Aspectual selection and negation in Mandarin Chinese

(1)

Mandarin Chinese*

JO-WANG LIN

Abstract

This paper argues that the distributions of bu ‘not’ and mei ‘not’ in Mandarin Chinese can be accounted for in terms of their aspectual selec-tions. I propose that while mei aspectually selects an event as its comple-ment, bu aspectually selects as its complement a stative situation that requires no input of energy in order to obtain that situation. I show that this approach to the distributions of bu and mei better accounts for the data than previous analyses such as Huang (1988), Ernst (1995), and Lee and Pan (2001).

1. Introduction

Bu ‘not’ and mei ‘not’, both translated as English not, are the two most frequently used negation markers in Mandarin Chinese. Although they sometimes may appear in the same linguistic contexts, they are mostly in complementary distribution. (1) and (2) are two sets of examples that illustrate their basic usages.

(1) a. Ta bu qu shangxue he not go school

‘He does not want to go to school./He will not go to school.’ b. Ta mei qu shangxue

he not go school

‘He did not go to school.’ c. Wo bu chou yan

I not smoke cigarette. ‘I don’t smoke.’

d. Wo mei chou yan I not smoke cigarette ‘I didn’t smoke.’

Linguistics 41–3 (2003), 425–459 0024–3949/03/0041–0425 ©Walter de Gruyter

Brought to you by | National Chiao-Tung University Authenticated | 140.113.38.11

(2)

(2) a. Ta mei/*bu nong-dong zhe-ge lilun he not/not make-understand this-CL theory ‘He hasn’t understood this theory.’

b. Ta bu/*mei dong zhe-ge lilun he not/not understand this-CL theory ‘He does not understand this theory.’

Special attention has been paid to two central facts about the above two negation markers. One is the inability of bu to occur with the perfective aspect marker le and the other is the incompatibility between bu and a postverbal de phrase (Huang 1988; Ernst 1995; Lee and Pan 2001). The two facts are illustrated by (3) and (4).

(3) a. Wo chi-le pingguo

I eat-ASP apple ‘I ate apples.’ b. Wo bu chi pingguo

I not eat apple ‘I do not eat apples.’ c. *Wo bu chi-le pingguo

I not eat-ASP apples (4) a. Ta pao de hen kuai

he run DE very fast ‘He runs very fast.’ b. Ta pao de bu kuai

he run DE not fast ‘He does not run fast.’ c. *Ta bu pao de kuai

he not run DE fast ‘He does not run fast.’

Several attempts have been made to account for the above two facts in the literature. Huang (1988) suggests that bu is a bound morpheme in Infl that forms an immediate construction with the first V0 element. Thus, when bu is cliticized onto the following verb, the de phrase, which is some kind of manner expression, takes scope over it. However, this leads to a semantic anomaly because a nonevent cannot be done in some manner. The same analysis also accounts for the inability of bu to modify a verb with the perfective marker le: a nonevent cannot be perfective.

Ernst (1995), on the other hand, argues that bu is not a clitic attached to verbs but is a proclitic on the following adjacent word. Furthermore, it aspectually requires an unbounded situation. According to him, the incompatibility between le and bu results from a conflict between the

(3)

unboundedness requirement of bu and the boundedness requirement of le. On the other hand, by assuming that a de phrase is base-generated at a preverbal adjunct position and is later moved to the postverbal position, he argues that the XP trace left by the de phrase blocks cliticization of bu onto a lexical word.

Still another diﬀerent account for the same facts is proposed by Lee and Pan (2001), who argue that bu is not a clitic element but a focus-sensitive unselective binder. Roughly, they argue that bu is incompatible with le, assumed to be a selective binder of an event variable, because le, having scope over bu, has no free variable to bind, the event variable being already bound by bu. Their account for the incompatibility between bu and a de phrase is along the same line of reasoning.

From the above brief summary of the different existing accounts for the distribution of bu, it is clear that there is still no consensus among Chinese linguists concerning how to analyze bu and its syntactic distribu-tion. The lack of consensus arises not only because different authors have made different assumptions for bu but also because some authors — for example, Lee and Pan (2001) — even question the correctness of the cooccurrence constraint between bu and le on the one hand and bu and the de phrase on the other hand. The different assumptions made in the above-mentioned analyses of bu are summarized below:

(5) a. Bu is a bound morpheme or clitic element (Huang 1988; Ernst 1995).

b. Bu aspectually selects a certain type of situation as its complement ( Ernst 1995).

c. Bu is an unselective binder for free variables (Lee and Pan 2001).

d. Bu is a focus-sensitive element that can be associated with a variety of constituents (Lee and Pan 2001).

Among the four assumptions in (5), the assumption (5d ) is possibly one that most Chinese linguists would agree with even if they may not explicitly say so. I will not question this assumption, either. But for the other three assumptions, questions can be asked as to whether a combina-tion of two or more assumpcombina-tions is really needed in order to successfully account for the distribution of bu. It seems quite possible that one single assumption alone might be suﬃcient to cover all relevant examples. Of course, there is no a priori answer to the question of which assumption is correct. Choice of the diﬀerent assumptions is a pure empirical issue. Indeed, the main purpose of this paper is to reexamine the distribution of bu and to argue that the assumption (5b) alone is able to explain a wide range of data, including those already observed in the literature,

(4)

those cited as counter-examples to the two central facts about bu, and many other new data that are not discussed in the literature and that give rise to problems with the above-mentioned analyses of bu.

The remaining sections of this paper are organized as follows. Section 2 discusses the aspectual requirement of bu and section 3 the aspectual requirement of mei, as well as the inability of bu to occur with the perfective aspect marker le. Section 4 deals with the problem of the incompatibility between bu and postverbal de phrases. Section 5 discusses the problems faced by previous analyses that are avoided by the aspectual restrictions proposed in this paper. Section 6 concludes this article.

2. Aspectual selection of bu

Although like Ernst (1995) I will utilize aspectual requirement to account for the distribution of bu, the content of my proposal is diﬀerent from his original analysis. The aspectual selections of bu and mei that I will adopt are the following:

(6) Aspectual selections of bu and mei

a. Bu aspectually selects as its complement a stative situation that requires no input of energy in order to obtain that situation. b. Mei aspectually selects an event as its complement.

Before showing how the distinction between states vs. nonstates (events) is a better notion to characterize the licensing condition on the use of bu, let me first briefly review the notions of states and nonstates (events). It is now well known that eventualities can be subdivided into states and nonstates (or dynamic events), and many finer distinctions can be drawn from the two categories. Here, I will only focus on the states vs. nonstates distinction. A consensus about the characteristics of states that linguists seem to generally agree on is that states do not develop or change in time. They simply obtain in time and no conscious eﬀort is required in order for the states they describe to obtain or go on obtaining (cf. Comrie 1976; Smith 1991; among many others). In contrast to states, nonstates or dynamic situations involve change over time and require input of energy. Dynamic situations can be further divided into processes and events (accomplishments or achievements) (Bach 1986; Vendler 1967; among others). It is the above sense of states vs. nonstates that I will adopt in what follows in this paper.

Given the above definition of states and nonstates, let us now see what the assumption in (6a) can buy for us. To begin with, this assumption straightforwardly accounts for familiar examples in which bu negates

(5)

stative adjectives, verbs, or modal auxiliary verbs. Sentences with these words are typical stative situations that require no input of energy. Some illustrating examples are given below.1

(7) a. Ta bu laoshi he not honest ‘He is not honest.’

b. Wo bu ai ni

I not love you ‘I don’t love you.’

c. Ta bu hui/yinggai/keyi shuo huang he not will/should/may tell lie ‘He does/should/may not tell lies.’

It is worth noting that some examples that do not contain an overt stative modal may get a volitional or future modality reading and allow bu to appear. I assume with Huang (1988) that such sentences actually contain an empty volitional or future modal verb, and what bu negates in such sentences are the empty modal verbs. So examples like (8) actually obey the proposed aspectual restriction on bu.2

(8) a. Wo bu qu Taibei I not go Taipei

‘I do not want to go to Taipei/I will not go to Taipei.’ b. Ta bu bang wo xie zuoye

he not help me write homework

‘He does not want to help me with my homework.’

Having seen how the proposed aspectual restriction accounts for some basic data, I turn now to some more troublesome examples.

It has been sometimes claimed or assumed that progressive sentences such as He is writing a book are statives ( Vlach 1981, 1993; Mufwene 1984; Saurer 1984; Dowty 1986; Langacker 1987; Parsons 1990; Kamp and Reyle 1993). If progressive sentences describe true states like other typical stative sentences, then the proposed aspectual requirement predicts that bu should be able to appear in such sentences. Interestingly, the fact seems not to be the case. The following sentences are ungrammatical.3

(9) a. *Ta bu zai xi zao

he not PROG take bath ‘He is not taking a bath.’

b. *Ta bu zai gai fangzi

he not PROG build house ‘He is not building a house.’

(6)

The negation marker for progressive sentences is mei, not bu.

(10) Ta mei zai xi zao

he not PROG take bath ‘He is not taking a bath.’

Examples such as those in (9) thus seem to constitute a problem for the proposed aspectual restriction for bu.

Fortunately, not every linguist agrees that progressive sentences are stative sentences. Counterarguments have been produced by Smith (1991), Bertinetto (1994), and Glasbey (1998). In particular, Bertinetto (1994) has a very detailed review of the pros and cons and rejects treating progressive sentences as statives. In his view, progressive sentences describe dynamic situations that require input of energy in order to obtain the states they describe. Indeed, this is also the view put forward by Nordlander (1997), who explicitly says that the situation described by a progressive sentence constitutes ‘‘a number of consecutive phases’’ that need ‘‘an ongoing supply of energy.’’ I agree with Bertinetto (1994) and Nordlander (1997) that progressive sentences are not truly statives in the same sense as sentences such as He is clever, which requires no input of energy at all in order to obtain the situation. I also agree with Nordlander (1997) and Comrie (1976) that such situations are processes. If a situation described by a progressive sentence needs input of energy in order to obtain that situation, then it is predicted by the proposed aspectual selection of bu that such a sentence should be incompatible with bu. In fact, the Chinese sentence (10) can be taken as evidence for the position that progressive sentences are dynamic rather than stative as Bertinetto (1994) and Nordlander (1997) have argued.

A second type of example that might challenge the proposed aspectual restriction for bu has to do with locative-inversion sentences like (11), where the aspectual marker zhe is attached to the verb.

(11) Qiang shang gua-zhe yi fu hua wall on hang-ASP one CL picture ‘On the wall is hanging a picture.’

It has been argued that zhe is an imperfective durative aspectual marker or a stativizer that presents a continuous and stable situation without endpoints (Smith 1991; Jin 1991; Yuan 1993; Zhang 1996). According to Smith (1991), the basic meaning of zhe describes a resultative state. Thus, (11) describes the resultative state of a picture being hung on the wall. But such sentences are incompatible with bu. The negation marker for such sentences can only be mei. This is shown by the contrast between (12a) and (12b).

(7)

(12) a. *Qiang shang bu gua-zhe yi-fu hua wall on not hang-ASP one-CL picture ‘There is not a picture hanging on the wall.’ b. Qiang shang mei gua-zhe yi fu hua

wall on not hang-ASP one CL picture ‘There is not a picture hanging on the wall.’

Why are examples like those in (12) ungrammatical? Isn’t it true that examples like them describe states? To answer these questions, I would like to first point out two more observations about the interaction between bu and sentences with the aspectual marker zhe. The first observation is that not just locative-inversion sentences but many other non–locative-inversion sentences with zhe are incompatible with bu. This is illustrated by (13) and (14).

(13) a. Ta tui-zhe yi-liang jiaotache he push-ASP one-CL bicycle

‘He is (in the state of ) pushing a bicycle.’ b. Ta mei tui-zhe yi-liang jiaotache

he not push-ASP one-CL bicycle ‘He is not pushing a bicycle.’ c. *Ta bu tui-zhe yi-liang jiaotache

he not push-ASP one-CL bicycle ‘He is not pushing a bicycle.’ (14) a. Ta zai chuang-shang tang-zhe

he on bed-above lie-ASP ‘He is lying on the bed.’

b. Ta mei zai chuang-shang tang-zhe he not on bed-above lie-ASP ‘He is not lying on the bed.’ c. *Ta bu zai chuang-shang tang-zhe4

he not on bed-above lie-ASP ‘He is not lying on the bed.’

The second observation is that sentences containing the aspectual marker zhe are not always incompatible with bu. Illustrated below are two examples in which bu cooccurs with zhe.

(15) a. Women zhijian bu cunzai-zhe nimen suo shuo de

we in-between not exist-ASP you PAR say REL

wenti problem

lit. ‘In between us does not exist the problems that you said.’ b. Zhe liang dang de zhenghe bing bu yiwei-zhe

this two party DE integration therefore not mean-ASP

(8)

xunju tamen jiu hui ying election they then will win

‘The integration of these two parties does not therefore mean that they will win in the election.’

Returning to the question of why bu is not allowed in locative-inversion sentences and certain sentences with the aspectual marker zhe, consider (13a) first. This sentence clearly describes a situation that needs a con-tinuous input of energy in order for the situation to go on obtaining. In fact, traditional Chinese linguists call zhe as it occurs in (13a) dynamic zhe. Thus, according to the proposed aspectual selection of bu, (13a) should be ill-formed.

Next, consider (14a), which is an example of positional sentences. It might be correct to say that such sentences describe positional states. However, in order for a positional state to obtain, there should be an input of energy first. In the case of (14a), for example, one has to first do the action of lying before the state of lying can eventually obtain. Thus, even if positional sentences describe states, their nature is still diﬀerent from that of sentences such as He is clever, which needs no initial conscious eﬀort and energy in order to obtain the state.

As for the examples in (15), they contrast with (13a), (14a), and progressive sentences precisely because the former need no input of energy in order to obtain the states they describe. In these examples, the verbs are truly statives and the states they describe obtain in time rather than develop over time. They are therefore predicted to be compatible with bu even with the presence of zhe.

Locative-inversion sentences such as those in (11) and (12) are quite similar to positional sentences. The states they describe are result states of events. Without the action part of the event, there is simply no result state at all. Apparently, just like positional sentences, locative-inversion sentences require initial input of energy in order to obtain the intended state. Perhaps we can call such states derived states. Such states describe one phase of a situation and thus they are diﬀerent from truly stative situations, which do not involve change in time. Thus, rather than falsify-ing the proposed aspectual selection of bu, the interaction between bu and locative-inversion sentences actually lends support to it.

The final set of examples in support of the proposed aspectual require-ment of bu is generic or habitual sentences indicating permanent states or truth, as illustrated by the following examples.

(16) a. Wo bu chou yan

I not smoke cigarette ‘I do not smoke.’

(9)

b. Mao bu chi zhurou cat not eat pig-meat ‘Cats don’t eat pork.’

c. Taiyang bu rao diqiu xuanzhuan sun not around earth turn-around ‘The sun does not turn around the earth.’

If bu in such sentences is replaced by mei, the interpretation shifts to a single episode interpretation. For example, in contrast to (16a), wo mei chou yan is interpreted as ‘I didn’t smoke’, that is, the eventuality denoted by I smoke did not take place.

Following Bertinetto (1994), I will refer to sentences like those in (16) as ‘‘attitudinals’’ and distinguish them from habitual sentences. Like progressive sentences, generic or habitual sentences are often claimed to be statives, by scholars such as Leech (1971), Partee (1984: 270), Mufwene (1984: 31–33), Chung and Timberlake (1985: 215), Mittwoch (1988), and Rot (1987). Bertinetto (1994), however, argues that this suggestion is only partially correct. According to him, only a subset of habitual sentences, those that he calls ‘‘attitudinals’’ (or more precisely ‘‘attitudinal habituals’’) are truly stative. By attitudinals, he refers to ‘‘situations in which the repeated occurrence of a given event is taken to be the defining property of an individual or object.’’ Such situations consist of activities that correspond to permanent attributes of that individual or object. Attitudinal predicates are thus much like individual-level predicates, which characterize stable rather than transient properties. Attitudinal sentences can be illustrated by sentences such as John sings, which refers to John’s profession rather than his habit. In contrast, pure habituals, which can be illustrated by John often sings, are not purely stative situations but ‘‘may easily have a strictly eventive character’’ (Bertinetto 1994). Arguments for the distinction between pure habituals and attitudinals include the ability to be modified by agentive adverbs, the possibility of anchoring the situation to a single moment, and the possibility to occur in perceptual reports. I refer readers to Bertinetto (1994) for detailed argumentation.

One point that should be clarified is in order, with regard to the level to which the property of being eventive or stative is assigned. Bertinetto seems to have taken a habitual or an attitudinal sentence as a whole as eventive or stative. However, it does not seem unreasonable to claim that in a habitual sentence only the predicative VP is eventive while the sentence as a whole is stative, as the linguists mentioned above have argued. As far as I can see, this view is fully compatible with Bertinetto’s arguments and in fact I will take this position in what follows. Empirical arguments for this will be given later.

(10)

If Bertinetto’s distinction between attitudinals and pure habituals is correct, an interesting prediction by the aspectual constraint on bu is that only attitudinals but not pure habituals may be negated by bu. Is this prediction borne out? The answer seems to be yes. First, consider (17). (17) a. Ta bu chou yan

he not smoke cigarette ‘He does not smoke.’ b. Ta mei chou yan

he not smoke cigarette ‘He did not smoke.’

(17a) has an attitudinal reading. Therefore, the (negative) property denoted by the (negative) predicate is contextually turned into a perma-nent stative.5 Consequently, the negation marker is bu rather than mei. In contrast, with the negation marker mei, (17b) must describe an episode and the predicate here is eventive in character.

Now let us consider sentences in which a frequency adverb is combined with an activity, that is, habitual sentences. According to Bertinetto, such sentences describe pure habits and the VPs in such examples are eventive rather than stative. This then predicts that the negation marker in this type of sentence should be mei rather than bu. Very interestingly, both bu and mei may occur in habitual sentences. This is illustrated by (18). (18) a. Xiaoming chang(chang) bu xi zao

Xiaoming often not wash bath

‘It is often the case that Xiaoming does not want to take a bath.’

b. Xiaoming chang(chang) mei xi zao6

Xiaoming often not take bath

‘It is often the case that Xiaoming does not take a bath.’ However, the truth conditions of the above two sentences are diﬀerent. For (18a) to be true, it must be the case that in each occasion where Xiaoming does not take a bath, he purposefully refuses to do it. But for (18b) to be true, the will of Xiaoming is not necessarily involved. Imagine that Xiaoming is a kid and needs help from his mother with bathing. But his mother is often too busy to do it for him. Therefore, he often does not take a bath. In such a scenario, only (18b) is felicitous but not (18a).

With the above characterization of the semantic diﬀerence between (18a) and (18b) in mind, the use of bu and mei in these two sentences can now be accounted for as follows. (18b) is grammatical because the VP in a pure habitual sentence is eventive and hence must be negated by

(11)

mei. The semantics of this sentence only says that an event of the sort as denoted by Xiaoming xi zao ‘Xiaoming takes a bath’ does not often occur; it does not claim anything about Xiaoming’s volition. In contrast, the interpretation of (18a) involves the subject’s volition. As mentioned, I assume with Huang (1988) that such sentences involve an empty voli-tional modal. Thus what bu really negates in (18a) is the empty modal. Since modal verbs are specified as [+stative], bu is used in (18a).

Regarding the interaction between negation markers and frequency adverbs, it is very interesting to point out that in addition to appearing after the frequency adverb chang(chang) ‘often’, bu and mei may also be placed before it, as the examples in (19) illustrate.

(19) a. Xiaoming bu chang xi zao Xiaoming not often wash bath

‘It is not the case that Xiaoming often takes a bath.’ b. ?Xiaoming mei chang xi zao

Xiaoming not often take bath

‘It is not the case that Xiaoming often takes a bath.’ c. Xiaoming mei you chang xi zao

Xiaoming not have often take bath

‘It is not the case that Xiaoming often takes a bath.’ d. Xiaoming you chang xi zao ma?7

Xiaoming have often take bath Q ‘Does Xiaoming often take a bath?’

Unlike the contrast between (18a) and (18b), it is more diﬃcult to tell what truth-conditional diﬀerences are between (19a) and (19b). Both (19a) and (19b) can be used in nonvolitional readings. But syntactically, (19b) seems to be somewhat worse than (19a) unless the existential verb you ‘have’ is inserted as in (19c). The grammaticality of (19a) is expected if an habitual sentence as a whole is stative, as many linguists have argued, though the VP part is eventive, as mentioned earlier. As for (19b), I suspect that (19b) is well-formed because an empty you is in the construction. As mentioned, the verb you is always negated by mei rather than by bu, regardless of the fact that you is a stative verb. One possible piece of evidence for this analysis comes from consideration of the ques-tion (19d ), in which the existential verb you is used. (19b) is a very appropriate answer to (19d ). Another alternative analysis might be to say that pure habitual sentences are ambiguous between stative and eventive readings. This alternative needs further examination. I will not try to decide which alternative better accounts for (19b).

I conclude that the distribution of bu in generic sentences, including attitudinal and habitual sentences, supports the idea that bu selects a stative situation that requires no input of energy.

(12)

3. Aspectual selection of mei

The distribution of mei is closely related to the perfective aspectual marker le. Wang (1965) has proposed that le and you, which is optional in the presence of mei, are allomorphs of the same perfective morpheme. According to him, when the perfective morpheme is realized as you, bu changes to mei by a special morphological rule. Therefore, you and le are in complementary distribution.

(20) a. Ta mai-le fangzi he buy-ASP house ‘He bought a house.’ b. Ta mei (you) mai fangzi

he not have buy house ‘He did not buy a house.’ c. *Ta mei (you) mai-le fangzi

he not have buy-ASP house ‘He did not buy a house.’

I will not further pursue the issue of whether or not you and le are allomorphs, but it is clear that understanding the distribution of le will be very helpful in sorting out the interaction between le, mei, and bu.

The literature on the aspectual marker le is too rich to give a complete survey here (Liu 1988; Shi 1990; Heinz 1990; Lin 2000; Klein et al. 2000; among many others).8 What I will do is briefly summarize the situation types that occur with le and indicate what the resulting interpretation is. According to previous research, except for a small set of adjectives and verbs such as shuyu ‘belong’, xiang ‘resemble’ le may occur with most verbs and adjectives.9 Putting that small set of verbs and adjectives aside, when le occurs with a stative predicate, the sentence gets an inchoative reading. This is illustrated by (21).10

(21) a. Zhe ke mugua shou-le this CL papaya ripe ASP ‘This papaya has become ripe.’ b. Ta pang-le

he fat ASP

‘He has gotten fat.’

In other words, the function of le in (21) is to indicate a change of state. Some authors have treated inchoativity as an achievement. I will follow this treatment (cf. de Swart 1998; Heinz 1990).

When le occurs with an achievement, accomplishment, or activity situation, the sentence denotes a completed or terminated event. This is illustrated by the examples in (22).

(13)

(22) a. Ta ying-le na-chang qiu he win-ASP that-CL ball ‘He won that ball game.’ b. Ta gai-le yi-dong fanzi

he build-ASP one-CL house ‘He built a house.’

c. Ta he-le jiu, ye chang-le ge he drink-ASP wine also sing-ASP song ‘He drank alcohol and sang songs.’

If inchoativity is indeed a kind of achievement, then the two sets of examples in (21) and (22) indicate that the perfective aspectual marker le selects a bounded event as its complement. Indeed, this is exactly what Li and Thompson (1981) propose.

Returning to mei, it seems that the semantics of mei is the opposite of le. When mei occurs with a stative predicate, it indicates that a change of state does not occur. For example, the adjective lao ‘old’ can be negated by both bu and mei.

(23) Ta kan-shangqu yi dian dou bu lao he look-appear one little all not old ‘He is not old at all in appearance.’ (24) Ta kan-shangqu yi dian dou mei lao

he look-appear one little all not old ‘He hasn’t become old at all in appearance.’

When the negator is bu, the sentence means that the subject NP does not have the property denoted by the stative predicate. But when the negator is mei, the focus is not on lack of the property denoted by the stative predicate but on whether or not there is a change of state described by the stative predicate. Because of this, stable properties such as those denoted by individual-level predicates are not able to be negated by mei, as is shown by (25).

(25) *Ta mei congming he not clever

‘He has not turned clever.’

Very similarly, when mei occurs with an achievement, accomplishment, or activity situation, it indicates that an event of the relevant sort did not occur. This is illustrated by (26).

(14)

(26) a. Ta mei ying na chang qiu he not win that CL ball ‘He didn’t win that ball game.’ b. Ta mei gai yi dong fanzi

he not build one CL house ‘He didn’t build a house.’

c. Ta mei he jiu, ye mei chang ge he not drink wine also not sing song

‘He didn’t drink alcohol nor did he sing songs.’

If a change of state represents an achievement, as mentioned, then the following generalization can be reached: mei is the negation marker of nonexistence or nonrealization of an event. In other words, it is the negative counterpart of the perfective marker le. Thus, just like le, mei also aspectually selects an event as its complement.

With the above discussion of the aspectual requirements of le and mei, it should now come as no surprise that bu is incompatible with le. Bu aspectually selects a situation type as its complement that has no inherent end point — states are situations that have no inherent end boundary, whereas le selects a situation type that is dynamic and bounded — that is, events. Thus, the aspectual requirements of these two elements are in conflict with each other: states are not perfective events and vice versa. This explains one of the two central facts about bu widely discussed in the literature.

4. Postverbal de phrases and the distribution of bu

Having shown how the proposed aspectual selections of bu and mei account for a wide range of data, including the inability of bu to occur with the perfective marker le, I now turn to the other central fact about the distribution of bu and mei, namely, the incompatibility between bu and a postverbal de phrase.

In order for non-native speakers to have a better understanding of the Chinese de phrases, I begin with a brief discussion of verb–predicative complement constructions ( VPC constructions in short). Traditionally, Chinese VPC constructions are divided into two types, according to whether the verb is immediately followed by the predicative complement, as in (27a), or the predicative complement is separated from the verb by the marker de, as in (27b).

(15)

(27) a. Zhe jian jiaoshi dasao ganjing le this CL classroom sweep clean ASP ‘This classroom has been swept clean.’ b. Zhe jian jiaoshi dasao de hen ganjing

this CL classroom sweep DE very clean

‘This classroom has been swept such that it is very clean.’ Examples like (27a) are often referred to as verb-resultative constructions. As for constructions with the marker de, they are often subdivided into several types:

A. Potential complements (de is infixed to a verb–verb/adjective compound)

(28) a. Ta kan-de-dong zhe ben shu

he read-DE-understand this CL book ‘He can understand this book.’

b. Ta kan-bu-dong zhe ben shu

he read-DE-understand this CL book ‘He cannot understand this book.’

In (28), kang-dong ‘read-understand’ is a compound verb. The morpheme de is infixed to this compound verb to yield the potential reading. The negative counterparts of such constructions are derived by replacing de with the negation marker bu as is illustrated by (28b). Potential de complements do not allow any constituent to be inserted in between de and the resultative verb.

B. Resultative complements

Resultative de complements do not always take the form of VP or AP but can be a full clause. The negation marker for such constructions is mei, which is placed before the whole verb–de–XP rather than inside the de complement.11 This is illustrated by (29) and (30).

(29) a. Ta wan de hen lei he play DE very tired ‘He played till he got tired.’ b. *Ta wan de mei/bu (hen) lei

he play DE not very tired ‘He played till he didn’t get tired.’

(16)

c. *Ta bu wan de (hen) lei he not play DE very tired

‘He does/did not play till he get tired.’ d. Ta mei wan de hen lei

he not play DE very tired

‘It is not the case that he played till he got tired.’

(30) a. Ta qi de nian-se fa bai

he angry DE face-color turn white

‘He got so angry that his face turned pale.’

b. Ta mei qi de nianse fa bai

he not angry DE face-color turn white

‘It is not the case that he got so angry that his face turned pale.’

c. *Ta qi de nianse mei/bu fa bai

he angry DE face-color not turn white ‘He got so angry that his face did not turn pale.’

d. *Ta bu qi de nianse fa bai

he not angry DE face-color turn white

‘It is not the case that he got so angry that his face turned pale.’

C. Descriptive (or extent) complements

Descriptive complements are semantically very similar to manner adjuncts.12 In fact, Ernst (1995) has treated them as adjuncts. Sentences with a descriptive complement are usually ambiguous between a (past) episode and a generic reading. This is illustrated by (31).

(31) Ta pao de hen kuai he run DE very fast i. ‘He runs very fast.’ ii. ‘He ran very fast.’

Which negation marker is used in such constructions depends upon which reading is intended. On the generic reading, the negation marker can only be bu, and it must appear within the de complement rather than before the verb–de–XP constituent. On the other hand, if the episode reading is intended, both mei and bu can be used, but they appear in diﬀerent positions. Mei appears before the whole VP phrase, whereas bu must appear within the de complement. The distribution of the negation markers is illustrated by the following examples.

(17)

(32) a. Ta pao de bu (hen) kuai he run DE not very fast i. ‘He did not run very fast.’ ii. ‘He does not run (very) fast.’ b. *Ta bu pao de (hen) kuai

he not run DE very fast i. ‘He does not run very fast.’ ii. ‘He did not run very fast.’ c. Ta mei pao de hen kuai

he not run DE very fast ‘He did not run very fast.’

Returning to the central question of why bu is not licensed before a VP consisting of verb–de–XP, it is of great help to first clarify what the real issue is in such constructions. To start with, I assume that potential complements are frozen expressions, which should be dealt with probably in morphology rather than in syntax. So in what follows, I will ignore potential complements and focus instead on resultative and descriptive complements, which have received much attention in the literature (see Huang 1988 and references cited there).

As we have seen above, if the de phrase is a resultative complement, the negation marker can only be mei, and it must appear before the whole verb–de–XP constituent rather than inside the de–XP complement. This fact has an easy explanation. As the name ‘‘resultative’’ suggests, the de–XP complement describes a resultative state of the action (event) denoted by the (main) verb. Thus, situations denoted by such construc-tions must be dynamic. Consequently, the negation marker can only be mei or the aspectual constraint of bu would be violated. Also, because the function of a resultative de complement is to describe the result state of an event, it is pragmatically less plausible to say that the result state does not exist. Therefore, the ungrammaticality of examples such as (29b) can be ascribed to semantic incoherence.

As for descriptive complements, the fact that bu is not licensed before the whole VP phrase on the episode reading is predicted by the aspectual selection of bu, because on this reading, bu is associated with a dynamic event rather than a stable state. What is more surprising and curious is the fact that bu is not licensed on the generic reading. This question is obviously related to the question of whether the predicates in generic situations are permanent statives like those in attitudinals or are eventive like those in pure habits. To this I turn my attention in what follows.

Recall that attitudinals refer to activities that are contextually turned into permanent statives. An important property of permanent statives is

(18)

that they allow no exception in the sense that the stative property must always hold of an individual or object irrespective of time and space. For example, if a man is tall, he is tall every time you see him. But this is not the case for generics. It is well known that generics allow exceptions. Thus, a man can be claimed to have the property of generally running fast even though he fails to do so in several occasions. Thus, activities associated with generic situations must not represent permanent statives or a defining property of an individual. This view is supported by the fact that, unlike permanent statives, generic sentences with a de comple-ment can be combined easily with temporal modifiers such as zongshi ‘always’ or jingchang ‘often’ as pure habits can:

(33) Ta zongshi/jingchang pao de hen kuai he always/often run DE very fast ‘He always/often runs very fast.’

Another diﬀerence between permanent statives and generic sentences is this. Permanent statives, as Carlson (1982) has noticed, have even tighter restrictions on their cooccurrence with locative modifiers than temporal modifiers. Thus, it is very odd to say the sentence

(34) ??John is intelligent in France but is not in America.

But sentences with a descriptive de complement have no such restrictions. Compare (35) with (34).

(35) Zhangsan zai guo nei pao de hen kuai, keshi bu zhidao Zhangsan in country inside run DE very fast but not know weishenme laidao guowai bisai de-shihou pao de zheme man

why come abroad contest when run DE this slow

‘Zhangsan runs very fast in his own country but it is unknown why he runs so slow when he comes abroad to attend the contest.’ The above evidence shows that though sentences with a descriptive de complement express a property and never report a specific event, the property is only a generalization over events that is not contextually turned into a permanent stative. In view of this, it is not unreasonable to say that, unlike eventive predicates in attitudinals, which are turned into permanent statives through the loss of the event component by a special type of conversion, the eventive predicates in sentences with descriptive de complements do not undergo such a conversion.

It is important to note that by the above analysis I am not by any means claiming that generic sentences are not statives. I only intend to claim that the predicates in such sentences are eventive, but the whole sentences themselves can be stative, as I have indicated earlier. More

(19)

precisely, I propose that generic sentences become stative only after their event variables are quantified over by the generic operator. Before the closing of the event variables, the predicates themselves are eventive rather than stative. If this suggestion is correct, we now have a straightfor-ward answer to the question of why bu may not modify a [VP verb–de–XP] constituent: its aspectual constraint cannot be respected in such environments, because the VP is eventive in character.13

To make my above argumentation clearer, I will make the following assumptions. I assume that the Chinese phrase structure contains a TP projection whose head specifies tense information. Moreover, the generic operator is located in tense just as are other kinds of tense operator. Temporal modifiers such as zongshi ‘always’ are treated as the specifier of TP and the negation marker bu heads NegP, which is located above VP and below TP by default. But NegP may also occupy a position above TP when there is overt evidence that forces it to do so.

On the above assumptions, the impossibility for bu to modify the VPs in generic sentences with a descriptive complement is explained schematically as follows:

(36) a. _{*[AgrP Ta [TP[+stative] [NegP bu [VP[−stative] pao de hen}

he not run DE very

kuai ] ] ] fast

b. _{[AgrP Ta [TP[+stative] [VP[−stative] pao de [NegP bu [AP[+stative] hen}

he run not very

kuai ] ] ] ] ] fast

c. _{[AgrP Ta [TP[+stative] [NegP mei [VP[−stative] pao de hen kuai]]]]}

he not run DE very fast

In (36a), the complement of bu has the feature [−stative] rather than [+stative]; therefore its selectional restriction is not satisfied. So (36a) is ill-formed. In contrast, the selectional restrictions of bu and mei in (36b) and (36c) are satisfied. Therefore, these two examples are well-formed.

The examples in (36) do not contain a temporal modifier. Now let us also consider what happens when a temporal quantifier appears. Interestingly, grammatical judgments of negative sentences with a tempo-ral modifier and a postverbal de complement depend upon the relative order between the negation marker and the temporal modifier. When the temporal modifier appears before the negation marker, the sentence is ill-formed; but when the former follows the latter, the sentence is well-formed. The situation here is very similar to what I discussed in section 2.

(20)

(37) a. *Ta chang/zongshi bu pao de hen kuai he often/always not run DE very fast

‘It is often/always the case that he does not run very fast.’ b. Ta bu chang/zongshi pao de hen kuai

he not often/always run DE very fast

‘It is not often/always the case that he runs very fast.’ The contrast between (37a) and (37b) is predicted by the analysis that I am proposing. The schematic representations of the above two sentences are the following:

(38) a. _{*[AgrP Ta [TP[+stative] zongshi [NegP bu [[−stative] pao de hen} kuai ] ] ] ]

b. _{[AgrP Ta [NegP bu [TP[+stative] zongshi [[−stative] pao de hen} kuai ] ] ] ]

In fact, the same account may also be extended to my earlier discussion of progressive and locative-inversion sentences. Recall that progressive and locative-inversion sentences do not license bu. Interestingly, the gram-matical judgments are not changed even if a temporal modifier such as zongshi ‘always’ is inserted before bu forcing the whole sentence to be interpreted generically.

(39) *Ta zongshi bu zai dushu he always not PROG study

‘It is always the case that he is not studying.’ (40) *Qiang shang zongshi bu gua-zhe yi fu fa

wall on always not hang-ASP one CL picture

‘It is always the case that there is not a picture hanging on the wall.’

However, like the examples in (37), if the relative order between the temporal modifier and bu is reversed, the sentences become well-formed: (41) Ta bu zongshi zai dushu

he not always PROG study

‘It is not always the case that he is studying.’ (42) Qiang shang bu zongshi gua-zhe yi fu fa

wall on not always hang-ASP one CL picture

‘It is not always the case that there is a picture hanging on the wall.’ On my assumptions, the logical representations of (39), (40) and (41), (42) are (43a), (43b), and (44a), (44b) respectively:

(43) a. _{*[AgrP Ta [TP[+stative] zongshi [NegP bu [[−stative] zai du shu]]]]} b. _{*[AgrP Qiang shangi [TP[+stative] zongshi [NegP bu}

[[−stative] gua-zhe yi fu hua]]]]

(21)

(44) a. _{[AgrP Ta [NegP bu [TP[+stative] zongshi [[−stative] zai du shu]]]} b. _{[AgrP Qiang shangi [NegP bu [TP[+stative] zongshi}

[[−stative] gua-zhe yi fu hua]]]

Summarizing, I have argued in this section that the interaction between bu, mei, and postverbal de phrases can be nicely accounted for in terms of the proposed aspectual selectional restrictions of the negation markers. This result is very desirable, because it indicates that a unifying account can be given for all the relevant data on the basis of one single assumption, namely, that the negation markers bu and mei have aspectual selectional restrictions.

5. Remarks on previous analyses

Having proposed my own analysis of the negation marker bu and mei, in this section, I will make some detailed remarks on the treatments of bu proposed in Huang (1988), Ernst (1995), and Lee and Pan (2001).

Huang’s (1988) account for the two central facts about bu is based upon a principle — referred to as ‘‘principle P’’ — that requires that it ‘‘form an immediate construction with the first V0 element following it.’’ According to him, principle P can be derived as a theorem on the assumption that ‘‘bu is base-generated as a bound form in an Infl node which, if containing no lexical material, triggers Koopman’s verb-raising rule’’ (Huang 1988: 286–287). On this analysis, sentences like (3c), reproduced as (45a), form a negated predicate as represented in (45b). (45) a. *Wo bu chi-le pingguo

I not eat-ASP apple b. _{[ [V0 bu [V0 chi]] le]}

not eat PERF

However, (45b) is a semantically uninterpretable structure, because it is absurd to assert completion of an event that does not happen. The same analysis also explains why bu is incompatible with a postverbal descriptive or resultative de phrase in a similar vein: nonexistence of an event cannot have a result or manner.

Huang’s above analysis, though quite interesting, has empirical prob-lems. Consider (46), in which the negative morpheme bu is separated from the adjective (or verb) by a degree word.

(46) Wo hai bu zenme er I yet not very hungry ‘I am not very hungry yet.’

(22)

If bu in (46) triggered the raising of the adjective (or verb) er ‘hungry’ to Infl, a structure similar to (45b) would be derived. This then falsely predicts that (46) should be ill-formed parallel to (3c) or (4c), because it is absurd to ascribe a degree to a nonproperty. Examples like (46) thus suggest that bu cannot be a verbal clitic, a conclusion that Ernst (1995) also reaches on the basis of the fact that bu is not always adjacent to the verb of the sentence, as (47) shows.14

(47) Jinrong bu mashang huida Jinrong not immediately answer ‘Jinrong doesn’t answer immediately.’

In addition to examples like (47), Ernst also observes that it is not true that a perfective marker is semantically incompatible with a non-event, as Huang assumes. For example, sentences such as John has [not come], where the negation marker has narrow scope with respect to the perfective experiential marker, are perfectly acceptable. I agree with this point.

Finally, Ernst oﬀers evidence showing that bu clearly takes scope over not just the verb but other constituents:

(48) Ta yiban (dou) bu shui san-ge xiaoshi he generally all not sleep three-CL hour ‘He generally doesn’t sleep for three hours.’

According to him, the duration phrase in (48) must be within the scope of negation, so the scope of bu cannot be restricted to the verb as Huang’s analysis assumes.

Given the above problems with Huang’s analysis, Ernst proposes that bu is a proclitic on the following word rather than a verbal clitic. Moreover, it aspectually requires an unbounded situation as its comple-ment. As mentioned in the introduction section, Ernst’s aspectual require-ment accounts for the inability for bu to occur with le because a conflict in (un)boundedness exists between them. On the other hand, the proclitic analysis explains the incompatibility between bu and a postverbal phrase, because the trace left by the postverbal de phrase, assumed to have moved from a preverbal adjunct position, blocks bu from cliticizing onto the adjacent lexical word.

However, Ernst’s solution to the distribution of bu has a conceptual weakness as well as many empirical problems. The conceptual problem is that unlike the unifying analysis proposed in this paper, his analysis uses two diﬀerent mechanisms to account for two sets of facts, which seem to be reflections of the same phenomenon. As for empirical prob-lems, reconsider the interaction between bu and progressive sentences

(23)

such as (9)–(10), locative-inversion sentences such as (11)–(12) and sentences with the durative aspectual marker zhe such as (13)–(15), discussed in section 2. These sentences are all unbounded sentences but their negation marker is not necessarily bu. The contrast between the examples (13), (14), and (15) is particularly interesting because they all describe unbounded situations but only the examples in (15) license the use of bu. This indicates that the crucial notion governing the distribution of bu cannot be the notion of (un)boundedness as Ernst suggests.

The most recent analysis of bu is provided by Lee and Pan (2001), who argue against Huang’s and Ernst’s clitic approaches to the negation marker bu. They point out several problems with this approach.

First, Huang’s nonevent account relies on whether the relevant event exists or not, but this seems to be wrong, because sentences containing a potential de phrase need not describe a real event.

Second, bu is a focus element that may focus or negate a constituent other than the verb or the constituent immediately following it. For example, in (49), bu focuses kuai ‘fast’ rather than pao ‘run’, so it negates the former rather than the latter, a fact that cannot be correctly predicted by the clitic approach.

(49) (Lee and Pan 2001: 708)

Zuotian ta yaoshi bu pao-de name [kuai ]f, jiu hui yesterday he if not run DE that fast then will

wu-le huoche

miss-ASP train

‘Yesterday, if he had not run that fast, he would have missed the train.’

Third, bu can actually cooccur with a postverbal de phrase and perfec-tive le, and this is not restricted to conditional clauses. Lee and Pan’s (2001: 709–710) examples are the following:

(50) Ta yaoshi bu liu de [kuai ]f, jiu gei jingcha zhua qu le he if not run DE fast then get police caught away SFP ‘If he had not run that fast, he would have got caught by the police.’ (51) Zhangsan guyi bu ba [suoyou ]f de lanpingguo dou

Zhangsan deliberately not BA all DE rotten-apple all reng-le weile re ni shengqi

throw-LE for make you angry

‘Zhangsan deliberately did not throw away all rotten apples so as to make you angry.’

(52) Ta zai jia bu shuo de kuai, ni na ta yidian banfa dou he at-home not speak DE fast you do him some thing all

(24)

meiyou not-have

‘He doesn’t speak fast at home, and you can do nothing to him.’ (53) Ta neng bu pao de bi Lisi man ma?

he can not run DE than Lisi slow SFP ‘Can he not run slower than Lisi?’

(54) Ta changchang bu shuo de hen qingchu he often not say DE very clear ‘He often does not speak very clearly.’

Finally, even when bu directly focuses the verb, the sentence is still well-formed. According to them, Huang’s and Ernst’s analyses fail to predict the grammaticality of (55).15

(55) Yaoshi ta bu [shuo]f-de hen kuai, ta [ xie]f-de hen kuai, ni if he not say DE very fast he write DE very fast you yao bu yao ta

want not want him

‘If he does not speak very fast but he writes very fast, do you want him?’

Although not every argument and example of Lee and Pan’s are convincing, I agree with them that bu should not be treated as a clitic. This is particularly clear when we consider the fact that bu may focus and hence negate constituents in various positions.

Given the above problems with Huang’s and Ernst’s approaches, Lee and Pan (2001) propose to account for the inability of bu to occur with the perfective marker le and the postverbal de phrase on the basis of the three assumptions in (56) and the interpretation condition in (57). (56) a. Bu is a focus-sensitive operator and an unselective binder that

can bind any variable.

b. Perfective le is a selective binder that binds an event or situation variable.

c. A manner phrase requires a sentential subject that contains a free event or situation variable.16

(57) The Interpretation Condition (IC) (Lee and Pan: 717):

The negator bu associates with the focus if there is one to its right and thus introduces a tripartite structure; otherwise it negates the adjacent word.

With the above three assumptions and IC, Lee and Pan explain the incompatibility between bu and le as follows. In examples like (3c), reproduced below, there is no focus, so bu negates the adjacent verb and

(25)

takes narrow scope with respect to le, which is assumed to have sentential scope. This then yields a Davidsonian’s or Parsonian’s type of logical structure like (58).

(3c) *Ta bu chi-le mugua he not eat-ASP papaya

(58) _{Le [Bue [chi(e) m Subject(e, ta) m Object (e, mugua)]]}

In (58), the event variable is bound by the unselective binder bu, leaving the selective binder le nothing to bind. The representation (58) thus violates the ‘‘prohibition against vacuous quantification’’ ( Kratzer 1991; de Swart 1993).

On the other hand, if there is a focus as in (51), bu is compatible with le, because partition of the structure into tripartite structure forces le to take narrow scope with respect to the focusing bu, which induces a set variable of alternatives (Rooth 1985). The logical representation of (51) is thus (59).

(59) bu {[Zhangsan guyi ba P de lanpingguo dou reng-le], p=suoyou} In (59), le selectively binds the event variable of the verb and bu binds the set variable P. Therefore (51) is well-formed.

Lee and Pan’s explanation of the incompatibility between bu and de complement is quite similar to their account for the incompatibility between bu and le. In examples like (4c), reproduced below, there is no focus, so its logical representation should be (60).

(4c) *Ta bu pao de kuai he not run DE fast ‘He does not run fast.’

(60) _{Kuai ([Bue [Pao(e) m Subject(e, ta)]])}

In (60), the event variable is bound by bu, producing a closed proposition. Consequently, the manner phrase kuai ‘fast’ has no free event or situation variable to be predicated of. This explains the ungrammaticality of (4c). On the other hand, if there is a focus of bu as in (50), the story will be similar to the case of (51).

Lee and Pan’s analysis is very interesting but it has no fewer problems than Huang’s and Ernst’s analyses. To begin with, the assumption that bu and le may close an event or situation variable seems problematic. Consider the following example with no particular focus.

(61) Ta chi-wan-le wanfan yihou, youshi/zongshi hui chuqu he eat-finish-ASP dinner after sometimes/always will go-out sansanbu

take-a-walk

‘After he eats his dinner, he sometimes/always goes out to take a walk.’

(26)

It is now well known that adverbs of quantification such as always, sometimes may bind situation variables (Lewis 1975; Heim 1982; von Fintel 1994). Thus, (61) can be paraphrased as, ‘Every situation in which he eats his dinner can be extended to a situation in which he goes out to take a walk.’ However. if le is able to close a situation variable as Lee and Pan have assumed, then the adverb of quantification would not be able to bind the situation variable. Therefore, on Lee and Pan’s assump-tion, (61) should also be a case in violation of the prohibition against vacuous quantification, just like his analysis of (3c) and (4c), which is obviously wrong.17

The assumption that bu closes a proposition also runs into problems in examples with multiple negation:

(62) a. Ta bu hui bu lai18 he not will not come

‘It will not be the case that he won’t come.’ b. _{Bu [ Hui [Bue [lai(e) m Subject(e, ta)]]]}

As shown by (62b), when the event variable introduced by lai ‘come’ is bound by the closest negation operator, it will no longer be available for binding. Consequently, the outside bu, and possibly hui ‘will’ as well, has nothing to bind, in violation of the prohibition against vacuous quantification. But (62) is a perfect example.

Still another problem with Lee and Pan’s analysis of bu is related to the fact that it can negate constituents other than the verb, as in the sentence Ta bu chang lai ‘he not often come’ with no particular focus. In such examples, it is hard to imagine that the adverb chang ‘often’ intro-duces a variable that is bound by bu, because chang itself is a quantifier that binds variables. It should also be noted in passing that in this example bu cannot bind the event variable introduced by the verb, because this will give a wrong semantics for the sentence.

Lee and Pan’s explanation of (3c) also has a flaw. Recall that they assume that bu is a focusing element and is able to unselectively bind any variable. In particular, they also assume that bu can bind variables introduced by bare nouns in Chinese. Now, imagine that (3c) is uttered with a focus on the object NP pingguo ‘apples’. On this reading, the logical representation should be as follows:

(63) bu {[ta chi-le p], p=apples}

In this logical representation, le can bind the event variable introduced by the verb and bu binds the set variable p introduced by the object NP. Thus, the prohibition against vacuous quantification is not violated, but

(27)

(3c) with a focus on the object NP is as ill-formed as when there is no focus at all.

As for Lee and Pan’s example (51), in which le occurs with bu, it is not clear that this is a true counter-example. It has been observed that in addition to marking the aspect of a sentence, le has another usage, similar in function to the resultative complement diao ‘off ’, as in mai-diao ‘sell off ’ or chi-diao ‘eat off ’ (Heinz 1990, among others). According to Heinz (1990), only a small set of verbs, including wang ‘forget’, reng ‘throw away’, chi ‘eat’, etc., that is, verbs with a ‘‘disposal meaning,’’ can have this usage. Evidence for the nonaspectual usage of le comes from several considerations. Here I will only give two, discussed by Heinz (1990). First, replacement of le with diao ‘off ’ in such examples does not change the meaning of the sentence. Second, the aspectual marker le is in complementary distribution with mei ‘not’ ( Wang 1965; Chao 1968), but le occurring with the above set of disposal verbs is not subject to this restriction, just as normal resultative complements are not. Compare the following examples:

(64) a. Lisi da-le Zhangsan Lisi hit-ASP Zhangsan ‘Lisi hit Zhangsan.’ b. *Lisi mei da-le Zhangsan

Lisi not hit-ASP Zhangsan ‘Lisi did not hit Zhangsan.’ c. Lisi mei da Zhangsan

Lisi not hit Zhangsan ‘Lisi did not hit Zhangsan.’

(65) a. Lisi da-si-le Zhangsan

Lisi beat-dead-ASP Zhangsan ‘Lisi killed Zhangsan.’

b. Lisi mei da-si-(*le) Zhangsan Lisi not beat-dead-ASP Zhangsan ‘Lisi didn’t kill Zhangsan.’

(66) a. Ta (dao xianzai) hai mei wang-le nei jian shi he until now still not forget-LE that CL matter ‘Until now, he still hasn’t forgotten that matter.’ b. Ni dao xianzai hai mei reng-le nei jian yifu a

you until now still not throw-LE that CL cloth PAR ‘How come you haven’t thrown away that cloth until now?’ If Heinz is correct, then Lee and Pan’s example (51) is not a true counter-example to the generalization that bu may not occur with le, because the le in this example is a resultative complement equivalent to diao ‘oﬀ ’ in

(28)

meaning. Notice that if the verb reng ‘throw’ in (51) is replaced by some other verb, such as mai ‘buy’, the sentence becomes unacceptable. This confirms the analysis that the le in (51) is better treated as a resultative complement. Also notice that (51) has a strong volitional meaning.

In fact, I would like to claim that none of Lee and Pan’s examples in (49)–(55) is a true counter-example to the generalization that bu may not occur with le and a postverbal de complement. The examples in (49) through (55) all have a volitional or future modality reading and the actions denoted by the verbs are controllable by the subject NPs. Take (49), for example. The negation marker for the conditional clause can actually be mei or bu, but the implication is diﬀerent depending upon which negation marker is used. When the negation marker is bu, the subject NP’s volition is clearly involved, though this is not necessarily the case for mei. In view of this, I assume that (49)–(55) are all analyzed as containing an empty volitional or future modal verb, as Huang (1988) suggests.

Further evidence supporting the above view comes from the following consideration. When one replaces the [+controllable] verbs in (49)–(55) with a [−controllable] verb, making the volitional reading impossible, those sentences with an adverb of quantification all become unacceptable, no matter what constituent bu is intended to focus:

(67) *Ta zongshi bu ying de name piaoliang he always not win DE that beautiful

‘It is always the case that he does not win beautifully.’ (68) *Ta chang bu bing de name lihai

he often not sick DE that serious

‘It is often the case that he is not sick that seriously.’ (69) *Jiali de shi, ta zongshi bu zhidao de name qingchu

home DE matter he always not know DE that clear

‘As for matters of our(his) home, it is always the case that he does not know that clearly.’

(70) *Zhe zhong bing yiban bu quanyu de hen kuai19 this kind illness generally not recover DE very fast

‘It is generally the case that from this kind of illness one does not recover fast.’

Observe that all the ungrammatical examples in (67)–(70) have the adverb of quantification preceding the negation marker. As before, if the order is reversed, the sentences become fully grammatical:

(71) Ta bu zongshi ying de name piaoliang he not always win DE that beautiful

‘It is not always the case that he wins beautifully.’

(29)

(72) Ta bu chang bing de name lihai he often not sick DE that serious

‘It is not often the case that he is sick that seriously.’ (73) Jiali de shi, ta bu zongshi zhidao de name qingchu

home DE matter he not always know DE that clear

‘As for matters of our(his) home, it is always the case that he does not know them that clearly.’

Another relevant fact that should be pointed out here is that condi-tional clauses may seem to allow bu to modify a nonvolicondi-tional verb that takes a de complement, as (74) illustrates.

(74) Wo de jiao yaoshi bu tong de name lihai de-hua, wo jiu

I DE leg if not hurt DE that serious if I then

gen ni yiqi qu

with you together go

‘If my leg does not hurt that seriously, I will go with you together.’ However, such conditional clauses always have the same kind of modality force expressed by hui ‘will’.20 If we assume that there is indeed an empty hui ‘will’, as in Huang’s analysis, then examples like (74) are not a problem, because stative modals license the use of bu, as we have seen. There is no doubt that it is a very complicated issue when an empty modal can be licensed. However, discussing this issue is beyond the scope of this paper. So I will leave it to another occasion.

6. Conclusions

In this paper I have argued that the distributions of bu ‘not’ and mei ‘not’ can be accounted for in terms of their aspectual selections. I have proposed that while mei aspectually selects an event as its complement, bu aspectually selects as its complement a stative situation that requires no input of energy in order to obtain that situation. I have shown that this approach to the distributions of bu and mei not only avoids many problems that previous analyses such as Huang (1988), Ernst (1995), and Lee and Pan (2001) face but also provides a single unifying account for a wide range of data already discussed and not discussed in the literature.

Received 9 October 2001 National Chiao Tung University Revised version received

20 March 2002

(30)

Notes

* Part of this paper was presented at the Workshop on Formal Syntax and Semantics, National Chiao Tung University, Hsinchu, September 22, 2001. I would like to thank the participants at the workshop, especially Kuang Mei, Ting-chi Tang, Weitian Tsai, and Yongli Zhang for their useful comments and suggestions. I am also very grateful to two anonymous referees’ help with the reorganization of this paper, as well as their valuable suggestions for some points. As usual, the author alone is responsible for any possible error in this article. Correspondence address: Department of Foreign Languages and Literatures, National Chiao Tung University, 1001 Ta Hsueh Road, Hsinchu 30050, Taiwan. E-mail: [email protected].

1. An exception to the cooccurrence restriction between bu and stative predicates is the verb you ‘have’. When the verb is you ‘have’, mei must be used instead of bu. This is illustrated by (i).

(i) Ta mei/*bu you qian he not/not have money ‘He is not rich.’

The exception in question is due to an idiosyncratic property of you and should not be dealt with by general rules.

2. Notice that the volitional or future modality reading of bu is not available for just any predicate. For example, in contrast to the examples in (8a) and (8b), the example (ia) sounds quite odd.

(i) a. *Ni/ta bu gao qingchu wenti chu zai nali you/he not make clear problem out in where ‘You/he does/did not make clear what the problem is.’ b. Ruguo ni/ta bu gao qingchu wenti chu zai nail, ...

if you/he not make clear problem out in where ‘If you/he does not (want to) make clear what the problem is, ...’

But the same negative proposition becomes acceptable when it is embedded into a conditional clause as is shown by (ib). It is not clear to me why conditional clauses license a volitional or future modality reading more easily than normal declarative clauses and I will not further pursue this issue.

3. Ernst (1995) has cited (i) as a grammatical sentence to support his unboundedness requirement.

(i) Hongmei bu zai shuo hua Hongmei not PROG say speech ‘Hongmei isn’t speaking.’

However, he says that use of bu in sentences like (i) is only characteristic of southern speakers, while mei is more common for speakers of northern Mandarin. I know nobody from Taiwan who accepts examples like (i), though many of them speak the southern dialect; nor have I read any paper that says that sentences like (i) are gram-matical, but there are references that report that such sentences are ungramgram-matical, Guo (1997), a mainland speaker, for example. Therefore, in this paper, I will assume that sentences like (i) above are ungrammatical. Also because Ernst does not provide a way to explain the variation that he reports, I take my dialect to be a problem with Ernst’s analysis.

(31)

4. Examples like (i) are acceptable, but their interpretations are volitional. In other words, what bu negates here is an empty stative modal.

(i) Ni bu zai chuang shang tang-zhe, ni yao qu nali? you not on bed on lie-ASP you want go where ‘You don’t (want to) lie on the bed. Where do you want to go?’

5. It is interesting to point out that if a (negative) property is very unlikely to be contextu-ally turned into a permanent stative, then even if bu is used, the sentence cannot be understood as an attitudinal. For example, the property of not taking a bath is very unlikely to be contextually turned into a permanent property of an individual. Thus, (i) cannot be understood as an attitudinal, though the negation marker bu is used. Instead, it is construed as a volitional or future-tense sentence.

(i) Wo bu xi zao I not take bath

‘I do not want to take a bath.’

6. An anonymous reviewer says that (18b) is odd, because mei seems to modify quite specific episodes and is thus not compatible with adverbial modifiers. This is not true. (18b) can be easily understood as a generic sentence, when some more words are added, as in (i) below.

(i) Xiaoming changchang mei xi zao jiu qu shuijiao Xiaoming often not take bath then go sleep

‘It is often the case that Xiaoming does not take a bath and goes to sleep.’ 7. The use of you in this sentence is grammatically correct in the Taiwanese variety of

Mandarin Chinese but is not accepted by mainland speakers.

8. For a review of the most recent analysis of le put forth by Klein et al. (2000), see Lin (forthcoming).

9. Chinese has a sentence-final le, which is homophonous with the verbal le, as is illustrated in (i).

(i) Ta chi fan le he eat rice ASP ‘He has eaten his meal.’

In this paper, I will only discuss the verbal le.

10. It is also possible that the le in (21a) and (21b) is the sentence-final le. But it is diﬃcult to argue for or against it.

11. There are examples where bu appears in a resultative de complement, as (i) below shows.

(i) Ta qi de bu shuo hua he angry DE not say word

‘He was so angry that he did not want to speak.’

However, in such examples, mei can still negate the verb–de–complement as a whole. (ii) Ta mei qi de bu shuo hua

he not angry DE not say word

‘He was not so angry that he did not want to speak.’

12. Normally, adjuncts in Chinese only appear in a preverbal position, as is illustrated by (i).