 2013).
If their observations are correct, then it should not be venturesome to say that of a Chinese sentence, the first constituent can almost always be taken as the topic. That is to say, even though there is no requirement for (in situ) focusing of SUBJ constituents shown in (41) and (42), the topic position in Chinese still needs to be filled.

Focalization in Chinese.
Focalization is essentially one of the ways of focusing. Generally speaking, there are two kinds of focusing in Chinese: natural focus (also known as conventional focus) and contrastive focus.
Usually, the natural focus in Chinese is the end of a sentence (D. Liu & Xu, 1998;B. Zhang & Fang, 1996, p. 73). A more detailed discussion can be found later in the article.
The contrastive focus, in contrast, relies on some specific measures: phonological or syntactic. For example, (44) shows the phonological measure (cited from D. Liu & Xu, 1998)  In the literature, many studies can be found on syntactic focusing (e.g., R. L. Cheng, 1983;Chiu, 1993;Ernst & Wang, 1995;Huang, 1982;M. Li, 2007;Paris, 1979Paris, , 1998Paul, 2002Paul, , 2005D. Shi, 1994;Shyu, 1995Shyu, , 2001Teng, 1979;Tsai, 2000;N. Zhang, 1997). All scholars admit that there are two ways of syntactic focusing, namely, focus marking and focalization. Focus marking is shown in (45) (D. Liu & Xu, 1998) Focalization differs from focus marking in that it involves syntactic movement, as X. Liu (2004) states: The linear sequence in Chinese is a plate that is abstract, where there are certain positions which are pre-set for the focus. In the generation of sentences, adjusting word order (the linear sequence of constituents) can make certain constituents occupy the positions pre-set for the focus. This is called focalization. (p. 237) However, the syntactic status of focus has evoked heated discussion. When it comes to the landing site of the focalized phrase, until now, at least three kinds of hypotheses can be found: (1) In Spec of FocP, which is between TP and AspP (Shyu, 1995(Shyu, , 2001; (2) In Spec of AspP (M. Li, 2007, p. 25); (3) Under an inner TopicP (internal topic, lower than IP; Paul, 2002Paul, , 2005. These hypotheses are correct in some aspects, but once faced with (46), all of them are problematic.
Zhangsan, laoshi ta dang de hao. Zhangsan teacher he work DE good "As for Zhangsan, a good teacher he is." b.
Zhangsan, laoshi ta meiyou dang hao. Zhangsan teacher he have-not work good "As for Zhangsan, a good teacher he isn't." Split CP hypothesis, however, can accommodate the phenomenon illustrated by (46) better. Suppose Zhangsan in (46) is the topic (positioned in Spec,TopP), and ta "he" is the subject (positioned in Spec,IP), then there is not much doubt that laoshi "teacher" should be positioned in Spec,FocP. One may think that ta "he" can be in Spec,VP. However, (46b) excludes this possibility, because meiyou "have-not" occupies a position higher than VP, namely, NegP (cf. Zhuang & Liu, 2011).

A Syntactic Perspective on Baoyu qi-lei-le ma and Taotao zhui-lei-le Youyou
If all the above analyses are along the right lines, then let us now move on to explain the readings of Baoyu qi-lei-le ma (Y. Li, 1990) and the readings of Taotao zhui-lei-le Youyou (Y. Li, 1995), as shown in (47) and (48), respectively ("#" in Reading (47c) is to indicate that the reading is grammatically possible but semantically unacceptable).
In the model of the present study, Reading (a) and Reading (b) in (47) are derived in (49) and (50) Baoyu in (50f) is suppressed. It seems that this phenomenon can be accounted for in this way: The reason for Baoyu to drop is that it cannot get a Case from V 1 . We would call it a Pro-drop. It is necessary to note that the Pro here could be a pro or a PRO or a zero topic or none of them (but surely a phonetically empty), waiting for a substantial justification (see also Huang et al., 2009).
Generally speaking, a verb can assign at the most one accusative Case to its right. Note, again, that V 1 is the only verb that can assign a Case to it because, according to Sybesma and Shen (2006), the T (the lower one) is infinite; thus, it cannot assign a Case. It seems that dative shift verbs such as "give" are exceptions. According to Chomsky (1981), these verbs have the ability to assign a "secondary" Case to the second (non-adjacent) NP and a "primary" Case to the first (adjacent) NP. However, Baker (1988, Chapter 4) argues that dative shift is actually a type of Preposition Incorporation, namely, one of the Case is actually assigned by the preposition. Besides, Larson's (1988) Double Object Construction also denies the possibility of secondary accusative Case.
It is necessary to point out that the derivation in (47)  As is shown, ma "horse" in (53) is in a verb-copying construction, in (54) as a topic, and in (55) as a focus. What is the structure of (56)? Very possibly, it is derived from (54), whose structure is shown as (57): (57) TopP Recall that there is still another projection between TopP and TP, namely, FocP (Rizzi, 1997). If there is a [+Foc] in FocP attracting the constituent below to rise, it is not venturesome to assume that qi-lei-le "ride-tired-Perf." can be further moved up to the head of FocP. This idea has been applied by Zhuang (2013) to explain the derivation of the construction Wangmian sile fuqin "Wangmian's father died." Therefore, the derivation of (56) can be shown as (58): This also explains the phenomena in (23-24) and (30-31) above, which obviously can be derived through focalization. It is shown in (59): (59) ForceP With the discussion above, it is no longer difficult to explain all the readings of (48) except Readings (c) and (d) whose structures are shown as (60) and (61), respectively. As discussing Reading (c) of (47) seems unnecessary, here we discuss that of (48), whose derivation is shown as (60): (60) ForceP (60) shows the derivation of the reading "Youyou chased Taotao and Youyou got tired" and that in this case Taotao is topicalized while zhui-lei-le is focalized. One of the two Youyou's left at the end of the sentence is then omitted through a process called "haplology," which is a mechanism of sound change involving the loss of a syllable (or word) when it is next to a phonetically identical (or similar) syllable (or word).
The reading (48d), however, cannot be derived via a similar way as shown in (61). The reason is that in this case Taotao cannot be dropped or topicalized-According to Empty Category Principle, the Pro in this place must be bound by Youyou in Spec,TP.

(61) ForceP
Resultatives are syntactically different from embedded clauses. For one thing, the resultative clause is governed by the causative verb (V-DE), and an NP in the resultative clause is assigned an accusative Case, while the embedded clause is governed by C, and no constituent in it can be assigned a Case by the verb c-commanding it; further, if there is a Pro in the resultative clause, its binding condition is different. Let us consider (62) The two Pros (Pro 1 and Pro 2 ) are different in essence. Pro 1 is very much like a pronoun, which is free in its governing category, pointing to ta or an individual outside of the sentence, while Pro 2 can have only one reading, that is, ta ziji "himself" bound by ta "he." It has been shown that Pro 1 is actually the object of the verb in a V-DE construction. It has to retreat because of VDCO. Now the question arises: What is the nature of Pro 2 ? In accordance with the four types of empty categories defined in Chomsky (1982), namely, variable, NP-trace, PRO, and pro, it must be either a PRO or a pro. However, Pro 2 cannot be a PRO due to the fact that it is properly governed, as discussed earlier. However, Pro 2 is not as free as Pro 1 . Pro 1 in (62) may refer to ta or some individual outside of the sentence, depending on the context, while Pro 2 has only one reading: ta ziji "himself," referring to ta "he" in Spec,TP. This is further confirmed by Pro 2 , therefore, must be treated as an anaphor, to be more specific, a reflexive, although it is null.

Reexamining Li's Evidence Against a Syntactic Account
Li provides two pieces of evidence to argue against a syntactic account. The first is from duration phrases which can occur freely inside the embedded clause of the V-DE construction but are completely incompatible with resultative compounds, as shown below (Y. Li, 2005) In fact, this is not difficult to explain. As is known, duration phrases in many languages are represented by PPs. For example, in English, a preposition (e.g., for) is often employed to introduce a duration phrase, as shown in (66)  Suppose that every duration phrase, in order to be visible to get a θ-role (Visibility Hypothesis), must get a Case-It might be an oblique Case, as shown in (66), or an accusative Case. In English, a duration phrase usually appears in a PP, where it can receive an oblique Case and a θ-role. However, accusative Case cannot be excluded: As shown in the sentence He taught us three months of French, the accusative Case to be assigned to French (by the verb taught, following Chomsky's, 1981, suggestion, or by an incorporated preposition in the framework of Baker, 1988) is grabbed by the duration phrase, that is, three years. Therefore, Chinese has to resort to a preposition of to get an oblique Case. (Another example, He studied three years of French, where three years has consumed the accusative Case assigned by studied and forces French to appear in a PP.) In Chinese, however, duration phrases never appear in PPs. It means that a duration phrase does not get a Case from a preposition but from a verb directly. That is to say, once the verb moves away, the duration phrase is not able to get a Case anymore.
Another piece of evidence against a syntactic account is from Li (1997Li ( , 2000. Li argues that while most V-V compounds have corresponding V-DE constructions, a few of them do not. For example, (67) (67). This contrast can be explained by information structure. Nowadays, increasingly more scholars accept the idea that "some language-specific phenomena, such as focus and topic, concern information structuring" (Schwabe & Winkler, 2007, p. 1). Some of them even try to integrate these phenomena into the formal study of languages. Following this line of thought, some scholars explore the left periphery of the sentence (e.g., É. Kiss, 1998;Rizzi, 1997), while some the right periphery (Feng, 2003), that is, natural focus. Most languages, including Chinese, have the natural focus at the end of a sentence (D. Liu & Xu, 1998;B. Zhang & Fang, 1996, p. 73), which is different from certain languages, such as Japanese, Hungarian, and so on. In the latter, the constituent that precedes the verb at the end of a sentence immediately bear the natural focus (more detailed discussion can be found in Harlig & Bardovi-Harlig, 1988;Kim, 1988; and many others).
Taking the information structure into consideration, now we can see why (62) is kicked out. As is pointed out, in Chinese two kinds of resultatives coexist: V-DE constructions and V-V compounds. They "share the same D-structure but are different at PF" (Zhuang, 2014, p. 589), to make it specific, at the focused end. This difference is applied by Chinese speakers to convey information naturally (Zhuang, 2014). However, it is clear that (61) and (62) have the same focused end na shou ge "that song." This means that one of them must be redundant. As discussed earlier, V-V compounds are derived via V 0 -movement, while V-DE constructions are derived via DE-insertion. Obviously, the former is more economical, because insertion is always the last resort.
Actually, this kind of evidence provides some support for the present account. Consider (69): (69) Laoshi jiao-hui-le Youyou yingyu. teacher teach-know-Perf. Youyou English. "The teacher taught Youyou (English and Youyou came to) know English." Y. Li (1990) proposes that "when a compound verb heads the VP in a clause, only two structural Cases, nominative and accusative Case, can be assigned to its NP arguments" (p. 184). However, in (63), the V-V compound jiao-hui "teach-know" obviously assigns three Cases and four θ-roles (with both jiao "teach" and hui "know" assigning two θ-roles), and three of them are assigned to Case-marked arguments. The present account offers an explanation as this: one of θ-roles, namely, Patient assigned by jiao "teach" is absorbed by the causative morpheme, and the rest are assigned to three arguments, namely, Zhangsan, Lisi, and yingyu "English." The derivation of (63) is straightforward, as shown in (64) (Note that Zhangsan may move to Spec,TP or Spec,TopP, depending on one's theory. In this study, for consistency, it is assumed to move to Spec,TopP.): (70) VP

Conclusion
On the derivation of resultative V-V compounds in Chinese, there have been two approaches in the literature: the lexicalist approach and the syntactic approach. This article argues against the lexicalist one and proposes that all V-V compounds are derived via syntactic operations. It is also shown that the multiple readings of V-V compounds are actually results of object realization, focalization, and topic-prominence.
To study the V-V compounds, one needs to consider many factors, such as the Economy Principle, Information Structure, and oblique Case. Taking all these into account, we conclude that the evidence provided by the lexicalists does not constitute an argument against syntactic account.