漢語動結式中的中心語與論元體現 - 政大學術集成

全文

(1)國立政治大學語言學研究所碩士論文 National Chengchi University Graduate Institute of Linguistics Master Thesis. 指導教授：何萬順 Advisor: One-Soon Her. 立. 政治大. ‧ 國. 學. 漢語動結式中的中心語與論元體現. Headedness and Argument Realization in Mandarin Resultative Compounds. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. 研究生：吳郁賢撰 Student: Yu-Hsien Wu 中華民國九十九年六月 June, 2010. v.

(2) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. ii. i n U. v.

(3) HEADEDNESS AND ARGUMENT REALIZATION IN MANDARIN RESULTATIVE COMPOUNDS. 立. BY 政治大 Yu-Hsien Wu. ‧. ‧ 國. 學 Graduate Institute of Linguistics. n. al. er. io. sit. y. Nat. A Thesis Submitted to the. in Partial Fulfillment of the. v i n Ch Requirements for the Degree of engchi U Master of Arts. June, 2010.

(4) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v.

(5) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. Copyright © 2010 Yu-Hsien Wu All Rights Reserved. iii. v.

(6) Acknowledgements. I would like to give my greatest thanks to my advisor, Professor One-Soon Her, for his instruction and guidance during my study at the National Chengchi University.. 政治大. His patient guidance inspires many valuable ideas, leading me to the completion of. 立. my thesis. His insights and knowledge sharpen my perspectives about this research.. ‧ 國. 學. Also, I want to show my gratitude to my committee members, Professor Feng-fu Tsao. ‧. and Professor Hsun-huei Chang, who so kindly offer many helpful suggestions that I. Nat. io. sit. y. could improve the rationale and organization of my thesis and open up the horizon for. er. further research. In addition, I appreciate Professor Chun-Yin Doris Chen’s and. al. n. v i n C h encouragement and Professor Chiung-chih Huang’s constant e n g c h i U advice during the process of my research.. My thanks go to my wonderful family and friends. My friends are Patty Chen, Shiuan-Feng Huang, Wan-Chen Huang, Karen Liao, Pei-Shan Lu, Li-Hsin Ning, Jennifer Yang, Yi-Ting Xie, and many others, who share their ideas or thoughts with me about academic studies and everyday trifles. They help tackle my worries and cheer me up. Their company certainly adds some relief and fun to my days throughout. iv.

(7) the school years. As for my family, I would like to extend my thanks to my uncles and aunts, who care much about me, and also one of my cousins, Edward Wu, who shares many interesting moments in life with me. Finally, to my dear parents, I am deeply indebted. I am especially grateful for their effort of raising me and educating me and their generous and wholehearted support for the whole time.. 政治大. Without any of the people mentioned above, the completion of this thesis would. 立. not be possible. Therefore, I would like to dedicate my thesis to all of them.. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. v. i n U. v.

(8) Table of Contents. Acknowledgements .....................................................................................................iv. 政治大. Chinese Abstract ...................................................................................................... viii. 立. English Abstract ...........................................................................................................x. ‧. ‧ 國. 學. Chapter. 1. Introduction..............................................................................................................1. Nat. io. sit. y. 2. Literature Review ....................................................................................................5. n. al. er. 2.1 Characteristics and Classifications ..............................................................5. i n U. v. 2.2 Argument Realization....................................................................................9. Ch. engchi. 2.2.1 Li’s Analysis (1990, 1995) ...................................................................9 2.2.2 Her’s Analysis (2004, 2007) ..............................................................15 3. Formulating Argument Structures and Argument Realization.........................23 3.1 The Linking of Composing Roles ...............................................................23 3.2 Suppression in a Composite Role ...............................................................27 3.3 Causativity Distribution ..............................................................................29 3.4 Argument-Function Mapping .....................................................................31 3.4.1 Transitive-Transitive Verb Compounds..........................................32 3.4.2 Transitive-Intransitive Verb Compounds .......................................36 3.4.3 Intransitive-Transitive Verb Compounds .......................................39 vi.

(9) 3.4.4 Intransitive-Intransitive Verb Compounds ....................................40 4. Headedness .............................................................................................................43 4.1 V1 as Head ....................................................................................................43 4.1.1 Li’s Analysis (1990, 1993, 1995, 1999) .............................................43 4.1.2 Cheng and Huang’s Analysis (1994)................................................45 4.2 V2 as Head ....................................................................................................48 4.2.1 Shen’s Analysis (1992) ......................................................................48 4.2.2 Tai’s Analysis (2003)..........................................................................50 4.3 Double-headedness.......................................................................................53 4.3.1 Gu’s Analysis (1992)..........................................................................53. 政治大 4.4.1 Huang and Lin’s Analysis (1992).....................................................55 立 4.4.2 Li’s Analysis (2008, 2009) .................................................................59. 4.4 Headlessness .................................................................................................54. ‧ 國. 學. 4.5 An Alternative of Determining Headedness ..............................................61. ‧. 5. Argument Realization and Headedness in Lexical Mapping Theory ...............71. sit. y. Nat. 5.1 Lexical Mapping Theory .............................................................................71. io. er. 5.2 The Application ............................................................................................75 5.3 Locative Inversion in Resultative Constructions ......................................89. al. n. v i n Ch 6. Conclusion ..............................................................................................................95 engchi U References ...................................................................................................................97. vii.

(10) 國. 立. 政. 治. 大. 學. 研. 究. 所. 碩. 士. 論. 文. 提. 要. 研究所別：語言學研究所論文名稱：漢語動結式中的中心語與論元體現指導教授：何萬順教授研究生：吳郁賢. 立. 政治大. 論文提要內容：(共 1 冊，16,429 字，分 6 章 14 節). ‧ 國. 學. 有關漢語動結式的議題，中心語與論元體現皆呈現了複雜的現象。就中心語. ‧. 這部份，過去研究(Cheng & Huang, 1994; Gu, 1992; Huang & Lin, 1992; Y. Li, 1990,. Nat. io. sit. y. 1993, 1995, 1999; C. Li, 2008, 2009; Shen, 1992; Tai, 2003)的論點相當分歧，顯示. er. 漢語動結式中心語的位置仍有討論的空間；至於論元體現方面，由於論元與詞類. al. n. v i n Ch 間的互動會帶出許多不同的語意和句法表現，而先前的研究(Her, 2004, 2007; Li, engchi U 1990, 1995)尚未足以完整的解釋這些情形。. 本研究主要從詞彙的角度來探討中心語和論元體現這兩個議題。首先，依照 Her (2004, 2007)的分析方式，本研究嘗試排列出漢語動結式合法的論元結構，並利用這些論元結構描述各個動結式的論元體現，解釋可能的語意及句法結構。接著，同樣根據所提出的合法論元結構，以顯現的論元為主，並採用有關判斷中心語的假設(Zwicky, 1984; Y. Li, 1990, 1995; Chung, 2006; C. Li, 2008, 2009)，整理出. viii.

(11) 漢語動結式中心語的位置。最後，詞彙功能語法中的詞彙映照理論驗證了本研究前半部分對論元體現與中心語的討論，這個理論也另外解釋漢語動結式處所詞倒置的現象。. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. ix. i n U. v.

(12) Abstract. Two issues regarding Mandarin resultative compounds, headedness and argument realization, present a complex phenomenon. For one thing, extensive. 政治大. studies (Cheng & Huang, 1994; Gu, 1992; Huang & Lin, 1992; Y. Li, 1990, 1993,. 立. 1995, 1999; C. Li, 2008, 2009; Shen, 1992; Tai, 2003) concerning headedness fail to. ‧ 國. 學. critically determine the head of a Mandarin resultative compound. For another,. ‧. despite previous research (Her, 2004, 2007; Li, 1990, 1995), the interaction between. Nat. sit er. io. inconclusive.. y. arguments and grammatical functions of Mandarin resultative compounds remains. al. n. v i n To settle the foregoing matters,Cthe U is to probe into hpurpose e n gofc this h i thesis. argument realization and headedness from a lexicalist approach, aiming to provide a full account of both issues. In the first part of this thesis, following Her’s (2004, 2007) analysis, the thesis focuses on formulating systematic feasible argument structures for Mandarin resultative compounds, then examining the argument structures of a resultative compound to explain its possible readings and syntactic representations. Based on the available argument structures proposed in the first half, the second part. x.

(13) of the thesis investigates the headedness of Mandarin resultative compounds, suggesting that the head can be determined when arguments are overt. The criterion for headedness that is adopted in this thesis involves assumptions proposed by Zwicky (1984), Y. Li (1990, 1995), Chung (2006), and C. Li (2008, 2009). Finally, the thesis demonstrates that both issues of argument realization and headedness are well governed by Lexical Mapping Theory and that Lexical Mapping Theory further. 政治大. clarifies constructions with locative inversion of Mandarin resultative compounds.. 立. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. xi. i n U. v.


(15) Chapter 1 Introduction. A type of verb-verb compound is commonly observed in Mandarin Chinese. This type of V-V compound denotes an action-result sequence. The first verb typically represents the action, while the second verb refers to the result caused by the action. Viewed as a whole, the compound describes the event of an action followed by its. 政治大. outcome. Given this cause-consequence relation, such V-V compound is called the resultative compound.. 立. ‧ 國. 學. The headedness and argument realization of Mandarin resultative compounds. ‧. present a complex phenomenon. For one thing, since a resultative compound consists. Nat. io. sit. y. of two verbs, researchers have endeavored to determine which verb should stand as. er. the head by examining resultative compounds and their sentence constructions from. al. n. v i n different aspects. Sentences C in (1) evidence toU the complexity of headedness. The h elend ng chi compound in (1a) can be considered left-headed because the first verb subcategorizes for all arguments required by the whole compound. However, the verb on the left does not subcategorize for the sole argument in (1b), which casts doubt on the original left-headed perspective. If it is the verb on the right that serves as the center of the compound, still, this view fails to explain the structure in (1a) is right-headed. Consequently, the headedness of Mandarin resultative compounds remains a. 1.

(16) 2. debatable topic. (1) a.. Zhangsan. da-si-le. Lisi. John. hit-die-ASP. Lee. ‘John hit Lee to the extent of making Lee die.’. b.. ta-de. yanjing. he-GEN eye. ku-xia-le. 政治大. cry-blind-ASP. 立. ‘He cried so much he could barely see.’. ‧ 國. 學. For another, the readings below in (2) demonstrate apparent argument-function. ‧. mismatches. Among the four readings, three of them are valid, and only one of them. Nat. io. sit. y. is inaccessible. It seems arbitrary that only the reading in (2b) is ungrammatical while. n. al. er. the rest three readings are allowed. (2). Ch. e Lisi. ngchi. Zhangsan. zhui-lei-le. John. chase-tired-ASP Lee. i n U. v. a. ‘John chased Lee to the extent of making Lee tired.’ b. *‘Lee chased John and John got tired.’ c. ‘John chased Lee and John got tired’ d. ‘Lee chased John and Lee got tired.’ Additionally, not all resultative compounds are able to induce so many readings.

(17) 3. as zhui-lei ‘chase-tired’ does. Consider compounds pao-lei ‘run-tired’ and ku-shi ‘cry-wet’ described in (3) and (4). They both obtain only one reading respectively. Comparing their argument-function mappings with those of (2), the three sentences apparently differ.. (3) ta. pao-lei-le. he. run-tired-ASP. 政治大. ‘He ran and he got tired.’. cry-wet-ASP handkerchief. Nat. io. sit. ‘He cried and the handkerchief got wet.’. y. shoupa. ‧. he. ku-shi-le. 學. (4) ta. ‧ 國. 立. al. er. Therefore, it is possible to assume that resultative compounds provide evidence. n. v i n C hthematic argumentsUand grammatical functions. of a complex mapping between engchi. Plenty of research has been conducted using different frameworks, trying to explain such argument distribution. The purpose of this thesis is to investigate the argument realization and headedness of Mandarin resultative compounds from another perspective, a lexicalist point of view which regards resultative compounding as a morpho-lexical process. Regarding argument realization, this thesis analyzes resultative compounds by delving.

(18) 4. into their argument structures. This approach is further supported by Her’s (2004, 2007) analysis and Lexical Mapping Theory of Lexical-Functional Grammar. As for the headedness issue, we propose that only when arguments are overt can the head feature percolation condition (Y. Li, 1990, 1995; C. Li, 2008, 2009) and definitions of headedness (Zwicky, 1984; Chung, 2006) be applied to determine the head of a resultative compound.. 政治大. The remainder of the thesis is arranged in six parts. Chapter 2 presents an. 立. overview of resultative compounds and their characteristics. It also reviews four. ‧ 國. 學. studies regarding the argument-function realization of resultative compounds: Li. ‧. (1990, 1995) and Her (2004, 2007). Chapter 3, concentrating on argument realization. Nat. io. sit. y. of resultative compounds, provides an account of the argument-function assignment,. al. er. demonstrating how the reading of each resultative compound falls out. Chapter 4. n. v i n C hcompounds. AfterUreviewing previous focuses on the headedness of resultative engchi. different analyses, we argue that the head is to be determined when the argument structure is present. In Chapter 5, our analysis concerning argument-function realization is verified when it operates under Lexical Mapping Theory. This section also describes the case of locative inversion where a locative role is added to resultative compound constructions. Finally, Chapter 6 concludes the argument-function realization and headedness of Mandarin resultative compounds..

(19) Chapter 2 Literature Review. 2.1 Characteristics and Classifications In Mandarin, a resultative compound contains two verbs, with the former specifying the cause and the latter denoting the result. Moreover, the first verb usually serves as an action verb. The second verb tends to be a stative verb or an adjectival. 政治大. verb, which is specifically restricted to Mandarin (Thompson, 1973; Li, 1990). Li. 立. (1995) regards the first verb as Vcaus and the second as Vres. Furthermore, it is found. ‧ 國. 學. that Vcaus can be either transitive or intransitive whereas Vres is typically intransitive.. ‧. However, cases where Vres is transitive do nonetheless exist.. Nat. io. sit. y. Resultative compounds can be categorized based on different criteria including. er. semantic (Li & Thompson, 1981) and syntactic (Li, 1990; Cheng & Huang, 1994).. al. n. v i n Here we focus on two more C syntax-related h e n g cclassifications. h i U In terms of transitivity,. which Li (1990) uses to classified resultative compounds, Vcaus and Vres generate four different possible combinations, each listed in (5), with examples. (5) a. V1 and V2 are both transitive.. tamen da-ying-le they. zhechang zhanzheng. fight-win-ASP this. battle. ‘They fought this battle and they won it.’. 5.

(20) 6. b. V1 is transitive, while V2 is intransitive. i. Zhangsan ting-ni-le John. zheshou. listen-bored-ASP this. ge song. ‘John listened to this song (too often) and he got bored.’. ii. Zhangsan. da-si-le. Lisi. 政治大. 立 hit-die-ASP. Lee. 學. ‧ 國. John. ‘John hit Lee and Lee died.’. ‧. Nat. sit. wan-wang-le. gongzuo. n. al. tade. Ch. he play-forget-ASP his. e nwork gchi. er. io. ta. y. c. V1 is intransitive, while V2 is transitive.. i n U. v. ‘He played and he forgot his work.’. d. V1 and V2 are both intransitive. i. ta he. xiao-feng-le laugh-crazy-ASP. ‘He laughed so much it seemed like he went crazy.’.

(21) 7. ii. ta. ku-shi-le. he. shoupa. cry-wet-ASP handkerchief. ‘He cried and the handkerchief got wet.’ If V1 and V2 are taken as a whole, resultative compounds can be grouped, based on two dimensions, aspectuality and transitivity, into four classes: unergatives,. 政治大. transitives, ergatives, and causatives (Cheng & Huang, 1994). Examples of each class. 立. 學. ‧ 國. are given in (6).. (6) a. Unergatives. ‧. ta. he-zui-le. y. Nat. n. ‘He drank and he got drunk.’. Ch. engchi. b. Transitives. ta. he-zui-le. jiu. he drink-drunk-ASP alcohol ‘He drank alcohol and got drunk.’. er. io. al. sit. he drink-drunk-ASP. i n U. v.

(22) 8. c. Ergatives. ta. lei-si-le. he exhaust-dead-ASP ‘He got extremely tired.’. d. Causatives. 政治 ta 大. zhejian shi. 立exhaust-dead-ASP. matter. ‘This matter tired him out completely.’. him. 學. ‧ 國. this. lei-si-le. ‧. Nat. io. sit. y. From Cheng and Huang’s (1994) analysis, among these four classes, unergatives. al. er. and transitives represent activities, while ergatives and causatives suggest states or. n. v i n Cfor changes of state. Generally speaking, group and the h ethenunergative-transitive gchi U ergative-causative group, their respective associations with activities and. states/changes of state determine the category of V1. The category of V2 typically belongs to stative verbs. In other words, the former group reveals a relation in which the act from V1 causes the result from V2, while the latter group indicates a situation where V2 further modifies the result of V1. In terms of the unergative-transitive group, the unergative pattern appears to.

(23) 9. alternate with its accusative counterpart, the transitive pattern that requires an internal argument. Note that the transitive pattern sometimes changes the subject of V2. For example, in (6b), the subject of V2 remains to be ta ‘he’. However, in sentences such as ta qi-lei-le liangpi ma ‘he rode two horses, and the horses got tired,’ the subject of V2 lei ‘tired’ becomes liangpi ma ‘two horses’ instead of ta ‘he.’ Likewise, for the group of ergatives and causatives, the ergative pattern also alternates with the. 政治大. causative pattern. The distinction between the two can be attributed to the presence of. 立. 2.2 Argument Realization. 學. ‧ 國. a Causer.. ‧. In addition to those characteristics mentioned above, the realization from the. Nat. io. sit. y. argument structure to grammatical functions has been examined (Li, 1990, 1995, 1999;. al. er. Her, 2004, 2007). The link between argument roles and grammatical functions is. n. v i n C h construction along considered to explain the syntactic e n g c h i U with its interpretation. 2.2.1 Li’s Analysis (1990, 1995). Li (1990, 1995) analyzes the structure of resultative compounds based on three assumptions in Government and Binding (GB) theory: a structured theta-grid, theta-identification, and the head-feature percolation, aiming at correctly assigning theta-roles to its arguments. A structured theta-grid means that thematic roles are in a hierarchical order governed by the thematic hierarchy. The mapping starts from the.

(24) 10. least prominent role to the most prominent one, from the internal argument to the external one. The theta-grid of a verb like give is illustrated in (7a). The least prominent role is shown furthest to the right and assigned first. As the degree of prominence increases, the theta-role is placed more to the left and assigned later. As a result, the prominence in a theta-grid can be represented by numbers, as in (7b). (7) a. give <agent, goal, theme> b. give <1, 2, 3>. 立. 治政(Li, 1990:179) 大. The second assumption, theta-identification, concentrates on the identification of. ‧ 國. 學. two or more theta-roles. When two or more theta-roles together indicate the same. ‧. entity, these roles are said to be identified. The identified theta-roles function as a. Nat. io. sit. y. single one for argument assignment, in order to satisfy the Theta Criterion.. al. er. Finally, the head-feature percolation 1 postulates that the head of a compound. n. v i n C h Thus it is suggested reflects the main properties of the compound. e n g c h i U that the prominence hierarchy of theta-roles from the head should be maintained as the hierarchy for the compound. According to these assumptions, Li (1995) argues for ambiguous readings in the resultative verb compound construction zhui-lei ‘chase-tired’ presented in (8). The ambiguity lies in the difference of theta-role identification.. 1. The head-feature percolation will be discussed in detail in Chapter 4..

(25) 11. (8) Zhangsan. zhui-lei-le. John. Lisi.. chase-tired-ASP Lee. a. ‘John chased Lee to the extent of making Lee tired.’ b. *‘Lee chased John and John got tired.’ c. ‘John chased Lee and John got tired’ d. ‘Lee chased John and Lee got tired.’. 政治大. Vcaus zhui ‘chase’ subcategorizes two roles (represented by numerals), and Vres lei. 立. ‘tired’ has one (represented by alphabets). Through theta-identification, the role from. ‧ 國. 學. lei ‘tired’ binds with either of the roles from zhui ‘chase.’ With the constraint of. ‧. thematic hierarchy, theta-identification produces two possible theta-grids, given in (9a). Nat. (Li, 1995:265). n. al. sit. 2-a>. Zhangsan. Lisi. b. <1-a,. 2>. Zhangsan. Lisi. Ch. engchi. er. io. (9) a. <1,. y. and (9b). (9a) and (9b) correspond to the readings in (8a) and (8c) respectively.. i n U. v. (Li, 1995:265). In (9a), with the only role <a> from Vres and the role <2> from Vcaus identified,.

(26) 12. <2-a> selects the object position. The thematic hierarchy forces the most prominent role, <1>, to fill the subject position. Hence this theta-grid predicts that John chased Lee and that Lee is the one who got tired. By contrast, in (9b), the single role <a> from Vres is identified with <1> from Vcaus. After the less prominent role, <2>, occupies the object slot, the <1-a> role is linked to the subject. The reading of John chasing Lee and John getting tired is thus derived.. 政治大. After correctly predicting the argument structure and the meaning in (8a) and. 立. 學. ‧ 國. (8c), the three assumptions further lend evidence to the non-availability of the reading in (8b). The argument structure for such reading is organized as shown in (10). (Li, 1995:265). Lisi. io. al. er. sit. y. Nat. Zhangsan. 2-a>. ‧. (10) <1,. n. v i n C hrole <2-a> is assigned (10) shows that the less prominent e n g c h i U to the subject, while the more prominent role <1> is linked to the object, which apparently violates the thematic hierarchy. It also fails to follow the head-feature percolation. If Vcaus is the head as assumed, the theta-role prominence that the head presents does not correspond with that of the whole compound. Li’s proposal seems to successfully explain the argument structure of resultatives; however, it does raise some problems. First, the availability of (8d), whose argument.

(27) 13. structure is given in (11), cannot be accounted for by Li’s assumptions. 2>. (11) <1-a,. Zhangsan. (Li, 1995:265). Lisi. Similar to the mismatch in (10), (11) is not governed by the thematic hierarchy and the head-feature percolation, but it remains grammatical. If the thematic hierarchy. 政治大. is complied with, this reading should not be valid.. 立. In order to tackle this problem, Li (1995) proposes a causative hierarchy which. ‧ 國. 學. can override the thematic hierarchy. The causative hierarchy includes two causative. ‧. roles, Cause and Affectee, with Cause being more prominent than Affectee. For. Nat. io. sit. y. causative roles assignment, causative roles, also called c-roles, are only assigned. al. er. when a causal event is present overtly. With an overt causal event, c-roles are assigned. n. v i n C hto the conditions listed to syntactic positions according e n g c h i U below:. (12) a. The argument in the subject position receives the c-role Cause from a resultative compound only if it receives a theta role only from Vcaus. (Li, 1995:267) b. The argument in the object position receives the c-role Affectee from a resultative compound if it receives a theta role at least from Vres. (Li, 1995:267).

(28) 14. Moreover, Li (1995) points out that causative hierarchy can override the thematic hierarchy, proposing the following as (13) describes: (13) Theta roles can be assigned contrary to the thematic hierarchy if the arguments receiving them are assigned c-roles in ways compatible with the causative hierarchy. (Li, 1995:269) This assumption suggests that when a causal relation is observed, the causative. 政治大. hierarchy outweighs the thematic hierarchy, justifying readings such as (8d) which. 立. were originally considered invalid due to thematic hierarchy violation.. ‧ 國. 學. Given the conditions above, the problem with (8d) is solved, as repeated in (14).. ‧. First of all, given the fact that causal relation is available, based on the c-role. Nat. io. sit. y. assignment detailed in (12), the causative hierarchy assigns Cause to <2> and. al. er. Affectee to <1-a>. Regarding the thematic hierarchy, the agent role <1> is mapped to. n. v i n Cthe the object and the patient role <2> to U the thematic h esubject, h i violates n g cwhich. hierarchy. However, with the condition from (13) specifying that the causative hierarchy overrides the thematic hierarchy, the inversion of theta roles is allowed. Thus, <2>, as Cause, maps to the subject, and <1-a>, as Affectee, selects the object. The argument structure is therefore acceptable..

(29) 15. (14) Zhangsan. zhui-lei-le. John. Lisi.. chase-tired-ASP Lee. ‘Lee chased John and was made tired (by John).’ <1-a. 2>. SUB. OBJ. Zhangsan. Lisi. Cause. Affectee. 學. ‧ 國. 立. 政治大. ‧. Although the idea of causative hierarchy helps account for the reading in (8d), it. Nat. 2.2.2 Her’s Analysis (2004, 2007). al. er. io. sit. y. is a stipulation which appears to be specifically designed to handle such a reading.. n. v i n From the perspective ofCLexical-Functional Her (2004, 2007) explains U h e n g c h i Grammar,. the mapping from the argument structure to syntactic structure of the resultative compound zhui-lei ‘chase-tired’ by adopting Lexical Mapping Theory. First, Her (2004, 2007) makes several adjustments in the original Lexical Mapping Theory for constructions with resultative compounds. The mapping between arguments and grammatical functions is done based on the rules shown in (15) and (16)..

(30) 16. (15) Intrinsic Classification of Argument Roles for Functions (IC): a. patient/theme  [－r] b. secondary patient/theme  [＋o] (16) Default Morphosyntactic Classification of Argument Roles (DC): θ≠ ˆ , θ  [＋r] Second, to ensure a strict one-to-one argument-function linking, as the. 政治大. θ-Criterion does, Her (2004, 2007) proposes a unified mapping principle (UMP).. 立. (17) The Unified Mapping Principle (UMP):. ‧ 國. 學. Map each role in a-structure to the highest compatible* AF. ‧. available+. 2. Nat. io. sit. y. (*An AF is compatible if and only if it contains no. al. er. conflicting features. +An AF is available if and only if it is. n. v i n Cand no fully specified by a role to a higher role.) h enotnlinked gchi U. Despite the fact that a SUBJ should be present in a clause as the Extended Projection Principle or the Subject Condition requires, the UMP does not insist on this point. This point is further supported by other researchers (Babby, 1989; Simpson, 1991; McCloskey, 1999). A clause without a subject is allowed, simply considered marked or construction-specific. Still, SUBJ is the most prominent grammatical. 2. Her modified the UMP after a class discussion in 2009. The UMP in (17) is the latest version. AF: argument function. A-structure: argument structure..

(31) 17. function to map to thematic roles (Her, 2004, 2007). Next, the argument structure of zhui-lei ‘chase-tired’ is produced as arranged in (18) through three steps: linking of composing roles, suppression, and causative distribution. Note that roles subcategorized by zhui ‘chase’ are represented by alphabets and that those of lei ‘tired’ are represented by numerals. In the argument structures, hyphens specify linking of composing roles, suppression is indicated by. 政治大. crossing out of a role, and [caus] and [af] respectively stand for Cause and Affectee. 立. features resulting from causative distribution.. ‧ 國. 學. (18) zhui ‘chase’ <a, b> ＋ lei ‘tired’ <1>  i. <a b-1>. ‧. ii. <a[caus] b-1[af]>. Nat. iv. <a-1[af] b[caus]>. er. io. sit. y. iii. <a-1 b>. al. n. v i n Ch In order to create an argument-function i U by the simplified Lexical e n g c hmapping, Mapping Theory, the readings of sentences containing resultative compounds such as zhui-lei ‘chase-tired’ can be understood. (19) Zhangsan John. zhui-lei-le. Lisi.. chase-tired-ASP Lee.

(32) 18. a. ‘John chased Lee to the extent of making Lee tired.’ b-1>. <a IC CF. [－r] S/O/…. S/O. UMP SUBJ. ii. (Compatible Grammatical Functions). OBJ. b-1[af]>. <a[caus]. 立 [－r] S/O/…. agent, 1 = theme) 政 (a =治大. S/O OBJ. io. sit. y. Nat b. *‘Lee chased John and John got tired.’. al. n <a IC CF. v i n C h (a = agent, bU= patient, 1 = theme) b-1> engchi. [－r] S/O/…. UMP *OBJ. ‧. UMP SUBJ. 學. ‧ 國. IC CF. (a = agent, b = patient). er. i.. S/O *SUBJ.

(33) 19. c. ‘John chased Lee and John got tired.’ iii. b>. <a-1 IC. (a = agent, b = patient). [－r]. CF. S/O/…. S/O. UMP SUBJ. OBJ. 政治大. d. ‘Lee chased John and Lee got tired.’. [－r]. S/O. S/O SUBJ. io. sit. OBJ. ‧. UMP. [－r]. (1 = theme, b = patient). 學. CF. b[caus]>. Nat. IC. ‧ 國. <a-1[af]. y. 立. iv.. al. er. Despite the involvement of causativity and different suppressed roles in <b-1>. n. v i n C hsimilar in some ways. and <b-1>, (19i) and (19ii) are e n g c h i U Their syntactic assignments. behave alike and identical readings are derived. When the IC applies, the [－r] feature is associated with either the patient role <b-1> or the theme role <b-1>. Compatible functions for both cases (19i) and (19ii) remain the same. Based on the thematic hierarchy, the UMP allows the most prominent thematic role agent to link to the most prominent syntactic position SUBJ. After the SUBJ position is occupied, patient/theme only has the OBJ position to fill. Even though (19i) and (19ii) differ at.

(34) 20. some phases before grammatical functions surface, the final outcomes of the syntactic construction and semantic reading are identical. This account successfully predicts the causative reading in (19a). Regarding the reading in (19c), the suppressed <1> fails to receive causativity. Still, as patient, receives [－r] and can be associated with SUBJ or OBJ. Without any features, agent can be assigned to all possible grammatical functions. The UMP. 政治大. compels agent to occupy SUBJ and patient to link to OBJ.. 立. In (19d), <a> is bound to <1> and with <a> suppressed, the composite role. ‧ 國. 學. <a-1> receives [af] and receives [caus] according to the causativity assignment.. ‧. With both roles unsuppressed and receiving a [－r] feature, both theme and patient. Nat. io. sit. y. can be opted for by SUBJ and OBJ. A distinction between the two needs to be made at. al. er. this point. Dowty (1991) asserts that as prototypical properties for argument roles,. n. v i n C h[af], [caus] and [af]Uare typically assigned to with [caus] being more prominent than engchi agent and proto-patient respectively. Therefore, the UMP automatically associates the. more prominent [caus] with SUBJ and the less prominent [af] with OBJ. As for the unavailable reading in the zhui-lei ‘chase-tired’ compound, as repeated in (19b), the inaccessibility exists because the argument-function mapping violates the UMP. However, the above analysis is limited to certain types of resultative compounds,.

(35) 21. leaving the argument structures of other resultative verb compounds unspecified. In order to interpret various readings behind sentence constructions with different resultative compounds in Mandarin, in our analysis, the following firstly concentrates on the overall formation of all argument structures by adopting ideas about linking of composing roles, suppression in a composite role, and causativity distribution. By employing the principles and constraints, the next section demonstrates how argument. 政治大. roles correspond to grammatical functions, further proving how different readings can. 立. be realized.. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v.


(37) Chapter 3 Formulating Argument Structures and Argument Realization. In this section, we aim at formulating a complete set of argument structures for four different types of resultative compounds based on the transitivity of each composing verb. These four types of resultative compounds include. 政治大. transitive-transitive verb compounds, transitive-intransitive verb compounds,. 立. intransitive-transitive verb compounds, and intransitive-intransitive verb compounds,. ‧ 國. 學. as categorized by Li (1990).. ‧. 3.1 The Linking of Composing Roles. Nat. io. sit. y. As mentioned earlier, resultative compounding enables a role from Vcaus to. er. combine with a role from Vres. The binding, also known as theta-identification,. al. n. v i n C hby the fact that theUtwo roles index the same target merging, or fusion, is triggered engchi (Higginbotham, 1985; Li, 1990, 1995; Chang, 1998; Her, 2004, 2007; Huang, Li, & Li, 2009; Randall, 2010). The combination of the two roles is thus understood as a composite role; the two roles that form a composite role are called composing roles (Her, 2004, 2007). From the various combinations of one composing role bound to the other, we could predict that one resultative compounding construction would bring forth various composite roles which may cause the reading to differ from one to. 23.

(38) 24. another. In linking composing roles, we indicate argument structure arguments of Vcaus by alphabets, such as <a, b> while arguments of Vres are represented by numerals, <1, 2>, for example. A universal constraint is proposed in (20), following Li (1990), who states that the theta-role prominence of each verb should not be violated on the compound level.. 政治大. A related idea is observed in other studies (Goldberg & Jackendoff, 2004; Randall,. 立. should control the syntactic representation at a higher level.. ‧. (20) The Argument Hierarchy Preservation Principle:. 學. ‧ 國. 2010), indicating that the aspectual or the conceptual structure of each constituent. Nat. morpho-lexical operations.. al. er. io. sit. y. The relative hierarchy in the argument structure must be preserved in. n. v i n Corders It is concluded that the reversed h e nof g<a,cb>hfrom i Uthe first verb or <1, 2> from the second verb are impossible because either <b, a> or <2, 1> destroys the relative hierarchical positions in an argument structure. This constraint predicts that any argument structure with a reversed order should be ruled out. Therefore, feasible composite roles of each type of resultative compound are formulated as in (21)-(24). (21) transitive-transitive verb compounds Vcaus <a, b> ＋ Vres <1, 2>  <a-1 b-2>.

(39) 25. (22) transitive-intransitive verb compounds Vcaus <a, b> ＋ Vres <1>  i. <a-1 b> ii. <a b-1> (23) intransitive-transitive verb compounds Vcaus <a> ＋ Vres <1, 2>  i. <a-1 2> ii. <1 a-2>. 政治大. (24) intransitive-intransitive verb compounds. 立. ii. <a 1>. ‧. iii. <1 a>. 學. ‧ 國. Vcaus <a> ＋ Vres <1>  i. <a-1>. Nat. io. sit. y. In addition to the Argument Hierarchy Preservation Principle, note that possible. al. er. combinations are also restricted since roles of a Mandarin resultative compound are. n. v i n limited to two at most (Her, C 2007). h eHuang, i ULi (2009) also observe that under n g cLi,hand typical circumstances the Case filter assigns Cases to two arguments at a maximum; that is, the subject and the object. They thus assert that theta-identification is triggered when thematic roles are more than two and that theta-identification is not in demand when the total number of thematic roles is identical with that of available arguments with Cases. Based on these rationales, therefore, even though composing verbs subcategorize more than two arguments in total, only two argument roles can surface.

(40) 26. on the compound level. For the same reason, it is acceptable to surface two separate roles without being linked for (24ii) and (24iii) from intransitive Vcaus and Vres. Furthermore, as the first component verb of a resultative compound denotes the cause whereas the second one implies the result, the argument role <a> licensed by Vcaus is presumed to be more prominent than the argument role <1> subcategorized by Vres. <a> tends to be agent, while <1> is likely to be theme. This tendency is intuitive.. 政治大. According to the Argument Hierarchy Preservation Principle, argument structures. 立. where <1> precedes <a> are prohibited. Possible composite roles of each type of. ‧ 國. 學. resultative compounds are thus revised in (25)-(28).. ‧. (25) transitive-transitive verb compounds. Nat. (26) transitive-intransitive verb compounds. n. al. Ch. Vcaus <a, b> ＋ Vres <1>  i. <a-1 b>. engchi. ii. <a b-1> (27) intransitive-transitive verb compounds Vcaus <a> ＋ Vres <1, 2>  <a-1 2> (28) intransitive-intransitive verb compounds Vcaus <a> ＋ Vres <1>  i. <a-1> ii. <a 1>. er. io. sit. y. Vcaus <a, b> ＋ Vres <1, 2>  <a-1 b-2>. i n U. v.

(41) 27. As we proceed to argument-function mapping, note that all argument structures (25)-(28) are not supposed to map to syntactic functions immediately because there are two roles in one composite role. According to the UMP or the θ-Criterion, only one argument role is allowed to map to one syntactic position. The next section explains how this problem can be solved. 3.2 Suppression in a Composite Role. 政治大. The composite role which in fact consists of two roles links to only one. 立. grammatical function instead of two. Even so, such linking does not seem to violate. ‧ 國. 學. the one-to-one correspondence between argument roles and grammatical functions. ‧. that either the UMP or the θ-Criterion enforces.. Nat. io. sit. y. Based on Higginbotham’s (1985) statement about theta-assignment, Li (1990). al. er. argues that both composing roles are given the same index and thus function as a. n. v i n Ch single theta-role. Later at a derivational as a single theta-role, it fills a syntactic e n gstage, chi U slot without violating the θ-Criterion. This stipulation therefore satisfies the θ-Criterion. Furthermore, in order to fulfill the θ-Criterion, Randall (2010) provides the Bound Argument Condition, stating that between two co-indexed conceptual arguments only the higher one is allowed to link to a syntactic representation. This condition indicates that two arguments which are co-indexed can link and that the.

(42) 28. unexpressed lower argument might be suppressed, an idea similar to Her’s perspective. Her (2007) suggests a rather intuitive solution called suppression to clarify this seemingly “two-to-one” mapping between the composite role and its corresponding grammatical function. The idea of suppression assumes that one of the roles in the composite role should be suppressed so that the role which is not suppressed can map. 政治大. to only one syntactic slot. Hence only one role is linked to one grammatical function,. 立. which satisfies the strict one-to-one mapping required by the UMP or the θ-Criterion.. ‧ 國. 學. Moreover, suppression aims at preventing a role from surfacing as a separate. ‧. grammatical function, not intending to omit its meaning. On the one hand,. Nat. io. sit. y. syntactically, the suppressed role is implicitly linked to the syntactic argument that its. n. al. er. unsuppressed counterpart is explicitly connected to. On the other hand, semantically,. Ch. engchi. the meaning of the suppressed role remains intact.. i n U. v. If two linked composing roles, one of which is suppressed, do not violate the UMP or the θ-Criterion, possible composite roles in (25)-(28) would be modified as (29)-(32) present, with the suppressed role being crossed out. (29) transitive-transitive verb compounds Vcaus <a, b> ＋ Vres <1, 2>  i. <a-1 b-2> ii. <a-1 b-2>.

(43) 29. iii. <a-1 b-2> iv. <a-1 b-2> (30) transitive-intransitive verb compounds Vcaus <a, b> ＋ Vres <1>  i. <a-1 b> ii. <a-1 b> iii. <a b-1> <a b-1> 政 iv. 治大. 立. 學. ‧ 國. (31) intransitive-transitive verb compounds Vcaus <a> ＋ Vres <1, 2>  i. <a-1 2>. ‧. ii. <a-1 2>. Nat. n. al. Ch. e ii.n <a-1> gchi. sit er. io. Vcaus <a> ＋ Vres <1>  i. <a-1>. y. (32) intransitive-intransitive verb compounds. i n U. v. iii. <a 1> 3.3 Causativity Distribution Some writers have claimed that causativity is achieved when a resultative compound is formed, and that it needs to be assigned in resultative compounding (Li, 1995; Her, 1997, 2007). An argument from Vcaus initiates the causing event, while an argument from Vres is involved with the result. Thus the former is related to Cause,.

(44) 30. whereas the latter to Affectee (Li, 1995; Her, 1997, 2004, 2007). Different from Li’s (1995) c-roles which are considered distinct linguistic entities independent from theta roles, Her (2007) treats Cause and Affectee as semantic features attached to theta roles. Given the intuition that Affectee refers to an entity which is affected by Cause, we assume Affectee bears a theme-like theta role, a role that goes through a change of state, therefore modifying Her’s (2007) causativity distribution in resultative compounds, as (33) shows:. 立. 政治大. (33) Causativity Assignment in Resultative Compounding - revised. ‧ 國. 學. An unsuppressed theme-like role from Vres receives [af] (Affectee) if and. ‧. only if an unsuppressed role from Vcaus exists to receive [caus] (Cause).. Nat. sit. n. al. er. io. finalized as in (34)-(37).. y. Consequently, the argument structures of all resultative compounds can be. Ch. engchi. (34) transitive-transitive verb compounds. i n U. v. Vcaus <a, b> ＋ Vres <1, 2>  i. <a-1 b-2> ii. <a[caus]-1 b-2[af]> iii. <a-1[af] b[caus]-2> iv. <a-1 b-2> (35) transitive-intransitive verb compounds Vcaus <a, b> ＋ Vres <1>  i. <a-1 b>.

(45) 31. ii. <a-1[af] b[caus]> iii. <a b-1> iv. <a[caus] b-1[af]> (36) intransitive-transitive verb compounds Vcaus <a> ＋ Vres <1, 2>  i. <a[caus]-1 2[af]> ii. <a-1 2>. 政治大. (37) intransitive-intransitive verb compounds. 立. 學. ii. <a-1>. Nat. io. sit. 3.4 Argument-Function Mapping. y. iii. <a[caus] 1[af]>. ‧. ‧ 國. Vcaus <a> ＋ Vres <1>  i. <a-1>. al. er. With one composing role suppressed to satisfy the UMP or the θ-Criterion and. n. v i n C hwe continue to analyze with the addition of causativity, e n g c h i U each of the four different. types of resultative compounds, investigating their argument-function linking and how their readings are yielded. Before we move on to the mapping between argument roles and grammatical functions, two hierarchies or principles which are related to grammatical function assignment should be taken into consideration. First, the thematic hierarchy has been proposed by numerous researchers; that is, the more prominent role must fill the.

(46) 32. structurally more prominent position (Jackendoff, 1972, 1990; Carrier-Duncan, 1985; Higginbotham, 1985; Baker, 1988; Bresnan & Kanerva, 1989, 1992; Grimshaw, 1990; Li, 1990, 1995; Her, 2004, 2007; Randall, 2010, and many others). For example, agent is more prominent than patient; the subject is structurally ranked higher than the object. Consequently, the thematically most prominent role, such as agent, must be assigned to the structurally most prominent position, the subject of a sentence. Second,. 政治大. supported by the idea of proto-role that Cause is prototypically associated with agent. 立. and Affectee is prototypically theme (Dowty, 1991), the prominence of causativity. ‧ 國. 學. demonstrates that Cause is more prominent than Affectee (Li, 1995; Her, 1997, 2004,. ‧. 2007). Due to this hierarchical prominence, in Li’s (1995) and Her’s (1997, 2004,. Nat. io. sit. y. 2007) studies, Cause fills the subject position and the Affectee the object position in a. al. er. sentence. Such argument-function linking is consistent with the linking rules in Levin. n. v i n and Rappaport Hovav (1995) whichC specify cause is expressed by the external h e nthatgthe chi U argument and that the affected entity corresponds to the internal argument. 3.4.1 Transitive-Transitive Verb Compounds With the foregoing considerations in mind, we are able to first probe in resultative compounds with two transitive verbs, such as da-shu ‘hit-lose.’ Potential combinations for the argument structure of transitive-transitive verb compounds are given in (38)..

(47) 33. (38) Vcaus <a, b> ＋ Vres <1, 2>  i. <a-1 b-2> ii. <a[caus]-1 b-2[af]> iii. <a-1[af] b[caus]-2> iv. <a-1 b-2> In terms of the sentence containing da-shu ‘hit-lose’ as in (39), three out of the four argument structures mentioned above predict two correct readings and the. 政治大. 立da-shu-le. John. hit-lose-ASP. Nat. io. sit. b. ‘Lee hit John and Lee lost to John.’. Lee. ‧. a. ‘John hit Lee and John lost to Lee.’. Lisi. 學. ‧ 國. (39) Zhangsan. y. structure of the sentence.. al. er. For the reading in (39a), argument structures (38i) and (38iv) cover this reading.. n. v i n C h lead to the sameUconsequence. In (38i), the agent Note that both argument structures engchi role from Vcaus identifies with the theme role from Vres, both referring to Zhangsan ‘John;’ the patient role from Vcaus also combines with the patient role from Vres because they both denote Lisi ‘Lee.’ All roles of Vres <1> and <2> here are. suppressed after composing roles are linked, with <a> and left unsuppressed. As to what the thematic hierarchy requests, <a> as the most prominent agent naturally occupies the subject position and the less prominent patient role takes the object.

(48) 34. position. In (38iv), likewise, agent <a> and theme <1> form a composite role; so do patient roles and <2>. The only difference from (38i) is that in (38iv) all roles of Vcaus are suppressed. In terms of the unsuppressed roles, note that theme and patient appear at the same time. The thematic hierarchy does not specifically treat theme and patient any differently. In addition, no extra features are available for differentiating. 政治大. the two. The argument-function assignment may be confusing. To avoid such. 立. confusion, a distinction therefore must be made. Here we assume that the patient is. ‧ 國. 學. secondary to theme when the two exist without further information available for. ‧. grammatical function assignment. Evidence of this is that when the verb shu ‘lose’. Nat. io. sit. y. stands alone, the theme role bears a priority over the patient role to map to the SUBJ.. n. al. er. Therefore, based on this assumption, the primary <1> and the secondary <2> select SUBJ and OBJ respectively.. Ch. engchi. i n U. v. As we can see, although these two argument structures, (38i) and (38iv), have different unsuppressed roles, their mappings to grammatical functions behave exactly the same. They generate an identical sentence construction and meaning. The fact that two argument structures lead to only one reading may seem redundant. However, such many-to-one correspondence is in fact quite common in approaches to language analysis. For instance, tree structures for an adjunct-stacking phrase like ‘a little cute.

(49) 35. child on the stage’ are more than one, but they all refer to the same functional structure or meaning. From this perspective, it is hence legitimate to allow two argument structures to illustrate one syntactic representation and one reading. As for the reading in (39b), as discussed in previous research (Lin, 2007), it is correctly and easily explained by the argument structure (38iii). Lisi ‘Lee’ gets agent <a> from Vcaus and theme <1> from Vres; Zhangsan ‘John’ is assigned patient and. 政治大. <2> from both Vcaus and Vres. Since Vres’s theme <1> and Vcaus’s role patient are. 立. left unsuppressed, causativity is activated, contributing the [af] feature to Lisi ‘Lee’. ‧ 國. 學. and the [caus] feature to Zhangsan ‘John.’ Given Dowty’s (1991) claim that [caus] is. ‧. typically more prominent than [af], it is possible to link <b[caus]-2> to SUBJ and. Nat. io. sit. y. <a-1[af]> to OBJ and to produce the meaning in (39b).. n. al. er. Furthermore, another argument structure with causativity distribution (38ii). Ch. engchi. corresponds to a sentence like (40). (40) Zhangsan John. i n U. zuo-pei-le. maimai.. do-lose-ASP. business. v. ‘John did business and he lost it.’ In (40), <a> and <1> refer to Zhangsan ‘John,’ whereas and <2> indicate maimai ‘business.’ Given the causativity assignment, when the unsuppressed role <a> of Vcaus features [caus], the unsuppressed theme <2> of Vres takes the [af] feature..

(50) 36. Both the causativity hierarchy and thematic hierarchy require that agent Zhangsan ‘John’ and patient/theme maimai ‘business’ map to SUBJ and OBJ respectively. 3.4.2 Transitive-Intransitive Verb Compounds In addition, the argument-function mapping for the second type of resultative compound which consists of a transitive Vcaus and an intransitive Vres can also be accounted for based on their argument structures, which are set out in (41), as discussed in Her (2007).. 立. 政治大. 學. ‧ 國. (41) Vcaus <a, b> ＋ Vres <1>  i. <a-1 b>. ii. <a-1[af] b[caus]>. ‧. iii. <a b-1>. Nat. io. sit. y. iv. <a[caus] b-1[af]>. n. al. er. Again, the classic sentence containing zhui-lei ‘chase-tired,’ for instance, is. Ch. engchi. investigated in (42) according to Her’s (2007) analysis. (42) Zhangsan John. zhui-lei-le. i n U. v. Lisi.. chase-tired-ASP Lee. a. ‘John chased Lee to the extent of making Lee tired.’ b. *‘Lee chased John and John got tired.’ c. ‘John chased Lee and John got tired’ d. ‘Lee chased John and Lee got tired.’.

(51) 37. The reading in (42a) may result from either (41iii) or (41iv). Note that two slight distinctions exist through the mapping process in these two argument structures. First, the unsuppressed composing roles differ. <1> from Vres is suppressed in (41iii), while from Vcaus is suppressed in (41iv). The second difference lies in the presence of causitivity. The suppressed role from Vres prevents argument roles in (41iii) from receiving causitivity. In (41iv), on the other hand, the theme role from Vres is not. 政治大. suppressed so that causativity is registered. Nevertheless, no matter which composing. 立. role is left unsuppressed and whether causativity is present, their syntactic. ‧ 國. 學. assignments simply fall out and derive identical readings. In both (41iii) and (41iv),. ‧. <a> denotes the agent role Zhangsan ‘John’ and both composite roles <b-1> and. Nat. io. sit. y. <b-1> mean the patient and theme role Lisi ‘Lee.’ Thematic prominence predicts that. al. er. <a> is more prominent than <b-1> or <b-1>. The causative assignment in (41iii). n. v i n likewise justifies that <a> isC more <b-1>. Therefore, <a> h eprominent i U n g c hthan. automatically occupies SUBJ and <b-1> or <b-1> surfaces as OBJ. (41iii) and (41iv) bring forth the identical syntactic structure and reading. In terms of (41i), it corresponds to the reading in (42c). Here <a-1> indicates the person who chased and got tired, while is the one being chased. Since <a-1> preserves the role of agent and is undoubtedly theme, <a-1> and naturally map to SUBJ and OBJ respectively..

(52) 38. As for the last argument structure (41ii), <a-1> which refers to Lisi ‘Lee’ and which specifies Zhangsan ‘John’ both receive a patient/theme role. Note that to distinguish the thematic prominence between patient and theme can be ambiguous because different studies analyze this issue differently. Consequently, it is puzzling to determine which role should connect to SUBJ and which should connect to OBJ. However, the causativity distribution assigns <a-1> a [af] feature and a [caus]. 政治大. one. The causativity hierarchy links the more prominent <b[caus]> to SUBJ and the. 立. less prominent <a-1[af]> to OBJ, accounting for the reading in (43d).. ‧ 國. 學. Furthermore, a rather tricky case is introduced. In addition to the readings. ‧. mentioned in (39), another reading from da-shu ‘hit-lose’ is accessible and analyzed. Nat. io. sit. y. under the category of transitive-intransitive verb compounds. Instead of treating shu. n. al. er. ‘lose’ as a transitive verb, we regard it to be an intransitive verb in order to produce this reading, as presented in (43). (43) Zhangsan John. Ch. engchi. da-shu-le. Lisi. hit-lose-ASP. Lee. i n U. v. ‘John hit Lee and Lee lost (to John).’ Given that the second verb is intransitive, argument structures (41iii) and (41iv) could be applied to explain this interpretation. In both (41iii) and (41iv), <a> refers to Zhangsan ‘John,’ accepting the agent role. and <1>, identifying with each other,.

(53) 39. indicate Lisi ‘Lee.’ With either role suppressed, <b-1> and <b-1> receive the patient and theme role respectively. For (41iii), according to the thematic hierarchy, the more prominent agent is designated to fill the subject position, while the less prominent patient ought to occupy the object slot. As with (41iv), the causative distribution assigns <a> a typically more prominent [caus] feature and <b-1> a typically less prominent [af] feature. As a result, both the thematic hierarchy and the causative. 政治大. distribution request that the agent <a[caus]> surfaces as SUBJ and the theme. 立. <b-1[af]> as OBJ. This analysis of treating da-shu ‘hit-lose’ as a transitive-intransitive. ‧ 國. 學. verb compound fully accounts for the reading and syntactic structure of Zhangsan. ‧. da-shu-le Lisi ‘John hit Lee and Lee lost.’. Nat. io. sit. y. 3.4.3 Intransitive-Transitive Verb Compounds. er. After viewing the cases of zhui-lei ‘chase-tired’ and da-shu ‘hit-lose,’ we proceed. al. n. v i n to examine other resultativeC compounds. a resultative compound composed h e n gRegarding chi U of an intransitive verb and a transitive verb, with the first composing verb generating one argument and the second verb two arguments, possible argument structures of the resultative compounds are generalized as follows. (44) Vcaus <a> ＋ Vres <1, 2>  i. <a[caus]-1 2[af]> ii. <a-1 2> Take the compound pao-diu ‘run-lose’ in (45) for example. Based on these.

(54) 40. argument structures, both (44i) and (44ii) correctly predict the sentence’s meaning and its linking between arguments and grammatical functions (45) Zhangsan John. pao-diu-le. qian.. run-lose-ASP. money. ‘John ran and he lost the money.’ Both (44i) and (44ii) show that Vcaus’s only thematic role <a> is identified with. 政治大. Vres’s agent role <1>. Unlike (44ii), causitivity is involved in (44i). As the only role. 立. from Vcaus is not suppressed, the causativity assignment activates, revealing. ‧ 國. 學. causativity in this argument structure. With causativity or without it, however, either. ‧. argument structure demonstrates that the most prominent agent role. Nat. io. sit. y. <a-1>/<a[caus]-1> is mapped to the SUBJ, and theme role <2>/<2[af]> is mapped to. al. er. the OBJ, as the thematic hierarchy requires. The argument-function mappings of both. n. v i n argument structures lead to the sameCform of Zhangsan pao-diu-le qian h eand n gmeaning chi U ‘John ran and he lost the money.’ Hence these two argument structures suffice to illustrate this sentence with an intransitive-transitive verb compound. 3.4.4 Intransitive-Intransitive Verb Compounds A final type of resultative compound features two intransitive verbs. Both Vcaus and Vres are intransitive, subcategorizing for one argument only. Consequently, as supported by Her’s (2007) analysis, the argument structure can be organized as (46)..

(55) 41. (46) Vcaus <a> ＋ Vres <1>  i. <a-1> ii. <a-1> iii. <a[caus] 1[af]> Constructions introduced at the beginning in (3) and (4) are analyzed below as instances of this type. For both resultative compounds pao-lei ‘run-tired’ and ku-shi ‘cry-wet,’ the consequences of their argument-function mappings automatically. 政治大. emerge, as repeated in (47) and (48).. 立. pao-lei-le. he. run-tired-ASP. ‧. ‘He ran and he got tired.’. 學. ‧ 國. (47) ta. Nat. io. sit. y. Similar to the analysis from (39i) and (39iv), the two argument structures (46i). er. and (46ii) likewise result in the same consequence. They produce the same syntactic. al. n. v i n C hone result, no variation function assignment. With only e n g c h i U should be specified between the two. As (46i) and (46ii) both have only one composite role, SUBJ is definitely filled by this composite role whether it is the agent role or not. Thus the syntax and the meaning of this sentence simply fall out. (48) ta he. ku-shi-le. shoupa. cry-wet-ASP handkerchief. ‘He cried and the handkerchief got wet.’.

(56) 42. Furthermore, unlike that of (47), the reading of resultative compound ku-shi ‘cry-wet’ in (48) takes two separate roles along with causative assignment. The argument structure of such compound is given in (46iii). (46iii) shows that the two argument roles would surface in the syntax. <a>, referring to ta ‘he,’ serves as agent; <1>, indicating shoupa ‘handkerchief,’ functions as theme. Since these two roles come from Vcaus and Vres respectively and neither of them is suppressed, the. 政治大. causativity distribution becomes valid. Either due to the thematic hierarchy or the. 立. causative hierarchy, the more prominent <a>, which serves as agent and receives. ‧ 國. 學. [caus], selects the subject position and the less prominent <1>, which functions as. Nat. io. sit. succeeds in formulating the causative reading in (48).. y. ‧. theme and accepts [af], fills the object position. As a result, the argument structure. al. er. From the examples above, it can be concluded that not all feasible combinations. n. v i n of roles and composite roles on the C argument can be realized by the same h e nstructure gchi U. resultative compound. The compounding only provides possible argument structures to explain the syntactic assignment and its reading. Moreover, it is attested that the possible argument structures illustrated should suffice for predicting the grammatical function arrangement from the reading of various resultative compounds..

(57) Chapter 4 Headedness. Previous studies have debated whether Mandarin resultative compounds are headed. The headedness issue has been discussed from four perspectives. Some scholars posit that V1 is the head of a resultative compound (Li, 1990, 1993, 1995, 1999; Cheng & Huang, 1994), whereas some consider V2 to be the head (Shen, 1992;. 政治大. Tai, 2003). While evidence is presented to argue that both V1 and V2 serve as heads. 立. (Gu, 1992), a stance that resultative compounds are headless is also proposed (Huang. ‧ 國. 學. & Lin, 1992; Li, 2008, 2009).. Nat. io. sit. 4.1.1 Li’s Analysis (1990, 1993, 1995, 1999). y. ‧. 4.1 V1 as Head. er. Li (1990) firstly suggests that V1 is the head on the basis of the categorical. al. n. v i n features of the two verbs. InC hishexample of pao-leiU‘run-tired,’ native speakers would engchi intuitively assert that the resultative compound is form by a verbal V1 pao ‘run’ and an adjectival V2 lei ‘tired’. Since the whole compound functions as a verb rather than an adjective, it is possible to assume that the categorical feature of V1 seems to determine that of the whole compound. This left-head assertion is later attested in his analysis. Reported in section 2.2.1, among the three assumptions that govern the structure. 43.

(58) 44. of resultative compounds, the last assumption, the head-feature percolation (Li, 1990, 1995), playing a crucial role in determining the head of a resultative compound, postulates that the head of a compound would reflect the main properties of the compound. However, the head-feature percolation cannot explain (49). These examples militate against the assertion that V1 is the head. As pointed out by Li (2009), there. 政治大. are plenty of sentences whose roles from V1 never surface as arguments in the. 立. ta-de. yanjing. ku-xia-le. ‧. he-GEN eyes. 學. (49) a.. ‧ 國. syntactic construction.. cry-blind-ASP. Nat. sit. n. al. er. io. b.. shoupa. y. ‘He cried so much he could barely see.’. Ch. ku-shi-le. engchi. i n U. v. handkerchief cry-wet-ASP ‘The handkerchief was wet with tears.’ As the head, V1 is supposed to bring forth one or more roles from V1 to map to an argument (or more). Nevertheless, the surface arguments in (49), yanjing ‘eyes’ and shoupa ‘handkerchief’ do not receive a role from V1. The sentences in (49) apparently violate the head-feature percolation, refuting that V1 is the head..

(59) 45. Another problem from V1 as the head is discovered in sentences such as (50), given by Li (1995). In (50), the argument nachang jihuang ‘that famine’ occupying the subject position is neither the argument of V1 nor the argument of V2. Realized as the object, the argument henduo ren ‘many people’ is linked to a single thematic role subcategorized by both V1 and V2. (50). na-chang. jihuang. that-CL. famine. 立. e-si-le. 治政 hungry-dead-ASP 大. henduo. ren. many. people. ‘Many people starved to death in that famine.’. ‧ 國. 學. If V1 is the head, according to the head-feature percolation, the thematic. ‧. hierarchy of V1 should be maintained in the whole compound. In other words, as V1 e. Nat. io. sit. y. ‘hungry’ requires only one thematic role, this role ought to map to the argument. al. er. henduo ren ‘many people’ and surface in the subject position. This realization should. n. v i n C hHowever, for the U be maintained in the compound. e n g c h i compound, its subject position is replaced by an argument whose role does not belong to V1. The argument with the. role from V1 is somehow realized as the object. Sentences like (50) obviously do not adhere to the head-feature percolation and the V1-as-head hypothesis. 4.1.2 Cheng and Huang’s Analysis (1994) Cheng and Huang (1994) argue that V1 occupies the structural head position, which is related to the aspectual properties and event type. They claim that the.

(60) 46. aspectual properties of V1 determine those of the entire resultative compound. For resultative compounds, V2 generally describes a state or a change of state, while V1 can either specify an action or a state. When V1 is active, the whole compound denotes a transitive or unergative pattern. When V1 is stative, the whole compound indicates an ergative or causative use. In either situation, V2 stays stative or as a change of state. Therefore, the aspectual or event type of a resultative compound. 政治大. depends mainly on the aspectuality or event type of V1. V1 is considered the head of. 立. a resultative compound.. ‧ 國. 學. However, there are cases where V1 denotes an action but the entire compound. ‧. look-blurred-ASP. eye. sit. kan-hua-le. n. al. er. he-GEN. yanjing. io. ta-de. Nat. (51) a.. y. realizes in an ergative or causative pattern, given in (51).. i n C ‘He read so much that his blurred.’ heyes e ngot gchi U b.. v. na-ben. shu. kan-hua-le. ta-de. yanjing. that-CL. book. look-blurred-ASP. te-GEN. eye. ‘That book made his eyes blurred (from reading it).’ In the compound kan-hua ‘look-blurred,’ V1 undoubtedly refers to an activity. According to Cheng and Huang’s (1994) classification, the compound should be used.

(61) 47. in a transitive or an intransitive unergative pattern. Nevertheless, the entire compound presents the opposite, entering an ergative-causative pattern whose V1 in the compound is supposed to be stative. (51a) shows the compound’s ergative use; (51b) is the causative pattern. Cheng and Huang (1994) attribute such pair to the thematic property V1 assigned to its subject. In both (51a) and (51b), the subject of V1 kan ‘look’ receives an Affectee or an experiencer role instead of a pure agent role that is. 政治大. used in a transitive-unergative pattern. With a causee/experiencer role, the entire. 立. resultative compound kan-hua ‘look-blurred’ hence belongs to the ergative-causative. ‧ 國. 學. category. In this sense, it is apparently not the event type of V1 that decides the event. ‧. type of an entire resultative compound.. Nat. io. sit. y. Furthermore, examples are also available for activity-denoting V1s forming. n. al. er. resultative compounds that interact between transitive and causative uses, as shown in. Ch. engchi. (52) (Cheng & Huang, 1994:190; Huang, 2006). (52) a.. ta. ku-xing-le. xiaohai. he. cry-awake-ASP. child. i n U. v. ‘His crying awoke the sleeping child.’. b.. meng-li-de. na-jian. shi. ku-xing-le. ta. dream-in-MM. that-CL. matter. cry-awake-ASP. he.

(62) 48. ‘What happened in the dream made him cry himself awake.’ Again, V1 in the resultative compound ku-xing ‘cry awake’ indicates an action. Instead of entering a transitive-unergative category or the opposite category as the compound in (51), the whole compound ku-xing ‘cry awake’ is realized in a mixed category, the transitive pattern given in (52a) and the causative one presented in (52b). This evidence implies that an activity-denoting V1 does not necessarily block. 政治大. stative-related ergative and causative patterns for its compound.. 立. Despite V1’s activity event type, one compound is still able to form an. ‧ 國. 學. ergative-causative pattern. The aspectual and event type of V1 roughly generalizes the. ‧. possible aspectual and event type of the whole compound, not precisely predicting it.. Nat. sit. n. al. er. io. type account may be shaky.. y. Therefore, the ground for determining V1 as the head based on this aspectuality/event. 4.2 V2 as Head. Ch. engchi. i n U. v. 4.2.1 Shen’s Analysis (1992) The V2-as-head hypothesis is supported by Shen (1992), who counters the V1-headed view (Li, 1990, 1995, 1999) by asserting that the logical subject of V2, not that of V1, is realized as the subject of the entire compound. Examples are given in (53) and (54)..

(63) 49. (53) a.. shoupa. ku-shi-le. handkerchief cry-wet-ASP ‘The handkerchief was wet with tears.’. b. *ta. ku-shi-le. he. cry-wet-ASP. 政治大. Intended: ‘He cried and made something wet.’. kan-hua-le. eye. look-blurred-ASP. ‧. yanjing. 學. (54) a.. ‧ 國. 立. Nat. n. al. er. io. sit. y. ‘Someone read so much that his eyes got blurred.’. b. *ta he. Ch. kan-hua-le. engchi. i n U. v. look-blurred-ASP. Intended: ‘He read and made something blurred.’ In these sentences, neither the external argument of ku ‘cry’ nor that of kan ‘look’ is allowed to surface as the subject of the whole compound. Instead, it is the external argument of V2 that is realized as the subject of the entire compound. Given that the external argument of the head should be projected, Shen (1992) then concludes that.

(64) 50. resultative compounds are headed by V2. Nevertheless, this argument can be disproved because compounds with V1’s external argument realized are still available as shown in the previous section 4.1. Therefore, it is inappropriate to postulate this V2-as-head argument when counterexamples exist. 4.2.2 Tai’s Analysis (2003) From a relatively semantic perspective, one study (Tai, 2003) presumes that V2. 政治大. is the head of a resultative compound because it holds the main meaning of a. 立. predicate. For many compounds, V1 specifies how an action is done or a state is. ‧ 國. 學. achieved, while V2, as the “center of predication,” demonstrates the semantic focus of. ‧. the compound.. Nat. io. sit. y. Tai (2003) provides several verb compounds consisting of si ‘die’ as the second. al. er. verb to illustrate that V2 is the “semantic prime,” the semantic focus, of the. n. v i n C h by comparing Mandarin compound. First, he justifies his position e n g c h i U verb compounds with si ‘die’ to their equivalent English verb ‘kill.’ For example, nian-si ‘grind-die,’ da-si ‘hit-die,’ e-si ‘starve-die’ can all be translated into ‘kill’ in English. In this sense, V2 indeed functions as the center of predication, relating the semantic focus to si ‘die.’ As for V1, it serves more as a manner adverb, explaining the manner that causes dying. Thus, it is plausible to regard V2 as the head of the compound. However, Cheng and Huang (1994) refuted this hypothesis, offering other strings.

(65) 51. of examples such as ti-kai ‘kick-open,’ ti-dao ‘kick-fall,’ ti-bian ‘kick-flat.’ These compounds with ti- ‘kick’ describes the results caused by the action ti ‘kick.’ After comparing compounds with -si ‘die’ to those with ti- ‘kick,’ we observe that manners of killing are highlighted for the former and that results from kicking are emphasized for the latter. Based on these examples, Cheng and Huang conclude that this semantic/conceptual analysis can only show the semantic composition of compounds.. 政治大. The issue of headedness is not even involved.. 立. Rejecting both Tai’s and Cheng and Huang’s stances, we discover that the. ‧ 國. 學. concept of semantic prime as the head is too context-dependent. The semantic focus. ‧. of one compound can be either V1 or V2 based on the context or the focus a speaker. Nat. io. sit. y. aims to convey. Take ti-bian ‘kick-flat’ for instance. It is allowed to appear in a. al. er. sentence like wo mei-you ti-bian-le guan-zi, wo ya-bian-le guan-zi ‘I didn’t kick the. n. v i n C hthe manner of making can flat. I squashed it.’ to show e n g c h i U the can flat is the focus. In this case, V1 is the semantic focus. Nevertheless, if a sentence like wo mei-you ti-bian-le guan-zi, wo ti-kai-le guan-zi ‘I didn’t kick the can flat. I kicked it open.’ is presented, it is clear that the semantic focus lies in V2, the result from kicking. The example of ti-bian ‘kick-flat’ indicates that this semantic perspective relies much on the context and that the head varies too easily as the context changes. Second, from a syntactic perspective, the sentence from (55) in which V2 can.