A discourse Grammar of Mandain Chinese1

10.1. Between SENTENCEs 10.1.1. Lexical Cohesion 10.1.2. Rhetorical Relations 10.2. The Paragraph: Discourse Theme and Paragraph Boundary 10.2.1. The Discourse Theme 10.2.2. How Paragraph Boundaries Are Identified 10.2.2.1. Semantic Incompatibility 10.2.2.2. Appearance of 'Controlling' Phrase of the Same Category 10.2.2.3. Switch ofViewpoint..

10.2.2.4. Appearance of Anaphoric Phrase 10.2.2.5. Difference in Style of Writing 10.2.2.6. End-of-Scope Description 10.2.2.7. Summary

385 385 393

401 401 404 405 .405 407 407 408 .409 409

10.3. Discourse Markers in Spoken Mandarin 10.3.1. Miracle on Hao, Keshi (Danshi, Buguo) and Na(me) 1O.}.2. Yang on Na and Wang on Ranhou

411 411 .413

11.1. The Running Theme 11.2. Grammatical Roads That Lead to Discourse Structure 11.3. Further Issues and Outlook

416 418 419

INTRODUCTION: BETWEEN GRAMMAR AND DISCOURSE

Researchers in Chinese grammar have long recognized the limitations of the Western theoretical framework of syntax that has been imposed on the study of Chinese grammar since Ma Jianzhong's Ma Shi Wen Tong toward the end of the 19th century. For example, Chao (1968) adopts the notion of 'topic' instead of the Western 'subject' as one of the immediate constituents of the Chinese sentence.'

Since then, topic has become a familiar term to most grammarians in the characterization of Chinese and other similar languages in spite of the lack of a precise defmition for it. Through the development of Chinese grammatical study in the past twenty years or so, 'topic' has been utilized to create further functional notions, such as 'topic prominency' (Li and Thompson, 1976), 'topic chain' (Tsao, 1979), 'secondary topic' (Tsao, 1990), and many others. Topic structures in Chinese are also well recognized and worked on in the G-B theory (Li, 1990:197-200). They all represent a departure from the mainstream Western linguistic thought and a new direction for Chinese linguistics.

While the tremendous amount of work done on 'topic' has certainly helped answer many structural and functional questions in Chinese grammar, there are a lot more pr.oblems that can be raised concerning the further understanding and explanatIOn of the grammar of the language. For instance, are there other functions for the verbal suffixes, such as Ie and zhe, than their aspectual marking function?

Why is it that adverb sets such as you, zai, and hai are interchangeable in isolated clauses while they are not always interchangeable in larger contexts despite their near synonymity (Chen, 1993)? Why is it that the third-person pronouns ta and tamen tend to occur less frequently in Chinese than their counterparts in many Western languages? How can a Chinese 'sentence' be given a more precise definition than 'expressing a complete thought' while English and other Western lan?uages. seem to have ~ better structural defmition for it? Why is it that many native ~hmese grammarians have perceived the Chinese language as having a predommantly liushuiju (literally, 'flowing-water sentence') structure2 as distinguished fro~ other Western types of sentences? These are, of course, ju~t a few of the questIOns that may be raised within the context of current research in Chinese grammar. Attempts have been made to explain such facts but, unfortunately, many of them have only appealed to linguistic bias and conjecture through native intuition (Cf. Shen, 1988).

Underlying all those questions, however, there is one fundamental problem with

the traditional, and some of the current, grammatical theories. That is the strict adherence to 'sentence grammar'-grammar that limits itself within the confines of the sentence. In other words, most traditional and many currently dominant grammatical theories are only concerned with the structure of the sentence, without recognizing the fact that the form of a sentence may be affected by extra-sentential factors. When the criteria of such a sentence grammar are applied to a language like Chinese, it is immediately obvious that the model is far from being adequate for describing the structure of a linguistic system that lacks an elaborate formal apparatus of tense-aspect, case marking, voice, modal auxiliaries, etc. in terms of the familiar structural signals that prevail in Indo-European languages. Chinese, in particular, relies heavily on relative ordering of constituents, inter-clausal core ference, particles, and semantic correlates, among many others, to signal syntactic structure as well as discourse relations. It is therefore indispensable, on one hand, to account for the syntactic structure of Chinese in terms of signals different from the ones familiar to most Western grammarians and, on the other, to utilize discourse notions to uncover the inner workings of the clause/sentence structure of the language. The presence of the verbal suffix -Ie, for example, is not only to satisfy the need for marking the perfective aspect, but also to indicate the organization of clauses into larger units. (Cf. Chu and Chang, 1987.) Thus, the grammatical description of the suffix -Ie cannot be regarded as complete without consideration of its discourse functions.

Facts like the above have in many ways convinced the present author that a meaningful grammar of Mandarin Chinese must take into consideration what has generally been swept aside as irrelevant to syntax because of its functional and pragmatic nature. One of the purposes of this volume is, for example, to show that functional and pragmatic factors, such as information structure and the relative degree of foregrounding and backgrounding, are not only relevant but also crucial to the determination of the structure ofa clause in Chinese. Another such purpose is to demonstrate that there is not necessarily a clearcut dividing line between syntax and discourse but rather a continuum extending from one to the other.

It is on the basis of beliefs like the above that the present author has been working in the areas of Chinese syntax and discourse. He has thus called his work Discourse Grammar rather than the more prevalent term Discourse Analysis.

The term Discourse Grammar3 is here used to designate the area of linguistic inquiry that unites the study of the syntax of the clause on one hand and the study of the organization of clauses into larger structural units on the other. It covers some part of the syntax of the clause and the combination of single clauses into what might be perceived as similar to 'compound sentence' and 'complex sentence' in Western terminology as well as combining these 'sentences' into even larger units like the 'paragraph'. More importantly, however, it deals with the effect of such combinations on the internal structure of the clause.

An example of the discourse grammar of Mandarin Chinese may be taken, again, from the use of the verbal suffix -Ie, which has been regarded as the perfective aspect marker of the language. As a perfective aspect marker, however, it doesn't parallel many of the known uses for Western languages. Though it does often mark a past action or event in its entirety (Comrie, 1976:12); yet, even more often, such an action or event is NOT so marked in Mandarin. Otherwise, -Ie is used for something else. (Cf. Chu & Chang, 1987; Chang, 1986.) These latter cases cannot be simply dismissed as optional or be reasonably explained on the basis of perfectivity without taking into account the discourse functions of the suffix. A specific instance is the deletion of -Ie in a series of action verbs except the last one to signal their relatedness.

Another example concerns the relative positioning of the reason-consequence clauses containing the conjunctive adverbs yinwei 'because' and suoyi 'therefore'.

It is often said that Chinese grammar strictly observes the iconicity principle of cause-effectlreason-consequence and therefore the yinwei-clause as a statement of 'reason' invariably precedes the statement of 'consequence' in the form ofa clause with or without suoyi. As a matter of fact, there ARE cases where we find the reverse order of the consequence clause without suoyi before the reason clause with yinwei. These cases then are often conveniently labelled as Westernization.

Admittedly, there are cases of Westernization where no justification of reversing the order can be found. Ifwe take a closer look, however, some instances of the reverse order may be well justified on the basis of discourse cohesion, just like the preposed and postposed if- and when-clauses are in English (Ramsey, 1987).

Given the rough defmition of a discourse grammar proposed at the beginning of this section, the discourse grammar of Mandarin Chinese then must be able to relate the structural descriptions of such grammatical items as the verbal suffixes and the relative order of clauses to the cohesive organization of a block of discourse. In other words, it must account for how the structure of a clause may affect the cohesion of the linguistic form within which the clause occurs. Conversely, it must also be able to demonstrate in what ways discourse may require that a structural unit, such as a verb or a clause, take one form over any other for reasons of felicitous organization.

Thus, a discourse grammar in our sense is distinct, in some significant ways, from what is generally known as discourse analysis in linguistics. We address the following three areas of difference.

First, while discourse analysis is by and large communicatively oriented, discourse grammar is structurally oriented. The prevailing view of discourse analysis is that discourse is a process and it takes 'words, phrases and sentences which appear in the textual record of a discourse to be evidence of an attempt by a producer (speaker/writer) to communicate his message to a recipient (hearer/reader)' (Brown and Yule, 1984:24). The main purpose of a discourse grammar is, on the other hand, to examine the structure of discourse: what structural devices are used to perform what functions?

Secondly, while discourse analysis is concerned with both spoken and written

discourse (though for many researchers spoken discourse is the primary concern); discourse grammar at this stage mainly deals with the written text (cf. Liao, 1992).

This is not a distinction by nature but one by convenience. Since discourse grammar is at its infant stage, it has to start from something that is more tangible and better understood. By taking written language as its data base, discourse grammar may not have to worry about intonation, speaker attitude,4 performance factor, etc., all of which are less well understood than overt written signals, especially in terms of Mandarin Chinese.

Thirdly, discourse grammar emphasizes both the effect of discourse organization on the structural grammar of the clause and that of grammar on discourse, while there is no such emphasis discernible in the usual practice of discourse analysis.

This has to do with the primary motivation of the present author. He started with syntax in his academic career, but was disillusioned by the restrictive approaches in the syntactic theories at that time. He turned to discourse for solution to the syntactic problems that he had to deal with. He has now been fIrmly convinced that there exists in discourse a large body of reasonable explanations for syntactic problems, which would otherwise seem to be arbitrary or unaccountable. This naturally leads to why there is a need for a discourse grammar, especially for Mandarin Chinese.

Before we go on, some clarifIcation is in order. Our discourse grammar in one way or another bears quite some resemblance to one view of discourse analysis, which Brown and Yule (1985:24-25) terms discourse-as-product. Typical of this view is the cohesion approach such as exemplifIed in Halliday and Hasan (1976).

It is obvious that many of their notions on cohesive ties are adopted in this volume.

(A) Pronominalization. within a sentence, e.g.

Pronominalization,

for example, is supposed to occur

(l.a) When Dad! comes home, hej will be mad. b) When hej comes home, Dadj will be mad.

(2.a) Dadj will be mad when hej comes home. b) *Hej will be mad when Dadj comes home.

In the (a) sentences, forward pronominalization is said to have occurred and there is no restriction that forbids such forward process. In the (b) sentences backward pronominalization is said to have occurred, but there is some restriction that forbids such a backward process. Sentence (2.b) is ungrammatical just because it has violated the restriction. While it is true that the ungrammaticality of (2.b) is the result of violation of some rule, the acceptability of (l.b) may not be by the same token that the other sentences in such sets are acceptable. The following pair that Kuno (1980: 131) gives illustrates the problem: (3.a) *In John's portrait of Maryj, shej found a scratch. b) In John's portrait ofMaryj, shej looks sick.

Any syntactic constraint should be applicable to both or neither of(3.a) and (3.b).

As a matter of fact, the same argument also applies to forward pronominalization: (Kuno, 1980: 130) (4.a) *In John/s

dormitory, hej smoked pot.

b) In John!' s dormitory, only hej smoked pot.

In the early 70s, some syntacticians started to feel uncomfortable about the solutions that autonomous syntax had to offer for many structural problems. Among those problems were, just to name a few, pronominalization, reflexivization, defIniteness interpretation, tense-aspect marking, relativization, and subordination. What autonomous syntax like transformational grammar did was to set up a rule for each and every syntactic phenomenon to generate grammatical forms and rule out ungrammatical ones. While such rules in a grammar may certainly be made related to each other and most of them may be made general enough to subsume certain portions of related phenomena under them, they don't really offer any further rationale for why there are such forms. Thus, one can not feel satisfIed with such rules unless one is willing to accept them as the ultimate explanations. This dissatisfaction led to the formulation of many other theoretical frameworks such as generative semantics, case grammar, lexical function, etc. One of the further developments resulted in functional syntax, which paved the way for discourse grammar.

Some of the problems that fInd autonomous syntax insufficient are illustrated below.

According to Kuno, the problem involves whether the nominals in question are thematic or not. Pronominalization therefore is not merely an INTRAsentential operation and it has to go beyond the sentence boundaries to fInd a more plausible solution.

Pronominalization in Chinese is notoriously irregular in terms of J.Ylltax. Any account for the many uses and non-uses of third-person pronouns in Chinese must address a larger domain than what has usually been equated to the Western sentence. (See, for example, Cherry Li, 1985.)

(B) Reflexivization. The solution to reflexivization has long seemed to be the comer stone of formal syntax. In early generative grammar, it was often used as the starting point for transformational rules. Later in the G-B theory, it has formed part of the basis for the binding theory. One would think that it ought to have been solidly established and universally accepted as a grammatical rule. Recent research, however, has discovered that many cases of the reflexive pronouns in English and

other languages are not accountable by such rules. Zribi-Hertz (1989:707-8) cites the following examples as reflexivization on the principle of 'emphasis' and 'domain of point of view': (5.a) ARTHU~'s fulsomeness seemed to embarrass the Baron as much as it did HIMSELFj. b) MARYj eventually convinced HER SISTER SUSANj that John had better pay visits to everybody except THEMSELVESj&j'

In Mandarin Chinese, the problem seems to be even more complex. Chu (1989:101-6) presents a survey of60 native speakers on the use ofziji 'self and taziji 'him-/herself in various contexts. He finds that while the binding principle can correctly apply to the use of the reflexive pronouns in single clauses in Mandarin, in cases where more than one single clause are involved the notion 'theme' must be invoked to provide a better explanation. (Also cf. Xu, 1993.) The analysis of the Korean reflexive caki seems to involve a hierarchical ranking of the coreferential NPs. (See O'Grady, 1987, and Kang, 1990, for example.)

(C) Aspect Marking. In most theories of syntax, tense and aspect have merely been treated as a problem of cooccurrence of verbal inflections with appropriate auxiliaries within a clause. The more substantial problem of choosing one tense or aspect over another is relegated to rhetoric. Only recently has the use of tense and aspect attracted the attention of theoretical linguists such as Hopper (1979) and Givon (1984). Aspect, in particular, seems to bear more on discourse than on anything else. Hopper (1982:4) suggests for aspect 'a core function which is discourse-derived and in some sense universal, and a set of additive functions which are not universal (though some of them may be common) and which represent grammaticized semantic extensions of the discourse function.' Foreground in discourse seems to be closely related to the perfective aspect in French and Russian (Hopper, 1979:217-8). In Mandarin Chinese, the perfective aspect marker -Ie performs, among many others, a forward linking function (Li & Thompson, 1981:640-1) or a 'peak' and an anteriority function (Chang, 1986:105-114), all of which are discoursal in nature.

Especially disturbing is the problem of when the perfective aspect marker -Ie in Mandarin occurs or does not occur. E.g. in single clauses like the following: (7.a) Zuotian wanshang women kan dianying. yesterday evening we see movie 'Last night we saw a movie.'

b) Zuotian wanshang women kanle dianying. yesterday evening we see-PERFECTIVE movie 'Last night we saw a movie.' there doesn't seem to be any difference between them in isolation. The difference shows up only when one asks whether there is anything following the given statement. The statement in (7.a) doesn't seem to be finished, but the one in (7.b) sounds more complete. An explanation for this kind of phenomenon has to come from connected discourse instead of isolated clauses/sentences.

The problem of aspect in syntax can be further illustrated by the verbal suffix -zhe in Mandarin. Chu (1987) claims that there are three levels of function for this durative aspect marker: semantic, syntactic and pragmatic. While its semantic function is to signal a durative aspect and its pragmatic function is an interactive one, Chu's syntactic analysis of the suffix is to treat it as a means of subordination in both the simple and the complex sentence. This analysis, in fact, should be regarded as a discourse solution to the problem rather than a syntactic one. That is, the decision of when to use the subordination and what to be made subordinate is based on discourse rather than on syntax. E.g., the choice between the following forms depends more on discourse needs than anything else:5 (8.a) Ta tiaozhe wu chang ge. (s)he dance-zhe dance sing song '(S)he sings while (s)he dances.' b) Ta changzhe ge tiao wu. (s)he sing-zhe song dance dance '(S)he dances while (s)he sings.'

Despite the fact that (8.a) and (8.b) are syntactically different but semantically (i.e. propositionally) equivalent, the selection of one over the other depends entirely on whether chang ge 'singing' or tiao wu 'dancing' is taken to be the main theme in the discourse.

(D) Subordination.

As it was mentioned in the preceding sub-section (C), though subordination is a syntactic device, it anchors its motivation in discourse.

The anatomy of the syntax of subordination, therefore, cannot be fully understood unless a discourse view of the structure is incorporated.

In terms of discourse, a claim has been made and generally accepted that independent clauses are for main events in a narrative but subordinate clauses are for background descriptions or non-eventline happenings. This absolute dichotomy, however, has recently been challenged.

Thompson (1987:451) fmds that 'while a majority [of the temporally sequenced events in a written narrative] are coded by independent clauses, a significant subset are "subordinate".'

In each of these cases, she claims, 'the use of a subordinate

clause allows the writer to accomplish a text-creation goal in addition to the obvious one of maintaining the temporal line.' Some of her text-creation goals of the subordinate structures are: (I) to indicate dependency of one event on another (p. 446), (2) to signal simultaneity of one event with another (pp. 446-7), and (3) to perform a relating-back function (pp. 447-9).

Mathiessen and Thompson (1988:275) find that 'it is impossible to defme or even characterize "subordinate clause" in strictly sentence-level terms.' They claim that 'in written English discourse, a certain kind of what linguists have called "subordinate clause" ...can usefully be viewed as grammaticalization of a very general property of the hierarchical structure of the discourse itself.' \ibid.)

Fox and Thompson (1990) show that the form of a relative clause in English conversation (i.e. whether the head NP is a subject, direct object or object of a preposition, etc. and whether the relativized NP is a subject, direct object or object ofa preposition, etc.) can be reasonably explained 'by the information-flow patterns characteristic of English discourse .... ' For example, ' ...Existential-Head relative clauses tend to be S-relatives, since the grounding for the human Existential Heads is typically either main-clause grounding8 or proposition-linking, and the relative clause generally does not serve an anchoring function.' (p. 314) Finally, Myhill and Hibiya (1988) deal with the problem of 'subordination' in another light. Skeptical of the assumption that non-finite verbs as a form of subordination must be background, they fmd that, in terms of grounding, 'non-fmite forms occupy a position between fully finite forms and other subordinate forms.'

(p.395) It is quite obvious that demands of discourse heavily affect the structure of syntax. Despite the general view that syntax makes available formal devices for discourse, it can perhaps be more appropriately claimed that discourse provides a basis for such syntactic forms.

We have argued above that there is a close and indispensable relation between syntax and discourse. While discourse relies to a certain extent on grammatical markings to express its organization, many syntactic devices have their roots in discourse. It is therefore safe to say that one cannot fully understand grammar without also taking discourse into consideration and vice versa. But when we start to look at such a relation in Mandarin Chinese, we immediately fmd ourselves confronted with an entry problem in terms of methodology-that is, the problem of 9

beginning research activities in a relatively uncharted area. To start from Chinese discourse is unrealistic because there is virtually nothing to start from in that area. 10 It is then perhaps more practical to begin with the problems in Chinese grammar that have not been sufficiently dealt with in the current theory of syntax.

The present work, representing the research philosophy of the author, therefore adopts an approach that begins with a look at such syntactic problems. It first takes

up those syntactic features in Mandarin Chinese that seem most likely to be operating at the sentence level, yet have not had a reasonable explanation within that domain. This includes aspect marking and sentential adverbials. From there, the book moves on to areas that have less syntactic relevance. This covers clauseparticles and modality. Finally, the volume enters the purely discourse-oriented domain. This embodies such notions as topicality, information status, and grounding structure. These latter notions, however, necessarily involve syntactic structures such as pronominalization, subordination, and clause-combining. The studying of these notions will profitably result in a relatively precise defmition of what the unit above clause might be like in Chinese. In other words, is there a structural unit in Mandarin Chinese that can compare with the sentence in Western grammar? If yes, what is it like structurally? A positive answer to the questions should lead the study into another level of discourse structure-the definition of paragraph and its organization.

1.4. Mandarin Chinese Discourse Grammar-What Should Be Like

It

On the basis of the discussions in the preceding sub-sections, we are now in a position to layout a blueprint for a complete discourse grammar of Mandarin Chinese. At this point, the requirements that it must meet seem to be the following: (A) A discourse grammar of Mandarin Chinese must distinguish between grammatical markers which are for purely syntactic purposes and those which are also relevant to discourse. It is the latter markers that it must be able to treat reasonably and convincingly. They include, but are not limited to, such devices as aspect marking, pronominalization and subordination. All of them have traditionally been treated as purely grammatical devices or processes with little or no attention paid to their discourse functions.

(B) A discourse grammar of Mandarin Chinese must determine whether markers that are amphibious in nature are actually syntactically oriented or discoursally motivated. Only after such determinations have been made can any reasonable explanation be offered for why a given marker is used in both ways. The sentencefmal particle Ie seems to be a syntactically oriented marker that is extended to signal discourse functions, while other sentence-final particles like ne, me, and (y)a seem to be discoursally oriented but their interpretations are often restricted within the sentence.

(C) A discourse grammar of Mandarin Chinese must attempt to discover discourse signals that have not been grammaticalized. Such signals have traditionally been neglected in linguistic study, but their importance in a discourse grammar is self-evident. The relative order of clauses and the use of adverbs especially in terms of their position in a clause, seem to belong to this category i~ Mandarin Chinese.

(D) A discourse grammar of Mandarin Chinese must recognize grammatical

notions that have impact on discourse but have not so far been utilized in the study of Mandarin Chinese discourse. This covers the notions of state vs. action (and the fmer distinctions in-between), marked vs. unmarked verb forms, and the transitivity scale. Shades of subordination, for example, may be expressed in Mandarin by different forms of the verb in addition to the explicit subordinate conjunctions. Such functions, by which clauses are put together in an orderly manner, are inevitably relevant to the organization of discourse.

(E) Finally, a discourse grammar of Mandarin Chinese must propose an integrated framework that is based on commonly accepted theories of discourse in general but at the same time is compatible with the peculiarities of the language.

Such a proposal represents the ultimate contribution of a complete grammar to the field of Chinese discourse.

As a first attempt for a discourse grammar of Mandarin Chinese, this book does not aim to formulate a complete theory of Mandarin discourse, nor does it claim to present a comprehensive record of all that there is about discourse of the language.

It is rather a sampler of what has so far been done in the field as well as an effort to demonstrate what is most appropriate for a fuller theory of discourse grammar. A large portion of the book is a synthesis of past research in Mandarin discourse, in which the author and some of his graduate students have played an active part. The remainder of the book, the author may legitimately claim to be his own original contributions to the analysis of many of the facets of Mandarin discourse. In fact, such a combination of past research and original contribution can be found throughout the book. The second chapter, for example, is a recount of the grammatical analyses of the aspect markers together with their discourse functions discovered in more recent studies. The author's original contribution in this chapter lies largely in the discovery ofthe background/foreground hierarchy where each of the aspect markers may be assigned a position.

The main emphasis of the book is on the total organization of discourse. While there are specific innovative ideas in the chapters, each ofthem plays a vital part in the final definition of a structural unit above the Mandarin clause. Thus, for example, in addition to the grounding hierarchy of aspect markers mentioned above, Chapter 3 claims its originality in the treatment of the conjunctive nature of many modality adverbs; Chapter 4 proposes a 'personal involvement' factor for the particle alya; Chapter 5 incorporates the management of information and its impact on the interpretation of some types of topics; Chapter 7 identifies the three steps of topic formation; etc. In the end, however, they all converge in Chapter 9 to define a discourse unit in Mandarin to be called SENTENCE, which is vastly different from the Western sentence.

As was stated earlier, the main body of data for study in this book is from written discourse. Hopefully, further research will expand its scope to include spoken discourse.

For the benefit of those who do not have a prior knowledge of Mandarin Chinese, we give a sketch of the language below before we go on to describe its discourse structures.

1.6. A Sketch of the Language The following sub-sections are very short descriptions of modem Mandarin phonology, morphology, and syntax of basic simple sentences. The pinyin system is used in representing the Mandarin pronunciation.

The phonology of Mandarin Chinese generally starts with a discussion of the syllable structure followed by a description of the consonants, vowels and tones. In our presentation we will follow this same format. The Mandarin syllable structure is quite simple and can be represented in (9) below:

where C is a consonant, V is a vowel and N is a nasal consonant. Each pair of parentheses enclose an optional element and V 3 and N are alternates. The three vowels are not equal in prominence of articulation. While V 2 is the vocalic nucleus, V I and V 3 are less vocalic in nature and can be considered glides, i.e. semi-vowels.

There are, however, some occurrence restrictions. Only Iii, Iu/ and IUI may occur in the V I position and only Iii and 101may occur in the V 3 position. Furthermore, (10) Mandarin Consonants: labials

b

dentals

d

velars

g

alveolar affricates

j

retroflexes

zh

alveolar fricatives

z

not occur in the post vocalic position ofN and the velar nasal Ing/ does not occur in the prevocalic position of C. In the spelling system known as pinyin, which is more or less phonemic in nature, there are 22 consonants and 6 vowels. They are listed in (10) above and in (11) below, respectively.

A few explanations are in order. Each of the digraphs represents a single consonant, e.g. Ing/ represents a velar nasal and the graph h in 1M, Ich! and Ish! stands for retroflexion. Apart from the point and manner of articulation derivable from the labels in the fIrst column, there are other important distinctions of the phonemic sounds from their close counterparts in English. For example, the Mandarin /hi is a velar fricative rather than a glottal one. While 1M, Ich! and Ish! are affricate in nature, Irl is more like a fricative than a liquid. The second column contains unaspirated sounds but the third column contains their aspirated counterparts.

The vowel lilI is a rounded high front one, very much like its counterpart in French.

The pinyin system, however, has two more letters that are not included in the chart in (10) above: y and w. They are used to replace a Iii, Iu/ or lilI in the VI position when there is no initial consonant in the syllable. They are also added in front of a syllable that would otherwise begin with a Iii, Iu/ or lilI in the V 2 position.

Now we turn to the tonal system of the language. The tones are defIned by the height of voice pitch over the syllable. Chao (1968) distinguishes a scale of fIve pitch levels and labels them by numbers 1 through 5, with 1 being the lowest and 5 the highest. There are four distinctive tones which are generally numbered from 1st to 4th. The tones are presented in the chart in (12) above.

The fIrst tone starts at the highest pitch numbered 5 and stays on that pitch level all the way till the end of the syllable. It is marked by the symbol ,-, over the vowel of the syllable, e.g. ma 'mother'. The second tone starts at the mid-level pitch numbered 3 and rises to the high level 5. It is marked by the symbol ,', over the vowel, e.g 'hemp'. The third tone starts at the low pitch level 2 and dips even lower to level 1 before it rises to level 4. It is marked by the symbol ,~, over the vowel, e.g. mil 'horse'. Finally, the fourth tone starts at pitch level 5 and falls all the way to level 1. It is marked by the symbol ,', over the vowel, e.g. 'to scold'.

Thus, the tones serve a function of distinguishing meanings carried by the otherwise identical syllables.

ma

ma

Name

Description

Level/Contour

First

High-Level

5-5

Second

High-Rising

3-5

Third

Low-Dipping

2-1-4

Fourth

High-Falling

5-1

Symbol

There is another tone known as 'neutral tone', which generally occurs on an unstressed syllable and may vary in its pitch height depending on the neighboring syllable. To indicate the neutral tone, the vowel is left unmarked, e.g. chuanghu 'window', where the second syllable hu carries a neutral tone, which is indicated by leaving the vowel Iu/ unmarked.

Mandarin morphology is relatively simple in terms of inflection and derivation.

Furthermore, as 'the borderline between affixation and compounding is rather fuzzy in Chinese' (Tang, 1994:496), there are many cases where a form can be considered either as a complex or as a compound word. For this reason, we will just give a few less controversial examples for inflection and derivation. Compounding will be discussed as an important process of word-formation in an outline format.

Two categories of affIXes may be considered most genuinely inflectional in modem Mandarin. They are the plural suffIx -men and the aspect markers -Ie, -zhe, -guo, zai- and -qilai. A third category, the diminutives, is also presented in this section as inflectional in nature. ClassifIers, or measure words, will also be discussed in this section, though they are not really inflectional.

The 'Plural' -men. Strictly speaking, the suffix -men is a true plural marker only when it is added to personal pronouns w6 'I', ni 'you (sing.)' and 'he/she', e.g. wOOlen 'we', nimen 'you (pl.)' and tamen 'they'. When it is used with human nouns, its function is to mark them as collective nouns rather than to indicate the usual sense of plurality, e.g. haizimen 'children', xueshengmen 'students', etc.

Such nouns take on a collective meaning in that they cannot be further specifIed by a defInite number, i.e. one cannot say *sange haizimen for 'three children' or

ta

* liiingbiiige

xueshengmen for 'two hundred students'.

While the suffix -men is basically restricted to use with the pronouns and human nouns, there are further complications on one hand and some relaxation of the basic rule on the other. One of the complications is that some monosyllabic human nouns don't take the suffix easily, e.g. ?blhgmen for 'soldiers' but shiblhgmen 'soldiers', though the form renmen 'people (in general)' is easily acceptable. Others monosyllabic human nouns take the suffix only when another suffix -er is used fIrst, e.g. germen 'buddies' ÂŤgtH er + men). The relaxation of the basic rule has to do with the extension of its use to non-human nouns, e.g. niiiormen 'birds' ÂŤniiio + er + men), gOOrmen 'dogs' (goo + er + men). These forms are mainly used in modem literary writing. In short, the usages of -men as the plural suffix are still in a state of fluctuation.

Turning to the aspectual affixes, we fIrst note that only -Ie, -zhe and -guo are generally recognized and discussed as aspectual markers in most grammars. But we also include the other two, zai- and -qilai, for the reason that they fIt very well into the system of aspect-marking, though morphologically and syntactically they seem different from the other three. We discuss the fIve below.

The perfective Aspect Marker -Ie. The suffix -Ie expresses a 'perfective' viewpoint in the sense that the speaker views a specifIc event as an integral whole, e.g. Zuotian wanshang wo kanle yichang dianying, 'I saw a movie last night,' where kan 'to see' is the verb and the perfective -Ie indicates the speakers is treating the specifIc event of the subject's seeing a moving as a whole. It is much like the past tense in English. But the similarity between the Mandarin perfective and the English past tense stops here. While the English past applies to a state, e.g. I was very tired last night, the Chinese -Ie does not, e.g. Zuotian wanshang wo hen lei, 'I was very tired last night,' where hen lei 'very tired' is the predicate but no perfective -Ie need occur in it.

More detailed discussion ofthis and the other aspect markers can be found in Chapter 2.

The Durative Aspect Marker -zhe. The suffix -zhe expresses a durative viewpoint in the sense that the speaker views the progression of a specifIc event as the most salient, Women kanzhe dianshi (de shihou), huran tingdian Ie. '(When) we were watching TV, the power suddenly went out,' where the suffix -zhe highlights the progression of the event 'our watching TV.' Without another predicate, the one affixed with -zhe is felt to be incomplete. An explanation will be given when the suffix is compared with the progressive aspect marker zai- below.

The Experiential Aspect Marker -guo. The suffix -guo marks some past event(s) as relevant to the present state, e.g. Wo quguo Riben sand, 'I have been to Japan three times.' The -guo on the verb qu 'to go' indicates that the speaker regards the events of 'his/her going to Japan' as an experience. The main difference between the functions of the perfective and the experiential aspect is that the former

reports the happening of an event while the latter asserts the existence of an event and treats it as an experience. Thus, the use of -guo coincides with one of the uses of the English present perfect form.

The Progressive Aspect Marker zai-. The suffix zai- is not often treated as an aspect marker in Mandarin for several reasons. First, unlike the other aspect markers, it is a prefIx rather than a suffix. Secondly, it can be heavily stressed.

Finally, its function overlaps with that of the durative -zhe to a large extent. Just like -zhe, this marker can be similarly characterized as 'indicating that the speaker views the progression of a specifIc event as the most salient.' Thus, the sentence illustrate the use of -zhe can also be used to illustrate that of zai-, i.e. Women zai kan dianshi (de shihou), huran tingdian Ie '(When) we were watching TV, the power suddenly went out.' The most outstanding difference is that the clause with zai- is interpreted as a temporal adverbial but the one with -zhe is interpreted as a manner adverbial.

As a result, the phrase de shihou 'when' is completely optional with the zai- clause as far as meaning is concerned while it is necessary if the -zhe clause is intended to be a temporal adverbial. Another difference is that the zai- clause is more grammatically independent than the -zhe clause is. One piece of evidence for this independence is that the former can occur alone and serve as an answer to a question, e.g. Tamen zai chifan 'They are/were eating' in response to the question Tamen zai zuo shenme? 'What are/were they doing?' On the other hand, as a manner adverbial, a -zhe clause cannot stand alone as an answer to a question asking about the manner in which an action is taken or an event takes place, e.g. *Ta tiaozhe wu 'He dancing' in response to the question Ta zenme chang ge? 'How does/did he sing?' even though it is perfectly acceptable to say Ta tiao-zhe wu chang ge 'He sings/sang while dancing.' The acceptable form of the answer is Ta tiaozhe wu chang, where the verb of the other predicate must be repeated.

The Inchoative Aspect Marker -qilai. This suffix indicates that the speaker is concerned with the starting of an event, e.g. Neige xiaohai shuozhe shuozhe kuqilai Ie 'While talking, the child started crying,' where the verb ku 'to cry' with the suffix -qilai expresses the starting of the action/event. This suffix is not generally treated as an aspect marker for several reasons. First, unlike the fIrst three, it more often receives a stress. Secondly, the same suffix can be used for another meaning, e.g. zhan-qilai 'stand up', where it means 'up' when added to a posture verb like zhan 'to stand'. Finally, the two syllables of the suffix can be separated by a direct object, e.g. Tamen changqi-ge-Iai Ie 'They started singing song(s),' where -qi- and -Iai are separated by the direct object ge 'song'.

From a diachronic viewpoint, the last two affixes are undergoing a process of grammaticalization to become full-fledged aspect markers.

Diminutive Affixes. These are affixes that combine with a noun, basically to attach the meaning 'small' to it. They include the suffixes -er, literally 'son', and -zi, literally 'off-spring', and the prefIx xiao- 'little'. They are best illustrated by the

two words xiaohar (<xiao + hai + er) and xiaohaizi (<xiao + hai + zi). Note that the 'suffix -er is reduced to Irl in the process of affixation. Both of them mean 'child/children' and alternate with xiaohai and haizi and less frequently with har.

The three affixes, however, do not serve the diminutive function in exactly the same way. While the prefix xiao- is more often than not interpreted as carrying the meaning 'little', the two suffixes are looked upon as simply adding something to a short, mostly monosyllabic, noun to make it longer. Thus, to the root {-hail or {hai} 'child', any of the three affixes may be added to form a word, as was illustrated above. But the bare root *hai itself is not acceptable as a free form.

The addition of the suffixes -er and -zi to a noun, however, is not completely productive. Whether a noun can be suffixed with one or the other or neither depends very much on the dialect, though there is some consensus on a number of words.

The forms listed in (13) below are commonly acceptable.

(13) hua-r 'flower' ge-r 'song' yi-zi 'chair' dao-zi 'knife' jin-zi 'gold'

ma-r 'horse' ren-r 'person' Jang-zi 'house' xie-zi 'shoe' yin-zi 'silver'

jiao-zi 'dumpling' piao-zi 'ticket'

Generally spekaing, the addition of the suffix -er is largely optional. In the case of the other suffix -zi, it is obligatory for some, e.g. yizi 'chair' ,jinzi 'gold' andyinzi 'silver'; but is optional for others, e.g. dao(zi) 'knife', and piao(zi) 'ticket'.

Compare with the following, however: (14) ?cao-r 'grass' *chuang-zi 'bed'

*shu-r'tree' ?/ou-zi 'building'

*qiu-r 'ball' *tie-zi 'iron'

It is impossible to give any hard and fast rule governing the use or non-use of the two suffixes. A rule of thumb is not to use -er unless one is sure of it. But for the use of -zi, each form has to be learned individually.

The case of the prefix xiao- is quite different. Only a very limited number of forms with xiao- are available: xiaogui 'little devil', xiaojiahuo 'little fellow', etc.

And they border on compound words. This prefix is therefore not productive at all except in one instance, i.e. its addition to a surname, e.g. Xiao-Zhang 'Little Zhang', Xiao- Wang' Little Wang', etc. As part of a proper name, it generally has a very weak semantic content of' little/small'. The practice of adding this prefix to a surname has become more popular in mainland China after 1949. Parallel with this practice is the addition of /ao- 'old' to a surname. Thus, Lao-Zhang 'Old Zhang', Lao-Wang 'Old-Wang', etc. Just as Xiao-, Lao- in this use has little semantic content of 'old'. Lao- as a prefix to a common noun, on the other hand, is in fact more like the suffixes -er and -zi in its morphological function. It is added in many cases to a monosyllabic noun just to make it longer, e.g. laohu 'tiger',

laoshu 'mouse', laoshi 'teacher', etc. In these latter cases, it has become an integral part of the words.

Classifiers. They are a category of words used to individuate, specify or quantify a noun. Examples are: yige ren 'a person', sankuai qian "three dollars', neiben shu 'that books'. -Ge, -kuai and -ben are classifiers to individuate or specify the nouns ren 'person', qian 'money' and shu 'book', respectively, so that a numeral or a specifier can precede them. The most general or generic classifier in modem Mandarin is -ge which, with an obsolete meaning of 'individual', can be applied when no specific classifier is available for a given noun, e.g. anything newly introduced into the culture, e.g. yige dianzi chongwu 'a virtual pet'. But most nouns have their own specific classifiers. Some of them seem to be arbitrarily assigned, e.g. yikou guancai 'a coffin', where the classifier -kou means 'mouth'. But many of them are meaningful and have some semantic relation with the noun, e.g. yitiao bandeng 'a bench', yizuo Jangzi 'a building', liangtao jiaju 'two suites of furniture' , where the classifier -tiao indicates 'a long shape', -zuo being derived from the verb zuo 'to sit' refers to 'a large mass or object that sits on the ground', and -tao gives the idea of' a set' of things.

Mass and abstract nouns can also be individuated, specified or quantified by classifiers, e.g.yitan shui 'a pool of water (on the ground)', zhejian shi 'this matter (of business)'. The classifi~rs -tan 'shapeless pool' and -jian 'item' help specify the shape of water and individuate the matter. When a new noun is accepted into the language after the initial introduction, a classifier has to be found from the stock to match the noun, e.g. yitai/jia dianshi 'a television set', where -tai or -jia is for machines. Otherwise, the general -ge may be used improvisionally until a more sophisticated one is found, e.g. yige diannao 'a computer'.

Historically, most classifiers were nouns but have developed into a special class. Many of them can still be used as nouns or have retained some nominal properties. They are here considered inflectional in nature for the reason that they don't change the major grammatical class of the stem to which they are added.

Modem Mandarin has a number of derivational affixes. We discuss some commonly-used ones for illustrative purposes. They are the localizers, the genitive or manner de, and some nominalizing, verbalizing and adjectivalizing affixes. This section is more or less modelled on Li and Thompson (1987) and draws heavily on Chao (1968).

Localizers. They are suffixes added to a common noun for it to be used as a locative noun. A common noun can not serve as a locative noun in the sense that it cannot normally be preceded by locative prepositions (known as coverbs) like zai 'at', dao 'to', cong 'from' and wang 'to, toward'. Compare *zai shujia with zai

shujiashang 'on the bookshelf and *congfangzi chulai with congfangzili chulai 'come out of the house.' The suffixes -shang '(on the) top (ot)' and -Ii 'inside of are the two most commonly used localizers that turn common nouns into locative ones. There is another way to make locative nouns out of common nouns and even pronouns, i.e. by following the noun or pronoun with a locative adverb zher 'here' or ner 'there', e.g. dao wo jer lai 'come to me' «literally, 'to me here come'), cong tushuguan ner zou guochu 'walk past by the library' «literally, 'from library there walk past go'. The locative adverbs jer 'here' and ner 'there' make the pronoun wo 'I' and the common noun tushuguan 'library' into locative words, respectively.

Though the locative adverbs serve the same function as the localizers, they are not generally considered derivational suffixes.

The Genitive and Manner Adverbial -de. 12 The particle -de serves a number of functions, of which we select two for discussion: as a genitive marker and as a manner adverbial marker.

The genitive -de follows all nouns and pronouns to turn them into genitive forms, e.g. wode shu 'my book', nide xiaohai 'your child(ren)', zhuozide tui 'table's leg', kexuede mudi 'objective(s) of science' (literally, 'science's objective'),falude jichu 'foundation of law' (literally, 'law's foundation'), etc. When the noun to which this -de is suffixed is an abstract one, the resultant form can be interpreted as an adjective instead of a genitive, e.g. kexuede fangfa 'scientific method' ,falude wendi 'legal problem', etc. The adjectival -de, however, is often optional in this latter usage.

The manner adverbial -de is most frequently illustrated with a reduplicated form, e.g. manmiirde zou «manman 'slow-slow' + de + zou 'to walk') 'walk good and slow', Mohiiode zuo «MoMo 'good-good' + de + zuo 'to do/work') 'do it properly and well,' etc. In fact, it is not restricted to use with reduplicates. It can also occur with other types of words and phrases, e.g. xiaoxinde kan 'look carefully,' gankuaide mai 'hurry up and buy', etc., where xiaoxin 'careful' is a regular adjective and gankuai is a regular verb.

Affixes Corresponding to English -able, -ize, -ology, -ist, -er, -ness. These are derivational affixes in the same sense their English counterparts are. We discuss six of them in total: ke- '-able', -hua '-ize, -ization', -xue '-ology', -jia '-ist', -zhe '-er (indicating a doer)', -xing '-ness (indicating an abstract quality)'.

The prefix ke- is added to a verb to produce an adjective, e.g. ke'ai 'lovable', kexiao 'laughable', kehen 'hateful', kechi 'shameful', where the stems ai 'to love', xiao 'to laugh', hen 'to hate' and chi 'to feel ashamed' are prefixed with ke- to become adjectives. It is noted that in fact the Mandarin prefix ke- is even more productive than the English suffix -able/-ible.

The suffix -hua is added to a noun or adjective to produce a corresponding verb, e.g. meihua «mei 'beautiful' + hua) 'to beautify', chouhua «chou 'ugly' + hua) 'to uglify', kexuehua( <kexue 'science' + hua) 'to make scientific', Meiguohua 'to Americanize', yitaihua (<yitai 'one fetus' + hua) 'to make "one child per

family" into a policy'. This is again a very productive suffix, even more so than the corresponding -ize and -jj; in English put together.

The suffix -xue '-ology', originally meaning 'to learn', produces words that name a discipline of scholarship, e.g. wulixue «wuli 'principle(s) of matter' + -xue) 'physics', huaxue «hua 'change' + -xue) 'chemistry', shenxue (<shen 'deity' + xue) 'theology', lishixue «lisM 'history' + -xue) 'history (as a discipline)', yuyanxue (<yuyan 'language' + -xue) 'linguistics'. This is another suffix that is very productive.

The suffix -jia '-ist' indicates a person who is an expert in a certain field. Thus, any of the disciplines that end in -xue can take this suffix to mean an expert in that discipline, e.g huaxuejia 'chemist', yuyanxuejia 'linguist', etc. But its application is wider than just fields of learning. It can apply to skills such as playing the piano, writing the novel, etc., e.g. ganqingjia 'pianist', xiaoshuojia 'novelist', huajia '(artistic) painter', etc. The suffix, however, has a limited application to other fields of activities. For example, even thoughyundongjia (<yundong 'athletics' + -jia) 'athlete' is acceptable, *zixingchejia «zixingche 'bicycle' + -jia) for '(bi)cyclist' is not. Since the suffix always carries a tone of expertise, one does not use any term containing it to refer to oneself. For example, it is inappropriate for the present author to refer to himself as a yuyanxuejia, even though it is perfectly appropriate to use the English term linguist in this case.

The suffix -zhe13 'a person who V-es' is somewhat obsolete as a productive suffix, but it appears in many fixed combinations in modem Mandarin, e.g. zhangzhe '(venerable) old person', zuozhe «zuo 'to do, work, create' + -zhe) 'author', bianzhe «bian 'to edit' + -zhe) 'editor',jizhe (<ji 'to record' + -zhe) '(news) reporter',fangwenzhe (<fangwen 'to interview' + -zhe) 'inteviewer', bei fangwen-zhe «bei 'passive marker' +fangwen to 'interview' + -zhe) 'interviewee'.

The last one involves a syntactic process ofpassivization before the suffix is added.

An interesting comparison may be made between the two suffixes -jia and -zhe: While a zuojia is an expert or professional writer, a zuozhe is an author of a specific production or productions.

The suffix -xing basically means 'nature, quality'. It generally combines with bound morphemes to give basic vocabulary like naixing «nai- 'to last for a long time' + -xing) 'patience', duoxing «duo- 'lazy' + -xing) 'laziness',jixing (ji- 'to remember' + -xing) 'memory'. When it is added to a free morpheme, whether a noun, verb or adjective, to form an abstract noun, it takes on the function of a suffix much like the English -ness or -ity. Thus, kexuexing «kexue 'science' + -xing) 'the quality of being scientific', lianxuxing «lianxu 'to continue' + -xing) 'continuity', pubianxing (<pubian 'universal' + -xing) 'universality', kenengxing (keneng 'possible' + -xing) 'possibility'. By analogy, it applies to phrases, e.g. kerongxing «ke rong 'which is soluble' + -xing) 'solubility', kesuxing «ke su 'which is plastic' + -xing) 'plasticity'. The use of this suffix seems to be gaining ground rapidly in writing and it has now become a completely productive one.

Compounding, a very important process in Mandarin Chinese, has attracted a lot of attention in traditional as well as in modem linguistic studies. Almost all studies recognize the fact that there is no clear-cut boundaries between what might be called simple, compound and complex words in modem Mandarin. The usual .criterion of relationship between the meaning of the word as a whole and the meanings of the components is not able to make a clear distinction between the three categories of words, e.g. hudie 'butterfly' and huasheng ÂŤhua 'flower' + sheng 'to be born') 'peanut' unquestionably are single words, tuoniao ÂŤtuo'camel' + niao 'bird') 'ostrich' andjengniao (<feng 'bee' + niao 'bird') 'humming bird' may be conveniently considered complex or compound words. On the other hand, in the case of shuiguo (<shui 'water' + -guo 'fruit') 'fruit', xiangjiao (<xiang 'fragrant' + -jiao 'palm (tree)') 'banana' and such similar forms, are they complex or compound words? The problem is further complicated by differences between colloquial and written styles, between dialects and between plain and technical uses.

(Cf. Tang, 1994:496) For all these reasons, we will only give compounds for illustrative purposes without harping upon the definition of what a compound word is in Mandarin. The words in (IS) show a range of relationship between the meaning of the compound as a whole and those of its components. (Chu, 1983b:3) The meanings of some of the nouns in (IS) are not clearly related to those of the components therein. The closeness of this semantic relationship, as a matter of fact is in descending order from (a) to 0), i.e. it is closer on the top of the list and becomes farther and farther apart as one comes down to the bottom. All this shows (IS) Compound

Word

Component

Meanings

Gloss

a. xinfeng

letter + seal

'envelope'

b. qiiinbf

lead + writing-instrument

'pencil'

c. luguin

travel + hall

'hotel'

d.sushe

stay + lodge

'dormitory'

e. zidian

word + concordance

'dictionary'

f. wentE

ask + topic

'question, problem'

g. zazhi

sundry + journal

'magazine'

h. gangbf

steel + writing-instrument

'pen'

i. xiaoshuo

small + speech

'novel'

that the distinction between compound words and non-compounds in Mandarin is indiscrete and the many degrees of difference may even form a continuum.

In the following, we give a few structural schemata to characterize different types of compounds in modem Mandarin. Compound words may be made up of coordinate components. Under this category, there are several sub-types. All of them seem to apply mostly to nouns, sometimes to verbs, and occasionally to adjectives. (a) Synonymous

Compounds.

As nouns, e.g. pengyou 'companion + friend

= friend', shumu 'tree + wood = trees', jannao 'vexation + angry = vexation, worry', etc. As verbs, e.g. tiaoyao 'jump + leap = to jump', tingzhi 'stop + stop = to stop', gongji 'charge + attack = to attack', etc. As adjectives, e.g. meihao 'beautiful + good = good and beautiful', youliang 'excellent excellent', polan 'broken + rot = broken down', etc.

+ fine

=

good,

(b) Opposite Compounds. As nouns, e.g. maimai 'buy + sell = (commercial) business', kaiguan 'open + close = switch', daxiao 'big + small = size', etc. As verbs, e.g. huxi 'exhale + inhale = to breathe', laiwang 'come + go = to socialize, have dealings (with)', etc. (c) Category Compounds. As nouns, e.g. huacao 'flower + grass = plants', chema 'vehicle + horse = traffic,' jinliang 'catty + ounce = weight', youqi 'oil + paint = paint', etc. (d) Half-Transliteration Compounds. As nouns, e.g.jipuche 'jeep + vehicle luojixue 'logic + discipline = logic (as a field oflearning)' bingqilin 'ice + ice cream = ice cream', Jiazhou 'Ca(lifornia) + state = California', etc.

Compound nouns can be made up of elements with specific semantic relations.

These relations can vary widely and we only give a few of them: (cf. Li and Thompson, 1981:49-53) (a) The first noun denotes a location where the second noun is located or applied, e.g. taideng 'table + lamp', hema 'sea + horse = hippopotamus', yagao 'tooth + paste', mianshuang 'face + frost = facial cream', etc. (b) The first noun denotes the source of the second one, e.g. niunai 'cow + milk',jengmi 'bee + honey', etc. (c) The first noun denotes the material that the second is made of, e.g. caomao 'straw + hat', yinbi 'silver + coin', etc.

As one can easily imagine, such semantic relations are in fact too numerous to be listed exhaustively. Furthermore, they do not seem to be language specific, i.e. of the items in (a}-(c) above, most have quite natural semantic relations, which exit in other languages as well.

Another way of analyzing compounds is to look at the relations between their

= jeep',

components by analogy to syntactic structures. The following are some examples from the syntactic viewpoint, many of which are cited from Chao (1968, Chap. 6) and Tang (1994). (a) Subject-Predicate Compounds. The resultant form can be a noun, a verb or an adjective. As both nouns and verbs, e.g. dizhen 'earth + quakes = earthquake', touteng 'head + aches = headache', etc. As adjectives, e.g. mianshou 'face + is familiar = familiar-faced', mingku 'destiny + suffers = ill-fated (in life)', etc. (b) Verb-Object Compounds. The resultant form can also be a noun, a verb or an adjective. As verbs, e.g. chucha 'go-out + errand = to go on a business trip', guanxin 'to be concerned + heart = to be concerned', touji 'to pitch + opportunity = to take chances', etc. As nouns or adjectives, e.g. zhiji 'to know + self= intimate (friend)', xingzheng 'to carry out + government affairs = administration', etc. (c) Verb-Complement Compounds. The resultant form is always a verb. E.g. kanjian 'to look + to see = to see,' suoxiao 'to shrink + small = to shrink', dadao 'to beat + to tumble = to knock down,' shuopo 'to speak + to break = to break (a secret)', tingdong 'to hear + to underst~.nd = to hear and understand', maidao 'to buy + to obtain = to buy and obtain', ducheng 'to read + to become = to read (something wrongly) as', etc. The semantic relation between the components of this type can range from very close to very loose. The range is shown in the above examples from the close to the loose in that order. As we go down from kanjian 'to see' to ducheng 'to read wrongly as', the structures become more and more like phrases rather than words. There is again no clear-cut division between this verbcomplement compounds and their corresponding phrases. (d) Modifier-Head Compounds. The resultant form may be a verb or a noun.

As a verb or a noun, e.g. duli 'alone + to stand = independence or to be independent', qingshi 'light (in weight) + to look = (to) dispise', dongmian 'winter + to s!eep = 'hibernation or to hibernate', etc. As a verb only, e.g. xiangxin 'mutual + belIeve = to trust, believe',jiazhuang 'false + to dress up = to pretend', etc.

Besides the affixation and compounding discussed above, there are of course other interesting processes, such as reduplication and the addition of the particle de for other functions. But a sketch like this one just has to leave out many facts.

In this section, we will briefly discuss word order, topic vs. subject and some special constructions.

The unmarked basic word order of modem Mandarin is generally recognized as SVO, e.g.

(16) Wo maile (yiben) shu.

I buy-PFV (a-M) book 'I bought a booklbooks.'

(17.a) Shu, wo maile. book, I buy-PFV 'I bought the book(s).' b) W 0 shu maile.

I book buy-PFV 'I bought the book( s).'

In both (17.a) and (17.b), the object shu 'book(s)' precedes the verb for its definiteness.!

For existential/presentative sentences, the basic order is VS, just like that in English, e.g.

(18) Laile hen duo keren. come- PFV very many guest 'There came many guests.' where hen duo keren 'many guests' is the subject but follows the verb laile 'came'.

Word order becomes a little complicated when an additional object or a time duration adverb is involved, e.g.

(19.a) Wo geile ta yiben shu.

I give-PFV himlher a-M book 'I gave himlher a book.' b) Wo gei ta jile yifeng xin.

I to himlher send-PFV a-M letter 'I sent himlher a letter.' c) Wo jile yifeng xin gei ta.

I send-PFV a-M letter to himlher 'I sent a letter to himlher.'

and there is no alternative word order for the sentence. Sentences (l9.b) and (l9.c) are alternative forms of each other when the indirect object ta 'himlher' follows the preposition (or coverb) gei 'to, for'.

A time-duration adverb occurs after the main verb if there is no direct object, as in (20.a) below. When there is a direct object, the time-duration adverb may occupy either of two positions. It may occur after the object but the verb must be repeated before it, as in (20.b) below. Or it may occur between the verb and the direct object with a -de following it, as in (20.c) below.

(20.a) Tamen chile liangge zhongtou Ie. they eat-PFV two-M hour LE 'The have been eating for two hours.' b) Tamen chifan chile liangge zhongtou Ie. they eat-rice eat-PFV two-M hour LE 'They have been eating (meal) for two hours.' c) Tamen chile liangge zhongtoude fan Ie. they eat-PFV two-M hour-DE rice LE 'They have been eating (meal) for two hours.'

The structural difference between (20.a) on one hand and (20.b) and (20.c) on the other is that the former doesn't have a direct object/an 'meal'. The time-duration adverb liangge zhongtou 'two hours' thus follows the verb chile 'ate' in (20.a) while it occurs after the repeated verb in (20.b) and between the verb and the direct object in (20.c). Furthermore, the direct object may occur at the beginning of the sentence, such as in (20.a). Then, the only possible position for the time-duration phrase is the one following the verb directly, e.g.

(20.d) Fan, tamen chile liangge zhongtou Ie. meal, they eat-PFV two-M hour LE 'As for the meal, they have been eating it for two hours.'

Here, the object/an 'meal' is treated as a topic and thus the interpretation is quite different. We will discuss the structure for 'topic' vs. 'subject' in the next subsection.

1.6.3.2. 'Subject' vs. 'Topic' Mandarin Chinese has been recognized as a 'topic prominent' language vis-a vis a 'subject prominent' language, such as English. What this means is that while in English 'subject' is a very important structural notion, in Chinese 'topic' plays an equally, if not more, important role in terms of structure. In order to see how the Mandarin topic fimctions, it may be helpful to fIrst look at what a typical subject is

like in English. A typical subject in English may be characterized by three properties. That is, a nominal is considered a subject if it (a) bears a doing or being relationship with the verb/predicate, (b) agrees with the verb, and (c) occupies the preverbal position. Thus, the subjects in the sentences in (21) below become less typical, going from (a) to (c).

(21.a) John runs three miles a day. b) The door opens from the outside. c) There are three people there.

The subject of (21.a), John, bears a 'doing' relationship with the verb run, agrees with it, and is in the preverbal position. The one in (21.b) does not have a doing or being relationship with the verb, though it agrees with it and is in the preverbal position. The one in (21.c) just agrees with the verb.

On the other hand, since there is no agreement in Mandarin, there are only two subject properties for Chinese, e.g.

(22.a) Lao Zhang meitian paobu pao san yingli.

Old Zhang everyday run-step run three mile 'Old Zhang runs three miles a day.' b) Zheshan men cong waimian kai. this-M door from outside open 'This door opens from the outside.' c) N er you sange ren. there there-be three-M person 'There are three people there.'

The noun phrase Lao Zhang 'Old Zhang' in (22.a) bears a doing relationship with the verb paobu 'to run' and is preverbal. Its subject status is thus uncontroversial.

The noun phrase zheshan men 'this door' in (22.b) is preverbal but does not bear a doing or being relationship. Its subject status is quite controversial. The noun phrase sange ren 'three people' in (22.c) does not bear a doing or being relationship nor is it in the preverbal position. It is therefore not recognized as a subject at all.

A noun phrase like zheshan men 'this door' in (22.b) is sometimes considered a topic. Pragmatically, a topic is something that the speaker brings up to be talked about. (But see Chapter 7 for more discussion.) If so, any nominal mentioned at the beginning of a clause, including the subject, can unquestionally be treated as a topic, e.g. the sentence-initial nominals in (21.a), (21.b) and (22.a), as long as they are followed up on by something else. Conceivable follow-ups for (21.a) and (21.b) are:

While the two nominals John and the door are subjects within the boundaries of the first clause, they serve as topics across the boundaries because they are what is brought up by the speaker to be talked about. When a topic occurs in the second clause, it is usually referred to by a pronoun, as in (21.b'), or by a zero anaphor, i.e. deleted as in (lO.a'). Similarly, sentences (22.a) and (22.b) can be expanded as (22.a') and (22.b'), respectively.

(22.a') Lao Zhang meitian paobu pao san yingli, suoyi shenti hen hao.

Old Zhang everyday run-step run 3 mile, so body very good 'Old Zhang runs three miles a day, so (he is) in good health.'

(22.b') Zheshan men cong waimian kai, suoyi ni dei you yaoshi. this-M door from outside open, so you must have key 'This door opens from the outside, so you must have a key (to open it).'

The nominals Lao Zhang 'Old Zhang' and zheshan men 'this door' are brought up first and then talked about. They are clearly the topics in (22.a') and (22.b'). As the topic is what is brought up to be talked about, the rest of the clause that is used to talk about the topic is known as comment.

For the sole purpose of showing the differences between the topic and the subject, most grammars of Chinese mainly discuss another kind of topic. This kind is what we will later call 'marked topics' in Chapter 7. They are marked by phonological, morphological and/or syntactic devices. For example,

in (23.b) is phonologically marked by a pause, which is graphically represented by a comma. Zuo maimai dejihua 'the plan of doing (commercial) business' in (23.c) is morphologically marked by the preposition guanyu 'concerning, as to' and may be optionally marked by a phonological pause. Zhe sanshige xuesheng 'these thirty students' in (23.d) is syntactically marked by being the first of two consecutive nominals in front ofthe verb and may also be optionally marked by a phonological pause. Example (23.a) is supposed to be an utterance by a parent. Then, a literal translation of 'We are two boys and a girl' doesn't make sense. Therefore, the sentence-initial pronoun women 'we (the parents)' has to be treated as a topic-i.e. 'as for us'. Though this topic is not overtly marked, it is completely possible for a pause to follow it. If a pause does follow it, it is then considered to be marked phonologically. In all cases, a pause may always be accompanied by the particle a/ya or ne to make the marking more prominent.

Mandarin Chinese is recognized as a topic-prominent language for the reason that there is a preponderance of occurrences of topics and that the subject is not required for a sentence. (Also see the discussion under 'Double Subject' and 'Comparison' later in the following sub-section.) It is to be noted, however, that the subject is a syntactic notion while the topic is mainly a discourse notion. For this reason, we will discuss the notion oftopic in some detail in a separate chapter later.

1.6.3.3. Special Constructions This sub-section discusses five special constructions: the ba-construction, construction, the 'double subject', verb-copying, and comparison.

the bei-

(23.a) Women shi liangge nanhai, yige nilhai. we be two-M boy, one-M girl 'As for us, there are two boys and a girl.' b) Ren, zuihao bie tai tanxin. person, best don't too greedy 'Better not be too greedy.'

c) Guanyu zuo maimai de jihua, women yihou zai tan. concerning do business DE plan, we later again talk 'As to the plan of doing business, we will discuss (it) later.' d) Zhe sanshige xuesheng, ershibage bu jige. this 30-M student, 28-M not pass 'Of these thirty students, twenty-eight failed (the test).'

where ba is the preposition (or coverb) under discussion and Comp stands for 'complement of the verb', e.g. (Chu, 1983b:206-7) (24.a) Women ba gongke zuowan Ie. we BA homework do-finish PFV 'We have finished (doing) the homework.' b) Qing ni ba yizi nachuqu. please you BA chair take-out-go 'Please take the chair(s) out.' c) Wo bu yuanyi ba zheme duo qian doujie gei ta.

I not willing BA this much money all lend to him/her 'I am not willing to lend him/her all this money.'

d) Haizimen ba jiali nong de luanqibazao. children BA house-in made DE in-a-big-mess 'The children made a big mess of the house.'

The complement may take various forms. It may be a state verb, e.g. wan 'finish' in (a), a directional complement, e.g. chuqu 'out(-go)' in (b), a prepositional phrase, e.g. gei ta 'to himlher' in (c), or a whole predicate introduced by the particle de, e.g de luanqibazao 'DE (be) in-a-big-mess' in (d). Other forms are also possible, though less frequently used. Besides, there are quite a few semantic restrictions. The subject must be an agent, the object must be a patient and the complement must be a result ofthe action indicated by the verb. Since the construction usually encodes a specific event, the object is generally interpreted as defmite or, at least, referential.

Thus, the following sentence is not quite acceptable: (25) ?Women zuotian wanshang ba dianying kanle. we yesterday evening BA movie watch-PFV 'We watched a movie last night.' (intended meaning) The sentence is not readily acceptable just because no result generally comes out of watching a movie. But, some situations may sanction the use of (25). For example, we had wanted to see a movie for a long time and finally we got a chance to see it last night. Then, (25) is interpreted as expressing a result of realizing a long-cherished wish, even though the result is only implied by the perfective aspect marker Ie. In this scenario, the object dianying 'movie' must be interpreted as definite.

This structure has been known as the 'disposal' construction and Thompson (1973) characterizes it as capable of answering the question 'What did X do to Y?' where X is the agent subject and Y is the patient object. Its pragmatic importance is to indicate that some action is/was taken in regard to a specific thing, person or matter and that a result has/had been produced due to the action. In technical terms, the construction is mainly to increase the 'transitivity' of the statement. (Cf. Hopper and Thompson, 1980; Sun, 1995.)

The Bei-Construction. This construction is known as the passive and takes the following syntactic form:

where the object is optional. In this construction, the subject is the patient and the object is the agent, just like the English passive. The preposition bei is therefore often equated to the English counterpart by. The following examples are adapted from Chu (1983b:216).

(26.a) Qiaokeli tang dou bei (xiao didi) chiwan Ie. chocolate candy all BEl (little younger-brother) eat-fmish PFV 'All the chocolate was/has been eaten up (by Little Brother).' b) Neizhang hua bei (ren) banchuqu Ie. that-M Painting BEl (person) move-out-go PFV 'That painting was/has been moved out (by somebody).' c) Shu dou bei (ta) song gei bieren Ie. book(s) all BEl (himlher) give to other-people PFV 'All the books were/have been given away (by him/her).' d) Jiali bei (haizimen) nong de luanqibazao. house-in BEl (children) make DE in-a-big-mess 'The house was/has been made a big mess (by the children).'

As with the ba-construction, the complement can be of various forms. As far as the semantic restrictions are concerned, the complement must similarly be the result of the action indicated by the verb. The semantic roles played by the subject and object are, however, just the reverse of those of the ba-construction, i.e. the subject must be the patient and the object must be the agent, if present at all. As a result, the patient subject is generally interpreted as definite or referential, though the agent object mayor may not be so interpreted.

So far, the bei-construction may appear quite similar to the English actionpassive (i.e. a passive that involves an action) except that there must be a complement. But there is a major difference between the two: the Chinese beiconstruction has a strong pejorative reading, i.e. the event is unfortunate. In this sense, it is very close to the get-passive in English. In some speakers, especially the younger generations in Taiwan, the pejorative reading is less binding, i.e. the beiconstruction in those speakers is not necessarily interpreted as having a pejorative sense.

The 'Double Subject'. This structure is characterized by two or more preverbal nominals without any formal marking to relate them to each other.

Sentence (23.d), which is repeated below, is an example ofthis so-called 'double subject'.

(23.d) Zhe sanshige xuesheng, ershibage bu jige. this 30-M student, 28-M not pass 'Of these thirty students, twenty-eight failed (the test).'

The term, however, is a misnomer because it recognizes both nominals as subjects, in this case, zhe sanshige xuesheng 'these thirty students' and ershibage 'twentyeight (of them)'. In fact, as we have seen above, at least one of them must be treated

as topic instead of as a subject, although the other can be a genuine subject. In (23.d), zhe sanshige xuesheng 'these thirty students' is the topic and the rest of the sentence is the comment. The comment, however, is a full clause by itself. It contains the nominal ershibage 'twenty-eight (of them)' as the subject and bujige 'not pass' as the predicate. There is a chance that ershibage may also be used as a topic at the same time. For example, (23.d) may very well be followed by another clause like the following: (27) yiding dei bukao. definitely must make-up-exam '(they) must take a make-up exam.'

Verb-Copying.

This structure involves the repetition of a verb in the same clause under certain conditions:

Strictly speaking, (23.d) has a structure of 'Topic + [com Subject + Predicate]', where the subscript com indicates a comment. But when it is followed by (27), the subject in the comment performs the dual function of being a topic at the same time.

The topic-comment structure of(23.d) plus (27) can be indicated by the following:

The bold square brackets designate the first layer in the topic-comment structure and the regular square brackets designate the second layer. As indicated, the topic in the first layer is zhe sanshige xuesheng 'these thirty students' and all the rest is its comment. In the second layer, ershiba ge 'twenty-eight (of them)' is the topic and it has two comments: bujige 'fail (an exam)' andyiding dei bukao 'definitely must take a make-up exam' .

It can now be seen that what has been called 'double subject' actually consists of a topic and another nominal which may either be a subject or another topic or both.

In fact, more than two preverbal nominals can occur in a clause, e.g.

(29) Women, liangge nanhai, yige basui, yige shisui. we, two-M boy, one-M 8-year-old, one-M ten-year-old 'We (have) two boys~ne (is) eight years old, the other ten.'

(30)

women] [com [top liangge nanhai] [com (top yige) (com shisui)]]

[top

other) ten years old'. In the third layer, which has two parallel topic-comment substructures, the first yige 'one' is a topic which takes basui 'eight years old' as its comment, and the secondyige 'one' is another topic which takes shisui 'ten years old' as its comment.

We have so far illustrated a very interesting fact that the topic-comment structure in modem Mandarin can in theory have an infinite number of layers, just as embedding in the syntactic structure of a sentence. But, of course, anything more than three or four layers will be out of the reach of a regular speaker's cognitive domain.

[com (top

yige)

(com

basui)]

There are three layers in the topic-comment structure in (30). They are designated by bold square brackets, regular square brackets and parentheses, respectively. In other words, women 'we' is the topic ofthe top layer and the rest of the utterance is its comment. In the second layer, liangge nanhai 'two boys' is the topic, which has two comments:yige basui 'one (is) eight years old' andyige shisui 'one (i.e. the

Actually, we have already seen an example of verb-copying in (20.b), which is repeated below.

(2o.b) Tamen chifan chile liangge zhongtou Ie. they eat-meal eat-PFV two-M hour LE 'They have been eating (meal) for two hours.'

In this example, the verb chi 'to eat' is copied (i.e. repeated) after the object -fan 'meal' so that a time-duration phrase can follow it directly. This is a typical case of when the verb needs to be copied. There are other circumstances where verbcopying is necessary. E.g.

(31.a) Ta fuqin xiezi xie de haojile. he father write-character write DE good-extremely-LE 'His father writes characters extremely well, i.e. is a good calligrapher. ' b) Women zhege libai malCal maile sanci Ie. we this week buy-groceries buy-PFV three-times LE 'We did grocery-shopping three times this week.' c) Nimen song huo tongchang song dao naxie difang? you deliver goods generally deliver to what-PL place 'What kind of places do you deliver (goods) to?'

In (31), the verbs are copied after the direct to follow them directly. The verbs, direct write', zi '(Chinese) character(s)' and haojile cai 'groceries' and sanci 'three times' in (b);

objects in order for the complements object and complements are: xie 'to 'extremely well' in (a); mai 'to buy', and song 'to deliver', huo 'goods' and

dao naxie difang 'to what all places' in (c). The complements are of different kinds.

Haojile 'extremely well' is a manner/result adverbial, sand 'three times' is a frequency adverb, and dao naxie difang 'to what all places' is a directional phrase.

It is to be noted that with the latter two kinds of complements, i.e. frequency and directional, verb-copying is optional.

Another interesting fact about verb-copying is that if there is a need for an aspect marker, it is affixed to the copied verb rather than to the fIrst one. Compare the following pairs.

(20.b) Tamen chifan chile liangge zhongtou Ie. b') *Tamen chilefan chi liangge zhongtou Ie. (31.b) Women zhege libai maicai maile sanci Ie. b') *Women zhege libai mailecai mai sanci Ie. What this means is that the portion of the structure containing the copied verb (i.e. chile liangge zhongtou Ie 'ate for two hours' or maile sand Ie 'bought three times' in our case) is actually the main predicate rather than just a copied image of the fIrst verb. And the other portion containing the frrst verb (i.e. tamen chifan 'their eating (meal)' or women zhege libai maicai 'our grocery shopping this week' in our case) is the topic to be commented on. An interpretation of this kind naturally explains why the aspect marker must be attached to the second form of the verb.

Comparison. Li and Thompson (1981:564) give the following as the basic pattern of comparative sentences in Chinese:

(32.a) Tamen bi ni congming. they compare-with you intelligent 'They are more intelligent than you.' b) Tamen meiyou ni (neme) congming. they not-have you (that) intelligent 'They are not as intelligent as you are.' c) Tamen gen ni yiyang congming. they with you same intelligent 'They are as intelligent as you are.'

The X and Y that are being compared in the examples are tamen 'they' and ni 'you'.

The dimension is congming 'intelligent'. Three different comparison words with or

without an adverb are used to represent three types of comparative expressions. The 'more than' type in (32.a) only uses the comparison word bi 'compared with', the 'less than' type in (32.b) uses the comparison word meiyou 'not have' with an optional adverb neme 'that (much)', and the 'same as' type in (32.c) uses the comparison word gen 'with' plus the obligatory adverb yiyang 'same'.

There is, however, one property that distinguishes the comparison structure in Mandarin from its counterpart in other languages. That is, only topics can be compared. In other words, anything in the non-topic position cannot enter into the comparison structure. For example, English may compare direct objects: He speaks English better than (he does) French. A corresponding Mandarin sentence is the following: (33) Ta shuode Yingwen bi (ta shuode) Fawen hao. he speak-DE English compared-with (he speak-DE) French good where Yingwen 'English' and Fawen 'French' are made into topics. It literally means 'The English he speaks is better than the French (he speaks).' Furthermore, if what is being compared cannot be made into topics, such a comparison has to be expressed in a round-about way. The Chinese version of the English sentence I love truth more than I love my teacher is (34) below: (34) Wu ai wu shi, wu geng ai zhenli.

I love my teacher, I even-more love truth 'I love my teacher, (but) I love truth even more.'

It cannot be expressed by a comparison sentence. This is perhaps one of the reasons why Mandarin Chinese is felt to be topic-prominent.

In this section, we have presented a very short sketch of modem Mandarin Chinese. For a more complete description of the language, the reader is referred to Chao (1968), Li and Thompson (1981) and Chu (1983b). A short introduction of the writing system is to be found in Appendix B.

1. The word 'sentence' is used here as a pre-theoretical term for Mandarin Chinese, as there has not been any rigorous defmition for it in the literature of Chinese linguistics. We will attempt to defme what a sentence is like in Mandarin in more tangible terms than just 'expressing a complete thought' in Chapter 9. Until then, the term will be used in a very loose sense.

2. Liushuyu is meant to denote a sentence structure where clauses are lineally arranged without a hierarchical structure. It is said to be the characteristic way of combining Chinese clauses into sentences. (See, for example, Shen, 1988:447.)

This view is based on the preoccupation that Chinese is non-inflectional and

therefore any notion expressed by inflection in Western languages is not applicable to Chinese. We will have more comments on such views in Chapter 9.

3. The term 'discourse grammar', from now on, will be used in the sense as defmed at each progressive stage. As far as I know, the term has not been utilized by any other author with a specific definition. I have used it as the title of a graduate linguistics course at the University of Florida and the National Chengchi University in Taipei to designate an area of study delineated here.

4. Of course, there is a dilemma when modality is considered. Although modality (which is speaker attitude and, sometimes, a guess on the hearer attitude) is explicitly expressed by sentence particles and modality adverbs in Mandarin, there is a lot more than those structural signals.

5. But see Chen (1986 and 1987).

6. The term 'theme' is being used in its loose sense to show 'what is being talked about as the main idea.' We will defme terminology of this kind as we go

on.

7. One of my students and Liu and Chu (1993) have found that one of the major differences between Chinese and English in grammar and discourse is that many discourse markers in Mandarin are not as much grammaticalized as their counterparts in English.

8. 'Grounding' is here used to designate certain ways of relating one term or proposition to another in a given discourse. Fox and Thompson (1990:300) recognize three central kinds of grounding: anchoring, main-clause grounding and proposition-linking.

It is to be distinguished from the term 'grounding' in connection with foreground and background in narrative structure.

9. Cf. Sanders and Wirth (1985:9).

10. There is certainly some pioneering work done in this area, such as by Liao Qiu-Zhong in the 1980s, which can be considered purely discourse. Most of his publications reappeared in Liao (1992). Other works such as by Tsao, Li & Thompson, and Chen Ping mostly deal with syntactic problems seeking discourse solutions.

11. Phonetic changes in the root due to the addition of the -er suffix are not accounted for in this list.

12. We here ignore the different character used in orthography for the manner adverbial in some styles of writing.

13. This suffix has a different graphic representation from the durative aspect -zhe. Furthermore, it is accented and carries the third tone: zhe.

14. Defmiteness, of course, has to do with given information and topicality. But here we only present the syntactic structure of a sentence in the traditional manner.

VERBAL AFFIXES: ASPECT AND DISCOURSE FUNCTION Verbal affIxes in Mandarin Chinese have, in the past thirty years or so, been treated mostly as aspect markers. The prefix zai- has been recognized as the progressive aspect marker and the suffixes -guo, -Ie, -qilai and -zhe have been regarded as marking experiential, perfective, inchoative and durative aspects, respectively.)

As aspect markers, the affixes have been given quite a bit of attention in grammatical studies, especially in the past decade. Most of the treatments are semantically based, largely within the theoretical model of Western aspectual system. Beginning in the mid-70s, works like Chu (1978), Tai (1984), Szeto (1989), He (1992) and Smith (1993) have made some effort to develop a framework to accommodate the peculiarities of the Chinese system. More recently, as the pragmatic notion of 'point of view' is incorporated in the theory as a means for arriving at a better understanding of the aspectual functions in general, Mandarin Chinese researchers have also taken advantage of this development and followed the two-component theory of 'viewpoint' and 'situation type' in their study of the aspectual system. (E.g. see Smith, 1993.)

In this chapter, we will first summarize some of the recent aspectual studies of the Mandarin verbal affixes. Many issues are found to either remain unsolved or have been unsatisfactorily treated in those studies. We then try to look at the problems from a different perspective-a discourse one. We hope to show that from the viewpoint of discourse, some of the unsolved issues can be given a head start and some others may find a more natural explanation than before. Finally, we try to construct a network within which the Mandarin Chinese aspect markers interact to produce discourse cohesion.

2.1. Situation Types: Verbal Meaning Since grammatical aspects are viewed in terms of two components: situation type (verb semantics) and viewpoint (pragmatic perspective), it is admittedly impossible to attain a thorough understanding of the workings of aspect without an accurate knowledge of either component. We will thus review the semantics of Mandarin Chinese verbs before the discussion of their aspectual consequences. In our review, we mainly follow Smith (1993), which is ultimately based on Vendler (1967).

Chinese verbs may be categorized into five situation types: State, Activity, Semelfactive, Achievement, and Accomplishment. They are defmed on the basis of

therefore any notion expressed by inflection in Western languages is not applicable to Chinese. We will have more comments on such views in Chapter 9.

3. The term 'discourse grammar', from now on, will be used in the sense as defmed at each progressive stage. As far as I know, the term has not been utilized by any other author with a specific defmition. I have used it as the title of a graduate linguistics course at the University of Florida and the National Chengchi University in Taipei to designate an area of study delineated here.

4. Of course, there is a dilemma when modality is considered. Although modality (which is speaker attitude and, sometimes, a guess on the hearer attitude) is explicitly expressed by sentence particles and modality adverbs in Mandarin, there is a lot more than those structural signals.

5. But see Chen (1986 and 1987).

6. The term 'theme' is being used in its loose sense to show 'what is being talked about as the main idea.' We will define terminology of this kind as we go on.

7. One of my students and Liu and Chu (1993) have found that one of the major differences between Chinese and English in grammar and discourse is that many discourse markers in Mandarin are not as much grammaticalized as their counterparts in English.

8. 'Grounding' is here used to designate certain ways of relating one term or proposition to another in a given discourse. Fox and Thompson (1990:300) recognize three central kinds of grounding: anchoring, main-clause grounding and proposition-linking.

It is to be distinguished from the term 'grounding' in connection with foreground and background in narrative structure.

9. Cf. Sanders and Wirth (1985:9).

10. There is certainly some pioneering work done in this area, such as by Liao Qiu-Zhong in the 1980s, which can be considered purely discourse. Most of his publications reappeared in Liao (1992). Other works such as by Tsao, Li & Thompson, and Chen Ping mostly deal with syntactic problems seeking discourse solutions.

II. Phonetic changes in the root due to the addition of the -er suffix are not accounted for in this list.

12. We here ignore the different character used in orthography for the manner adverbial in some styles of writing.

13. This suffix has a different graphic representation from the durative aspect -zhe. Furthermore, it is accented and carries the third tone: zhe.

14. Defmiteness, of course, has to do with given information and topicality. But here we only present the syntactic structure of a sentence in the traditional manner.

VERBAL AFFIXES: ASPECT AND DISCOURSE FUNCTION Verbal affixes in Mandarin Chinese have, in the past thirty years or so, been treated mostly as aspect markers. The prefix zai- has been recognized as the progressive aspect marker and the suffixes -guo, -Ie, -qUai and -zhe have been regarded as marking experiential, perfective, inchoative and durative aspects, respectively.)

As aspect markers, the affixes have been given quite a bit of attention in grammatical studies, especially in the past decade. Most of the treatments are semantically based, largely within the theoretical model of Western aspectual system. Beginning in the mid-70s, works like Chu (1978), Tai (1984), Szeto (1989), He (1992) and Smith (1993) have made some effort to develop a framework to accommodate the peculiarities of the Chinese system. More recently, as the pragmatic notion of 'point of view' is incorporated in the theory as a means for arriving at a better understanding of the aspectual functions in general, Mandarin Chinese researchers have also taken advantage of this development and followed the two-component theory of 'viewpoint' and 'situation type' in their study of the aspectual system. (E.g. see Smith, 1993.)

In this chapter, we will first summarize some of the recent aspectual studies of the Mandarin verbal affixes. Many issues are found to either remain unsolved or have been unsatisfactorily treated in those studies. We then try to look at the problems from a different perspective-a discourse one. We hope to show that from the viewpoint of discourse, some of the unsolved issues can be given a head start and some others may find a more natural explanation than before. Finally, we try to construct a network within which the Mandarin Chinese aspect markers interact to produce discourse cohesion.

2.1. Situation Types: Verbal Meaning Since grammatical aspects are viewed in terms of two components: situation type (verb semantics) and viewpoint (pragmatic perspective), it is admittedly impossible to attain a thorough understanding of the workings of aspect without an accurate knowledge of either component. We will thus review the semantics of Mandarin Chinese verbs before the discussion of their aspectual consequences. In our review, we mainly follow Smith (1993), which is ultimately based on Vendler (1967).

Chinese verbs may be categorized into five situation types: State, Activity, Semelfactive, Achievement, and Accomplishment. They are defmed on the basis of

temporal semantic features of'telic' vs. 'atelic', 'durative' vs. 'instantaneous' and 'static' vs. 'dynamic'. Table 1 gives the feature composition of each ofthe verbal categories. Notationally, [-Dynamic] replaces 'State', [-Telic] replaces 'Atelic', and [-Durative] replaces' Instantaneous'. It is quite clear from the feature specifications that the general distinction of State Verb vs. Event Verb is made on the basis of the feature [+Dynamic] vs. [-Dynamic] and that Event Verb is further categorized into several sub-types.

The situation types may thus be characterized in the following way. Examples are given at the end of each type. (i) A State Verb represents a state that simply exists (i.e. [-Dynamic]) over a duration oftime (i.e. [+Durative]) with no change or result involved (i.e. [-Telic]).

E.g. you 'have, exist', zhidao 'know', gaoxing 'be happy', hao 'be good" xihuan 'like', etc.

All other types represent events that happen (i.e. [+Dynamic]). (ii) An Activity Verb represents an event that happens (i.e. [+Dynamic]) and lasts for some time (i.e. [+Durative]). The event itself does not involve any change or result (i.e [-Telic]). E.g. pao 'run', kan 'look', shuo 'talk', gaosu 'tell', chi 'eat', etc. (iii) A Semelfactive Verb represents an event that happens (i.e. [+Dynamic]) but does not last for any discernible stretch of time (i.e. [-Durative]). The event itself does not involve any change or result (i.e. [-Telic]). E.g. ti 'kick', qiao 'knock', fang 'put', etc. (iv) An Achievement Verb represents an event that happens (i.e. [+Dynamic]) but does not last for any discernible stretch of time (i.e. [-Durative]). The event itself involves some change or result (i.e. [+Telic D. E.g. dapo 'break', shuizhao

'fall asleep', etc.2 (v) An Accomplishment Verb represents an event that happens (i.e. [+Dynamic]) and lasts for some time (i.e. [+Durative]). The event itself involves some change or result (i.e. [+Telic]). E.g. gai yisuo fangzi 'build a house', zou huijia 'walk back home', etc.

With this semantic picture of the verbs in mind, we are ready to discuss the 'point of view' component of the aspectual system of Mandarin Chinese.

Before going into the discussion of 'point of view', we must mention another important work, i.e. He (1992), on the aspects and situation types of Mandarin verbs. It is a culmination of research done through early 1990s in this area. It contains a detailed classification of Mandarin verbs and explains how they do or do not fit into the conventional situation types. Its main theme is the interactions between the aspetual meanings and the classes and sub-classes of verbal situation types. The major difference between this work and our present study is one of approach and focus. He (1992) takes a semantic approach and its focus is on the interpretation of the structure ofa clause/sentence. We take a discourse approach and our focus is on discovering relationships between clauses. In spite of the difference and some overlapping, the two works often complement rather than conflict with each other. The reader is therefore strongly advised to refer to He (1992) whenever more detail about verb classification and situation type is desirable.

The 'point of view' component of the aspectual system is a complex device for presenting a situation or event as the speaker/writer views it. A situation (i.e. a state) may be viewed as existing at present (e.g. I am busy), having existed in the past (e.g. I was busy), or to be existing in the future (e.g. I will be busy). Similarly, an event may be viewed as happening at present (e.g. I am walking), having happened in the past (e.g. I walked/was walking) and to be happening in the future (e.g. I will walk/be walking).

Though a situation generally lasts for some time, the speaker/writer may still want to make an explicit statement about its existence over a duration of time (e.g. I have been busy). On the other hand, some events may by nature last for some time (e.g. those expressed by activity and accomplishment verbs), others start and terminate instantaneously (e.g. those expressed by semelfactive and achievement verbs). The former may thus be viewed as continuing (e.g. I am walking; I am writing a letter) or as having continued for some time (e.g.

I have been walking, I have been writing a letter). The latter may not be viewed as continuing over time. If the verbs expressing semelfactive and achievement occur in the 'continuing' form at all, they take on various interpretations (e.g. They are arriving late 'Predicting that they will arrive late'; They have been arriving late 'Stating that they have repeatedly arrived late'). Furthermore, a situation or event may be viewed at its beginning point (e.g. I began/will begin to be busy; I

began/will begin to read) or at its end point (e.g. I stopped/will stop being busy; I finished/will finish reading).

From the brief sketch above, it is clear that, in English, points of view are most often expressed by varying the forms of the verb and/or the auxiliaries associated with it. But some viewpoints, such as the beginning and end points, are expressed by additional verbs. The Mandarin aspectual system works under the same principle, as far as form is concerned.

Aspects in Mandarin Chinese, as in all other languages, are subject to constraints on the basis of the semantics of the verb. For example, English state verbs are known to be incompatible with the progressive aspect in English (e.g. *He is knowing me). Ifany such form occurs at all in a specific context, it is interpreted as taking on some dynamic sense, i.e. 'He is beginning to know me'. (Cf. Binnick, 1991 :282ft) Similarly, Chinese state verbs typically do not occur in the progressive or durative aspect, e.g. * Wo zai dong/dongzhe zhege wenti 'I am understanding this problem'.

In this section, we will look at some of the previous analyses of the meanings of each of the aspect markers in Mandarin, explain some of the constraints, and decide what residual problems there are in the treatments.

2.2.1. The Experiential -Guo and the Inchoative -Qilai The suffixes -guo and -qilai are treated in the same section here not because they are semantically or pragmatically similar, but because there is less controversy on their analyses. The suffix -guo is generally recognized as 'experiential' or 'indefinite past' and the suffix -qilai is generally recognized as 'inchoative'. Not only is the terminology widely accepted, but its interpretation is also uncontroversial. Though -guo may be better discussed together with the perfective -Ie for their similarities, it is presented here in its own right and is to be followed by a comparison with -Ie.

2.2.1.1. Experience as a Viewpoint The suffix -guo has been acknowledged as marking the experiential aspect (Smith 1993; Mangione and Li, 1993) or the indefmite past (Chao, 1968:251). We choose the term experiential' for its brevity and ease of comprehension.

In general, an experience is a prior event or events (rarely a state or states) that is/are relevant to the present or speech time. Linguistically, an experiential aspect is the viewpoint of 'specifYing the occurrence of a situation as one of a class of occurrences' (Smith, 1993:10).3 This characterization can be more plainly stated as specifYing that an event happened at least once prior to a reference time (Li and Thompson, 1981 :228-9; Mangione and Li, 1993:83-86). The happening of the event, however, is viewed as stative rather than dynamic (Smith, 1993:13). It is perhaps due to this last view that the aspect is used to refer to an experience.

The following examples illustrate the main characteristics and constraints of the experiential aspect marked by -guo in Mandarin: (In some of the examples, adjustments are made for easy reading and comprehension.)

(2.a) Ni chi-guo yuchi meiyou? (Chao, 1968:251) you eat-EXP shark's-fin not 'Have you ever eaten shark's fin?' b) Lao Zhangjie-guo hun meiyou? (L&T, 1981:226) Old Zhang marry-EXP marriage not 'Has Old Zhang ever been married?' c) Wo shuaiduan-guo tui. (L&T, 1981:227) I fall-break-EXP leg 'I once fell and broke my leg.' d) Bill qu-guo Meiguo.

Bill go-EXP U.S. 'Bill has been to the U.S.' (to be distinguished from 'Bill has gone/went to the U.S.') It is clear from the translations that 'experience' is generally denoted by the present perfect in English, though there may be some variations.

B. The experiential 1968:439)

-guo is incompatible

with the perfective

-Ie: (Chao,

(3.a) Ta qu-guo (*-Ie) Xianggang. (L&T, 1981:204) she/he go-EXP (-LE) Hong Kong 'She/He has been to Hong Kong.'

b) Ni zou-guo-Ie neitiao qiao jiu dao Ie. you walk-past-LE that-M bridge then arrive LE 'When you walk past that bridge, you are there.'

There is, however, another -guo which almost always occurs with -Ie. Fang (1992:458-9) distinguishes it from the experiential aspect -guo by its inability to cooccur with the adverb cengjing 'once; ever'. E.g.

c) Guangju kan-guo-Ie xin, ba Lao Yang Tongzhi rang dao gongsuo.

Guangju read-GUO-PFV letter, BA Old Yang Comrade let-in to office 'After Guangju read the letter, he asked Comrade Yang into the office.'

This -guo seems to have the same meaning of 'past' but its scope is over the event rather than over any physical object as in (b) above. Together with the perfective -Ie it helps strengthen the temporal sequencing of the events.

C. The experiential -guo differs from the perfective -Ie in that the former explicitly claims that the state of being associated with the happening is over: (L&T, 1981 :228-9)4 (4.a) Ta qunian dao Zhongguo qu-Ie. shelhe last-year to China go-LE 'She/He went to China last year.'

b) Ta qunian dao Zhongguo qu-guo. shelhe last-year to China go-EXP 'She/He went to China last year.'

Sentence (4.b) says that herlhis going to China happened at least once last year and that the state ofherlhis being in China is over now. On the other hand, (4.a) says that herlhis going to China last year happened but it doesn't say whether the person is still in China or not. Thus, only (4.a) may be followed by hai meiyou huilai 'has not come back yet'. (See further discussion under Section 2.2.1.2. below.)

D. The experiential -guo has restricted occurrence with certain verbs: (Chao, 1968:668; L&T, 1981:230)

d) Yingwen wo hui shuo-guo, keshi xianzai wangle.

English I can speak-EXP but now forget-LE 'I used to be able to speak English, but have forgotten it.'

L&T claim that (5.a & b) are not acceptable because dying and being old are not repeatable. However, the following utterances actually occurred in speech and writing: (6.a) Wo si-guo haojici.

I die-EXP quite-a-few-times 'I died quite a few times.' b) Women dou cengjing nianqing-guo. we all once young-EXP 'We all were young once.'

Sentence (6.a) occurred in a personal story of several brushes with death. Though it seems like a counter-example of(5.a), si 'death' actually is used in a metaphorical sense. But (6.b) is a real counter-example ofL&T's explanation, because nianqing 'young' is not anything repeatable, either. What underlies the acceptability of sentence like (5.a & b) and (6.a & b) is whether the situation still pertains. If it does, it cannot be taken as an experience. Once one is old, one cannot become 'not old' any more. But one is no longer young after youth and thus being young can be taken as an experience. These examples in fact supports the statement that the -guo form explicitly states that the happening is over.

Chao thinks that (5.c & d) are rare examples where -guo occurs with a stative verb (hao 'good') and a stative predicate (hui shuo 'able to speak'). As a matter of fact, there is no reason why it is hard for the experiential -guo to occur with any stative verb or predicate. The problem may also be whether the state still exists or not at the speech time.

(5.a) *Ta si-guo. shelhe die-EXP '*Shelhe died before.' b) *Ta lao-guo. shelhe old-EXP '*She/He has been old before.' c) Tamen dui wo conglai mei zeme hao-guo. they to me ever not so good-EXP 'They have never been so good to me.'

Smith (1993: 13) claims that experiential sentences are stative in situation type. She gives the following to characterize the nature of -guo.

She further maintains that as such, experiential sentences don't advance discourse.

In contrast to event, they don't take time. She doesn't give any clear-cut examples to support her claim. Whether or not it advances discourse and takes time, we will evaluate it in terms offoreground vs. background in Section 2.3. But the issue of the experiential sentence being stative in situation type is immediately relevant to

what has been said in the preceding subsections.

Comrie (1978:58) also describes the experiential perfect as indicating that 'a given situation has held at least once during some time in the past leading to the present.' [boldface added by C.C.C.] It seems reasonable then to interpret the Mandarin experience marker -guo as a device for turning a previous event or repeated events into a situation/state that, though no longer existing at the speech time, is relevant to the topical entity involved in the event. 5 This interpretation reasonably explains why (5.a & b) are not acceptable under normal circumstances: Once one is dead, one remains dead; and once one is old, one stays old. It also explains why sentences like (5.c & d) have restricted occurrence: Stative verbs and stative predicates represent states by nature and don't need to be further turned into states by -guo. In other words, with stative verbs/predicates, -guo can only be partially utilized: its function of turning an event into a state becomes vacuous though the other function to indicate that the state no longer exists remains in force. And, of course, its main function has to be performable in all cases to indicate that this past state is relevant to the topical entity at the speech time.

In the same vein, we take another look at sentences (4.a & b), repeated below: (4.a) Ta qunian dao Zhongguo qu-Ie. shelhe last-year to China go-LE 'She/He went to China last year.' b) Ta qunian dao Zhongguo qu-guo. shelhe last-year to China go-EXP 'Shelhe went to China last year.'

L&T's explanation (1981:229) for the difference between the -guo and -Ie above goes like this: Sentence (4.b) 'assumes that slhe went to China and claims that this took place at least once during last year and is over now; this is why (b), but not (a), implies that slhe is now back from China. The subject's return is not part of the meaning of -guo, but it is the message of the -guo sentence because we can infer it from the meaning of -guo: if something has been experienced, it is over.' In fact, the event ofherlhis going to China is implied as over in both statements (a) and (b).

What may not be over in (a) but must be over in (b) is not the event itself; it is rather the necessary result of the event: herlhis being in China.

Let's look at some other examples: (8.a) Wo kan-Ie neiben shu.

I read-LE that-M book 'I read that book. '

b) Wo kan-guo neiben shu.

I read-EXP that-M book 'I read that book.'

Both of them imply that the event of reading that book is over at the speech time.

As the act of reading a book does not have any necessary result, neither of the sentence can say or imply anything about such a result. Therefore, there is very little discernible difference in interpretation between (8.a & b) in isolation.

Similarly, there is very little difference between the members of the pair in (9) below in terms of the fact that the event is over: (9.a) Wo mai-Ie baoxian.

I sell-LE insurance 'I sold insurance.' b) Wo mai-guo baoxian.

I sell-EXP insurance

(lO.a) Wo mai-Ie baoxian.

I buy-LE insurance 'I bought (some) insurance.' b) Wo mai-guo baoxian.

I buy-EXP insurance 'I bought (some) insurance.'

While (lO.a) implies that the insurance policy might still be in force, (lO.b) definitely says that the policy is no longer in force. Here, we note that dao Zhongguo qu 'to go to China' and m;Ii baoxian 'to buy insurance' are of the accomplishment event type while kan shu 'read (a book)' and mai baoxian 'to sell insurance' are of the activity type.

From the examples and their discussion above, it can be safely concluded: (11) The experiential marker -guo indicates that (a) a prior event (which occurred at least once) is being viewed as a state, (b) this state includes whatever necessary result the event may entail, (c) the state existed at a prior time but is no longer existent at a reference time or the speech time, and (d) the state is relevant to the topical entity involved in the event, i.e. as an experience.

Mangione and Li's analysis (1993) seems to support (b) and (c) above. The pragmatic implications and the discourse consequences of this conclusion will be

Very few grammarians have treated -qilai as an aspect marker. Among the few that have done so, Fang (1992:493-499, 526-533) has probably the most detailed discussion.6 Though he does not call it an inchoative aspect marker, the meanings that he gives for the suffix all converge on the beginning of an event or state. The only exceptions to this general observation are expressions with 'directional' (quxiang in Chinese) verbs, where the suffix -qilai 'up' is generally considered a resultative or directional complement, e.g. zhanqilai 'stand up', baoqilai 'hold (something or someone) up',juqilai 'lift up', etc. He states that for most 'nondirectional' verbs, -qilai indicates the beginning and continuation of an action.

These verbs, according to him, consist of 58.2% of the 1,000 most frequently used verbs; they furthermore make up 71.6% of all non-directional verbs. He then mentions the following seven meanings, each illustrated with a group of verbs having something in common. Thus, -qilai indicates (12) An action or state that has developed and changed: biatihua 'change', chengshou 'ripe', haipa 'be frightened', xiguan 'be used to', xihuan 'like', etc.

(13) An action that has realized through progress:jianli and organize', zhong 'plant', etc.

'build', choubei 'plan

(14) An action that has realized and come into a state of being opened up: buzhi, 'decorate', shenzhan 'extend', zhanlan 'exibit', etc.

(15) An action that has realized and come into a state of convergence: jihe 'collect', zuzhi 'organize', luanjie 'unite', etc.

(16) An action that has realized and come into a state of covered up: cang 'to hide', maifu 'ambush', zhegai 'to cover', feng 'to seal', etc.

(17) An action that has realized and produced a result: kongzhi 'control', liyong 'utilize', huixiang 'recollect', etc.

(18) An inference or estimate that has been made through one's senses: chi 'eat', kan 'look', ling 'listen', mo 'feel', wen 'smell'.

It is easy to see that, except perhaps those in (18), the types of the verbs actually contribute to the different meanings that -qilai is said to have with each group of the verbs. One single meaning may suffice to provide an appropriate interpretation for each of the groups: 'to start to'. In fact, this interpretation applies to group (18) as

(19.a) Zhezhong shuiguo wen-qilai hen xiang, chi-qilai hen suan. this-kind fruit smell-start very sweet, eat-start very sour 'This fruit, if you start to smell it, is sweet; but, if you start to eat it, it is sour.' (i.e. it smells sweet but tastes sour.) b) Zheliangge zi, ting-qilai, jiu kepa. these-two-M word, listen-start, then frightening 'These two words, if you start to listen, are frightening.' (i.e. they sound frightening. ) And 'to start to' is exactly what inchoative means. (Cf. Comrie, 1967:15).

The difficulty with this suffix seems to lie in the same form occurring as a complement and as a verbal suffix. But, this difficulty pertains to the written language only; in the spoken form, a distinct stress may be heard on the complement.

A complication for this form is that the two syllables of the suffixes are separable from each other by a grammatical object: (19.c) Ni zeme chang-qi-ge-Iai Ie? you how sing-QI-song-LAI LE 'Why did you start singing?'

2.2.2. The Progressive Zai- and the Durative -Zhe The prefix zai- was perhaps the last of all Mandarin aspect markers to be recognized for its aspect-marking function. As an aspect marker, it is often lumped together with -zhe under the same name of 'durative,.7 In this subsection, we will treat them as two distinct ones, designating -zhe as the durative marker and zai- as the progressive marker. Reasons for taking this approach will become clear as we comment on some of the previous analyses ofthem. When examples are cited from the authors, some slight modifications may be made for easy comprehension.

Abbreviations and notations follow either the general practice or the particular author if no confusion arises.

Chao (1968) does not consider zai- a verbal prefix. He only treats it as a verb (pp.

312 & 337) and as a preposition (p. 754). He regards -zhe as a progressive suffix (pp. 248-251), as in Ta da-ZHE dianhua ne 'He is talking on the phone'.

(20) in a special intensive form -zhe ne, as in Wo xiang ni-ZHE NE 'I miss you terribly'.

(21) in a command, as in Ji-ZHE! 'Remember!'

(22) in fIrst position as a subordinate form to the second verb, as in Ta qi-ZHE ma zhao ma 'He looks for a horse riding a horse'.

(23) in the meaning of 'so far as V-ing is concerned', as in Zhe juzi wen-ZHE xiang, chiqilai suan 'This tangerine smells sweet but tastes sour' (i.e. as far as smelling is concerned ...).

(24) where a few verbs require it, as in tuo-ZHE 'to hold up on the palm', xiangZHE 'to favor (the side 00', chong-ZHE 'to face (the direction 00'.

Li and Thompson (1981 :217-226) treat both zai- and -zhe as 'durative aspect' markers. They seem to distinguish between the two markers mainly by the different constraints under which each of them occurs. Below are some oftheir examples for the meanings, uses and constraints:

(25.a) Zhangsan zai-da Lisi.

Zhangsan DUR-hit Lisi 'Zhangsan is hitting Lisi.' b) *Zhangsan zai-pang.

Zhangsan DUR-fat B. Verbs of posture may occur with the durative aspect marker -zhe to signal an on-going posture or physical disposition of an entity at a location.8 (26.a) Ta zai fangzili zuo-zhe. she at house-in sit-DUR 'She's sitting in the house.' b) Chezi zai waimian ting-zhe. car at outside park-DUR 'The car is parked outside.'

(27.a) Zai menkoude bolishang xie-zhe sige zi. at door-mouth-DE glass-top write-DUR four-M character 'On the glass in the doorway are written 4 characters.' b) Wo wen ta qian dou zai nali ge-zhe.

1 ask himlher money all at where put-DUR 'I asked himlher where all the money was kept.'

(28) Neige fangjian hei-zhe ne. that-M room dark-INT NE 'That room is pretty dark.'

E. In a complex sentence, the verb with -zhe signals that the event is a durative background for another event.

(29.a) Xiao gou yao-zhe weiba pao Ie. small dog wag-DUR tail run LE 'The small dog ran away wagging its tail.' b) Ta nao-zhe yao mai dayi. shelhe fuss-DUR want buy coat.

'She/He made a fuss about wanting to buy a coat.'

We supply the following pair to contrast the difference between the two markers in this environment: (30.a) Ta ting-zhe shouyinji shuizhao Ie. shelhe listen-DUR radio sleep-achieve LE 'She/e fell asleep listening to the radio.' b) ?Ta zai-ting shouyinji shuizhao Ie. shelhe DUR-listen radio sleep-achieve LE Sentence (30.b) is dubious in its grammatical status in spite of the fact that Ta zai ting shouyirifi 'She/He is listening to the radio' is completely acceptable in isolation.

Another grammarian, Chu (1983:92-95), treats the two markers separately, calling zai- the progressive marker and -zhe the concomitative marker. He defInes 'progressive' as 'indicating an action going on at a given point of time' and

'concomitative' as 'referring to an action or event existing or occurring concurrently with or accompanying what is denoted by the main verb in the sentence'. The reason for his reluctance to lump them together under either 'progressive' or 'durative' is obvious when he cites the following examples: (31.a) Na-/Man-/Deng-zhe! take-/slow-/wait-ZHE 'Hold (on to it)/Hold it/Wait (up).' b) Taiyang xie-zhe ne. sun slant-ZHE NE 'The sun is at a slant.'

Neither of them can be comfortably accommodated within the general definition of either 'progressive' or 'durative'.

Other treatments of the two aspect markers include: Su (1974), Chung-yu Chen (1986, 1987a, 1987b), Ma (1985), Chu (1987), and Magionne (1987). Some of them will be discussed in Section 2.2.3. on the discourse functions of the aspect markers.

Reviewing the past writers on zai.- and -zhe, especially Chao (1968), Li and Thompson (1981) and Chu (1983), we find the following general questions: (a) Do the notions 'progressive', 'durative', and 'concomitative', the authors, adequately apply to the facts being described?

as defined by

(b) How can zai- and -zhe be made more distinct, without neglecting their similarities in the description of the facts related to them?

(c) While L&T recognize the occurrence of zai- with activity verbs, no mention is made of -zhe in this connection, i.e. whether it occurs with activity verbs only or it occurs with other types of verbs as well. If both of them are durative, what precludes -zhe from occurring in that same environment?

(e) There is some similarity between L&T's Band C. Can they be generalized to cover a wider range of facts in a more natural way?

(f) If both markers are for durative aspect, why can't zai- be used for the durative background?

With regard to the question posed in (a) above, obviously it is not quite feasible to find a universal definition for each of the notions and apply it to Mandarin Chinese without any difficulty. (Cf. Binnick, 1991, specifically, Chaps. 5, 6 and 7.)

Our strategy is therefore to adapt the most basic defmitions available and try to apply to Chinese. In the process, we also try to make a distinction between what is semantic and syntactic in nature and what is discourse in function.

The most plain and basic defmition of 'progressive' seems to come from Comrie (1978: 12): ' ...progressiveness is the combination of continuousness with nonstativity.'

That is, when an event is viewed as continuing, it is basically expressed by a progressive form of the verb. In our terms (cf. Table I above), an event is [+Dynamic] and continuousness is [+Durative]. Ifwe recognize zai- as a progressive-aspect marker, then it should be compatible with semantically [+Dynamic, +Durative] verbs. The following examples bear out this designation: (32.a) Ta zai-paobu. shelhe PROG-run 'She/He is running.' b) Tamen zai-gai fangzi. they PROG-build house 'They are building a houselhouses.' where pao(bu) is an activity verb and gai(fangzi) is an accomplishment verb, both of which are [+Dynamic, +Durative].

On the other hand, 'durative' is usually defmed in opposition of 'instantaneous' (Binnick, 1991:144). Another term for the same notion is 'continuous', which is subdivided into 'progressive' and 'non-progressive' (Comrie, 1978:25). In other words, 'durative' or 'continuous' is actually a cover term for both 'progressive durative' and 'non-progressive durative'. Now that we have identified zai- as a progressive (and durative) marker, is it possible that -zhe is just its non-progressive counterpart in the durative family? If so, -zhe should be semantically compatible with all [+durative] verbs: (33.a) Ta xihuan-zhe ni ne. shelhe like-DUR you NE 'She/He likes you very much.' b) Xiaohaizi xihuan pao-zhe shang xue. children like run-DUR go-to school 'Children like to run to school.'

where xihuan is a state verb and pao is an activity verb, both of which are [+Durative].

There are, however, two problems with this semantic account. One is that -zhe doesn't normally occur with state verbs like xihuan. If it does, there is a special interpretation: it serves as an intensifier. An explanation can be offered like this.

Since [+Durative] is the only positive feature of a state verb, there is no need for the speaker/writer to explicitly indicate such a viewpoint. However, if a speaker/writer chooses to do so, she/he means, 'I want you to pay special attention to this viewpoint even though it is the only possible one.' This then is no longer a semantic issue but a pragmatic one. (Cf. Chu, 1978.) The second problem is that -zhe does not seem to occur with accomplishment verbs very easily. Perhaps because accomplishment verbs by and large have their focus of information on the result rather than on the process, it is then discoursally infelicitous to impose a durative viewpoint over the result. E.g.

(34.a) ?Tamen gai-zhe neisuo fangzi.... they build-DUR that-M house b) Tamen gai neisuo fangzi, gai-zhe gai-zhe, jiu they build that-M house, build-DUR build-DUR, then daoxialai Ie. fall-down-come LE 'When they were building the house, as the construction was going on, it collapsed. ' According to Smith (1993), gai neisuo jangzi 'to build that house' is an accomplishment verb. But in (34.a), there doesn't seem to be a context that could accept the form. On the other hand, in (34.b) where 'building that house' is made a topic, the durative aspect of the building process (i.e. the construction) can be expressed as the speaker/writer's viewpoint. Therefore, the problem is again not a semantic one, but one of discourse perspective.

The third notion is 'concomitative'. This term is actually functional rather than semantic in nature. We will comment on it in Section 2.2.3.

Having looked at the definitions of 'progressive' and 'durative', we feel ready to state our position on the two aspect markers. On the basis of the linguistic facts and semantic considerations, it is justified to designate the two aspect markers by different names in order to indicate not only their differences but also their similarities. The prefix zai- is a 'progressive durative' and can be designated by the name 'progressive' for short; the suffix -zhe is a 'non-progressive durative' and can be designated by the name 'durative' for short. 9 With the understanding that they are both durative, it is easy to see why they behave similarly

when only the semantic feature 'durative' is in question. This is where most ofthe confusion arises in the literature. On the other hand, they may act differently when the semantic feature 'progressive' is also at issue. Our decision above is reached, however, on the basis of the semantics of verbs and the viewpoints of the aspects only. When pragmatic factors such as focus and topic are considered, there will be complications. We will discuss such complications in Section 2.2.3.

Another point that must be clarified is the functions that zai- and -zhe serve in complex sentences. Zai- tends to be temporally oriented while -zhe tends to be manner-oriented. We will fully discuss this issue with the other discourse and pragmatic problems later in the next section.

At this juncture it may be appropriate to stress a fact about the relationships between aspectual marking and verbal affixes in Chinese as well as in other languages. While aspect is by and large marked by verbal forms, the marking relationship is not necessarily exclusive. In other words, the verbal affixes may very well perform other functions than marking aspect and, conversely, some aspectual meanings may not necessarily be expressed by varying the form of the verb. We have seen above that -zhe may be used for some pragmatic purposes. On the other hand, the attemptive aspect is expressed by try to in English, i.e. I tried to call you last night.

In the preceding section, we tried to explore the similarities and differences of the affixes -zhe and zai-. It was found that they are similar in terms of being durative but are different in terms of being progressive. As a result, it was decided that zaishould be designated as a progressive-aspect marker and -zhe as a durative-aspect marker to reflect their nature, leaving the descriptive terms 'durative' and 'nonprogressive' out of the names for short, respectively. There remain, however, many problems in interpreting some of the uses of zai- and -zhe. We will try to account for those problems in this section from the perspective of discourse.

One of the main differences between the two affixes is a discoursal one-one that goes beyond the boundaries of a sentence: zai- indicates that the clause with the prefix in it is temporal in nature while -zhe indicates that the clause with the suffix in it is treated as a manner adverbial (cf. Ma, 1985). Evidence can be found in the fact that a zai- clause is more naturally used to answer a 'when' -question than a -zhe clause is.

(35.a) Ni shenmo shihou tingdao zhege xiaoxi de? you what time hear this news DE 'When did you hear this news?'

b) Wo zai-kan dianshi de shihou.

I PROG-watch TV DE time 'When I was watching TV. ' c) Wo kan-zhe dianshi de shihou.

I watch-DUR TV DE time 'When I was watching TV ... .'

As an answer to the question in (35.a), the (b) sentence is natural while (c) is not.

Moreover, the latter sounds unfinished regardless of whether as whether it is used an answer or otherwise.

On the other hand, the claim that the -zhe clause behaves like a manner adverbial fmds evidence in the following fact: a 'how'-question rather than any other type of question is used for clarification if a verb with a -zhe is not clear to the hearer. A 'how'-question asks about the manner in which something is done.

(36.a) Xiaomao changchang V-zhe yanjing kan woo kitten often V-DUR eye look-at me 'The kitten often looks at me, V-ing its eyes.' b) Xiaomao zenmo-zhe yanjing kan ni? kitten how-DUR eye look-at you 'How does the kitten look at you?'

The verb with -zhe is blurred in (36.a) and the hearer wants the speaker to clarify what the verb is by asking (36.b). Note that the verb slot is filled by zenmo how' rather than any other question word. The same principle applies to simple sentences as well: (37.a) Xiaomao zai ner V-zhe. kitten at there V-DUR 'The kitten is V-ing there.' b) Xiaomao zai ner zenmo-zhe? kitten at there how-DUR 'What's the kitten doing there?'

Just like in (36), the verb with -zhe in (37.a) is blurred and the hearer asks for a clarification by using azenmo 'how' question (37.b). This fact pretty much reflects the manner adverbial nature of the -zhe clause. (But see Section 2.2.3.2. below.)

The following pair provides a sharp contrast between the two affixes: (Ma, 1985:42)

(38.a) Ta zai-baifang pengyou de shihou bing Ie. shelhe PROG-visit friend DE time sick LE 'Shelhe got sick during herlhis visit with a friend.' b) ?Ta baifang-zhe pengyou (de shihou) bing Ie. shelhe visit-DUR friend (DE time) sick LE where 'visiting a friend' cannot be interpreted as the manner in which one gets sick and therefore (38.b) is questionable.

While the usage of the zai- clause seems straightforward (except, perhaps, for its relation with the locative preposition zai); that of the -zhe clause is quite complex. Simply put, its semantic import is 'durative' with a stativizing effect, its syntactic function is 'subordination', and its discourse function is 'backgrounding'.

Yet it interacts with the semantics of the verb and the pragmatics of an unfinished utterance to produce a variety of possible interpretations. Below, we summarize the recent research on this suffix, mainly following Chu (1987).

2.2.3.1. Semantic Interpretation of -Zhe and Verb Type Chu (1987:2-13) recognizes the suffix -zhe as basically 'durative' in meaning: It 'largely indicates an on-going act in a simple sentence and a simultaneous occurrence of two acts in a complex sentence.' But this durative meaning interacts with different types of verbs to produce different effects on the meanings of the verbs. He differentiates four types of verbs: action, posture, placement and state.

(The first three are all event verbs in our present framework. Roughly, action corresponds to activity; and both posture and placement are subclasses of the semelfactive verb.) The most typical durative interpretation of the suffix is with the action verb. For example: (adapted from Ma, 1985) (39.a) Henduo ren zaijieshang zou-zhe. very-many person at street-on walk-DUR 'Many people are walking on the street.' b) Huanghe de shui riyede liu-zhe.

Yellow-River DE water day-night-DE flow-DUR 'The Yellow River flows day and night.' c) Ta jiao-zhe kouxiangtang shuohua. shelhe chew-DUR chewing-gum talk . 'Shelhe talks/talked while chewing gum.'

The verbs in (39), zou 'walk', liu 'flow' andjiao 'chew', are all action verbs and the durative marker -zhe indicates in all cases either an on-going act or a simultaneous

occurring of two acts.

With posture and placement verbs, the interpretation of the suffix is a little skewed because the event represented by the verb is non-durative itself. To visualize a non-durative act as durative, one has to incorporate its resultant state into the event. This is exactly the effect of the durative aspect marker with non-durative posture and placement verbs in Mandarin. E.g. (from L&T, 1981; and Ma, 1985)

(40.a) Tamen zai menkou zhan-zhe. they at doorway stand-DUR 'They are standing in the doorway.'

simple sentences-those with only one verb and the verb is suffixed with -zhe-there obviously is a problem: How can a simple sentence be subordinated to anything else? The key lies in the fact that a simple sentence with -zhe almost always sounds unfinished or incomplete, (even though we have been giving complete English sentences in translation by following the usual practice). Chu believes that it is exactly this subordinating force that makes them sound unfmished or incomplete.

For example: (40.a) Deng-zhe! wait-DUR 'Wait (and) ... .'

b) Ta zai dishang gui-zhe. shelhe at floor kneel-DUR 'She/He is kneeling on the floor.'

b) Zuo-zhe! sit-DUR 'Be seated (and) ....'

c) Qian zai yinhangli cun-zhe. money at bank deposit-DUR 'The money is deposited in the bank.'

c) Na-zhe!' take-DUR 'Hold on to it (and) ... .'

d) Zazhi zai shujiashang fang-zhe. magazine at bookshelf-on place-DUR 'The magazine(s) is/are placed on the bookshelf.'

The verbs_zhan 'stand' in (40.a) and gui 'kneel' in (40.b) are posture verbs and the verbs cun 'deposit' in (40.c) andfang 'place' in (40.d) are placement verbs. All of them are semelfactive, i.e. once the act occurs, it ends instantaneously. To interpret them as durative, the only way is to incorporate their resultant states. When this is done, the durative aspect is perceived as having a stativizing effect on these verbs, i.e. an effect that makes non-state verbs appear stative. This same stativizing effect is also exhibited in the verbs like kai 'open' and guan 'close' (as in Men kai-/guanzhe 'The door is open/closed'); chuan 'wear' and dai 'wear' (as in Tamen dou chuan-zhe hong yifU 'They are all dressed in red.'), gua 'hang', etc. Some ofthese are neither posture nor placement verbs but they are semelfactive ones.

With state verbs, the suffix -zhe doesn't generally occur, as state verbs are by defmition already durative. If it occurs, then its function is not to impose a specific aspectual viewpoint or to make the verb more stative. It is rather used for syntactic and pragmatic reasons. We discuss this problem in the following sections.

2.2.3.2. Syntactic Interpretation of -Zhe Chu (1987:14-22) maintains that the main syntactic function of the suffix -zhe is subordination. This claim is well accepted for complex sentences with -zhe. For

The utterances in (40) all imply that if you wait, be seated, or hold on to it, something else will happen. (That is why the -zhe form is often used to express a threat, warning or advice.) The implication can not be derived from anything but the interpretation that -zhe is a subordinating suffix.

The recognition of -zhe as a subordinating suffix accounts for quite a few other issues which would otherwise appear to be unrelated. We will discuss the following problems pertaining to -zhe as a subordinating suffix: (i) its use with state verbs, (ii) its optional nature with other verb-like words, (iii) its interpretation as a manner adverbial marker, (iv) its backgrounding function, and (v) its emphatic use in conjunction with ne.

We will first look at the use of -zhe with state verbs: (42.a) Ta mang-zhe dao chao (Chen, 1986:9) shelhe busy-DUR pour tea 'She/He is busy serving tea.' b) W 0 fanzheng xian-zhe ye mei shi.

I anyway in-Ieisure-DUR also not-have work 'I am not doing anything anyway.'

The verbs mang 'busy' and xian 'in leisure' are stative. The reason for them to occur with -zhe can hardly be to focus on their durative nature. The only explanation for this use is to subordinate the predicate in question to another

predicate.

The next few examples illustrate the optional -zhe with some verb-like words: (43.a) Bie dang-zhe haizide mian chaojia. (Chen, 1986:11) don't in-front-of child's face quarrel 'Don't fight in front of (the) child(ren).'

It was mentioned above that a main difference between zai- and -zhe as aspect markers is that the latter functions to mark the clause as a manner adverbial and the former, as a temporal clause. The manner adverbial function of -zhe, however, is not one of its fundamental characteristics but a derived one. It is derived from the fact that the clause is subordinated to another and that the verb in the main clause is most often an activity verb. An action subordinated to another is very easy to be interpreted as 'manner'. E.g. (Adapted from Ma, 1985) (45.a) Ta chuan-zhe gaogen xie pa shan. she wear-DUR high-heel shoe climb mountain 'She mountain-climbed in high heels.'

b) Shun-zhe zheitiao lu zou. follow-DUR this-M road walk '(Go by) follow(ing) this road.'

b) Dajia dou chang-zhe ge tiao wu. everybody all sing-DUR song dance dance 'Everybody danced while singing.'

c) Ta dui-zhe jingzi zai-zuo guilian. shelhe toward-DUR mirror PROG-make devil-face 'She/He is making faces before the mirror.'

c) Ta mang-zhe dao chao (Same as (42.aÂť shelhe busy-DUR pour tea 'She/He is busy serving tea.'

d) Zhao-zhe wode hua zuo. follow-DUR my word do 'Do as I tell you to.'

The suffix -zhe in (43) may be deleted without affecting the acceptability in all cases. A closer examination reveals that all the words having the suffix are 10

coverbs-words that play an indeterminate role between verb and preposition.

Treated as verbs, they need some marking for subordination; treated as prepositIons, they don't. That seems a good explanation why the suffix is optional.

The next group of sentences illustrate backgrounding by the suffix -zhe: (from Li and Thompson, 1981:233-4 and Dou, 1983:163) (44.a) Xiao gou yao-zhe weiba pao Ie. (Li & Thompson) small dog wag-DUR tail run LE 'The puppy ran away wagging its tail.' b) Ta xie-zhe yan xiao-zhe kan woo (Li & Thompson) shelhe slant-DUR eye smile-DUR look I 'Smiling, shelhe looked at me out of the comer ofhislher eyes.' c) Tamen shuo-zhe shuo-zhe xiao-qilai-le. (Dou) The talk-DUR talk-DUR laugh-INCHO-LE 'As they went on talking, they started laughing.'

The backgrounding function is derived from the predicate being subordinated to the main predicate. (For more detail, see Chapter 6.)

While chuan-zhe gaogen xie 'wear-DUR high heels' in (45.a) can be interpreted as a manner adverbial; the other two, chang-zhe ge 'sing-DUR song' in (b) and mangzhe 'busy-DUR' in (c) can hardly be interpreted in the same way. 'Singing' cannot be said to be the manner of 'dancing' and 'being busy' cannot be said to be the manner of 'serving tea'. 'Singing' is actually used to denote an activity going on at the same time as another, but is less important than the other. 'Being busy' is used to describe the state in which the person is. As a matter of fact, there is another interpretation for (45.c): 'She/He hurried to serve tea', which is even farther from having a manner adverbial. Therefore, the -zhe clause being a manner adverbial, though a very prominent identifying feature, is not a primary characteristic by which the function of the suffix can be defmed. The following example defmitely does not involve a manner adverbial: (46) Van kan-zhe fangzi jiu shaoguang Ie. (Chu, 1987:28) eye look-DUR house at-once bum-up LE 'The house got bumt up in no time, as (we) watched.'

The emphatic use of -zhe, usually accompanied by a clause-final ne will be discussed in the next section.

interpretation of simple-predicate sentences with -zhe and the emphatic meaning of -zhe in conjunction with ne.

The simple-predicate -zhe sentence has been independently treated and regarded as a progressive form, implicitly or explicitly, by many grammarians. Its incompleteness has thus been ignored. Once it is recognized as incomplete or unfinished, a natural link may be established between the syntactic function of subordination by -zhe and the incomplete reading of such a sentence: A subordinate clause by itself is not complete. The next task then is to take care of the interpretation of the incompleteness in each and every case. Let us look at some simple cases: (47.a) Deng-zhe! (Same as (41)) wait-DUR 'Wait (up)!' b) Zuo-zhe! sit-DUR 'Sit (down)!' c) Na-zhe! take-in-hand-DUR 'Hold (on to it)!'

(48.a) Ni ke xiaoxin-zhe. you however careful-DUR 'You'd better be careful.' b) Shu zai shujiashang fang-zhe. book at bookshelf-on put-DUR 'The book(s) is/are on the shelf.'

(50.a) Tang re-zhe ne. (Chao, 1968:248) soup hot/heat-DUR NE 'The soup is awfully hotlbeing heated.'

The reason that the utterances in (47) can be interpreted as commands is precisely because they are unfinished. The implication is, 'Do as you are bein told to and something else will (or will not) happen, which you can figure out.' I Utterance (48.a) is a piece of advice also because of what is unsaid: Something might happen if you weren't. The sentence in (48.b) is not usually considered incomplete because it can stand alone, for example, as an answer to the question Shu zai nar? 'Where is/are the book(s)?' But when we consider an alternative form as an answer to the question:

r

(49) ÂŤShu) fang) zai shujiashang. ÂŤbook) put) at book-shelf-on '(They are) on the bookshelf.'

we fmd that (48.b) says a lot more than (49). While (49) is a plain straightforward answer without any implication, (48.b) has a variety of possible implications: 'You should know that'; 'How can it/they be anywhere else?' 'Why are you asking?', etc.

12 All such possibilities may be attributed to the incompleteness of the sentence.

The emphatic use of -zhe has long been noted but has not often been linked to its fundamental functions of durative aspect marking, subordination, and backgrounding. (Cf. Section 2.2.2.1, Item D, above.) Chu (1978) tries to find relations between the pragmatic emphatic force of the -zhe... ne form and the aspect marker's syntactic functions of subordination and its semantic function of stativizing an action verb. He further explored them in another article (1987). Here, we synthesize the previous fmdings and try to conclude on the inner workings of the emphatic use of -zhe in non-technical plain language.

The fact that -zhe in this use often cooccurs with ne warrants a look at the function of the latter. King (1985:27) justifiably claims that ne 'is used by the speaker as a device for highlighting or evaluating certain portions of background information in the discourse and bringing them to the attention of the hearer in the speaker/hearer world. Utterances with ne are of particular relevance to the point the speaker wants to make. In using ne the speaker is making a metalinguistic comment and, by extension, expressing his attitude toward the content of the utterance.' In other words, by using ne, the speaker is saying, 'You are familiar with what I am saying, though it may have slipped your mind. But, anyway, this is relevant to the discourse and I want you to pay attention to it.' Chu, therefore, calls ne a particle of continuation (1983) and of relevance (1985b). This discourse interpretation of ne seems to fit very well with its frequent cooccurrence with -zhe for an emphatic expression.

Let's look at some of the examples that have been cited many times by grammarians:

b) Neige fangjian hei-zhe ne. (Li & Thompson, 1982:222) that-M room dark-DUR NE 'That room is pretty dark.' c) Wo xiang ni-zhe ne. (Chao, op. cit.)

I think you-DUR NE 'I miss you terribly.'

According to Chu (1987 :31), there are three forces working together here to contribute to the emphatic or intensifying effect of the construction -zhe ... ne: '(a) the use ofa subordinating suffix to signal the unfmished nature of the utterance, (b)

the employment of ne to indicate the topical relationship to the unmentioned context, linguistic or otherwise, and (c) the addition of a durative aspect marker to a stative verb which is intrinsically durative, to begin with.' We have discussed the first two in Section 2.2.3.2. and above, respectively. Why should Š also contribute to the intensifying effect? Chu (op. cit.) further invokes Grice's Maxim of Relation to explain the durative aspect marker with a state verb, i.e. the redundant [and intentional] use of the durative marker must also be relevant.

Putting the three forces together, we obtain a complex message like this: (a) figure out what is not mentioned but strongly implied, (b) take the information expressed here as topical to the missing piece of information, and (c) mind that the state of being durative is specifically relevant. Translating these three forces into everyday language by fitting the three examples in, we get: (51.a) Let me remind you that the soup is hot/being heated and you know what to do with it. b) Let me remind you that the room is dark and you know what to do about

it. c) You know all along that I will miss you and I am just telling you once again that I do miss you (and you figure out what's next). 13

In this section, we have tried to distinguish between the syntactic functions of the progressive zai- and the durative -zhe. The former is found to be temporal in nature and the latter is found to perform the role of a manner adverbial. This distinction is easy to detect, yet -zhe being a manner adverbial is not its primary function. It is derived from the one for subordination.

While the progressive marker appears quite straightforward in its usage and interpretation, the durative marker is surprisingly complex. We have explored the functions of -zhe on three levels: semantic, syntactic, and pragmatic. We find that the functions may interact with each other and with other grammatical structures to produce many different interpretations of the suffix -zhe in usage.

In particular, -zhe is found to be basically durative in semantic terms. This durative meaning, however, is typical only with activity verbs ('action verbs' in Chu, 1987). With semelfactive verbs, the durative meaning does not pertain to the event itself, but is extended to the resultant state of the event. This accounts for the interpretation of the so-called 'posture' and 'placement' verbs (L&T, 1981; Ma, 1987; Chu, 1987). As a matter of fact, the extension is valid with many other types of semelfactive verbs as well. It is this extended function that has been recognized as a stativizing effect. With state verbs, the addition of -zhe is primarily for syntactic or pragmatic reasons rather than for semantic purposes.

Achievement and accomplishment verbs are not discussed because they are telic in nature (i.e. their focus is on the end product of an event) and are not directly relevant to the durative meaning.

Syntactically, -zhe functions to indicate subordination. This subordinating function accounts for a host of issues that would otherwise appear to be unrelated.

It accounts for (a) the general incomplete reading of a simple clause with -zhe, (b) the possible occurrence of -zhe with a state verb despite of semantic redundancy, (c) the optional use of -zhe with coverbs and other similar words, (d) the interpretation of the -zhe clause as a manner adverbial, though it is not its basic function, (e) the source of backgrounding by -zhe, and (f) the emphatic use in conjunction with ne.

The interpretation ofthe incompleteness of a simple clause with -zhe, however, has to come from pragmatics. The intentional use of a dangling subordinate clause usually means that the speaker /writer wants the hearer/reader to complete it by him/herself Similarly, the emphatic interpretation of the -zhe... ne construction finds its root in pragmatics, too. Three forces converge to make the emphatic meaning: (a) incompleteness, (b) topical nature of the information contained in the utterance, and (c) the double signaling of durativeness by the state verb and the suffix. These three combine to produce a complex message of what is usually regarded as 'emphatic'.

What is obvious in this section is the complex interface of the suffix -zhe with the different types of verbs on the semantic level and the interaction of the functions of -zhe between all three levels-semantic, syntactic and pragmatic. These interrelations work together to move the grammar from the sentential domain to the discourse world.

The perfective -Ie has been a source of great interest for linguists, but it also has been one of tremendous puzzlement and confusion for language teachers and students. It behaves sometimes like the past tense and other times like the present perfect in English; but it is not quite the same as either of them. Especially intricate and puzzling is the non-use of -Ie where the theory predicts otherwise. Recent research has discovered that the interpretation of -Ie is not only closely interwoven with the situation types of verbs, but it is also affected by discourse and pragmatics.

In this section we will summarize the most recent research on -Ie as an aspect marker, point out problems and try to solve some of them. Many of them, however, will have to find answers beyond the semantic and viewpoint model.

It has been recognized that -Ie is mainly used for indicating a bounded situation, focusing on the endpoint of the situation. Thus, Shi (1991 :95) states that 'perfectivity is the result of bounded situations viewed as relatively anterior, ... ' Smith (1993 :6) makes a similar claim that 'the perfective morpheme -Ie presents closed, non-stative situations ....the perfective spans the single point of instantaneous events, and the initial and terminated fmal endpoint of durative events.' A bounded

(55.a) Wo kanwan-Ie bao, jiu shui.

I read-finish-PFV paper, then sleep 'When I finish reading the paper, I will go to sleep.'

or closed situation is usually understood as one that is viewed in its entirety regardless of time.

In spite of the different wordings used by Shi and Smith, their analyses are consistent with Li and Thompson's interpretation of the aspect marker and its relationship with boundedness (1981:185fi): 'An event is viewed in its entirety if it is bounded temporally, spatially or conceptually. There are essentially four ways in which an event can be bounded.'

A. By being a quantified phrase following -Ie.) v

event: (The event is quantified by the underlined

(52.a) Wo zai nali zhu-Ie liangge yue.

I at there live-PFV two-M month '1 lived there for two months.'

pengdao Lin Hui. Ta shuo mingtian lai.

I yesterday ran-into Lin Hui she said tomorrow come 'I ran into Lin Hui yesterday. She said she would come tomorrow.'

, (53.a) Wo pengdao-Ie Lin Hui. I bump-into-PFV Lin Hui 'I ran into Lin Hui' (where the important information in the content is whom I ran into). b) Wo xiangchulai-Ie neige zi.

I think-out-PFV that-M word 'I remembered that character.'

v

(54.a) Wo wang-Ie tade dizhi.

I forget-PFV herlhis address.

'I forgot herlhis address.' b) Ta shuizhao-Ie shelhe sleep-succeed-PFV 'Did shelhe fall asleep?'

ma?

Q

bounded

(56.a) Wo zai ner zhu liangge yue, jiu huilai Ie. I at there live two-M month then return-come LE 'I stayed there two months and came back.'

v b) Wo zuotian

B. By being a definite or specific event: (The event is made defmite or specific by the underlined phrase following -Ie.)

verbs with inherent

We note here that while the sentences in (52)-{53) all represent bounded events-events viewed in their entirety, their boundedness does not necessarily come from the quantification, definiteness or specificity of the event. For example: I;

b) Diandeng liang-Ie henduo. electric-light bright-PFV very-much 'The electric light got a lot brighter.'

C. By containing underlined.)

b) Ta kai-Ie men, m JIll jinqu. shelhe open-PFV door, you then enter-go 'When shelhe opens the door, you go in.'

meanings:

(The verbs are

Compare (56.a) with (52.a). Both of them contain the quantifying phrase liangge yue 'two months', but one is interpreted as a bounded event by the presence of -Ie and the other is not. Similarly, both (56.b) and (53.a) contain the defmitizing personal name Lin Hui, but only one ofthem is presented as a bounded event by the presence of -Ie. It thus seems that while there is some correlation between boundedness on the one hand and quantification and defineness/specificity on the other, the use of the perfective aspect marker -Ie does not completely depend on the presence of the latter. The relationship between the three-boundedness, the presence of expressions for quantification, definiteness and specificity, and the use of the perfective-aspect marker -Ie-is not as straightforward as it has usually been pictured. It will be further discussed in Section 2.2.5. below.

For the sentences in (54), the meaning of the verb certainly plays a part in determining the boundedness of the event, i.e. wangle lade dizhi 'forget herlhis address' has to occur in its entirety as an event. But, that's not the whole story. If wang 'forget' is replaced by a synonym wangji in (54.a), the verbal suffix -Ie becomes optional, espceially when the sentence ends in a final k: Wo wangji lade dizhi Ie 'I forgot herlhis address.' Again, there is something other than the meaning of the verb that contributes to the determination of the boundedness of the event and, thus, that of the use or non-use of -Ie. The 'first event in a sequence' condition illustrated by (55) works well, but it can perhaps be better described in Shi's notion ofanteriority. This notion, however, needs to be extended from semantics to discourse (cf. Chang, 1986). We will not further comment on this point until we discuss the other two problems in more

use of -Ie. Unfortunately, he falls into the same trap as we pointed out above. One example suffices: It is very interesting to note that in each of the three studies that we will review both the verbal suffix -Ie and the sentence-fmal particle ~ are treated at the same time.

The reason is perhaps that sometimes it is hard to tell one from the other. Our discussion here, however, will concentration on the verbal -Ie only.

Chronologically, Spanos (1979) is the first systematic study of -Ie for its pragmatic functions. His approach is one of speech act. Andreasen (1981) analyzes the use and non-use of -Ie in terms of foreground vs. background in narrative discourse. Chang (1986) integrates the previous research and sets up a framework to account for the various possibilities of -Ie as a discourse marker beyond aspect.

Spanos conducted a survey of 62 native speakers of Mandarin in order to 'determine how native speakers employ LE when constructing sentences and paragraphs involving the variety of meanings associated with LE' (p. 34). Chang (1986:55) summarizes his major findings as follows: (58.a) Use of LE is subject to a principle of non-redundancy, i.e. LE tends to be omitted in contexts where the time, aspcet, phase, or modality of action, 14 process or state of affair is already specified. b) A rule of consistency appears to operate for some speakers. That is, if LE is used in a specific structure, it will always be used in that structure regardless of redundancy. c) In instances where the context is insufficient to determine the time, aspect, phase or modality of an action, process or state of affairs, a rule of caution appears to increase the likelihood that LE will be used.

Spanos then uses Grice's maxims of conversation to justify the rules that he has formulated on the basis of the statistical figures obtained in the survey. While the numbers are reliable and the formulations are consistent with the facts, he doesn't give due consideration to the linguistic properties of both the perfective aspect -Ie and the sentence-fmal particle Ie. (Cf. critique by Chang, 1986:55-65.)

In particular, he jumps directly from morphology into pragamtics without looking at any consequences that syntax, semantics and discourse may have on the use and non-use of the two particles.15 Andreasen (1981) claims that the perfective-aspect in Mandarin has the additional function of marking foreground in narrative discourse. In the study he shows that certain foregrounded clauses are marked with -Ie. There is a very disturbing fact, however. At least as many, ifnot more, forgrounded clauses occur without -Ie. He resorts to L&T's principle of perf ectivi zing phrases preempting the

(60.a) Hua Laoshuan huran zuoqi 0 shen Hua Laoshuan suddenly sit-up body b) ca-zhaol6 huochai strike-light match c) dianshang 0 bian-shen youla de dengzhan light-on whole-body grease DE oil-lamp d) chaguan de liangjian wuzi-li bian miman-Ie qing-bai de guang. tea-house DE two-M room-in then fill-PFV green-white DE light 'Hua Laoshuan suddenly sat up, struk a match, and lit the completely greasecovered lamp. The two rooms in the tea-house then were filled with a greenish white light.' (From Lu Xun, fao'Medicine') Following Chang (1986:88), we have indicated two positions with 0 where a -Ie could potentially occur in (60). Andreasen's explanation (1981:66-72) why -Ie does not occur in those positions goes as follows. Clause (60.a) contains the directional complement -qi 'up' and the temporal phrase huran 'suddenly', both of which make the event 'perfectivized' and 'viewed as an unanalyzable whole'. Thus, the non-use of -Ie. Clause (6Q.c)c211tains another perfectivizing complement -shang, which is sufficient to signal the endpoint of the event and makes the use of -Ie unnecessary.

On the other hand, (60.d) needs the -Ie because there is a highly qualified object qing-bai de guang 'greenish-white light'.

The explanation seems reasonable enough. But Chang (1986:90) raises this crucial question: Ifthe highly qualified object qing-bai de guang requires a -Ie for its verb, why is it that an equally highly qualified object bianshen youla de dengzhan 'oil lamp completely covered with grease' in (60.c) does not demand the same? To solve contradictions of this kind, Chang proposes to look at the problem of use and non-use of -Ie from the discourse perspective. His conclusions are based on a careful examination of written texts from both Taiwan and mainland China and are verified by extensive surveys of native speakers.

Sections 2.2.5.2 throgh 2.2.5.5 below are largely based on Chang's view on the 17

discourse functions of the perfecitve -Ie in Mandarin Chinese. additional examples are contributed by the present author.

Comments and

importance within a segment.,18 A segment, according to Hinds (1979), is a consituent which stands between a paragraph and each individual sentence. Within each segment there is typically one sentence that is functionally more prominent.

This sentence Hinds calls a peak. In Japanese procedural discourse, a peak representing the culminating event is marked with a fmite verb whereas non-peaks representing preliminary events are marked with either a stem or a particle (ga or to). Chang goes on to claim that 'one of the discourse functions -Ie serves in Chinese narrative is to [overtly] mark the peak clause of a segment, much like the use of a finite verb form in Japanese procedural discourse.' Thus, (6 1.a) Yushi Mingdi bian paiqian 0 Caiyin he Qinjing liangge consequently Ming-emperor then send Caiyin and Qingjing two-M guanyuan dao Yindu qu qiuqu fojing. official to India go search Buddhist-sutra b) taman zoudao 0 xianzai Afuhan de yige difang they walk-to now Afghanistan DE a-M place c) dedao-Ie fojing he foxiang. acquire-PFV Buddhist-sutra and Budda's-statue (From Bai Mai Si 'The White Horse Temple') 'Thereupon, Emperor Ming sent two officials, Caiyin and Qinjing, to India to search for the Buddhist sutra. When they came to a place known today as Afghanistan, they found the Sutras and the Statues.'

(62.a) Jinnian lai suizhe Zhongwai wenhuajiaoliu Shiye recent-year come follow-DUR Chinese-foreign culture exchange business de fazhan DEdelopment b) Beijing kaoya he 'Quanjude' dianhao ye piao 0 yang guo 0 hai Beijing roast-duck and Quanjude store-name also float Ocean cross sea c) chuandao 0 guowai transmit-arrive country-outside d) zhe jiu shi gengduo de ren changdao-Ie this then make even-more DE people taste-reach-PFV

wel-xlang-se-mei de Beijing kaoya Ie. taste-sweet-Iook-pretty DE Beijing roast-duck LE (From Beijing Kaoya 'Beijing Roast Duck') 'Recently, along with cultural exchanges between China and other countries, the Beijing roast duck and the store name Quan-Ju-De went abroad, crossing the seas and oceans. This made it possible for more people to (be able to) taste the delicious and splendid-looking Beijing roast duck.'

In (61) and (62) above, the positions indicated by 0 are all potential slots for a perfective -Ie. The reason that they are not filled, according to Chang, is that they are in non-peak clauses. On the other hand, the appearence of -Ie in (61.c) and (62.d) is exactly for the purpose of overtly marking the peak event in each series of events.

The discovery of this discourse function based on factual data effectively solves many of the problems that have been raised here and elsewhere concerning the nonuse of -Ie in a potentially possible position. For example, the passage in (60) fro~ Lu Xun may be evaluated by this 'overt peak-marking principle' without going into the complicated explanation of Andreasen's or incurring any self-contradiction: The potential -Ie positions are not filled because they are not peak events while -Ie occurs in the final clause because it represnets a peak event.

This same principle may also apply to cases where quantification, definiteness/specificity, and perfectivization by other phrases are involved in L&T's claim. We repeat some of their examples and our counter-examples below: (52.a) Wo zai nali zhu-Ie liangge yue.

I at there live-PFV two-M month 'I lived there for two months.'

(56.a) Wo zai nar zhu-O liangge yue, jiu huilai Ie. I at there live two-M month, then return-come LE 'I lived there for two months and came back.'

Rather than saying that the presence of the quantified noun phrase liangge yue 'two months' in (52.a) makes the event bounded so that the verb requires a -Ie (which doesn't occur in the other example), we can now claim that the verb zhu 'live' in (52.a) is in a peak clause but the one in (56.a) is in a non-peak clause. There is a piece of strong evidence for the claim that the first clause in (56.a) is a non-peak one: Invariably, native speakers feel that without a second clause, a portion like wo zai nar zhu liangge yue doesn't sound fmished. The incompleteness comes from the fact that the clause is treated as non-peak. If so, the hearer/reader naturally expects a peak clause to follow. Some grammarians attribute the incompleteness to the indeterminacy of time. But even when a past time such as qunian 'last year' is

added, it still sounds unfmished. A counter-argument might be offered, 'Of course, it doesn't sound fmished. It lacks a perfective marker to indicate its boundedness.'

The argument is circular in itself. Why is it then that it becomes complete whenjiu huilai Ie is added to it? Until there is a better explanation, the theory stands firm that the perfective -Ie serves to mark a peak event in Chinese narrative discourse.; Another pair of previously discussed sentences illustrate the same point: (53.a) Wo pengdao-Ie Lin Hui. I bump-arrive-PFV Un Hui 'I bumped into Un Hui.'

(64) Shangxingqi tamen bang-Ie wo ban xingli, jintian wo qing tamen chi last-week they help-PFV I move baggage, today I invite then eat wanfan. dinner

(56.b) Wo zuotian pengdao-O Lin Hui, ta shuo mingtian laL I yesterday bump-arrive Un Hui, she said tomorrow come 'I bumped into Lin Hui yesterday. She said she would come tomorrow.'

The absence of -Ie in (56.b) can be similarly explained by the overt peak-marking principle. Note that the same degree of incompleteness would be felt by native speakers for (56.b) if the second clause is left out. The presence or absence of zuotian 'yesterday' is inconsequential.

Finally, the peak-marking function of -Ie also serves to answer a nagging question posed by Lu and Ma (1985:3) in their insightful discussions ofa number of function words: Why is it that Sentence (a) below cannot take -Ie after the verb bang 'help' while Sentence (b) must take one after the same verb?

(63.a) Wo xia che yihou, Zhongguo tongxue reqingde bang-Ol*-Ie I get-off vehicle after, Chinese school-mate warmly help-Ol*-PFV wo ban xingli.

I move baggage 'When I got off (the train), some Chinese school-mates my baggage.'

peak-marking is necessary at all, ban rather than bang is more likely to get marked.

Nevertheless, the statuses and relative positions of the two verbs do not necessarily prevent bang 'help' from becoming the peak as long as there is a discourse need.

E.g.

helped me with

b) Zuotian nimen zhenshi bang-Iel*-O wo hen da de mango yesterday y'all really give-PFV/*-O I very big DE help 'Yesterday you guys really gave me a big help.'

There are two reasons why (63.a) cannot t~ke -Ie. One is structural, the other pragmatic; but both of them are based on the peak-marking function of the perfective marker. Structurally, the verb phrase bang wo ban xingli 'help me move baggage' consists two verbs: bang 'help' and ban 'move'. The first verb bang is less dynamic semantically and less prominent positionally than the second verb ban and therefore is less likely to serve for a peak event than the second verb. So, if a

'Last week they helped me with my baggage and I am inviting them to a dinner tonight. ' where 'helping' is the main event that constitutes the reason why 'I am inviting them to a dinner.' Pragmatically, (63.a) as a whole doesn't serve the peak (Le. foreground--ef. Section 2.3) function in a narrative. It usually sets up a background or circumstance where a narrative is to take place.

On the other hand, (63.b) is obviously a commentary on a past single event. As a single event and nothing else is to follow, the event verb must be marked with a PFV marker.

In addition to observing and analyzing such facts as discussed in (61) and (62), Chang (1986) also provides rationales for why there is a need for an overt peak marking device in Mandarin and why there are exceptions to the rule. Three facts are advanced as related to the overt peak-marking principle and other uses: iconicity of word order, lexcial collocation, and the semantic nature of Chinese action/event verbs.

Tai (1985) convincingly argues that Mandarin Chinese utilizes word order to express temporaVlogical sequence extensively. In other words, juxtaposition of phrases or clauses almost always means temporal sequencing, if not otherwise indicated. If so, then a series of events verbally represented one after another would be signaling their occurrence in that order without singling out which of the events is the culminating one. This extensive use of word order for temporal and logical sequencing, therefore, makes it necessary for Mandarin Chinese to employ an overt peak-marking device. The selection of -Ie for this use is perhaps because of its foregrounding function as a perfective aspect marker.

The other two facts will be presented in connection with some other discourse functions of -Ie in the next two section.

Halliday and Hasan (1978) point out lexical cohesion as one of the najor cohesive ties in English. Lexcial cohesion has three types: repetition, synonymy and

collocation. The same devices work in other languages as well, including Mandarin Chinese. Chang argues that in order for events to form a genuine event line, there must be lexical cohesion. In (62) above, for example, the verb phrases piao yang 'float the ocean', guo hai 'cross the seas' exhibit synonymy. Futhermore, the verbs piao 'float', guo 'cross', and chuan 'transmit' all involve motion and direction, exhibiting lexical collocation. When no lexical cohesion is evident in a series of verbal expressions, something other than a sequential order is needed to make their relation explicit. Oddly enough, -Ie is again called upon for this job. This time it is for its function for marking 'anteriority' that -Ie is given the job.

The following examples illustrate this point: (65.a) Wo chiwan-Ie fan jiu lai.

I eat-finish-PFV rice then come 'I'll come when I dinish eating.' b) Ta zai ner zhu-Ie liangge yue cai zou de. she/he at there live-PFV two-M month then leave DE 'She/he stayed there for two months before she/he left.'

In (65), neither chiwan 'fmish eating' and lai 'come' nor zhu 'live' and zou 'leave' exhibit any lexical collocation. Their sequential word order would only denote that one happens or happened after the other. There is a need for some marking to explicitly indicate that they are not just some unrelated events which happen to occur or have occurred one after the other in time. They are in fact closely related.

The perfective -Ie comes in here to perform the function, which Chang calls 'anteriority'. By 'anteriority' he doesn't just mean that one event precedes the other but also that the first one holds a cause-effect to the one that follows. Note that the conjunctive adverbsjiu 'then' in (65.a) and cai 'then' in (65.b) are used for the same purpose of relating the two events. Leaving out either the conjunctive adverb or -Ie in the sentences in (65) would result in an odd-sounding though acceptable sentence, whereas omitting both would make the two sentences downright unacceptable.

Here are some longer discourse fragments originally cited by Chang (1986: 100 and 113, respectively): (66.a) Dang wo jiaoxing kaoshang-Ie Bei-shi yisuo shiH gaozhong when I luckily admit-PFV Taipei-city one-M municipal high-school b) jiali ranfang-Ie yichuan bianpao home-in set-off-PFV one-string firecracker c) zuzu gaoxing-Ie completely happy-PFV

haoji tian. several day

'When I was luckily admitted into a Taipei municipal high school, my folks set off a string of firecrackers and we were completely immersed in happiness for several days.'

(67) Jiali sui qiong, ye henxiaxin chou-Ie home-in though poor, yet cruel-decide-mind pull-together-PVF

shikuai ten-M

qian, gei mai-Ie shuang biejiao de pixie. money for buy-PFV pair cheap DE leather-shoe 'Although we were poor, (my parent) nevertheless came to the difficult decision. (They) pulled together ten bucks and bought me a cheap pair of leather shoes.'

In (66), the first two -Ie's are for marking anteriority, i.e. kaoshang 'be admitted into' and ranfang 'set off' are explicitly marked as temporally and logically preceding the third verb gaoxing 'happy'. The third -Ie, however, serves the peakmarking function, as the event is a culminating one in the series. In other words, the three events didn't just happen to have occurred one after another; they rather occurred in a cause-effect relationship. 'To be admitted into a high school' is the cause for 'setting off a string of crackers'. Furthermore, 'to be completely immersed in happiness for several days' is both the effect of 'to be admitted into a high school' and the culmination of the celebration activities.

Similarly, in (67), the -Ie after chou 'pull together' marks the anteriority status of this verb in relation to the next verb mai 'buy'. The -Ie with mai, however, marks the peak-the culmination of the events.

The fact that the perfective aspect marker -Ie is used for two entirely different functions might seem to be puzzling. In fact, there is a very good explanation for it. The peak-marking comes from the fact that the event is the final stage of a series of related events and that, as such, it is the finale of this portion of narrative discourse. Recall that the perfective aspect is defined as 'indicating a bounded situation with a focus on the endpoint of the situation' above. The peak-marking function is obviously derived from the notion of perfectivity by taking the whole series of events as a single event. On the other hand, the anteriority-marking function is derived from the same notion, but by a different route. Here, each of the events in the same sentence is taken separately first, but together they are interpreted as 'one doesn't begin until the end of another'. That is, the end of the first event is marked by -Ie as anterior to the next event.

Support for -Ie marking relative anteriority can be found from semantic studies of the aspect, too. Shi (1990:107-109) states: 'Since completion is the result ofa bounded situation marked as relatively anterior, the verb LE can be identified as a marker of relative anteriority.' Also see Huang and Davis (1989).

Besides marking the peak: and relative anteriority, the aspect marker -Ie is involved in some other non-syntactic issues. One ofthem is the priority for -Ie to occur with a past event expressed in a monosyllabic activity verb.

Chang (1986: 1Oland 114-116) observes that there is a strong tendency for the perfective -Ie to occur with a monosyllabic activity verb for a past event despite other conditions ..

(68.a) Ni Sanshu YIJmg dao-Ie Guilin, wo yao ba ni songdao you third-uncle already arrive-PFV Guilin, I want BA you send-arrived Guilin quo Guilin go 'Your Third Uncle has already arrived in Guilin and I want to send you there.' b) Zhege cun qunian hua-Ie sanwan duo yuan, gai-Ie this-M village last-year spend-PFV 30,000 more dollar, build-PFV shijian fang banqi xueqian ban he xiaoxue ban. ten-M room run-start pre-school class and primary-school class 'Last year this village spent more than 30,000 dollars, built ten rooms, and started (running) pre-school and primary school classes.'

The -Ie's with monosyllabic verbs in (68) are almost mandatory in spite ofthe fact that they represent non-peak: events. Of course, with some changes in the context, some of them may become deletable: e.g. a clause-fmalle may be added to the end ofthe first clause in (68.a), then the perfective -Ie mayor may not be used. This option will be discussed in Chapter 4.

The reason that there is such a strong tendency for -Ie to occur with a monosyllabic activity verb for a past event can be found in the semantic nature of Mandarin Chinese verbs. Chang cites Chu (1978) as stating that Chinese action [Le. activity] verbs don't even express an active attempt without a -Ie or any resultative complement. Now, consider monosyllabic activity verbs. They are by nature verbs without any resultative complement. Therefore, the only way for them to express ' an event having actually taken place is at least to have a -Ie affixed to them. This requirement is so strong that it cannot be waived except in one case where there is another Ie at the end of the clause. This exception is not surprising since the perfective -Ie and clause-final Ie are related to each other, both historically and synchronically.

2.2.5.5. The Non-Occurrence of -Le: Verbs of Saying and Verbs with a Classical Flavor Two other non-syntactic issues that involve the perfective aspect marker -Ie are its non-occurrences with verbs of saying when a quote follows and with verbs having a classical flavor.

The following examples are from Chang (1986:104-105): (68.a) Heizi de niang shaoshao chenmo-Ie yihuer, ba kuangdeng digei Heizi's mother a-little silent-PFV a-while, BA mine-lamp hand-to zhanggu, you zhufu shuo(*-Ie): 'Mingtian ling bu ling husbadn, then mindingly say tomorrow receive not receive mianfen, dou zao diar huilai. ' flour, all early a-little return 'Heizi's mother remained silent for a while. She handed the miner's lamp to her husband and remindingly said, "Whether or not you get the flour tomorrow, (try to) come home early.'" b) Di-er tian Mingdi wen(*-Ie) dachen zhe shi zenmo yihui second day Ming-Emperor ask(*-PFV) minster this be how one-M shi, yiwei dachen shuo(*-Ie), na shi fo de fali. matter, one-M. minister say, that be Buhha DE power 'The next day, Emperor Ming asked (his) ministers what that was all about.

One of them said that it was Buhhda showing his power.'

(69) Jige yue hou, wo sui Sanshu quan jia zi Hainan some-M month after, with Third-Uncle whole family from Hainan Dao ru(*-Ie) TaL Island enter-PFV Taiwan 'A few months later, I went to Taiwan from Hainan Island with (my) Third Uncle's family.'

The excerpts in (69) illustrate that a -Ie after a verb of saying (or asking) would render the clause unacceptable ifthere is either a direct or indirect quotation.

While it is easy to see the non-occurrence of -Ie in this position, it is not so easy to explain what motivates such a constraint. Chang (1986: 117-118) thinks that

since the message is contained in a quotation, direct or indirect, the quotation carries more important information than the verb of saying does. As a result, the quotation rather than the 'main' verb of saying is interpreted as the peak or culminating clause in Mandarin. This explanation is quite reasonable despite the fact that in English only the direct quotation is construed as in coordinate construction with the verb of saying or asking (Halliday, 1986:227-234). The argument can be further strengthened by apparent exceptions: (71) A: Ni bu qu, yinggai xian gen tamen shuo yisheng. you not go, should first to them say one-M 'If you are not going, (you) should let them know first.'

B: Wo gen tamen shuo-Ie wo bu qu, keshi tamen yiding yao wo quo I to them say-PFV I not go, but they certainly want me go 'I told them I wasn't going, but they insisted that I go.'

The mini-dialog in (71) may seem to be a glaring counter-example to the claim, but in fact it only shows that the non-occurrence of -Ie with verbs of saying is certainly not a hard and fast rule on the basis of any syntactic formulation. In (71) Speaker A suggests that the listener tell the people involved that she/he is not going. Speaker B gives a response where he/she presupposes that Speaker A has already known she/he is not going. Therefore, the main message in the response is not what the speaker told the other people involved but whether the speaker did the telling. As a result, the verb shuo 'to tell' gets marked with -Ie. On the other hand, the message wo bu qu 'I wasn't going' is almost incidental, and it is there just to provide a contrast with the next clause tam en yiding yao wo qu 'they insisted that I go'. In this case, the use of -Ie, instead of its non-use, contributes to the cohesion of the dialog.

Indeed, the mini-dialog in (71) demonstrates that the same principle for -Ie to mark the most salient message in a discourse underlies both the use and non-use of -Ie with verbs of saying/asking.

The excerpt in (69) shows that a classical verb like ru 'to enter' is not compatible with the perfective -Ie. This constraint is again very reasonable owing to the late emergence ofthe affix. (Cf. Mei Tsulin, 1981.) Further evidence for the classical flavor of the verb lies in its object: Instead of Taiwan, the shorter form Tal is used.

entirety with a focus on the endpoint of this event.

A monosyllabic activity verb demands a -Ie if it represents a past event because of its inability to express an actual occurrence of the event without such a complement. Verbs of saying and asking are NOT suffixed with -Ie if they are followed by a direct or indirect quotation, i.e. if they don't carry the main message of the discourse. Another non-use of -Ie is with an activity verb that has some classical flavor.

We have examined the verbal suffix -Ie on different levels. Semantically, it signals the boundedness ofan event (Li and Thompson, 1981); pragmatically, it indicates fore grounding (Andreasen, 1981); and in terms of discourse, it marks peak event and relative anteriority (Chang, 1986). All these functions are derivable from the basic notion of perfectivity and they complement each other in their contribution to the communicative mechanism of the language. While each of them serves some specific purpose(s) at a given level, together they carry out the complex mission of expressing the complete idea of what 'perfectivity' means in Mandarin Chinese.

The implication here is that the analyst can not just look at the verbal suffix -Ie from one perspective without considering the ramifications from the other perspectives.

Otherwise, the understanding of the Mandarin 'perfectivity' would not be complete.

To show why it is so, we take another look at an earlier example in (66).

(66.a) Dang wo jiaoxing kaoshang-Ie Beishi yisuo shili gaozhong, when I luckily admit-PFV Taipei-city one-M municipal high-school b) jiali ranfang-Ie yichuan bianpao, home-in set-off-PFV one-string firecracker c) zuzu gaoxing-Ie haoji tian. completely happy-PFV several day 'When I was luckily admitted into a Taipei municipal high school, my folks set off a string of firecrackers and we were completely immersed in happiness for several days.'

As we mentioned above, Chang (1986:100) accounts for the three occurrences of

-Ie by saying that the first two mark 'anteriority' and the last one marks 'peak'.

In this section, we have claimed that the main discourse functions of the perfectiveaspect marker -Ie are (a) to indicate the peak event in a sequence of related events viewed as sub-events, and (b) to mark anteriority of an event in a temporal series of events that would otherwise be regarded as unrelated. The seemingly opposite functions are actually derived from the same notion of viewing an event in its

While his account is sufficient for the purpose of the discourse organization of the passage, it has to be complemented by analysis at the other levels for a total view of the structure. The verb gaoxing 'happy' in (c) is a state verb, which doesn't usually permit the coocurrence of the perfective aspect. (For example, the appropriate translation of 'He/She was very happy yesterday' is Ta zuotian hen

gaoxing without a -Ie following the verb.) But with the following duration phrase haoji lian 'several days', the predicate is made to express a bounded event and becomes eligible for the suffix -Ie for perfectivity. On the other hand, the -Ie after kaoshang 'to be admitted through an entrance exam' in (a) is more likely to be left out than the other two because the conjunction dang 'when' indicates that the clause is in some sense a subordinate one and is thus more of a background than the other two clauses. A background clause may not have the overt marking of the perfective -Ie. The optional use of -Ie in this case may certainly be accommodated by one of Spanos' principles, Le. the principle of non-redundancy in (59.a).

The example above illustrates that a full understanding of the suffix -Ie can not come from any single perspective but it must cover consist in all the functions that it performs at different levels. Though 'perfectivity' has mostly been associated with aspect-the viewpoint superposed over the verb semantics, its pragmatic and discourse functions must also be viewed as part and parcel of the total system. More recently, Zhang (1996) reiterates this view ina an aptly argued article on Mandarin aspect in general.

occurs in a subordinate structure, it is negatively associated with 'indpendence'. others are all positively associated with it. They are listed in (72) below.

-Ie

+

-qilai

+

-guo

+

+

+

+

+

+

+

The

+

zai-zhe

2.3. Aspects and Foreground/Background As far as I know, there has been no wcrk done on the discourse functions of the experiential-aspect marker -guo and inchoative-aspect marker -qilai. It is possible, however, to figure out their relative strengths vis-a-vis other aspect markers in terms of foregroundinglbackgrounding force. 19 Hopper and Thompson (1980) claim that the higher a clause is in transitivity, the more likely it serves the foreground function. Of the ten features of transitivity that Hopper and Thompson give, we discuss three that are relevant to the evaluation of the aspect markers in this chapter: (a) Kinesis: -Le and -qilai are positively associated with this feature as they are used to indicate the actual happening of an event. -Guo, zai- and -zhe are negatively associated with it because they are not for indicating the actual happening of an event, but for viewing an event as an experience (-guo) and for its duration (zai- and -zhe). (b) Telicity: Only -Ie is positively associated with this feature as it expresses the endpoint of an event. All the others are negatively associated with it. (c) Punctuality: -Le, -qiali and -guo are capable of occurring with punctual verbs and they are positively associated with punctuality; but zai- and -zhe are by definition incapable of doing so and are therefore negatively associated with it.

To the three features, we could add 'subordination' as a fourth one for background. But, because foregrounding features have been used above, we will add 'independence' as a feature in opposition to 'subordination'. Since -zhe always

The result of the evaluation is tabulated in (72) above. It is clear from the configurations of the feature specifications that the perfective-aspect marker -~e is the highest of all in its fore grounding power. The others can be ranked ill a descending order: -qilai, -guo, zai-, and -zhe. Thus, a scale of the degree of foreground/background for the aspect-markers in Mandarin Chinese can be established as follows:

('-------------)

Foreground

Background

With this scale as a point of reference, we will sketch a discourse network of the aspectual system of Mandarin Chinese in the next section.

2.4. The Discourse Network of Aspect Marking The discourse functions ofthe Mandarin Chinese aspect markers are recapitualted below in Section 2.4.1, together with their basic semantic features.

2.4.1. Summary of Aspectual Functions

endpoint; (b) indicates a peak event and therefore appears in foreground; and (c), due to its focus on the endpoint, also serves to explicitly express temporal and logical sequence of events which would otherwise be less obviously so related.

B. -Qilai: (a) marks the beginning of an event; (b) introduces new information; and (c) thus, appears in the foreground. Since it views the event only in its initial stage, it usually has to be followed by something else to bring the larger scene to a finish.

C. -Guo: (a) marks a past occurrence or past occurrences of an event viewed as an experience; and (b), since an experience is by defmition related to the present is more likely to supply background information for the discourse. ' D. Zai-: (a) marks an event viewed in its progression, and (b) as such, is timedependent. When reference time is the present or the discourse time (which is the default time), zai- appears to be time-independent and thus may appear in foreground. Otherwise, it appears in background.

E. -Zhe: (a) marks an event with a focus on its duration and viewed as a state' (b) is dependent on the occurrence of another event; (c) thus, indicate~ subordination and often serves as a manner adverbial; and (d) appears in background.

We give some statistical data and a few longer discourse excerpts to further illustrate the discourse network of the aspectual system. Further justification and detailed discussion will be found in later chapters.

Fang (1992:493-499) gives a total of 109 examples to illustrate the use of -qilai. ~n27 out of the 109 cases, the predicate with -qilai is followed by another predicate In the same sentence; in II others the predicate has either a perfective -Ie or a clause-fmalle. The rest of them end with the -qilai predicate having no indication of whether it is the peak or not. The fact can be stated in another way: While in the majority ofthe examples it is not clear whether or not the -qilai predicate is in the fo.reground, it is nevertheless obvious that in II cases it appears in the foreground With the help of -Ie or Ie and in 27 others it appears in the background with another predicate following each. This confirms our claim that the inchoative aspect marker -qilai is capable of occuring in the foreground.

The following excerpts (74)-(76) are taken from Zhu (1984:106-8). They are parts of a composition written by a junior high student. The composition is used as on~ ~f ~he samples in the section on 'Comment and Improvement on (Student) Wntmg . The suggested changes have been incorporated in the following but they

(74.a) Dang wo fu-zhe menba xiangwai zou shi, during I hold-DUR door-knob toward-out walk time b) zong hui chuanlai yizhenzhen niu jiao sheng ..., always will transmit-come one-M-M cow call sound c) niubeishang qi-zhe yige fangniuwa. cow-back-on ride-DUR one-M cowherd-child d) Ta jiao Acong, he be-called Acong e) shi wojia yige yuanqinde haizi. be my-family one-M distant-relative-DE child 'When I walked out holding the doorknob, there always came some cowmooing ...on the back of the cow was riding a child-cowherd. He was Acong, the child of a distant relative of our family.'

The two instances of -zhe in the excerpt are in background:juzhe 'holding' in (74.a) serves as a manner adverbial for the predicate xiangwai zou 'walk toward the outside' and qizhe 'riding' in (74.c) provides a background by introducing the childcowherd in anticipation ofa fuller description of him in the next two clauses. The period at the end of(74.c) is a little misleading: it makes the clause look like the culmination of a series of events. In fact, the first sentence, (a)-(c), just introduces the cowherd in order for him to be talked about. The switch from the full noun phrasefangniuwa 'child cowherd' at the end ofthe first sentence to the pronoun fa 'he' at the beginning of the second sentence is a clear indication of continuation from what has just been introduced by adding new information. The same narrative could very well be coded in English as ' ...riding on the cow was a child-cowherd whose name was Acong, the child of a distant relative of our family.' The English relative clause at the end is by no means subordinated to the other predicate. (For further detail of relative clauses as subordinate structure, see Fox and Thompson, 1990 and Section 6.1.2.1.1. of this volume.)

Next, we look at another passage for -guo: (75.a) Kexi wo hen pa niu, it's-a-pity I very fear cow b) suoyi yizhi meiyou qi-guo ta. so all-the-time did-not ride-EXP it

t) Zheshi, Acong pao-Ie guolai, this-time, Acong TUO-PFVover-come It is interesting to note that the English translation has to code the predicate with -guo in a subordinate clause. A closer look at the Mandarin conjunctive adverb kexi 'it's a pity' reveals the same structure: it actually takes the whole sentence as its domain and should likewise be interpreted as the main predicate. It becomes clearer when we consider what goes before the sentence. It is a long passage where the girlwriter admires the young cowherd so much that she implicitly wishes she had had more chances to be with him. Finally, she regrets that she didn't have the courage to ride on the cow of the cowherd. The culminating statement cannot be that she didn't ride on the cow, but it has to be her regret that she didn't. A possible paraphrase of (75) is (75') below, which gives a better indication of what is the foreground and what is the background: (75'.a) Wo hen pa niu, I very fear cow b) suoyi yizhi meiyou qi-guo ta, so all-the-tie did-not ride-EXP it c) zhen kexi. real pity

h) jide zhi nao ta de naodai. worry-DE straight scratch he DE head.

'The little fish swimming back and forth looked very lovely. I liked them very much. Mom promised to catch some for me; but she failed to catch any, so I started crying. At this time, Acong ran over and, seeing I was crying, was so worried that he just scratched his ...head.'

The clause with -qilai in (e) above can certainly be taken as the culminating event, but at the same time it can also be treated as the casue of a following event. It is therefore appropriate to rank it higher in foregrounding than -zhe and -guo.

The next excerpt is from Xie (1992:24). The discourse is a story told by a native speaker of Chinese on the basis of a pitcure about a man coming home.

(77.a) ...na shi tade muqin, ...that be his mother

'(That) I was afraid of cows and never rode on the cow was a real pity.'

This excerpt, therefore, serves to demonstrates that the -guo predicate appears in the background.

Next we look at another passage for -qilai: (76.a) Xiao yu youlaiyouqu, little fish swim-come-swim-go,

g) kanjian wo ku, see I cry

feichang ke'ai. very lovely

b) taitai he haizi dou zai ner deng ta ne.20 wife and child all at there wait him NE c) Tade nu'er zheng zai nabian gen nainai yiqi kan yiben his daughter just at there with grandma together read a-M huace. picture-book

b) Wo hen xihuan ta, I very like it/them

d) Nainai jiao ta nian shu, gmadma teach her read book

c) Mama daying gei wo nongjitiao lai, Ma promise for me get some-M come

e) taitai zheng zai-da dianhua, wife just PROG-call telephone

d) keshi meiyou zhuodao, but didn't catch-hold

t) xunzhao you meiyou geng-duo de gongzuo jihui, look-for have not-have more DE job opportunity

e) wo jiu da ku-qilai.

I then big cry-INCHO

g) wen ta jintian zenmoyang, ask him today how

h) zhaodao shenme ...shi you-Ie ma. find what...work have-PFV Q

h) Zai zhe yi shunjian, Shen Yawei bei dadong-Ie. at this one moment, Shen Yawei BEl move-PFV

' ...that's his mother, (his) wife and (his) child, (are) all there waiting for him. His daughter is there reading a picture book with her grandma.

Grandma is teaching her to read. His wife is making a phone call (to) look for some more job opportunity. (She) asks him how the day was and ifhe found any ...ifthere is some work.'

'Song Bingbing, talking with tears in eyes, looked at Shen Yawei eagerly.

The breeze wasc aressing her silk-like long hair. Twinkling tear drops hanging on her cheeks as pale as the moonlight, the pink fleshy lips slightly open, the tilted-up face pushed her warm and sweet breath toward Shen Yawei. At this moment, Shen Yawei was moved.'

Clauses (b) through (t) above all serve to provide a descriptive background for the story .. The peak e~ent in the narrative is the wife's inquiry of her husband upon returnmg home. It IS therefore clear that the zai- predicate in (e) is the background relative to the -Ie predicate in (h).

Shen (1987: I) cites the following passage to illustrate the relationships between the three original 'sentences': clauses (a)-(c), (d)-(g) and (h).

Shen believes that the three original 'sentences' are linked by their meanings. The first one, (a)-(c), 'vaguely tells that Song Bingbing was eagerly looking at Shen Yawei.' The second one, (d}-{g), 'specifically describes how eager she was.' The third one, (h), 'points out the result.' Therefore, 'the first and the second sentences are linked by a general-specific relationship and the third one has an activity-result relationship with the other two.' Regardless of whether Shen is right in his description of the meaning relationships, the ties between the three 'units' can be made much clearer by the formal markers of -zhe and -Ie. All the instances of the durative marker -zhe are for background through subordination, though their interpretations may be different from one another in terms of specific discourse functions. Both the first and the second occurrences of the perfective marker -Ie are for indicating foreground through a peak event, though the second one may also be regarded as marking the end of discourse. (See Section 4.3. I for the latter function.)

We will take up this passage again in our discussions of subordination in Chapter 6 and of the definition of the Chinese 'sentence' in Chapter 9.

The passages in (74}-{78) illustrate the relative strengths of the aspect markers for marking foreground and background in discourse. This function of theirs, however, is not independently performed; it is achieved through their syntactic and semantic functions such as subordination, peak-marking, temporality, etc. They seem to be intervowen into a network that relates all levels of the language in terms of interpretation and usage. What we have done in this chapter is to extend from the traditional syntactic and semantic studies of the aspects in Mandarin to the study of their discourse (and some other pragmatic) functions. By doing so, we hope that we have made some contibution to the better understanding of this intricate and intriguing portion of the language.

(78.a) Song Bingbing shuo-zhe, Song Bingbing say-DUR b) han-zhe lei hold-DUR tear c) qidaide wang-zhe Shen Yawei. eagerly look-at-DUR Shen Yawei d) F~ngr chuidong-zhe ta rousi yiyangde chang fa, wmd blow-move-DUR she soft-silk same-DE long hair e) xiang yuese yiyang cangbai de lianjia shang gua-zhe liangke like moon-color same pale DE cheek on hang-DUR two-M liangliangde leizhu, bright-bright-DE tear-pearl f) na gei ren yi rougan de hong chun weiwei he-zhe that to person with flesh-feel DE red lip slightly c1ose-DUR g) angqide lian shi Shen Yawei xiudao-Ie ta wenxin tilt-up-DE face cause Shen Yawei smell-get-PFV she warn-sweet fenfanf de bixi fragrant DE nose-breath

I. There is another form which may be considered an aspect marker: the reduplication of some event verbs, e.g. kankan 'try to look'. This form is sometimes recognized as marking the tentative aspect. But because of its similarity in form to the syntactic construction of 'verb + cognate object', e.g kan(yi}kan 'take a look' and its dissimilarity from the other aspect markers, it is not always recognized for

its grammatical function. (Cf. Chao, 1968:204-5) 2. Situation types (iv) and (v) are actually a verb plus a complement of some sort. The verbs without a complement, i.e. da, shui, gai, and zou, should probably be better categorized under 'activity'. On the other hand, an activity or semelfactive verb may shift into another type when a complement (including an object) is added to it: tingjian 'hear' and tihuai 'kick (and) damage' as semantic units may be viewed as achievement verbs.

Single achievement and accomplishment verbs in English generally find their Chinese counterparts in the form of ' V + COMP'. (See Chu, 1976; He, 1992.) This is one of the peculiarities that deserves serious consideration in applying a theoretical model to the description of Chinese. It seems that the Chinese verbal system may simply have three situation types: State, Activity and Semelfactive. The other two types are expressed by the combination of a single verb plus a complement, though some of such combinations have already been lexicalized, e.g. kanjian'see'.

3. The quote is originally from Robert Iljic (1987) L 'Exploitation Aspectuelle deja Notion de Franchissment en Chinois Contemporain. Paris: L'Harmattan.

4. L&T's original wording is: 'Sentence ...assumes that s/he went to China and claims that this took place at least once during last year and is now over', where what is over is obviously an event. We think that the event is over in both the perfective and the experiential sentences. The difference is that the experiential sentence explicitly claims that the state associated with the event is over. In the example in (19), the state of being is the result from going to China, i.e. being in China. There will be some more discussion on this problem in Section 2.2.2.2.

5. A topical entity is an animate or inanimate discourse participant presented as the topic of the discourse.

6. Some others who have discussed the suffix as an inchoative aspect marker are Su (1974) and Chu (1983).

7. As far as I know, Su (1974) was the first to treat zai- as the progressive marker. Some PRC scholars do recognize zai as having a progress meaning, but it is usually treated as a variant of an adverb (zheng)zai 'in progression'. (Cf. Lu and Ma, 1985:1 16-117; Beida, 1982:529.)

On the other hand, -zhe is considered a progressive-phase marker in Li et al (1984:329).

8. Verbs of posture, according to L & T (1981:219), are ones that 'denote postures or physical dispositions of an entity at a location.' They include: zuo 'to sit', zhan 'to stand', dun 'to squat', xie 'to rest', gui 'to kneel', tang 'to lie', ting'to stop', shui 'to sleep', etc. While the defmition may exclude others likejang 'to put', gua 'to hang', na 'to hold in hand', /iu 'to save or stay', cun 'to save or deposit', etc.; they behave the same way as the other group with regard to the aspect marker -zhe. In fact, these are exactly the ones that L&T address in another subsection under the title 'Activity Verbs Signaling States Associated with Their Activity Meanings'. The use of the term 'posture' thus appears less justifiable syntactically

than semantically.

9. Smith (1993) calls -zhe the stative-imperfective. In a theoretical sense, the term is more descrpitive than '(non-progressive) durative'.

10. For more detail about the indeterminate nature of coverbs, see Chang (1977).

11. Chu (1987:31) attributes the possibility of such an implication to the Maxim of Quality of Paul Grice's Cooperative Principle in specch act.

12. I was told by some northern speakers that they are not very keen on the perception of this difference between (68.b) and (69), though southern speakers seem to be more aware of it. Thus, there may be some change going on with the subordinating function of -zhe.

13. The plain paraphrasing here is a little different in wording from Chu (1987:32).

14. LE represents the form Ie without differentiating between the perfective suffix and the sentence-final particle.

15. I owe this comment to one of the anonymous reviewers of an earlier version of this book manuscript.

16. In Andreasen's transliteration, the zhao is rendered as the druarive suffix -zhe, for which the same grapheme is used as for zhe. With zhao, there can also can be a potential -Ie after cazhao.

17. For a summary view of the same theme, see Chu and Chang (1987), which is based on a joint paper at the 1985 CL TA meeting.

18. The term 'sentence' apparently is used in a loose sense. As we know, the notion 'sentence' in Chinese is not as well defmed as in English or any other Western languages. One of the ultimate aims of writing this book is to more rigorously define a Mandarin Chinese structural unit that is closely similar to the Western sentence. See Chpater 9 for more detail.

19. For detailed treatment offoregrounding and backgrounding, see Chapter 6.

20. The status of the two instances ofzai in (b) and (c) is unclear as to whether they should be considered prepositions or the progressive aspect markers. But whichever way, they don't affect our point that these predicates are in the background.

(l.a) I wish I had been in Shanghai. (subjunctive) b) I remember I was in Shanghai. (indicative)

Modality Adverbs in Discourse

Mandarin adverbs have largely been studied for their modifying function. Many of them, however, also function as connectives. Li and Thompson (1981:635ff & 653ft) attribute two connective functions to a few dozen movable adverbs: Forward and backward linking. As movable adverbs, their linking function is quite transparent and they are generally recognized as conjunctions at the same time.

There are, however, other adverbs, mostly non-movable monosyllabic ones, whose clause-linking function in discourse is not as obvious because of their non-initial position in the clause. Until recently they have always been treated in terms of their modifying capacity only. Such a treatment, from our viewpoint, is at best incomplete. To remedy this shortcoming in Chinese grammar, we will try to look at the clause-linking function of those adverbs in this chapter.

There are some other noteworthy facts about the non-movable monosyllabic adverbs that perform linking functions in discourse. Many of them are used to express modality and they often form sets whose members are usually similar in certain ways but are quite different in other ways. Those facts may seem to be irrelevant to their discourse function at first glance. But, upon closer examination, they can be considered one of the sources for their discourse function. We will therefore examine some modality adverb sets (which are commonly used but insufficiently understood) to better understand their behavior as connectives.

(2.a) You must fmish the homework in three hours. (obligation) b) You may leave now. (permission) As a result, modality has often been said to be expressed by verbal morphology or modal verb under the grammatical category 'mood' in English and in many other Western languages. In fact, however, modality can also be expressed by other means. One of them is by particles, which we will discuss in the next chapter. Still another is by adverbs. Lyons (1977:451-2) also recognizes as part of modality some sentential adverbs that express the speaker's opinion or attitude 'towards the proposition that the sentence expresses or the situation that the proposition describes.' The following examples are taken from Lyons (1977:451).

(3.a) Frankly, he doesn't stand a chance. b) Fortunately, no one was hurt. c) Possibly, it will rain. d) Wisely, he said nothing.

Palmer (1986:2) also believes that though 'the notion of modality ...is much more vague [than that of mood] and leaves open a number of possible definitions, ...something along the line of Lyons' "opinion or attitude" of the speaker seems promising. ' The treatments of both authors confirm that adverbs genuinely constitute a main portion of modality that expresses the speaker's opinion or attitude. We will thus examine some Mandarin adverbs' linking function in discourse on the basis of the modality that they express.

3.1. Modality, Modality Adverbs and Connectives In this section, we try to explain what is meant by modality in general, why some Mandarin adverbs are called modality adverbs, and how they serve the connective function in the language.

Lyons (1977:787-849) distinguishes between two main categories of modality: ' epistemic and deontic. The former has to do with the truth of the proposition, which is mainly realized in factual vs. counter-factual expressions (e.g. subjunctive vs. indicative); the latter is concerned with obligation and permission, which are mostly expressed by modality verbs in English (e.g. must for obligation and may for

Mandarin Chinese has many adverbs that express modality. The following examples are adapted from Chu (1983:46ft):

shi: For emphatic assertion. E.g. Wo shi yao quo 'I DO want to go.' zhenshi: 'really'. E.g. Zhege ren zhenshi bu xianghua. 'This man is REALLY impossible. ' keshi: For contrast on the nominal immediately preceding it. E.g. Women keshi xihuan zai haibian dujia. ' ...but WE like to have our vacations at the beach.'

b) Adverbs of Evaluation: ke(shi): 'but; however'. E.g. Ke(shi) Ribenhua gen Zhongguohua yiyang nanxue. 'BUT, Japanese is as hard to learn as Chinese.' jiu(shl): For contrast on the predicate following it. E.g. Wo jiu(shl) bu xihuan zheyangde mafan. 'I ruST don't like this kind of trouble.' ye: 'necessarily' in a negative context. E.g. Alasijiaye bu tai leng. 'Alaska isn't very cold (AS YOU MIGHT THINK).'. .. . dou: 'even'-often used with /ian. E.g. Zheyang rongyl de wentl (han) sansUl de xiaohai dou hui. 'An easy question like this, EVEN a 3-year-old knows how (to solve it).' you:'again'-for disapproval. E.g. Ni zuotian cai kanle yichang dianying, zenmo you yao qu kan Ie. 'You just saw a movie yesterday and you are going to see one AGAIN.' ... . . juran: 'unexpectedly'-with disapproval. E.g. Dlanymg mmgxm Juran dangle zongtong. 'Wow, a movie actor became the President! (Could you believe it?)' ben/ai: 'in the fIrst place; to begin with. E.g. Wo ben/ai jiu bu xihuan chi niupai. 'I don't like to eat steaks, TO B~~IN 'YITH.' .. , . . jianzhi: 'downright; just'. E.g. Zhezhong shlpanzhl buke Slyl. ThIS kmd of thing is DOWNRIGHT incredible.' guoran: 'just as expected'. E.g. Kelindun ~oran danxuan Ie. 'ruST AS EXPECTED, Clinton was elected (President).' qishi: 'as a matter offact. E.g. Qishi, Mai-a-mi ye hui xiangdang leng. 'AS A MATTER OF FACT, Miami can be pretty cold.' c) Adverbs of Judgment: yiding: 'defmitely; surely'. E.g. Mingtian yiding hui xiayu. 'It DEFINITELY will rain tomorrow.' juedui: 'absolutely'. E.g. Nijuedui dei kanwan zheliangben shu cai ne~gjige.

'You ABSOLUTELY must fmish reading these two books (otherwIse you won't be able) to pass (the course).' yexu: 'perhaps; maybe'. E.g. Xiage xingqi tianqi yexu hui hao yidiar. 'Next week the weather will PERHAPS be better.'

Though the adverbs listed above are labeled differently, they all express modality. Those in (a) assert the truth of the propositions; those in (b) express the speaker's opinion or attitude, and those in (c) signify necessity/obligation. or possibility. This list, of course, is illustrative in nature rather than an exhaust~ve listing. For example, in the same semantic fIeld as yexu 'perhaps' are others hke dagai 'probably' and keneng 'possibly' and disjunctives like huozhe 'or' and haishi '( either) ...or'. . .

We have given a relatively large number of 'adverbs of evaluation', WhICh

express the speaker's opinion or attitude, because it is this kind of adverbs that most prominently perform the discourse function of linking clauses and even sentences.

3.1.3. The Connective Nature of Modality Adverbs The modality adverbs listed in (4) are mostly illustrated with single sentences. In some cases, it is obvious that single sentences can not satisfy what is needed to show the gist of the adverb. Thus, additional meanings are given in parentheses in the translations, e.g. for ye andjuran. But, even in other cases where no additional meanings are explicitly given, it is quite clear that some implications are present.

For example, a sense of disapproval is implicit in you as well as in its English translation 'again'. The example for ben/ai clearly cannot serve as a complete statement in itself; it must either follow from some other remarks or be followed by a further statement. It is in this sense that the adverbs serve as connectives.

It seems obvious that restrictions on the use of connectives should form an important portion of the grammar. As a matter of fact, while restrictions on the Chinese connectives that may be easily equated to English conjunctions have been 1 treated to some extent, those on others like the ones listed in (4) have not been 2 studied until very recently. We will thus illustrate a few of the restrictions that must be imposed on some of the adverbs in order to understand them properly.

The adverb keshi, which may occur either clause-initially or between subject and predicate, is examined in the following: (5) Wo xiang mingnian xuan Riwen, keshi Riwen tai nan Ie. I think next-year take Japanese, but Japanese too hard LE 'I want to take Japanese next year, but it's too hard.'

The utterances in (5) and (5') differ only in the position of the adverbial connective keshi, but one is felicitous while the other is not. One explanation might be that there is no need to mention the topic Riwen after it has just been introduced. But the explanation is refuted by the presence of the two instances of the noun Riwen in (5).

Another explanation might be that the repeated forms are too close to each other in (5'). It might be so. But there can be other reasons why (5') is infelicitous. Let us look at some more examples.

(6) A: Wo xiang mingnian xuan Riwen.

'I want to take Japanese next year.'

B: (i) Keshi Riwen hen nan ou! but Japanese very hard OU 'But Japanese is very hard (let me warn you)!'

(ii) ?Riwen keshi hen nan ou!

'Japanese, however, is very hard (let me warn you)!'

The mini-dialog in (6) shows that the two consecutive occurrences of Riwen are still unacceptable even when they occur in two separate speakers. The mini-dialog in (7) below may clarify the problem.

(7) A: Wo xuele yixueqi de Xibanyawen, hen meiyisi; suoyi xia I study-LE a-semester DE Spanish, very boring; so next xueqi xiang xuan Riwen. semester think take Japanese 'I took Spanish for a semester and it was boring; so I want to take Japanese next semester.'

B: (i) Keshi Riwen hen nan ou! (ii) Riwen keshi hen nan ou!

The responses ofB in (7) are the same as those in (6), yet (i) and (ii) are equally acceptable here. The real explanation here is that A provides a contrast in (7) while no such explicit contrast is available in (6). Contrast on the preceding nominal seems to be exactly what the non-clause-initial keshi is supposed to express.

From the examples, we conclude that, to understand the real differences between the clause-initial and the non-clause-initial keshi, it is not sufficient just to use the structural term 'movable' to describe their different positions. Nor is it sufficient to translate one as 'but' and the other as 'however' in the fashion of English grammar.

It is rather the presence or absence of contrast on the nominal in the context that calls for the use of one or the other. It is, therefore, the discourse context that distinguishes them from each other.

Below are some more discourse segments to illustrate the connective nature of another adverb benlai 'to begin with'.

(8) A: Zhejia guanzi de niupai hen hao, ni zenme bu jiao yike? this-M restaurant DE steak very good, you how not order one-order 'This restaurant has very good steaks; why aren't you ordering one?'

B: Wo benlai jiu bu xihuan chi niupai; zheme gui, wo dangran I to-begin-with just not like eat steak; this expensive, I of-course bu jiao Ie. not order LE

(9) A: Zhejia guanzi de niupai hen hao, ni jiao yike ba. this-M restaurant DE steak very good, you order one-order BA 'This restaurant has very good steaks; why don't you order one?'

B:?Wo benlai jiu bu xihuan chi niubai; tai gui Ie, wo bu I to-begin-with just not like eat steaks; too expensive LE, I not xiangjiao. think order '?I don't like steaks, to begin with. (Theirs) are too expensive for me to order one.'

The dialogs above contrast with each other. In (8), Speaker A assumes that B is not taking any action about the steaks and Speaker B follows up on it by using the statement with benlai. In (9), however, Speaker A doesn't assume anything about Speaker B's action or non-action towards the steaks and B doesn't have anything ofthat kind to follow up on. Thus, the use of benlai is inappropriate.

In order for the use of benlai to be appropriate, it seems that the speaker must agree with something that has been said or implied. Furthermore, benlai does not only indicate an agreement but also serves to follow up on the point by supplying further facts or reason for it. Negative evidence can be deduced by omitting the clause with benlai in B's rejoinder in (9). As soon as it is left out, what remains becomes an appropriate response to A's suggestion. As a matter of fact, just the frrst portion of response without benlai jiu would also be a felicitous answer to the suggestion. Speaker B could even retain the use of benlai by choosing to follow up on Speaker A's assumption of 'the restaurant having good steaks.' All these possibilities are illustrated below in (10). (l0) A: Zhejia guanzi de niupai hen hao, ni jiao yike ba.

'This restaurant has very good steaks. Why don't you order one?'

B: (i) Tai gui Ie, wo buxiang jiao. too expensive LE, I not-think order 'They are too expensive for me.' (ii) Wo bu xihuan chi niupai.3 I not like eat steak 'I don't like to eat steaks.'

(iii) Zher de niupai benlai hen youming; keshi tai gui here DE steak to-begin-with very well-known; but too expensive

Ie, wo bu xiangjiao.

LE, I not think order 'I know the steaks here are well-known; but they are too expensive forme.'

It should be clear by now that the adverb benlai serves to match the present clause to a previous assumption by both agreeing with it and adding further support to it. Any simple translation like the ones in (4) or any purely semantic explanation will not able to do justice to its connecting function between clauses and sentences.

In this section, we have discussed several topics: (i) What is meant by modality, (ii) How modality is expressed, (iii) Adverbs as a means of modality expression, and (iv) Chinese modality adverbs having the tendency to link clauses. In the discussion, an illustrative list of modality adverbs is introduced and a few detailed examples are given to show how important the analysis of discourse context is in understanding such adverbs.

The next section will discuss some sets of modality adverbs in terms of their linking function in discourse.

3.2. Modality Adverbs and Their Discourse Function In this section, we will discuss in detail the discourse functions of three sets of modality adverbs. They are: (i)jiu and cai; (ii) bing, dao andye; and (iii) you, hai and zai. No translation is given for any of them because the best one can do is a partial translation. Set (ii) might be glossed 'on the contrary' and Set (iii), 'again'.

But, the glosses only show some of the similarities among the members and no difference is revealed between them. This non-distinction of the differences between them has been one of the main difficulties of treating modality adverbs in Mandarin grammar, especially in grammars for foreign learners. The situation with Set (i) is even worse. It simply defies any translation without specific contexts. This is perhaps the main reason why quite some attention has recently been given to this set. Representative work can be found in Biq (1987) and Mei-chun Liu (1991 & 1993).

The following descriptions are mostly adapted from previous researchers cited wherever appropriate.4 Reanalysis in terms of discourse is added by the present author.

Before starting our discussion, there are a few disclaimers to make. The number of adverbs to be treated are minuscule in comparison with the total number in the language. Beida (1982) lists 466 main entries for adverbs. Our treatment only serves as a demonstration for what can be better done with such adverbs. The reason we put the adverbs in sets is that we can limit our treatment to the similarities and differences between the members of the sets. Some of the adverbs may have other meanings and functions, which are not treated or even mentioned unless they are relevant to our discussion of the relations between the members. For example, ye is not treated for its function to connect two events/situations for their categorial similarity.

The most intriguing aspect of this pair of adverbs is, as Biq (1988:72) puts it, that they sometimes appear synonymous with each other and sometimes antonymous.

In her analysis, Biq unravels the puzzle by treating both as 'quantification adverbs which place four types of focus on some element involved in speech: parametric, limiting, emphatic and temporal.' The examples below (from Biq, 1988) serve to explain her analysis.

(a.i) Zhangsan lai, Lisi cai quo ZS come, LS CAI go 'Only if Zhangsan comes will Lisi go.' ii) Zhangsan lai, Lisijiu quo ZS come, LS nu go 'IfZhangsan comes, Lisi will go.' (b.i) Zhangsan chi sange pingguo cai bao.

ZS eat three-M apple CAI full 'Only after eating three apples does Zhangsan get full.' ii) Zhangsan chi sange pingguo jiu bao.

ZS eat three-M apple JIU full 'When he eats (as few as) three apples, Zhangsan gets full.' (c.i) Zhangsan mingtian cai zou.

ZS tomorrow CAI leave 'Zhangsan won't leave until tomorrow.' ii) Zhangsan mingtian jiu zOU. ZS tomorrow leave

nu

In (11), both cai andjiu are said to be used for 'focus on condition'. In (a), Lisi's going is conditioned on Zhangsan's coming; in (b), Zhangsan's getting full is conditioned on eating three apples; and in (c), Zhangsan's leaving is conditioned on the time 'tomorrow'. The difference between the (i) and (ii) sentences is that cai indicates that the condition is necessary while jiu indicates that the condition is sufficient.

(a) Zhangsan cai/jiu kanwanle diyi zhang.

ZS CAIIJIU read-finish-LE first chapter 'Zhangsan only fmished reading the first chapter.' (b) Cai/Jiu sange ren lai canguanle huazhan.

CAIIJIU 3-M person come visit-LE painting-exhibit 'Only three people came to (visit) the painting-exhibit.'

In (12), both cai andjiu are said to serve the function of putting a focus on a 'limited set of choices within a specific domain' and they are synonymous. This translates into the English adverb 'only'.

(a) A: Women mingtian qu kan neige dianying, zenmeyang? we tomorrow go see that-M movie, how-about 'How about going to this movie tomorrow?'

B: Zhezhong dianying wo cai/*jiu bu kan ne! this-kind movie I CAI/*JIU not see NE 'This kind of movie, I (for one) won't see it!' (b) A: Women dou qu kan dianying, ni weishenme bu qu? we all go see movie, you why not go 'We are all going to a movie. Why aren't you going?'

B: Wo bu guano Wo *cai/jiu bu quo I not care I *CAIIJIU not go 'I don't care. I am just not going (as you well know).'

In (13), both cai andjiu put a focus on the proposition of the sentence where they occur. The difference is that cai emphatically denies the proposition as a

presupposed fact while jiu emphatically confirms it as such. This difference explains the unacceptability ofjiu in (a) and cai in (b).

Recall the analysis of ben/ai in Section 3.1.3, where it is recognized as an adverb for agreeing on a previous assumption and adding further information to support the assumption. We noticed that it conveniently cooccurs withjiu. When it is deleted, the followingjiu has to be deleted as well. The reason that they come and go together, though redundantly, is that they share the function of confirming a previous assumption/presupposition. That way, they reinforce each other.

(a) A: Wanfan shenme shihou hao de? dinner what time ready DE 'When was the dinner ready?'

B: (Wanfan) cai/*jiu hao. (dinner) CAI/*JIU ready 'It was ready just now.' (b) A: Wanfan shenme shihou hao? dinner what time ready 'When will dinner be ready?'

B: (Wanfan) *cai/jiu hao. (dinner) *CAIIJIU ready 'It will be ready soon.'

In (14), both cai andjiu assert about an inunediacy between the speech time and the time of the narrated event in the sentence. But cai points to a PAST time andjiu points to a FUTURE time.5 Looking through the examples in (lIHI4), we find that, with a few apparent exceptions, the use of cai andjiu has to do with the linking of clauses. They link 'one person's coming' to 'another person's going' in (lO.a), and 'a person's being full' to 'hislher eating three apples' in (lO.b). They link the present statement to a previous one in a conversation by confirming or denying an assumption present in the previous statement in (13) and by focusing on the immediacy of the event time relative to the speech time in (14).

The apparent exceptions are (lI.c) and (12). Though they are simple sentences in structure, they all involve some expectation. In (I2.c) cai expresses a time later than expected andjiu expresses one sooner than expected. Both cai andjiu in (12) indicate that the number mentioned is smaller than expected. In other words, those sentences must occur in a context where an expected time/number/quantity exists, explicitly or implicitly.

What we have seen in the above examples and their discussion is the fact that

the two adverbs do not just occur within the context of a simple proposition. They must follow from some condition, assumption, or expectation. Conditions are explicitly mentioned in the same sentence, as in (II). Assumptions may be explicitly expressed, as in (l3.b) and (14), or less explicitly expressed, as in (B.a).

Expectations seems to be a default explanation, as in (12).

On the basis of the contexts, generalizations can be arrived at concerning the interpretations of the two adverbs. When a default expectation is involved, both cai andjiu carry the meaning of 'less than expected', though there are a few cases where only jiu can be used.6 They are largely synonymous. (But, see examples in (17) below.) When either a condition or an assumption is involved, cai andjiu are antonyms. If the condition is non-temporal, cai asserts that the condition is necessary whilejiu asserts that it is sufficient, as in (I I.a & b). When the condition is temporal in nature, cai means 'later than the expected time' and j iu means 'sooner than the expected time', as in (I I.c). If the assumption is non-temporal, cai indicates a strong disagreement with the assumption while jiu indicates a strong confirmation, as in (13). Finally, if the assumption is temporal, cai means an immediacy between the event time in the past and the speech time (or a reference time) whilejiu means an immediacy between the future event time and the speech time (or a reference time), as in (14). For this reason, cai is strongly compatible with gangang 'just now' andjiu is strongly compatible with mashang or like 'at once'. What is noteworthy here is that, except the default meaning, all the interpretations are contingent on their contexts, whether linguistic or non-linguistic.

Summarizing our discussion of the meanings of the two adverbs, we give the chart in (15) below:

Expectation: (See Note 6)

less than expected, as in (12): scalar contrast

less than expected, as in (12): simple contrast

sooner than expected, as in (11.c.ii) sufficient, as in (I I.a.ii) & (I I.b.ii)

Temporal: Non-Temp:

immediate past, as in (l4.a) strong disagreement, as in (B.a)

immediate future, as (l4.b) strong confirmation, as in (l3.b)

Some more examples are cited from Beida (1982) to further verify the interactions between the contextual environment and the semantic interpretations of the two adverbs. (The number after each example indicates the page number where it is found in Beida.)

(a) Chabuduo chile ershi nian de ku, cai ba tamen daidao zheme almost eat-LE 20 year DE suffering, CAI BA they bring-to this da. (110) big 'Only through nearly 20 years' suffering, did (i) bring them (i.e. the children) up.' (b) Ta biyele, jiu dao biede difang qu gongzuo. (292) he/she graduate-LE, JIU to other place go work 'He/She will leave for work after graduation.'

The cai in (16.a) indicates that only after nearly 20 years of suffering did (i) bring them uIJ---:-anecessary condition. Thejiu in (l6.b) indicates that upon graduation, he/she wIll leave for work-a sufficient condition. But, if the conditions are ~te~ret.ed as tem~or~l, then (l6.a) wi~1be t~en to mean 'nearly 20 years is a long time -I.e. the bnngmg up of the chIldren IS accomplished later than one would like. Likewise, (l6.b) will be interpreted as 'his/her leaving will soon happen after graduation' .

(a) Wo yujian yige qingnian zhanshi,jinnian cai/*jui ershiyi-sui. (109) I meet a-M young soldier, this-year CAI/*JIU 21-years-old 'I came across a young soldier, who was only 21 years old.' (b) Guniang, erzi, guye dou zoule, *cai/jiu shengxia wo zhege lao daughter, son, son-in-law all leave-LE *CAI/JIU remain I this-M old gui. (290) ghost '(My) daughter, son, and son-in-law are all gone, (and) there remains only this wretched old me.'

interchangeable. The expectation in (17.a) is that an average soldier is perhaps in mid- or late-twenties. Thus, the range of ages provides a scale of contrasts and cai, rather thanjiu, indicates such a scalar contrast. On the other hand, the expectation in (17.b) is for the family to be together. The contrast is between togetherness and separation-which is a simple one. Thus,jiu, instead of cai, is used to indicate such a contrast.

(a) Lao Bao jielijieba shuoshang lao bantian, cai shuochu tade daoli.

Old Bao mumblingly talk-up quite long-time, CAI talk-out his reason 'Lao Bao mumbled a long time before he made clear his reason why.'

(109) (b) Liang sui de shihou, wojiu chengle gu'er. (189) two year-of-age DE time, I 1IU become-LE orphan 'I became an orphan at the age of two. ' As 'Old Bao's mumbling' is quantified by a time-duration adverb lao bantian 'quite a long time', (18.a) is interpreted as containing a temporal condition. The cai takes on a 'later/longer than expected time' meaning. But, of course, the sentence may just as well be interpreted as containing a non-temporal condition. Then, the cai will take on the 'necessary condition' meaning and the translation will become 'Only after Old Bao mumbled for a long time did he make his reason clear.' In (18.b), the time connective de shihou rules out the non-temporal condition interpretation and thejiu can only be viewed as indicating a 'sooner than expected' condition-i.e. 'At the young age of two, I became an orphan.'

(a) Ni cai xia shuo. (111) you CAI blind talk 'You (not anybody else) are telling lies!' (b) Wojiu zhidao ta bu ken lai. (292) I 1IU know he/she not willing come 'I know he/she wouldn't come.' (Didn't I tell you that?)

The remark in (19.a) is a retort following an accusation that someone (mostly likely the present speaker) is telling lies. The cai serves to deny the assumption. The statement in (19. b) confirms either that he/she would not come or that the speaker knows that he/she would not come. The jiu indicates such a strong confirmation.

As Beida (1982) does not provide dialogs like those in (13), no examples can be found for illustrating the 'temporal assumption' category.

In this section, we have discussed in detail the contextual implications for the set of modality adverbs cai and jiu. If most of the interpretations of the adverbs depend on the context in which the clause occurs, there can be no denial that the adverbs are mostly used for discourse cohesion/coherence. It is no wonder then that grammars generally fail to capture the important uses of the adverbs by trying to describe them within the confines of isolated sentences ..

Finally, we give a graphic representation of the discourse functions of cai and jiu in the following. The numbering corresponds to that of the examples used above.

Each formula represents the propositions linguistically realized in that order.

Parentheses enclose what is implied but not physically present. The arrow with cai orjiu in the middle means that the two propositions are linked by that adverb, which is understood to occur in the second proposition.

a) Necessary Condition -cai-> result b) Sufficient Condition -jiu-> result

a) (Expectation) -cai-> scalar contrast b) (Expectation) -jiu-> simple contrast

a) Condition -cai-> later-than-usual realization b) Condition -jiu-> sooner-than-usual realization

a) (Assumption) -cai-> strong disagreement b) (Assumption) -jiu-> strong confirmation

a) Past time -cai-> immediate past b) Future time -jiu-> immediate future The direction of the arrow indicates that the proposition on the left is linked by the adverb to the one on the right. The implication here is that the first proposition is subordinated to the second one. Clause order and subordination will be taken up in Chapter 5.

The reader is also referred to Shao (1997) for an insightful analysis of the semantic and syntactic ramifications ofthe adverb cai.