a discourse grammar of mandain chinese16

29
Information has generally been treated in terms of two major categories: Given (or Old) vs. New. Given information is often defined as definite or presupposed and new information, as indefinite or asserted. While they are closely correlated to the semantico-syntactic notions of defmite vs. indefinite and the logical concept of presupposition and assertion, the three sets do not completely overlap and therefore can hardly be used to define each other. Indeed, there are frequent cases of given information in the form of a non-definite expression occurring in the asserted portion of a clause and cases of new information in the form of defmite expressions occurring in the presupposed portion of a clause. In spite of efforts to refme the categorization of given and new information by further subcategorizing them, problem of the kind have not been satisfactorily resolved in the literature. In this chapter, we treat 'given' and 'new' as categories on the basis of the source where information comes from. For the actual use of information, we propose the notion of management of information-i.e. whether a piece of information is used to inform or not. In other words, a piece of information, regardless of its source, may be signalled for high or low informative value within a given context. Thus, the structure of information is viewed as consisting of two tiers: source and management. These two tiers are independent of each other, though there is a strong correlation between given information and low informative value on the one hand, and between new information and high informative value on the other. In the sections that follow, we will look at the problems inherent in the dichotomy of given and new information as categories irrespective of source or management. We will try to find solutions to the problems. In pragmatics, information is generally classified into two categories: Given (or Old) and New. This tradition can be traced back to the Prague School view, where new Information is defmed as 'information that the addressor believes is not known to the addressee,' and given Information as information 'which the addressor believes is known to the addressee (either because it is physically present in the context or because it has already been mentioned in the discourse.)' (Brown and Yule, 1983:154). Halliday (1967) elaborated on the notions by applying them to spoken English and established correlates between intonation and information status. Later scholars, notably Chafe (1976) and van Dijk (1977) among many others, tried to link information status to syntactic structure with some success. But, after a lengthy discussion of both the prosodic and structural approaches, Brown and Yule conclude: (1983: 189) It is certainly the case, as Halliday has always insisted, that information status is determined, not by the structure of discourse but by the speaker. It is also certainly the case that there are no 'rules' for the specification of 'new' or 'given' status by the speaker. There are, however, regularities .... regularities which permit us to make statements like 'speakers usually introduce new entities with indefinite referring expressions and with intonational prominence' and 'speakers usually refer to current given entities with attenuated syntactic and phonological forms.' The regularities that Brown and Yule refer to are just correlations between information status and the various signals for notions like definiteness, presupposition and assertion. Within the existing frameworks of information status, to look for perfect correspondence between them would be futile and would only lead to disappointment and confusion. We will therefore work out a different system to accommodate the general correspondences as well as the occasional discrepancies in a more cohesive way. Before doing that, however, we need to look into the problems first. 5.1.1. Information Status and Syntactic Correlates ~rown ~d Yule (1983:170-1) present five different 'syntactic forms' for given mformatIon. In fact, these five can hardly be recognized as different 'forms', but they are rather five sources which given information may be said to come from. They are illustrated below in (1)-(5). (La) Yesterday I saw a little girl get bitten by a dog. b) I tried to catch the dog, but it ran away. (2.a) Mary got some picnic supplies out of the car. b) The beer was warm. (3.a) I saw two young people there. b) He kissed her. (4.a) Look out. b) It's falling.

Upload: eva-li

Post on 28-Mar-2016

264 views

Category:

Documents


6 download

DESCRIPTION

 

TRANSCRIPT

Page 1: A discourse Grammar of Mandain Chinese16

Information has generally been treated in terms of two major categories: Given (orOld) vs. New. Given information is often defined as definite or presupposed andnew information, as indefinite or asserted. While they are closely correlated to thesemantico-syntactic notions of defmite vs. indefinite and the logical concept ofpresupposition and assertion, the three sets do not completely overlap and thereforecan hardly be used to define each other. Indeed, there are frequent cases of giveninformation in the form of a non-definite expression occurring in the assertedportion of a clause and cases of new information in the form of defmite expressionsoccurring in the presupposed portion of a clause. In spite of efforts to refme thecategorization of given and new information by further subcategorizing them,problem of the kind have not been satisfactorily resolved in the literature.

In this chapter, we treat 'given' and 'new' as categories on the basis of thesource where information comes from. For the actual use of information, wepropose the notion of management of information-i.e. whether a piece ofinformation is used to inform or not. In other words, a piece of information,regardless of its source, may be signalled for high or low informative value withina given context. Thus, the structure of information is viewed as consisting of twotiers: source and management. These two tiers are independent of each other,though there is a strong correlation between given information and low informativevalue on the one hand, and between new information and high informative value onthe other.

In the sections that follow, we will look at the problems inherent in thedichotomy of given and new information as categories irrespective of source ormanagement. We will try to find solutions to the problems.

In pragmatics, information is generally classified into two categories: Given (orOld) and New. This tradition can be traced back to the Prague School view, wherenew Information is defmed as 'information that the addressor believes is not knownto the addressee,' and given Information as information 'which the addressorbelieves is known to the addressee (either because it is physically present in thecontext or because it has already been mentioned in the discourse.)' (Brown andYule, 1983:154). Halliday (1967) elaborated on the notions by applying them tospoken English and established correlates between intonation and information

status. Later scholars, notably Chafe (1976) and van Dijk (1977) among manyothers, tried to link information status to syntactic structure with some success. But,after a lengthy discussion of both the prosodic and structural approaches, Brownand Yule conclude: (1983: 189)

It is certainly the case, as Halliday has always insisted, that informationstatus is determined, not by the structure of discourse but by the speaker.It is also certainly the case that there are no 'rules' for the specification of'new' or 'given' status by the speaker. There are, however, regularities ....regularities which permit us to make statements like 'speakers usuallyintroduce new entities with indefinite referring expressions and withintonational prominence' and 'speakers usually refer to current givenentities with attenuated syntactic and phonological forms.'

The regularities that Brown and Yule refer to are just correlations betweeninformation status and the various signals for notions like definiteness,presupposition and assertion. Within the existing frameworks of information status,to look for perfect correspondence between them would be futile and would onlylead to disappointment and confusion. We will therefore work out a different systemto accommodate the general correspondences as well as the occasionaldiscrepancies in a more cohesive way. Before doing that, however, we need to lookinto the problems first.

5.1.1. Information Status and Syntactic Correlates

~rown ~d Yule (1983:170-1) present five different 'syntactic forms' for givenmformatIon. In fact, these five can hardly be recognized as different 'forms', butthey are rather five sources which given information may be said to come from.They are illustrated below in (1)-(5).

(La) Yesterday I saw a little girl get bitten by a dog.b) I tried to catch the dog, but it ran away.

(2.a) Mary got some picnic supplies out of the car.b) The beer was warm.

(3.a) I saw two young people there.b) He kissed her.

(4.a) Look out.b) It's falling.

Page 2: A discourse Grammar of Mandain Chinese16

The parts in italics in the examples all carry given information. That is, theinformation they carry is assumed to be known to the addressee. The pieces ofinformation, however, come from five different sources. The dog in (1.b) is thesecond mention of the entity that has been previously introduced, Le. a dog in (l.a).The beer in (2.b) within the semantic field ofa previously mentioned entity, i.e.some picnic supplies in (2.a). He and she in (3.b) back to the previous mentionedentities two young people in (3.a). It in (4.b) refers to an entity in the physicalcontext of situation. Do in (5.b) represents the same notion as does the previouspredicate work in Manchester in (5.a). Despite the fact that each piece of giveninformation in (1)-{5) comes from a different source, there are only three syntacticforms: definite noun phrase in (1) and (2), pronominal in (3) and (4), and pro-verbial in (5). Pronominals and pro-verbials are considered defmite forms.

On the other hand, the information carried by a dog in (1), some picnic suppliesin (2) and by two young people in (3) is considered new because it is assumed to beunknown to the addressee at that point oftime. All these expressions are indefinite.

From the above facts, it is claimed that there is a strong correlation betweengiven information and defmite expressions (Le. defmite noun phrase and pro-form).An equally strong correlation can also be claimed between new information andindefinite expressions.

The problem, however, is not that straightforward. A number of questions canbe raised with regard to (i) the adequacy of the dichotomy of given vs. new, (ii) thechoice between pronominal and defmite noun phrase, and (iii) the degree ofconsistency in the correlation between given and definite and especially betweennew and indefmite. While the choice between pronominal and definite noun phrasecan very well be attributed to the organization of discourse units and will be furtherdiscussed in Chapter 8, we will take up the other two problems: the degree ofcorrelational consistency and the adequacy of the dichotomy of given vs. new in thenext two sections.

As is well known, defmite and indefinite are semantic notions and they don'tnecessarily have a one-to-one formal representation in a given language. Besides,they may be represented at more than one level of linguistic expression, e.g.morphological and syntactic. Chinese as well as English appears to be no exceptionto such complexity. For example, the same definite expression may represent bothdefinite and indefinite concepts in the following:

(6.a) The bald eagle we saw yesterday was a rare sight.b) The bald eagle is almost distinct.

!he bal~ eagle ~ (~.a) is defmi.te in b~th ~orm and meaning but the same expressionm (6.b) ISdef~lte ~ form.and mde~mlte m meaning, which is known as the genericuse. The EnglIsh mdefimte form IS even more notoriously ambiguous. 1 On theoth~r hand, Mandarin Chinese doesn't rely as much on morphology for defmitenessas It ~oe~ on word order. The same nominal expression may be interpreted asdefimte . m a preverbal position but as indefinite in a postverbal position.2Contras!mg Mandarin with English thus yields the two outstanding differences: InMandarm th~ preverbal position is a device for marking definiteness and a genericnoun phrase ISnot overtly marked. Both will be discussed at some length in Section5.2. For our present purpose, we only need to keep in mind that definiteness inMandarin.is marked either morphologically by zhe-/nei- and possessive pronounsor syntactically by preverbal position.

Now ~ur task is .to find ~ut how much correspondence there is represented~etw~e~ gIVen/new m~ormatlOn and the notions of definite/indefinite through~mgUlStIC.form. We will first examine the five sources of given information,Illustrated m (1)-{5), t~ see if they also apply to Chinese. The examples in (7)-{14),all taken from Sel~ctlons /rom .Zhong Fu, Vol. 20, except the one without a pagenumber reference m (13), provide data for our examination and discussion.

(7.a) Dangshi Zhongyang Ribao you liangge fukan:that-time Central Daily-News has two-M supplement

b) yige shi...Zhongyang-Fukan, lingyige ze shione-M be ...Central-Supplement, another-M whereas be

Zonghe-Fukan.General-Supplement

c) Dangchu sheji de shihou, Zong-Fu de neirongat-beginning design DE time, Zong-Fu DE contents

yuan yi zhishi-quwei wei zhu,originally take knowledge-interest be major

d) buguo jigeyue yihou, Zhong-Fu yu Zong-Fu debut a-few-M-month later, Zhong-Fu and Zong-Fu DE

neirong jianjiande chongfu qilaL (p.3)contents gradually overlap begin

'At that time, the Central Daily News had two supplements: one was theCe~tral Supplement and the other, the General Supplement. The initialdeSign was for the General Supplement to be mainly for general

Page 3: A discourse Grammar of Mandain Chinese16

knowledge and interest, but in a few months the contents of the twosupplements started to overlap.'

In (7.c), Zong-Fu is old information because it is the second mention of the entityreferred to by a previous noun phrase Zonghe-Fukan 'General-Supplement' in (b).In (d), Zhong-Fu and Zong-Fu are mentioned for another time as old information.Two likely candidates that might be recognized as devices for signalling oldinformation here are proper names and the preverbal position, both of which are fordefinite concepts. Proper names, however, do not figure prominently for oldinformation in this case. Note that proper names Zhongyang-Fukan and Zonghe-Fukan are used even when the two supplements are mentioned for the first time andtreated as new information. As new information, they are placed in the post-verbalposition. To see how much the preverbal position figures in signalling oldinformation, we present another example in (8) below.

(8) (Shop-Keeper Luo and Hu Sao are next-door neighbors. Hu brought Luoa plate of steamed stuffed buns a few hours ago and now she appears atLuo's door unannounced:)

a) L: Wuli zuo, wuli zuo.house-in sit, house-in sit

'Come on in, come on in!'

b) H: Wo bu zuo. Bapanzina gei wooI not sit BA plate get for me.'I won't. Just give me the plate.'

c) L: Shenmo panzi?what plate'What plate?'

d) H: .. .langge3 chile wode baozi bu ren,how-come eat-PFV my steamed-stuff ed-bun not acknowledge,

panzi dou bu huan gei wooplate even not return to me

'How come you don't remember eating my stuffed buns? And youwon't even return the plate to me?'

e) L: Aiya! Yuanlairuci. (Luo Laoban xiaozhe ba panzi di geiShoot! Oh-I-see (Luo Owner smile-DUR BA plate hand to

Hu Sao, qianyide shuo:) Duibuqi,O hai mei xi. (p.233)Hu Sao, apologetically say:) 'sorry yet not wash

'Shoot! I plain forgot. (With a smile, Shop-Keeper Luo handed theplate to Hu Sao and said apologetically:) Sorry but I haven't cleanedit yet.'

In (8.b) of the dialog, panzi 'plate' is mentioned for the first time by Hu Sao. But,as it is placed in front of the main predicate na gei wo 'get for me', it is construedas given information. (The presence or absence of the preposition ba doesn't affectthe interpretation.) The intentional placement of the noun phrase in the preverbalposition is highly visible because semantically it is the direct object of the verb na'to fetch,.4 But the basic order of 'V + DO + Prep-IO' (i.e. na panzi gei wo) isinfelicitous here because a post-verbal panzi would be construed as indefinite andthus as new information. In (c), L asks' What plate?' to show that he is not familiarwith the given information 'the plate'. But as soon as Hu mentions baozi 'steamedstuffed bun' and repeats panzi as given information in (d), L remembers what plateshe is talking about and accepts it as given information by using 0 to refer back toit in his speech in (e). The writer of the dialog, knowing the plot well, also acceptsit as given information in the narration of the event of Shop-Keeper Luo handingthe plate to Hu Sao in (e).

This passage explicitly illustrates the important role that preverbal positionplays in signalling given information. At the same time, it also shows the use of zeroanaphora with regard to information status. This latter problem will be furtherdiscussed in Chapter 8.

Next we look at how given information is derived from a previously mentionedentity that defines a semantic field. The passage in (9) serves as an example.

(9) Yao kaifan Ie, ta jinjinchuchude cong chufang daogoing-to serve-dinner LE, she in-in-out-out-DE from kitchen to

canting, yu qitade taitai bangmang ba cai duan daodining-room, with other lady help BA food bring to

zhuoshang.(p.208)table-on

'It's dinner time. She and other ladies, going back and forth between thekitchen and the dining room, help bring the food to the table.'

The noun cai 'food, dish' is treated as given information, again, by placing it infront of the predicate duan dao zhuoshang 'bring to the table'. And it is understoodas a defmite noun. It can be so treated because it is within the semantic field of Ian

Page 4: A discourse Grammar of Mandain Chinese16

in kaifan 'to serve dinner'. The case is pretty straightforward with regard to thesemantic field as one of the sources for given information and one of the ways giveninformation is represented.

Another way of representing given information is by anaphoric forms-Leoforms that refer back to a previously mentioned entity. (But see Chapter 8 for amore accurate definition.) The example in (10) illustrates the anaphoric pronounsignalling given information.

(10) Ying men chulai kai men de ...shi yiwei sanshi zuoyou deanswer door out-come open door DE ...is a-M 30 about DE

nanshi, fa dui ta weixiaozhe shuo, faxing Yan, ta yegentleman, he to her smile-DUR say, he surname Van, she also

weixiaozhe xiexie fa kai men.5 (p. 208)smile-DUR thank him open door

'A gentleman about 30 years old opened the door. With a smile, hetold her that he was surnamed Van. She smiled back and thankedhim for opening the door for her.'

The three instances of the underlined pronoun fa refer back to the gentleman whoanswered the door. They all carry given information. As a pronoun, each instanceis construed as definite. This is another straightforward case of given informationderived from a previously mentioned entity and represented by a defmite pronoun.

There is another form of anaphora, which borders on the personal and deicticpronoun, zhe-/nei- 'this/that'. The following example illustrates it.

(11) Zhangbeimen changjiang zhimindi shidai de kuchuold-folk-PL often talk colonial period DE suffering

danshi neixie dou dui wo tai yaoyuan Ie. (p.13)but those all to me too far-away LE

'Older folks often talk about their sufferings during the colonial period, butthose all seem too far away to me.'

Neixie 'those' refers back to the sufferings mentioned previously and thus carriesgiven information. Both zhe- and nei- are deictic (or demonstrative) in nature butcan be used anaphorically.

Given information can also be derived from the physical context of situation.The example in (12) below illustrates this source well.

(12) (A retired teacher is showing a photo to her former student. The teacheris smoking in the photo.)

Zhe shi wo qishiwu sui shi, wo nei da erzithis be I 75 year-of-age time, my that old son

gei wo pai de. 0 Buhui xiao laoshi chouyan ba? (p. 67)for me take DE. won't laugh teacher smoke BA

'This picture was taken by myoid son when I was 75 years old. Youwon't laugh at my smoking, will you?'

In the second sentence in (12), the subject 'you' does not show up. In ourrepresentation of the text, a 0 is used to fill its position. This non-occurrence or 0indicates given information. Here, its source is not from a previous mention, butfrom the physical context of situation, the presence of the listener.

Finally, instead ofa pro-verb as in English, Mandarin Chinese makes use of theadverbial conjunction ye 'also' to indicate that the predicate contains oldinformation. This ye must be followed by a copy of the first portion of thepreviously mentioned verbal expression. The example in (13) below illustrates thisdevice.

(13) Nimen jintian wanshang qu kan dianying, wo ye quoy'all today evening go see movie, I also go

'(It) y'all are going to the movie tonight, I'm going, too.'

Ye 'also' signals that the predicate presents the same information as a previous onedoes. Thus, what it presents is given information. Note that there is no pro-verb inMandarin and, instead, the first portion qu 'go' of the verbal expression qu kandianying 'go see movie' in the first clause must be repeated after ye.

Another case ofye 'also' seems quite different from the one in (14) below.

(14) Ta rendechu qizhong you yibu Kaitelai, haishe recognize among-them there-is a-M Cadillac, still

you yibu Benchi, qitade kanqilai ye dou shithere-be a-M Benz, others appear also all be

zhanxin faliang. (p.208)brand-new shining

Page 5: A discourse Grammar of Mandain Chinese16

Unlike in (13), the ye in (14) does not introduce a shortened verbal repetition of aprevious predicate. What it does is rather introduce a full predicate, the content ofwhich is implied in a previous situation-i.e. the Cadillac and the Mercedes arebrand new and shining. Since the information has been implied, it is recognized asgiven.

From the observation and discussion of the data in (7)-(14), we conclude thatthere is a high degree of consistency in the representation of given information bydefinite expressions in Mandarin Chinese. Unlike English, which relies almostexclusively on morphological markings for definiteness and given information,Chinese makes use of both morphological markings of zhe-/nei- and syntacticpositions. To sum up, given information may come from the following sources:

(i) A previously mentioned entity, as in (7) and (8), where a preverbal positionfor a repeated noun is the signalling device.

(ii) A semantic field set by a previously mentioned entity, as in (9), where apreverbal position for a new noun is the signalling device.

(iii) A previously mentioned entity, as in (10) and (11), where an anaphoric ordeictic (or demonstrative) pronoun is the signalling device.

(iv) The physical context of situation, as in (12), where a 0 is the signallingdevice.

(v) A previously mentioned or implied predicate, as in (13) and (14), wherethe adverbial conjunction ye 'also' plus a repetition of the first portion of thepreviously mentioned predicate or a full predicate for the implied one is thesignalling device.

So far, we have discussed the representation of given information by definiteexpressions. The other side of the coin is the representation of new information bynon-definite expressions. We will look at it briefly. The following expressions arenon-definite and they occur in post-verbal positions in the examples cited above.They represent new information.

(9) Yao kaifan Ie,....going-to serve-dinner LE, ....'Dinner is going to be served, ....'

(11) Zhangbeimen changjiang zhimindi shidai de xuduo kuchu, ....old-folk-PL often talk colonial period DE many sufferings, ....

'Old folks often talk about the many sufferings during thecolonial period, .... '

(14) Ta rendechu qizhong you yibu Katelai, haishe recognize among-them there-be a-M Cadillac, still

you yibu Bengchi, ....there-be a-M Benz, ....

'She is able to tell that one of them is a Cadillac and another is aMercedes, .... '

They seem to argue forcefully that there is also a high degree of correspondencebetween new information and non-definite expressions in Mandarin, if thepostverbal position is considered a device for non-definiteness. (But see Section5.2.)

What we have done up to this point is to find positive evidence for thecorrespondences between given/new information and definite/non-definiteexpressions. One would inevitably ask: Is there any negative evidence against thishigh degree of correspondence? The answer is a resounding 'Yes'. The nextquestion is then: How are we going to reconcile between the positive and negativeevidence? We will illustrate the problem and try to find a solution to it in the nextsection.

5.1.3. Can Given Information Be Used To Inform?-Sourcevs. Management

It has often been pointed out that a sharp division between given and newinformation is not feasible because there is a certain degree of cross-over fromeither side. Some linguistic forms which ought to be recognized as giveninformation are used as new information and vice versa. E.g.

where the defmite expressions the boss and your performance are judged to beintrinsically given information due to their definite meanings which signalfamiliarity to both the speaker and the hearer. In fact, however, the two nouns areused to inform the listener and thus should be regarded as performing the functionof new information. An even more puzzling case is found in (16) below:

which serves as an answer to a question like 'Who took my car?' As did is aproverb, it has to carry given information. But, for the statement in (16) to be usefulas an answer, there must be some new information. The only element that can carryit is your husband, which is morphologically marked as defmite and should be

Page 6: A discourse Grammar of Mandain Chinese16

given information as the information is derivable from the common stock ofknowledge shared by the speaker and the listener. On the other hand, when thephrase your husband is heavily stressed, the sentence is equivalent to a cleftstructure in writing: It was your husband who did. The clefted noun phrase issupposed not only to carry new information but it is also the focus of the message.As a result, there has been this problem of whether to designate expressions like theboss, your performance and your husband in the above contexts as carrying givenor new information.

Conversely, in (17), the non-definite expression yige 'one-M' is judged to becarrying new information but is used to set the scene and thus serves more likegiven information.

(17) Yige yikuai qian.one-M one-dollar money

'Each (piece) is a dollar-i.e. a dollar apiece. '

These and other similar examples have long confounded the analyst. (See, forexample, Brown & Yule, 1983, Chapter 5; Renkema, 1993, Chapter 13.) Thisconfusion, however, is unnecessary. It all comes from the use of the same set ofterminology 'given/new information' for two different circumstances-where theinformation comes from and how information is used. To see the difference, oneneeds only to recognize the following facts. As an isolated term, a defmite/indefmitenoun phrase signals given/new information by its form. This has to do with wherethe information comes from, i.e. its source, usually referred to as the status ofinformation. In Section 5.1.1, we discussed how the designation of given or new isarrived at. On the other hand, in a context, the same noun phrase may be part of alarger chunk of information which as a whole functions to set the scene for furtherinformation or to inform the reader/hearer. If a noun phrase is used only to set thescene for further information, it is low in informative value. If a noun phrase is usedto inform the reader/hearer, then it is high in informative value. In this sense, howa noun phrase is used is distinct from where its information comes from. Todistinguish the former from the latter, we use the term management of informationin opposition to the source of information. Thus, the boss as a definite noun phrasecarries given information in terms of its information status-i.e. considering itssource. But when it occurs in a context like (15), it becomes part of a larger chunkof information ('talked to your boss about your performance'), which is used toinform the reader/hearer. In this latter capacity, the noun phrase the boss possessesa high informative value in terms of the management of information.

In other words, what we are trying to do is to separate two levels of informationstructure. One is the phrase level where an expression may intrinsically carry givenor new information. This is the SOURCE tier of the system. At this level, there isa high degree of correspondence between given/new and definite/non-definite, ashas been shown in Sections 5.1.1 and 5.1.2. The other tier is the clause level where

expressions are put together for larger information units. This is theMANAGEMENT tier of the system. At this level, information is further organizedfor informative value (or informativeness). Naturally, a noun phrase with giveninformation tends to be used for low informative value and a noun phrase with newinformation tends to be used for high informative value. Furthermore, since thetopic/subject position generally corresponds to low informativeness and thecomment/ predicate position generally corresponds to high informativeness, giveninformation tends to occur in the former position and new information tends tooccur in the latter position. However, this is not always the case. A piece of newinformation may be intentionally placed in the topic/subject position and its valueof informativeness is decreased. Conversely, if a piece of given information isintentionally placed in the comment/predicate position, its value of informativenessis increased. The following examples illustrate this placement of information unitsto increase or decrease their informative value.

(18) Dangshi Zhongyang Ribao you liangge fukan: yige shithat-time Central Daily-New has two-M supplement: one-M is

zhongsuozhouzhide Zhongyang-Fukan, /ingyige shieverybody-know-DE Central-Supplement, other-one-M is

Zonghe-Fukan. (Cf. (7) above)General-Supplement

'At that time, the Central Daily News has two supplements: one is thewell-known Central Supplement and the other, the General Supplement.'

Both yige 'one-M' and /ingyige 'another-M' are non-definite forms. At the nounphrase level, they represent new information. In fact, however, their interpretationmust follow from what has just been mentioned /iangge fukan 'two supplements' .Indeed, they each refer back to a part of it. As a result, they are placed preverballyto reduce their informative value. In past literature, such expressions have beenarbitrarily interpreted either as definite or as given information despite their form.Our interpretation, however, is that the forms are still indefinite in isolation. Thoughthey intrinsically carry new information, the information does not have a high valueof informativeness because of their syntactic positions. Another piece of evidencefor their being low in informative value is their inability to be directly preceded bya you in the same context. This you has the effect of causing the following nounphrase to become highly informative. Compare (19.b) with (19.a) below:

(19.a) Keren lai Ie.guest come LE'The guest(s) is/are coming.'

Page 7: A discourse Grammar of Mandain Chinese16

b) You keren lai Ie.there-belhave guest come LE

'A guest is/Some guests are coming.'

The preverbal keren in (a) is not as high in informative value as the postverbalkeren in (b). As a consequence, the former is generally interpreted as definite andthe latter, as indefinite.

Returning to (18), the two noun phrases Zhongyang Fukan 'CentralSupplement' and Zonghe Fukan 'General Supplement' in (18) are definite by wayof their being proper names. By themselves, they carry given information. Yet, asthey are occurring for the first time in that piece of writing and by virtue of theirposition following the main verb shi 'be', they are interpreted as highly informative.

The example in (18) illustrates very well (i) how an indefinite expression,which carries new information in terms of the source, can be placed in a preverbalposition so that it is interpreted as low in its informative value and (ii) how adefmite expression, which carries given information, can be placed in a postverbalposition so that it is interpreted as high in its informative value. Thus, what has beenviewed as a dilemma of a form carrying given information but used as newinformation is now resolved by distinguishing two levels of informationstructure-{a) status of information which has to do with the source and (b) valueof informativeness which has to do with the management. In spite ofthe fact thatgiven information tends to be used for low informative value and new informationtends to be used for high informative value, we have seen defmite forms with giveninformation treated as having a high informative value and indefinite forms treatedas having a low informative value.

The following two passages taken from Shen (1987) serve to further illustratehow the new system works:

(20) You YIPI nongjiade ma zhan zai heli,there-be one-M farmer-house-DE horse stand at river-in,

shui qizhao xi, lanyangyangde zai nali yaodongwater level-up knee, lazily at there wag

shilinlinde weiba, (p. 36)dripping-wet tail, .

'There's a farming horse, standing in the river, with the water covering upto the knees, and wagging its tail. '

The nouns xi 'knee' and weiba 'tail' should be interpreted as given informationbecause they refer to parts of the horse that has just been mentioned. In other words,they carry given information by virtue of their reference to things in the same

semantic field as a previously mentioned entity. The facts that they are in postverbalpositions (after verbs qizhao 'level-up; come as high as' andyaodong 'wag') andthat they occur without any morphological marking (such as fade 'its', neige'that'}-these facts clearly indicate that they have a high informative value in spiteof the given information they carry with them. It seems that in terms of semantic~elations. with a ~reviously mentio~ed entity, the two nouns do represent ~ivenmformatlOn. But m terms of syntactIc structure, they are highly informative.

In (21) below another example illustrates the same situation:

(21) Zhenli mianqian renren pingdeng, shi xueshu-taolun-zhongtruth front everybody equal, be scholarly-discuss ion-middle

de yige zhongyao yuanze. Jianchi zhege yuanze, jiuyaoDE one-M important principle. uphold this-M principle, must

qiu fayang shehuizhuyi minzhu, guanche ....require develop socialism democracy, thoroughly-carry-out.. ..

'The most important principle in scholarly discussion is that everybody isequal in front of the truth. To uphold this principle, one must developsocialist democracy and thoroughly carry out....'

The noun phrase zhege yuanze 'this principle' is definite in form and carries giveninformation since it refers to the same entity as a previously mentioned noun phraseyige zhongyao yuanze 'an important principle'. Yet, since it directly follows thetransitive verb jianchi 'uphold', it is interpreted as part of a construction with highinformative value: 'to uphold this principle.' Again, in terms of its form and source,zhege yuanze 'this principle' carries given information, but in terms of its context,it is part ofa larger chunk of highly informative material.

From the discussion above, we conclude that, at least in Mandarin Chinesegiven information CAN be used to inform, i.e. used as highly informative. Thi~claim can be made, however, only when two levels of structure are distinguished.~ form s~ch as a defmite expression or a proper name may intrinsically carry givenmformatlOn. Or, a form may be interpreted as carrying given information due to itsrelation with a previously mentioned entity. Both these interpretations operate at thephrase level. They pertain to the tier of source in the structure of information. At~e cla~se level, however, the same form carrying given information may becomehIghly mformative by virtue of its postverbal position regardless of whether it ismarked by the definite marker zhe-/nei- or any such form. This pertains to the tierof management in the structure of information. The reverse applies to an indefiniteexpr~ssion, though an indefinite expression is not as often used preverbally to signal!ow mformativeness as is a definite expression used postverbally to signal highmformativeness.

Page 8: A discourse Grammar of Mandain Chinese16

This two-tiered system of the structure of information does not only interpretsthe behavior of defmite and indefinite expressions in a clause, it has positiveconsequences on other areas of linguistic import. They are (i) the justification forthe dichotomy of given vs. new information, (ii) the interpretation of the functionof word order in Mandarin, and (iii) the distinction between marked and unmarkedtopic structures. For issue (i), we will just say that with the two-tier structureinstalled, there is no more need for further subcategorization of given and newinformation. For details of such subcategorizations, see Prince (1981 and 1992).The other two issues will be addressed in the rest of the chapter.

Word order in Mandarin Chinese became a topic of intense interest in mid-1970 when Li and Thompson (1975) brought it up by examining its semanticfunction. Since then, many interesting articles and theses have appeared on thetopic. We will briefly discuss some of the more important ones and follow them upwith a hypothesis about the function of word order from a purely discourse point ofview. The papers that will be discussed are Li and Thompson (1975), Lu (1984),Sun and Givon (1985), Wang (1988), Hu (1991) and Xing (1993).

Li and Thompson (1975: 170-185) treat the correlation between the positionof a noun in a clause and the definiteness of the noun as a tendency, which wepresent in (22) below.

(22) Nouns preceding the verb tend to be definite, while those following theverb tend to be indefinite.

(23.a) The noun in postverbal position will be interpreted as indefmite unlessit is morphologically, inherently, or anaphorically defmite.

b) A sentence-initial noun must be interpreted as defmite; it may not beinterpreted as indefinite even if preceded by the numeral yi 'one'.

c) The noun following bei (the agent-marker in the passive construction),although preverbal, is immune to Tendency A [i.e. the tendency in (22)above].

In terms of isolated sentences, the tendency and refmements seem to work verywell. But why is there such a tendency and how meaningful is it? These questionsare essentially what Sun and Givon (1985) asked when they challenged thetendency. After a quantitative analysis of both written and spoken discourse bynative speakers, Sun and Givon (p. 336) conclude that, while it is true that about90% of the preverbal nouns are defmite, an overwhelming number of definite nounsalso occur in postverbal position. According to their count, the proportion of

definite nouns vs. non-defmite ones in the postverbal position in the written textsis 870:321 and the proportion in the spoken texts is 289: 150.7 Their defmite'nouns' are actually defmite expressions including pronouns, names, definite nounswithout marking, nouns marked by demonstratives and nouns with restrictivemodification. Some of their examples (pp. 334-5) follow in (24)-(28).

(24) Qing fa dian xi.ask him order play

'Ask him to select the program (ofa show).'

(25) Ta ning-kai neizhi chang bozi bai/andijiuping.he twist-open that-M long neck brandy bottle

'He twisted open the long-necked bottle of brandy.'

(26) Sun Zhongshan ...rang sheyingshi paixia zhegeSun Yat-sen ... let photographer take-down this-M

nanwang dejingfou.unforgettable DE scene

'Sun Yat-sen ...had a/the photographer take a picture of this unforgettablescene.'

(27) Yigege liewu dou diaojinle fayuxian shehao de xianjing.one-M-M hunt-thing all fall-enter-PFV he advance set-ready DE trap'The game, one by one, all fell into the trap that he had set up in advance.'

(28) Wan Shenshi you kanlekan shoushangde dianwen chiyide shuo:Wan Shenshi again look-a-Iook hand-in-DE telegram hesitantly say:

'Ci dianwen dui Dashuai... duo you buxun zhi ci.''this telegram toward Marshall ...much have insulting ZHI word

Zhang Xun yiba duoguo dianwen....Zhang Xun one-grab seize-over telegram

'Wan Shenshi took another look at the telegram in his hand and saidhesitantly: "This telegram has many disrespectful words directed at YourExcellency." Zhang Xun grabbed the telegram in a quick swoop.'

The italicized portions in (24)-(28) are all defmite expressions. Ta 'he' in (24) isa personal pronoun; chang bozi bai/andijiuping 'long-necked bottle of brandy' in(25) is a noun, which is modified by neizhi 'that-M', a demonstrative; nanwang de

Page 9: A discourse Grammar of Mandain Chinese16

jingtou 'unforgettable scene' in (26) is also modified by a demonstrative zhege'this-M'; xianjing 'trap' in (27) is a noun modified by a restrictive relative clauseta yuxian shehao de 'which he had set up in advance'; dianwen 'telegram' in (28)is definite by virtue of reference to the entity in immediately previous context. Themost problematic is the last item, dianwen 'telegram' in (28). Why does the authorchoose to use an otherwise indefinite form and rely on the context for its definiteinterpretation?

Wang (1988:44), in his comments on Sun and Givon, presents his fmdings onthe correlation between word order and definite vs. indefmite as follows: Out of atotal number of259 tokens, 33% of the objects in the va construction are defmiteand 67% are indefmite. Out ofa total number of360 tokens, 20% of the objects inthe SVO construction are definite and 80% are indefinite. For the OV order,including the constructions ofOV, SOY, OSV and the ba-construction, less than2% (2 out of 120 tokens) are indefinite. He attributes the differences between hisresults and those of Sun and Given to different interpretations of their data.

Whatever the reason and the results, there do exist a large number of postverbaldefinite nouns, though the number of preverbal indefinite nouns is small enough tobe disregarded. Prince's (1981) taxonomy of information status has a category of'unused given', which seems a good candidate capable of providing a viableexplanation for the occurrence of those definite nouns in (24)-(27). Yet, withoutany previous context, we can not say for sure that 'unused' is the explanation.

The most difficult problem, however, lies with the example in (28). Theunderlined noun phrase dianwen 'telegram' toward the end of the example isdefmite in meaning because it refers to the same telegram designated by ci dianwen'this telegram' two lines above it. If so, it is expected to be either preverbal inposition or precede by a morphological marker such as zhe- 'this', nei- 'that', ci-'this', etc. But it is neither in a preverbal position nor is it preceded by any suchmorphological marker. One would naturally ask why it is possible for a noun phraseto be interpreted as anaphorically definite without any overt marking. The answercannot come from the relationship between definiteness and word order; nor can itcome from the correspondence between definiteness and information status. It hasto be found in the management of information. While in Mandarin Chinese,definiteness and given information play an important role in setting the scene orforming the topic; they don't play any part in indicating informativeness. In (28),the second mention of dianwen 'telegram' has nothing to do with the topic. It ismentioned only as part of the new message. Its definiteness and given informationstatus can thus be disregarded and simply relegated to the minimally noticeableanaphoric relation with a previously mentioned noun phrase-i.e. postverballywithout any morphological marking.

All this indicates that viewing definiteness and information status withoutlooking at how a particular piece of information is used-managed-in a clause isa futile exercise. In other words, the explicit morphological form and/or thedeliberate placement of a noun phrase in a certain position must not be regarded

simply as a function of definiteness or given information. It must be regarded as aresult of the interaction of all three-definiteness, information status, andinformation management.

There are other theories of the function of word order in Mandarin Chinese.They are reviewed briefly in the following just for easy reference.

Lu (1984) claims that word order in Mandarin enjoys five distinctive functions.It marks the agent-patient relation, the topic, reference [to second or third person],definiteness vs. indefiniteness, and temporal sequence. Hu (1992) believes thatword order is the only means to code topic in Mandarin Chinese, though there areother coding devices in English such as pronouns and articles.

Besides, two Ph.D. dissertations appeared on the topic of Mandarin word orderby Hu (1991) and Xing (1993). Both of them recognize word order as a generaldiscourse device to achieve textual cohesion and as a specific means toaccommodate patient-fronting. Textual cohesion and patient-fronting are closelyrelated to coding a certain portion of a message as given or new information. Inturn, the coding of given information is the first step toward the selection of topicin discourse. The following passage is taken from Hu (1991:204) to illustrate ourpoint:

(29.a) Xu Fengxian shi chunjie shihou cai jia dao zhe cunXu Fengxian be spring-festival time only marry to this village

lai de xifu,come DE wife

b) 0 shi ge neng shuo hui gan de renwu,be (a-)M able speak able work DE personality

c) 0 jia dao zheli bu ji tian,marry to here not a-few day

d) 0 bei xuan wei fu dui zhang Ie.BEl elect be deputy team leader LE

'Xu Fengxian is a new wife who was married into this village as recentlyas the Spring Festival [Lunar New Year]. (She) is an able speaker andhard-worker. Just a few days after (she) came to the village, (she) waselected deputy team-leader [ofthe commune].'

The three O's in (b)-(d) represent the deleted pronoun ta, which appears as she inparentheses in the translation. Deletion of a noun phrase, or zero anaphora, is a verycommon device to indicate topic in Mandarin.8 As topics, the three deletedpronouns in a row here serve to organize all four clauses into one unit known as a

Page 10: A discourse Grammar of Mandain Chinese16

topic chain. Note that in clauses (a)-(c) the word order is SV ..., that is, ta shi 'shebe ...'in (a) and (b) and tajia 'she was married ...' in (c). The word order in (d),however, is shifted to OV : ta bei xuan ... 'she by (X) elect...' This shift isnecessary to maintain cohesion-Leo the four clauses as a discourse unit known asa topic chain. But the reason that she can play the role of topic is that it carriesgiven information. In other words, word order is a reflection of information statuson the one hand and on the other the explicit device for topic, which in turn servesto organize discourse.

The bei-form in (d) is translated as a passive voice. Its function, Xing (1993)claims, is to move the patient, the 0 in this case, to the front of the clause. Itcertainly is a device to explicitly mark the OV order. But, without it, an OV ordercan still be achieved if a different verb is used: dangxuan 'be elected' instead ofxuan 'elect'.

In our new framework, the reason that 'definiteness', 'given information' and'topic' tend to be placed in the preverbal or clause-initial position is because theyare intrinsically low in informative value. If a definite noun phrase or any piece ofgiven information is meant to be treated as highly informative, it can be placed ina postverbal position or specifically marked. We have discussed the positioningstrategy of raising the informative valu~ in this and the preceding section. In thefollowing section some syntactic structu' ~swill be discussed as functioning to raisethe informative value of a definite noun phrase or any piece of given information.

To sum up, the relative word order of S, V and 0 in Mandarin Chinese hasgenerally been regarded as a reflection of the dichotomy between definite vs.indefinite or given vs. new information. There certainly is a high degree ofcorrelation between preverbal position and defmiteness or given information on theone hand and between postverbal position and indefmiteness or new information onthe other. But the correlation is only a rough tendency and there are many cases ofdiscrepancy which detY explanation. The two-tiered information structure of sourceand management seems to offer an explanation for such discrepancy. Owing to thenature that given information (which is expressed by defmite forms) is intrinsicallyless informative than new information (which is expressed by indefinite forms), theformer tends to occur in the preverbal position while the latter, in the postverbalposition. The positioning of a piece of information in a clause, however, has to dowith the management of information. The preverbal position is for setting the sceneor forming the topic and the postverbal position is for informing the reader/hearer.When a piece of new information is intended to set a scene or forming a topic(though not very often), it can be placed in the preverbal position to do so.Similarly, when a piece of given information is intended to inform, it can be placedin the postverbal position. This is the cases in (20), (21) and (28). Furthermore, aspecific slot of the preverbal position is reserved for an important device for textualcohesion in Mandarin. That slot is the clause-initial position and the importantdevice is TOPIC. There is a lot more to say about topic and topic chain in laterchapters.

Incidentally, it is true that the ba- and the bei-constructions are explicit markersfor (S)OV and O(S)V orders, respectively. (A better label is to use .Agent for S ~ndPatient for 0.) Their primary functions, however, are not to achieve the speCificword order but for ba to increase transitivity (Hopper and Thompson, 1980) and forbei to indicate pejorativity (Chu, 1983 :217).

5.3. Theme, Topic, Focus and Contrast in InformationStructure

Besides defmiteness and word order, there are other sets of notions that have oftenbeen correlated to, or even identified with, given and new information. These aretheme, topic, focus and contrast. Their correlations were pointed out mostpronouncedly in Chafe (1976). In subsequent years, a number of papers werepublished on similar subjects. Among them, was a collection of articles on focusand topicalization in Chinese in Tang, Cheng and Li (1983). This collection consistsofChu (I 983a), Lu (l983b), Cheng (1983) and Tang (1983), among others. Thegeneral consensus ofthe papers is that topic and presupposition are correlated andthat there are certain constructions in Chinese that can be considered devices formarking focus. But none of them specifically treat the information status of topicor focus, though Chu (I 983a) states that specificness, presupposition and topic ascorrelated notions actually belong to three different levels.

In the following subsections, we first claritY some terminological confusion,examine the problems in dealing with the notions, and finally try to find a betterapproach by applying the notion of information management.

5.3.1. Theme, Topic, Contrast and Their Informative Value

There has been quite a bit of confusion between theme and topic. Brown and Yule(1983:68-133) and Renkema (1993:63-66) try to claritY these notions, but therestill remain many questions as to the distinctions between them.9 In Chineselinguistics, however, topic is the more usually accepted term to designate agrammatical unit that serves to link one clause to another by occupying the clause-initial position and taking various forms. Theme, then, is a term more or lessreserved for its use in rhetoric, such as in 'theme of a paragraph or chapter.' Butthere are exceptions. Ho (1993) uses theme and thematic structure in the same senseas most Chinese linguists would topic and topic structure. Following the generaltrend, we will use the term topic to designate the structural unit that serves as a linkbetween clauses. In the unmarked case, this unit is a nominal and occurs in theclause-initial position.

Tsao (1990:191,199,237) tacitly accepts the topic as carrying old (i.e. given)information. Citing from Tang (1980), he recognizes the differences between themembers of each pair in the following as a matter of expressing given information

Page 11: A discourse Grammar of Mandain Chinese16

on different constructions: (Some adjustments have been made in the translation toreflect the differences.)

(30.a) Tade tui duan Ie.his leg break LE'His leg broke.'

b) Ta duanle tui le.1o

he break-PFV leg LE'He broke his leg.'

(31.a) Tade mama si Ie.his mother die LE'His mother died.'

b) Ta sile mama Ie.he die-PFV mother LE'As for him, his mother died. '

The members of each pair differ in topic. Sentences (30.a) and (30.b) have tade tui'his leg' and ta 'he' as their respective topics. And (3 1.a) and (31.b) have tademama 'his mother' and ta 'he' as their respective topics. The reason why differenttopics are selected in each pair is that tade tui 'his leg' and tade mama 'his mother'are treated as given information in the (a)-sentences. And selecting them as topicsis a way to indicate that they carry given information. The same may be said of ta'he' in the (b)-sentences.

On the other hand, Ho (1993) notes the fact that, in the unmarked cases, a'theme' (i.e. our topic) carries given information, but 'local conditions may overridethe globally unmarked pattern of the given within the theme.' (p. 89-90) Unmarkedcases of topic are ones like those in (30) and (31). The following example ofHo'sillustrates local conditions overriding the global tendency of topics carrying giveninformation:

(32) Qian wo meiyou. Ming dao you yitiao.money I not-have. Life rather have a-M'Money I have not; life, I do have one.'

Normally, the topics qian 'money' and ming 'life' would carry given informationand the comments would carry new information. Here, according to Ho, the topicscarry 'new information' and the comments wo meiyou 'I have not' and dao youyitiao 'rather have one' carry 'given' information. While Ho doesn't spell out whatthe local condition is that reverses the information status of topic and comment, wecan easily figure out that it is the contrast on qian 'money' and ming 'life' that

constitutes the local condition. Ho's account of reversing the information status oftopic and comment is at best confusing, if not downright misleading.

However, if we distinguish the informativeness of a form from its status ofgiven/new information, then there is no need to stipulate the overriding principle.The reason that topics are marked by special devices such as contrast is not reversetheir information status, but to increase their informative value. In this way, suchtopics can retain their given information status but can be said to be used to informas part ofthe contrast.

To example (32), we can add a well-known topic structure where the topic maycarry 'new' rather than 'given' information. It is the comparison construction, asfollows:

(33.a) Ta bi wo gao.he compare-to I tall

'He is taller than I.'

b) Ta shuo de Yingwen bi wo (shuo de) hao.he speak DE English compare-to I (speak DE) good.

'He speaks better English than I do'-literally, 'the English he speaks isbetter than the English that I speak.'

In (33.a), what is being compared is ta 'he' and wo 'I'. They are topics carrying'given' information by way of the pronominal forms. In (33.b), however, what isbeing compared is ta shuode Yingwen 'the English he speaks' and wo (shuode 0)'that which I speak'. (Note that yingwen 'English' could not be compared if it werecoded as the direct object of shuo 'speak', as in the English version He speaksbetter English than I do. It has to be made into a topic in order to be compared.)Here, what is being compared is exactly what the speaker wants to inform thelistener of. In this case, the speaker must have already said or implied that he speaksEnglish. Thus, the fact that he speaks the language is given information and notinformative in and by itself. What IS informative is rather the comparison beingmade, including the two entities being compared. The entities in the comparison,represented as the two topics, are therefore part of the essential information beingconveyed. They are highly informative despite their given information status. Inother words, topics in a marked structure like comparison are not ordinary topics.They set the scene as well as serve to inform.

In sum, topic and contrast have generally been considered opposite notions:One carries new information; the other, given information. But when they convergeon the same constituent, as in (32) and (33), the dichotomy of given and new alonedoesn't provide an adequate account for how they can work together. The additionaldimension of information management proves to be an effective tool for the analysisof topic and contrast present on the same constituent.

Page 12: A discourse Grammar of Mandain Chinese16

5.3.2. Focus and Contrast as Means oflnfonnationManagement

We noted above that contrast in Mandarin Chinese is expressed in the topicposition. In spite of the given information status that a topic is usually associatedwith, a contrast is interpreted as high in its informative value. In the same vein,other structures, though less obviously contrastive in nature, can likewise beinterpreted as highlighting highly informative topics. One such structure is the/ian ...dou/ye construction, generally known as a focusing structure. Others are thosemarked by such connectives as guanyu/zhiyu 'concerning, regarding, as to', shuodao 'talking about', etc. These latter ones express a mild contrast. Here, we willonly discuss the /ian ...dou/ye construction.

Again, there is some terminological confusion to clarify.Tsao (1990:249-278) claims that 'the /ian constituent is always a topic ...and

the stress that is associated with it is due to the implicit or explicit contrastassociated with the /ian constituent.' E.g.

(34) Lian Xingqitian ta dou qu shangban.including Sunday he all go work

'Even on Sunday, he goes to work.' (Tsao, 1990:270)

Here 'Sunday' is implicitly contrasted with other days of the week. He denies thatit carries a focus. His reason for the denial may very well stem from the generalbelief that focus carries new information and is in direct opposition to topic whichcarries given information. Therefore, its presence in a topic would be acontradiction in and by itself.

In another recent work, Ho (1993:125) treats the lian ...dou/ye construction asa marked focus. E.g.

(35) Xihuan chi la de /ian zi dou daizhe ye mei guanxi.like eat hot DE even seed all retain also not matter

'For people who like to eat spicy hot food, they don't mind even if theseeds (of the hot peppers) are in there.'

He thinks that zi dou daizhe 'with seeds all retained' is a marked focus by virtue ofthe /ian ...ye form around it. He uses 'marked focus' vis-a-vis 'unmarked focus'. Tohim, an unmarked focus invariably occurs at the end of a sentence, such as in (36)below.

(36) Wo gebo teng.I arm sore'I have a sore arm.'

Because an 'unmarked focus' must occur at the end, the Chinese sentence is codedin such a way that the most informative lexical item teng 'sore' occurs in the finalposition while in English the focus point may be expressed by an attributive, s~rein a sore arm in this case. Furthermore, Ho believes the members of the followmgpair is supposed to differ only in focus: (Ho, 1993:113)

(37.a) Shui liu xia shan.water flow down mountain'The water flows down the mountains.'

b) Shui cong shan shang liu xia.water from mountain top flow down'The water flows down the mountains.'

Obviously, what Ho means by foeus is essentially the unit of the most i?ro~ativeconstituent of the structure that normally occurs at the end of a clause. It IS notthe usual sense offoeus in general or Chinese linguistics.

Here we have just seen an example of how the term may be used variously bydifferent analysts. Then, what IS focus?

There doesn't seem to have been much definitive work on this notion. If it isdiscussed at all, it is almost always in terms of how it is marked, e.g. by the pseudo-cleft sentence, parallel construction, contrast, etc. Among the few linguists whodiscussed focus in its own right are Chu and Cheng. Chu (1983:21) states that 'whenemphasis [important information to be transmitted] is concentrated on a smallportion of a structure, it becomes focus.' Cheng (1983 :55) identifies focus as 'anelement on which the addressee, according to speaker's judgment, will focus hisattention for its significance of message.' In thee sense Chu anc Cheng, both (34)and (35) have a focus because there is a piece of concentrated importantinformation to be transmitted (i.e. the /ian ...dou/ye construction) and the addresseewill focus his/her attention on it for its significance of message. For (36) and (37),however, it is hard to decide where the important information is 'concentrated' andwhere the addressee will focus his attention for the significance of message.Relative to the topic wo 'I' in (36), gebo teng 'arm sore' may carry the importantinformation and catch attention. But relative to the topics wo 'I' and gebo 'arm', thestative verb teng 'sore' alone may do so as well. Without a larger context, it issimply an indeterminate case of whether there is a focus. Similarly, in (37), it is truethat the important information is more concentrated in (b) than in (a), but is itconcentrated enough to be called a focus?

Below we cite some of the focalizing devices from Cheng (1983) for furtherdiscussion:

Page 13: A discourse Grammar of Mandain Chinese16

Ta zuotian meiyou lai shi yinwei shengbing.he yesterday didn't come SHI because sick'The reason he didn't come yesterday was because he was sick.'

Dianying ta yao kan, wo bu yao kan.movie he want see, I not want see'He wants to see the movie, but I don't.'

Shei yao chi dianxin?who want eat snack'Who wants to eat a snack?'

Ta (shi) zuotian laide.he (SHI) yesterday come-DE'It was yesterday that he came.'

Treating (38.c) as possessing a focal structure is problematic and we will notdiscuss it any further. The other sentences, however, can all claim to have a genuinefocus (indicated by underlining). Each focus carries a significant message, whicheasily catches the attention of the addressee. At this point, we are tempted to ask,By what means is the addressee's attention attracted to this particular portion ofmessage? The mechanism to make it possible is the markers prominently employedin the structures. Sentence (38.a) is marked by shi and by the reverse order of cause-effect; (38.b) is marked by a contrast; and (38.d) is marked by (shi) ...de. We canthus conclude that a focus is the constituent that carries a significant message whichthe readerlhearer can easily identify by some special marking without any difficulty.By this criterion, (34) and (35) each do have a genuine focus but (36) and (37)don't.

The motivation to encode information in a focus structure, however, is drivenby a difficulty to arrange information in the topic-comment configuration-Le lowbefore high informative value. The statement in (38.a), for example, presupposes'he didn't come yesterday', which is low in informative value and should occur asa topic. But, as a proposition, this presupposed information here cannot be easilymade into a topic without adding any special marker to it. Thus, a shi is used toindicate it as a topic and low in informative value. Similarly, (38.b) and (38.d)

presuppose 'go to the movie' and 'he came' , neither of which can be made into atopic by simply putting them in front of the comment. Thus, a contrast and a(shi) ...de construction are used to indicate their low informative value.

Conversely, the special markings can just as well be regarded as markers forthe high informative value of the rest of the sentences. Since the rest of the sentencein each case is a single constituent, it can be uniquely labelled as a focus. In thissense, a focus is nothing other than a specially-marked highly-informativeconstituent in a clause. And it has nothing to do with whether it carries given or newinformation. As a result, there shouldn't be anything to prohibit a focus from beinga topic at the same time. The removal of any such prohibition should free theanalyst from their unwillingness to designate certain topics as focus.

In Sum, we have looked at some treatments of focus in Chinese. Havingpresented the relevant facts, we conclude that focus is a piece of (concentrated)information specially marked for the addressee to easily recognize it as the mostinformative portion of the utterance. Whether it is given or new information is quitea separate matter. On the other hand, contrast, such as in (34) and (35), is just aspecial case of focus where the focus is also a topic. It is a case in which the topicis highly informative (and thus usually receives a stress). Here, we are in partialagreement with Tsao (1990:255-278), Ho (1993:124-9) and other previousresearchers. But we believe we have presented a consistent view of all the notionsinvolved: contrast, focus and topic in relation to information structure and to oneanother. Contrast and focus may be expressed as marked topics, and such topicshave their informative value raised from low to high. Therefore, contrast and focuscan be viewed as means of information management to increase the informativevalue of the form in question.

Background and foreground (or grounding) form another set of notions that arebelieved to correlate with given and new information. The two notions are usuallydefined in terms of the advancement of the event line of a narrative, though asimilar definition is believed to be applicable to other genres of discourse.Foreground materials are said to directly contribute to the advancement of the eventline and thus are coded in high transitivity verbs (e.g. activity verbs); backgroundmaterials don't contribute to the advancement and are coded in low transitivityverbs (e.g. stative verb) and/or in subordinate structures. The correlations ofbackground and subordination will be discussed in detail in Chapter 6.

Furthermore, grounding is also believed to bear a high degree of correlation toinformation status. Background materials tend to carry given information andforeground materials tend to carry new information. E.g.

(39) Lao Zhao ba ta shudiao de liangbaikuai qian yingOld Zhao BA he lose-off DE 200-M money win

Page 14: A discourse Grammar of Mandain Chinese16

huilaile jiu zoule.back-PFV then leave-LE

In (39) the event line consists of two sub-events: Old Zhao won back the 200 bucks'and 'he left.' They directly contribute to the advancement of the event line and arethus foreground materials. Sure enough, they are coded as main predicates, whichare said to carry 'new information'. On the other hand, the third sub-event that hehad lost 200 bucks doesn't directly contribute to the advancement of the event lineand thus is background material. Just as expected, it is coded in a relative clausewith a preverbal head NP. Both the relative clause as a subordinate construction andthe preverbal NP are said to mark 'given information'. Example (39) constitutespositive evidence for a high correlation between grounding and information status.

The same event can be told in a different way, as in (40) below:

(40) Lao Zhao xian shudiaole Iiangbaikuai qian, houlaiOld Zhao first lose-off-PFV 200-M money, later

(ba qian) ying huilaile, jiu zoule.(BA money) win back-PFV, then leave-LE

The three sub-events in (40) are the same as in (39) but are organized differently.The sub-event of Old Zhao's losing 200 bucks is coded as foreground and is saidto be 'new information' in an independent clause. By the time when the second sub-event occurs (i.e. when he won it back), the lost money has become 'giveninformation' (i.e. ba qian), and it is coded as background. So far, there is nodiscrepancy in the correlation.

However, there is a third way of reporting the event, as given in (14) below. Itis not considered as felicitous as the other two above.

(41) Lao Zhao ying huilaile ta shudiao de Iiangbaikuai qian,Old Zhao win back-PFV he lose-off DE 200-M money,

jiu zoule.then leave-LE

in the postverbal position. If we rely solely on the dichotomy of given vs. newinformation and defmite vs. indefmite, we will have to account for the infelicity bypointing out the following fact: The relative clause is associated with definiteness,background and given information, but its postverbal position is for indefmitenessand new information. Therefore there is a clash in information status or defmitenessdesignation. This account, however, is untenable in either way. First, preverbaldefmiteness and postverbal indefiniteness is still a controversy, as we discussed inSection 5.2. Secondly, we have seen cases of 'given information used as newinformation'-e.g. a defmite NP used to inform. Now, if we subscribe to thehypothesis that the postverbal position is not for indefiniteness or new informationbut for high informative value, then a new explanation can be offered. We can juststipulate that subordinate structures like relative clauses can only serve asbackground and cannot be raised for high informative value. Therefore, they can notoccur in the postverbal position. Indeed, the infelicity in (41) is a pragmaticproblem rather than a grammatical one.

The relationship between subordinate structure and background will be thetopic of investigation in the next chapter. (Cf. example (13) in Section 6.1.2.1.1.)

This chapter addresses the problem of relations between information and conceptssuch as defmiteness, word order, topic, focus and contrast. It begins with the usualassumption that there are some correlations between information status anddefiniteness. The correlations are found to be very strong but not absolute. Thoughdefiniteness and given information come from the same source, a defmiteexpression sometimes has to be interpreted as 'carrying new information'. Thistreatment is not only confusing but also puzzling. That is, an expression is definiteowing to the given information it carries on the one hand; but, on the other, it ismade to 'carry new information'. In search ofa more reasonable explanation, anexamination of the relationship between definiteness and information status isconducted. Despite the fact that defmite vs. indefmite does strongly correlate to theinformation status of given vs. new, one must add another tier to the structure ofinformation. That is, the tier 'management of information' must be added to theoriginal tier 'source of information'. The source of information-i.e. where theinformation comes from- determines the status of being given or new, while themanagement of information designates an expression as high or low in itsinformative value within a certain message. Thus, in spite of the fact that a defmitenoun phrase with given information is generally placed preverbally in a Mandarinclause owing to the low informative value of the given information it carries, it maylegitimately be placed in a postverbal position to raise its informative value. Thisdoes not only circumvents the awkward way of stating that a definite noun phrasecap be made to 'carry new information'. It also has positive consequences on thefollowing issues: (i) the positive support for the dichotomy of given vs. new

Page 15: A discourse Grammar of Mandain Chinese16

information without any further frivolous subcategorization, (ii) a realisticexplanation for the function of word order in Mandarin, and (iii) a reasonableinterpretation for such structures as comparison, focus and contrast as opposed tothe plain topic structure.

The problem of the dichotomy of given vs. new is found to be an adequatecategorization without any further refinement of the classification if a two-tieredview of information structure is adapted. If, as has been proposed in the precedingsections, defmiteness and information status are viewed at the level of the sourceand, at the level of management, a form is assessed of its value of informativenessindependently of its information status; then such sub-categories as 'unused','inferable', etc. are no longer necessary.

The syntactic device of word order in Mandarin has been said to determine thedefinite or indefinite interpretation of nominals to a large extent. Preverbal nominalstend to be interpreted as definite and postverbal ones, as indefinite. There are,however, numerous exceptions. Definite nominals may occur in the postverbalposition and may not be overtly marked by morphology as such. The discrepancycan, again, be resolved by a discourse need, i.e. the demand for definite entities toappear as highly informative. It is therefore best to look at word order not asdirectly correlated to definiteness, but as a means of information management. Thegeneral unmarked word order is one where materials of low informative value areplaced preverbally and materials of high informative value are placed postverbally.As definite expressions carry given information, which is intrinsically low ininformativeness; they tend to occur preverbally. As indefmite expressions carry newinformation, which is intrinsically high in informativeness; they tend to occurpostverbally. Quite often, however, there is a need for a defmite expression to be(part of) the carrier of a new message; then this defmite expression is placed in thepostverbal position to increase its value of informativeness. Marked word orderdoes not follow the above format but it has to be signalled by syntax or prosody asunusual in its interpretation. The focus and contrastive constructions are among thesyntactic structures to signal such a marked word order.

As a corollary, the term theme is discussed as to how it is used differently thanthe term topic in Chinese linguistics. Theme and topic are similar but distinctconcepts. Chinese linguistics has long adopted topic as the term for the grammaticalform (including zero anaphor) that serves to link clauses by occupying the clause-initial position. Mandarin topics are generally defmite and carry given information.In marked topic structures such as comparison and contrast, however, a topic maybe highly informative despite its defmiteness and given information.

After some clarification in the use of terminology, contrast and focus arediscussed in terms of their information value. It is observed that both of them arehighly informative and their informative value is concentrated on one constituent.While focus is expressed by marked structures like the cleft sentence shi ...de,contrast is a special case of focus that must occur in the topic position.

The correlation between information status and grounding falls under the same

area of discussion as those between information and the other sets of notions. Ourshort discussion indicates that there indeed is a certain degree of correlationbetween background vs. foreground and given vs. new information. But thecorrelation should be better represented as one between background vs. foregroundand low vs. high informative value. The relationship between grounding andsubordination will be discussed in detail in the next chapter.

1. Celce-Murcia and Larsen-Freeman (1983: 177) presents a simplified way ofhow definite and indefinite are distinguished on the basis of speaker's and hearing'sknowledge of specific reference.

2. See Chu (1983:137-143) for the different ways of expressing definite vs.indefinite in Mandarin and a comparison between Chinese and English.

3. Langge is the Sichuan dialectal form of zen me 'how come?'4. The so-called ba-sentence is known as the disposal construction in the sense

that something is (to be) done to the object. The use of ba to prepose the directobject can be better viewed as a device to increase the transitivity of the wholesentence. For the defmition of transitivity see Hopper and Thompson (1980). ButHo (1993) thinks that the function of the ba-structure is to free up the sentence-fmalposition for something more informative.

5. The third person pronouns in this passage are distinguished in gender inwriting but not in pronunciation. Thus, there is no difficulty in translating them. Butif read aloud, the passage would be very hard to understand.

6. In some style of writing one does fmd something like yaodong TADEshilinlinde weiba, literally 'wag ITS dripping-wet tail', but it sounds more like adirect translation from a Western language where its is required.

7. The article addresses other problems such as the function of ba (which the~uthors call Object Marker), the referential distance and potential referentialmterference. They are not our immediate concern here and will not be discussed inthis connection or with regard to Wang's (1988) comments on them.

8. For details of topic and topic chain, see Chapters 7 and 8.9. It has now become especially confusing since theme is used in a totally

different sense in GB.10. The sentences in (41.b) and (41.d) are not quite acceptable to me, if they

stand alone. But judgment of acceptability in Taiwan seems to be quite differentfrom mine and that of many others from mainland China.

11. In fact, Ho's principle of end focus is more like Ernst's (1988) proposal toallow one unit of new information after the main verb.

Page 16: A discourse Grammar of Mandain Chinese16

Subordination is a powerful device of discourse organization. Linguistics, however,has often treated subordination merely as a structural form in syntax without muchconsideration of its function. As such, subordination is usually defmed in terms ofsuch structures as nominalization, relative clause, conjunction, non-fmite verb, etc.This approach may be appropriate in autonomous syntax but is far from adequateif syntax is regarded as part of an integral linguistic system. The inadequacy maybe reflected in questions like the following: What are the purposes for each andevery subordinate structure? Are there different degrees of subordination t~at ~einherently associated with the different structures? Those and other sundarquestions have recently been raised in linguistic literature, and vigorous rese~chhas been carried out to fmd answers to them. (e.g. Thompson, 1987) Some fmdmgsare very revealing and insightful for the better understanding of the functionalnature of subordinate structures in English and some other languages.Unfortunately, very little has been done in this area for Mandarin Chinese. Thischapter makes an effort to examine the discourse functions of subordination inMandarin Chinese-the role that subordination plays in the combination of clausesinto larger units.

More specifically, this chapter will try to clarify the relationships betweensubordination, information status (Le. given or new) and grounding status (Le.background or foreground). In terms of such relations, major subordinate structuresof Mandarin Chinese will be examined to verify their grounding status. Thestructures examined include the relative clause, conjunctions, nominalization, and'non-finite' verbs. Finally, whole passages are used to illustrate how subordinatestructures contribute to the organization of Mandarin Chinese in this new light.

6.1. Grounding, Information, Subordination, and TheirCorrelations

This section examines questions about whether subordinate structures carry giveninformation and therefore serve as background.

6.1.1. Background and Foreground

though they are frequently used and generally accepted by discourse analysts inconnection with narrative discourse. Certain characteristics of grounding have beenmentioned and have often been taken for granted without serious discussion. In theliterature, for example, foreground is considered to be the material that representsthe event line (or story line) of a narrative and thus is temporally ordered, takes non-stative verbs, and is coded in perfective aspect. Background is considered to be thematerial that represents sidetracks and thus does not have to be in temporal order,generally takes stative verbs, and is usually coded in imperfective aspect.Furthermore, background is believed to be correlated with presupposition,defmiteness, given information and subordination while foreground is believed tobe correlated with assertion, indefiniteness, new information and main predicate.

Such descriptions of the correlations, however, are not precise and there arenumerous exceptions. Thompson (1987), for example, fmds that some event-lineevents may occur in subordinate structures for 'the multiple purposes to which awriter puts temporally sequenced predicate.' (p. 451) Among the multiplepurposes are (i) dependency of one event on another, 1 (ii) simultaneity of one eventwith another, and (iii) adverbial and participial clauses for the relating backfunction. The following sentences from Thompson (1987:437-440) illustrate thethree, respectively.

(l.a) Only after he stopped smiling and shrieking did he go to Stephanie andhug her.

c) I picked him up and threw him away from me. [Three interveningsentences] Getting up, he half-smiled and signed 'play'.

In (l.a) the event 'stopping smiling and shrieking' is temporally sequenced with'going and hugging'. Both events should be foreground. Yet, the first event is codedin a subordinate clause, according to Thompson, just to show that there is adependence relationship of this event to the other. In (l.b) 'seeing' and 'signing' aretwo consecutive events. But the former is coded in an adverbial clause to signal itssimultaneity with the preceding event of 'Nim's being excited' rather than as beingconsecutive with 'signing "play'''. In (l.c) 'getting up' has every reason to be in theevent line, but it is coded in a subordinate structure just to relate back to 'throwinghim away' four sentences ago.

Givon (1987), on the other hand, adopts a more dynamic view of grounding,which is quite different from some other views agreed upon in the field. He believesthat the grounding status of a given clause is relative to those of its neighboringclauses. He states:

Page 17: A discourse Grammar of Mandain Chinese16

becomes-in the absence of challenge from the hearer-a sharedpresupposition ('background') at point n plus J. The determination ofwhat is foreground and what is background in actual discourse must thenbe relative to aparticular point in the discourse, the particular frame. Andthe frame in discourse tends to shift, to be reassembled, to be reframed. (p.176)

He obviously claims that what has happened constitutes a background for what'shappening, which in turn becomes background for what's going to happen, if allthem are related-Leo in the event line.

In the same article, Givon (1987:186) goes one step further to challenge thebinarity of foreground and background. To him, the foregroundlbackgrounddistinction 'is useful in carrying us the fIrst step toward a function-based defInitionof an important strand in the thematic coherence of discourse.' But it is dangerousif we treat it with undue rigidity and do not look for fmer distinctions betweenforeground and background on empirical basis.

The above observations suggest that there is a general dissatisfaction with thestatus quo of the binary distinction offoregroundlbackground. Some analysts havetried to fInd more accurate relations between grounding and such other notions asdefIniteness, information status and subordination. Others have attempted toexamine the notions with more empirical evidence. In this chapter we will try toanswer some of the problems raised above on the basis of Chinese discourse.

In linguistic literature there is an assumption that background is correlated to giveninformation, though foreground is not explicitly mentioned as directly correlated tonew information. This assumption has recently been called into question becausesome background structures represented by subordination do not necessarily carrygiven information. To clarify this correlation, we need to go back to what exactlyis meant by given/new information.

As we have noted in Chapter 5, the distinction between given and newinformation has to do with the source of information and is mainly signalled at thelevel of a phrase, especially that of a noun phrase. Yet, the determination ofbackground and foreground depends on relations between events, which arerealized as clauses in form. (See preceding subsection.) In other words, groundingpertains to the inter-clause level. Therefore, structurally speaking, information statusand grounding status operate at two completely separate levels, with an interveninglevel of clause between them. On one hand, it is certainly true that there is a closecorrelation between given/new information and background/foreground, as hasgenerally been observed. But on the other hand, there is no reason to expect a 100%correspondence between them because the representation of information at onelevel of structure does not, and should not, warrant a parallel representation of

grounding in a structure two levels removed from it. In Chapter 5 we noted that inspite of the close correlation between information status and informative value, aconstituent carrying given information can often be placed in a certain position toraise its informative value. It should be equally possible, then, that in a similar waya constituent carrying new information can be placed in a particular structure tobecome background. If this assumption is correct, it will not be surprising at all tofInd certain story-line events serving as background in subordinate clauses, as in (1)above, though they actually carry new information, are highly informative, andwould normally be foreground. Let's take (1.a) for example. As indefmite gerunds,smiling and shrieking in (La) carry new information at the phrase level and theyoccur in a postverbal position for high information value at the clause level. Thepredicate stopped smiling and shrieking is non-stative and advances the story. Itshould normally be foreground. But at the inter-clause level, it is placed in asubordinate structure after he stopped smiling and shrieking to serve as backgroundto did he go to Stephanie and hug her. This example fully illustrates that giveninformation and background don't have to coincide in one constituent. In otherwords, as soon as one recognizes the separate levels of operation, the non-correspondence between the sets of notions is no longer a problem in spite of thegenerally observed correlations.

To contrast the backgrounding in (La), the sentence can be recast as (La')below:

(1.a') He stopped smiling and shrieking and then went to Stephanie and huggedher.

In (1.a'), all the predicate verbs carry new information and they are all foreground.The pronouns he and her carry given information. He occurs in the preverbalposition and thus is low in informative value while her occurs in the postverbalposition and is thus high in informative value. But both of them occur in foregroundclauses. All these facts clearly indicate that there is not any necessary relationshipbetween given information and background on one hand and between newinformatiop and foreground on the other.

To summarize, we claim that a crucial difference between given/newinformation and background/foreground is that the former pertains to thelevel of a phrase but the latter operates between clauses. Thus, one cannot talkabout given or new information carried by a clause, nor can one talk about whethera nominal or a verb is background or foreground to some other constituent. One canonly assign information status to phrasal constituents because information statuspertains to the phrase level. Similarly, one can only assign grounding status toclauses, whether in full or reduced form, in relation to another clause becausegrounding operates at the inter-clause level. Just as we claimed in the precedingchapter, any information unit (whether with given or new information) may beplaced in a position for high or low informative value. Similarly, any constituent

Page 18: A discourse Grammar of Mandain Chinese16

(whether high or low in information value) may occur in a structure for backgroundor foreground. As a result, the terms background information and foregroundinformation, which crop up very often in linguistic literature, can only be taken intheir non-technical senses within our framework.

Integrating all that we have discussed, we offer the following characterizationof background and foreground:

Grounding is a relation between events/situations represented byclauses. As foreground material generally pushes a narrative forward, it isin the event line. It tends to be temporally ordered, takes non-stative verbsand is coded in perfective aspect. As background material doesn'tgenerally push a narrative forward, it is not in the event line. It doesn'thave to be in temporal order, may take stative verbs and is usually codedin imperfective aspect (e.g. zai- and -zhe in Mandarin). Since backgroundis the basis on which new events/situations (foreground) can be built, aprogression from background to foreground is assumed if a combinationof clauses is not marked otherwise. This is the unmarked sequence and wewill use BFP to represent this principle of Background-to-ForegroundProgression. A clause may also be explicitly marked as background bysubordination (such as relativization, nominalization, verbal form,conjunction, etc.) so that other events/situations may stand out asforeground more prominently against it.

Note that we have incorporated 'situation' as well as 'event' in the characterizationin order for it to apply to non-narrative discourse.

In essence, we hypothesize that there are three pragmatic relations thatcontribute to backgrounding. They are: (i) the event line, which also incorporatestemporal order, (ii) the basis for the next event or situation to build on, and (iii) thereduction of discourse weight to make other events or situations stand out moreprominently. We will use the following shorter terms for them: (i) event-line, (ii)scene-setting, and (iii) weight-reduction, respectively.

The remainder of this subsection will be devoted to the confIrmation of theabove hypothesis. It will also serve to illustrate the consequences of adopting thisthree-component hypothesis when grounding is separated from information statusbut is integrated with subordination. The net effect is a natural interpretation of thesmooth organization of discourse. We fIrst look at its application to the analysis ofEnglish discourse. The passage in (3) is cited from Thompson (1987:439).

(3.a) I PICKED him UP and THREW him away from me.

c) I had thrown him so hard that he ENDED UP HITTING the cinderblock wall and not the carpeted floor, as I had intended.

d) I quickly DISCOVERED that there was no reason to feel concernedthat I might have hurt Nim.

This passage is the full version of (l.c), with the three missing sentences fIlled upin the middle. The main participants of the discourse are the author himself andNim, a chimp that he has kept and taught to use a specially designed sign language.We follow Thompson here in using italics to indicate background and capitals toindicate foreground. Her criteria for grounding are temporal sequencing and eventline. Of course, we could add state vs. non-state verbs and perfective vs.imperfective aspect to the criteria. All of these criteria seem to fall into place nicelyexcept the participial form getting up in (e).2 Nim's getting up after he was thrownaway is genuinely part of the event line being reported, but it is coded in asubordinate structure (i.e. a present participle), which, by conventional wisdom,should represent background. Thompson argues that not all subordinate structuresrepresent background but that some subordinate structures can be used to codeforeground events if they serve some other discourse purposes. As we noted in thepreceding sub-section, she gives three such purposes and one of them is for thesubordinate structure to revert back (i.e. to link backward) to an earlier event, whichin this case is Nim's having been thrown away in (a). The explanation soundsreasonable enough. But, there is a problem.

If coding the event of Nim's getting up in the subordinate structure of theparticipial phrase getting up as background is to serve the purpose of reverting backto an earlier event, it would then entail that the full form would not be as able toperform the same function. But, replacing the non-subordinate structure (3.e')below for (3.e) above doesn't seem to make any difference in its ability to revertback to the event ofNim's being thrown away in (a).

In fact, the difference between (3.e) and (3.e') is not, as Thompson claims, in theirability to revert back to an earlier event. It is rather in how much weight is assignedto each of the three events in the same sentence.3 It is well known that it isuninteresting to recount all the events one after another in a narrative by giving eachthe equal weight. Some of them should be signalled as more central to the story thanothers. Ifbackground is interpreted narrowly as non-event-line material, then therewould be no way for a narrator to make the story anything other than a series ofevents in their temporal order. Subordination obviously is one powerful way toreduce the weight of some events so that other events may stand out more

Page 19: A discourse Grammar of Mandain Chinese16

prominently. Thus, subordinating 'Nim's getting up' makes it less central and carryless weight than other events in pushing the story forward. On the other hand, aspart of the event line, Nim's getting up still has to be temporally ordered withregard to the other events: his half-smiling and signing 'play'. Moving it toanywhere else in the sentence makes it incongruous. Obviously, there is an apparentcontradiction: if, as we have claimed, sequencing of clauses automatically indicatesa background-to-foreground progression, why is there a need to use a subordinatestructure to signal background for a clause preceding a foreground clause?

Ifwe recall the characterization in (2), grounding may actually encompass threedifferent pragmatic relations: (i) Event-line, which is the most traditional view offoreground and background, (ii) Scene setting, which forms the foundation for theprinciple of natural background-to-foreground progression, or BFP, and (iii) Weightreduction of a main-line event or situation to make other events or situations standout more prominently. The three relations may, of course, converge and be presentin the coding of one and the same event. The episode in (3) provides a goodexample of all three existing at the same time in the coding of 'Nim's getting up'as a participial phrase at the beginning of the sentence. Getting up is a foregroundclause because it represents one of the events in the narrative. Yet, it is backgroundto the following two clauses (half-smiled and signed "play") because it serves asa basis on which to build the other two events. It is further coded as background ina subordinate structure (Le. present participial), because the author wants to reduceits importance as an event and to give more prominence to the other events that arecoded in the fmite forms.

Incorporating three different pragmatic relations into the notion 'grounding',as we just did, may be cumbersome, but it does not only make the groundingrelationships more concrete, it also gives the classical rhetorical device ofsubordination a linguistic validation. For convenience of discussion, we will startusing the abbreviates names for the three pragmatic relations that we have proposedearlier: event-line, scene-setting, and weight- reduction, in their technical senses.

We have just seen an instance of treating an event-line event as background inspite of its inherent foreground status. Its foreground status, however, is preservedby its temporal ordering with the other events and by its non-stative verb. At thesame time, it also observes BFP in the sense that it serves as a basis for the ensuingevents to happen. Its background status comes from the form of the verb, aparticipial form to signal subordination. Numerically, the form getting up in (3.e)fulfills two foreground criteria but fails one. It is therefore less foreground than anythat fulfills all three but is more foreground than any that fulfills one or none. Wethus conclude, that a binary distinction of foreground vs. background is at bestimprecise. As a result, we claim that grounding is actually of a gradient nature.

The following Chinese example is adapted from the one in (l2.a) of Chapter5. It is given here to further examine the three pragmatic relations under the rubricof grounding and the correlations between grounding and information status.

(4.a) Ni maile zhege diannaoyou buy-PFV this-M computer

b) bu yong,not use

c) fangzheleave-DUR

d) gan shenmo?do what

'You bought this computer and are not using it. What' s (the purpose of)leaving it idle?'

The sentence is dissected into four numbered clauses for ease of discussion. Interms of event-line, each numbered clause is foreground in its own right. But interms of scene-setting (Le. relative to the following one), a given clause isbackground. Because they are not marked otherwise, we canjust follow BFP in theinterpretation of their grounding status. Clauses (a) and (b) form a structural unit.The temporal relation of the event (a) and the state (b) is implicit in the linear orderof the clauses. But it is also explicitly marked by the perfective -Ie, which indicatesanteriority, Le. the event of buying the computer precedes the state of not using it.(Cf. Section 2.3 .1.3.) Clauses (c) and (d) form another unit. The logical relation ofthe two events (scene-setting) is implicitly marked by the linear order of the clauses.The first event of' leaving (it idle)', however, is explicitly marked by the durative-zhe as the background to the second of 'what for?' (Cf. Section 2.3.2.3.) In otherwords, 'leaving it idle' is made less prominent in the discourse than the question'what for?'(weight-reduction).

In evaluating the degree of foregrounding of each clause above, we come tothis conclusion. Since all of them are in the natural order and all the verbs are non-stative, clause order and stativeness of the verb are not consequential in theevaluation. The only one criterion that is applicable is then subordinate structure.In this case, it is expressed in different forms of the verbs, Le. plain verb, with theperfective -Ie and with the durative -zhe. At this point, let us recapitulate thediscourse functions of the verbal suffixes. Recall we ranked the verb forms in termsof their grounding force in Section 2.3.3. In accordance with the ranking, (4.c) canbe considered the most backgrounded. The -Ie in (4.a), however, might be regardedas a foreground marker, but actually it is used here for indicating anteriority andthus rather marks (4.a) as background to (4.b). The relationship between (4.a)-(4.b)as a unit and (4.c)-( 4.d) as another unit seems to be one of coordination, though onehas to follow the other in a linear order to express the cause-consequence relation.

In terms of information status, only the noun phrase zhege diannao 'this

Page 20: A discourse Grammar of Mandain Chinese16

computer' in (a) carries given information. But, there should be a zero-anaphor inboth clauses (b) and (c) to refer back to 'the computer'. The zero-anaphors alsocarry given information. While there is no perfect correspondence between giveninformation and background, it makes sense for clauses (a), (b) and (c) to containgiven information because they are background relative to each of their followingclauses. Only clause (d) doesn't contain any given information, because it is thefmal clause in the discourse and it plays the role of foreground only in its relationwith the other clauses.

In the preceding section, we have mentioned the grounding function of some of theChinese subordinate structures at one point or another. Now that we have clarifiedthe basic notions of grounding, we will more systematically examine subordinationin Chinese. Four subordinate structures will be discussed: relative clause,nominalization, conjunctions and 'non-finite' verb forms. Specifically, they will beevaluated on the basis of the three pragmatic relations of grounding: (i) event-line,(ii) scene-setting, and (iii) weight-reduction. The information status and informativevalue of any component structure will be discussed only as an aside.

Relative clauses in Mandarin Chinese are restrictive. When they have a defmitehead NP, they serve as background and carmot help advance a narrative. Thedefinite head NP mayor may not be explicitly marked by zhe- or nei- . But as longas it occurs preverbally, it is low in informative value and is interpreted as definite(Le. carrying given information). When a morphologically unmarked head NPoccurs postverbally, there are some complications. We illustrate them with thefollowing sentences adapted from Tsao (1986).

(5.a) Gangcai da dianhua fai de shi shei? (p. 25)just-now call telephone come DE be who'Who was it that just called?'-Literally, 'the one who just called waswho?'

b) Zhe shi Li Xiaojie nian shu de difang. (p. 27)this be Li Miss read book DE place

'This is a/the place where Miss Li studies/studied.'

c) Chunjuan shi dajia xihuan chi de shiwu. (p. 29)egg-roll be everybody like eat DE food

'Eggrolls are (a kind of) food that everybody likes to eat.'

The relative clause gangcai da dianhua fai de 'the one who just called' in (a)occurs with a deleted head NP in front of the main verb shi. By its position in themain clause, the head NP is interpreted as definite and the information it carries isunderstood as given information. In our framework, the deleted head NP is low ininformative value by its preverbal position in the main clause.

Since 'someone called' is the only event in the sentence, it is not relevant toconsider whether anything is in the event line or not. However, since the relativeclause represents an event while the main clause represents a non-event, it isforeground to the main clause. On the other hand, it does seem to set the scene forthe main-clause speech act of asking the question 'Who is it?' In this sense, then,it is in the background. Moreover, the weight of the event is reduced by beingplaced in a subordinate relative clause. So it explicitly serves as background to themain clause 0 shi shell 'Who is (the one ...)?' Everything taken together, weconclude that the relative clause in (5.a) is not completely foreground, nor is itcompletely background. It has one foreground feature: it is an event vis-a-vis thenon-event in the main clause. But it has two background features: scene-setting andweight-reduction.

But, by changing the subordinate structure, we can certainly make the eventexpressed by the relative clause more foreground than it is in (5.a). E.g.

(5.a') Gangcai you (yige) ren da dianhua lai, shi shei?just-now there-be (a-M) person call telephone come, be who'Someone called a while ago. Who was it?'

In (5.a') the clauses are equally foregrounded. The first clause has one foregroundfeature: an event vis-a-vis the other non-event. The second clause also has oneforeground feature: it is built on the basis of the preceding one. Thus, they can beregarded as equally foregrounded and thus as forming a coordinate structure.Another strong piece of evidence for the claim that the first clause in (5.a') is moreforeground than its counterpart in (5.a) is the fact that the clauses in (5.a') is lessreversible than those in (5.a). Thus,

(6) Shi shei, gangcai da dianhua lai de? [Reverse of (5.a)]be who, just-now call telephone come DE

'Who is it that just called?'

(7)?Shi shei, gangcai you (yige) ren da dianhua lai?be who, just-now there-be (a-M) person call telephone come

'?Who is it? Somebody just called.' [Reverse of(5.a')]

Both (6) and (7) violate the BFP principle. The principle says that the clauseorder should be from background to foreground if not marked otherwise; in bothexamples, 'Who is the one ...?' precedes 'Somebody called a while ago.' But, there

Page 21: A discourse Grammar of Mandain Chinese16

is one difference. In (6) the violation in compensated by the subordinate relativeclause, which strongly demands that, despite the clause order, the relative clause beinterpreted as background. Thus, the reversed order of the clauses is onlyinterpreted as afterthought. In (7), however, the violation is not compensated by anymarking and therefore the combination of the clauses in reversed order is notacceptable.

Turning to (5.b), the relative clause with its head NP Li Xiaojie nianshu dedifang is potentially ambiguous-It may be interpreted to mean 'a place where MissLi studies/studied' or 'the- place where Miss Li studies/studied'. In the firstinterpretation the head NP carries new information and in the second, it carriesgiven information. By its postverbal position, the head NP is interpreted as high ininformative value regardless of its information status. All this, however, is notdirectly relevant to the grounding of the clauses because grounding status isdetermined by other pragmatic factors. As this sentence is not a narrative, there isno event line to begin with and therefore the first pragmatic factor of event-line doesnot apply. The second pragmatic factor of scene-setting, however, does apply, Le.the event 'Miss Li studied here' or the state 'Miss Li studies here' does provide abasis for the identifying clause Zhe shi ...difang 'This is a/the place.' To express thisrelation, the most basic device is by clause order-Li Xiaojie zai yige difang nianshu 'Miss Li studied or studies at someplace' precedes zhe Oiu) shi niege difang'This is the place. ,4 Putting them together in that order gives us a coordinatestructure that simply links the event and the identifying act without involving thethird factor-weight-reduction. One may, of course, choose to consider the eventor the identifying act as more weighty than the other. Sentence (5.b) is an exampleof assigning more weight to the identifying clause than the event/situation bysubordinating the latter in the form of a relative clause. Thus, Li Xiaojie nianshu deis placed in a subordinate structure to be background to Zhe shi ...difang.

On the other hand, one might choose to give more weight to the event/situationof 'Miss Li studies/studied at someplace'. In this case, the identifying clause wouldbe made subordinate. Instead of a subordinate clause of any sort, an adverbialphrase zai zhege difang or zai zher would be sufficient for the subordination.5 Inother words, the fmal product of subordinating the identifying act would come outas (5.b') below:

(5.b') Li Xiaojie zai zer nianshu.Li Miss at here study'Miss Li studies/studied here.'

Thus, the main difference between (5.b) and (5.b') is one of grounding, specificallyin terms ofweighting-Le. of the event/situation of Miss Li's studying at a placeand the act of identifying the place, one is more prominent than the other indiscourse. One of the other two pragmatic factors of backgrounding-the event-line-is not relevant here. The other factor-scene-setting-is generally expressed

by clause order. In the absence or any other linking devices, this order has to befollowed.

Turning to the relative clause in (c), it again has a postverbal head NP. Thereare two interpretations. It may be used to identify 'eggrolls' as the popular kind offood that the listener is familiar with, Le. 'Eggrolls are the kind of food thateverybody likes to eat.' If so, the head NP carries given information and may bepreceded by the morphological marker neizhong 'the/that kind'. On the other hand,it may also be used to describe eggrolls, Le. 'Eggrolls are a kind of food everybodylikes to eat.' In this case, the head NP does not carry any given information in andby itself and the morphological marker yizhong 'a kind' may appear in front of theNP. Whichever interpretation, the relationship between the two clauses isconstant-dajia xihuan chi chunjuan is background to chunjuan shiyizhong/zhezhong shiwu, where the former carries less weight than the latter.Conversely, one might choose to give dajia xihuan chi chunjuan more weight thanchunjuan shi yizhong/zhezhong shiwu. Then, the resulting sentence would be theone in (5.c').

(5.c') Dajia xihuan chi yizhong/zhezhongjiao chunjuan de shiwu.everybody like eat a-M/this-M call egg-roll DE food'Everybody likes to eat a/the kind of food called egg-roll.'

Now let us consider another case. In connection with (5.b), we mentioned thatwithin a scene-setting relation, the first event/situation can precede the secondwithout any special marking. Or, either may be made subordinate to the other byspecial marking. This so because (5.b) did not involve any temporal order. Whenthe events are also temporally ordered (i.e. in the event line), there is somecomplication. While clause order remains to be a device for the expression of thisscene-setting relation on the BFP principle, there is some restriction on the relativeclause as a subordination for reducing the weight. E.g.

(8.a) Renjia diudiaole yizhang yizi, tajianhuilaile.people throw-out-LE a-M chair, he pick-back-come-LE'Somebody threw out a chair, he brought it home.'

b) Tajianhuilaile yizhang renjia diudiao de yizLhe pick-back-come-LE a-M other-people throw-out DE chair'He brought back a chair that somebody threw out.'

In (8), Event A 'somebody threw out a chair' has happened before Event B 'hebrought it back'. Therefore, A naturally precedes B without any special markingdevice, as in (a), or A may be subordinated to B, as in (b). In other words, the firstevent in the event-line may either be expressed as scene-setting by clause order orby the relative clause.

Page 22: A discourse Grammar of Mandain Chinese16

Now, given an English sentence like (9.a) below, one is irresistibly tempted totranslate the relative clause into its structural counterpart in Chinese, as in (9.b).

(9.a) I checked out a book from the library which turned out to be boring.

b)?Wo cong tushuguanjiele yibenjieguo bu haokan de shu.I from library borrow-LE a-M turn-out not interesting DE book

Yet, such a direct translation is not acceptable in Chinese because the event 'thebook turned out to be boring' is not the fIrst event in the event line. In other words,the event of 'the book turning out to be boring' can not be made into backgroundbecause it neither precedes the other event nor does it form the basis for the otherevent to build on. It is completely foreground to the event of 'my checking out abook from the library.' Therefore, the Chinese counterpart of(9) can not containa relative clause for it, but must have it as an independent clause added on to thefIrst clause. This is what Chu (1983 :273) calls an elaborative clause and it is givenbelow in (9.c).

(9.c) Wo cong tushuguanjiele yiben shu, jieguo bu haokan.I from library borrow-LE a-M book, turn-out not interesting'I checked out a book from the library (and it) turned out to be boring.'

Here in (9.c), the two clauses form a coordinate structure, with the fIrst clause beingthe background to the second one. The grounding relation is merely indicated byclause order. On the other hand, one may want to reduce the weight of 'my checkingout a book from the library' by subordinating it to the other event throughrelativization. The result is in (9.d) below.

(9.d) Wo cong tushuguanjiede neiben shu jieguo bu haokan.I from library check-out-DE that-M book turn-out not interesting

'The book that I checked out from the library turned out to be boring.'

The example in (9.a) shows that in English a foreground event can be encodedin a relative clause as long as the background clause precedes it. This is notpermitted in Mandarin Chinese. The reason is quite obvious: Given that the Chineserelative clause must precede its head NP, there is no way for the entire main clauseto precede the relative clause in order to serve as background. This accounts for theunacceptability of(9.b).

In this section, we have looked at the backgrounding function of the relativeclause as a subordinate structure in Mandarin. It is found that, as a pragmatic devicefor weight-reduction, the relative clause works in conjunction with the other twopragmatic factors: event-line and scene-setting. All three must be considered in thedetermination of the grounding strength of a clause. That is, a clause may be

background in only one or two of the features or in all three of them. Thus, thegrounding status ofa clause relative to another is not a dichotomy ofbackgr?undvs. foreground. Rather, it may vary in degree. The relative clause as a subor~matestructure accounts for only one of the three features of background. In particular,we fInd that, unlike its English counterpart, the Chinese relative clause cannot codea foreground event because of its structural limitations.

Subordinate conjunctions in Chinese primarily function as a warning to thereaderlhearer that there is an order offoreground-to-background, a violation of theBFP principle. They also mark logical relations but it is their secondary func~ion.Another set of conjunctions-generally known as coordmateconjunctions-primarily mark logical relation and they only occur in the second oftwo iconically ordered clauses-Leo in temporal or logical sequence. The examplesin (10) and (11) below illustrate the behavior of subordinate conjunctions withregard to grounding. (Adapted from Li and Thompson, 1981 :644)

(10.a) (Yinwei) wo shi Meiguoren, suoyi wo qu Zhongguo xuyao huzhao.(because) I be American, so I go China need passport'Because I am an American, I need a passport to go to China.'

b) (*Suoyi) wo qu Zhongguo xuyao huzhao, yinwei wo shi Meiguoren.(so) I go China need passport, because I be American'I need a passport to go to China because I am an American.'

The clauses in (lO.a) are in a the logical order of cause and effect. They also holda scene-setting relation, going from background to foreground since the effectbuilds on the basis of the cause. The two conjunctions yinwei 'because' and suoyi'so' serve different functions. The coordinate conjunction suoyi indicates that theclause it occurs in is the effect and the preceding clause is the cause. In other words,it marks the logical relation and leaves the scene-setting relation to the BFPprinciple. Thus, the use of suoyi in (10.a) is sufficiently clear for both relations andthe other conjunction is not necessary and becomes optional. On the other hand, thesubordinate conjunctionyinwei 'because' indicates that the clause it occurs in is thecause. At the same time, its subordinating force demands that the following clausebe recognized as foreground, thus as the effect. The use ofyinwei alone in (10.a)should even more sufficiently mark both of the relations between the clause. Yet,the other conjunction suoyi is NOT optional. The reason has nothing to do withgrounding but is that Mandarin strongly demands a conjunction in the secondclause.6 Thus,

Page 23: A discourse Grammar of Mandain Chinese16

(lO.a') Yinwei wo shi Meiguoren, ?O/suoyi wo qu Zhongguo xuyao huzhao.because I be American, O/so I go China need passport

Turning to (lO.b), the reverse order of the clauses makes quite a difference in theappropriateness of the uses of the conjunctions. First of all, suoyi 'so' is marks onlythe logical relation of cause-effect and leaves the scene-setting relation to the BFPprinciple. When the BFP principle is violated by the reversed order, the scene-setting relation is also reversed. Thus, suoyi is out of place in (IO.b). On the otherhand, yinwei 'because' indicates that the clause it occurs in is the cause. By itsbackgrounding force, it also indicates that the other clause is the effect. Since in thiscase the grounding status of the 'effect'-clause is not determined by the BFPprinciple but by the subordinate structure of the 'cause'-clause, the reversed clauseorder does not disturb the interpretation of the grounding relation between theclauses.

Now let us examine another set of examples in (II).

(ll.a) Wo (yao(shl}) shi Meiguoren,jiu bu hui shuo zhemo zao deI (it) be American, then not will speak this bad DE

Yinwen Ie.English LE

b) Wojiu bu hui shuo zhemo zao de Yingwen Ie, wo yao(shi) shiI then not will speak such bad DE English LE, I if be

Meiguoren de hua.American if

In (Il.a), the two clauses are in the logical order of condition and result, with theresult building up on the condition. They are again in the scene-setting relation. Thesubordinate conjunctionyao(shi) 'if' is optional since the clause order follows theBFP principle. In (ll.b), the order is reversed where the result precedes thecondition. In order to mark that the clauses have violated the BFP principle, thesubordinate conjunctionyao(shl} 'if' is required; it has to cooccur with the final dehua as a highlighted compound conjunction. The presence of the coordinateconjunctionjiu 'then' in (Il.b) must remain mandatory becausejiu here is at thesame time an adverb. (Cf. the adverbial connective geng 'even more' in (26.1) laterin this chapter.)

Nominalized clauses in Mandarin occur primarily as subjects and objects. Whatevergrammatical function they perform, a preverbal nominalized clause is both low ininformative value and serving as background to the main clause. A postverbalnominalized clause is high in informative value despite the fact that it may containgiven information in it. Its grounding relationship with the main clause is not clearto us.

The following sentences serve to illustrate the points.

(I2.a) A: Haizimen yigegede dou yao gaozhong biye Ie womenchildren one-by-one all will high-school graduate LE we

haowu jixu, zenmo ban?not-at-all savings, how do

'The children are going to finish high school one after another. Whatare we to do with no savings at all?'

B: Haizimen shang daxue shi tamen ziji de shi.children go-to college be their own DE matter'Children going to college is their own business.'

b) Wo xiwang dajia dou neng fai.I hope everybody all can come'I hope you all can come.'

c) A: Ta zou de yiqian, weishenme bu gen ta yao qian ne?he leave DE before, why not from him ask money NE

'Then, why didn't you ask him for the money before he left?'

B: Women houlai cai faxian ta yijing zou fe.we later not-until find he already leave LE'We didn't find out until later that he had already left.'

In (I2.a), the nominalized clause haizimen shang daxue 'children going to college'is used as the subject of shi tamen ziji de shi 'is their own business' and is placedpreverbally. It defmitely is low in informative value and serves as background to thepredicate. In (I2.b), the clause dajia dou neng fai 'everybody can come' is theobject of the verb xiwang 'hope' and is placed postverbally. It is placed there toindicate its high informative value in spite of the fact that it contains some giveninformation in dajia 'everybody', which refers to the hearers and some newinformation in dou neng fai 'all can come'. In (I2.c), ta yijing zou fe 'he had

Page 24: A discourse Grammar of Mandain Chinese16

already left' is the object ofjaxian 'fmd out' and is placed postverbally. Followingfrom the preceding text, it contains nothing but given information. Yet, by itsrelative position in the sentence, it is high in its informative value.

Because the nominalized clause in (12.a) both carries a low informative valueand serves as background to the main clause, it has to stay in front of the mainclause, in this case, preverbally. On the other hand, the nominalized clause in (12.b)has a high informative value, it has to stay postverbally. Moving it to the frontwould make the main predicate sound like after-thought, as in (l2.b') below:

(12.b') Dajia dou neng lai, wo xiwang.everybody all can come, I hope'Everybody will come, I hope.'

The situation in (12.c) is a little more complex. The entire nominalized clause fayijing zou Ie 'he has already left' obviously contains given information followingfrom what Speaker A has just said but is placed here for its high informative valuein relation to the main verb. Moving it to the preverbal position, as in (12.c') below,would not change the given information it contains but then it would not be as highin its informative value.

(12.c') Ta yijing zou Ie, women houlai cai zhidao.he already leave LE, we later not-until know

'The fact that he had already left, we didn't fmd it out until later.'

In this order, the nominalized clause is no longer part of the sentence to inform. Itis low in informative value and seems to serve as background to the main clausewomen houlai cai zhidao 'we didn't find (it) out until later' . The reason for placingthe nominalized object in the preverbal position, especially in the utterance-initialposition, is the same as placing a non-fmite verbal clause in the initial position. Thatis, it is most likely for what Thompson terms 'reverting back to a previous event'.(Cf. Section 6.1.2.) We will have more detailed discussion on this issue in the nextchapter.

There are other object clauses that behave differently than the ones illustratedin (12). They are presented in (13) below.

(13.a) Women genben bu zhidao fa hui zhemo bu jiang daoli.we at-all not know he will this-way not talk reason

'We didn't know at all that he would be so unreasonable.'

b) Women genben bu zhidao fa hui bu hui zhemo bu jiang daoli.we at-all not know he will not will this-way not talk reason'We simply don't/didn't know ifhe will/would be so unreasonable.'

c) Ta hui (bu hui) zhemo bu jiang daoli, women genben bu zhidao.he will (not will) this-way not talk reason, we at-all not know

In (13.a), the nominalized clause fa hui zhemo bu jiang daoli 'he would be sounreasonable' is the object of the verb zhidao 'know' and is placed postverbally. Bythe presence of the adverb zhemo in the clause, we can infer that it carries giveninformation, but by the position it occupies in the sentence, we can tell that it is highin informative value. Because of the given information that it carries, it isinterpreted as an event that already happened and the main verb zhidao 'know' isinterpreted as in the past. Thus, the English translation. Comparing (13.b) with(I3.a), one can easily find out that, because of the question form hui bu hui 'will orwill not' in (b), the whole nominalized clause is interpreted as carrying newinformation. It even affects the interpretation of the main verb zhidao 'know'-i.e.it can be regarded as in the past or the present. Sentence (13.c) shows that thenominalized clauses in (a) and (b) can both be moved to the utterance-initialposition even though they grammatically function as objects. Once they are placedutterance-initially, they can only be interpreted as low in informative value and theycan only serve as background to the following main clause. The purpose of placingthem in utterance-initial position is, just as in (12.c'), to revert back to someprevious event or statement.

In Chapter 2, we observed that verb affixes serve the discourse function of rankingverbs in their degree of grounding. Altogether, five affixes were discussed. Whilethe ones in the middle range (i.e. the inchoative -qiali and the experiential -guo) arenot very clear-cut as to their capacity of ranking the verbs, the other three (i.e. theperfective -Ie, the progressive zai- and the durative -zhe) are quite clear in theirgrounding function. The perfective -Ie has the dual function of marking bothbackground and foreground. When used to mark anteriority, it explicitly marksbackground. Otherwise, it marks foreground. The other two mark background only.Thus, a verb marked with either the progressive zai- or the durative -zhe can beconsidered a 'non-fmite' form subordinated to another verb that is more 'fmite'. Itis in this sense that we will discuss the 'non-fmite' verb forms.

In terms of discourse, the non-fmite verb forms in Mandarin are used mainlyto mark a clause as background to another clause. The events in the clauses usuallydevelop in the direction of background- to-foreground progression. Occasionally,they may be organized in the reverse direction, background being expressed solelyby the verbal form. The sentences below illustrate the grounding function of thedurative -zhe.

(14) Ta banzhe lian, juezhe zui, dizhe tou jinleshe stiff-DUR face, pout-DUR mouth, hang-DUR head enter-PFV

Page 25: A discourse Grammar of Mandain Chinese16

zhengwu. (Fang, 1992:456)main-hall

(15) Ta zuo zai yige xiao zhuo pangbian, zhuoshang puzhe shenhe sit at a-M small table side, table-top cove-DUR dark

luse de rongtan, fangzhe yige hen guya degreen DE velvet-cloth, place-DUR a-M very ancient-elegant DE

huaping, ping zhong chazhe yizhi qiu hua. (op. cit.)vase, vase middle insert-DUR a-M autumn flower

'He is sitting at a small table covered with a dark green velvet table-cloth,on which there is an elegant vase with some autumn flowers in it.'

Sentence (14) contains four verbs: ban 'to be stiff (in the face)' ,jue 'to pout', di 'tohang (low)' andjin 'to enter'. The first three are all marked with the durative -zheas background to the last one, which is marked by the perfective -Ie. Thebackground here cannot be interpreted as a basis on which to build a further event.With their reduced weight, they are to be interpreted as subsidiary to the main eventof her entering the main hall. As the grounding relationship does not involve anevent line, the BFG principle doesn't have to be followed. Therefore, the relativeorder of these four clauses is of no consequence. They can almost be randomlyscrambled and still make good sense. Sentence (15) also contains four verbs: zuo'to sit', pu 'to cover' ,fang 'to place' and cha 'to insert, put inside'. The last threeare marked by the durative -zhe as background to the first one, which is not markedby any affix. As their grounding relationship does not involve an event line, therelative ordering of the clause should be of no consequences, either. Yet, none ofthe clauses can be moved to any other position for another reason. That is, the topic(in the form of a preverbal phrase) in each of the clauses depends on a precedingclause. The phrase zhuoshang 'table-top' in the second clause cannot be a topicwithout xiao zhuo 'small table' having been introduced in the first clause. The thirdclause shares zhuoshang as its topic with the preceding clause. The phrasepingzhong in the fourth clause cannot be a topic without huaping having beenintroduced in the preceding clause. In this sense, each clause provides a basis forthe following to build on and each serves as background to its following clause.

The examples have illustrated the relative grounding relation between -zhe and-Ie on the one hand and between -zhe and an unmarked verb on the other. It is clearthat -zhe is a background marker. The grounding relation, however, is not atemporal one. It is rather one of downgrading the importance of the event and tomake the other stand out more prominently in the narrative. (Cf. Chu, 1987a.)

(16) lintian xiawu women zai kan dianshi (de shihou), hurantoday afternoon we PROG see TV (DE time), suddenly

tingdian Ie.stop-power PFV

'(When) we were watching TV this afternoon, the power suddenly wentout.'

(17 .a) Dang you j ishiwan dangyuan,party has hundreds-of-thousands members

b) tamen zai lingdao renmin,they PROG lead people

c) xiangzhe direnface-DUR enemy

d) zuo jiankuzhuojue de douzheng. (Beida, 1986:529)do difficult-extremely DE fight

'The Party has hundreds of thousands of members who are leading thepeople to face the enemy in an extremely difficult war.'

In (16), the first verb kan 'watch' is marked by the progressive zai- and thus is madebackground to the other verb tingdian 'stop electricity', which is marked by theperfective -Ie and is interpreted as the main event. The grounding relationship hereis twofold. One is the scene-setting; the other, weight-reduction. Therefore, theorder of the clauses cannot be altered. The expression de shihou 'when' is opticalas the verb is already explicitly marked as subordinate. If there were no zai- beforethe verb kan, then de shihou would be necessary. In the latter case, anothersubordinate structure is in use for backgrounding-the relative clause.

The example in (17) is dissected into four numbered clauses for ease ofdiscussion. Clause (a) forms a discourse unit by itself; the other three form anotherunit as they share the same topic tam en 'they' in (b). Relative to the other unit, (a)is the background merely by its position. Within the three-clause unit, (b) and (c)serve as background to (d) by both their relative order and their verbal marking.The verb lingdao 'to lead' in (b) is marked by the progressive zai- and the verbxiang 'to face' in (c) is marked by the durative ~zhe while the verb zuo 'to do' in (d)is not marked by any affix. The clause order follows a logical sequence of doing the

Page 26: A discourse Grammar of Mandain Chinese16

three things: leading the people, facing the enemy, and fighting a fight. Thissequence is also reflected in the English translation. Even though there is anindication of the relative weight of back grounding between zai- and -zhe in Chapter2, the weight is not clear in this example.

The sentences in (16) and (17) show the grounding function of the progressiveaspect marker zai- in relation to the perfective aspect marker -Ie and the unmarkedverb. Zai- serves to mark background relative to the other two verbal forms. If thereis an event line (Le. temporal sequence) involved, the clauses must reflect this order.

There are some structures that come somewhere between subordination andcompound word on the one hand and between subordination and coordination onthe other. In the former category are constructions like serial verbs, e.g.

(18) Wo xihuan chi Zhongguo cai.I like eat Chinese food'I like to eat Chinese food.'

where xihuan 'to like' and chi 'to eat' si. lply occur one after the other without anyformal marking. The order of the two verbs seems to follow BFP. In the lattercategory, are constructions like (19) below.

(19) Ta you yige meimei, hen xihuan kan dianying.he has a-M younger-sister, very like see movie

'He has a younger sister who likes to see movies.'(Li and Thompson, 1981 :617)

where two clauses are used one after the other, with a shared NP meimei 'youngersister'. The order of the two clauses also seems to follow BFP. Another type in thiscategory is known as the pivotal construction (Chao, 1968:124-129) where an NPis at the same time the object of the first verb and the subject of the second verb, asin (20).

(20) Wo quan ta nian yi. (Li & Thompson, 1981:607)I advise him/her study medicine'I advised him to study medicine.'

In (20), the pronoun fa is the object of the first verb quan 'to advise' and the subjectofthe second verb nian 'to study'. In this example, the two verbs represent a seriesof events in the order they may occur, if the second occurs at all. Thus, the firstevent is the background to the second one. But, in other cases, the clause order maynot represent the actual temporal order, e.g.

(21) Gongxi ni de jiang Ie.congratulate you receive award PFV'(i) congratulate you on receiving the award.'

In (21) gongxi 'to congratulate' and de jiang 'to receive (an) award' are the twoevents involved. Their order in the sentence does not reflect the actual temporalorder of occurrence. The final Ie may be better interpreted as marking anteriorityand thus making 'receiving an award' the background. Li and Thompson(1981:608-610) distinguish between the two types of pivotal construction by thecategory of the first verb. If the verb is a factive one (like gongxi 'to congratulate'in (21)), the event represented by the second verb is 'realized', i.e. it has happenedbefore the time of the first verb. The order of the verbs is the reverse of thetemporal order of the events. If the first verb is a non-factive one (like quan 'toadvise' in (20)), the event represented by the second verb is 'unrealized', i.e. itmight happen. The order of the verbs is the same as the temporal order of the eventsif the second one occurs at all.

This chapter started with a general discussion on the relationship betweensubordination and information status. It was found that there is some closecorrelation between them but since they operate at different structural levels,background does not necessarily entail given information or vice versa. Thenbackground was carefully examined in terms of its relationship with subordination.Background was found to consist of three pragmatic components: (i) event-line, (ii)scene-setting, and (iii) weight-reduction. They interact each other. On the otherhand, subordination is found to be a general device for marking background. Thesubordinate structures, however, differ in the details of their background-markingfunctions. Among the subordinate structures discussed, the relative clause and thenon-fmite verb forms are used for weight-reduction only. Subordinate conjunctions(e.g. yinwei 'because') are used to explicitly indicate backgrounding, especiallywhen there is a violation of the BFP principle. Nominalized sentential subjects area device for background, but sentential objects may vary in their grounding status,depending on the nature of the main predicate verb.

In the following, we will examine longer passages to see the interactions ofsome of the subordinate devices working together in the organization of a discourseunit. The first passage is taken from Ke and Xu (1990: 12).

(22.a) Neishan Laoban dagai KANCHU dian shemo miaotou,Neishan Owner probably see-out a-little some clue,

b) jiu XIAOzhe HUITOU dui limian SHUOle yiju Ribenhua,then smile-DUR turn-head toward inside say-PFV a-M Japanese,

Page 27: A discourse Grammar of Mandain Chinese16

c) yuanxian he Neishan SHUOHUA de neige laorenoriginally with Neishan talk DE that old-man

YAOzhe yanzui ZOUle CHULAI.hold-in-mouth-DUR pipe walk-PFV out-come

'The owner Neishan, probably seeing some clue, turned around with asmile and, toward the inside, said something in Japanese. The old manwho had been talking with Neishan walked out, holding a pipe in hismouth.'

d) Tade miankong huangli dai bai, shoude jiao ren danxin,his face yellow-in with white, thin-DE make people worry,

e) danshijingshen hen hao, meiyou yidian tuitang de yangzi.. ..but spirit very good, not-have a-little despirited DE look ....

'His face was ofa whitish yellow color, (he was) so thin that you wouldnaturally worry (about him), but he looked full of energy.'

t) "Ni yao mai zheben shuT' ta KANle wo yiyan."you want buy this-M book?" he look-PFV me a-look"'You want to buy this book?" he took a look at me.'

g) Nazhong zhengzhi cixiang de yanguang, SHI wo likethat-kind upright kind DE glance, make me at-once

GANDAO shenshang SHOU/e fuqin de fumo-feel-reach body-on receive-PFV father DE caress-

h) yansu he cixianjiaozhizhe de fumo side.solemn and kind interweave-DUR DE caress as-if

'The upright and kind glance made me feel as if (I were) receiving afatherly caress on my body-a caress with a mixture of solemnity andkindness.'

The fIrst-person participant in the discourse is a poor student trying to buy a bookand fmding that he can not afford it. But an old man is to sell it to him for less thanhalf its price. This passage is a narration of the participant's fIrst encounter with theoldman.

We have divided the passage into four sections, each being considered a'sentence' and punctuated with a full-stop in the original text. The fIrst section

contains (aHc), which is in narrative style. The second contains (d) and (e), whichis a description. The third section contains (t) only, which is another narrative. Thelast section contains (g) and (h), which is a mixture of narration and description.The event-line verbs are in capitals and the non-event-line verbs are in italics.

The initial narrative, (aHc), is further divided by two topics: Neishan Laoban'Owner Neishan' and neige laoren 'that old man'. Within the fIrst topic, there arefour verbs: kanchu 'to fInd out', xiao 'to smile', huitou 'to turn (the) head' and shuo'to say, speak'. Xiao 'to smile' is affixed with the durative aspect -zhe. It is thusmarked as less important than the other events and interpreted as a manneradverbial. The next two verbs kanchu 'to fInd out' and huitou 'to turn (the) head'are without any affix and the last one shuo 'to say, speak' is affixed with theperfective -Ie. According to Chang (1986: 105-11 0), only the last of a series of sub-events is marked with the perfective -Ie to indicate that they form an integral largerepisode. The marked one is known as the 'peak' event. Here we see three sub-events 'seeing some clue', 'turning his head' and 'saying something in Japanese' ina series, but only the last sub-event of 'speaking' is marked with the perfective -Ie.This is then exactly the way Chang has claimed that sub-events are organized intolarger episodes, with the culminating peak event occurring at the end. The otherportion of the initial narrative is headed by the topic neige laoren 'that old man' in(c). There are three verbs in this portion: shuohua 'to talk', yao 'to hold in themouth' and zou ...chulai 'to walk out'. The fIrst verb is in a relative clause yuanxianhe Neishan SHUOHUA de 'who had talkedlbeen talking to Neishan' and is thusmarked as a background event that happened before. The second verb yao 'to holdin the mouth' is suffixed with -zhe to show that it is treated not as important as aforeground event but as a manner accompanying another verb. The last verbzou ...chulai 'to walk out' is suffixed with -Ie to signal it as the peak event.

Summarizing, the grounding structure of(aHc) in (22) can be diagrammed asfollows:

Topic!: Neishan Laoban (KANCHU, [manner xiaozhe] HUITOU,SHUOle)

The parentheses indicate the domain of the topic. The square brackets enclosebackground materials. The verbs in uppercase represent events that remain in thestory-line after the discourse structuring has assigned background status to the otherevents. The subscripts describe the subordinate structures: manner = manneradverbial; reI = relative clause.

Page 28: A discourse Grammar of Mandain Chinese16

The descriptive portion in (22), (d}-(e), contains stative verbs only. Though theclause jiao ren danxin 'to make people worry' in (e) may be less stative than theother clauses, yet it represents a result of the preceding verb shou 'to be thin' andcan be regarded as a digression from the description. Being descriptive in nature,the clauses do not bear any temporal relationship to each other. But there is anotherorder that the description follows: from the most obvious face to the entire body andthen to the less obvious spirit of the old man. In some sense, the clauses do followan order of some sort.

The portion in (f) is obviously a story-line event, as the verb kan 'look' isaffixed with -Ie. The direct quote can simply be treated as the direct object of shuo'to say', which could occur at the very end.

The final section in (g}-(h) is a mixture of narration and description. Portion(g) contains three verbs: shi 'to make', gandao 'to feel' and shou 'to receive'. Theyare in a serial-verb construction. Furthermore, while both gandao and shou are non-state verbs and are eligible for perfective aspect marking, only the last one ismarked to indicate that it is the peak of a series of sub-events. Portion (h) containsthe verb jiaozhi 'interweave', which is suffixed with -zhe to make it less prominent.As a matter of fact, the suffix is not necessary because the verb is already in asubordinate structure marked by the relative pronoun de. But, the writer may simplybe using the suffix to make it doubly sure that the subordination is unmistakablyrecognized.7 The diagram in (25) below shows the grounding structure of(g}-(h)in (22).

Topic: Yanguang (SHI, GANDAO, SHOUle fumo<[rei jiaozhizhe de] fumo»

The pair of the angled brackets indicate an apposition structure to the objectfumo'caress'.

The analysis of the passage in (22) reveals that the subordinationdevices-relative clause, fmite vs. non-finite verbs and clause order-work togetherto organize discourse in a hierarchical grounding structure. To cover the other twosubordination devices that we have discussed earlier (conjunction andnominalization), we quote another passage from Liang (1989:40).

(26.a) Women Zhongguoren laozao jiu renshi Kangnaixin niunaishui,we Chinese old-early soon know Carnation milk-water,

b) haoxiang yiban ren cheng zhi wei sanhuapai naishui,seen general person call it be three-flower-brand milk-water,

c) yinwei guantou biaoqianshang huazhe sanduo hua,because can label-on paint-DUR 3-M flower

d) er neizhong hua de mingzi bu shi women yibanbut that-kind flower DE name not be we general

ren suo xizhi de.people SUO familiar DE

'We Chinese have known Carnation Condensed Milk for a long time; itseems that it was generally called Three-Flower Condensed Milk,because there were three flowers printed on the can label but the nameof the flowers was not familiar to most people.'

e) finci wo daole zhejia niunai gongsi qu canguan,for-this I arrive-PFV this-M milk company go visit,

f) bei jue qinqie,double feel friendly

g) haoxiang shi wuyizhong zoudaole yige shou pengyoudeseem be accidentally walk-arrive-PFV a-M familiar friend's

laojia.old-home

'For this reason, when I went to visit this milk factory, I felt especiallywelcomed; it seemed like I accidentally walked into an old friend'shome.'

h) Yige gongsi hanghao fei wanbudeyi buhui guachu 'Xiejuea-M company firm not cannot-help not-will hang-out 'Please-no

canguan' de paizi,visit' DE sign,

i) geng buhui haobukeqide gaobai 'Xianren mian jin,'even-more not-will rudely announce 'Loiterer don't enter.'

j) zhaodai canguan zhengshi jigaomingde guanggao shouduan.entertain visit just-be very-smart-DE advertise means

Page 29: A discourse Grammar of Mandain Chinese16

compelling reasons, nor would (any company) rudely announce "NoLoiterers"; to welcome visitors is just a very smart way ofadvertisement. '

k) Kangnaixin gongsi menkou jiu shule paishi, zhidianCarnation company entrance just stand-PFV sign, direct

canguanren ying caiqxu de luxian, ...visitor should take DE route, ...

'The Carnation Company has a sign standing at its entrance, (which)gives visitors directions, ... '

The passage is also divided into four parts: (a}-(d), (e}-(g), (h}-(i) and U}-(k). Thedivision coincides with the 'sentences' indicated by full-stops in the original.

What we are interested in here are the conjunctions yinwei 'because' in (c) andyinci 'for this (reason)' in (e), the adverb geng 'even more' in (i), and the sententialsubject zhaodai canguan 'to welcome visit(or)s' in 0).

The conjunction yinwei is a subordinate one. It is optionally used to indicatethe background status of the clause it is in when the clause order is BFP. But if theorder is reverse, the conjunction is obligatory. Here, it is the latter case thatnecessitates the presence of the conjunction. If the order of clauses (b) and (c) werereversed, then this subordinate conjunction yinwei would become optional, but therewould be a need for a coordinate conjunction suoyi 'therefore' to go with thesecond clause for the logical relatiem between them. (Cf. Section 6.2.2.)

The coordinate conjunctionyinci 'for this (reason)' in (e) actually connects thewhole section made up of(e}-(g) to the preceding section consisting of(a}-(d). Therelationship between these two sections is one of cause-effect and the sections areordered in BFP. No special subordinate conjunction is needed.

The adverb geng in (i) is one of those that work both as an adverb and aconnective. It connects two parallel events, with the second being a step further thanthe fIrst. The clause order of (h) and (i) is one ofBFP. Therefore, only a coordinateconjunction for the logical relation is called for. The clause order can not bereversed in this case, because there doesn't exist a subordinate conjunction for thislogical relationship.

The sentential subject zhaodai canguan 'to welcome visit(or)s' refers back towhat has been mentioned in the same section and in the previous section. It carriesgiven information, is low in informative value and serves as background to theclause with the predicate 'is a smart way of advertisement' .

One additional interesting construction is the relative structure ying caiqu deluxian 'route that (one) should take'. The relative clause, as a rule, serves as thebackground to the main clause. But, since the head NP is postverbal, the entirerelative structure is high in informative value. It is oflittle consequence whether the

head NP carries given or new information. Both the grounding relation ~etween therelative and main clauses and the informative value of the head NP rem~m the same.In this particular case, the head NP luxian 'route' seems to be pragmatIcally ratherthan grammatically defInite because when a sign gives directions to visitors, itwould usually suggest only one route as the best and thus the route is defmite by itsuniqueness. This kind of defmiteness doesn't necessarily carry given informationand it doesn't allow defInite marking by morphology in Mandarin.

1. The dependence relationship that Thompson talks about here is one wherethe subordinate structure is dependent on the main clause. It is structural in natureand is very different from a pragmatic relationship which we will later propose as'scene-setting' in Section 6.1.2.

2. Clause (c) is still a problem, but see Section 6.2.3.3. Reverting back will be further discussed in the next chapter.4. The adjustments in wording do not directly bear on our discussion. This also

applies to the variations in the following examples.5. Nominalization, expressing a proposition in the form ofa nominal, is also

a subordinating device.6. For details, see Chu (1991a), Uu and Chu (1993) and Uu (1996).7. Another reason for the use of -zhe may have to do with a different perception

of the structure by native speakers. The de, though interpreted as a relative pronounby the present author, may be more comfortably recognized as an adjectival suffixby a non-linguist. If so, the verb jiaozhi 'to interweave' would not be appropriateunless it is turned into a state verb by adding the suffix -zhe to it. Cf., for example,Icaizhe de men vs. ?kaide men 'the open door/the door that is open'.