"Dou" in Wh-Questions

“Dou” in Wh-Questions Hongyuan Dong Manuscript, 2008 Cornell University In this paper, I argue that in Mandarin Chinese the adverb “dou” can be associated with a wh-phrase to its right in surface syntax, and the semantics of this “dou” is equivalent to the exhaustivity marker as proposed by Beck and Rullmann (1999). This marker operates on a question denotation and returns the weakly exhaustive answer. I also argue that the Leftness Condition can be maintained even in this case of seemingly rightward association, if we assume LF-movement of whphrases in Mandarin Chinese. Then it is shown that this “dou” is not possible in an embedded wh-question with a verb that does not take questions as arguments, arguably due to lack of LF-movement. Then I discuss examples with both a universal quantifier and a wh-phrase in one sentence, and point out that there are certain conditions on which element this “dou” can be associated with, and I formulize this condition as an adjacency condition, based on the usual c-command relation. 1. Introduction In Mandarin Chinese, the adverb “dou” can be used in a wh-question, for example: (1) Zhangsan dou mai-le shenme? Zhangsan all buy-ASP what What all did Zhangsan buy ? The issue here is whether this “dou”, as in (1), is associated with the wh-phrase “shenme” to its right. There has been an abundance of work on this issue, and the general agreement is that “dou” can only be associated with an element to its left. This is called the Leftness Condition. Therefore accordingly, in sentences like (1), the “dou” should also be associated with some element to its left, or to put it in a milder way, this “dou” cannot be associated with the wh-phrase to its right. One representative proponent of this view is Jiang (1998). He argues that in such cases the 1 “dou” is associated with some element in the presupposition. First, there is a retrievable presupposition that says something like “Zhangsan has bought some things”, and then “dou” can be associated with the plural NP “things” in the presupposition. Of course this does not amount to saying that this “dou” is associated with something to its left. But at least it would not be associated with the wh-phrase to its right. This is what he wants to argue for. Moreover the presupposition is in some sense to the left of “dou”. For one thing, the plural NP is aforementioned. Second, if the presupposition is written in parentheses to the left of the sentence, which is a way of indicating the content of the presupposition, then the associated element is again to the left of “dou”. Jiang (1998) also cites Xu (1985) as saying that the reason why “dou” cannot be associated with a wh-phrase is that they do not denote a set of individuals. Appealing as it is, I think there are many problems with this argument. First, why should “dou” be associated only with “a set of individuals”? Intuitively, “dou” implies that there are multiple things to consider, but these things can be either simple entities such as individuals, or complex semantic objects. Even if wh-phrases do not denote a set of individuals, nothing prevents them from being associated with “dou” as long as there is plurality on some level of semantic representation. Second, in surface syntax, “dou” is indeed associated with the wh-phrase. If the whphrase is to the left of “dou”, then the sentence wouldn’t be a question at all. E.g. (2) Shei dou xihuan Zhangsan. Who all like Zhangsan Everyone likes Zhangsan In (2), the wh-phrase “shei” is to the left of “dou” in surface syntax, and the whole sentence cannot be interpreted as a question. Therefore it seems that if “dou” is ever used in a question, then the wh-phrase has to be to its right, even when the wh-phrase is in the subject position. For example: (3) Dou (you) shei xihuan Zhangsan? All have who like Zhangsan Who all likes Zhangsan? 2 In (3), the “dou” has to move to the front of the wh-phrase. Although it sounds better with the existential verb “you”, it is not indispensable in fast speech. Thus at least in cases like (3), “dou” is right next to the wh-phrase. If “dou” is not associated with the wh-phrase, as argued by Jiang (1998), then why should the wh-phrase be to the right of the “dou” in surface syntax? Actually, the rigid position of the wh-phrase with respect to “dou” suggests that they are at least related somehow. Third, the presuppositional approach is rather less constrained, since the presupposition can be retrieved in various ways. The content of the presupposition seems to be quite elusive too. If there is no systematic way of deriving the content of this presupposition, I don’t know how the semantics of such wh-questions can be derived compositionally. Therefore it seems to me that the presuppositional approach is only an effort to maintain the view that “dou” cannot be associated with something to its right. But it is rather unsuccessful. The problem lies in the fact that none of these scholars have taken into consideration the denotation of questions. Without a proper denotation of questions, there is indeed no way of associating the “dou” with a wh-variable. In this paper, I will argue first that this “dou” is actually an exhaustivity marker, as proposed by Beck and Rullmann (1999). It takes a question denotation and returns the weakly exhaustive answer. The structure of the paper is as follows. In section 2, I introduce the semantics of the exhaustivity marker, and see how the Chinese data can be accounted for. I also point out some differences between the exhaustivity marker in Dutch and Mandarin, which can be explained by LF-movement, as argued in section 3. Then in section 3, I argue that even in these cases of seemingly rightward association of “dou” with some element in surface syntax, they can also be regarded as obeying the Leftness Condition, at LF. Section 4 discusses some more examples of “dou” in adjunct whquestions, and point out that the presupposition of “dou” conflicts with the presupposition of adjunct wh-questions. Section 5 deals with cases where a universal quantifier and a wh-phrase occur in the same sentence. I formulize an adjacency 3 condition to account for the restrictions of the association of “dou”. Then section 6 summarizes the paper. 2. “Dou” as an Exhaustivity Marker. Before talking about the semantics of the exhaustivity marker, we need to know what the semantics of questions is, and the notions of complete answers and weak exhaustivity. 2.1. Semantics of questions The denotation of questions has been a popular topic in semantics and philosophy of language. There are two major theories: proposition set semantics of Hamblin (1973) and Karttunen (1977), and partition semantics of Groenendijk and Stokhof (1984). Since we are only going to use the proposition set semantics of Hamblin and Karttunen, I will not introduce the partion semantics here. Hamblin (1958, 1973) argued that the semantic value of a question is a set of its corresponding possible answers. For example: (4) a. Who went to the party? b. { that Adam went to the party; that Bill went to the party; that Chris went to the party; that Dan went to the party} c. λp∃x [person’(w)(x) ∧ p = λw’party’(w’)(x)] The question in (4a) denotes a set of propositions which are possible answers to that same question. Suppose there are four individuals in the domain of discourse, and they are Adam, Bill, Chris, and Dan. The possible answers would be those in (4b). The set of propositions can also be written as (4c). I use “party” as a shorthand 4 notation for the predicate “went to the party” in (4a). The free world variable w in person’(w)(x) is the contextually salient world, which is usually the actual world. Karttunen(1977) modifies Hamblin’s proposal, and argued that denotations of questions contain only true answers. For example in (4a) if in the actual world only Adam and Dan went to the party, then the denotation of (4a) only includes the two propositions “that Adam went to the party” and “that Dan went to the party”. Correspondingly, the formula in (4c) can be revised as (5): (5) λp∃x [p(w) ∧ person’(w)(x) ∧ p = λw’party’(w’)(x)] The new conjunct p(w) filters out the false answers. It guarantees that only those propositions that are true in the actual world w are in the set. 2.2 Complete Answers and Weak Exhaustivity If we ask the question in (4), the interlocutor can give various answers, such as: (6) a. Adam went to the party. b. Adam and Bill went to the party. If in the real world, Adam and Bill actually went to the party, then (6a) is a partial answer, and (6b) is a complete answer. In most cases, a complete answer is desirable. If the question is embedded in some epistemic verbs like “know”, then there is the issue of strong and weak exhaustivity. For example: (7) John knows who went to the party. Adam went to the party. Therefore John knows that Adam went to the party. (8) John knows who went to the party. Elaine didn’t go to the party. Therefore John knows that Elaine didn’t go to the party. 5 In (7) John knows of each one who went to the party that they did, but he is ignorant about those who didn’t go to the party. Thus he knows the complete answer with knowing it is complete, or in another term, exhaustive. This is called weak exhaustivity. On the other hand, in (8) John not only knows who went to the party but also knows who didn’t. He knows the complete answer and he knows that it is complete, or rather, exhaustive. This is called strong exhaustivity. 2.3 Exhaustivity marker McCloskey (2000) observes that most varieties of English allow questions as shown in (9). (9) McCloskey (2000:58) a. What all did you get for Christmas? b. Who all did you meet when you were in Derry? c. Where all did they go for their holidays? The word “all” indicates that an exhaustive answer is needed. Take (9a) as an example. If I got only three things for Christmas: an iPod, a Mozart CD, and a book by Dan Brown, then the appropriate answer would be that I got an iPod, a Mozart CD and a book by Dan Brown. By contrast a question without this “all” does not require the most exhaustive answer, and the hearer can give as much information as he is willing to, and there is no obligation of any sort. Such a phenomenon is also found in many other languages, e.g. German, Dutch. (10) German invariant alles, Reis (1992: 465) Wen alles hat Hans besucht? Whom all has Hans visited Who all has Hans visited? (11) Dutch allemaal, Beck & Rullmann (1999: 287) 6 Hij weet wie He knows who er allemaal op het feest waren. there all at the party were He knows who all were at the party. In their work, Beck and Rullmann (1999) call this use of “all” the exhautivity marker. In a question-answer dialog, the interlocutor is not obliged to give an exhaustive answer all the time, and an exhaustive answer is not actually needed in some cases. But if an exhaustive answer is required, then the use of “all” indicate this requirement. In Mandarin Chinese, the use of “dou” in a wh-question has the same function. For example, by asking a question like (1), we want to know the exhaustive answer of what Zhangsan bought. Then what is the semantics of such an exhaustivity marker? Beck and Rullmann (1999) give the following sematnics: (12) a. answer1(w)(Q) = ∩{ p: Q(w)(p) ∧ p(w)} b. all(w)(Q) = λp. [p=answer1(w)(Q)] In (12a), answer1 equals to the conjunction of all the true answers in the Karttunen set. If we take propositions as sets of possible worlds, then the conjunction actually corresponds to the intersection of all these sets of worlds, as shown in (12a). Answer1 is the weakly exhaustive answer. Then the exhaustivity marker takes a question as its argument, and returns this weakly exhaustive answer. Thus its function is to build a weakly exhaustive answer out of the Karttunen set of propositions. Correspondingly in Mandarin Chinese, “dou” does not operate on the denotation of the wh-pharse directly, but rather influences the denotation of the question. Take the question in (1) as an example. If there are four things in the universe of discourse, i.e. a, b, c, and d, and Zhangsan bought two things, i.e. a and b, then the denotation of (1) without the “dou” would be: (13) {Zhangsan mai-le a; Zhangsan mai-le b} 7 Upon hearing such a question, we can answer either that Zhangsan bought a or Zhangsan bought b. If “dou” is used, then the denotation of (1) is: (14) {Zhangsan maile a he b} Upon hearing such a question, we should give the exhaustive answer “Zhangsan bought a and b”. Therefore, it seems that “dou” sums up all the true propositions in the Karttunen set. This function of the “dou” reminds us of the sum operator of events proposed by Huang (1996). Thus what “dou” requires is only the plurality of semantic objects in its domain, but not that all these semantic objects have to be individuals. There are important differences, however, between the kind of exhaustivity marker considered by Beck and Rullmann and the Chiense “dou”. Note that in (11), the Dutch “allemaal” is in an embedded question. It seems to me that the Chinese “dou” cannot be embedded this way. For example: (15) ??Zhangsan zhidao Lisi dou mai-le shenme. Zhangsan know Lisi all buy-ASP what Zhangsan knows what all Lisi bought. Although such sentences are judged grammatical by Jiang (1998), I don’t agree. It seems to me that such sentences are very odd, hence the two question marks. If we use a question-embedding verb like “to ask” or “to wonder”, then “dou” is ok. For example: (16) Zhangsan wen Lisi dou mai-le shenme. Zhangsan ask Lisi all buy-ASP what. Zhangsan asks what all Lisi bought. (17) Zhangsan xiangzhidao Lisi dou mai-le shenme Zhangsan wonder Lisi dou mai-le shenme Zhangsan wonders what all Lisi bought. 8 There is a sharp contrast between (15) and the two sentences in (16) and (17). In Dutch, the exhaustivity marker can be used in an embedded question with verbs that do not take questions as arguments. But in Mandarin Chinese, the “dou” can be embedded in a question with question-embedding verbs like “wen” and “xiangzhidao”, but not with verbs like “zhidao”. I will solve this problem in section 3, resorting to LF movement. 3. Leftness Condition and LF movement Generally speaking the “dou” in Mandarin Chinese is associated with some phrase to its left, and this is called the Leftness Condition. The effort to account for the “dou” in wh-questions in terms of presupposition e.g. Jiang (1998) or paraphrases, e.g. Xu (1985) is motivated by this Leftness Condition. Therefore if we could also move the wh-phrase to the left of “dou” somehow, then we can still maintain the leftness condition here. Actually this is an easy argument to make. Chinese is a wh-in-situ language, and there are currently two theories of wh-in-situ. The first one was originally proposed by Huang (1982). He argues for a LF movement analysis of whin-situ. The second one is a revival of an old idea of unselective binding. The theory says that the Q morpheme unselectively binds all wh-variable in its domain, without resorting to LF movement, or chain formation, as convincingly argued by Tsai (1999). But if we assume the LF movement analysis of wh-in-situ, then we get a very desirable result of “dou” association. For example, the LF of the question in (1) would be: (18) [shenme [Zhangsan dou mai-le t ] ] At LF, the wh-phrase moves to the spec,CP position as usual to check the [wh] feature. In the configuration of (18), the wh-phrase “shenme” c-commands “dou”. I will formulize the Leftness Condition as: (19) the Leftness Condition “Dou” must be associated with one element that c-commands it. 9 This condition can be paraphrased as follows. First, “dou” must be associated with something. It cannot be vacuous. Therefore the presuppositional approach seems a little less good, since it is not visibly in the sentence. Second, “dou” must be associated with something that c-commands it. This is a structural realization of the Leftness Condition. Third, “dou” must be associated with only one element. It cannot be associated with more than one element at the same time. I will talk about this in more details in section 5. Therefore even in cases where the “dou” is associated with a wh-phrase to its right, the Leftness Condition is still obeyed at LF after LF movement. Another issue is how to distinguish (18) from the universal construction, such as (20): (20) Shei dou xiahuan Zhangsan Who all like Zhangsan Everyone likes Zhangsan In fact the LF of (20) should be different from (18), since (20) is not a question, and hence no LF movement. Therefore the LF of (20) should be: (21) [Op [ shei dou xihuan Zhangsan ] ] Suppose there is some kind of an operator in the C of (21), and “shei” is bound by this operator in-situ without resorting to movement. Then there are two ways of distinguishing sentences like (20) from (1). First, they have totally different LF structures. Second, they are interpreted via different operators. (1) is interpreted via the question morpheme Q, while the universal construction is interpreted via another operator, whatever that is. Now let’s look at examples (15)-(17). First lets characterize the verbs in the following ways: (22) Know C-wh Wonder C+wh 10 The non-question-embeddeding verbs take a [-wh] C, where the question-embedding verbs take a [+wh] C. If LF movement is motivated by feature checking, then only in the “wonder” type sentences will the embedded wh-phrase move to check the [+wh] features, while in the “know” type sentences there is no movement for lack of feature checking. Therefore the LFs of (15) and (17) would be (23a) and (23b) respectively: (23) a.[Zhangsan zhidao [ C-wh [ Lisi dou mai-le shenme ]]] b.[Zhangsan xiangzhidao [ shenme C+wh [ Lisi dou mai-le t ]]] In (23a), the wh-phrase does not c-command the adverb “dou”, thus violating the Leftness Condition. Therefore it is ruled out. In (23b), the wh-phrase “shenme” has moved to the spec,CP position, hence c-commanding “dou” in accordance with the Leftness Condition. Therefore it is a perfectly grammatical sentence. Our problem raised in section 2 is thus solved here. Now let’s look at the Dutch example in (11) again. There is wh-movement in Dutch, and indeed the type of sentences corresponding to the Chinese (15) is perfectly grammatical in Dutch. Therefore it seems that the c-command relation is important for the kind of semantic interpretation of the exhaustive marker. 4. Exhaustivity Marker in Zenme questions So far we have only considered the exhaustivity marker “dou” in questions with wharguments, i.e. shei, shenme. If we look at other types of wh-questions, then we get a different picture. For example, consider zenme-questions. (24) Zhangsan dou zenme tiao-de wu? Zhangsan all how dance-DE dance *How all did Zhangsan dance. Such sentences are ungrammatical in English and the other languages that have this exhaustivity marker, e.g. German and Dutch. What about Mandarin Chinese? On one of the readings, (24) is perfectly ok. Suppose we are talking about three parties that Zhangsan attended this week. He danced on these three occasions. Then we want to 11 know how he danced on these occasions. He could have danced in the same way on these occasions or he could have danced differently on each occasion. But the sentence is ambiguous in this aspect. Therefore it seems that we are talking about different times, and “dou” is associated with these times, such as t1, t2, and t3. In this reading “dou” is not associated with the “zenme”. This is not the reading we are concerned with here. Suppose we are talking about yesterday’s party where Zhangsan danced. I want to know about all the ways that Zhangsan danced. Such a reading corresponds to the English translation, which is not grammatical in English. It seems to me that such a reading is quite hard to get in Mandarin as well. Of course it is possible to imagine a scenario where this sentence is ok, but uttered alone without context, the sentence in (24) seems quite odd. The “dou” doesn’t really add anything new to the sentence. Without the “dou”, the sentence seems to mean the same thing. Now we need to solve this problem. I will argue that zenme-questions have a different denotation than the questions with wh-arguments. In this section, I will give a semantics of how-questions in general and show how it interacts with the exhaustivity marker “dou”. But before giving the semantics of such questions, I need to introduce the event semantics. 4.1 The event semantics Davidson (1967) argued for an ontology which includes events, in addition to the usual entities of concrete material objects and etc. It has since been shown that a number of linguistic phenomena such as nominalization, adverbial modification, tense and aspect, causal statements, and temporal reasoning, can be successfully accounted for within an event-based semantics. Take the sentence (25) as an example: (25) Brutus stabbed Caesar violently. The sentence is about an event of stabbing. The agent of the event is Brutus, and the patient of the event is Caesar. The event is a violent one. Thus the logical form of (25) can be represented as (26) in the neo-Davidsonian event semantics of Parsons (1990): (26) ∃e [ Stabbing(e) ∧Agent(e, Brutus) ∧Patient(e, Caesar) ∧Violent(e)] 12 There are many other respects about event semantics, but this little introduction is enough for our current purposes. 4.2 Manner “how” questions The question (27a) is adapted from Davidson’s (1967) example. (27) a. How did Jones butter the toast? b. He buttered the toast slowly. c. He buttered the toast slowly with a knife in the kitchen. d. *He buttered the toast slowly, and he buttered the toast with a knife, and he buttered the toast in the kitchen. (27b) is a perfect answer to (27a); (27c) is also an acceptable answer to (27a). But (27d) is not a felicitous answer to (27a), for it is not clear whether the three conjuncts are about the same event of “toast buttering” or not. They could be three separate events. Then what is the denotation of (27a)? In the Hamblin-style proposition set approach of questions, the denotation of a question is the set of propositions as possible answers to that same question. Since (27b) is a good answer, other “model” answers should be similar and we can posit a set of possible answers as shown in (28): (28) {that Jones buttered the toast slowly, that Jones buttered the toast quickly, that Jones buttered the toast absentmindedly, that Jones buttered the toast happily, …… } 13 The wh-phrase “how” ranges over adverbial modifications like “slowly” and “happily”, which are properties of events in the event semantics. (28) can be represented as (29): (29) {p|∃P [p = ^∃e. [Buttering(e) ∧Agent(e, Jones) ∧Patient(e, toast) ∧P(e)]]} In (29), the variable P ranges over properties of events, and the carat sign turns the formula into a proposition. Therefore (29) seems to be a good candidate for the denotation of manner questions. If we replace the variable P with all kinds of adverbial modifiers, we get the same set of possible answers as (28). Then what is the true answer set in this case? Suppose that in the actual world Jones buttered the toast slowly and absentmindedly. Should we represent the true answer set as (30)? (30) a. {that Jones buttered the toast slowly, that Jones buttered the toast absentmindedly.} b. {^∃e. [Buttering(e) ∧Agent(e, Jones) ∧Patient(e, toast) ∧Slow(e)], ^ ∃ e. [Buttering(e) ∧ Agent(e, Jones) ∧ Patient(e, toast) ∧ Absentminded(e)]} But (30) is probably not the right denotation. As has been shown in (27d), these separate answers seem to be about different events. This is further shown in (30b). There is an existential quantifier before each event variable. There is no guarantee that the two events are the same one. For lack of further specification, e.g. time and location of the event, these events can be treated as separate different events. If (30) is wrong, how can we represent such an answer? If Jones buttered the toast slowly and absentmindedly, then there is only one such event and there are multiple adverbial modifiers, which can be simply treated as equal conjuncts, for example: (31) a. {that Jones buttered the toast slowly and absentmindedly} 14 b. {^∃e. [Buttering(e) ∧Agent(e, Jones) ∧Patient(e, toast) ∧Slow(e) ∧ Absentminded(e)] The difference between (30) and (31) is that (31) is a singleton set with just one proposition about one event of toast-buttering which can be described in different ways, such as being slow and absentminded. (31) is the only correct representation of the denotation of manner questions. I further argue that “how” questions are about just one event and they ask about the manner predicate in which the event was carried out. But in questions with wharguments, there can be multiple events. Let’s call this fact the Event-Singularity Presupposition (ESP for short) as stated in (32). (32) Event-Singularity Presupposition (ESP) In wh-adjunct questions, e.g “how”, there is a single presupposed event. (32) might seem very bland and stipulative if we only look at English. In fact (32) corresponds to a nice distinction between two different constructions for the past tense in Mandarin Chinese. One of the constructions uses the perfective aspect marker –le to indicate a past event, the other uses “shi…de” (“it is… that…”) similar to the English cleft sentences. For argument questions only –le is possible, while the use of “shi…de” results in uninterpretable sentences. They are not just ungrammatical, but they have no interpretation at all. For example: (33) a. Ni You zuotian zuo shenme le? yesterday do what ASP What did you do yesterday? b. *Ni zuotian shi zuo de shenme? You yesterday be do DE1 what. 1 DE is a structural morpheme in Mandarin Chinese which can be used in many different constructions. For ease of explanation, it is usually glossed just as DE. 15 (34) a. Ni zuotian kanjian shei le? You yesterday see who ASP Who did you see yesterday? b. *Ni zuotian shi kanjian shei de? You yesterday be see who DE On the other hand, wh-adjunct questions with “how”, “when” and sometimes “where”2 can be used with “shi…de” only, and similarly, the use of –le results in uninterpretable sentences. For example: (35) a. Ni shi zenme qu You be how de Niuyue? go DE NYC How did you go to NYC? b. * Ni zenme qu-le You how Niuyue?3 go-ASP NYC The “shi…de” construction can be translated into English as “it is… that”. The Chinese construction is often said to emphasize the part between the two words. Yet another prerequisite of using this construction is that there should be a topic event, which is the part of the sentence minus the immediate constituent right after the “shi”. For example in (35a), the emphasis is on “zenme”, but the rest of the sentence “ni qu Niuyue” (“You went to NYC”) is the topic event. If the Chinese wh-adjunct questions are to be uttered out of the blue, they sound very odd, because they carry a presupposition that there is a single topic event. Therefore the hearer will have to accommodate such a presupposition, by assuming that there is a relevant event. Therefore the distinction between the uses of these two different constructions confirms our ESP in (32). This presupposition together with the semantics of wh2 “Where” can be either an argument or an adjunct, depending on the predicate used. Therefore some “where” questions are argument questions, some are adjunct questions. “Why” involves causation between multiple events, and hence does not fall into this category. “Why” questions can be used both with the aspect marker –le and with “shi…de”. 3 This sentence is uninterpretable if zenme means “how”. This word, however, can mean “why” in some cases, which make it compatible with –le. 16 adjunct questions will give us the right answer set. First let me revise the representation in (29) to better capture the fact in (31). The new formula is shown in (36): (36) {p | ∃P .[p=^∃e. [Buttering(e) ∧Agent(e, Jones) ∧Patient(e, toast) ∧[∀P∈P.P(e)]]} In (36) the curly P ranges over sets of predicates of events. Thus a possible illustration of (36) would be (37): (37) { that Jones buttered the toast slowly and beautifully, that Jones buttered the toast quickly, carelessly, and absentmindedly, that Jones buttered the toast slowly, happily, clumsily and amusedly.} Now if we want the true answer set derived from (36), this is when the ESP comes into play. According to the ESP, adjunct questions should be about a single event, and hence the different propositions in (37) cannot be adjoined, and we can only choose one of them according to the actual state of affairs of the actual world. Suppose in the actual world Jones buttered the toast slowly and beautifully. Then the true answer set will just be a singleton set containing the first proposition in (37). 4.3 “dou” in zenme-questions in Mandarin. Now let’s focus on the sentences like (24), in which the “dou” is not appropriate if considered to be associated with “zenme”. The reason is that the plurality presupposition of “dou” conflicts with the ESP of zenme-questions. To be specific, first by using “dou”, we are assuming that there are multiple semantic objects to 17 consider. Second, the denotation of zenme-questions contains only one proposition. For example, the denotation of (24) could be: (38) {Zhangsan xinggaocailie-de, zhangyawuzhao-de tiao-de wu4} Therefore by using a zenme quesiton, we already know that there is only one true proposition. If we already know that there is no plurality to consider, then we wouldn’t use “dou”. Suppose we know Zhangsan only bought one thing, and then we wouldn’t ask a question like (1) either. “Dou” can be used if we don’t know if there is only one thing to consider. 5. Adjacency condition on the association with “dou” “Dou” can be associated with many different things, e.g. universal quantifiers, plural NPs, and a wh-phrase to its left. What if there is more than one element that “dou” can be associated with in one sentence? Presumably, in such cases, “dou” can be associated with only one of these elements, and there are certain conditions as to which one the “dou” can associate with. In this section, I will consider some of these cases, but not all of them. I will just concentrate on the interaction between “dou” and the universal quantifier in wh-questions. For example: (39) Meigeren dou xihuan shei? Everyone all like who Who does everyone like? (widescope for “every” only) How many readings does (39) have? If “dou” is associated with “shei”, then the question is asking about all of the persons that each person likes. The interlocutor has to give an exhaustive list of the persons that a certain person likes. But this reading is definitely not available here. If “dou” is associated with “meigeren”, then the question is asking about the person or persons that each person likes. Indeed this is the only natural reading available here. Therefore it seems that in such cases “dou” can be associated only with “meigeren” but not with “shei”. Then what is the condition on 4 Zhangsan danced happily and in a funny, exaggerated way. 18 this kind of association? I will propose that it is due to the usual adjacency conditions, such as (40): (40) Adjacency condition on “dou” association “Dou” can be associated only with the closest c-commanding NP. This condition says that if there are more than one candidate for “dou” association, then only the closest c-commanding NP can be associated with “dou”. Let’s see how this works in the example (39). Here we have to resort to LF movement again. Arguably the LF of (39) could be: (41) [ shei [ meigeren dou xihuan t ] ] In (41), both the wh-phrase “shei” and the “meigeren” c-command “dou”. But “meigeren” is closer to “dou” than “shei” is. Therefore the only possible association is between “dou” and “meigeren”. Now let’s look at more examples: (42) a. Shei meigeren dou xihuan Who everyone all like Who is liked by everyone? b. [shei [ t [ meigeren dou xihuan t ] ] In (42a) “shei” has been topicalized in surface syntax, and in the LF, according to the adjacency condition, “dou” can be associated with only “meigeren”, and this is indeed the only reading of this sentence. (43) a. Dou shei xihuan zheli-de meiyigeren All who like here-DE everyone Who all likes everyone here? b. [ shei [dou t xihuan zheli-de meiyigeren ]] 19 In (43a), “dou” is only associated with “shei”. In LF, “shei” c-commands “dou”, but “meigeren” does not. (44) a.Zhangsan dou ba mei-ben-shu gei-le shei ? Zhangsan all BA every-book give-ASP who For each book, who all did Zhangsan give it to? b. [shei [ Zhangsan dou ba mei-ben-shu gei-le t ] ] This is a very complicated example. First let’s concentrate on the “dou” association. There is no doubt that “dou” is associated with “shei”. This is the only reading that I can get. At LF, “shei” c-commands “dou”, and according to the adjacency condition, only “shei” can be associated with “dou”. On the other hand, “meigeren” has wide scope over “shei”. If we look at (44a), then indeed the “meigeren” takes wide scope over “shei”. This seems to suggest that quantifier scope is interpreted in surface syntax in Mandarin Chinese, i.e. as shown in the configuration in (44a), while “dou” association is interpreted at LF. Actually this is a well-known phenomenon called the Isomorphism of quantifier scope in Mandarin Chinese. This refers to the fact that quantifier scope in Mandarin Chinese corresponds to their surface syntax configuration. Some scholars propose LF quantifier raising for English quantifiers, but in Mandarin, there is no such need, since the scope is resolved in syntax. Therefore the isomorphism of quantifier scope and the lack of quantifier raising combined with our adjacency condition can successfully explain (44). 6. Conclusions In this paper, I start out by arguing that the Mandarin Chinese “dou” in wh-questions can be considered an exhaustivity marker, in light of cross-linguistic data from English, German and Dutch. This exhaustiviy marker takes a question denotation and returns the weakly exhaustive answer. However, unlike the Dutch exhaustivity marker, the Chinese “dou” cannot be used in a embedded question with a matrix verb like “zhidao”. To explain this, I argue that the Leftness Condition can be extended to this use of “dou” by resorting the LF movement, and verbs like “zhidao” do not take question arguments, and in their embedded clauses the wh-phrases do not move at LF 20 due to lack of feature checking, thus violating the Leftness Condition. This explains why “dou” is not grammatical in such embedded questions. Then I continue to consider the incompatibility of “dou” and “zenme”, arguing that it is due to the presupposition conflict between the plurality of “dou” and the singularity of “zenme”. Then I discuss examples where there are multiple NPs for “dou” association and propose an adjacency condition which says “dou” can be associated only witht the closest c-commanding NP. This condition successfully explains a series of data. Reference Beck, Sigrid and Hotze Rullmann. 1999. A Flexible Approach to Exhaustivity in Questions. Natural Language Semantics 7:249-298. Davidson, Donald. 1967. The Logical Form of Action Sentences. In N. Rescher (ed.), The Logic of Decision and Action, 81-95. University of Pittsburgh Press. Engdahl, E. 1986. Constituent Questions: the syntax and semantics of questions with special reference to Swedish. Dordrecht/Boston/London: Kluwer Academic Publishers. Groenendijk, Jeroen (1984) Studies on the Semantics of Questions and The Pragmatics of Answers. PhD dissertation, University of Amsterdam. Groenendijk, Jeroen and Stokhof, Martin (1989) Type-shifting Rules and the Semantics of Interrogatives. In Gennaro Chierchia, Barbara H. Partee, and Raymond Turner (eds), Properties, Types and Meanings. Vol 2: Semantic Issues, pp.21-68. Kluwer. Reprinted in Portner, Paul and Partee, Barbara H. (eds) (2002). Groenendijk, Jeroen and Stokhof, Martin (1997) Questions. In J. van Benthem and A. ter Meulen (eds), Handbook of Logic and Language, pp 1055-1124. Amsterdam: Elsevier, and Cambridge, MA: the MIT Press. Groenendijk, Jeroen (1999) The Logic of Interrogation, Classical Version. In T. Matthews and D. L. Strolovitch (eds), SALT IX: Semantics and Linguistic Thoery, pp. 109-126, Ithaca: Cornell University Press. Hamblin, C. L. (1958) Questions. The Australasian Journal of Philosophy 36: 159-168. Hamblin, C. L. (1973) Questions in Montague English. Foundations of Language 10:41-53. Heim, Irene and Kratzer, Angelika (1998) Semantics in Generative Grammar. Malden. Higginbotham, James, and Fabio Pianesi, Achille C. Varzi. (eds) 2000. Speaking of Events. Oxford University Press. 21 Huang, Shi-zhe (1996). Quantification and Predication in Mandarin Chines; A Case Study of Dou, UPenn Dissertation, Philadelphia, PA. Karttunen, Lauri (1977) Syntax and Semantics of Questions. Linguistics and Philosophy 1:344. Reprinted in Portner, Paul and Partee, Barbara H. (eds) (2002). Lahiri, Utpal(2002) Questions and Answers in Embedded Contexts. Oxford UP. McCloskey, James. 2000. Quantifier Float and Wh-Movement in an Irish English. Linguistic Inquiry 31: 57-85 Reis, Marga. 1992. The category of invatiant alles in wh-clauses: On syntactic quantifiers vs. quantifying particles in German. In Who climbs the grammar tree?,ed. Rosemarie Trace. 465492. Tuebingen: Max Niemeyer Verlag. Rullmann, Hotze. 1995. Maximality in the Semantics of Wh-Constructions. UMass Dissertation. Tsai, Wei-Tien Dylan. 1999. On economizing the theory of A-bar dependencies. New York, Garland. 程美珍(1987) , 关于表示总括全部的“都”, 蒋 (1998), 语用推理与 都 的 语言教学与研究 法/语义特征 现代外语 兰宾汉(1988) , 副词“都”的语义及其对后面动词的限制作用, , 第2 期, 27 - 36 页 ，第一期，10-24页 语言教学与研究 , 第2 期, 46 - 51 页 马 真(1983) , 关于“都/ 全”所总括的对象的位置, 马真(1985) , 现代汉语虚词散论 个问题, 汉语学习 语言学论丛 王 还(1983) ,“All”与“都”, 王 还(1988a) , 再谈谈“都”, 王 还(1988b) , 三谈“都”, 徐 杰(1985) ,“都”类副词的总括对象及其隐现 杰(1993) , 语言教学与研究 语言教学与研究 世界汉语教学 汉语描写语法十论 , 第1 期 载陆俭明 , 98 - 105 页, 北京大学出版社, 北京 史锡尧(1990) , 副词“都”语义语用综合考察, 苏培成(1984) , 有关副词“都”的 汉语学习 , 第4 期, 6 - 10 页 , 第十三辑, 57 - 61 页 , 第4 期, 24 - 28 页 , 第2 期, 52 - 53 页 , 第2 期 位序, 汉语学习 , 第一期 ,77 - 87 页, 河南教育出版社 徐颂列(1993) , 表总括的“都”的语义分析, 语言教学与研究 袁毓林(2005), 都 的语义功能和关联方向新解 22 中国语文 , 第四期, 75 - 86 页 第二期，99-109页 载徐