A-not-A as a Reduplicative Inflectional Morpheme

1. Introduction 2. A-not-A as a reduplicative inflectional morpheme 2.1. The NQ morpheme 2.2. The role of negation 3. Movement of NQ and parallels to weishenme ‘why’ 3.1. A-not-A questions can be embedded 3.2. A-not-A questions cannot appear in islands 3.3. A-not-A questions are like weishenme ‘why’ questions 3.4. A-not-A qu can be embedded: A further note 4. Focus marking and A-not-A questions 4.1. Restrictions on focus marking 4.2. Focus markers in A-not-A questions 4.3. Parallels to weishenme ‘why’ questions 4.4. Discourse conditions on A-not-A questions 5. Further surface scope restrictions on NQ 5.1. Adverbs 5.2. Indefinite wh-words 5.3. Intervening quantifiers 6. Negative particle questions and VP-not-V questions 6.1. Distributional similarities and differences 6.2. The location of Neg(-V) 6.3. Distributional similarities and differences revisited 7. VP-not-VP questions as VP-VP questions 8. Crosslinguistic notes 9. Closing remarks 1. Introduction An A-not-A question is a type of question with a function similar to that of a yesno question, the surface form of which involves two copies of a predicate with one copy negated. A-not-A questions have been investigated most thoroughly in Mandarin Chinese, with (1) and (2) below being characteristic examples. One noteworthy aspect of (1) is that it has reduplicated only the first syllable of the verb; xi here is not a word on its own, it is simply the first syllable of xihuan. (1) ta xi-bu-xihuan zheben shu? he li-not-like this book ‘Does he like or not like this book?’ (A-not-A) (2) ta xihuan-bu-xihuan zheben shu? (A-not-A) he like-not-like this book ‘Does he like or not like this book?’ The A-not-A questions above represent one of several ways that one can inquire about the truth of a proposition in Mandarin, and much of the recent literature on the topic has been concerned with distinguishing such questions from one another. A-not-A questions like (1) and (2) are not really yes-no questions of the familiar sort; yes-no questions are usually formed using the sentence-final question particle ma, as in (3). Apart from the distributional differences to be discussed at length below, perhaps the clearest difference between A-not-A questions and yes-no questions is that only yes-no questions can be naturally answered with shi ‘yes’, dui ‘correct’, or bu ‘no’; the questions in (1) and (2) are answered by repeating either the questioned predicate xihuan ‘like’ or its negation bu xihuan ‘not like’. (3) ta xihuan zheben shu ma? he like this book Q? ‘Does he like this book?’ (YNQ) With respect to how they are answered, A-not-A questions are similar to alternative questions, such as (4) below. Although it is tempting to assume that (4) is simply a more explicit version of (2), the distribution of questions containing the word haishi ‘or’ differs from that of questions lacking haishi. (4) ta xihuan zheben shu haishi ta bu xihuan zheben shu? he like this book or he not like this book ‘Does he like this book or doesn’t he like this book?’ (AltQ) In addition to the three classes of questions described so far, there are several other ways in which this question can be asked, each of which has been claimed at some time or another in the literature to have a distinct structure. In the first set of examples, (5–7), the negative morpheme comes after the VP; what differentiates the forms is how much of the second copy is repeated. In the VP-not-VP type in (5) the whole VP is repeated, in the VP-not-V type in (6) only the verb is repeated, and in the negative particle question (NPQ) in (7), the negation appears alone at the end of the question. (5) ta xihuan zheben shu bu xihuan zheben shu? (VP-not-VP) he like this book not like this book ‘Does he like this book or doesn’t [he] like this book?’ (6) ta xihuan zheben shu bu xihuan? he like this book not like ‘Does he like this book or doesn’t [he] like [it]?’ (VP-not-V) (7) ta xihuan zheben shu bu? he like this book not ‘Does he like this book or not?’ (NPQ) The second set of examples illustrates A-not-A-type questions formed with an auxiliary, a modal, or the focus morpheme shi. In (8) is an M-not-M question in which a modal is reduplicated, and in (9) is a B-not-B question in which the focus morpheme/copula shi is reduplicated. (8) ta hui-bu-hui tiaowu? He can-not-can dance ‘Can he dance?’ (9) ta shi-bu-shi xihuan zheben shu? he be-not-be like this book ‘Is it the case or not that he like this book?’ (M-not-M) (B-not-B) Given that Mandarin allows omission fairly freely, the null hypothesis should be, and traditionally was, that (1–2) and (5–7) are simply versions of (4) that have undergone varying degrees of ellipsis, and that (8) and (9) are simply extensions of the A-not-A pattern of (1–2). By far the most influential work on A-not-A questions was that of Huang (1982a, 1991), who challenged the traditional assumption by showing, among other things, that A-not-A questions, yes-no questions, and alternative questions have different distributions. Probably the most well known observation that Huang made about A-not-A questions like (1–2) is that they, unlike alternative questions like (4), resist appearing in movement islands. On the basis of this, Huang hypothesized that the A-not-A surface sequence reflects an abstract question morpheme that must undergo movement to a scope position in CP, and drew a parallel between this A-not-A morpheme and adjunct wh-words like weishenme ‘why’. In the remainder of this article, we will review the distributional characteristics of A-not-A questions and the rest of the family of questions introduced in this section. The main goal here is not so much to exhaustively present and compare the different theoretical analyses as it is to try to present the types of data that have been crucial in formulating the analyses. The first few sections will concentrate on the A-not-A question “proper”, like (1–2), after which we will expand to consider similarities and differences between A-not-A questions and the other forms, such as VP-not-V questions, NPQs, and VP-notVP questions. A relatively brief review of some related cross-linguistic data will close out the discussion. 2. A-not-A as a reduplicative inflectional morpheme 2.1 The NQ morpheme Huang (1991) proposed that the A-not-A sequence in A-not-A questions is the morphological exponent of a reduplicative interrogative morpheme located in the inflectional system, an “interrogative INFL”. (10) IP qp DP I′ ni qp you I VP NQ qp [+Q] V DP xihuan zheben shu like this book In light of subsequent developments in syntactic theory, particularly the widely assumed need to separate IP into several separate projections (following Pollock 1989), the structure in (10) requires elaboration—but it illustrates the basic idea: the A-not-A sequence is essentially a word formed morphologically by an abstract question morpheme and a following predicate. For expediency, we will refer to this morpheme as NQ (intended to reflect both its negative and interrogative character). A simple argument in favor of a structure like (10) can be made from the fact that NQ must appear on the modal or auxiliary when one is present in the sentence. Thus, while (11) is an acceptable question, (12), with NQ on the main verb, is impossible (Li & Thompson 1979). (11) ta hui-bu-hui tiaowu? He can-not-can dance ‘Can he dance?’ (12) *ta hui tiaowu-bu-tiaowu? He can dance-not-dance (‘Can he dance?’) It is also worth noting that NQ is not limited to attaching to verbs; An Anot-A question can also be formed, for example, with an adjective (13), a preposition (14), an adverb (15), or a modal/auxiliary as above (11). (13) zheben shu hao-bu-hao? this book good-not-good ‘Is this book good?’ (14) ta cong-mei-cong Beijing lai? he from-not-from Beijing come ‘Did he come from Beijing?’ (15) ta chang-bu-chang lai? he often-not-often come ‘Does he come often?’ If NQ is indeed something like a VP-proclitic, these facts follow; it simply reduplicates the word (or syllable) that it attaches to, with a negative morpheme between the copies. Huang (1991) also provides a crosslinguistic argument for this analysis, pointing out that Taiwanese has a question particle kam that appears in the same contexts in which NQ is hypothesized to be in Mandarin. Huang cites an extensive cross-dialectal survey by Zhu (1985) that concludes that kam-type questions and A-not-A type questions are essentially in complementary distribution across dialects: a language either has kam-type questions or A-not-A questions but not both. Huang interprets this as evidence that NQ and kam are different morphological exponents of the same underlying morpheme. Other kam-type questions are found in Shanghai and Early Mandarin; Huang’s examples are given in (16–18). (16) li kam beh lai? (Taiwanese) you Q want come ‘Do you want to come?’ (17) nong a le va? you Q come prt ‘Will you come?’ (18) ni ke you kong? (Early Mandarin) you Q have time ‘Do you have time?’ (Shanghai) 2.2 The role of negation The fact that the A-not-A question in (1) involves a syllable that does not carry any meaning of its own (it is simply the first syllable of the questioned predicate) suggests that the surface form arises from a morphological process, and not from ellipsis or any purely syntactic process, since neither syntax nor ellipsis can generally target just part of a word. As discussed above, Huang (1991) proposed that NQ is realized as a copy of (some part) of the following predicate, and an inserted negative morpheme bu. It is far from clear what role negation plays in A-not-A questions, because it is a definitional characteristic of an A-not-A question; a question without negation is not an A-not-A question and has no NQ, and all questions with NQ show negation. There is clearly an ‘or not’ meaning associated with these questions, but it is impossible to say whether this is inherent in the interpretation of NQ or whether this is contributed by the negative morpheme. Note, for example, that the kam and ke type morphemes in other dialects (16–18) do not appear with a negative morpheme, although they are interpreted in the same way as A-not-A questions in Mandarin. Still, there are some respects in which the negative morpheme in A-not-A question behaves like meaningful negative morphemes elsewhere in Mandarin. McCawley (1994) provides several arguments to show that bu is not simply an unanalyzed expletive element, but is rather an actual negative morpheme. His arguments are reviewed below. In Mandarin, the form in which the negative morpheme appears depends on the aspectual structure of the sentence. A fair approximation is that bu is used in clauses denoting states (19) and mei(you) is used in clauses denoting bounded and episodic events (20); see Ernst (1995), Lin (2003) for a more precise characterization. The choice of bu or mei in an A-not-A question is governed by the same conditions; an A-not-A question about a state interposes bu between the copies of the predicate (21), and an A-not-A question about a bounded event appears with mei (22). This pattern by itself rules out the possibility that bu is an unanalyzed connective morpheme; as far as the morphology is concerned, NQ is a negative morpheme. (19) wo bu chi mugua. I not eat papaya ‘I do not eat papaya.’ (20) wo mei-you chi mugua. I not-have eat papaya ‘I did not eat papaya.’ (21) ta xi(huan)-bu-xihuan neiben shu? he like-not-like that book ‘Does he like that book?’ (22) ta kan(jian)-mei-kanjian neiben shu? he see-not-see that book ‘Did he see that book?’ McCawley further observes that lexical restrictions on the appearance of negation are observed in A-not-A questions. Where a predicate cannot be negated, as is the case for some lexically negative predicates like weijue ‘unsolved’, neither can the predicate be questioned using NQ. Not only are absolute co-occurrence restrictions observed, but intermediate judgments are also parallel; in each of the examples (23–26), the acceptability of the (a) sentences matches the acceptability of the (b) sentences. While judgments seem to vary on the sentences in (23–26), McCawley’s point is secure so long as the (a) examples have the same status as the (b) examples. (23) (24) (25) (26) a. *neige wenti bu/mei weijue. that problem not unsolved (‘That problem isn’t unsolved.’) b. *neige wenti weijue be/mei weijue? that problem unsolved not unsolved (‘Is that problem unsolved?’) a. ??zheiyang zuo bu wuli. thus do not without-reason ‘It isn’t unreasonable to do this.’ b. ??zheiyang zuo wu bu wuli? thus do without not without-reason ‘Is it unreasonable to do this?’ a. ?zheiyang zuo bu feifa. thus do not illegal ‘It isn’t illegal to do this.’ b. ?zheiyang zuo fei-bu-feifa? thus do il-not-illegal ‘Is it illegal to do this?’ a. ta bu wuchi. he not without.shame ‘He isn’t shameless.’ b. ni shuo ta wu(chi)-bu-wuchi? you say he without.shame-not-without.shame ‘Do you say he’s shameless?’ McCawley’s argument shows that, as far as the morphology is concerned, the negation in an A-not-A question is a real negation, but does not decisively show that the negative morpheme is part of the syntax and interpretation. The negation in NQ could as easily be similar to the supportive verb do in English, which inflects like a real verb, but is arguably not part of the syntax (Halle & Marantz 1993, Chomsky 1957). It could be that NQ carries a negative feature, one that is realized in the Mandarin morphology, but has no exponent in the hypothetically parallel kam-type questions in Taiwanese. 3. Movement of NQ and parallels to weishenme ‘why’ Unlike yes-no questions formed with ma, A-not-A questions can be embedded, and can take scope beyond their own clause. However, a long-distance interpretation is blocked if the NQ and its would-be scope position are separated by an island boundary. On the basis of these facts, Huang (1982a, 1991) argues that A-not-A questions involve covert movement of NQ to its scope position for interpretation. Below, we will first review the evidence that A-not-A questions can be embedded when no island interferes, and then the evidence that islands can interfere. The distributional characteristics of NQ are parallel to those of nonnominal adjunct question words like weishenme ‘why’, for which a movementbased analysis is relatively uncontroversial. 3.1 A-not-A questions can be embedded The statement in (27) has an embedded A-not-A question and it is grammatical, indicating that A-not-A questions are not exclusively a matrix clause phenomenon. (27) [ta qu bu qu Meiguo] bu qingchu he go not go America not clear ‘Whether he will go to America is not clear.’ The question in (28), due to Huang (1982a), can also make this point, in a slightly more sophisticated way. The main verb in (28) is the question-embedding verb xiang-zhidao ‘wonder’, and the interpretation of this question has the wh-word in the embedded clause taking wide scope. While the English translation of this question would be somewhat marginal (due to extraction of the wh-word from the embedded interrogative), Huang rates the Mandarin question as fine, and the embedded clause is clearly an A-not-A question. Care is required in judging this sentence; the intended interpretation is one in which a person’s identity is sought, and some speakers find it helpful to emphasize the embedded wh-word. (28) ni xiang-zhidao [shei xi-bu-xihuan ni] (ne)? You want-know who li-not-like you ‘Who is the person x such that you wonder whether x likes you or not?’ The ability to embed A-not-A questions is fairly restricted—enough so to lead Li & Thompson (1979) to view the situation in the opposite way: they claim that A-not-A questions are only possible in matrix clauses except under special circumstances. According to Li & Thompson, a certain class of “semantically bleached” verbs (including xiang/kan ‘think’, shuo ‘say’, and juede ‘feel’) is semantically transparent enough to allow the embedded verb to “count as” a matrix verb, and thereby permit the attachment of NQ morphology. McCawley (1994) makes essentially the same observation, putting forth the idea that those verbs that allow “embedded A-not-A questions” are susceptible to a parenthetical interpretation, and in that case are not truly embedding the A-not-A question. However, the examples above are sufficient to show that it is at least in principle possible to embed an A-not-A question. 3.2 A-not-A questions cannot appear in islands The dominant view in the literature on A-not-A questions, arising from the influential arguments presented by Huang (1982a, 1991), is that NQ is something akin to a wh-word, related by movement to a CP-level scope position. In addition to giving us a window into the syntactic structure of A-not-A questions, these facts also provide evidence that alternative questions, with haishi ‘or’, are different from A-not-A questions. To see the sensitivity of A-not-A questions to islands, we can compare them to alternative (haishi) questions, which are very similar in meaning but are syntactically relatively unconstrained. In particular, alternative questions can occur inside islands without difficulty. In (29) we have haishi-questions inside sentential subjects, and in (30) we have haishi-questions inside relative clauses. The questions in (29–30) differ in how much material is elided, but all have the disjunctive haishi. (29) (30) a. [wo qu Meiguo haishi bu qu Meiguo] bijiao hao? (AltQ) I go America or not go America more good ‘Is it better that I go to America or that I do not go to America?’ b. [wo qu Meiguo haishi bu qu] bijiao hao? I go America or not go more good ‘Is it better that I go or not go to America?’ c. [wo qu haishi bu qu Meiguo] bijiao hao? I go or not go America more good ‘Is it better that I go or not go to America?’ a. ni xihuan [renshi ni haishi bu renshi ni] de ren? (AltQ) you like know you or not know you DE person ‘Do you like people who know you or people who don’t know you?’ b. ni xihuan [renshi haishi bu renshi ni] de ren? you like know or not know you DE person ‘Do you like people who know or don’t know you?’ c. ni xihuan [renshi ni haishi bu renshi] de ren? you like know you or not know DE person ‘Do you like people who know you or don’t?’ The questions in (31–33) are A-not-A questions, the counterparts to (29–30) but without the word haishi, and all are ungrammatical as a result of being embedded in an island. The questions in (31–33) are intended as matrix questions, with NQ inside a sentential subject, a relative clause, and a because clause, respectively. (31) *[wo qu bu qu Meiguo] bijiao hao? I go not go America more good (‘Is it better that I go to America or not?’) (32) *ni xihuan [ren(shi) bu renshi ni de ren]? you like kn(ow) not know you DE person (‘Do you like people who know you or don’t know you?’) (33) *ni [yinwei Lisi lai-bu-lai] shengqi ne? you because Lisi come-not come angry Q (‘Are you angry or not because Lisi came?’) Example (27) in the previous section showed a grammatical A-not-A question inside a sentential subject, but there the scope of the question is itself within the island, hence the hypothesized movement need not cross the island boundary. In addition to the islands noted by Huang (1982a), verbs that take nominal complements, including yihan ‘regret’, jide ‘remember’, and tongyi ‘agree’, also prevent A-not-A questions from appearing in their complements (Tsai (1994)). Note that to judge (35) requires some care, since jide ‘remember’ can also be read parenthetically. (34) *ni hen yihan [Lisi xi-bu-xihuan Meiguo]? you very regret Lisi li-not-like America (‘Do you regret that Lisi likes or doesn’t like America?’) (35) *ni jide [Lisi xi-bu-xihuan Meiguo]? you remember Lisi li-not-like America (‘Do you remember that Lisi likes or doesn’t like America?’) 3.3. A-not-A questions are like weishenme ‘why’ questions The observed sensitivity to movement islands suggests that NQ undergoes covert movement, presumably to a position in the CP system. In fact, the distribution of NQ is quite parallel to that of adverbial wh-words like weishenme ‘why’, also generally assumed to undergo covert movement to CP: The class of constituents from which A-not-A questions are excluded appears to be the same as the class of constituents from which non-nominal adjunct wh-phrases like weishenme ‘why’ are excluded. To capture this, parallel, Huang (1982a, 1991) proposed that NQ is, in essence, a non-nominal adjunct wh-phrase. Examples showing the parallel ill-formedness of weishenme in islands are given below. Of course, weishenme questions are grammatical when weishenme is not embedded in an island. (36) *[women weishenme chuli zhejian shi] bijiao hao? we why handle this matter more appropriate (‘What is the reason x such that it is more appropriate [for us to handle this matter for x]?’) (37) *ni bijiao xihuan [[ta weishenme zhu] de cai]? you more like he why cook de dish (‘What is the reason x such that you like better [the dishes which he cooks for x]?’) (38) *Lisi hen yihan [ta weishenme cizhi]? Lisi very regret he why resign (‘What is the reason x such that Lisi regrets [that he resigned for x]?’) (39) *ni jide [Lisi weishenme cizhi]? you remember Lisi why resign (‘What is the reason x such that you remember [that Lisi resigned for x]?’) A-not-A questions and weishenme questions pattern together in several other ways as well, as will be noted in upcoming sections. For example, both must have scope over focus-sensitive adverbs like ye ‘also’ and zhi ‘only’ (section 4.3), and over quantifiers like meiyouren ‘nobody’ (section 5.3). 3.4. A-not-A questions can be embedded: a further note To complete the argument that movement is implicated in the ill-formedness of Anot-A questions within islands, it is important to establish that A-not-A questions can be embedded within non-islands. This was shown in section 3.1, but there has been some debate on this point. Most prominently, McCawley (1994) objected that Huang’s examples used matrix verbs that lend themselves to a parenthetical interpretation. Countering Huang’s examples, McCawley provided the example (40), in an attempt to choose a matrix verb that would not be likely to receive a parenthetical interpretation. Indeed, using the verb xihuan ‘like’ as a matrix verb results in ungrammaticality when an A-not-A question is embedded as or in its complement. (40) *Lisi xihuan [women he bu he pijiu]? Lisi like we drink not drink beer (‘Does L like for us to drink beer?’) However, given that verbs that take nominal complements (as xihuan clearly does, cf., e.g., (1)) induce islands for both weishenme (39) and A-not-A questions (35), the ungrammaticality of (40) is unsurprising, and does not undermine the generalization that A-not-A questions can be embedded, just not in islands. 4. Focus marking and A-not-A questions Focus constructions in Mandarin interact with A-not-A questions in an interesting way. In this section we will first review some restrictions on these focusing constructions in general, and then turn to their interactions with and implications for A-not-A questions. 4.1 Restrictions on focus marking The copula shi is used in Mandarin as a focus marker, resulting in a cleft-like interpretation, as illustrated by the examples in (41–43). (41) shi wo mingtian yao mai neiben shu be I tomorrow want buy that book ‘It is I that want to buy that book tomorrow.’ (42) wo shi mingtian yao mai neiben shu I be tomorrow want buy that book ‘It is tomorrow that I want to buy that book.’ (43) wo mingtian shi yao mai neiben shu I tomorrow be want buy that book ‘I do want to buy that book tomorrow.’ Huang (1982a) observes that only one shi is allowed per clause (44–45), and that shi is not possible in wh-questions (46–48) except when the wh-word itself is focused by shi (49–50). Huang analyzes this as a consequence of there being a single focus position (high in the clause) to which both foci and wh-words must move for interpretation (see also Rizzi 1997 for essentially the same claim). Since both clefted constituents and wh-phrases compete for the same position, only one or the other is allowed. (44) *shi Zhangsan shi mingtian yao lai be Zhangsan be tomorrow want come (‘It’s Zhangsan, tomorrow that t will come t.’) (45) shi Zhangsan shuo Lisi shi mingtian lai be Zhangsan say Lisi be tomorrow come ‘It’s Zhangsan who said that it’s tomorrow that Lisi will come.’ (46) *shi Zhangsan da-le shei? be Zhangsan beat-perf who (‘Who is it Zhangsan who beat t?’) (47) shi Lisi xiang-zhidao [shei da-le ta] be Lisi wonder who beat him ‘It is Lisi that wonders who beat him.’ (48) *shi Lisi xiangxin [shei da-le ta] be Lisi believe who beat him (‘Who is it Lisi that believes that t beat him?’) (49) shi shei da-le ta? be who beat-perf him ‘Who is it that beat him?’ (50) ta xiang-zhidao [shi shei da-le ta] he wonder be who beat him ‘He wonders who it is that beat him.’ In further support of the idea that the focused constituent must undergo covert movement, Huang (1982a) observes that the focus marker shi cannot occur within movement islands (much like the NQ morpheme introduced in section 2), e.g., the relative clause in (51). (51) *[Zhangsan shi mingtian lai] mei guanxi. Zhangsan be tomorrow come not matter (‘It’s tomorrow that [that Zhangsan will come t] does not matter.’) 4.2 Focus markers in A-not-A questions Focusing elements interfere with the formation of A-not-A questions in revealing ways. Focus-sensitive adverbs such as zhi ‘only’ and ye ‘also’ are incompatible with NQ in cases where the adverb takes scope over the NQ on the surface. A simple example is given in (52). (52) *ni zhi xi bu xihuan yige ren? you only li-not-like one person (‘Do you only like one person?’) Recall that a modal accepts NQ (53), and notice that zhi can precede a modal (54) or follow it (55). However, putting the two together is fine only if NQ precedes zhi (56), and not if NQ follows zhi (57) (see Soh (to appear) for further discussion). (53) ta hui-bu-hui tiaowu? He can-not-can dance ‘Can he dance?’ (54) ta zhi hui shuo Yingyu. he only can speak English ‘He can only speak English (no other languages)’ (55) ta hui zhi shuo Yingyu. he can only speak English ‘He can only speak English (since none of us speak Chinese)’ (56) ta hui-bu-hui zhi shuo Yingyu? he can-not-can only speak English ‘Can he only speak English?’ (57) *ta zhi hui-bu-hui shuo Yingyu? he only can-not-can speak English (‘Can he only speak English?’) The focus marker shi can take NQ (a B-not-B question), enabling NQ to take scope over zhi on the surface. Again, when this obtains, zhi is acceptable in an A-not-A question (58). (58) ta shi-bu-shi zhi hui shuo Yingyu? he be-not-be only can speak English ‘Is it the case that he can only speak English?’ It is not, however, possible to have NQ realized on zhi itself (59). (59) *ni zhi bu zhi xihuan yige ren? you only not only like one person ‘Do you only like one person?’ This interaction with focus-sensitive adverbs is peculiar to A-not-A questions; zhi does not cause similar problems in ma-questions (60), nor do the effects appear in full alternative questions with an explicit haishi ‘or’ (61). (60) ni zhi xihuan yige ren ma? you only like one person Q ‘Do you only like one person?’ (61) ni zhi xihuan yige ren haishi bu zhi xihuan yige ren? you only like one person or not only like one person ‘Do you only like one person or not only like one person?’ 4.3 Parallels to weishenme ‘why’ questions The surface scope restrictions on NQ seem to be for the most part parallel to the restrictions on weishenme ‘why’, as observed by Soh (to appear). Just as zhi ‘only’ and ye ‘also’ interfere with the association of NQ to its scope position, so do they interfere with the association of weishenme to its scope position (62), unlike the association of nominal argument wh-phrases like shei ‘who’ (63). (62) *ni zhi renwei Lisi weishenme kan zhentan-xiaoshuo? you only think Lisi why read detective-novel ‘What is the reason x such that you only think Lisi reads detective novels for x?’ (63) ta zhi renwei shei you qiantu? he only think who has prospect ‘Who does he only think has prospect?’ With respect to the interaction with clefting using shi, weishenme acts differently from other wh-words, as Hu (2002) observes. Whereas a nominal argument wh-word like shenme ‘what’ cannot co-occur with a clefted constituent but can itself be focused (46–50), the distribution of weishenme is the reverse: weishenme can co-occur with a clefted constituent (64), and cannot usually itself be focused with shi (65–66). (64) weishenme shi Zhangsan bixu cizhi? why be Zhangsan must resign ‘Why is it Zhangsan that must resign?’ (65) *women shi weishenme nianshu? we be why study (‘Why is it that we study?’) (66) *shi weishenme Zhangsan bixu cizhi? be why Zhangsan must resign (‘Why is it that Z must resign?’) Inspired by the other parallels we have seen between NQ and weishenme, we might expect to find that the same properties hold of NQ. However, it turns out to be difficult to test. Hu (2002) attempts to show this by noting that NQ in Anot-A questions repels focusing by shi (67), as weishenme does (65–66). (67) *ni shi qu-bu-qu meiguo? you be go-not-go America ‘Do you go or not go to America?’ However, it is not clear that we can draw any real conclusions from the ungrammaticality of (67). First, as was noted earlier (11–12), the NQ morphology necessarily appears on a modal/auxiliary when present, and shi in (67) is arguably such an element, ruling (67) out independently. The other property of weishenme noted above is that it can co-occur with shi, which arguably holds between NQ and shi in B-not-B questions like (58). We will return to some other parallels in section 5.3 as well. It seems that the distribution of NQ is at least consistent with that of weishenme, even if it is difficult to show the parallelism directly. 4.4. Discourse conditions on A-not-A questions A-not-A questions and yes-no questions differ with respect to the types of discourse they fit into. One of the earliest discussions of this was in Li & Thompson (1979), who observed that A-not-A questions have an essential “neutrality” causing them to seem inappropriate when the questioner has a predisposition toward one of the alternatives; in such a circumstance, a yes-no question with ma is used. A more recent study by Schaffar & Chen (2001) examined some of the pragmatic requirements of A-not-A, B-not-B, and maquestions in some detail. Schaffar & Chen (2001) put Li & Thompson’s observation into more precise terminology by proposing that A-not-A questions are questions with a “wide focus”, where neither the entire proposition nor its negative alternative are presupposed. A-not-A questions contrast with B-not-B questions, formed using the narrow focus marker shi, and which are only felicitous where narrow focus is appropriate. Both types of questions differ from ma-questions, which do not have any information structure restrictions. The examples and contexts that follow are from Schaffar & Chen (2001). The first context to consider is an “out of the blue” question, with no presumption about what the answer is. An example is given in (68), where either an A-not-A question or a ma-question is felicitous. (68) Context: After spending an evening at the cinema, a friend and I are hungry and are driving by a restaurant called “Tabula Rasa”. I ask my friend: a. ni lai-mei-lai guo Tabula Rasa? you come-not-come asp Tabula Rasa ‘Have you been to the Tabula Rasa before?’ b. ni lai guo Tabula Rasa ma? you come asp Tabula Rasa Q ‘Have you been to the Tabula Rasa before?’ In a more “slanted” context like (69), where one alternative does have the upper hand, an A-not-A question is not appropriate, but either a B-not-B question or a ma-question (with sentence stress, indicated by capitals) is possible. Schaffar & Chen characterize this as “verum focus”, or contrastive narrow focus on the truth value, taking as background the proposition (either affirmative or negative) as a whole. (69) Context: I invite a friend to Tabula Rasa, which I have recently discovered. As we enter the restaurant, a waiter I don’t know breaks into a smile and waves. I ask my friend: a. ni shi-bu-shi lai guo Tabula Rasa? you be-not-be come asp Tabula Rasa ‘Have you been to the Tabula Rasa before?’ b. ni LAI guo Tabula Rasa ma? you come asp Tabula Rasa Q ‘Have you been to Tabula Rasa before?’ Narrow focus is of course possible on other constituents as well, for example in (70). In (70a), the focus is expressed entirely by the sentence stress, and in (70b) it is expressed with the help of the focus marker shi. The example in (70b) is interesting in that the sentence accent takes priority in marking the focused constituent, which in fact precedes shi. Often the use of the focus marker shi is characterized as focusing any constituents following it in the sentence, but (70b) suggests that this is not an absolute constraint. (70) a. Peitela ZUOTIAN zuo-le suxi ma? Petra yesterday make-perf sushi Q ‘Did Petra make sushi YESTERDAY (not today)?’ b. Peitela ZUOTIAN shi-bu-shi zuo-le suxi? Petra yesterday be-not-be make-perf sushi ‘Did Petra make sushi YESTERDAY (not today)?’ Based on facts like this, Schaffar & Chen (2001) propose that B-not-B questions unambiguously indicate narrow focus, either verum focus if there is no special sentence accent, or on the constituent marked with sentence accent. Anot-A questions, on the other hand, unambiguously indicate wide focus. Yes-no questions with ma do not have any special focus constraints; the same focus options are available to yes-no questions as to declaratives. Judgments on what can be focused are quite difficult, and it is not surprising that there are conflicting judgments reported in the literature. To give an example, Zhang (2001) reports that the subject and any adverbials that precede NQ are unavailable to focus (71). Although Schaffar & Chen only address the A-not-A and B-not-B cases (71a–b), their analysis predicts that no narrow focus reading should be possible in (71a), and that all narrow focus readings should be possible in (71b). (71) a. Lisi gangcai zheng-mei-zheng jidan? Lisi just steam-not-steam egg b. Lisi gangcai shi-bu-shi zheng-le jidan? Lisi just be-not-be steam-ASP egg c. Lisi gangcai zheng jidan meiyou? Lisi just steam egg not d. Lisi gangcai zheng jidan mei-zheng? Lisi just steam egg not-steam ‘Did Lisi STEAM eggs just now?’ ‘Did Lisi steam EGGS just now?’ ‘Did Lisi STEAM EGGS just now?’ *‘Did LISI steam eggs just now?’ *’Did Lisi steam eggs JUST NOW?’ Nevertheless, the basic characterization of A-not-A questions as being strictly restricted to neutral contexts (in which neither alternative is presupposed), as opposed to ma-questions (that allow—but do not require—a more slanted context), is undisputed. Future research may provide more clarification of the focus properties of sentence accent as compared to focusing with shi, and better diagnostics for discerning what can and cannot be focused in Mandarin questions. The null hypothesis is that shi has the same properties whether it carries NQ or not. 5. Further surface scope restrictions on NQ Along with the restrictions discussed above on the surface scope of focussensitive adverbs and NQ in A-not-A questions, there are a number of other similar restrictions that have been discussed in the literature. We will review several such restrictions in this section, on adverbs, indefinites, and quantifiers. 5.1 Adverbs In one of the earliest serious discussions of A-not-A questions, Li & Thompson (1979) pointed out that certain adverbs interfere with A-not-A formation, giving as examples man-man-de ‘slowly’ or jing-jing-de ‘quietly’. Ernst (1994) explores these restrictions in much greater detail, showing that there is a structural explanation for this interference. Ernst considers three classes of adverbial modifiers. The first class includes adverbs like yiding ‘definitely’. Such adverbs seem to require an interpretation that puts them within the scope of a question. To give analogs from English, a question like ‘Is he definitely going?’ is fine, whereas ‘Definitely: is he going?’ is anomalous. A second class of adverbs includes the adverb daodi, which Ernst glosses as ‘after all’. This adverb seems to require interpretation outside the scope of the question: an interpretation like ‘After all: is he going?’ is fine, whereas ‘Is he after all going?’ is anomalous. The third class, including adverbs like jintian ‘today’ do not seem to have any particular scope requirements. Ernst shows that the scope requirement for any given adverb must be satisfied by the surface scope relation between the adverb and the NQ morphology. Ernst makes the further assumption (adopting Huang’s (1982b) Isomorphism Principle) that the surface scope relations must be preserved in the LF structure that feeds interpretation, so that if the adverb outscopes the NQ on the surface, it must also be interpreted outside the scope of the question (at LF). Assuming, along with an essentially unanimous literature, that the question particle ma occupies a complementizer position high in the structure taking scope over the rest of the clause, we have a straightforward explanation for why adverbs like yiding ‘definitely’ are compatible with ma-questions like (72) but are incompatible with A-not-A questions like (73). In (72), yiding is in the scope of ma, so the adverb is interpreted within the scope of the question, as required for this class of adverbs. In (73), yiding outscopes NQ on the surface and thus would be forced to take scope outside of the question operator at LF, in violation of the requirements of this adverb class. Where the scope of NQ and yiding is reversed, as in the B-not-B question in (74), the question is fine. (72) ta yiding qu ma? (YNQ) he definitely go Q ‘Is he definitely going?’ (73) *ta yiding qu-bu-qu?(A-not-A) he definitely go-no-go (‘Is he definitely going?’) (74) ta shi-bu-shi yiding qu? (B-not-B) he be-not-be definitely go ‘Is he definitely going?’ The class of adverbs that includes jintian ‘today’ is compatible with both Anot-A and ma-questions, as in (75) and (76). (75) ni jintian qu-bu-qu? (A-not-A) you today go-not-go ‘Are you going today?’ (76) ni jintian qu ma? (YNQ) you today go Q ‘Are you going today?’ The remaining class, exemplified by the adverb daodi ‘after-all’ shows the opposite pattern from the first. The adverb daodi is allowed in A-not-A questions (77), but not in ma-questions (78). The explanation of the patterning of daodi is the same as the explanation of yiding, but for the fact that daodi and yiding have opposite requirements. Since daodi must take scope outside the question operator at LF in order to be interpreted, (78) is ruled out because daodi is in the scope of ma on the surface. On the other hand, (77) is fine, since daodi has scope over NQ and if the scope relation is maintained at LF, the interpretive requirements of this adverb class are satisfied. Testing B-not-B questions shows that daodi is permitted only when it outscopes NQ on the surface (79–80). (77) Jinrong daodi qu-bu-qu? (A-not-A) Jinrong after-all go-not-go ‘Is Jinrong going or not?!’ (78) *Jinrong daodi qu ma? (YNQ) Jinrong after-all go Q (‘Is Jinrong going or not?!’) (79) *ta shi-bu-shi daodi qu? He be-not-be after-all go (‘Is he really going?!’) (B-not-B) (80) ta daodi shi-bu-shi qu? He after-all be-not-be go ‘Is he really going?!’ (B-not-B) The conclusion is that there is something interpretively significant about the surface position of NQ in A-not-A questions, and moreover, NQ is invariably in a lower structural position in A-not-A questions than the question particle ma is in yes-no questions. It is also possible for NQ to appear on adverbs themselves, such as on chang ‘often’ in (15) above. Not all adverbs can appear with NQ, however; Hsieh (2001) notes that some adverbs, such as guyi ‘deliberately’, cannot be marked with NQ (81). (81) *Ta guyi-bu-guyi qu? he deliberately-not-deliberately go (‘Did he not go deliberately?’) It is generally assumed that frequency adverbs like ‘often’ are lower in the structure than subject-oriented adverbs like ‘deliberately,’ (see, e.g., Cinque (1997), Ernst (2002)), and extending that line of thought, one avenue to explore would be to suppose that guyi ‘deliberately’ necessarily has a higher place in the structure than NQ, as does daodi ‘after-all’, while other adverbs (possibly including chang ‘often’, yiding ‘definitely’, and jintian ‘today’) are more flexible and (may) attach lower than NQ—in fact, future research might be profitably done looking for subtle differences in meaning depending on the relative positions of NQ and such adverbs. 5.2 Indefinite wh-words It is well known that wh-words in Mandarin act something like “polarity items”, acting like a variable bound by question or negation operators elsewhere in the sentence (see, e.g., Li 1992, Lin 1996 for discussion). In a simple affirmative sentence, wh-words are disallowed (82), but they receive an indefinite interpretation under negation (83) and in conditionals (84). In wh-questions, of course, wh-words function as wh-words (85). (82) *wo xihuan shenme ren. I like what person ‘I like somebody.’ (83) wo mei mai shenme (dongxi). I not buy what (thing) ‘I didn’t buy anything.’ (84) yaoshi shenme ren qifu ni, … if what person bully you ‘If someone bullies you, …’ (85) ni xihuan shenme ren? you like what person ‘Who do you like?’ The indefinite interpretation for wh-words is also available in other contexts, for example in (86), with the sentence-final particle le indicating an inference, rather than a direct observation, and in yes-no questions formed with the sentence-final particle ma (87). (86) ta kandao shenme le he see what LE ‘(It seems that) he saw something.’ (87) ta xihuan shenme ma? he like what Q ‘Does he like something/anything?’ The contexts in which a wh-word can be interpreted as an indefinite are characterized by Li (1992) as being ‘uncertain’ and by Lin (1996) as not entailing the existence of a referent for the indefinite. What allows the indefinite interpretation in (87), for example, is the presence of the question particle ma. The prevailing opinion has been that ma will license an indefinite interpretation of a wh-word in any position, including the subject position. However, conflicting judgments have been reported in the literature (and in my own informal consultations), which may undermine the use of indefinites in subject position as a useful diagnostic for structure in A-not-A questions. The paradigm examples are these: In A-not-A questions, a wh-word in object position (88), but not in subject position (89), can be interpreted as an indefinite. However, in B-not-B questions, a wh-word in subject position can receive an indefinite interpretation (90). (88) ta xi-bu-xihuan shenme ren? he li-not-like what person ‘Does he like someone?’ (89) *shenme ren xi-bu-xihuan ta? what person li-not-like him (‘Does someone like him?’) (90) shi-bu-shi shenme ren xihuan ta? be-not-be what person like him ‘Does someone like him?’ Li (1992) interprets this paradigm as an indication that NQ is the licensor of the indefinite interpretation, and can only license wh-words in its scope. It has also been pointed out (see Lin 1996, Schaffar & Chen 2001) that a B-not-B question in which the NQ appears question-finally also allows a subject wh-word to be interpreted as an indefinite (91). Assuming that this sentence-final position is within CP, at or near the position in which ma is found in yes-no questions, this again shows an indefinite interpretation licensed in the scope of NQ. (91) shei/shenme ren xihuan ta shi-bu-shi? who/what man like him be-not-be ‘Does someone/anyone like him?’ The judgments above are not much in dispute, but whether the yes-no question operator ma can license an indefinite interpretation of a wh-word in subject position, and by extension, the validity of the generalization that a question operator (NQ or ma) can always license such an interpretation for whwords in its scope, is less clear. Li (1992) gives (92) as an example of an indefinite in subject position licensed by ma, glossing it as shown, yet Huang (1982b) gives a nearly identical question (93), rating it as ungrammatical, a judgment endorsed by Cheng (1997). (92) shenme ren xihuan ta ma? (Li 1992: 128) what person like him Q ‘Does someone like him?’ (93) *shei xiang chi pingguo ma? (Huang 1982b:244, Cheng 1997:108) who want eat apple Q ‘Does someone/anyone want to eat apples?’ Lin (1996) provides (94) as an example like (92), but it is not completely clear whether ma or le is responsible for its acceptability. (94) shei you qifu ni le ma? who again bully you ASP Q ‘Did somebody bully you again?’ Schaffar & Chen (2001) rate (92) as grammatical, but suggest that it actually has a reading much more like a wh-question, not allowing an answer of shi (xihuan) ‘yes’ or bu (xihuan) ‘no’, but only an answer like Lisi xihuan ta ‘Lisi likes him’, and suggesting a meaning more like ‘Who likes him, doesn’t somebody?’ This interpretation is unusual in light of the fact that the question particle ma is almost universally judged to be incapable of appearing in wh-questions. Schaffar & Chen (2001) and Hua (1999), in discussing examples like (91), observe when a modal or auxiliary element bears NQ, even when it appears sentence internally, it can license an indefinite interpretation of a wh-subject. They provide examples like (95) and (96), to be compared to (89). (95) shei/shenme ren shi-bu-shi xihuan ta? who/what man be-not-be like him ‘Does someone/anyone like him?’ (96) shei you-mei-you du-guo zhe-ben shu who have-not-have read-asp this-cl book ‘Has someone/anyone read this book?’ We cannot further resolve or clarify the disagreements in the judgments here, although this is clearly an area in need of more thorough investigation. It appears that judgments are particularly volatile on the availability of indefinite interpretations of wh-subjects. This may be due to independent factors affecting the subject position, which under most circumstances is construed as a topic and hence incompatible with an indefinite interpretation. In this connection, it is worth mentioning that placing you ‘have’ before a wh-subject in a sentence like (92) will uncontroversially permit an indefinite interpretation of the wh-word, and which may interfere somewhat with judgments of such sentences lacking you. The whobject examples in the first part of this section suggest that NQ is capable of licensing an indefinite interpretation, but several issues remain unresolved. 5.3 Intervening quantifiers Wu (1997a) observes that in contrast to yes-no questions formed with the question particle ma (97), A-not-A questions have a restriction against having quantifiers outside the NQ. In (98), the quantifier is outside the A-not-A morphology, and the examples are degraded or ungrammatical. (97) (98) a. meigeren dou pao ma? everyone all run Q ‘Does everyone run?’ b. you ren pao ma? someone run Q ‘Does someone run?’ c. meiyou ren pao ma? no body run Q ‘Does nobody run?’ a. ?meigeren dou pao ba pao? everyone all run-not-run ‘Does everyone run or not?’ b. *you ren pao bu pao? someone run-not-run (‘Does someone run or not?’) c. *mei you ren pao by pao? no body run-not-run (‘Does nobody run or not?’) The fact that (98a) is rated better than (98b–c) is attributed to the possibility of meigeren dou getting a group interpretation, parallel to (99). (99) tamen pao bu pao? they run not run ‘Do they run or not?’ In a B-not-B question, such subject quantifiers are fine (100), so long as NQ is actually outside (101). In the absence of subject quantifiers, either position for shi-bu-shi is fine (102). (100) a. shi-bu-shi meigeren dou pao? be-not-be everyone all run ‘Is it the case or not that everyone runs?’ b. shi-bu-shi you ren pao? be-not-be someone run ‘Is it the case or not that someone runs?’ c. shi-bu-shi meiyou ren pao? be-not-be no body run ‘Is it the case or not that nobody runs?’ (101) a. *meigeren dou shi-bu-shi pao? everyone all be-not-be run (‘Is it the case that everyone runs?’) b. *you ren shi-bu-shi pao? someone be-not-be run (‘Is it the case that someone runs?’) c. *meiyou ren shi-bu-shi pao? no body be-not-be run (‘Is it the case that nobody runs?’) (102) (shi-bu-shi) Zhangsan (shi-bu-shi) pao? be-not-be Zhangsan be-not-be run ‘Is it the case or not that Zhangsan runs?’ With respect to meigeren dou ‘everyone’, Wu observes that the scope requirements are actually on dou, as can be seen by comparing (101a) to (103). In the grammatical (103) and (100a), shi-bu-shi takes scope over dou, but in the ungrammatical (101a), the scope relation between them is reversed. (103) meigeren shi-bu-shi dou pao? everyone be-not-be all run ‘Is it the case that everyone runs?’ It is worth noting that these same elements seem to block association of weishenme ‘why’ with its scope position (104), in contrast to nominal wh-words like shei ‘who’ (105) (Soh, to appear). Sentence (106) makes the same point as (98), but is provided for parallelism to (104–105), showing the incompatibility of these subject quantifiers in A-not-A questions. Thus, if movement of the same type is implicated for both weishenme and NQ, this might be viewed as a type of weak island phenomenon, further supporting the view that NQ undergoes movement. (104) *meiyouren/henshao ren/zuiduo liang-ge ren weishenme cizhi? nobody/few person/at most two-cl person why resign (‘What is the reason x such that nobody/few people/at most two people resigned for x?’) (105) meiyouren/henshao ren/zuiduo liang-ge ren gan gen shei dajia? nobody/few person/at most two person dare with who fight ‘Who is the person x such that nobody/few people/at most two people dare(s) to fight with x?’ (106) *meiyouren/henshao ren/zuiduo liang-ge ren qu-bu-qu? nobody/few person/at most two-cl person go-not-go (‘Is nobody/few people/at most two people going?’) 6. Negative particle questions and VP-not-V questions There are two other question forms that appear on the surface to be very similar to A-not-A questions. The first of these question types is the negative particle question, which ends in a negative morpheme (107); the second is what we will refer to as the VP-not-V question, which ends with a negative morpheme and a second copy of the verb (108). (107) Hufei kan-wan-le nei-ben shu meiyou (NPQ) Hufei read-finish-perf that-cl book not-have ‘Has Hufei finished reading that book?’ (108) ta xihuan neige dianying bu-xihuan? he like that movie not-like ‘Does he like that movie?’ (VP-not-V) A natural null hypothesis is that both NPQs and VP-not-V questions are derived from A-not-A questions; that is, we might suppose that a VP-not-V question arises from movement of the not-V from the V-not-V complex, and that an NPQ arises from additionally omitting the copy of the verb. However, appearances can be deceiving, and there has been debate as to which of these constructions, if any, are related to the others. Much of the discussion in this section draws upon work by Cheng, Huang, and Tang (1996), N. Zhang (1997), and Hsieh (2001). 6.1. Distributional similarities and differences We will begin by looking at ways in which these question types are similar to one another. One property they have in common is that the verb being questioned cannot already be negated. While it is perfectly acceptable to ask a negative yesno question with ma (109), such questions cannot be asked in the VP-not-V (110), NPQ (111), or A-not-A forms (112). (109) ta bu chang jing-ju ma? he not sing Beijing-Opera Q ‘Does he not sing Beijing-Opera?’ (110) *ta bu chang jing-ju bu-chang? he not sing Beijing-Opera not-sing (‘Does he not sing Beijing-Opera?’) (111) *ta bu chang jing-ju bu? he not sing Beijing-Opera not (‘Does he not sing Beijing-Opera?’) (112) *ta bu qu qu? he not go go (‘Isn’t he going?’) Just as in simple declaratives and in A-not-A questions, the choice of the negative morpheme in NPQs depends on the aspectual structure of the clause. Recall that there are two forms negation can take in Mandarin, bu (for bare verbs and modals, generally unbounded events), and mei(you) (for verbs that are marked with aspectual particles, and generally for bounded events), demonstrated in (113). NPQs can be formed with either (114), as can A-not-A questions (115), but which form of negation is used depends on the aspectual properties of the sentence in the same way it does in declaratives. (113) a. b. (114) a. b. (115) a. b. Hufei bu/*meiyou hui qu Hufei not/not-have will go ‘Hufei wll not go.’ Hufei meiyou/*bu qu-guo Hufei not-have/not go-exp ‘Hufei has not been (there).’ Hufei hui qu bu? Hufei will go not ‘Will Hufei go?’ ta qu-guo meiyou/*bu he go-asp not.have/not ‘Has he been (there)?’ Qiaofeng qu-bu-qu Qiaofeng go-not-go ‘Is Qiaofeng going?’ ta you-meiyou qu-guo meigo? he have-not-have go-exp America ‘Has he been to America?’ c. Qiaofeng qu-mei-qu? Qiaofeng go-not-go ‘Did Qiaofeng go?’ (Beijing Mandarin) VP-not-V questions show the same type of aspect agreement; for example, if meiyou is used instead of bu with an unbounded event, the result is ungrammatical (116); compare to (108). (116) *ta xihuan neige dianying mei(you)-xihuan? he like that movie not-like (‘Does he like that movie?’) A natural question to ask is whether the three kinds of questions pattern together with respect to sensitivity to islands. Initially, this appears to be the case: As Hsieh (2001) observes, NPQs, like A-not-A questions, cannot take scope out of movement islands such as sentential subjects (117), relative clauses (118) or reason adjuncts (119). (117) *[you qian meiyou] bijiao hao? have money not.have more good (‘Whether to have money is better?’) (118) *ni bijiao xihuan [you qian meiyou de ren]? you more like have money not.have de person (‘Who do you like better, the people who have money or the people who do not?’) (119) *Zhangsan [yinwei ni chi-le fan meiyou], suoyi hen shengqi? Zhangsan because you eat-asp rice not.have therefore very angry (‘Zhangsan got angry because whether you ate or not?’) This would be a suggestive similarity, but for the fact that, as N. Zhang (1997) notes, neither NPQs nor VP-not-V questions are content inside such clauses even when they take scope inside. While A-not-A questions can be embedded in islands if the scope of the question is contained within the island as in (120), neither NPQs (121) nor VP-not-V questions (122) permit even this type of embedding. While the incompatibility with islands in A-not-A questions has to do with establishing a relation between NQ and the scope of the question, this cannot be the explanation for the ungrammaticality of (121–122) nor can it be the only explanation for the ungrammaticality of (117–119). This, then, seems to be a true difference between A-not-A questions on the one hand and NPQs and VPnot-V questions on the other. (120) a. wo zai kaolu [ta mai-bu-mai shu] de wenti. I at consider [he buy-not-buy book] de question ‘I am considering the question whether he will buy the books.’ b. ta qu-mei-qu-guo Beijing gen wo wuguan. he go-not-go-asp Beijing to me irrelevant ‘Whether he has been to Beijing is irrelevant to me.’ c. wulun/buguan ni xi-bu-xihuan ta, wo dou yao jia ta. no-matter/regardless you like-not-like he, I all want marry he ‘Regardless of whether you like him or not, I want to marry him.’ (121) a. *wo zai kaolu [ta mai shu bu] de wenti. I at consider [he buy book not] de question (‘I am considering the question of whether he will buy the books.’) b. *ta qu-guo Beijing mei-qu-guo gen wo wuguan. he go-asp Beijing not-go-asp to me irrelevant (‘Whether he has been to Beijing is irrelevant to me.’) c. *wulun/buguan ni xihuan ta bu-xihuan, wo dou yao jia ta. no-matter/regardless you like he not-like, I all want marry he (‘Regardless of whether you like him or not, I want to marry him.’) (122) a. *wo zai kaolu [ta mai shu bu-mai] de wenti. I at consider [he buy book not-buy] de question (‘I am considering the question of whether he will buy the books.’) b. *ta qu-guo Beijing meiyou gen wo wuguan. he go-ASP Beijing not to me irrelevant (‘Whether he has been to Beijing is irrelevant to me.’) c. *wulun/buguan ni xihuan ta bu, wo dou yao jia ta. no-matter/regardless you like he not, I all want marry he (‘Regardless of whether you like him or not, I want to marry him.’) There are other reasons to differentiate A-not-A questions from NPQs and VP-not-V questions as well. N. Zhang (1997) observes that, while the NQ morphology is possible on different kinds of predicates in A-not-A questions, for example the preposition in (123a), only verbs can participate in VP-not-V questions (123b). (123) a. b. ni gen-bu-gen ta shuohua? you to-not-to him speak ‘Do you speak to him?’ *ni gen ta shuohua bu-gen? you to him speak not-to (‘Do you speak to him?’) Moreover, whereas in A-not-A questions the predicate copy can consist only of the first syllable (124a), the reduplicated verb cannot be truncated in a VP-not-V question (124b–c). Likewise, the negative morpheme meiyou may appear in its full form in an NPQ, but must be truncated to mei in an A-not-A question (125). (124) a. ta xi-bu-xihuan zheiben shu? he like-not-like this book ‘Does he like this book?’ b. *ta xi-zheiben shu bu xi-(huan)? he like-this book not like (‘Does he like this book?’) c. *ta xi(huan) zheiben shu bu xi-? he like this book not like (‘Does he like this book?’) (125) a. ta chi-mei-chi zaofan? he eat-not-eat breakfast ‘Has he eaten breakfast?’ b. *ta chi-meiyou-chi zaofan? he eat-not-eat breakfast (‘Has he eaten breakfast?’) c. ta chi zaofan mei(you)? he eat breakfast meal not ‘Has he eaten breakfast?’ Cheng, Huang, and Tang (1996), citing a historical study by Zhang (1990), note that NPQs are much older than the other types of yes-no questions. In Classical Chinese, such questions were formed with a negative particle and an interrogative particle in final position. Over time the interrogative particle disappeared, leaving the negative particle alone to mark the question. Such evidence suggests that we may not want to derive NPQs from other A-not-A constructions, since in Classical Chinese only the NPQs existed. It also suggests that in Classical Chinese, and perhaps by extension in modern Chinese, the negation in an NPQ is doing “double duty”, carrying features of both negation and interrogativity. Cheng et al. (1996) bring up a further potential difference between these question types. Recall that certain preverbal adverbs are disallowed in A-not-A questions (discussed in section 5.1), such as chang ‘often’ and yijing ‘already’ in (126). Cheng et al. observe that these adverbs are also ungrammatical in VPnot-V questions (127). Further, they rate these adverbs as being fine with NPQs (128). (126) a. b. (127) a. b. (128) a. b. *ta chang qu-bu-qu? he often go-not-go (‘Does he go often?’) *ta yijing you-meiyou kan-wan shu? he already have-not-have read-finish book (‘Did he already finish reading the book?’) *ta chang pian-ni-bu-pian? he often cheat-you-not-cheat (‘Does he often cheat you?’) *ta yijing kan-wan shu mei-kan-wan? he already read-finish book not-read-finish (‘Did he already finish reading the book?’) ta chang qu bu? he often go not ‘Does he go often?’ ta yijing kan-wan shu mei(you)? he already read-finish book not-have ‘Did he already finish reading the book?’ Subsequent literature has not unanimously agreed with the judgments reported by Cheng et al. for (128), however; Hsieh (2001) judges (128a) to be ungrammatical, and N. Zhang (1997) rates (128b) the same. At least for dialects in which all of the examples in (126–128) are ungrammatical, they cannot serve to support an argument for differentiating NPQs, VP-not-V questions, and A-notA questions. There may in fact be prosodic constraints that have not been thoroughly explored influencing the acceptability of these sentences. N. Zhang (1997) agrees with Cheng et al. that (128a) is acceptable, but finds that replacing the monosyllabic adverb chang with the disyllabic synonym jingchang ‘often’ renders it ungrammatical, unless the sentence is longer, e.g. with an expressed goal argument (129). Pending further systematic investigation, the diagnostics based on the acceptability of adverbs with NPQs are of uncertain status. (129) a. b. *ta jingchang qu bu? he often go not (‘Does he go often?’) ta jingchang qu Shanghai bu? he often go Shanghai not ‘Does he go to Shanghai often?’ 6.2. The location of Neg(-V) Cheng, Huang, and Tang (1996) propose that the negative morpheme at the end of NPQs (in Mandarin) is located in a complementizer position, like the question particle ma in yes-no questions. In support of this conclusion, they observe that the negative particle in NPQs cannot co-occur with the question particle ma. Yet, ma is generally not allowed with alternative questions of any sort, so the inability of ma to co-occur with NPQs may well be due to an independent semantic constraint. Cheng et al. (1996) also judge that the sentence-final particle ne, which can appear both in wh-questions and A-not-A questions, cannot appear in NPQs. However, both N. Zhang (1997) and Hsieh (2001) take issue with this judgment, indicating that for them questions such as (131–132) are grammatical. (130) ta xihuan ni bu-xihuan ne? he like you not-like Comp ‘Does he like you?’ (131) ni gei-le ta qian meiyou ne? you give-asp he money not.have prt ‘Did you give him money?’ (132) ta chi-le fan meiyou ne? he eat-asp meal not.have prt ‘Has he eaten?’ James Huang (p.c.) suggests that (131–132) might be acceptable with ne if the final meiyou allows them to be analyzed as VP-not-V questions (negating the auxiliary you) rather than as NPQs; he notes that if meiyou is shortened to mei (rendering the questions unambiguously NPQs), these examples degrade considerably. Hsieh (2001) puts forth two further arguments against the claim that the negative morpheme in NPQs is in the same position that sentence-final particles such as ne and ma occupy. First, NPQs are possible in an embedded clause (133), while sentence-final particles are not (134–135). Examples (133b–c) show that the embedded clause is in fact an embedded clause, by showing that it can be circumscribed by lian…dou (‘even’) and can be replaced by zhe-ge ‘this’ (136). (133) a. b. ta xiang zhidao [ta lai le meiyou] he want know he come asp not.have ‘He wants to know whether he came.’ ta bu zhidao ta ziji pang le meiyou he not know he self fat asp not.have ‘He does not know whether he gained weight.’ c. ta lian ta ziji pang le meiyou dou bu zhidao he even he self fat asp nothave all not know ‘He does not even know whether he gained weight.’ (134) ta xiang zhidao [shei lai-le (*ne)] he want know who come-asp prt ‘He wants to know who came’ (135) wo bu zhidao ta qu-guo Beijing le (*ma). I not know he go-asp Beijing asp Q ‘I do not know whether he has been to Beijing’ (136) ta lian zhe-ge dou bu zhidao. he even this-cl al not know ‘He did not even know this.’ Hsieh’s (2001) second argument has to do with the distribution of daodi ‘after-all’, which (as discussed in section 5.1) seems to be prohibited from the scope of the interrogative morpheme. Recall that this explained why daodi is not allowed in yes-no questions with the question particle ma: ma is so high in the structure that it takes scope over everything else. If the negative/interrogative particle in NPQs were also that high in the structure, it should have the same effect on daodi. However, as Cheng et al. themselves observe, daodi is just fine in NPQs (137), as it is in the other A-not-A questions. (137) ni daodi gei-le ta qian meiyou? you indeed give-asp he money not.have ‘Did you indeed give him the money?’ Assuming that the constraint on where daodi can appear is accurately characterized, this, like (133) above, tells us that the position of the negative morpheme in NPQs is not as high in the structure as the position of ma in yes-no questions. It is of course a separate question whether the negative morpheme in NPQs moves to its sentence-final position from lower in the clause. The arguments above suggest that there are at least two sentence-final positions, the position occupied by negation in NPQs being lower than the position occupied by ne and ma. Cheng et al. (1996) argue that in Mandarin NPQs, the negative morpheme has moved to its surface position based on the aspect agreement facts reviewed next. Which negative morpheme is used in an NPQ depends on the aspectual properties of the sentence, just as in declaratives. Assuming that such agreement must be determined locally, this negative particle must have originated within the clause being questioned. As evidence for this, they observe that the form negation takes can disambiguate which verb is being questioned when there is an embedded clause. The question in (138) is ambiguous, since both the matrix and embedded clauses would use the bu form in the negative, and the question in (139) is ungrammatical, since meiyou is not compatible with either clause. Only a reading in which the matrix verb is being questioned is possible for (140), since bu is incompatible with the aspect-marked embedded verb, and only a reading in which the embedded verb is being questioned is possible for (141), since meiyou is incompatible with the modal hui ‘will’ in the matrix clause. (138) ta yiwei ni qu bu? he think you go not ‘Does he think or not think that you are going?’ ‘Does he think that you are going or not going?’ (139) *ta hui yiwei ni yinggai qu meiyou? he will think you should go not-have (‘Will he think or not think that you should go?’) (‘Will he think that you should go or not go?’) (140) ta yiwei ni qu-guo bu? he think you go-exp not ‘Does he think or not think that you have been (there)?’ (141) ta hui yiwei ni qu-guo meiyou? he will think you go-exp not-have ‘Will he think that you have been (there) or you haven’t been (there)?’ The question in (141) is of particular interest because, although the agreement is with the embedded verb, it is still a matrix question. Cheng et al. take this as an indication that the sentence-final particle has moved from the embedded clause to the matrix CP system, responsible for both interrogative and negative features. One last point we might consider to help determine where in the structure neg-V appears in VP-not-V questions is the distribution of another clause-final particle, the (inchoative, perfective) le. N. Zhang (1997) observes that nothing can follow neg-V in a VP-not-V question, and furthermore that neg-V must follow the otherwise clause-final (inchoative) le (142). To this, I add (143), which shows that the same is true of NPQs. (142) ta chi fan le mei chi? he eat meal LE not eat ‘Has he eaten the meal?’ (143) ta chi fan le mei(you)? he eat meal LE not ‘Has he eaten a meal?’ 6.3. Distributional similarities and differences revisited Collecting together the observations from this section, the picture we arrive at is that there are at least three distinct clause-final positions. The highest of these is where the yes-no question marker ma appears, and the lowest is where the sentence final perfective/inchoative le appears. Between the two is the position in which neg and neg-V appear in NPQs and VP-not-V questions. For convenience, I will call these maP, leP, and npqP, respectively. That npqP is higher than and distinct from leP is clear from the surface forms, e.g. (142–143). The evidence from the distribution of daodi ‘after-all’ indicates that npqP is distinct from and structurally lower than maP: NPQs and VP-not-V questions are compatible with daodi, but yes-no questions with ma are not, indicating that there is a place in the structure for daodi that is between npqP and maP. The fact that NPQs and VP-not-V questions cannot be embedded within islands, regardless of the scope of the question (121–122), might be attributed to a syntactic impossibility: Clause-embedding verbs embed structures smaller than a maP (135), and for clausal adjuncts, even an npqP is too much structure (121–122). Following Cheng et al. (1996), we can suppose that the long-distance agreement between the form of negation in NPQs and VP-not-V questions and the internal aspectual properties of the clause diagnoses an overt movement of NQ from its clause-internal position (presumably the position in which it would appear in an A-not-A question) to npqP. This sketch of an analysis leaves unaddressed some of the facts in the preceding sections; in particular, it provides no clear explanation of the restriction to verbs in VP-not-V questions (as opposed to A-not-A questions, which allow NQ on adjectives, prepositions, and adverbs), nor does it add any immediate understanding of the prosodic requirements and restrictions on truncation of the predicate. However, it can serve as a simple model to work from in further exploration of these questions, and does at any rate serve as a reasonable synthesis of the facts so far accumulated. 7. VP-not-VP questions as VP-VP questions There is still one further question type that has appeared in the literature in the context of A-not-A questions: the VP-not-VP question. These are questions in which the entire VP is repeated after the negative morpheme. It is tempting to think that these are simply alternative questions with haishi omitted, but there are some distributional differences between VP-not-VP questions and alternative questions. As McCawley (1994) observes, VP-or-not-VP alternative questions are possible (even if somewhat unusual) when the order of the disjuncts is reversed: (144) is fine, but if haishi is missing, only the positive-negative order is possible (145–146). (144) ni bu xihuan ta haishi xihuan ta? you not like him or like him ‘Do you not like him or like him?’ (145) ni xihuan ta bu xihuan ta? you like him not like him ‘Do you like him or not like him?’ (146) *ni bu xihuan ta xihuan ta? you not like him like him (‘Do you not like him or like him?’) Huang (1991) points out that it is possible to ask alternative questions without haishi; e.g., (147) and (148) receive essentially the same interpretation. The interpretation they receive is as a straightforward alternative question, not asking for a choice between two alternatives that differ only in truth value, but between two distinct propositions. (147) ta chi fan haishi chi mian? he eat rice or eat noodle ‘Does he eat rice or noodles?’ (148) ta chi fan chi mian? he eat rice eat noodle ‘Does he eat rice or noodles?’ Zhang (1997) makes the interesting observation that even these affirmative disjunctive questions, when they lack haishi, cannot appear in islands (149–150). The implications of this observation are beyond the scope of this survey article, but it does suggest that the island sensitivity we see in A-not-A questions may not be specific to the properties of NQ, but rather to formation of this kind of alternative question in general (though not to alternative questions with haishi). A similar effect arguably occurs in English with the association between either and or (Larson 1985): In (151), the disjunction is easily interpretable outside of the embedded clause, but in (152) this reading disappears (leaving only the irrelevant reading comparing noodles to people who eat rice). Questions remain about what allows haishi its freedom from this kind of scope constraint. (149) ni xihuan [chi fan haishi chi mian] de ren? you like eat rice or eat noodle DE person ‘Do you like people who eat rice or who eat noodles?’ (150) *ni xihuan [chi fan chi mian] de ren? you like eat rice eat noodle DE person (‘Do you like people who eat rice or who eat noodles?’) (151) I either like to eat rice or noodles (but I can’t remember which). (152) I either like people who eat rice or noodles. James Huang (p.c.) also points out that disjunctive VP-VP questions (without haishi) are subject to a peculiar constraint requiring the two VPs to bear some phonological similarity: either the verb or the object must be identical between the two VPs; synonyms or pronouns will not satisfy the constraint. The VP-not-VP question has not been much discussed compared to NPQs, VP-not-V questions, and A-not-A questions. From these initial observations, it seems that the VP-not-VP questions are most properly considered to be essentially VP-VP disjunctive questions like (149), but for the incidental property that the second VP is the negated version of the first. 8. Cross-linguistic notes Although Mandarin Chinese is the language in which this family of question types has been most thoroughly studied, such questions appear in other languages as well. A full cross-linguistic and cross-dialectal comparison of question types is well beyond the scope of this review, and in fact only exists in the literature in bits and pieces, but we will briefly note some of the more prominent observations that have been made in this connection. Huang (1991) noted that Taiwanese and several other dialects feature a question particle that appears to be parallel to the NQ of Mandarin A-not-A questions, and hypothesized that it is an alternative realization of the same underlying morpheme (recall section 2.1). In Taiwanese, kam appears in the same place that NQ does in Mandarin (153). Given that the evidence for the existence of NQ as a single morpheme is rather indirect (even if convincing) in Mandarin, the fact that NQ is realized as a free morpheme in other dialects lends credence to the NQ hypothesis in Mandarin. Huang (1991) provides more support for the parallel by noting that kam in Taiwanese can take long distance scope (154), and cannot appear in islands (155–156). (153) li kam u chi:? you Q have money ‘Do you have money?’ (154) li siu: i kam e lai? you think he Q will come ‘Do you think he will come?’ (155) *[i kam u lai] kha hou? he Q have come more good (‘Is it better that he has come or has not come?’) (156) *li kha ai [kam u chi: e] lang? you more like Q have money REL person (‘Do you like the person who has money or doesn’t?’) Huang (1991) also cites Zhu’s (1985) cross-dialectal study, which showed, among other things, that A-not-A questions and kam-type questions generally do not co-occur in the same dialect, further supporting the idea that NQ and kam realize the same underlying morpheme. It should be noted that much subsequent literature has questioned Zhu’s (1985) conclusion about the complementary distribution of kam and A-not-A questions across dialects (see, e.g., Crosland (1998) for discussion), but it remains unclear to what extent this is theoretically problematic. In this connection, we can even point to Taiwanese as discussed by Huang (1991) which, in addition to the kam-type questions, does have a limited form of A-not-A questions that appear with a highly restricted subset of verbs. Hsieh (2001) provides a short list, including si ‘be’, bat ‘know’, and kaN ‘dare’, but notes that there does not seem to be a unifying property to the verbs which allow A-not-A questions and concludes that they must be lexically marked. Cole & Lee (1997) explore the properties of Singapore Teochew (ST) questions which also seems to have both A-not-A and kam-type questions, which, moreover, can co-occur in the same question. In the ST example in (157), the kam-type particle is ka, and it precedes an A-not-A sequence. (157) Ah Meng ka su m suka ji bun zi? (ST) Ah Meng Q li- not like this CL book ‘Does Ah Meng like this book?’ Cole & Lee review a number of syntactic tests, which show that the A-not-A forms are relatively unconstrained, able to appear within sentential complements and within movement islands. On the other hand, the ka question particle cannot be contained in movement islands, even though it can appear in embedded clauses when not within an island. These facts suggest that not everything that looks like an A-not-A question is the same: the A-not-A questions in Mandarin appear to be the analogs to kam questions in Taiwanese and to ka questions in ST, but crucially not analogs to the A-not-A questions in ST. Rather, ST A-not-A questions behave more like Mandarin alternative (haishi) questions. Although Cole & Lee essentially take this line, that ST A-not-A questions are just disjunctive questions, it is worth pointing out one fact that leans away from such a view: Notice that in (157), the A-not-A formation shares with Mandarin the property that the first copy of the affected predicate is reduced to only its first syllable. This same phenomenon was used as an argument in favor of the existence of NQ in Mandarin, given that ellipsis of the non-initial syllables of a word is not generally a possible syntactic operation. Given the syntactic evidence, we seem to be forced to conclude that ST A-not-A questions are basically just alternative questions, but that leaves us in the position of having to allow this type of “sub-word-level” ellipsis into the grammar. The facts just reviewed from ST reveal that the sort of typological study embarked upon by Zhu (1985) is more daunting than it might at first appear. Specifically, Cole and Lee’s evidence shows that identifying what qualifies as an A-not-A question in the relevant sense is not straightforward from the surface forms and requires more detailed investigation into the distributional properties within a given language/dialect. Cheng, Huang, and Tang (1996) explore the properties of NPQs across Mandarin, Taiwanese, and Cantonese. In the discussion above (section 6), the conclusion for Mandarin was that NPQs and VP-not-V questions appear to result from movement of NQ into a clause-final position we called npqP. The movement analysis was primarily supported by the long-distance aspect agreement shown on the negative morpheme. Neither Taiwanese nor Cantonese, however, show the same kind of long-distance agreement in NPQs, despite the fact that there is agreement in both languages between clause-internal (in situ) negation and aspect. In Taiwanese, the negative morpheme used in NPQs is always the same (mei), and cannot be used in A-not-A questions. In Cantonese, on the other hand, NPQs can freely use any of the negative morphemes without regard to the aspect of the clause. Cheng et al. conclude from this lack of agreement that in Taiwanese and Cantonese, NPQs do not involve movement to a clause-final npqP position, but instead simply have a negative particle base-generated in its surface position. 9. Closing remarks In the sections above, we have reviewed a number of the characteristics of several kinds of questions that have fallen under the label “A-not-A question”, and most of their major distributional characteristics as they have been described in the generative literature. For the most part, we have steered clear of minute syntactic analysis (e.g., debates about whether a [Neg] feature or a [+Q] feature is attracted to a functional head Σ in overt syntax), but the broad characteristics that such an analysis must have and the facts it is responsible for should nevertheless be clear. To review, it is clear that there are a number of different types of questions to consider. Alternative questions, which appear in Mandarin with an overt disjunctive word (haishi), are almost completely unconstrained, while questions formed by reduplication of a predicate in situ (A-not-A questions) are rather tightly constrained. Both kinds of questions are distinct from true yes-no questions, formed in Mandarin with the sentence-final question particle ma, as is clear from the kinds of answers they allow, the contexts they are allowed in, and from the distinctness of the syntactic constraints on ma questions and A-not-A questions. The VP-not-V questions and NPQs are not clearly different from one another, but have some properties that distinguish them from A-not-A questions, even if they may be derived from an underlying form that is very similar to A-notA questions. Crosslinguistically, the existence of a question type parallel to Mandarin Anot-A questions is relatively well-attested, e.g., the kam-questions in Taiwanese, but we also saw that not all questions with an apparent A-not-A form should properly be considered analogs to Mandarin A-not-A questions. There is much more to say on the cross-linguistic implications and parallels, as well as on language-internal implications, some of which exists in the literature already but not covered here, but much of which remains to be systematically explored. Our understanding of this family of “yes-no-like” questions has advanced a great deal over the last twenty years, and the field is now ready for more systematic, broad-ranging, and deeper cross-linguistic comparison. Also important for future theoretical understanding will be a concentration on points where judgments differ; as noted in a few places above, there are conflicting reports in the literature on certain crucial data points, some of which may be cleared up by more careful control of regional dialects and registers. One area that remains only relatively cursorily explored in the literature to date is the (admittedly difficult) determination of the pragmatic properties of the different types of questions: what can and cannot be focused, whether a certain type of question can be used in a pragmatically “slanted” context. Such facts are increasingly widely assumed to have structural correlates (e.g., movement to a functional “focus phrase” projection) and are at least potentially a fairly rich source of data.