西班牙语系_中国人民大学外国语学院.pdf

BEI-passive revisited: a constraint-based approach Fengrong Yang Renmin University of China Proceedings of the LFG’20 Conference On-Line Miriam Butt, Ida Toivonen (Editors) 2020 CSLI Publications pages 363–383 http://csli-publications.stanford.edu/LFG/2020 Keywords: BEI-passive, argument realization, Mandarin Chinese Yang, Fengrong. 2020. BEI-passive revisited: a constraint-based approach. In Butt, Miriam, & Toivonen, Ida (Eds.), Proceedings of the LFG’20 Conference, On-Line, 363–383. Stanford, CA: CSLI Publications. Abstract This paper revisits Chinese BEI-passive constructions by analyzing two lessstudied passive structures: the passive in which the patient argument of the passivized verb maps onto a post-verbal position, as well as the passive with intransitive verbs. I claim that there is no subject in these constructions, by showing that the post-V patient, if there is one, is the object and not the dislocated subject of the clause. I propose that BEI is a raising verb and a passive marker in Chinese and that a grammatical subject is not necessary for every clause. 1 Introduction The topic of this paper is Mandarin Chinese BEI-passive constructions. The paper is organized as follows. In section 2, I introduce basic BEI-passive structures and review current studies of BEI-passives. I will especially concentrate on the analyses of BEI and point out some problems of these analyses by presenting relevant facts. I then propose my own analysis of BEI and BEI-constructions in section 3. The main conclusions are summarized in section 4. 2 Current analyses of BEI and BEI-passive The patient/theme argument in Mandarin Chinese BEI-passive clauses can appear either in a pre-BEI position or post-verbally. I first introduce the basic passive BEI structures in Mandarin in 2.1 and then review the previous studies, in 2.2. The grammatical function of the post-V argument is discussed in 2.3. 2.1 Basic passive structures in Mandarin Chinese The commonly known and most studied passive construction in Mandarin Chinese is the passive ordered as NP-BEI(-NP)-V, in which the patient or theme argument maps onto the pre-BEI position, whereas the immediate postBEI element – if there is one – corresponds to the agent argument. Constructions with and without an overt post-BEI agent are named long passive and short passives (Huang et al. 2009), respectively. For example: (1) a. Zhang1 san1 bei4 Li3 si4 da3 le0. 张三被李四打了 Zhangsan BEI Lisi hit PERF ‘Zhangsan was hit by Lisi.’ b. Zhang1 san1 bei4 Ø da3 le0. 张三被打了 Zhangsan BEI hit PERF ‘Zhangsan was hit.’ (Huang et al. 2009) (1a) represents the typical long passive construction in Mandarin Chinese. The agent argument (i.e., Lisi, the hitter) is overtly expressed and it maps onto the 364 post-BEI position. (1b) represents the typical short passive construction, with no agent argument overtly expressed in the clause. In both cases, the patient argument (i.e., Zhangsan, the one being hit) is located before BEI. Apart from the construction represented in (1), in which the patient argument of a passivized transitive verb is located before BEI, the patient argument can also appear post-verbally, as illustrated by (2): (2) a. Bei4 ta1 pian4 le0 wu3 ge4 da4 huo2 ren2. 被她骗了五个大活人 BEI she cheat PERF five CLASS big living man (Literally) ‘There were cheated five living men by her.’ (Lu 2004) b. Bei4 ta1 cai3 zhao2 she2 le0. 被她踩着蛇了 BEI she step-on snake PERF (Literally) ‘There was stepped on a snake by her.’ c. Bei4 feng1 chui1 dao3 le0 yi4 ke1 shu4. 被风吹倒了一棵树 BEI wind blow down PERF one CLASS tree (Literally) ‘There was blown down a tree by the wind.’ Surprisingly, despite the huge bulk of research on Chinese BEI-passive constructions, with the patient argument mapping onto the pre-BEI position, little attention has been paid to the situation in which the patient argument is not located before BEI, but appears post-verbally, such as the NP wu3 ge4 da4 huo2 ren2 ‘five living men’ after the verb pian4 ‘cheat’ in (2a), the NP she2 ‘snake’ after the verb cai3 ‘step on’ in (2b), or the NP yi4 ke1 shu4 ‘a tree’ after the verb chui1 dao3 ‘blow-down’ in (2c), respectively.1, 2 The same order is also widely used in Shanghainese and Cantonese, as is shown in (3) and (4), respectively: 1 Though some studies have mentioned them in passing, as in Yu (1989), Lu (2004), Her (2008), or Huang et al. (2009), among others. 2 For me, when the patient argument maps onto the post-verbal position, the agent argument should be overtly expressed, such as the ta1 ‘she’ in (2a) and (2b), or the agent feng1 ‘wind’ in (2c). However, after consulting other native speakers, they pointed out that omitting the agent argument is also acceptable: (i) Bei4 pian4 le0 wu3 ge4 da4 huo2 ren2. 被骗了五个大活人 BEI cheat PERF five CLASS big living man (Literally) ‘There were cheated five living men.’ (ii) Bei4 chui1 dao3 le0 yi4 ke2 shu4. 被吹倒了一棵树 BEI blow down PERF one CLASS tree (Literally) ‘There was blown down a tree.’ (https://www.oursteps.com.au/bbs/archiver/?tid-727247.html&page=2, visiting time: 16:32, 05/12/2020.) 365 (3) a. Bah4 da3 khah4 tsho1 zaon3 wa3 theh4 ih4 khu1 zy3. 拔大客车撞坏脱一棵树 BEI big coach knock broken PERF one CLASS tree (Literally) ‘There was smashed a tree by a big coach.’ b. Bah4 yi1 sah4 theh4 sae1 eh4 nyin1. 拔伊杀脱三个人 BEI she kill PERF three CLASS person (Literally) ‘There were killed three people by her.’ (Yu 1989) (4) a. Bei2 keoi5 sik6 zo2 jat1 go3 lou5 baak3 gung1. 畀佢食咗一个老伯公 BEI it eat PERF one CLASS old uncle (Literally) ‘There was eaten an old man by it.’ b. Bei2 keoi5 sik6 zo2 gei2 go3 lei4. 畀佢食咗几个梨 BEI she eat PERF some CLASS pear (Literally) ‘There were eaten some pears by her.’ (Yu 1989) Similarly, though BEI is also widely used with intransitive verbs, little attention has been given, either. In such a construction, the sole argument of the intransitive verb maps onto the post-BEI position. Also, when BEI combines with intransitive verbs, only unergatives can appear in this construction, such as pao3/pao3 diao4 ‘run/run away’, tao2/tao2 zou3/tao2 pao3 ‘flee/flee away’, cheng2 gong1 ‘succeed’, etc. For instance:3 (5) a. Bei4 ta1 pao3 le0. 被她跑了 BEI she run PERF (Literally) ‘It was run by her. (She ran.)’ b. Bei4 ta1 cheng2 gong1 le0. 被她成功了 BEI she succeed PERF ‘It was succeeded by her. (She succeeded.)’ Unaccusative verbs, such as dao4 ‘arrive’, or diao4 ‘fall’, cannot appear in the BEI-passive construction: (6) a. *Bei4 ta1 dao4 le0. 被她到了 BEI she arrive PERF (Intended) ‘It was arrived by her. (She arrived.)’ 3 However, when passivizing intransitive – or rather, unergative – verbs, all the native speakers that I consult, including myself, agree that dropping the agent argument does not feel natural. 366 b. *Bei4 ta1 diao4 le0. 被她掉了 BEI she fall PERF (Intended) ‘It was fallen by her. (She fell.)’ To reorient a bit, BEI-passives can be used with transitive verbs as well as with intransitive verbs in Mandarin Chinese. When used with transitive verbs – as in (1) and (2) –, the agent argument maps onto the post-BEI position, whereas the patient argument can either appear before BEI (as in (1)), or map onto the post-verbal position (as in (2)). When used with intransitive verbs, only unergative verbs make a grammatical construction. This is illustrated by the contrast between the grammatical cases in (5) and the ungrammatical cases in (6), with BEI combining with unergatives and unaccusatives, respectively. A construction with the agent argument not overtly expressed is acceptable when transitive verbs are passivized, but does not feel natural with intransitive verbs. To give an analysis of the less studied passive constructions with BEI, especially the BEI-passive in which the patient argument of the transitive verb maps onto the post-verbal position, three questions need to be answered. First, about the grammatical function of the post-V NP: is it an object, or a subject that is somehow dislocated? Second, about the grammatical function of the immediate post-BEI NP: is it an object or an oblique? Third, about BEI itself: is it a preposition, a matrix verb taking thematic arguments, or something else? 2.2 Previous analysis of BEI-passive constructions To solve these puzzles raised in the last paragraph in 2.1, let us first survey some previous analyses of Chinese BEI-passive constructions. Attempts at analyzing Chinese BEI-passives treat BEI either purely as a passive marker (Xiao et al. 2006, Chow 2018, etc.), as a preposition (Zhu 1982, Li 1990, etc.), as a matrix verb taking two or three arguments (Ma 1985, Her 1989, 2009, Guo et al. 2007, etc.), or as a coverb (Kit 1998). Few words can be said about paths that simply treat BEI as a passive marker of the clause because they are untenable. In 2.2.1 we give a brief discussion on studies treating BEI as a preposition. In 2.2.2 we review two representative proposals that treat BEI as the matrix verb taking thematic arguments and discuss the problems they may face. 2.2.3 talks about approaches that treat BEI as a coverb and concludes. 2.2.1 BEI as a preposition Approaches treating BEI as a preposition mainly appear in early studies, such as Zhu (1982) or Li (1990). The claim is made by considering that BEI resembles the English preposition by in that both are followed by the agent argument in their respect passive construction. Now it is generally agreed that classifying BEI into preposition is not appropriate (Hsu 2009, Kit 1998, Guo 2007, Liu 2016, among others), given compelling evidence that preposition stranding is not allowed in Mandarin 367 Chinese (Huang 1991, Li 1990, among others)4 and that dropping the post-BEI element – at least in the passivization of transitive verbs – will not cause any ungrammaticality. Example (7) illustrates a case in which the verb da3 jia4 ‘fight’ in its active form needs a prepositional complement introduced by gen1 ‘with’. As one can see, the NP ta1 ‘he’ following the preposition gen1 ‘with’ cannot be dropped. In contrast, in a passive construction in (8), the agent argument following BEI can be freely dropped: (7) Wo3 mei2 you3 gen1 *(ta1) da3 jia4. 我没有跟他打架 I not with he fight ‘I did not fight with him.’ (Huang 1991) (8) Zhang1 san1 bei4 (Li3 si4) da3 le0. 张三被李四打了 Zhangsan BEI Lisi hit PERF ‘Zhangsan was hit (by Lisi).’ Therefore, though the “BEI + NP” sequence seems to resemble the English byphrase in introducing the agent argument of the passivized predicate, it is not logical to analyze BEI as the Chinese counterpart of the English preposition by. 2.2.2 BEI as a (thematic) argument-taking predicate Apart from assuming BEI to be a preposition, most studies adopt the approach of analyzing BEI as the matrix verb of the clause (Ma 1985, Her 1989, 2009, Guo et al. 2007, Hsu 2009, among others). As for the subcategorization of BEI however, no agreement has been reached. Ma (1985), Her (1989, 2009), among others, propose that BEI is a three-place predicate that selects a SUBJ, an OBJ, and a VCOMP (or an XCOMP, depending on different assumptions). For example, Her (1989) assumes that BEI subcategorizes for three functions in its PRED value and that it introduces two control equations. The lexical form of BEI as well as the control equations by Her (1989) are given in (9). 4 Notice that, circumpositions in Mandarin Chinese exist. These are a type of adposition combining a preposition that precedes an NP and a postposition following that NP (Liu 2002). In this case, “circumposition stranding” is allowed. For example: (i) Wo3 men2 yong4 qi4 che1 lai2 jie1 song 4 ke4 ren2. 我们用汽车来接送客人 we with car to pick-up client ‘We pick up clients with cars.’ (ii) Qi4 che1 wo3 men2 yong4 lai2 jie1 song4 ke4 ren2. 汽车我们用来接送客人 car we with to pick-up client ‘As for the car, we use it to pick up clients.’ (Liu 2002) (ii) without any element inside the circumposition yong4 lai2 ‘with … to’ is perfectly acceptable in Mandarin Chinese. However, preposition stranding is not allowed, as (7) shows. 368 (9) BEI, V (↑PRED) = ‘BEI <(↑SUBJ) (↑OBJ) (↑VCOMP)>’ (↑SUBJ) = (↑VCOMP OBJ) (↑OBJ) = (↑VCOMP SUBJ) (Her 1989) The f-structure in (10) gives a straightforward illustration of the control equations proposed by Her (1989): (10) Zhang1 san3 bei4 Li3 si4 ma4 le0. 张三被李四骂了 Zhangsan BEI Lisi curse PERF ‘Zhangsan was cursed by Lisi.’ PRED SUBJ OBJ VCOMP ‘BEI < SUBJ, OBJ, COMP >’ [PRED ‘Zhangsani’] [PRED ‘Lisi k’ ] PRED ‘ma ’ SUBJ [PRED ‘PRO k’ ] OBJ [PRED ‘Zhangsan’] (Her 1989) Concerning the subcategorization of BEI proposed by Her (1989), Hsu (2009) points out that the object-to-subject equi is doubtful within the existing theories of raising (Alsina 1996, Falk 2001, among many others). In addition, one has to explain “why Mandarin in particular allows this sort of object equi construction in addition to more standard cases of equi, and why other languages do not seem to do so at all.” Wong and Hancox (1999) in turn observe from the perspective of current Lexical Mapping Theories (LMTs, Bresnan and Kanerva 1989, Bresnan and Moshi 1990, etc.) that, if BEI should be treated as the matrix verb subcategorizing for an OBJ that maps onto the post-BEI position, then a clash would occur because the post-BEI element is both an OBJ (which is classified as [+o]) and an agent (which is intrinsically [-o]). Apart from the studies that assume BEI to subcategorize for three grammatical functions, some studies also assume that BEI only subcategorizes for a SUBJ and an XCOMP, as Guo et al. (2007) or Hsu (2009), among others. For example, Guo et al. (2007) propose that, for a passive sentence like the one in (11), the lexical form of BEI and the f-structure can be represented as in (12) and (13), respectively: 369 (11) Zhe4 xie1 shu4 ju4 bei4 wo3 hu1 lve4. 这些数据被我忽略 these data BEI I ignore ‘These data was ignored by me.’ (Guo et al. 2007) (12) BEI, V (↑PRED) = ‘BEI <(↑SUBJ) (↑COMP)>’ (↑SUBJ) = (↑COMP OBJ) (Guo et al. 2007) (13) F-structure for (11) PRED ‘BEI < SUBJ, COMP >’ SUBJ PRED ‘data’ DET [ PRED ‘these’ ] COMP 1 □ PRED ‘omit ’ SUBJ [PRED ‘I’ ] OBJ 1 □ (Guo et al. 2007) Assuming BEI to only subcategorizes for a SUBJ and a COMP avoids the feature conflicting problem in Her (1989), but still faces the object-to-subject equi problem: as one can see, the OBJ of COMP bears the same tag as the SUBJ of the matrix f-structure in (13) and again, an object-to-subject raising is quite suspicious. 2.2.3 BEI as a coverb Some words need to be said about previous analyses that treat BEI as a coverb (Li and Thompson 1989, Ramsey 1989, Kit 1998, etc.). Li and Thompson (1989) observe that BEI cannot be a (normal) verb in any context, since it must appear in a passive sentence together with another verb and cannot occur in a sentence alone. (14a) shows a normal passive clause in which BEI combines with a transitive verb pi1 ping2 ‘criticize’, whereas (14b) only contains BEI, leading to an ungrammatical construction: (14) a. Wo3 bei4 ma1 ma1 pi1 ping2 le0. 我被妈妈批评了 I BEI mother criticize PERF ‘I was criticized by mom.’ b. * Wo3 bei4 ma1 ma1. 我被妈妈 I BEI mother (Intended) ‘I was done something by mom.’ Li and Thompson (1989:365) 370 Further evidence supporting the assumption that BEI should not be fully treated as a verb comes from three verb tests in Li (1990:100): first, verbs can be attached to by aspect markers such as LE; second, verbs can appear in “Vnot-V” form in wh-questions; third, verbs can be used as a simple answer to a question. (15), (16) and (17) show that BEI behaves differently from normal verbs in that it can only pass the second test (i.e., the V-not-V test): (15) a. Ta1 mai3 le0 hen3 duo1 shu1. 他买了很多书 he buy PERF many book ‘He bought many books.’ Li (1990:100) b. * Zhang1 san1 bei4 le0 Li3 si4 da3. 张三被了李四打 Zhangsan BEI PERF Lisi hit ‘Zhangsan was hit by Lisi.’ (16) a. Ta1 mai3 bu4 mai3 shu? 他买不买书 he buy not buy book ‘Is he buying books?’ Li (1990:100) b. Zhang1 san1 bei4 mei2 bei4 Li3 si4 da3? 张三被没被李四打 Zhangsan BEI not BEI Lisi hit (Literally) ‘Was Zhangsan hit by Lisi or not?’ (17) a. – Ta1 mai3 shu1 ma0? 他买书吗 he buy book Q-PART ‘Is he buying books?’ – Mai3/ Bu4 mai3. 买 / 不买 buy / not buy ‘Yes/No.’ Li (1990:100) b. – Zhang1 san1 bei4 Li3 si4 da3 le0 ma0? 张三被李四打了吗 Zhangsan BEI Lisi hit PERF Q-PART ‘Was Zhangsan hit by Lisi?’ – *Bei4/ *Mei2 Bei4. 被没被 BEI / not BEI (Intended) ‘Yes/No.’ 371 The fact that BEI only shows the second property indicates that it should not be treated as a normal verb denoting action. Out of such considerations, Kit (1998) proposes that BEI is a coverb that shares properties of both prepositions and verbs. As for its lexical entry, Kit (1998) assumes that it subcategorizes for a SUBJ, an OBJ, and an XCOMP. For a passive clause like (10), the lexical entry of BEI is considered to be: (18) BEI, CV, PRED ‘BEI<(↑ SUBJ) (↑ OBJ) (↑ XCOMP)>’ (Kit 1998) As one can observe, approaches as such are not essentially different from those that treat BEI as a normal verb (like Her 1989), except that BEI is named as “coverb” to indicate the awareness that BEI has some properties that differentiate itself from normal verbs. In conclusion, previous studies treating BEI as a pure passive marker, a preposition, or the matrix verb taking thematic arguments all seem to be somehow untenable. First, one can consider BEI to be a passive marker to a certain extent, since it must appear – together with another verb – in a passive sentence. But BEI should not only be treated as a passive marker, because it shows partial verbal properties, as observed by Li (1990), Kit (1998), among others. Second, BEI is not a preposition, for the immediate post-BEI NP can be dropped without causing ungrammaticality and that Mandarin Chinese does not allow preposition stranding. Third, assuming BEI to be a (thematic) argument-taking verb also faces some problems: these arguments are in fact the arguments of the verb that BEI combines with, and control equations are needed to identify the functions subcategorized by BEI and the functions subcategorized by the verb combining with BEI. Then an object-to-subject raising is doubtful within existing theories of raising and equi, as in Her (1989) or Guo et al. (2007). Moreover, a SUBJ is assumed in all these studies and it is not clear how this approach would help to analyze the data described in (2)(5), in which the intuition is that no subject exists. Finally, studies calling BEI as a coverb show the awareness that BEI is not a normal verb, but contribute no essential difference from previous studies that treat it as the matrix verb. 2.3 Grammatical function of the post-V argument Before giving an analysis of BEI, let us first survey the grammatical function of the patient argument that maps onto the post-verbal position, such as those represented in (2). One may therefore wonder if the post-V NP in (2) is a dislocated subject. Based on a subjecthood test and an objecthood test, I argue that the grammatical function that this patient argument maps onto is an object. The subjecthood test I adopt is the floating quantifier test. Floating quantifier is a valid subjecthood test for many languages, like Catalan (Alsina 1996) or Tagalog (Kroeger 1993). It is also valid for Mandarin. For instance, the universal quantifier quan2 bu4/suo3 you3 ‘all’, when modifying the subject, 372 can appear either before the subject, left-attaching to it, or float to the position before the verb phrase: (19) a. Quan2 bu4 tong2 xue2 zai4 kan4 shu1. 全部同学在看书 all students PRES read book ‘All the students are reading books.’ b. Tong2 xue2 quan2 bu4 zai4 kan4 shu1. 同学全部在看书 students all PRES read book ‘The students are all reading books.’ By contrast, when modifying an object, such a quantifier can only appear right before the object that it modifies. Floating it to anywhere else is not allowed: (20) a. Tong2 xue2 zai4 kan4 quan2 bu4 shu1. 同学在看全部书 students all PRES read book ‘The students are reading all the books.’ b. * Tong2 xue2 quan2 bu4 zai4 kan4 shu1. 同学全部在看书 students all PRES read book (Intended) ‘The students are reading all the books.’ As we can see, only (20a) is grammatical. The universal quantifier that modifies the object shu1 ‘book’ appears right before it. In (20b), the quantifier floats to a pre-VP position. The sentence is grammatical when meaning “the students are all reading books”, as in (19b), but it is not grammatical when meaning “the students are reading all the books”. People may wonder if the phenomenon might be explained in terms of thematic roles – that is, if being patient/theme disallows the floating of its modifier –. For such a discussion, let us see a typical unaccusative clause with lai2 ‘come’. In Mandarin, unaccusative verbs such as lai2 ‘come’ allows its patient argument to appear either before or after it,5 as is shown in (21): (21) a. Ke4 ren2 lai2 le0. 客人来了 guest come PERF ‘Guests came.’ b. Lai2 le0 ke4 ren2. 来了客人 come PERF guest ‘Came guests.’ Examples in (22) illustrate the case in which the patient argument of the unaccusative verb come appears before the verb and allows quantifier floating: 5 Whereas the sole argument in unergative clauses can only appear pre-verbally. 373 (22) a. Quan2 bu4 ke4 ren2 lai2 le0. 全部客人来了 all guest come PERF ‘All guests came.’ b. Ke4 ren2 quan2 bu4 lai2 le0. 客人全部来了 guest all come PERF ‘Guests all came.’ However, when the argument appears post-verbally, floating its modifying quantifier to anywhere else is ungrammatical, as (23) shows: (23) a. Lai2 le0 quan2 bu4 ke4 ren2. 来了全部客人 come PERF all guest ‘Came all guests.’ b. * Quan2 bu4 lai2 le0 ke4 ren2. 全部来了客人 all come PERF guest (Intended) ‘Came all guests.’ The grammaticality of (23) with the patient argument being immediately preverbal rejects the assumption that quantifier floating can be settled in terms of thematic roles. A reasonable explanation is that, the patient ke4 ren2 ‘guests’, when being pre-verbal, is realized as the subject, thus allowing its quantifier to float. In contrast, when mapping onto the post-verbal position, the same patient NP is realized as the object and to float its quantifier is not allowed. Therefore, the fact that the pre-verbal patient NP in (24a), i.e., quan2 bu4 pan4 tu2 ‘all traitors’ allows its quantifier to float to the pre-verbal position, as in (24b), is evidence that it is the subject of the clause. (24) a. Quan2 bu4 pan4 tu2 bei4 ta1 dai4 bu3 le0. 全部叛徒被他逮捕了 all traitor BEI he arrested PERF ‘All traitors were arrested by him.’ b. Pan4 tu2 bei4 ta1 quan2 bu4 dai4 bu3 le0. 叛徒被他全部逮捕了 traitor BEI he all arrested PERF ‘Traitors were all arrested by him.’ Similarly, the fact that the post-verbal patient NP in (25a), which is also quan2 bu4 pan4 tu2 ‘all traitors’, does not allow its quantifier to float, is evidence that it is the object of the clause: (25) a. Bei4 ta1 dai4 bu3 le0 quan2 bu4 pan4 tu2. 被他逮捕了全部叛徒 BEI he arrest PERF all traitor ‘All traitors were killed by him.’ 374 b. * Quan2 bu4 bei4 ta1 sha1 le0 pan4 tu2. 全部被他杀了叛徒 all BEI he kill PERF traitor ‘Traitors were all killed by him.’ The second test is proposed by Zhang (2000), and I name it as “focus SHI test”. SHI is a multifunctional word with wide use in Mandarin Chinese. When it is used as a focus marker to introduce new information, it appears right before the element that needs to be introduced. (26b) and (26c) illustrate cases in which the SUBJ and the VP are marked as informationally new, respectively: (26) a. Ta1 xiu1 hao3 le0 zi4 xing2 che1. 他修好了自行车 he repair good PERF bike ‘He has repaired bike.’ b. Shi4 ta1 xiu1 hao3 le0 zi4 xing2 che1. 是他修好了自行车 SHI he repair good PERF bike ‘It is he that has repaired bike.’ c. Ta1 shi4 xiu1 hao3 le0 zi4 xing2 che1. 他是修好了自行车 he SHI repair good PERF bike ‘What he has done is repair bike.’ However, not all elements can be preceded by SHI, such as the case with objects. By contrast, a subject can be marked by SHI when being introduced as new information, as we have seen above. (27) shows such a case in which SHI precedes the post-verbal object and results in an ungrammatical structure: (27) Ta1 xiu1 hao3 le0 (*shi4) zi4 xing2 che1. 他修好了是自行车 he repair good PERF SHI bike (Literally) ‘What he has repaired is bike.’ (Zhang 2000)6 People may wonder if (27) is ungrammatical because the post-verbal element is marked as discourse old via its position and is therefore incompatible with SHI.7 The fact is that, the post-verbal NP can be either discourse old or new. It is incompatible with SHI even if it is discourse-new, shown in (28): (28) – Ta2 xiu1 hao3 le0 shen2 me0? 他修好了什么 he repair good PERF what ‘What has he repaired?’ 6 I have slightly adjusted the form of this example to give a unified presentation for all the examples in this paper. 7 I thank the anonymous reviewers for bringing this point out. 375 – Ta1 xiu1 hao3 le0 (*shi4) zi4 xing2 che1. 他修好了是自行车 he repair good PERF shi bike (Literally) ‘What he has repaired is bike.’ The same happens with unaccusative constructions. When the patient argument appears pre-verbally, it can be proceeded by SHI. In contrast, when it maps onto the post-verbal position and is proceeded by SHI, the construction is ungrammatical: (29) a. *Lai2 le0 shi4 ke4 ren2. 来了是客人 come PERF SHI guest (Intended) ‘It is the guests that came.’ b. Shi4 ke4 ren2 lai2 le0. 是客人来了 SHI guest come PERF ‘It is the guests that came.’ The difference between (29a) and (29b) further confirms our explanation of the quantifier floating contrast between (25a) and (25b). Here in (29b), the patient ke4 ren2 ‘guests’, being immediately pre-verbal, is realized as the subject, thus allowing the SHI marker. In (29a), the same patient is realized as the object when mapping onto the post-verbal position, thus the SHI marker is disallowed. Therefore, the fact that in passive constructions with transitive verbs, in which the patient argument is post-verbal and cannot be marked by SHI, provides evidence that this NP is the object and not the subject of the clause: (30) a. *Bei4 ta1 pian4 le0 shi4 wu3 ge4 da4 huo2 ren2. 被她骗了是五个大活人 BEI she cheat PERF SHI five CLASS living man (Intended) (literally) ‘What was cheated by her were five living men.’ b. *Bei4 ta1 cai3 zhao2 shi4 she2 le0 被她踩着是蛇了 BEI she step-on SHI snake PERF (Intended) (literally) ‘What was stepped on by her was a snake.’ c. * Bei4 feng1 chui1 dao3 le shi4 yi4 ke1 shu4 被风吹倒了是一棵树 BEI wind blow down PERF SHI one CLASS tree (Intended) (literally) ‘What was blown down by the wind was a tree.’ Before concluding this section, some words need to be said about the constructions in which no subject exists, as in (21b), (23a), or (25a), etc. (21b) reminds us of the example given by Kibort (2001), i.e., a locative inversion construction without a locative: (31) And then, came those visitors. (Kibort 2001) 376 To account for this, Kibort (2001) proposes a “demotion” approach within LMT.8 (31) is subjectless because, since there is no locative element, nothing can be “promoted” to the SUBJ. The Subject Condition as an inviolable constraint in traditional LMTs (Bresnan and Kanerva 1989, Bresnan and Moshi 1990, etc.) is thus challenged. If it is so, then it is not so surprising that some Mandarin constructions also lack a subject, as in (21b), (23a), or (25a), etc. 3 A unified approach to BEI constructions 3.1 The lexical entry of BEI By now it is clear that BEI is not a preposition, nor does it behaves totally like normal verbs. Now it is time to rethink BEI as a coverb, though previous analyses apparently name it as coverb but essentially treat it like other normal verbs (as in Kit 1998). Exploring BEI as a coverb (or, a light verb, in words of But 1993) that forms a complex predicate with the main verb seems to be a viable option. Light verbs in complex predicate constructions are studied in many languages, such as Catalan (Alsina 1996) or Urdu (Butt 1993). When analyzing Catalan causative constructions as a complex predicate, Alsina (1996) suggests that causative verbs (which equals light verbs) have two arguments: a causer and a causee. The causee binds an argument of the base verb (i.e., the main verb with which the light verb forms a complex predicate), mapping together onto the same function. Butt (1993) in turn considers the predicate composition to be a fusion process that melds the matrix patient with an argument of the embedded base verb. Once fused, the embedded argument is no longer available for mapping, and linking rules will only be used with the remaining arguments. Though formal representations in these two approaches differ, their essential spirit is the same: both of them assume permissive or causative construction to have a monostratal f-structure involving a single complex predicate formed by a light verb and the main verb, and both imply that predicate composition and argument mapping take place in the syntax, rather than in the lexicon. Let us see a causative construction in Catalan, represented in (32): (32) El mestre fa llegir un poema al nen. the teacher make read a poem to-the boy ‘The teacher is making the boy read a poem.’ (Alsina 1996:190) The corresponding a-structure is: (33) A-structure of fa-llegir ‘make-read’: 8 That is, to add the [+o] feature to the theme argument, thus demoting it to OBJ. 377 ‘cause < [P-A]2 [P-P]1 read <[P-A]1>>’9 (Alsina 1996:191) One can observe from previous studies on complex predicate – like Alsina (1996) or Butt (1993), etc. – that in these constructions, both the light verb and the main verb contribute their own thematic arguments. Then one can see an essential difference between BEI and light verbs (or coverbs). That is, whereas light verbs are at least partially predicated by taking an external argument, BEI does not have thematic arguments at all. All the thematic arguments one can find in a BEI-passive come from the verb that BEI combines with. This suggests that treating BEI as a coverb or light verb is not appropriate. If no thematic arguments can be contributed by BEI, nothing can be used to bind (in words of Alsina 1996) or to fuse with (in words of Butt 1993) an argument of the main verb.10 However, this is quite suggestive of raising verbs such as seem, which also takes no thematic arguments. For a sentence such as: (34) The geneticist seemed to clone dinosaurs. (Falk 2001:128) The lexical entry for seem is suggested to be represented as: (35) Seem: V (↑ PRED) = ‘seem <(↑ COMP)> (↑ SUBJ)’ (Falk 2001:128) This analysis can easily be adapted to the lexical entry of BEI and Mandarin passives. That is, to treat BEI as a raising verb and to treat Chinese passive to be a raising construction. In addition, given that BEI can only occur in passive constructions, and must co-occur with another (normal) verb (Li and Thompson 1989), I propose that BEI should also be the passive marker of the clause. As a raising verb, BEI lexically specifies both the grammatical function and the grammatical category of their single argument: it is an object and a CP:11 (36) Lexical entry of BEI: Bei, < Arg > | OBJ | CP As a passive marker, BEI blocks the linkage of the agent argument of the main verb to a direct grammatical function, but allows it to map onto an oblique. The agent argument can either go unexpressed or be expressed as a post-BEI NP. In this paper, I represent the argument-to-function mapping by using the 9 [P-A] is short for proto-agent, whereas [P-P] is short for proto-patient (Dowty 1991). By “main verb” I refer to the verb with which BEI combines, for it contributes the semantic content of the clause denoting action. 11 Here I follow Alsina et al. (2005), Forst (2006), and Patejuk and Przepiórkowski (2016), among others, in not assuming a COMP in the inventory of grammatical functions, and I assume that the syntactic function of the clausal phrase is OBJ, which maps onto the clausal constituent in c-structure. 10 378 same subscripted integers. The passivization is indicated by circling the subscripted index of that argument and the raising process is represented by marking two functions with the same tag. Then the f-structures of (1a) and (1b) can be represented as in (37) and (38), respectively: (37) Zhang1 san1 bei4 Li3 si4 da3 le0. 张三被李四打了 Zhangsan BEI Lisi hit PERF ‘Zhangsan was hit by Lisi.’ (repeating (1a) PRED ‘BEI < Arg1 >’ 1 SUBJ □ PRED ‘hit < Arg○ 2 , Arg3 >’ OBJ [PRED ‘Lisi’ ]2 OBL 1 [PRED ‘Zhangsan’] 3 SUBJ □ (38) 1 Zhang1 san1 bei4 Ø da3 le0. 张三被打了 Zhangsan BEI hit PERF ‘Zhangsan was hit.’ (repeating (1b)) PRED ‘BEI < Arg1 >’ 1 SUBJ □ PRED ‘hit < E○ 2 , I3 >’ OBJ 1 [PRED ‘Zhangsan’] 3 1 SUBJ □ In (37) and (38), the mapping of the agent argument of da3 ‘hit’ to a direct grammatical function is blocked. However, nothing prevents this argument to optionally be realized as an oblique. In this way, the optionality of the postBEI patient gets a natural explanation. As for the patient argument of da3 ‘hit’, it maps onto the SUBJ of the passivized predicate, which then raises as a nonthematic SUBJ of BEI. Note that here we are assuming an NP to bear the oblique function. This is not a problem for Mandarin Chinese, for obliques do not necessarily require a preposition/postposition in this language, as Ma (1985) and Tan (1987) observe. Evidence for this claim comes from NPs denoting location. That is, both NPs and PPs can be used to denote location in Mandarin Chinese: (39) a. Yi4 jian1 fang2 zhu4 lia3 ren2. 一间房住俩人 one CLASS room live two person ‘In one room live two persons.’ 379 b. Zai4 yi4 jian1 fang2 zhu4 lia3 ren2. 在一间房住俩人 in one CLASS room live two person ‘In one room live two persons.’ Therefore, assuming that the post-BEI agent bears the OBL function is not a problem for Mandarin. A similar assumption about the post-BEI agent is also adopted by Chow (2018) for Cantonese. 3.2 Patient in post-V position Let us then turn to the passive construction in which the patient argument of the transitive verb maps onto the immediate post-V position. I have argued in 2.3 that when appearing post-verbally, the patient argument is the object of the clause. In (39), the internal argument of BEI is expressed as the SUBJ of the passivized transitive verb cai3 ‘step-on’, but when raising to the matrix fstructure, it is the non-thematic OBJ of BEI: (40) Bei4 ta1 cai3 zhao2 she2 le0. 被她踩着蛇了 BEI she step-on snake PERF ‘A snake was stepped on by her.’ (repeating (2b)) PRED ‘BEI < Arg1 >’ 1 OBJ □ PRED ‘step-on < Arg○ 2 , Arg3 >’ OBJ OBL [PRED ‘pro’ ]2 1 [PRED ‘snake’] 3 SUBJ □ 1 When BEI combines with intransitive verbs, the linkage of the sole argument is suppressed and it maps onto an oblique function.12 Then no argument is left to map onto a direct grammatical function, thus no function is raised to the matrix f-structure. The f-structure in (41) illustrates this point. (41) Bei4 ta1 pao3 le0. 被她跑了 BEI she run PERF ‘She ran.’ (repeating (5a)) 12 Recall that the agent argument in passive constructions with intransitive verbs is always expressed, though theoretically, it is optional. The reason is not clear at the moment. I leave this issue for further study. 380 PRED ‘BEI < Arg1 >’ PRED ‘run < Arg○ 2 >’ OBJ OBL [PRED ‘pro’ ]2 1 Recall that, when combining with intransitive verbs, only unergatives make the construction grammatical. Unaccusative verbs cannot be passivized by BEI, which contributes evidence to confirm the Unaccusative Hypothesis. I will not go into details here for the space limitation. One may wonder if there is a thematic null subject in (40) and (41), as opposed to our assumption that no subject exists in the matrix clause, since “a lot of things called passive in the literature turn out not to be passive in any useful sense. Some so-called subjectless constructions turn out to have thematic null subjects. (Maling 2010)” Yet in the very same paper, the author points out that in real impersonal active constructions, an agentive by-phrase is impossible. Here in (40) and (41) however, the agent argument can appear as an oblique,13 and there are no other thematic arguments left to map onto a “thematic null subjects”. Constructions such as (40) and (41) are real subjectless constructions, which implies that a grammatical subject is not necessary for every clause. 4 Conclusions This paper analyzes the less studied BEI-passive constructions in Mandarin Chinese, i.e., passive constructions in which the patient argument of the transitive verb appears post-verbally, as well as passives with intransitive verbs. BEI is proposed to be a raising verb and the passive marker that blocks the linkage of the agent argument of the verb that it combines. An implication is that a subject is not necessary for every clause. The Subject Condition as an inviolable well-formedness condition in current mapping theories should therefore be reconsidered. Two points remain unsolved and are left to further studies. First, in passive constructions with unergative verbs, it is not clear why it is obligatory to express the agent argument as an oblique function, which is theoretically an optional operation. Second, it is not clear why a post-V patient in a passive clause (with transitive verbs) maps onto an OBJ whereas a pre-BEI patient maps onto the SUBJ of the construction. Given the importance of word order in analytic languages, the role of word order in the mapping process needs to be further explored to get a better comprehension of argument realization issues in Mandarin Chinese. 13 And in fact, must appear, in (41), though the reason remains unknown to us at the moment. 381 References Alsina, Alex. 1996. The Role of Argument Structure in Grammar: Evidence from Romance. Stanford: CSLI Publications. Alsina, Alex, K. P. Mohanan, and Tara Mohanan. 2005. How to get rid of the COMP. In Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG05 Conference, 21–24. Stanford: CSLI Publications. Bresnan, Joan and Jonni Kanerva. 1989. Locative inversion in Chicheŵa: a case study of factorization in grammar. Linguistic Inquiry 20(1), 1-50. Bresnan, Joan and Lioba Moshi. 1990. Object asymmetries in comparative Bantu syntax. Linguistic Inquiry 21(2), 147-186. Butt, Miriam. 1993. The Structure of Complex Predicates in Urdu. PhD dissertation. Stanford University. Chow, Pui-Lun. 2018. Extending the applicability of lexical mapping theory. In Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG18 Conference, 191-207. Stanford: CSLI Publications. Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67(3), 547-619. Falk, Yehuda. 2001. Lexical-Functional Grammar: an Introduction to Parallel Constraint-Based Syntax. Stanford: CSLI Publications. Forst, Martin. 2006. COMP in (parallel) grammar writing. In Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG06 Conference, 222– 239. Stanford: CSLI Publications. Guo, Yuqing, Josef van Genabith and Haifeng Wang. 2007. Treebank-based acquisition of LFG resources for Chinese. In Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG07 Conference, 214-232. Stanford: CSLI Publications. Her, One-Soon. 1989. An LFG account for Chinese BEI sentence. Journal of the Chinese Language Teachers Association 24(3), 67-89. Her, One-Soon. 2009. Unifying the long passive and the short passive on the Bei construction in Taiwan Mandarin. Language and Linguistics 10(3), 421470. Hsu, Brian. 2009. A Long-distance Dependency Analysis of Mandarin BeiPassive in LFG. Haverford College. Huang, Cheng-Teh James. 1991. Modularity and Chinese A-not-A questions. In Carol Georgopoulos and Roberta Ishihara (eds.), Interdisciplinary Approach to Linguistics: Essays in Honor of Yuki Kuroda, 305-332. Kluwer Academic Publishers. Huang, Cheng-Teh James, Audrey Yen-Hui Li, and Yafei Li. 2009. The Syntax of Chinese. Cambridge: Cambridge University Press. Kibort, Anna. 2001. The Polish passive and impersonal in Lexical Mapping Theory. In Miriam Butt and Tracy Holloway King (eds.), Proceedings of the LFG01 Conference, 163-183. Stanford: CSLI Publications. Kit, Chunyu. 1998. Ba and bei as multi-valence prepositions in Chinese – an 382 LFG perspective. In Benjamin K. T’sou (ed.) Studia Linguistica Sinica, 497522. City University of Hong Kong: Language Information Sciences Research Centre. Kroeger, Paul. 1993. Phrase Structure and Grammatical Relations in Tagalog. Stanford: CSLI Publications. Li, Charles N. and Sandra A. Thompson. 1989. Mandarin Chinese: a Functional Reference Grammar. Berkeley: University of Califonia Press. Li, Yen-Hui Audrey. 1990. Order and Constituency in Mandarin Chinese. Dordrecht: Kluwer Academic Publishers. Liu, Danqing. 2002. Hanyu zhongde kuangshi jieci (Circumpositions in Chinese). Dangdai Yuyanxue (Contemporary Linguistics) 4(4), 316-317. Liu, Na. 2016. The structures of Chinese long and short Bei passives revisited. Language and Linguistics 17(6), 857-889. Lu, Jianming. 2004. Youguan beidongju de jige wenti (Some issues on passive). Hanyu Xuebao (Chinese Linguistics) 2, 9-15. Ma, Lizi. 1985. The classical notion of passive and the Mandarin ‘bei’. Ms. Stanford University: Department of Linguistics. Maling, Joan. 2010. Unpassives of unaccusatives. In Donna B. Gerdts, John Moore, and Maria Polinsky (eds.), Hypothesis A/Hypothesis B: Linguistic Explorations in Honor of David M. Perlmutter, 275-292. The MIT Press. Patejuk, Agnieszka and Adam Przepiórkowski. 2016. Reducing grammatical functions in LFG. In Doug Arnold, Miriam Butt, Berthold Crysmann, Tracy Holloway King and Stefan Müller (eds.), Proceedings of the Joint 2016 Conference on Head-driven Phrase Structure Grammar and Lexical Functional Grammar, 541–559. Stanford: CSLI Publications. Ramsey, S. Robert. 1989. The Languages of China. Princeton: Princeton University Press. Tan, Fu. 1987. The predicate argument structure of bei. Proceedings of Berkeley Linguistics Society 13, 285-295. Wong, Shun Ha Sylvia and Peter Hancox. 1999. What is the lexical form of ‘Bei’? Language, Information and Computation. Proceedings of the 13th Pacific Asia Conference, 169-176. Xiao, Richard, Tony Mcenery and Yufang Qian. 2006. Passive construction in English and Chinese: a corpus-based contrastive study. Language in Contrast 6(1), 109-149. Yu, Guangzhong. 1989. Lingzhuyu beiziju. (BEI sentences with no subject). Yuyan Yanjiu (Studies in Language and Linguistics) 2, 95-104. Zhang, Ning. 2000. Object shift in Mandarin Chinese. Journal of Chinese Linguistics 28(2), 201-246. Zhu, Dexi. 1982. Yufa Jiangyi (Lectures on Syntax). Beijing: The Commercial Press. 383