A SYNTACTIC-SEMANTIC ACCOUNT OF CHINESE CAUSATIVE RESULTATIVE V-V

A SYNTACTIC-SEMANTIC ACCOUNT OF CHINESE CAUSATIVE RESULTATIVE V-V Jiaojiao Yao Center of Linguistics, University of Lisbon, Portugal jiao199@163.com Abstract. The Chinese Causative Resultative V-V, also referred to as “resultative verb compound”, is a construction that expresses caused-result meanings. A prominent feature of this construction exists in its word order: the cause-denoting V and the result-denoting V are in adjacency. While exhibiting compound nature, this construction also shows productivity, semantic compositionality, and semantic ambiguity. Within the theoretical framework of the Minimalist Progrram, under the general assumptions of Distributed Morphology, this study proposes a structure for this construction, attempting to account for its properties holistically. We propose that these V-Vs involve root-selecting causativization and Manner Conflation. In spite of being syntactically formed, each V-V functions in a similar way as a V0. 1. Introduction The type of construction in Mandarin Chinese (Chinese, henceforth) we will address in this study has been referred to under different names in the literature, such as “Resultative verb compounds” (Cheng and Huang 1994; Huang 1984; C. Li 2007; Thomson 1973) and “Resultative V-V compounds” (Y. Li 1990). This construction denotes a caused-result event, with two verbal components in adjacency. The order of the two compoents is iconic: the one indicating the causing eventuality precedes the one denoting the result. Consider the following example: 1 (1) Ta chang ku haizimen le. he sing cry children ASP ‘He sang, and this made the children cry.’ We will call this type of construction Causative-Resultative V-Vs (CR V-Vs), where “Causative” (C) and “Resultative” (R) refer to the involved semantic meanings, and the notion “V-V” simply refers to the fact that two verbal elements are in adjacency without any overt morpheme signaling the relation between them. Note that our notion of “V” has a broader sense than the traditional lexical category of verbs – adjectives will also be included. It is commonly known that there is not a clear line between verbs and adjectives in Chinese due to the fact that Chinese lacks a morphological system to distinguish word classes, and the adjectives can also directly function as predicates without the occurrence of any copula. Following the authors who include adjectives into verbs (e.g. Chao 1968; Y. Li 1990), we assume that if an adjective occurs in a CR V-V, it functions as a V (i.e. a stative V) to simplify our terminology, leaving the word category a separate issue. In fact, the category ambiguity will not affect the results of this study. As will be presented in the following sections, we posit that roots enter the computational system directly. Before conducting an extended analysis, we first would like to define what constructions are included in CR V-Vs according to our definition. In the previous studies, different ranges of examples have been included under a particular name. For example, the name “resultative verb compound” oftentimes not only covers what we call CR V-Vs, but also directional V-Vs (e.g. shang qu ‘go up’) and phase V-Vs (e.g. chi wan ‘finish eating’). By our definition, CR V-Vs meet the following requirements: (2) a. They denote a caused result event, consisting of two subevents: a causing subevent and a result subevent; 2 b. They are semantically compositional, with V1 denoting the cause and V2 the result, thus it can be paraphrased as ‘the result represented by V2 occurred due to the eventuality denoted by V1’; c. They are productive and formed with free constituents (in opposition to bound morphemes). First of all, we exclude the lexicalized compounds such as shuo-ming ‘say-clear: to explain’, since this combination has lexicalized and does not exhibit productivity (e.g. *han-ming ‘yell-clear: to explain by yelling’). Secondly, the directional V-Vs such as Vlai ‘come’, V-qu ‘go’, V-chu ‘exit’ and V-kai ‘away’, which are included under the name of “resultative verb compounds” in Thomson (1973), will not be included in CR V-Vs, because they do not convey a caused-result meaning, but rather a directional meaning. Moreover, the direction-denoting V2 in these instances may not be a free constituent. For example, kai in zou kai ‘walk away’ cannot be used independently with the meaning of ‘away’. Another group that will be excluded is the phase V-Vs, such as V-wan, V-jian, V-zhu and V-cheng, which denote the completion or achievement of the action denoted by V1, e.g. chi wan ‘eat finish: finish eating’. They only denote one event, which cannot be further subdivided. Moreover, V2 in these instances is usually a bound constituent and cannot independently form a predicate with the meaning of ‘complete’ or ‘finish’. The verb-complement constructions in which the second element is a modifier of the preceding verb are also excluded from CR V-Vs. For example, zhan wen ‘stand still’, where the two constituents do not enter in a cause-result relation, cannot be paraphrased into ‘being still is caused by the action of standing’ and thus is not considered a CR V-V instance. In contrast to CR V-Vs, it only contains one simplex event. 3 In spite of the quantity of works that have been conducted, it appears that a better account of CR V-Vs is still needed to account for the properties of this construction holistically. Some previous accounts consider CR V-Vs atoms in the lexicon (e.g. Cheng and Huang 1994; C. Li 2007; Y. Li 1990 et seq.), and with such account, the productivity of this construction would be a mystery. There have been various syntactic accounts (e.g. Fan 2013; Lin 2004; Sybesma 1999), but problems exist. With the proposed structures, it is usually difficult to explain why the two Vs are strictly in adjacency (3a), why individual modifier is not allowed (3b), why each V tends to be monosyllabic (3c), and why V2 is semantically constrained (3d-e). (3) a. *Ta ti men kai le. he kick door open ASP Intended: ‘He kicked the door open.’ b. Cai (* tai) la ku haizi le. dish (*too) spicy cry child ASP Intended: ‘The dish is (too) spicy, and this made the child cry.’ c. Ta nong ku/*kuqi he make cry le haizimen. ASP children ‘He made the children cry.’ d. Ta xia sha/*tiao haizimen le. he scare stunned/*jump children ASP ‘He scared the children, and this made the children stunned/*jump.’ e. Ta nong pao he make escape/*run fanren le. prisoner ASP ‘He made the prisoner escape/*run.’ 4 Furthermore, it has been observed that CR V-Vs show semantic ambiguity (4a), and some structures exhibit constraints (4b). Argument linking rules have been proposed (e.g. Y. Li 1990 et seq.), but an agreed-upon structure would be needed to make any argument linking rule workable. (4) a. Ta qi lei ma le. he ride tired horse ASP ‘He rode a horse, and this made the horse tired.’ ‘He rode a horse, and this made him tired.’ b. Ta he zui (* zhe ping) jiu le. he drink drunk (* this bottle) alcohol ASP ‘He got drunk by drinking (* this bottle of) alcohol.’ Another worthily noting issue is the transitivity alternation: while some CR V-Vs exhibit alternation (5), others do not (6). Authors such as Liu (2019) attempt to account for the argument structure of CR V-Vs with an event decomposition model, but problems remain. For example, with an event decomposition approach, it would be difficult to account for the compound nature of CR V-Vs (e.g. V-V adjacency is obligatory, and no individual modifier of V1/V2 is allowed). (5) a. Mama da sui beizi le. mom hit break cup ASP ‘Mom broke the cup.’ b. Beizi da sui le. cup hit break ASP. ‘The cup got broken.’ 5 (6) a. Zhangsan sha si Zhangsan kill le Lisi. die ASP Lisi ‘Zhangsan killed Lisi, and Lisi died.’ b. *Lisi sha si le. *Lisi kill die ASP Intended: ‘Lisi got killed and died.’ Within the theoretical framework of Minimalist Program (Chomsky 1995) and under the general assumption of Distributed Morphology (Halle and Marantz 1993, 1994; Harley and Hoyer 1999; Marantz 1997, 2013), this study will propose a syntactic structure of CR V-Vs, which can account for the properties mentioned above, such as the productivity, V-V adjacency, V-V integrity, semantic constraint on V2, semantic ambiguousness, and the mono-syllabic tendency on V1 and V2. Mostly importantly, our proposal takes a cross-linguistic perspective. Instead of proposing a language-specific structure, we will show how CR V-Vs fit into the algorithm of world languages – the essential idea of Universal Grammar and Principles and Parameters (Chomsky 1981). 2. The little v In the Minimalist Program, Chomsky (1995) articulated the hypothesis that Universal Grammar has only one computational system, and language variation reduces to differences in the lexical items that enter the computational system. That is to say, Universal Grammar makes available an inventory of functional elements from which each particular language makes its selection. When cross-linguistic variation occurs, the burden of explanation is then on the primitive building blocks of syntactic derivations. 6 Distributed Morphology (DM, henceforth; Halle and Marantz 1993, 1994; Harley and Hoyer 1999; Marantz 1997, 2013) is a framework that proposes an architecture of grammar in which the syntax is the only generative system responsible for both word structure and phrase structure. Within the DM framework, words are not atomic elements for syntactic computation. Rather, DM hypothesizes two classes of terminal nodes that enter in the syntactic derivation: “lexical roots” and “bundles of grammatical features” (functional elements), which together constitute the “Narrow Lexicon” (Marantz 1997). The lexical roots contain ‘encyclopedic’ semantic content and are acategorial; they can only get categorized and interpreted by merging with a categorizing functional head little x, such as v, n or a, which can categorize a lexical root as a verb, noun or adjective respectively, as illustrated in (7)1. The function of the roots is to provide encyclopedic meaning to modify the meaning of syntactic structures. (7) a. dance (v.) b. dance (n.) v v n √dance n √dance In the DM framework, the “Encyclopedia” lists special meanings of particular roots, relative to the syntactic context of the roots within local domains2. The “Vocabulary”, containing Vocabulary Items (VIs), provides the phonological forms for the terminal nodes from the syntax (for roots as well as bundles of grammatical features), and determines the connection between terminal nodes from the syntax and their phonological 1 The functional heads may be phonologically indistinguishable in some languages, but distinguishable in others. For example, the functional heads v and n are indistinguishable in English but are distinguishable in Finnish (Pylkkänen 2002) and Romance languages. 2 The syntactic head that projects agents defines a locality domain for special meanings. That is to say, nothing above this head may serve as the context for the special meaning of any root below this head (Marantz 1997). 7 realization. DM assumes Late Insertion: syntactic categories have no phonological content, and morpho-phonological material only enters a given derivation postsyntactically, at Morphological Structure, when VIs are inserted into the nodes containing the conglomerations of features created by syntax. In particular, the core structure of a verb phrase contains a little v head and a root, as illustrated in (7a): the little v semantically introduces an eventuality, and the root modifies this event by contributing semantic content. According to Folli and Harley (2005), the different nature (e.g. causative, unaccusative, stative, unergative) of events is not determined by the semantic information contained in the lexical roots, but it is rather determined by the different “flavors” of v heads that contain specific event-semantic content, such as vCAUSE, vBECOME, vDO3. In particular, vDO is an agentive v, which requires an animate Agent subject; vDO can take a straightforward Incremental Theme as its complement, and is a true verb of creation, as exemplified in (8a). In contrast, vCAUSE is a causative v, which only requires that the subject be a possible Cause; vCAUSE takes a state as its complement, creating essentially a resultative structure, as in (8b). (8) a. John ate the apple. b. The sea ate the beach away. vP vP DP John DP v’ v ate DP the sea v’ (CAUS) ate the apple SC v DP P the beach away (Folli and Harley 2005) 3 However, Marantz (2013) holds a different view and claims that the meaning conveyed by syntactic structure is not implemented by the various primitive meanings of little v, such as CAUSE, but rather, it is the result of the interpretation of the configuration. 8 Inspired by Folli and Harley’s different flavors of v, we claim that Chinese CR V-Vs involve a v head with CAUSE feature, namely vCAUSE. Note that CR V-Vs do not require agentivity, and the Causer is just a possible cause. As in (9a), the Causer, namely N1 ‘this song’, is not an Agent (but rather interpreted as the Theme) of the causing eventuality denoted by V1 ‘sing’. In (9b), the Causer N1 ‘puppy’s death’ does not have semantic relation with V1 ‘cry’ at all, but is just a possible cause of the whole caused-result event. (9) a. Zhe-shou ge this chang ku wo le. song sing cry I ASP ‘This song’s singing (by me or someone else) made me cry.’4 b. xiaogou de si puppy ’s ku yun wo le. death cry faint I ASP ‘The puppy’s death made me faint from crying.’ However, while assuming with Folli and Harley’s different flavors of v, we posit that the complement of vCAUSE is not limited to a Small Clause as in (8b), but can also be a root, a VP or phase. In the next section, we discuss the functional head vCAUSE in more detail. 3. Root-selecting vCAUSE In previous studies, CR V-Vs have been mostly analyzed in comparison with English resultatives such as John hammered the metal flat, and have been called resultatives in general. We use both “causative” and “resultative” in our terminology because CR V-Vs 4 Note that this sentence exhibits semantic ambiguity. The Agent of ‘singing’ can be interpreted as ‘I’ or someone else. Explanation is presented in §5. 9 are tightly related to causatives. Diachronic evidence shows that Chinese CR V-Vs did not exist from the beginning of Chinese language history, but rose in accompany with the fade-away of morphological and lexical causatives, when Chinese underwent a typological change (see Shi 2002; D. Xu 2006). In Old Chinese, causative meanings can be expressed via morphological or phonological means, as exemplified in (10). The morphological/phonological mechanism for causative formation faded away along the language development, and only some relics can be found in Modern Chinese, such as hǎo ‘good’ – hào ‘fond of’. In Old Chinese, lexical causatives can be formed from adjectives, nouns/nominals or verbs, as exemplified in (11), but become much less common in Modern Chinese5. (10) Voice/voiceless distinction (Shi 2002) k- g- jiao ‘see’ xian ‘make something visible’ jie ‘fall into pieces’ xie ‘dissect’ ji ‘be tied’ xi ‘tie’ (11) Hou qi qiangyuan (from Zuozhuan6) thick the wall ‘Make the wall thick.’ 5 Although being less common, lexical causatives do exist in Modern Chinese. Consider the examples: a. Keji fanrong le shichang technology prosper ASP market jingji. economy ‘Technology made the market economy prosper.’ b. Zhe this bu dianying huo le nan zhujue. movie famous ASP male lead ‘This movie made the male lead famous.’ 6 Zuozhuan, generally translated as The Zuo Tradition or The Commentary of Zuo, is an ancient Chinese narrative history. Most scholars believe that it was composed during the 4th century BC. 10 Chinese CR V-Vs are in line with morphological and lexical causatives that they tend to express the causation of a state/result, instead of causation of an activity. With the fadeaway of morphological/phonological mechanism and the “zero causative” strategy (i.e. the lexical causatives), the CR V-Vs occurred to compensate7. We hypothesize that they are generated from the same base structure, namely a structure involving a root-selecting vCAUSE, based on Pylkkänen’s (2002) proposal of root-selecting, verb-selecting and phaseselecting CAUSE. Pylkkänen argues that causative constructions essentially “involve the head CAUSE which combines with noncausative predicates and introduces a causing event to their semantics” (2002: 75). According to Pylkkänen (2002), crosslinguistic variation in causative constructions are attributed to two sources: Selection and Voice-bundling. Assuming with Marantz (1997) that verbs, nouns and adjectives all decompose into a category-neutral root (√) and a category-defining functional head, Pylkkänen proposed that causative heads are divided into three types: Phase-selecting, Verb-selecting, or Root-selecting. In root-selecting, CAUSE composes directly with an acategorial root (√), as illustrated in (12a). According to Pylkkänen, English zero causatives or Japanese lexical causatives8 belong to this type. In Verb-selecting, CAUSE composes with VPs lacking an external 7 It can also be understood in a reversed way: when CR V-Vs gradually became the preferred form due to the disyllabification tendency, the mono-syllabic words which were able to express complex meanings (i.e. causation of a result/state) came to have division of labors, serving as V1 or V2 in CR V-Vs. Over the time, the verbs which frequently occur at V1 position tend to denote only the activity, while those frequently occurring at V2 position tend to denote only the result/state. In this way, these verbs gradually lost the ability to denote a caused-result events independently. 8 Japanese “lexical causatives” are in opposition to the “productive causatives” in that any intervening material between CAUSE and the root is prohibited. 11 argument (12b). In Phase-selecting, CAUSE composes with a phase, which, according to Pylkkänen, is a syntactic unit that contains an external argument (12c). (12) a. Root-selecting CAUSE CAUSE √Root b. Verb-selecting CAUSE CAUSE c. Phase-selecting CAUSE CAUSE v √Root θEXT (Pylkkänen 2002) We hypothesize that the root-selecting CAUSE is involved in Chinese CR V-Vs and the existing lexical causatives (see note 5): the result-denoting root is merged with a functional head vCAUSE. Chinese CR V-Vs differ from the lexical causatives in that it takes two roots. In other words, a cause-denoting root and a result-denoting root are both present in CR V-Vs, whereas lexical causatives only contain the result-denoting root. We leave this matter to be discussed in the next section. For now, we claim that CR V-Vs also contain structure like (b). For example, the CR V-V nong zang ‘make dirty: to stain’ involves the structure in (13). (13) … vCAUSE √ zang ‘dirty’ Another important proposal by Pylkkänen (2002) is the Voice-bundling parameter. Assuming with Kratzer (1996) that external arguments are syntactically introduced by Voice, Pylkkänen proposed that while CAUSE and Voice are separate pieces in the universal inventory of functional heads, they can bundle together into a morpheme in the lexicon of a particular language. For example, English is claimed to be a voice-bundling language, in which the causative relation and the external theta-role are bundled into one 12 syntactic head and one morpheme. In contrast, in Japanese and Finnish, CAUSE and Voice are not bundled together. The contrast between non-Voice-bundling and Voicebundling is illustrated in (14). Pylkkänen claims that this proposal correctly predicts the possibility of forming unaccusative causatives (i.e. causative constructions without an external argument) in a particular language – while unaccusative causatives are possible in Japanese and Finnish, they are not possible in English. (14) a. Non-Voice-bundling causative b. Voice-bundling causative (e.g. Japanese, Finnish) (e.g. English) x x Voice [Voice, CAUSE] CAUSE (Pylkkänen 2002) Pylkkänen further hypothesizes that a causativized unergative structure should be impossible with the English-type zero causative head due to the root-selecting and Voicebundling properties, since there is not possible position for the Causee (the agent of the unergative) to be realized. This seems to be correct, as shown by the ungrammaticality of *John run the child (Intended: ‘John made the child run.). According to Pylkkänen, causativized unergative in root-selecting causatives should then be allowed in a language where Cause and Voice do not bundle together, because the separation of Voice and Cause into two different functional heads leaves more syntactic positions for argument licensing. For example, Japanese is a language which exhibit root-selecting causatives and non-Voice-bundling. Pylkkänen uses examples such as the one in (15) to show that causativization of unergatives is possible in Japanese root-selecting causatives. 13 (15) John-ga kodomo-o nak-asi-ta. John-NOM child-ACC cry-CAUSE-PAST ‘John made the child cry.’ (Pylkkänen 2002) However, Tubino Blanco (2011, Chpt. 3) argues that ‘cry’ in Japanese is in fact unaccusative, by presenting data from Tomioka (2006) showing that the Japanese verb nak ‘cry’ exhibits unaccusative properties. Furthermore, Tubino Blanco shows that in Hiaki, a language similar to Japanese in terms of Voice Bundling, unergatives do not root causativize. Based on these facts, this author points out that the fact that cry is unergative in languages like English does not entail that roots with similar encyclopedic meaning in other languages need to have the same thematic properties. In many studies (e.g. Levin and Rappaport-Hovav, 1995; Burzio, 1986; Rosen, 1989), it has been observed that there are verbs which exhibit similar semantics as unergatives in some languages, but unaccusatives in other languages. Therefore, Tubino Blanco (2011) argues that any apparently unergative root that allows root-selecting causativization has the ability to syntactically behave unaccusatively, as the case with Japanese root nak ‘cry’ in (15) above. Such ability is idiosyncratic, and this is why this phenomenon is not productive. When this happens, it is because the internal structure of these roots is compatible with unaccusative syntax. Therefore, Tubino Blanco holds that unergatives cannot by causativized by root-selecting CAUSE, regardless of the Voice-bundling feature of a particular language. This observation on root-selecting causation coincides with our observation in Chinese CR V-Vs. In Chinese CR V-Vs, the result-denoting root (i.e. V2) cannot be unergative and is rarely transitive, as illustrated by (3d-e) above and (16). Assuming with the 14 structure of root-selecting causatives, this restriction can be easily explained – there is no position for the agent of the causativized unergative/transitive to appear. (16) a. *Ta nong tiao Zhangsan le. he make jump Zhangsan ASP Intended: ‘He made Zhangsan jump.’ b. *Ta nong chi dangao haizimen le. he make eat cake children ASP Intended: ‘He made the children eat the cake.’ Interestingly, similar to Japanese lexical causatives, the Chinese verb ku ‘cry’ can be causativized in CR V-Vs, as in (4b) and (9a) above, as well as (17a). Given our hypothesis that CR V-Vs involve root-selecting CAUSE, this means, the Chinese ‘cry’ can also undergo root-selecting causativization. We follow Tubino Blanco (2011) and assume that this root is compatible with unaccusative syntax in both Chinese and Japanese. When being root-selecting causativized, they are indeed unaccusatives. We can also explain this observation in terms of telicity. Root-selecting Causation, including CR V-Vs denote causation of a result/state, in opposition to causation of an activity (e.g. John made Mary run) 9 . This implies that the causativized root should be able to convey an end result meaning, namely being telic. For example, in CR V-Vs, verbs of action such as tiao ‘jump’ and chi ‘eat’ are not allowed to serve as the result-denoting V2 (16), but the verbs ku ‘cry’ 9 This distinction can also be translated as direction vs indirect causation. Causation of an unergative is an indirect causation. For example, in John made Mary run, John did something to make this happen, but the final decision maker was Mary. That is to say, John was not able to yield the outcome without Mary’s internal decision or desire, thus it is an instance of indirect causation. In contrast, the causation of state, such as John broke the cup, tends to be a direct causation, as John is the sole trigger and is able to yield the result without the intervention of any other party. 15 and xiao ‘laugh’ are allowed (17). We believe that this is because ku ‘cry’ and xiao ‘laugh’ are able to denote an end state, being ‘sad/emotional’ or ‘happy/amused’ respectively. That means, these two verbs have both atelic and telic uses: in atelic use, they denote the action of crying or laughing; in telic use, they denote an emotional state. We suggest that it is the telic use of ku ‘cry’ and xiao ‘laugh’ that legitimates them to serve as V2 in CR V-Vs. (17) a. Ta chang ku wo le. he sing cry I ASP ‘He made me cry by singing.’ b. Ta dou xiao wo le. he amuse laugh I ASP ‘He made me laugh by amusing me.’ Transitive verbs are very restricted to occur at V2 position in Chinese CR V-Vs, but some are allowed. These verbs are mostly psych verbs and denote telic meanings, such as dong ‘understand’, hui ‘know’ and ying ‘win’. Again, it is the telicity of these verbs that legitimates them to be causativized in CR V-Vs, contrasting with the action verbs such as chi ‘eat’. Given these, it seems to be plausible to claim that a root should involve telicity to be able to be causativized by a root-selecting vCAUSE. We can summarize the constraints as follows: (18) In root-selecting causativization, the causativized root: a. should not involve agentivity b. should involve telicity 16 We argue that Chinese CR V-Vs involve at least the structure in (19), where the resultdenoting root (represented by √2) is merged with the functional head vCAUSE to bring encyclopedic meaning to the eventuality. In line with other instances involving rootselecting CAUSE, such as Japanese lexical causatives and English zero causatives, the caused eventuality tends to denote a result/state instead of an action. However, Chinese CR V-Vs differ from those instances in that the cause-denoting root (i.e. the V1) also occurs. In the next section, we will explain how the cause-denoting root is generated. (19) … vCAUSE √2 4. Manner conflation Assuming Hale and Keyser’s (1993 et seq.) l-syntactic approach, Harley (2005) proposed that the Activity verbs which are named after the instrument, such as those in (20), involve Manner Incorporation applying to the l-syntactic structure that would normally give rise to a verb of contact, involving a complement headed by an Eventdenoting root, as shown in (21)10, which represents the structure of (20a). According to Harley, through Manner Incorporation, a v is named by a Root describing the Manner in which it is accomplished. (20) a. Sue hammered the metal. b. Sue brushed the dog. c. Jill raked the leaves. 10 In Harley (2005), the provisional representation of the Manner Incorporation is via a ‘thought balloon’ applying to the v, as in (21). 17 (21) … vP DP v’ Sue √P v hammering √ (hit) DP the metal (Harley 2005) Harley further proposes that for verbs with a resultative PP, as in John pushed the cart to New York, the head v does not embed a complement-taking root as in (21), but rather embeds a Small Clause, which consists of the DP ‘the cart’ and the PP ‘to New York’. ‘Pushing’ is relegated to a mere Manner element, which is used as a pronunciation for the v via Manner Incorporation. A good paraphrase would be something like John caused the cart (to go) to New York by pushing. Hence, John pushed the cart to New York should have the following structure: (22) … vP DP John v’ Small Clause v DP pushing the cart PP P DP to New York Harley’s proposal is further fine-grained by Haugen (2009), who revised Hale and Keyser’s (2002) distinction between Incorporation (à la Baker 1988) and Conflation. According to Haugen, Incorporation involves head-movement, accounted for by Move (i.e. Copy), whereas Conflation is simply the equivalent of compounding (< 18 Merge). Following Hale and Keyser (1993), Haugen claims that denominal verbs such as dance involve Incorporation, as illustrated in (23). (23) a. b. VP V NP VP V NP dancei N à N ti dance (Haugen 2009) In contrast, the instrumental denominal verbs such as hammer in (20a), is claimed to involve Manner Conflation (a process in opposition to Incorporation): the root of hammer is merged (or conflated) as an adverbial directly into v. Other verbs of this type include brush, paddle, string, whistle, saw, anchor and comb in verb use. The root does not come from a complement position (i.e. Incorporation), but rather, is adjoined to v directly, in a similar sense to Harley’s (2005) proposal, but this process is under the notion of Conflation instead of Incorporation. We thus can present the Manner Conflation as in (24), where a Manner-denoting root adjoins to the functional head v. (24) √ v Although Haugen’s (2009) arguments were built upon analysis of denominal verbs, Mateu (2012) shows that Haugen’s (2009) distinction between Incorporation and Conflation can also apply to resultative constructions, accounting for Talmy’s (1985, 1991) Path conflated and Manner conflated typology and Washio’s (1997) distinction between strong and weak resultatives. 19 According to Talmy (1985, 1991), an event consists of a framing event and a supporting event, and the latter bears a support relation to the former. The framing event is further composed of some cross-linguistically universal semantic components – figure entity, ground entity, activating process, and association function. For example, in motion events, the framing event is composed of Figure, Ground, Motion and Path, while Manner and Cause constitute the supporting event. In a motion event, the Figure moves from a certain place via a certain Path and ends in another place denoted by Ground. Languages then differ in how the different semantic entities are encoded in surface verbs: for example, a verb can encode Motion + Path, Motion + Manner, or Motion + Figure. According to Talmy’s binary typology, Romance, Semitic, Japanese, Tamil, Polynesian, most Bantu, most Mayan, Nez Perce and Caddo are classified as Path conflated and verb-framed languages; whereas Finno-Ugric, Chinese, Ojibwa, Warlpiri, and Indo-European languages except for Romance 11 are Manner conflated and satellite-framed languages (Talmy 1991). We will first present the distinction between Path conflated type and Manner conflated type. For example, according to Talmy, Romance languages such as Spanish belong to the Path conflated type, where the co-event (e.g. Manner) being an external subordinate unit to the framing event, as shown in (25a). In contrast, English corresponds to the Manner conflated type, where the co-event (e.g. Manner) fused with the framing event, as shown in (25b). The Manner conflation pattern is not allowed in Spanish, as shown by the ungrammaticality of (25c). 11 However, Folli and Harley (2005) presented some counterexamples in Italian, arguing that some verbs of motion in Italian can form a goal of motion interpretation, although this possibility does not extend to all verbs of motion. 20 (25) a. La botella entró a la cueva (flotando). the bottle moved-in to the cave (floating) [Figure Ground] Motion+Path (Spanish) [Manner] ‘The bottle floated into the cave.’ b. The bottle floated [Figure Motion+Manner c. * La botella flotó into the cave. Path Ground] a la cueva. the bottle floated to the cave [Figure Path Ground] Motion+Manner (English) (Spanish) Intended: ‘The bottle floated into the cave.’ (adapted from Talmy 1985) On the other hand, from the perspective of the motion-framing typology, languages can be classified as either verb-framed or satellite-framed. In verb-framed languages, the core schema (i.e. Path) is encoded in the main verb; in satellite-framed languages, the core schema (i.e. Path) is syntactically realized by the “satellites”12. Spanish is a verbframed language, and as illustrated in (25a), Path is encoded in the main verb. In contrast, Path is realized by “satellites” in English, as in (25b). As shown in (25c), the satelliteframed pattern is not allowed in Spanish. As pointed by Talmy (1991), although English also has Path verbs that can directly gloss the Spanish Path verbs, their use is less colloquial and are mainly borrowed from Romance languages, such as enter, exit, ascend, descend, pass, cross, return, etc. 12 Satellites are defined by Talmy as “certain immediate constituents of a verb root other than inflections, auxiliaries, or nominal arguments”. The satellites relate to the verb root as periphery (or modifiers) to a head, and “[a] verb root together with its satellites forms a constituent in its own right, the ‘verb complex’” (1985: 102). 21 This binary typology also has impact on resultatives. In this case, the core schema (Path+Ground) has an abstract meaning (i.e. the end state). As shown in (26a), in English, a Manner conflated and satellite-framed language, the Manner ‘hammering’ is encoded in the main verb, and the result ‘flat’ appears as satellites; whereas in Spanish, a Path conflated and verb-framed language, the result ‘flat’ is encoded in the main verb aplanar ‘to flatten’, and the Manner has to occur periphrastically (26b). As expected, the Manner conflated pattern is not allowed in Spanish, as illustrated in (26c). (26) a. Mary hammered the metal flat. b. María aplanó el (English) metal ({con un martillo /martilleándolo}). (Spanish) María flattened the metal ({with a c. *María martilleó el metal hammer /hammering.it}) plano. (Spanish) María hammered the metal flat (adapted from Mateu 2012) Based on this typological distinction, Mateu (2012) proposes that the Path conflation pattern involves Incorporation of the root into P(ath) en route to V. As illustrated by the Japanese example in (27), the root √hui ‘wipe’ first incorporates into Path and then to V. (27) a. kare-wa teeburu-o he-top table-acc kirei-ni hui-ta clean wipe-past ‘He wiped the table clean.’ 22 (Japanese) b. vP P v √huii DP teeburu P P √huii X X √huii (A) √kirei In contrast, the Manner conflation pattern involves Conflation of the root with the light verb head v. As illustrated by the English example in (28), the root √dance adjoins to the light verb head and modifies the eventuality by specifying the Manner. According to Mateu, since there is no structural relation between the conflated root and the adjective, Washio’s (1997) notion of strong resultatives can be accounted for – i.e. the meaning of the verb and the meaning of the adjective are independent of each other. (28) a. The boy danced his feet sore. b. vP P v √dance v DP his feet P P A √sore (adapted from Mateu 2012) According to Mateu (2012), Romance languages such as Spanish allow Path conflation pattern (25a, 26b) because no Manner conflation is involved. In contrast, they do not allow the Manner conflation pattern, as shown previously in (25c, 26c). Therefore, Talmy’s typology can be translated syntactically: Manner conflated and satellite-framed languages such as Chinese and Germanic languages have the feature [+Manner 23 Conflation], while Path conflated and verb-framed languages such as Japanese and Romance languages exhibit the feature [-Manner Conflation]. In contrast to the unacceptability of the resultatives such as (26c), Romance languages accept “simple resultatives”, as exemplified in (29) (see Mateu (2012) for Spanish; Duarte and Oliveira (2010) for Portuguese). Note that the main verb is a light verb in both examples – poner ‘to put’ in (29a) and fazer ‘to do, make’ in (29b). We posit that these light verbs are phonetic realization of the functional head vCAUSE in these languages13. The simple resultatives are possible because no Manner Conflation is involved. (29) a. Juan puso a María John put (Spanish) nerviosa. María nervous ‘John got María nervous.’ b. O cozinheiro fez the cook o guisado queimado. made the stew (Portuguese) burnt ‘The cook made the stew burnt.’ Mateu (2012) further proposes that Chinese CR V-Vs (V-V compound in Mateu’s terminology) can be claimed to involve Manner Conflation and embedded Small Clause, the same as English resultatives such as (28). The structure proposed by Mateu for (30a) is simplified as (30b), where the result-denoting root shi ‘wet’ is encoded in an embedded Small Clause, and the cause-denoting root ku ‘cry’ is conflated/compounded with the null light verb head. 13 However, in Mateu’s (2012) proposal, these light verbs are generated inside Path, similar to the Japanese example in (27). 24 (30) a. Lisi ba Lisi shoujuan ku shi le. BA handkerchief cry wet ASP ‘Lisi cried the handkerchief wet.’ b. vP SC v √ku v ‘cry’ shoujuan √shi ‘handkerchief’ ‘wet’ (adapted from Mateu 2012) Mateu (2012) points out that in the proposed structure, word order details have been omitted for the sake of clarity. One may claim that the adjacency of the cause-denoting root ku ‘cry’ and the result-denoting root shi ‘wet’ is achieved when the DP ‘handkerchief’ moves up triggered by BA. However, note that even without the co-occurrence of BA, the V-V adjacency is still obligatory, which is one essential property of CR V-Vs. With Mateu’s structure, it is unclear how the adjacency is yielded 14 . While assuming that 14 An apparent counterargument of V-V adjacency may be the resultative V-de, since the two Vs in CR V-Vs may form V-de constructions, surfacing as V1-de-V2, or V1-bu-V2, producing the meaning of ‘able or not able to yield the result V2 by means of V1’. However, we argue that V-de constructions and CR V-Vs do not have derivational relations (pace Huang 2006; Thomspon 1973), since not all resultative V-de constructions can have counterparts in CR V-V form (i-ii), and there are CR V-Vs which do not have Vde counterparts (iii-iv). More importantly, the result part in V-de constructions is a clausal complement, but is a root in CR V-Vs. We hold that V-de and CR V-Vs are formed independently. Sometimes the same pair of Vs may occur in both constructions, and this is due to the semantic similarity between the two constructions – they both denote a caused result event. i. Ta mai de/bu qi zhe-ge bao. (V-de) she buy up this bag DE/not ‘She can/cannot afford to buy this bag.’ ii. *Ta mai qi zhe-ge bao She buy up this bag le. (CR V-V) ASP Intended: ‘She afforded to buy this bag.’ 25 Chinese CR V-Vs also involve Manner Conflation, we argue that their structure is different from the English resultatives. In particular, English resultatives involve embedded Small Clause containing the Causee and the result-denoting root, but Chinese CR V-Vs embed a result-denoting root directly, as we have proposed in the previous section. Therefore, we propose that a Chinese CR V-V involves at least the structure in (31): the cause-denoting root √1 is conflated with the head vCAUSE as a Manner adjunct to modify the causing eventuality; the result-denoting root √2 is incorporated into vCAUSE as its complement, denoting the caused result. Given this structure, the V-V adjacency is naturally yielded, since √1 and √2 are directly conflated/incorporated to the same functional head vCAUSE. (31) vCAUSEP √2 vCAUSE √1 vCAUSE In the next section, we will show how this proposed structure is able to account for the properties of CR V-Vs. 5. Explanatory power of this account iii. Zhe ge this meng ku xing wo le. (CR V-V) dream cry wake I ASP ‘This dream made me cry, and this made me wake up.’ iv. *zhe ge meng ku de wo xing le. (V-de) this dream cry DE I awake ASP ‘*This dream’s crying made me wake up.’ 26 5.1 Semantic constraint on V2 The V2 in Chinese CR V-Vs tend to be stative or unaccusative, but not unergative and rarely transitive, as illustrated in (3d-e) and (16) previously, as well as (32). This constraint is easy to explain with our account – as shown in (31), CR V-Vs involve a rootselecting vCAUSE head, and therefore exhibit the properties of root-selecting causativization that are observed cross-linguistically. As evidenced in English zero causatives, as well as Japanese and Hiaki root-selecting causatives (see Tubino Blanco 2011), root-selecting causativization does not take agentive root as complement (e.g. *John danced the child, intended ‘John made the child dance’), because there is no position for the embedded Agent to be realized. As we summarized in §3, root-selecting causativization is expected to disallow agentivity and require telicity in the caused eventuality. (32) Ta chang kun/*jiao he sing haizimen le. sleepy/*scream children ASP ‘He sang, and this made the children sleepy/*scream.’ 5.2 V-V adjacency and integrity Given the structure in (31), the V-V adjacency is naturally yielded. The cause-denoting root conflates with the functional head vCAUSE to modify the causing eventuality, and the result-denoting root incorporates into the same head. According to the Minimalist Program (Chomsky 1995), the syntactic derivation and transfer to PF and LF are conducted in a cyclic manner, based on the phase domain. That is, once a phase head is merged, its complement domain is shipped to PF and LF for phonological and semantic interpretation. Under the assumption that the categorizing 27 head little v also functions as a phase head (see Marantz 2007), once the merge with little v is completed, the root or a more complex constituent being complement of this head will be sent to PF and LF for pronunciation and interpretation. Any further operation would not have access to the individual root, but only the derived constituent headed by little v. In the context of Chinese CR V-Vs, this means that any further operation in the syntax will not have access to either √1 or √2, but rather, the whole V-V as a single unit. This can well predict the V-V integrity: as illustrated in (33), neither V1 nor V2 is extractable by wh-movement. (33) a. Ta chang ku meimei le. he sing cry sister ASP ‘He sang, and this made the sister cry.’ b. *Ta chang zenme meimei he sing how le. sister ASP ‘What happened to the sister because of his singing?’ c. *Ta zenme ku meimei le? he how cry sister ASP ‘How did he make the sister cry?’ 5.3 “Small size” constraint on V1 and V2 Another argument in favor of this structure is the tendency that each V in CR V-Vs contains one syllable only, as illustrated in (3c) above and (34). In our account, although the V-V is derived syntactically, as in (31), it is in fact a V0, just like how dance (v.) is derived (7a): the functional head v introduces the eventuality, and the roots add encyclopedic meanings to it. Since each V-V is a V0, they may exhibit some generic 28 properties observed in verbs or words. In fact, some authors have suggested that the rise of CR V-Vs during Chinese language development was driven by the disyllabification tendency of words (see Shi 2002; D. Xu 2006): in Old Chinese, verbs are mostly monosyllabic and may convey a caused-result meaning, but later due to the disyllabification, two mono-syllabic roots adjoined each other to form a unit, resulting in a V-V surface. This can be translated syntactically in our account: in Old Chinese, a result-denoting root may incorporate into a vCAUSE head to form mono-syllabic morphological or lexical causatives (without Manner Conflation); due to the disyllabification tendency, a causedenoting root conflated into this functional head to yield a disyllabic unit, and gradually this conflation became obligatory. As a result of the disyllabification tendency, in Modern Chinese, most words are disyllabic. Given that CR V-Vs are V0s, they also follow this tendency, which means, the V-V is disyllabic15, and thus each V tends to be mono-syllabic. (34) Ta he pao/*paobu lei run le. tired ASP ‘He got tired by running.’ 5.4 Scope of adverbial modifier and negation Our proposal can also correctly predict the scope of adverbials in CR V-Vs. Given the structure in (31), there is only one place to locate a verbal modifier (after the roots have been turned into a V0). That is, there can be only one possible scope for VP-modifiers – 15 However, although not common, CR V-Vs with more than two syllables do exist, where one component contains more than one syllable, for example: Ta xi ganjing le yifu. she wash clean ASP clothes ‘She washed the clothes clean.’ 29 the scope over the whole V-V. As shown in (3b) above and (35a-b), Chinese CR V-Vs do not allow adverbials to modify V1 or V2 individually. The adverbials that can cooccur with CR V-Vs should have scope over the whole V-V, as in (35c). (35) a. *Ta dashengde chang ku haizimen le. he loudly sing cry children ASP Intended: ‘He sang loudly, and this made the children cry.’ b. *Ta chang ku hen jiu he sing very long children ASP cry haizimen le. Intended: ‘He sang, and this made the children cry for a long time.’ c. Ta hen kuai he very fast he zui le. drink drunk ASP ‘Very quickly, he got drunk by drinking.’ *‘He got drunk by drinking very fast.’ Similarly, embedded negation is not possible16, as illustrated in (36), since there can be only one scope of negation – the scope over the whole V-V. 16 One apparent counterexample may be the negative potential form V-bu-V, as illustrated in the following example. We have presented previously in note 14 that the potential forms do not have derivational relations with CR V-Vs, thus the existence of negative potential form should not be considered violation of V-V integrity. Furthermore, although the negative marker 不 bu occurs between the two Vs in the negative potential form, it is not a case of embedded negation. This negative potential form expresses the meaning ‘he is not able to make the children cry by singing’, while an intended embedded negation should produce the meaning ‘he made the children not cry (anymore) by singing’. Ta chang bu ku haizimen. he sing not cry children ‘He is not able to make the children cry by singing.’ 30 (36) *Ta chang meiyou ku haizimen. he sing not cry children Intended: ‘He sang, and this made the children not cry (anymore).’ 5.5 Construction-specific constraint The constraint observed in (4c), here repeated in (37), can also be explained with our account. This constraint may appear to be related to definiteness, allowing a non-definite NP ‘alcohol’ while disallowing a definite NP ‘this bottle of alcohol’. However, the contrast of acceptability in (38) shows that it is not only about definiteness, since N2 is indefinite in both examples. (37) a. Ta he zui jiu le. he drink drunk alcohol ASP ‘He got drunk by drinking (alcohol).’ b. *Ta he zui zhe ping jiu le. he drink drunk this bottle alcohol ASP ‘He got drunk by drinking this bottle of alcohol.’ (38) a. Ta chi bao le fan. he eat full ASP rice ‘He got full by eating (rice).’ b. *Ta chi bao le mianbao. he eat full ASP bread ‘He got full by eating bread.’ 31 Now we will show how the structure proposed here can account for it. In (37a) and (38a), if N2 is removed, the sentences will maintain to be acceptable, as (39) shows. More importantly, the semantic meanings are also maintained. This is because ‘drinking’ can have the same semantic reference as ‘drinking alcohol’ does; ‘eating rice’ and ‘eating’ can also convey the same meaning17. (39) a. Ta he zui le. he drink drunk ASP ‘He got drunk by drinking.’ b. Ta chi bao le. he eat full ASP ‘He got full by eating.’ We posit that in (39), it is a simplex root that is conflated to the functional head, but in (37a) and (38a), the conflated root is a complex one. For example, (40a) represents the structure of (39a), where the conflated root is ‘drink’; (40b) represents the structure of (37a), where the conflated root is a complex one, ‘drink alcohol’. (40) a. b. vCAUSEP vCAUSE √he ‘drink’ vCAUSEP √zui ‘drunk’ √ vCAUSE √he ‘drink’ 17 √zui ‘drunk’ vCAUSE vCAUSE √jiu ‘alcohol’ Since rice is regarded as an essential part of a meal in traditional Chinese culture, chifan, which literally means ‘eating rice’, expresses the meaning of ‘having a meal’. In actual use, chifan denotes the generic meaning of ‘eating food’, regardless the type of food and whether rice is involved. 32 Note that in (40b), the root √jiu ‘alcohol’ is not a syntactic argument of √he ‘drink’, although their encyclopedic meanings imply such semantic relation. They are simply merged together to form a complex root which modifies the functional head vCAUSE. Then, why ‘drink a bottle of alcohol’ cannot form a complex root as ‘drink alcohol’ does? We hypothesize that “roots”, as shown by the name, should denote “basic” encyclopedic meanings. ‘Drinking alcohol’ denotes the generic meaning of ‘drinking’, and thus can be considered “basic”. In contrast, ‘drinking a bottle of alcohol’ specifies the form of the drink, namely, a bottle instead of a glass or something else, and thus is not “basic”. However, the notion of “basic” can be idiosyncratic and may be subject to language variation. For example, ‘eating rice’ is “basic” in Chinese, since it denotes the same meaning as the generic ‘eating’. However, this does not imply that ‘eating rice’ is “basic” in other languages. 6. Argument realization and semantic ambiguity 6.1 The Inchoative, Causative and Accusative CR V-Vs We hypothesize that the Causee of a CR V-V is always generated at the internal argument position, regardless of whether the CR V-V is transitive (with overt Causer) or intransitive. By the structure in (31), a CR V-V basically denotes a caused change of state: the causing eventuality is denoted by √1, and the resulted state is denoted by √2. When a CR V-V takes only one argument, this is an internal argument which receives the role of Causee. Semantically, the Causee entered into the state denoted by √2 due to the eventuality denoted by √1. Without an overt Causer, this internal argument moves to the matrix Subject position, surfacing as the Subject of the CR V-Vs. For example, in (41), 33 ta ‘he’ is an internal argument, which receives the role of Causee from the V-V. Since no Causer is present, this argument moves up to become the Subject on the surface. In this case, the structure is very similar to the “unaccusatives” (Perlmutter 1978; Burzio 1986) such as the English verb melt, which denotes a change-of-state event, and takes an internal argument, which may serve as the Subject when the verb is in intransitive use (e.g. The ice melted). We will call intransitive CR V-Vs like (41) the “Inchoative” type, echoing the intransitive use of melt-type of verbs. (41) a. Ta chang ku le. he sing cry ASP ‘He got to cry by singing.’ b. vP VP vCAUSEP (V0) vCAUSE √chang DP √ku ta ‘cry’ ‘he’ vCAUSE ‘sing’ Based on our observation, the Inchoative CR V-Vs can all have transitive counterparts, with an overt Causer. Some Inchoative CR V-Vs may seem to be “internally caused” (terminology from Levin and Rappaport 1994), but the Causer can still be added to yield transitive forms. For example, Liu (2019) claims the V-Vs zou-lei ‘walk tired’ and kuyun ‘cry dizzy’ do not have transitive uses. We argue that the transitive use is possible, as shown in (42). 34 (42) a. Zhe duan lu this zou lei wo le. road walk tired I ASP ‘This part of road made me tired from walking (on it).’ b. Na jian shi that ku yun wo le. thing cry dizzy I ASP ‘That thing made me dizzy from crying.’ We will call these transitive counterparts “Causative CR V-Vs”, echoing the transitive use of melt-type of verbs (e.g. The heat melted the ice). Following Krazter (1996), we assume that the Causer is generated at Spec, VoiceP. In this case, the Causee stays in situ. For example, (43) is a possible transitive counterpart of (41). (43) a. Na shou ge that chang ku ta le. song sing cry he ASP ‘That song’s singing made him cry.’ b. VoiceP DP vP VP na shou ge ‘that song’ vCAUSEP (V0) vCAUSE √chang DP √ku ta ‘cry’ ‘he’ vCAUSE ‘sing’ 35 Among the unaccusative verbs, it has been observed that some do not allow causative uses, such as the English verbs arrive and appear18: (44) a. *John arrived a group of people. Intended: ‘John made a group of people arrive.’ b. *John appeared his parents. Intended: ‘John made his parents appear.’ Within our knowledge, we have not found any Inchoative CR V-V that does not have causative counterpart. We hypothesize that the Inchoative and Causative CR V-Vs echo only a subtype of unaccusative verbs. In particular, this subtype of verbs should denote “inherently externally caused” eventuality, the essential idea in Levin and Rappaport Hovav (1994) on what is required for causative alternation. However, not all transitive CR V-Vs belong to the Causative type. That means, there are some transitive CR V-Vs that do not have intransitive counterparts. These CR V-V instances echo the telic action verbs such as the English verbs kill and cut, which convey the meaning of an activity with an end result. We call this type of CR V-Vs the “Accusative” type. Consider the following example: (45) a. Ta sha si le yi zhi xiaoqiang. he kill die ASP one cockroach ‘He killed one cockroach to death.’ 18 Note that both arrive and appear allow a locative to occur at Subject position, for example: There arrived a group of people. There appeared a letter. However, these uses do not involve causative meanings, and therefore are not related to our discussion here. 36 b. VoiceP DP vP VP ta ‘he’ vCAUSEP (V0) vCAUSE √sha DP √si yi zhi xiaoqiang ‘die’ ‘one cockroach’ vCAUSE ‘kill’ The example in (45) does not have intransitive counterpart, as shown by the unacceptability of (46)19. The Accusative CR V-Vs are agentive, and always require an Agent. Therefore, the Accusative type of CR V-Vs always surface in transitive forms. (46) *yi-zhi xiaoqiang sha si one le. cockroach kill die ASP ‘*One cockroach killed to death.’ 19 However, this sentence would become more acceptable if the NP is definite, especially if we insert a pause following it, as (i) shows. This is in fact a case of topicalization. With the simplex verb sha ‘kill’, topicalization also works, as in (ii). The topicalization strategy is commonly used in Chinese, and is pragmatic-motivated. The instances of topicalization should not be considered counterargument to our claim that the Accusative CR V-Vs do not have intransitive uses. (i) na-zhi xiaoqiang, that cockroach sha si le. kill die ASP ‘That cockroach, got killed to death.’ (ii) na-zhi xiaoqiang, sha le. that cockroach kill ASP ‘That cockroach, got killed.’ 37 We attribute to the possibility or impossibility of causative alternation to semantic reasons, following Levin and Rappaport (1994 et seq.). Each CR V-V combination is a Vo, and in line with the simplex verbs, their encyclopedic meanings (partially) determine the possible number of arguments and the (im)possibility of causative alternation. Due to the limited scope of this study, we do not address in detail what exactly are the rules that connect encyclopedic meanings with argument realization. Readers may refer to Levin and Rappaport’s works for their insights on this matter. Our main point is that any CR VV would behave either like melt or kill, depending on its encyclopedic meanings. This seems to be correct, because as a Chinese native speaker myself, I “know” whether a particular CR V-V can alternate or not, being either a conventional V-V or a novel V-V that I have never heard. For example, if we define a novel verb di as a verb denoting the action of ‘honking a car horn’, and a novel verb pa with the meaning of ‘falling face down’, I would be able to tell that the novel CR V-V di-pa can only be transitive (47a), and the intransitive counterpart is not acceptable (47b). These two combining verb roots are inexistent in Chinese, and it is not possible that I have ever heard such V-V combination. However, as a Chinese native speaker, I have the intuition about its causative alternation. This is possible if we assume that the encyclopedic meanings play an essential role in determining whether a particular verb or verb compound can alternate or not. (47) a. Ta di pa le haizi. he DI PA ASP child ‘He made the child fall face down by honking the car horn.’ b. *Haizi di pa le. child DI PA ASP 38 Intended: ‘The child fell face down due to car horn’s honking.’ To sum up, we have shown that Chinese CR V-Vs can be subcategorized into three types: the Inchoative type, the Causative type and the Accusative type. The Inchoative CR V-Vs are in intransitive form, echoing the intransitive use of the melt-type of verbs, taking an internal argument (i.e. the Causee), which then moves to the position of Subject. They denote externally caused eventuality, and therefore allow causative alternation. The transitive counterparts of the Inchoative CR V-Vs are called Causative CR V-Vs, in parallel with the transitive use of the melt-type of verbs, where the Causee would stay in situ since an overt Causer is generated at the Subject position. The Accusative CR V-Vs are similar to the kill-type of verbs, which involve agentivity and disallow intransitive uses. The structures of these three types of CR V-Vs can be simplified as those in (48). (48) a. Inchoative CR V-V b. Causative/Accusative CR V-V VoiceP vP VP vCAUSEP (V0) √2 vCAUSE √1 DP DP vP Causer Causee VP vCAUSEP (V0) vCAUSE vCAUSE √1 √2 DP Causee vCAUSE 6.2 Semantic ambiguity Semantically, the structures in (48) convey the meaning that ‘the Causee underwent a change into the state denoted by √2, and this is caused by the eventuality denoted by √1’. 39 When a Causer is present (48b), it expresses that the whole caused-result event is brought about by it. This is all that we know from this structure, and as a result, semantic ambiguity or “thematic flexibility” might occur, which has drawn much attention in previous studies (e.g. Her 2007; C. Li 2007; Y. Li 1990, 1995; Sybesma 1993). For example, the sentence presented previously in (43) is semantically ambiguous. This example can fit in the two-argument structure in (48b), where na shou ge ‘that song’ occurs at the Causer’s position, ta ‘he’ is the Causee, chang ‘sing’ occurs at position of √1, and ku ‘cry’ takes the position of √2. the conveyed meaning is that the Causee ‘he’ underwent a change of state into ‘crying’ due to the eventuality of ‘singing’, and the while change-of-state event was caused by ‘that song’. What is unclear is who sang that song to make him cry – whether it was someone else or himself. Through world knowledge, we infer that the NP ‘that song’ is a plausible Causer because it is a legit participant in the causing eventuality ‘singing’. However, that is all we know about the causing eventuality from the structure. The Agent of the activity ‘singing’ is suppressed, and that is why ambiguousness occurs. The example in (4a), here repeated as (49), also exhibits ambiguity. This sentence contains two NPs, and thus one possible structure can be the two-argument structure (48b), with ta ‘he’ at Specifier, VoiceP, and ma ‘horse’ at Complement, vCAUSEP. That means, N1 is the Causer and N2 is the Causee, yielding the interpretation in (i). (49) Ta qi lei ma le. he ride tired horse ASP i. ‘He rode a horse, and this made the horse tired.’ ii. ‘He rode a horse, and this made him tired.’ 40 However, there is another possibility, which might be a bit more “difficult” to yield – the interpretation (ii). Recall that the cause-denoting root may be a complex one, under the condition that it fulfills the requirement of conveying a “basic” meaning, such as he jiu ‘drink alcohol: drink’ (37a) and chi fan ‘eat rice: eat’ (38a). In (49), qi ma ‘ride a horse’ may also be considered to convey a “basic” meaning, referring to one category of activity or sport. Therefore, the structure (50) is also possible, where the complex root qi ma ‘rice a horse’ adjoints to the vCAUSE head to denote the causing eventuality, and lei ‘tired’ incorporates to vCAUSE to express the result. In this case, ta ‘he’ is generated at the internal argument position, receiving the role of Causee. Therefore, the conveyed meaning is ‘he became tired because of the activity of riding a horse’ – the interpretation in (ii). (50) VoiceP vP VP vCAUSEP (V0) vCAUSE √ √qi ‘ride’ √lei ‘tired’ DP ta ‘he’ vCAUSE √ma ‘horse’ Similar to (37a) and (38a), this complex root should denote a “basic” meaning, thus disallowing ‘ride this horse’ or ‘ride that horse’ to be the conflated complex root. As shown in (51), with the NP na pi ma ‘that horse’, the interpretation (ii) becomes unacceptable. 41 (51) Ta qi lei na pi ma he ride tired that le. horse ASP i. ‘He rode that horse, and this made that horse tired.’ *ii. ‘He rode that horse, and this made him tired.’ Another case of ambiguity is presented in (52). The interpretation (i) corresponds to the two-argument structure (48b), with zhe nühai ‘this girl’ generated at Spec, VoiceP, being the Causer, and wo ‘I, me’ at the Complement position of the vCAUSE head, receiving the role of Causee, as illustrated in (53a). By this structure, the conveyed meaning is that ‘I became tired because of the eventuality of chasing, and it was this girl who made all this happen.’ Note that this structure does not necessarily imply that ‘this girl’ is the Agent of ‘chasing’. We only know that ‘this girl’ is a Causer of the whole caused-result event. This means, there may be another possibility, namely the interpretation (iii) that the causing eventuality is ‘this girl is chased’ instead of ‘this girl is chasing’. This is possible in reality: imagine that you chased a girl, and as a result you got tired. Therefore, the structure (53a) yields two possible interpretations, (i) and (iii). (52) Zhe nühai zhui this girl lei wo le. chase tired I ASP i. ‘This girl chased me, and this made me tired.’ ii. ??‘This girl chased me, and this made her tired.’ iii. ‘I chased this girl, and this made me tired.’ iv. *‘I chased this girl, and this made her tired.’ 42 (53) a. Structure of (i) and (iii) b. Structure of ??(ii) VoiceP DP VoiceP vP vP VP zhe nühai ‘this girl’ vCAUSEP (V0) vCAUSE √lei ‘tired’ VP DP vCAUSEP (V0) wo ‘me’ vCAUSE √zhui vCAUSE ‘chase’ √ √zhui √lei ‘tired’ DP zhe nühai ‘this girl’ vCAUSE √wo ‘chase’ ‘me’ The interpretation (ii) is quite marginal. This interpretation is based on a structure similar to (37a), (38a) or (50), where the conflating root is a complex one, as illustrated in (53b), where zhui wo ‘chase me’ is supposed to form a complex root to conflate to the vCAUSE head. Marginality occurs because the meaning of zhui wo ‘chase me’ is not so “basic” as that of chi fan ‘eat rice: eat’ or he jiu ‘drink alcohol: drink’, thus is hardly to be considered a legit complex root. This interpretation is very difficult to yield, and may not be accepted by some native speakers. Regarding the interpretation (iv), which is unacceptable, zhe nühai ‘this girl’ is supposed to take the role of Causee but occurs at the Subject position, thus the corresponding structure is supposed to be the Inchoative one in (53b), involving a complex root. The interpretation (iv) would imply that zhui wo conveys the meaning of ‘I chase’ instead of ‘I am chased’. This imposes problem, since the linear order of zhui ‘chase’ preceding wo ‘I, me’ cannot yield the interpretation of ‘I chase’, but ‘chase me’. However, regardless of the semantic meaning, whether zhui wo can serve as a complex root is already problematic. Therefore, the interpretation (iv) is ruled out. 43 To sum up, from the structure of a CR V-V, we only know what is the causing eventuality (without knowing the roles of the participants) and what is the result (always predicated of the Causee), and that is why semantic ambiguity occurs. If the Subject is interpreted as Causee, the structure should be an Inchoative one, as in (48a); if the Object is interpreted as Causee, the structure is a Causative/Accusative one, as in (48b). Sometimes, √1 in (48a) may be a complex root, but this is subject to constraint – this complex root denotes a somewhat “basic” meaning, namely one generic category of human activity. When this constraint is violated, the structure would be marginal or unacceptable at all. Moreover, a possible interpretation should be culturally recognized, which means, such caused-result event should be possible in reality. 7. Advantages of this approach In contrast to the lexicalist accounts (e.g. Cheng and Huang 1994), which treat each CR V-Vs as atoms in the lexicon, we posit that they are syntactically formed via incorporation and conflation of roots into a functional head vCAUSE. Compared to the lexicalist approach, our proposal can better account for the productivity and semantic compositionality of CR V-Vs, while maintaining their compound nature – although they are syntactically formed, they are actually V0s and thus exhibit properties of words and compounds. Our proposal also contrasts with the previous syntactic decomposition accounts of CR V-Vs, which propose that CR V-Vs decompose into two or more functional heads (e.g. Fan 2013; Lin 2004; Liu 201920). Such approach may be able to account for the semantic 20 According to Liu (2019), the two roots in V-Vs such as kan dao ‘cut fall’ are generated under different heads, vCAUSE and vBECOME but the two roots in instances such as he zui ‘drink drunk’ are both generated under the head vBECOME. 44 compositionality and argument realization, but there is one problem: these separate heads would yield more than one scope for adverbial modifiers. However, according to our observation, CR V-Vs do not allow individually modifying adverbials, and any possible adverbial should have scope over the whole cause-result event. In this sense, our proposal is advantageous because it correctly predicts this – each CR V-V only involves only one functional head, namely vCAUSE, which is modified by two category-neutral roots. Consequently, there is only one place to locate a verbal modifier, and the only possible scope for VP-modifiers is over the whole caused-result event. Therefore, compared to the syntactic decomposition accounts, our proposal can better account for the “compound” nature of CR V-Vs, while maintaining the semantic compositionality. Some previous syntactic accounts propose that V1 in a CR V-V is generated at the position of the main verb, while V2 is generated within an embedded Small Clause, resembling the English resultatives (e.g. Sybesma 1999). Note that English resultatives exhibit a non-contiguous word order (e.g. John kicked the door open), whereas Chinese CR V-Vs require the two verbal components to be in adjacency. The Small Clause accounts could correctly yield a word order for English resultatives, but find it difficult to explain why it is a contiguous word order in Chinese. With our account, given the structure in (31), the V-V adjacency is naturally yielded. While assuming that English resultatives involve embedded Small Clause (54a), we hypothesize that in Chinese CR V-Vs, the functional head vCAUSE directly embeds a result-denoting root, as illustrated in (48), here summarized in (54b). Moreover, this structural difference not only can account for the word order differences in English and Chinese resultatives, but can also explain why intransitive resultatives are possible in Chinese, but very restricted in English. 45 (54) a. English resultative b. Chinese CR V-V VoiceP Causer VoiceP vCAUSEP (V0) √1 (Causer) vP vCAUSE vP Small Clause DP VP √2 vCAUSEP (V0) Causee vCAUSE √1 √2 DP Causee vCAUSE As shown in the following examples, in general, English does not allow intransitive resultatives (55a, c), in contrast to Chinese CR V-Vs (55b, d). Our proposal can correctly predict this contrast. The Chinese CR V-V structure (54b), in spite of involving vCAUSEP, is very much like a simple verbal predicate with an internal argument (and an external argument). When the meaning contained in vCAUSEP has accusative feature, resembling verbs like kill, the position Spec, VoiceP is obligatorily occupied by a Causer. However, when vCAUSEP exhibits unaccusative properties, resembling verbs like melt, the Causer may be absent, and in this case, Causee moves to Spec, VoiceP and becomes the Subject. In the latter case, an intransitive CR V-V may be formed. In contrast, English resultatives (54a) involve a “complex predicate” structure, with an embedded Small Clause containing the Causee and the result-denoting root. Such a structure forbids the Causee to move up to the Subject position and requires the Causer to be present. For this reason, English resultatives are in general transitive. In fact, the sentences in (55a, c) would become acceptable if we add a reflective pronoun to “fill up’ both Causer and Causee positions (see Rappaport Hovav and Levin 2001), as in (56). 46 (55) a. *He ran tired. b. Ta pao lei he run le. tired ASP ‘He got tired from running.’ c. *She sang hoarse. d. Ta chang ya she sing le. hoarse ASP ‘She got hoarse from singing.’ (56) a. He ran himself tired. b. She sang herself hoarse. Although rare, there are some intransitive cases in English resultatives, as in (57), examples from Rappaport Hovav and Levin (2001). We hypothesize that the NP at the Subject position in fact is generated at Spec, VoiceP, receiving a Causer role, and a coidentified pro is generated inside the Small Clause. This seems to be correct, because in these apparent intransitive resultatives, the NP at Subject position is interpreted as Causer and Causee simultaneously, whereas in true intransitive resultatives, this NP may be interpreted merely as a Causee. As shown in the intransitive CR V-V in (58), the NP shoujuan ‘handkerchief’ is interpreted as Causee, but not Causer. (57) a. a man grabbed and groped her and tried to get under her clothing, but she kicked free and flet. (The Courier-Journal, 21 Apr. 1998, p. 05B) b. …one of his race cars wiggled loose inside the transporter… (Kansas City Star, 1 Aug. 1997, p. D11) 47 (58) Shoujuan ku handkerchief cry shi le. wet ASP ‘The handkerchief got wet because of (someone’s crying).’ 7. Summary Within the theoretical framework of the Minimalist Program and under the general assumptions of Distributed Morphology, this study proposed a syntactic structure for Chinese Causative Resultative V-Vs (CR V-Vs). We propose that CR V-Vs involve a root-selecting functional head vCAUSE. A result-denoting root is generated at the Complement position and incorporates into this head, whereas a cause-denoting root conflates to this head as an adjunct, modifying the Manner of the causing eventuality. In spite of being syntactically formed, each CR V-V in fact behaves like a single verb – some resemble the melt-type of verbs, while other are like the kill-type of verbs. The proposed structure is able to account for various properties of CR V-Vs, including the VV adjacency, V-V integrity, the productivity, semantic compositionality, the tendency of each component being mono-syllabic, the transitivity alternation, as well as constructionspecific constraints. Our account is based on cross-linguistic variation. It involves rootselecting causativization, which is also attested in languages such as English, Japanese and Hiaki. It also contains Manner Conflation, which is possible in Germanic languages and Chinese, but not in Japanese and Romance languages. While both English and Chinese exhibit resultatives, differences exist, namely the word order and the possibility of exhibiting intransitive instances, which our account have also provided an explanation. Compared to the previous accounts, our proposal seems to be more capable to account for the properties of CR V-Vs holistically. However, future works are needed to fine- 48 grain a few issues. For example, why the resulted word order of (40b) is he zui jiu ‘drink drunk alcohol’ instead of he jiu zui ‘drink alcohol drunk’? Should instances such as kan dong ‘read understand’ and xue hui ‘study know’, as presented in (59), be considered CR V-Vs? If so, how can they fit in the proposed structure? (59) a. Ta kan dong zhe-ben shu le. He read understand this book ASP ‘He got to understood this book from reading it.’ b. Ta xue hui yingyu le. he learn know English ASP ‘He got to know English from learning it.’ Another interesting issue to be explored is the co-occurrence of BA-construction with CR V-Vs such as the one in (30a). There are some V-V instances which might sound strange to Chinese native speakers’ ears with the canonical word order N1 V1 V2 N2, but become more acceptable when BA co-occurs. References BAKER, M. 1988. Incorporation. Chicago: University of Chinago Press. BURZIO, L. 1986. Italian Syntax: A Government-Binding Approach. Dordrecht: Reidel. CHAO, Y.-R. 1968. A Grammar of Spoken Chinese. Berkeley, CA: University of California Press. CHENG, L. L.-S. & HUANG, C.-T. J. 1994. On the Argument Structure of Resultative Compounds. In Honour of William S-Y. Wang: Interdisciplinary Studies on Language and Language Change, ed. M. Y. Chen & O. J.-.L. Tzeng, 187 - 221. Taipei: Pyramid Press. 49 CHOMSKY, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. CHOMSKY, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. CHOMSKY, N. 2000. Minimalist inquires. Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, ed. R. Martin, M. Michaels, & J. Uriagereka, 89-155. Cambridge, MA.: MIT Press. DUARTE, I. & OLIVEIRA, F. 2010. Particípios resultativos. Textos Seleccionados do XXV Encontro Nacional da Associação Portuguesa de Linguística, 397-408. Lisboa: Colibri. FAN, S.-Y. 2013. Argument structure in Mandarin Chinese: a lexical-syntactic perspective. PhD thesis, Universidad Autónoma de Madrid. FOLLI, R. & HARLEY, H. 2005. Consuming results in Italian and English: flavors of v. Aspectual Inquiries, ed. P. Kempchinsky & R. Slabakova, 95-120. Dordrecht: Springer. HALE, K. & KEYSER, S. J. 1993. On the Argument Structure and the Lexical Expression of Syntactic Relations. The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. K. Hale & S. J. Keyser, 53-109. Cambridge: MIT Press. HALE, K. & KEYSER, S. J. 2002. Prolegomenon to a Theory of Argument Structure. Cambridge, Mass.: MIT Press. HALLE, M. & MARANTZ, A. 1993. Distributed Morphology and the pieces of inflection. The View from Building 20, ed. K. Hale & S. Keyser, 111-176. Cambridge, MA.: MIT Press. HALLE, M. & MARANTZ, A. 1994. Some key features of Distributed Morphology. MIT working papers in linguistics 21, 275-288. 50 HARLEY, H. 2005. How do verbs get their names? Denominal verbs, manner incorporation and the ontology of verb roots in English. The Syntax of Aspect. Deriving Thematic and Aspectual Interpretation, ed. N. Erteschik-Shir & T. Rapoport, 42-64. Oxford: Oxford University Press. HARLEY, H. & NOYER, R. 1999. Distributed Morphology. Glot International 4, 3-9. HAUGEN, J. D. 2009. Hyponymous Objects and Late Insertion. Lingua 119, 242-262. HER, O.-S. 2007. Argument-function mismatches in Mandarin Resultatives: A lexical mapping account. Lingua 117, 221-246. HUANG, C.-T. J. 1984. Phrase structure, lexical integrity and Chinese compounds. Journal of Chinese Language Teachers Association 19(2), 53–78. HUANG, C.-T. J. 2006. Resultatives and unaccusatives: a parametric view. Bulletin of the Chinese Linguistic Society of Japan 253, 1-43. KRATZER, A. 1996. Severing the external argument from its verb. Phrase Structure and the Lexicon, ed. J. Rooryck & L. Zaring, 109-137. Dordrecht: Kluwer. LEVIN, B. & RAPPAPORT HOVAV, M. 1994. A preliminary analysis of causative verbs in English. Lingua 92, 35-77. LEVIN, B. & RAPPAPORT HOVAV, M. 1995. Unaccusativity: at the Syntax-Lexical Semantics Interface. Cambridge, MA.: MIT Press. LI, C. 2007. Mandarin resultative verb compounds: Where syntax, semantics, and pragmatics meet. Ph.D. dissertation, Yale University. LI, Y. 1990. On V-V Compounds in Chinese. Natural Language and Linguistic Theory 8, 177-207. LI, Y. 1995. The thematic hierarchy and causativity. Natural Language and Linguistic Theory 13, 255-282. 51 LIN, J. 2004. Event structure and the encoding of arguments: The syntax of the Mandarin and English verb phrase. Doctoral dissertation. Massachusetts Institute of Technology. LIU, J. 2019. The syntax of VV resultatives in Mandarin Chinese. Doctoral dissertation, University of Victoria. MARANTZ, A. 1997. No escape from syntax: Don't try morphological analysis in the privacy of your own lexicon. University of Pennsylvania Working Papers in Linguistics 4(2), 201-225. MARANTZ, A. 2007. Phases and words. Phases in the Theory of Grammar, ed. S. Choe et al., 191-222. Seoul: Dong In. MARANTZ, A. 2013. Verbal argument structure: Events and participants. Lingua 130, 152168. MATEU, J. 2012. Conflation and Incorporation Processes in Resultative Construction. Telicity, Change, and State: A Cross-Categorial View of Event Structure, ed. V. Demonte & L. McNally, 252-278. Oxford: Oxford University Press. PERLMUTTER, D. M. 1978. Impersonal passives and the Unaccusative Hypothesis. Proceedings of the 4th Annual Meeting of the Berkeley Linguistic Society, Vol. 4, 157-190. PYLKKÄNEN, L. 2002. Introducing arguments. Ph.D. dissertation, MIT, Cambridge, Mass. RAPPAPORT HOVAV, M. & LEVIN, B. 2001. An event structure account of English resultatives. Language 77(4), 766-797. ROSEN, S. T. 1989. Argument structure and complex predicates. Ph. D. thesis, Brandeis University. 52 SHI, Y. 2002. The Estabilishment of Modern Chinese Grammar: The Formation of the Resultative Construction and Its Effects. Amsterdam-Philadelphia: John Benjamins. SYBESMA, R. 1993. On the ambiguity of riding tired. Linguistics in the Netherlands 10(1), 129-140. SYBESMA, R. 1999. The Mandarin VP. Dordrecht: Kluwer Academic Publishers. TALMY, L. 1985. Lexicalization patterns: Semantic structure in lexical forms. Language Typology and Syntactic Description, Vol. 3: Grammatical Categories and the Lexicon, ed. T. Shopen, 36-149. Cambridge, UK: Cambridge University Press. TALMY, L. 1991. Path to Realization: A Typology of Event Conflation. Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on The Grammar of Event Structure, 480-519. THOMPSON, S. A. 1973. Resultative Verb Compounds in Mandarin Chinese: A Case for Lexical Rules. Language 49(2), 361-379. TOMIOKA, N. 2006. Resultative Constructions: Cross-Linguistic Variation and the Syntax-Semantics Interface. Doctoral dissertation, McGill University. TUBINO BLANCO, M. 2011. Causatives in Minimalism. John Benjamins Publishing. WASHIO, R. 1997. Resultatives, compositionality and language variation. Journal of East Asian Linguistics 6, 1-49. XU, D. 2006. Typological Change in Chinese Syntax. New York: Oxford University Press. 53