EVALUATIVE ADVERBS IN MANDARIN WH-QUESTIONS Xin-xian Rex Yu National Tsing Hua University 1. Introduction Evaluative adverbs (EAs), such as haozai 'luckily', xingkui 'fortunately' and jingran 'unexpectedly' in Mandarin Chinese (henceforth Mandarin) are defined to express speaker's judgments on the proposition or the fact (Palmer 1986; Hoye 1997; Cinque 1999; Tang 2000). EAs in Mandarin could be listed as following, regarding to the morphological suffix –de: 1 (1) (a) with identical suffix (b) without identical suffix xingyunde ‘luckily’ xingkui ‘fortunately’ xinghao ‘fortunately’ buxingde ‘unfortunately’ fanzheng ‘at least’ haozai ‘luckily’ churenyilaode ‘unexpectedly’ juran ‘unexpectedly’ lingrenyiwaide jingran ‘unexpectedly’ ‘unexpectedly’ kexi ‘it’s pity that…’ momingqimiaode ‘inexplicable’ haihao ‘not-bad’ …… pianpian ‘contrarily’ Starting from the distributions of EAs, I am going to clarify why EAs are so differential from other kinds of adverbs, such as daodi ‘the hell’: (2) a. Weisheme Zhangsan canjia-le huiyi ? why Zhangsan attend-LE meeting ‘Why did Zhangsan attend the meeting?’ 1 In Mandarin, the adverbial suffix –地 can be pronounced either as -de or -di, when it attaches to adverbs. Regardless of the dialectal disputes, I consistently mark with –de in this paper. Nanzan Linguistics: Special Issue 1, Vol. 1, 187-208 © 2007 Xin-xian Rex Yu Nanzan Linguistics: Special Issue 1, Vol. 1 b. Daodi weisheme Zhangsan canjia-le huiyi ? the hell why Zhangsan attend-LE meeting 'Why the hell did Zhangsan attend the meeting?' c. * Haozai weisheme Zhangsan canjia-le huiyi ? luckily why Zhangsan attend-LE meeting (3) a. Lisi ai-shang-le shei? Lisi loved who 'Who did Lisi fall in love with?' b. Daodi Lisi ai-shang-le shei? the hell Lisi loved who ‘Who the hell did Lisi fall in love with?’ c. Kanlai Lisi ai-shang-le shei. seemingly Lisi loved who ‘Lisi seems to have fallen in love with somebody.’ (Xu 1990:358) d. * Xingkui Lisi ai-shang-le shei ?/. fortunately Lisi loved who? When daodi ‘the hell’ occurs at the beginning in (2b) and (3b), the sentence are grammatical, expressing questions with strong emphasis on the answers. Following Kuo (1997), daodi should be interpreted as “the hell”, preceding wh-phrases in wh-questions, and as ‘after all’ in non-wh-questions. As to kanlai ‘seemingly’ in (3c), Xu (1990) considers it as a modal, and the whole sentence should be interpreted with indefinite reading. However, these situations would be totally different when EAs occur. Once haozai 'luckily' and xingkui 'fortunately' occur in the sentences, both (2c) and (3d) become ungrammatical. However, these effects do not take place when wh-questions are initialized with another type of EAs in (1): (4) Lingrenyiwaide, weisheme Zhangsan qu-le Riben ? unexpectedly why Zhangsan went Japan 'Why did Zhangsan unexpectedly went to Japan?' (5) Xingyude, ta ai-shang-le shei? luckily he loved who ‘Who did he luckily fall in love with?’ Focusing on the interaction between EAs and wh-phrases in questions found in Mandarin, I propose two main issues composed of four questions dealing with these phenomena: - 188 - Evaluative Adverbs in Mandarin Wh-questions (X.-X. R. Yu) (6) The properties of EAs and their relations with other elements in Mandarin What kinds of influences would they make on wh-phrases? Why can they determine the grammaticality of the whole sentence? (7) Sub-classification of EAs in Mandarin Are they all derived from corresponding adjectives? Does any difference occur among them? How do the differences affect their interactions with wh-phrases? At the beginning, we need to consider the specific semantic property of EAs, which results in the contrasts in (2) and (3) above. The syntactic distributions of EAs would then be discussed, based on Cinque's (1999) universal hierarchy of adverbs, as well as on Rizzi's (2004) split CP structure. Relativized Minimality, adjusted by Rizzi (2004) can properly predicate the behaviors of adverbs in Mandarin. With slightly different from Rizzi's predictions, some EAs in Mandarin contain different features, which results in the ungrammaticality displayed in (2c) and (3c). 2. Characteristic properties of EAs 2.1 Semantic properties of EAs In general, adverb is just a syntactic category, while adverbials refer to expressions that modify verbs or “sentential” objects (Ernst 2002). Following Jackendoff (1972) and Ernst (2002), EAs should be regarded as speaker-oriented adverbs, belonging to the same group with speech-act adverbs ‘frankly’ and epistemic adverbs ‘obviously’:2 (8) predicational—composed of an adverb and its argument. speaker-oriented: frankly, maybe, luckily, obviously subject-oriented: deliberately, stupidly exocomparative: similarly event-internal: tightly, partially (Ernst 2002:9) According to (8) above, EAs are classified as predicational adverbials, since EAs are sensitive to the following propositions, and presuppose the truth of that propositions or events: (9) a. Tingshuo Lisi canjia-le huiyi. (epistemic adverb) allegedly Lisi participated meeting ‘Allegedly, Lisi participated the meeting.’ 2 In Bellert’s (1977) classification, EAs, domain adverbs, pragmatic adverbs, modal adverbs and perhaps are grouped as predicational adverbs. - 189 - Nanzan Linguistics: Special Issue 1, Vol. 1 b. Xingkui Lisi canjia-le huiyi. (evaluative adverb) fortunately Lisi participated meeting ‘It is fortunate that, Lisi has participated the meeting.’ c. Wo houhui canjia-le huiyi. (factive verb) I regret participated meeting ‘I regretted participating the meeting.’ As the addressees of (9a), we are not exactly sure whether Lisi did or did not participate in the meeting. But the event in (9b) that Lisi has taken part in a meeting could definitely be interpreted as nothing but true. The truth of the event is presupposed by the existence of EAs. Moreover, this phenomenon is similar to those sentences created by factive predicates. Factive predicates, such as houhui ‘regret’ in (9c), also presuppose the realization of the event as well as the truth of that event. No one would feel regretful if nothing has happened. Identically, no speaker could judge any event if it has not occurred yet. 2.2. Syntactic distribution Generally, sentential adverbs have relatively free distributions, with regard to the distributions of Subject and Auxiliary (Ernst 2002:114): 3 (10) a. Pianpian xiaozhang ye xiang qu. contrarily principal also want go b. Xiaozhang pianpian ye xiang qu. principal contrarily also want go 'It is contrary that the principle also wanted to go.' (Zhang 2000b: 63) These alternative distributions on sentential adverbs have gotten much attention and elaborate assumptions from various perspectives. For instance, xingkui 'fortunately' in Mandarin has been analyzed either as a raising verb (Tsao 1996), or as a sentential adverb (Zhang 2000ab, Cinque 1999, and among others). Conveying speaker's judgment of fortunateness on the fact, Tsao (1996) treats xingkui 'fortunately' in Mandarin as a kind of modal verbs, which can clearly separate the sentence topics (elements with double-underline in (11)) and clausal topics (elements with single-underline in (11)). With relatively free ordering among these topics, various alternations are allowed: 3 Pianpian 偏偏 in Mandarin implies a situation in which the speaker slightly blames the Agent for his/ her stubborn persistence in his/ her opinions, which are contrary to that of speaker. In this thesis, I translate it as ‘contrarily’, though these two words are not perfectly correspondent. - 190 - Evaluative Adverbs in Mandarin Wh-questions (X.-X. R. Yu) (11) a. Ta A-hua qunian xingkui meiyou qu. he A-hua last year fortunately not-have marry b. Ta A-hua xingkui qunian meiyou qu. c. Qunian ta A-hua xingkui meiyou qu. 'It is fortunate that he did not marry A-hua last year.' (Tsao 1996:178) However, Tsao's assumptions would encounter some challenges on the distinction between verbs and adverbs. First of all, verbs can be negated by bu 'not' or meiyou 'not-have', or construct an A-not-A form in yes/no questions. For example, bu-qu 'not go', qu-bu-qu 'go or not' and bu-keneng 'can not', ke-bu-keneng 'can or can not'. But it is not the case for xingkui and haozai.4 Moreover, unlike normal verbs in Mandarin, xingkui does not have any verbal inflections, such as tense or aspectual markers. This criterion may not be precise for raising verbs in English, but it reliable for raising verbs in Mandarin. Due to these doubts on the verb perspective, I alternatively treat EAs as adverbs through this paper, and adopted Cinque’s universal hierarchy among adverbs. 2.3. Frameworks 2.3.1 Cinque’s universal hierarchy In Cinque (1999), adverbs are base-generated in the specifier of related functional heads. And the projection of each functional head is ranked with relatively strict hierarchical ordering. Crosslinguistically, EAs in Cinque's hierarchy occupy the specifier of the head named evaluative mood, surrounded by speech-act mood and evidential mood: (12) [Mood [Mood [Mood [Mod [T(Past)[T (future)… ]]]]]] speech act evaluative evidential epistemic (Cinque 1999:106) This universal hierarchy exhibits a strict ordering among adverbs without exchanges, based on the observations on the distributions of adverbs. And this rigorous ordering in adverbs can be attributed to the same inherited feature shared by adverbs. In Rizzi (2004), this shared feature is presumed as modifier feature, which is the key concept to build minimal configurations, according to Relativized Minimality. 2.3.2 Relativized Minimality Relativized Minimality (RM), in Rizzi’s (1990, 1997, 2001, 2004) definition, is a kind of locality principles, and emerges its significance in accounting for the issues of syntactic movements, which challenged previous principles, such as Subjacency Condition and Empty Category Principle (ECP) without satisfied results. 4 Some EAs, such as xingyunde 'luckily' and yiwaide 'unexpectedly', can be negated by bu 'not', but it is another story that we will discuss in section 4. - 191 - Nanzan Linguistics: Special Issue 1, Vol. 1 First of all, RM contributes in defining syntactic constructions for the properties of creativity and recursion of languages. "A fundamental discovery of modern formal linguistics is that if the length and depth of syntactic representations is unbounded, core structural relations are local." (Rizzi 2004:1) RM provides a smallest environment, in which relations between elements must be locally satisfied. In other words, RM sets up the boundaries for a core syntactic structure, and forbids elements connecting or building relations across these minimal configurations. These delimitations, however, are variable depending on the properties of the components. For labeling these boundaries of different relations, Rizzi defines a minimal configuration as follows: (13) a. X....Z....Y Y is in a Minimal Configuration (Mandarin) with X iff there is no Z such as that Z is of the same structural type as X, and Z intervenes between X and Y (Rizzi 2001:(4), 2004:(4)) As to the connection between the displaced element and its trace, the concept of chain is mentioned. In Rizzi’s assumption, chain relation, including Head chain, A chain and A’ chain, must be represented through RM at LF. 2.3.3 Adverbs under RM As locality constraints, these chain distinctions also account for the argument/ adjunct asymmetries (Rizzi revised of 1991NELS), as well as the hierarchical ordering among adverbs in Dutch (Koster 1978) and Italian (Rizzi 2004). For preposed adverbs in Italian, Rizzi observed that they behave like topics, but in distribution, they may not naturally precede the wh-phrases, while topics can (Rizzi 2004:15-16). From this phenomenon, the preposed adverbs are proved to occur lower than topics. Meanwhile, due to different interactions with wh-phrases, topics are claimed to determine the island effect on A’ movement, while preposed adverbs do not. For precisely indicating the position of these preposed adverbs, Rizzi created a phrasal slot between Force and Fin in the structure of CP system, named Mod(ifier): (14) Force Top* Int Top* Focus Mod* Top* Fin IP (Rizzi 2004:18) Following Rizzi’s assumptions, adverbs that move to Mod occupy the Spec position and - 192 - Evaluative Adverbs in Mandarin Wh-questions (X.-X. R. Yu) “licensed by the substantive featural content of their heads” (Rizzi 2004:19). According to these substantive features, Rizzi predicts the typology of specifiers: (15) a. Argumental: person, number, gender, case b. Quantificaitonal : Wh, Neg, measure, focus... c. Modifier: evaluative, epistemic, Neg, frequentative, celerative, measure, manner... d. Topic (Rizzi 2004:19) In (15), only the specifiers that belong to the same classification in (15) are stated to interact with each other in sentences. Take adverbs in Dutch as examples, waarschijnlijk ‘probably’ is not allowed to move across helaas ‘unfortunately’, as long as they co-occur: (16) a. Het is zo dat hij helaas waarschijnlijk ziek is ‘It is so that he unfortunately probably sick is’ b. * Het is zo dat hij waarschijnlijk helaas ziek is ‘It is so that he probably unfortunately sick is’ (17) a. Helaas is hij ___ waarschijnlijk zick ‘Unfortunately is he probably sick’ b. * Waarschijnlijk is hij helaas ___ zick ‘Probably is he unfortunately sick’ c. Waarschijnlijk is hij ___ zick ‘Probably is he sick’ (Koster 1978 in Rizzi 2004:10-11) This blocking situation can also be predicted in English and in Mandarin: (18) a. Fortunately, he had evidently had his own opinion of the matter. b. Evidently he had fortunately had his own opinion of the matter. (Cinque 1999:33) (19) a. Henbuxingde, tamen keneng dui ni you pianjian. unfortunately they probably to you have prejudice 'It is unfortunate that they might have prejudice in you.' b. * Keneng, henbuxingde tamen dui ni you pianjian. probably unfortunately they to you have prejudice (inspired from Cinque 1999) EAs in Mandarin, however, represent more complicated situations when they co-occur with wh-phrases: - 193 - Nanzan Linguistics: Special Issue 1, Vol. 1 (20) a. Weisheme haozai Zhangsan canjia-le huiyi ? why luckily Zhangsan attend-LE meeting 'Why is it lucky that Zhangsan attend the meeting?' b. * Haozai weisheme Zhangsan canjia-le huiyi ? luckily why Zhangsan attend-LE meeting (21) a. Kanlai ta ai-shang-le shei. seemingly he full-in-love-LE who ‘He seems to have fallen in love with somebody.’ b. * Pianpian ta ai-shang-le shei ?/. fortunately ta loved who If EAs in Mandarin only affect the adverbs in the same subclass, that is, they only have Modifier feature, how could we explain the interaction between EAs and wh-phrases in (20) and (21)? It seems that EAs and wh-phrases also share the same feature, which results in the crash of (20b) and (21b). RM, however, could not exactly account for these situations. These observations lead us to the first issue: (6) The properties of EAs and their relations with other elements in Mandarin i. What kinds of influences would they make on other elements in sentences? ii. Why can they determine the grammaticality of the whole sentence? 3. Interaction between EAs and wh-phrases Beyond Rizzi’s predictions, the blocking effects caused by EAs in Mandarin are more complicated: not only leaving effects on other adverbs for ordering, but they also interact with wh-phrases, including wh-adverbs and wh-nominals.5 Observing these interactions, I further assume that an EA should be considered as an intervener in wh-questions. 3.1 Interveners and intervention effect In Beck’s (1996) hypothesis, certain quantifiers interrupting in a sentence would construct a structure as barriers for intervening LF movements. This structure can be illustrated as following: (22) *[...X...[Q...[...tLF...]]] (Beck 1996:1) i i The Q in (22) represents the position of those specific quantifiers in Beck’s definition, and interrupts into the path between X and its trace at LF. With further observations, Pesetsky (2000) further notices a phenomenon that the intervention effect occurs, as long as wh-feature movement is restricted under the structure of (22). He then claims that once intervention 5 Only Adv-EAs, the adverbs without uniform suffix –de, are discovered to interact with wh-phrase, while Adj-EAs are not. The differences would be discussed in section 4. - 194 - Evaluative Adverbs in Mandarin Wh-questions (X.-X. R. Yu) effect takes place, LF feature movement could be detected. Following Pesetsky's prediction on intervention effect, Soh (2005) presumes that the existence of intervention effect on adverbial wh-phrases exactly implies the involvement of covert feature movement. Nominal wh-phrases, on the other hand, take a covert phrasal movement, which would not be affected by blocking of interveners. Furthermore, Soh claims that the interveners in Mandarin, such as zhi ‘only’, bu ‘not’, meiyouren ‘nobody’, henshao ren ‘few people’, zuiduo liang-ge ren ‘at most two persons’, chang ‘often’ and ye ‘also’, block the covert feature movement of weisheme ‘why’ at LF: (23) a. Ni weisheme shuo meiyouren cizhi? you why say nobody resign 'What is the reason x such that for x you said nobody resigned?' b. * Meiyouren weisheme shuo ni cizhi? Dealing with the relations between quantifiers and in-situ wh-phrases in interrogatives, as well as yes/no questions in Mandarin, Soh discovers that quantifiers are forbidden to precede the in-situ wh-phrase. According to Beck’s assumption and Pesetsky’s prediction, these intervention effects could be schematized as following: (24) *[Op…intervener…wh-adverb] 3.1.1 EAs and weisheme ‘why’ In Mandarin, weisheme ‘why’, as a wh-adverb that can not move out from island constructions, are discovered to make an interrogative sentence in two positions— preceding or following the subject: (25) a. Weisheme Xiaozhang hui canjia huiyi? why Xiaozhang will attend meeting b. Xiaozhang weisheme hui canjia huiyi? Xiaozhang why will attend meeting 'What did Xiaozhang attend the meeting for?' In Tsai (2000), weisheme 'why' in Mandarin, occurring between auxiliary and verb, should be interpreted with interrogative questioning on purpose. Alternatively, when weisheme precedes Auxiliary, the interpretation of asking reason would appear. And these two different distributions represent different results, when EAs occur: - 195 - Nanzan Linguistics: Special Issue 1, Vol. 1 (26) a. Xiaozhang weisheme neng canjia wuhui ? Xiaozhang why can join party 'Why can Xiaozhang join the party?' b. * Xinghao Xiaozhang weisheme neng canjia wuhui ? fortunately Xiaozhang why can join party c. * Xiaozhang xinghao weisheme neng canjia wuhui ? Xiaozhang fortunately why can join party (27) a. Weisheme Xiaozhang neng canjia wuhui ? why Xiaozhang can join party 'Why can Xiaozhang join the party?' b. ? Weisheme Xiaozhang xinghao neng canjia wuhui ? why Xiaozhang fortunately can join party 'Why it is fortunate that Xiaozhang can join the party?' c. ? Weisheme xinghao Xiaozhang neng canjia wuhui ? why fortunately Xiaozhang can join party 'Why it is fortunate that Xiaozhang can join the party?' For all the LF movement examples in (26), sentences fail to compose interrogatives, because of the interruption of xinghao ’fortunately’. As to those examples of covert movement in (27), wh-adverbs successfully attach to the Spec of CP, regardless of any interruption. Identical situations on weisheme are also discovered by Tsai (2000). He assumes that weisheme should take an abstract movement to the Spec of CP to construct an interrogative. "It would violate the relativized minimality, if this abstract movement takes place in the scope of auxiliary, frequency adverbs, temporal adverbials, modal adverbs and negations" (2000: 55). These blocking effects on LF movement can be schematized as following: (28) * [ weisheme.....[EA/ zhi/Aux/ Freq/ Temp/ Modal/ Neg.....t..]] LF CP i i 3.2 Interaction between EAs and wh-nominals in Mandarin Although great distinctions had discussed between wh-adverbs and wh-nominals, similar results would be discovered when EAs co-occur with wh-nominals, such as sheme ‘what’ and shei ‘who’ in Mandarin. With indefinite reading on wh-nominals, sentences could be interpreted as either interrogatives or declaratives. EAs, however, would fail both readings. 3.2.1 The indefinite interpretation of wh-nominals Wh-nominals shei ‘who’ and sheme ‘what’ in Mandarin can be interpreted with either interrogative or non-interrogative reading (Huang 1982; Li 1992; Aoun & Li 1993). Tsai - 196 -