pulleyblank-classical chinese grammar review

Derek Herforth

A propos de... PulleyblanK Edwin G. : Outline of classicalChinese grammarIn: Cahiers de linguistique - Asie orientale, vol. 30 n°2, 2001. pp. 215-256.

Citer ce document / Cite this document :

Herforth Derek. A propos de.. PulleyblanK Edwin G. : Outline of classical Chinese grammar. In: Cahiers de linguistique - Asieorientale, vol. 30 n°2, 2001. pp. 215-256.

http://www.persee.fr/web/revues/home/prescript/article/clao_0153-3320_2001_num_30_2_1584

http://www.persee.fr/web/revues/home/prescript/author/auteur_clao_283

http://www.persee.fr/web/revues/home/prescript/article/clao_0153-3320_2001_num_30_2_1584

A propos de...

PULLEYBLANK Edwin G. (1995). Outline of Classical Chinese Grammar. Vancouver : University of British Columbia Press.

0. INTRODUCTION

Professor Pulleyblank has produced what is no doubt the most clearly reasoned general account in any language of the structure of Late Zhou Chinese (LZC). Among well-known earlier studies of the language, Chou Fa-kao (1959-62) may be more complete in his coverage of previous scholarship, both Chinese and western, while Yang and He (1992) range more widely in terms of both topics covered and chronological span. Harbsmeier (1981) is also quite insightful on a number of important areas, but by his own admission did not aim in his dissertation to write a new grammar of the language. In the 160-odd pages of this Outline, however, Prof. Pulleyblank manages to treat most if not quite all of the major constructions of LZC in a way that can be recommended to any anglophone student familiar with traditional linguistic terminology and analysis. It is indicative of the depth of Prof. Pulleyblank's scholarship that, though he refers to the Outline as an "introduction", "pedagogical" in aim (p. xiv), there is much here for seasoned students of the language to reflect on as well.

The present reviewer will happily admit to being a long- term user of the Outline in its various prepublication versions. I had the good fortune of spending several years in the mid- to late seventies at Prof. Pulleyblank's home department where rigorous seminars in classical texts, Chinese historical linguistics and paleo-

Cahiers de Linguistique - Asie Orientale 30(2): 215-256 (2001) © CRLAO-EHESS 54, Bd Raspail 75006 Paris 0153-3320/2001/030-215

216

Herforth D. / Cahiers de Linguistique - Asie Orientale 30(2001) 215-256

graphy were the staple fare. Various typed and eventually word- processed editions of the "Outline", kindly provided by the author over the years, have been my dog-eared companions ever since.

In some cultures it would no doubt be considered presumptuous for a former student to review the work of a mentor, but I believe it safe to assume that Prof. Pulleyblank is at least as interested in cogent responses to his work, regardless of their source, as in the deference clearly due his long and distinguished scholarly and teaching career. One of the virtues of his book is that in the many places where original analyses are introduced, the presentation is admirably clear, allowing advanced readers to see how his new proposals open up further questions and suggest hypotheses that might be tested in future studies. As Prof. Pulleyblank admits in his preface, we are "still at the stage of struggling to work out the basic patterns of Classical Chinese syntax" (pp. xiii-xiv). While the Outline is in many respects state- of-the-art, it is at the same time a call for all of us amateurs of LZC to engage in that struggle together. Indeed, some of the struggle may well have found its way into the present review - a fact that should be seen as reflecting the richness and importance of Prof. Pulleyblank's work.

1. ORGANIZATION AND TERMINOLOGY

The book is laid out in chapter and section divisions drawn in terms of several different types of distinctions: phrase-, clause- and sentence-type ("noun phrases and nominalization", "noun predication", "pronouns", "complex sentences"), semantics ("numerical expressions", "adnominal and adverbial words of inclusion and restriction") and clause-level operations ("topicalization and exposure", "negation"). Since all grammars "leak", the task of deciding precisely how to organize the grammatical phenomena of a language into a series of coherent

217

A propos de... /CLAO 30(2001) 2 1 5-256

expository chunks is never an easy one; any single set of decisions is likely to evoke some demurrals from fellow analysts.

Thus, from a strictly synchronie view of LZC clause structure, "coverbs" (pp. 47-57) could well be treated as "adpositions"

introducing either adjuncts (e.g. most jft -phrases) or arguments (e.g. VÀ -phrases with many verbs of transfer) of the verb.1 The fact that coverbs tend to derive from full verbs (and are occasionally attested as such in LZC) seems to me insufficient motivation for treating the vast majority of LZC clauses in which they occur as "compound verbal predicates", title of the chapter in which coverb phrases are discussed.

More surprisingly, the section on coverbs includes an analysis of "descriptive complements with ru #P and you Щ" (p. 57), though neither of these morphemes can reasonably be regarded as a coverb. In fact, Prof. Pulleyblank himself presents one rather persuasive argument for claiming that there is nothing at all (co)verbal about Щ (p. 18).

A related issue concerns the use of the label "complement". "Descriptive complements with #И and Щ"аге certainly very different both syntactically and semantically from so-called "locative complements" introduced by the "coverbs" (?) yu Ť, yu Ш and hu -*ř(pp. 53-55). This latter phrase-type functions as either argument or adjunct of the verb, not as a "complement" in the traditional sense: "a non-argument, non-adjunct category occurring in a predicate and interpreted as describing or referring to another constituent in the sentence" (adapted from Trask 1993, p. 51, emphasis added; this use of 'complement' is still useful as it defines one type of phrase interrogated in situ in LZC). In terms of

syntactic processes, locatives behave like direct objects in that they "front" to preverbal positions for focusing (as in example (15)

below), question formation (Ъууап/ап/ап М/^/Ш or wu/wu/wuhu

The term 'adposition' is motivated by the fact that most "coverbs" are attested both pre- and post-positionally, with postpositional use restricted to preverbal sites in the clause. See below, esp. note 7.

218

Herforth D. /Cahiers de Linguistique -Asie Orientale 30(2001) 215-256

Ш/ Ml И-?", pp. 96-97) and the expression of wide scope. Thus, locatives and objects might best be treated as two sub-types of the category "oblique phrase" whose default position in the clause is postverbal, where "default" refers to the absence of semantic

operations such as interrogation, narrow focus, etc. In contrast to these obliques, phrases introduced by #P and Щ are themselves predicates, not mere arguments or adjuncts; furthermore, they are never found in any preverbal position. Towards the end of this review, we will have reason to return to these issues to suggest that there is room in a description of the LZC clause for a phrasal

category "secondary predicate", subsuming "descriptive complements" in #P and Щ, together with a number of other phrase-types

Prof. Pulleyblank treats separately. (See § 5. below.) While advanced readers of the Outline may thus find they

have some reservations about precisely how the map of the grammar has been drawn, no one, student or expert, will experience difficulty in finding her or his way around the territory. The table of contents is admirably detailed, including characters, and there are two indexes, one of LZC (mostly functional) morphemes and the other of proper names and English grammatical terms.

2. SYNTAX AND VERB CLASSES

The book is particularly illuminating on the internal structure of LZC phrases, showing very clearly how regular interpretive differences often reflected in English by morphologically marked distinctions in part-of-speech, voice, etc. correlate with strictly syntactic or [lexical+syntactic] features of ОС. Thus, the verb types intransitive, transitive, ditransitive and adjectival are laid out on the basis of facts like their interaction with the auxiliaries ke Ц, keyi Ц VX and keyiwei Щ |Д fa (pp. 26-30).

219

A propos de... /CLAO 30(2001) 215-256

(la)XoJ +V — » (passive ability/permission) 'X can be V-ed,

(lb) Y "5J Ц + V -» (active ability/permission) "Y can V,

( le) Z йЩ + Ц + NP/ADJ -> (categorizing) Z can be (a) NP/ADJ.

Prof. Pulleyblank proceeds to explain the presence of the instrumental VX in [оТУХ + VTRANS] as a grammaticized marker of

agency. As pattern (la) implies, the adjective oj passivizes the following transitive verb. In the [Y Щ\Х + V^^s] construction,

that verb is VÀ, 'take; use1, so that, morpheme-by-morpheme, xvang keyi sha Ï oj [Д |g means "the king can be used [Ï ЦТ 1Д] to kill > the king may be the instrument/agent in killing > the king can kill" (p. 24; also p. 43). This works nicely for transitive verbs with agentive subjects, but does not explain the use of ôj \X with (in)transitives and linking verbs like M whose subjects are not agents or instruments, as in:

(Mel4.2B.2.1) gua ren bu ke yi feng bereft man NA AA suffer. wind 'I cannot bear (exposure to) the wind.'

(3) (Me32.4B.23.1) ke yi si AA die '(When one) can die ...'

(4) loIVÀmm& (Me2.1A.3.1) shu kou zhi jia keyi wu ji yi number mouth GN family AA NE starve PL 'A household of several individuals can be without hunger.1

220

Herforth D. /Cahiers de Linguistique - Asie Orientale 30(2001) 215-256

(28) ЛШ-йПД^ИИ(Ме46.6В.2.1)2

ren jie keyi wei Yao Shun 'Men can all be a Yao or a Shun.1

(33) Щ\>Л%Ш¥ (Ме44.6А.8.1) keyi wei mei hu 'Could [Ox Mountain] be (i.e. remain) beautiful ...?'

While Prof. Pulleyblank's analysis of VX in цГ \>X may not be synchronically valid for all types of verbs which follow oj 1Д in

LZC, one could ask if it might be defensible historically by going to the corpus in search of support for the hypothesis that oj \X started out marking agentive ability, i.e. was first restricted to prototypical transitive verbs, and later generalized to all non-patient subjects by spreading to non-agentive transitives, intransitives and adjectival predicators.

The use of [й]"ЦА + ̂ ] with adjectivals like тегШ, as in

(33), suggests a second question. The copular use of M to link subject NPs to subject complements is well attested, so that чГ \X М NP, 'can be(come) NP\ as in (28), can be treated as a subtype of [чГ 1Д + VARANS]- However, the use of M to introduce predicate

adjectives is much less common, at least in the late Zhou, and when it occurs, "give[s] a superlative sense" to the adjective (p. 25). As there is no obvious contextual support for construing M Ш in (33) as 'the most / the especially beautiful1, it might be worth considering an analysis of Мм Shan ke yi wei mei /ш(ФШ) Ц [\ХМШ] ^řas

consisting of transitive predicator [VX x M y] 'take x to be y' passivized by цГ: 'could (Ox Mount, still) be considered beautiful?'

Examples cited from the Outline are identified by their original numbering in bold and are followed by Prof. Pulleyblank's translations and the reviewer's own text references. Other examples are numbered consecutively in plain type and include morpheme glosses. Details of textual sources which differ from those in the "Sources of Examples" (Outline, pp. 169-70) will be found together with a list of abbreviations in the bibliography below.

221

A propos de.../CLAO 30(2001) 215-256

Careful study of the evolution of the uses of M and the syntax of VA M would be needed to assess the validity of this idea.

These distinctions in verb-type are put to further analytical use in disambiguating polysemous yu Ш. Many of the commonest meanings (in English) of Ifc can be predicted on the basis of the type of verb with which the adposition occurs. It is these verb types which English and LZC share, allowing accurate translation between the two languages. Thus, "[l]ocative phrases introduced by Ш and hu ^ř are used to express comparison after adjectives ... and agency after passive verbs ..." (p. 55). In other words, Ш etc. is rendered by 'than' when it follows an adjective and by agentive 'by' when it follows a verb which has transitive meaning but lacks a surface object. Consider the following default patterns, the empty object position "held" by 0.

(6) [NP1 Adj #*NP2] = 'NPl BE ADJ-er than NP21

(7) [NP3 V^^ 0 Ш NP4] = 'NP3 BE V-ed by NP4'

A few refinements can be added. Some adjectives, such as zhuo Jih 'be. inept', take relational adjuncts introduced by Ш.

(8) : (Zh 2.1.37) fuzi gu zhuo yu yong da yi Master firm inept LP use big PL 'You, Sir, certainly are inept at using large (things).'

These could be put aside as an exceptional subclass of adjective. But surely a more general solution for the comparative construction as a whole would be to add a semantic precondition to rule (6) stipulating that the referents of NPl and NP2 be comparable, i.e. of the same general category. Clearly, in (8), the human referent of NPl, ^Ť 'you, Sir' is not of the same category as the abstract referent of the nominalized predicate NP2, Ш ýí 'using the large'. A

222


sentence along the lines of xiao ren gu zhuo yu fu zi

/Ь À HI #Ь Ш ̂ к ~F t!l would, by semantic rules, be interpreted as comparative: 'I am surely more inept than you, Sir (at some activity specified in the co(n)text)'.

Rule (7) is not quite precise enough either. Both (9) and (10) below are "unmarked passives" by virtue of the conspicuous absence of a direct object after the highly transitive shang Ш 'wound'.

(9) ШЙ1|0|^ (Zz 213 Ch 2.4.8) Xi Ke shang yu shi PN wound LP arrow 'Xike was wounded by an arrow.'

(10) ШШШ0ШШ (Zz 131 Xi 28.5.5) Wei Chou shang yu xiong PN wound LP chest 'Wei Chou was wounded in the chest.'

The rule states that after such core clauses Ш will introduce the oblique agent, as it does in (9). However, (10) shows that even in such syntactic contexts, the adposition may retain its locative meaning when its object is a locus NP like 'chest' rather than a (metaphorically) agentive NP like 'arrow'. Note that we are not relying on "co-text" for disambiguation here, but on our world

knowledge of the referent of *Л as an instrument (of potential harm) and of Ш as a location.

These problems hardly exhaust the intricacies of Ш and its interpretation. In another highly testable claim, the adposition is described as optional after verbs of motion and location such as ji

Ж 'reach' da iÉ 'extend to' and ум Щ 'dwell' (p. 29). In designing an

empirical test of the claim that with such verbs the presence of Ш correlates with no "apparent difference in meaning", one would best

start with the pronominalized forms, [V + zhi Ž.] vs. [V + у an M].

223

A propos de... /CLAO 30(2001) 215-256

Given that a scribe is less likely to have changed M to Ž. than to have omitted Ш in а [Ш + NP] phrase (especially if unmarked postverbal locatives were common in the vernacular of his time), the pronominalized examples will probably preserve better evidence of any distinction that may once have existed. The investigator would then attempt to isolate a statistically significant number of examples of [V + M ] which correlate with some

semantic, etc. feature not present in [V +Ž.], perhaps an

aspectual distinction along the lines mooted by Prof. Pulleyblank (1991, pp. 33-34) in his elaboration of Graham's (1983) claims about the aspectual meaning of yun ^and yue EJ in the preclassical language. The results of such a study, if statistically significant, could then be applied to cases of Ж, Ш. and Щ in which the object is not pronominalized, and then to other verbs where the adposition Ш can apparently "be omitted" (pp. 33, 55-56). Even if we restrict the domain of inquiry to the Zuozhuan and Lunyu, ready examples of this sort of alternation occur, not all of them verbs of motion or location: ci Щ {Ш) 'decline1, jian Я, (Ш) 'see; have an audience with' and jia t!W (Ш) 'add, apply to'. To my knowledge, the whole question of LZC [verb zblocative] patterning has yet to be seriously addressed. Nor should such a topic be thought a fusty philological exercise; much current work of considerable theoretical interest focuses on precisely these problems — the organization of lexical information and its integration with syntax. (See recent works by Levin, Goldberg, and Pustejovsky listed in the references below).

3. EUROCENTRIC DESCRIPTION

In what is perhaps a concession to anglophone students, Prof. Pulleyblank occasionally allows himself to describe phrases in terms of their "deceptive appearances". Thus, certain adjectives and intransitives are said to take NPs "which look like objects but which

224


are semantically like oblique cases in a language like Latin or prepositional phrases in English" (p. 25; see also p. 27).

(39) Itt$Ž (Me 36.5A.5.7) bai xing an zhi 'The common people were peaceful under him.1

The characterization "look like objects but which are semantically like oblique cases" begs the crucial question for the precise description of LZC structure, namely, do these "look-like objects" exhibit object-like behavior elsewhere in the syntax? The

evidence I am aware of (on question formation, Ят-relativization,

etc.) suggests that they do. Hence, on language-internal criteria, they are perhaps not merely "like objects", they are objects, just as the single oblique argument of verbs meaning 'wait' in Chinese (si

#, hou Ш, dai Щ and later deng Щ) and Japanese (matsu) are direct objects, a categorization that has nothing to do with the fact that in English wait is followed by a prepositional phrase, wait for x (cf. await). To be sure, Prof. Pulleyblank's informal characterization in terms of Latin and English, might help students assimilate this sort of "quirky" direct object; at the same time, however, his method might create the impression that the argument structure of a LZC verb is best described by reference to its translation equivalent in well-studied European languages.

In another example of this type of insufficiently motivated analysis, the NP which precedes the existentials you ÍÍ and wu Ш is described as "a kind of pseudo-subject", as opposed to the "personal

3 This particular case might be accounted for as a lexicalization of the putative transitivization of $: 'be. secure1 > líc-Ž. 'consider it secure': "The common people considered him [their ruler] secure > felt secure under him". There are, however, other instances of "quirky" objects less amenable to this sort of explanation. (Note further that Prof. Pulleyblank's use of 'oblique' is distinct from the sense in which we use the term to mean "all canonically postverbal object and adjunct phrases". In the more restricted sense in which it is employed in the Outline, 'oblique' refers only to adjunct phrases. Both uses are attested in the literature.)

225

A propos de... /CLAO 30(2001) 215-256

subjects" preceding the homophonous and graphically identical

"ordinary transitive verbs" of (non-) possession 'have' and 'have not' (p. 30). Here the analysis appears to be based rather too closely on the fact that the cited English and French data show separate constructions for expressing possession and existence, the latter lacking true subjects. {There and il are assumed not to be the "genuine", but rather the "dummy" subjects of there is and il y a respectively.) I would suggest that, in the LZC [NPl ^f / MNP2] construction, the "domain" NPl bears a "superset" relation to NP2. This "super- to subset" relation may have either a physical, purely spatial realization (as in (64), implying "containment by/inclusion in", both schematically related to the anthropocentric notion of possession), or a whole-to-part realization (again fundamentally locative, and also easily conceivable as a type of metaphorical possession), this partitive relation emphasized by Prof. Pulleyblank (p. 31). These sub-descriptions of the possible relations between NPl and NP2 are all implicit in the semantics of LZC W / $£and determined by the different sorts of reference found for NP 1 . All of this, however, is simply compositional semantics. Labelling a phrase a "pseudo-subject", on the other hand, implies syntactic distinctions because "subject" is fundamentally a syntactic and not a semantic notion. However, no clarification is offered of precisely how such "pseudo-subjects" differ in their syntactic behavior from more "genuine" exemplars of the category 'subject' in LZC.

4. THE FINE STRUCTURE OF THE LEFT PERIPHERY

Chapter VIII, "Topicalization and exposure", treats the area of grammar now often referred to as "information structure". It is widely recognized that discourse-related phenomena such as topic and focus have major effects on intonation and/or morphological marking and/or word order in all natural languages. It is to Prof. Pulleyblank's credit that he recognizes the importance of these notions for the accurate description of LZC syntax. That said, it

226

Herforîh D. /Cahiers de Linguistique - Asie Orientale 30(2001) 215-256

must be admitted that his analysis could be clarified considerably by abandoning the ambiguous term "exposure" and drawing instead a clear distinction between topic and narrow focus. Unlike the case of "(pseudo-)subject" mentioned above, here the syntactic

differences between two constructions are laid out clearly enough (pp. 69-71), but their distinct functions are given short shrift.

Topicalized obliques (objects and adjuncts) occur in a pre- subject position and may be resumed pos/verbally by the case- marked anaphors (áland M), while focused obliques are found between the subject and verb and are resumed preverbally by case- marked expressions - shi Ik, later Ž. for direct objects, posmominal adpositions (i.e. postpositions) or [preposition+ anaphor] for adpositional phrases.4 Thus, topic (T), subject (S), focused phrase (F) and the T and F resumptives (t, f) each have a distinct position within the LZC clause, as shown in (11). While the LZC corpus appears to lack clauses in which T, S and F are all instantiated, it is a simple matter to infer their relative order from many examples in which T always precedes S (as in 251) and S always precedes F (as in 239 and 240).

(11) T S Ff VERBt

4 There are special rules for the focusing (p. 72) and topicalization (pp. 73-75) of subjects, the former exemplified in (12) below, the topicalization of instrumentais (as in gong yi zhao shi *?, \lk 0 IS i 'a bow is for summoning a knight', a construction unnoticed in the Outline) and for the "possessor extraction" exemplified in (230). It might further be noted that, while not all of these phenomena are treated in Pulleyblank (1960), "Studies in early Chinese grammar" remains, to my knowledge, the best analysis in any language of what we would now describe as "narrow focusing of the subject and direct object" in LZC. Though our understanding of discourse-based notions like focus has advanced considerably in the years since it was written, one returns repeatedly to this study for the wealth of data treated there.

227

A propos de... /CLAO 30(2001) 215-256

(11) is simply a schematized version of Prof. Pulleyblank's description. Instead of making the important functional distinction between topicalization and narrow focus the basis of his analysis, however, Prof. Pulleyblank attributes the difference between pre- and post-verbal resumption of "exposed elements" to a syntactic change claimed to have occurred between late Preclassical/early Classical Chinese and LZC (pp. 14.c, 70). There is, in fact, good reason to believe that the structure [(S) F f VERB] was by Han times no longer productive; in other words, new constructions evolved in Late Zhou-Han Chinese for the expression of narrow focus. This development on the part of one of the two constructions, topicalization and narrow focusing, is quite independent of Prof. Pulleyblank's claim that the two constructions are related diachronically, namely that early classical [(S) F f VERB] somehow "became" late Zhou [T (S) VERB t]. The clearest evidence that there is unlikely to be an evolutionary link between the two is found in clauses in the Zuozhuan in which both constructions occur. Such co-occurrence is possible because the discourse functions of topicalization and narrow focusing are distinct without being incompatible.

Both of the following examples adhere to the very rough template for topicalization and narrow focus proposed in (11): in each, the direct object is topicalized and resumed postverbally by case-marked Ž., while focused constituents, subject in (12) and "heavy" Й" -phrase in (13), are resumed preverbally and case-

marked by shi Й and yu shi hu Ш H -f respectively.

(12) EŽftSífŽ (Zz 310 Xi 26.3.4) chen zhi lu jun shi you zhi vassal GN emolument lord RSXV 30 'A vassal's emolument, (contrary to what one might think) it is the lord who owns it.1

228


shan chuan zhi shen ze shui han li yi zhi zai mt. river GN spirit RS water drought pest GN disaster yu shi hu yong zhi LP DP LP sacrifice 30 'The spirits of the mountains and rivers, (contrary to what the addressee has assumed) it is in disasters of flood, drought and pestilence that we sacrifice to them.'

(Zz 344 Zol.vii.2. 10-11)

Specialist works on information structure, such as Vallduvi (1992) and Lambrecht (1994), must be consulted for detailed accounts of the functional differences between topic and narrow focus. Suffice it to say here that a focused constituent typically contrasts with another, assumed or anticipated semantic value for the same type of constituent, a value often already established or implied in preceding discourse. Thus, (12) is asserted following a narrative in which a vassal behaves as if he had proprietary rights to

his fief: Sun Linfu yi qi ru Jin ШМ^ЛШиШ '(After assassinating his lord,) Sun Linfu proceeded to Jin, taking (the inhabitants and other movables of) Qi (his fief) with him1. In this sort of context, narrow focus on jun Ш, marked by shi Ш, is used by the speaker to correct the likely inference that a vassal has owner's rights to his fief: 'on the contrary, it is the lord who retains such rights'. Similarly, (13) is asserted to correct the assumption implicit in the preceding narrative that one sacrifices to mountain and river spirits for relief from a personal illness - 'on the contrary, such spirits are sacrificed to for relief from plagues and other natural disasters, not for relief from one's own aches and pains'.

Narrow focus is thus the sort of "emphasis" found in the

second of two propositions which differ from each other in only one respect - the semantic value of one of the constituents, direct object in (12), circumstantial adjunct in (13). The rest of the asserted

229

A propos de... /ОЛО 30(2001) 215-256

material is presupposed "background". Thus, in (12), the presupposition is 'someone owns fiefs'; the conflicting beliefs (one of them implied by Sun Linfu's behavior) concern just who that someone is. In (13), 'certain types of adverse conditions should motivate sacrifices to mountain and river spirits' is presupposed; the question is precisely what those conditions are.

This sort of single-phrase contrast with a competing claim in the co(n)text distinguishes narrow focus very clearly from the twofold contrast found in pairs of assertions with so-called contrastive topics, where both the topics and (at least part of) the comments offset each other. Compare (13), repeated below as (14a), with (14b), the sentence which follows it in the Zuozhuan.

(14a)

(i4b)

COMME

F f

штт,

N

V Z t

ri yue xing chen zhi shen ze xue shuang feng sun moon celestial bodies GN spirit RS snow frost wind yu zhi bu shi yu shi hu yong zhi rain GN NA timely LP DP LP sacrifice 3O 'The spirits of the mountains and rivers, it is in disasters of flood, drought and pestilence that we sacrifice to them.' 'The spirits of the celestial bodies, it is when freezing weather and rainstorms occur out of season that we sacrifice to them.1

(Zz344Zol.vii.2.11-12)

In (14a, 14b), each of the focused phrases contrasts with the addressee's assumption that both groups of spirits mentioned should be petitioned for relief from personal illness. The speaker contrasts the two groups of spirits, the topicalized objects of the verb 'sacrifice', in terms of the proper conditions under which each

230


should be petitioned, those conditions contrasting in turn with the conditions assumed in the co-text: '(You have assumed one should sacrifice to spirits of groups w and x under conditions c.) In fact, group w is to be appealed to under conditions y, and group x under conditions z'. A phrase in narrow focus thus contrasts once with a contextual assumption. У in (14a) and z in (14b) each "override" the с part of the contextually implied assumption. In pairs of topic- comment structures, the contrast is two-way, each of the two constituents, topic and comment, contrasting with its counterpart in the paired sentence. It is presumably this difference in "contrastive valence" that permits narrow focus to be embedded semantically within the comment (never the topic) of a topic-comment structure,

as (14a, b) show.

Recognizing a preverbal narrow focus position in the structure of the LZC clause, as sketched in (11), has implications for the description of a number of other syntactic phenomena treated in the Outline. Thus, the list of "exceptions" to basic SVO constituent

order (p. 14) could be made more precise by distinguishing topicalization and focusing as two separate syntactic processes, each with its own "landing site" left of the verb and resumptive strategy. The claim implied there about a diachronic relation between different types of "exposure" may need to be reconsidered. In view of the widely attested syntactic similarities between constituent question ("WH") words and narrow focused phrases in languages with a fixed focus position, the rule about the fronting of interrogative oblique phrases (not just "pronouns", as on pp. 14, 27) should probably be integrated with the rule about the fronting of focused oblique phrases.5

5 The point of stressing that it is whole interrogative oblique phrases, not just interrogative pronouns, which front to preverbal positions (as demonstrated by 234 and 340) is to distinguish this "WH-rule" from the rule which fronts pronouns and only pronouns under matrix negation. (Even monosyllabic status NPs, like chen Ě '(your) vassal1, used with humble 1st person singular reference, and zi T'son', used with familiar 2nd singular reference, seem never to front when the verb is

231

A propos de.../ CLAO 30(2001) 215-256

Still on the issue of focus and its treatment in the syntax of LZC, Prof. Pulleyblank notes (pp. 71, 72, 122) that the canonically prepredicate morphemes wei Щ,/е1 ^, Ы $b and jiang Ш are also found introducing "exposed" NPs, including subjects (as in 426, with ffî). We can further note that in the Zuozhuan, Щ, at least, is attested preceding preposed adpositional phrases as well, the adposition (Ap) appearing posipositionally to mark narrow focus: [Ap NPl+focus -> [NP Ap].6

negated, because they are full NPs, not genuine pronouns.) The fronting of pronouns under negation has been plausibly explained as a type of cliticization, i.e. as completely unrelated to focus, while a rule that fronts entire (contrastive!) NPs can hardly be the result of cliticization. Xu and Li (p. 89) fail to appreciate this crucial difference in terms of information structure between the targets of these two rules and so mistakenly treat pronoun fronting under negation as a type of narrow focus. (Cf. Wei p. 294, note 68.)

In many languages with "focus movement" (which we propose for LZC, following suggestions by Xu and Li, Feng, and Wei), constituent questions are formed according to the same kind of rule that governs the placement of focused phrases, preverbal position being a well-attested site for both types of constituent. (For the widely recognized congruity in terms of information structure between constituent questions and assertions containing a narrow focus, see Lambrecht (pp. 282-84)). É. Kiss (1995, p. 23) lists the following languages in which constituent question phrases "must land in the focus position": Somali, Chadic, Aghem (Caucasian), Basque, Hungarian, Haida, Omaha, Quechua, Korean and modern Greek (main clauses only). Note that this claim leaves open the question of just what happens when a constituent question word co-occurs with a focused phrase. For relevant LZC data, see example (22) below. Needless to say, the unfamiliarity of such languages to most of us, sino- and anglo-phone students of LZC alike, explains why we have failed to appreciate the close relation between the rules governing oblique question formation and ШвгыШЙи/ Ш$£ I "contrastive exposure". These rules have typically been treated as two totally independent "exceptions" to the otherwise quite rigid SVO basic word order of LZC, though the phrases targeted by both rules are, in fact, of a closely related type in terms of information structure. 6 Note that this is the same sort of "head-complement inversion" ('complement' in its more modern, general sense; see again Trask, s.v.) used to mark direct objects as focused. Cf. (a) [Ap (NP+focus)] > [NP Adposition], as exemplified in (236). (b) [V(O+focus)] >[OI/ŽV], as exemplified in (232-235).

232


(15) ЩШ Ш Ш {han, modern standard |É) (Zz 374 Zo 1 1.4.3)7 wei Cai yu han only PN LP be. upset 'It is only with (us,) Cai, (rather than any other neighboring state) that (the king of Chu) is disgruntled.1

The list of such apparently "floating" particles can probably

be extended. The prepredicate use zeng Ц 'even' is exemplified in

the Outline (422). In (16) it precedes a focused object.

(16) ŽA&m^ iÉŽ^ftó (Xz 9.4.10) shi ren ye er zeng gou zhi zhi bu ruo ye DP man PL CO even dog pig RS NA be.like PL '. . .this is to be a man, yet not as good as even a dog or a pig (Graham 1977, p. 32b; emphasis added).8

What then is the most general description of the syntax of these particles which precede both predicates and focused arguments and adjuncts, but are never found in either of the two oblique positions to the right of the verb? LZC is a language in which all postverbal obliques must move out of the VP to become focused; hence, particles like wei Щ. and ft (zeng), whose syntactic placement is sensitive to narrow focus, will simply not be found in oblique positions.

What then about the unmarked, immediately prepredicate position for such particles? Here, a second notion of focus is needed. It is generally recognized that in basic word order clauses

7 On the basis of examples like (15) and (236), we can propose a general explanation for the observation of Sun (1996, p. 19) that LZC postpositional UA- phrases only occur preverbally. At the phrase level, the conversion of any prepositional phrase, not just [ilk NP], into a postpositional phrase, [NP VX, etc.], marks the NP as focused. At the clause level, focused arguments and adjuncts must occur in focus position, between (unfocused) subject and verb. 8 [ h*+NP] is found twice in Lunyu; in both cases, the NP functions as object of the following verb/adposition.

233

A propos de... /CLAO 30(2001) 215-256

of SVO languages (i.e. in the absence of narrow focus, etc.), the predicate itself is the site of broad focus, in the sense of "rheme", "comment", or "new information".9 The claim about LZC would then be that H, nfë, etc. are always positioned so that any focused phrase, whether the entire predicate, or a fronted argument or adjunct, will occur to their right, within their scope. Thus, while the most commonly attested position of modal >& 'must; always1 is between subject and predicate because the latter instantiates unmarked, broad focus, a narrow focused subject must occur within the scope of ift, i.e. to the particle's right.

(17) 1^ФШлЕ; MZ'%& (Zz 382 Zo 13Л.-5) Mie xing you luan bi ji shi li Chu zhi chang ye PN clan XV chaos MN cadet RS stand PN GN regular PL 'When the Mie clan has a succession dispute, (contrary to our ways in the north) it is always the cadet who accedes (to power); that is the regular practice in Chu.1

In the translation of (17) we have replicated the scope-sensitive syntax of LZC № by using the 'it-cleft' which allows 'always' to be positioned before the subject. (Cf. *'Always the cadet accedes to power.1) An important syntactic difference between English and LZC is that narrow focus can in English be expressed by stress alone in situ (i.e. without movement), as in 'When the Mie clan has a succession dispute, the cadet always accedes to power'. As this second, equally accurate translation of 4& Ф ífjL suggests, English adverbs associated with narrow focus, such as 'always' and 'only',

9 See Van Valin and La Polla (pp. 206-10) for the distinctions "narrow", "broad" and "predicate" focus. A fuller account is provided by Lambrecht (pp. 221-38). E. Kiss (1998) draws a number of clear semantic and syntactic distinctions between "identificational focus" (the "narrow", fronted type in LZC, Hungarian and the English cleft construction) and "information focus". Kiss claims the latter kind of focus never involves syntactic reordering, implying that it is part of the inherent information structuring of basic word order clauses in all languages.

234


do not invariably precede the focused constituent. In LZC, on the other hand, one just doesn't find sequences like "Ф It 0 al".

This analysis can be extended to account for the syntax of the wide-scope question words he (gu I wei) Щ/Щ (Ш/М) 'why'

and qi м. 'how could ... possibly ...?' (expecting negative response; pp. 142-43). (18) and (19) show these expressions in their canonical prepredicate position, the "predicate" including a focused direct object in (18) and modal $ in (19). See (515) for the "normal"

position of je between the subject and the verb.

(18) &ffifc£#:£*$ (Mo 81.46.54) wo he gu ji zhe zhi bu fu lp QW reason painful SR RS N A shoo 'Why is it the painful that I fail to ward off?'

(19) ШШШММтЙ (Gg29Hu 4.1.2) zhu hou he wei bi tian shou feudal lord QW BE MN hunt chase 'Why must the feudal lords hunt?'

In his description of qi jÉl, Prof. Pulleyblank notes that, while "[t]he subject, if it is expressed, is normally placed in front of jel [, ...1 qi weiaL Ш , qi tu sL Ш, etc., 'it is only ..." precede the subject" (p. 143,

516). This fact, however, remains an isolated, unexplained anomaly of LZC syntax, unless we recognize a connection between the meaning of 'only' and the notion of focus.

Words like 'only' and 'even' are associated with focused

constituents, marking two semantic subtypes of the focusing operation. "Only" makes plain, narrow focus 'exclusive', while "even" evokes a scalar construal of the focused phrase. Syntactically, LZC appears to ignore these distinctions, marking the two subtypes by the same means - phrase- internal inversion and/or resumption in pre verbal focus position, as shown in examples (15) with Щ. 'only' and (16) with ■§• 'even' above.

235

A propos de... / CLAO 30(2001) 215-256

If any phrase within the scope of a morpheme meaning 'only' is, by definition, already in focus, then the generalization we

have proposed above entails Prof. Pulleyblank's claim "jËUfl... precede[s] the subject" - namely, any focused phrase must appear

immediately to the right of the focus-sensitive particles Щ., Ф, il&,

In (516), exclusive focus on the subject is associated with tu

ÍÉ 'only1, which, predictably, can occur in preverbal position as

well. (20) is an example with t*t, while (21) suggests that the syntax apparently treats subject NPs construed with postnominal du Щ 'only' in precisely the same way, moving them into focus position to the right of wide-scope question expressions. Note that the rules exemplified here refer to the interaction with narrow focus of the general meanings 'why' and 'only', expressed by several distinct

lexical items in LZC.

(20) IWRÍlSa^^l^ti: (Zh 2.1.31) qi wei xing hai you long mang zai fu zhi yi QE only form skeleton XV deaf blind EP DP know PL you zhi XV 30

10 It is important to stress that these particles do not themselves perform the operation of focusing the phrases that follow them. Such phrases are already in plain, unmarked focus of the type exemplified in (232), (233), (235) and (237). Note that phrases preceded by the particles typically both occur in the same position and are resumed in the same way as plain focused phrases. The particles simply mark phrases with more specialized types of focusing, 'exclusive1, 'scalar', 'negative presupposition', etc. Thus, the mere occurrence of #, etc. does not entail focusing; rather, a phrase already in focus will assume a specific position with respect to such morphemes. That 1^ does not actually assign focus is clear from its occurrence as a modifier of NPs in oblique positions, where focus is not licensed, (a) fô&ËfèftSHÉ! (Hf 6.2.30)

gu zhong chen wei si yu fei zui thus loyal vassal risk death LP non fault 'Thus, loyal vassals will be in danger of losing their lives for "non- offences" (= on false charges).'

236


'Can it be only the physical body that suffers deafness and blindness? (Hardly!) The mind suffers them as well.'

(21) ЩиШШШЪ$:Ш& (Xz67.18.46) he wei Chu Yue du bu shou zhi ye QWBE PN PN alone NA accept constraint PL 'Why do Chu and Yue alone not accept these constraints?1

One further crucial example of the interaction of a focused phrase with other preverbal constituents will allow us tentatively to sketch, in a greatly expanded version of (11), the "fine structure of the left periphery" (Rizzi) for LZC.

(22) ĚmUJfcfiÉJŘ^fiiJíf (Ly 10.6.8) You ye guo yu cong zheng hu he you PN PL decisive LP follow government LP QW XV 'You is decisive. What (problems) would (he) have in following an official career?11

Here the object interrogative he Щ 'what' occurs to the right of the

focused adpositional phrase Ш ÍaÉ Й -?•, suggesting, together with the data cited above, the following tentative scheme for the syntax of focus in the LZC clause.

11 Here the narrow focused adpositional ШШШ¥- differs from the focused it- phrases of (14a, b) both in terms of its marking and its relation to the context. There appear to be three variants in the marking of focus on ]ït -phrases in LZC: [NP Ш], as in (15), [Ш NP ¥], as here in (22), and [NP П fi ¥], as in (14a, b). The three examples cited suggest that it might be the phonological weight of the NP which determines the marking pattern. (22) is Confucius' response to Zhong You ke shi cong zheng ye yu \$ й "1 1Í № Ш. & Щ 'Can Zhongyou be employed as an administrative officer?', a direct request for an assessment of Zhongyou's suitability for such a career. The packaging of the answer, with focus on Ш Îtë. J$ ■?■, implies that such employment is unlikely to present You with any sort of challenge.

237

A propos de.../CLAO 30(2001) 215-256

(23) 1 2 3 4 5 6 7 VERB . . . Щ/Щ íft Focused

Topic Subject (ЙС/ЙУ Щ и$/# argument/ Щ .. t .. Ж $ adjunct 'what?1

why? / (+f) how could?

It appears that as many as seven positions can be distinguished in this part of the clause, though no single clause I have ever seen has all seven slots filled. Nor, of course, is any single slot of the seven obligatorily filled in all LZC clauses. As clauses consisting of a bare predicate are not at all uncommon in LZC, we must characterize all seven prepredicate sites as "optionally filled". The claim is, however, that any fillers of these sites will always occur in the sequence given in (23). Thus, LZC 'why' seems always to follow an unfocused subject, but precede

modal ift, as in (19) above, and (ft, in turn, to precede any focused phrase, including a subject, as in (17) above. This suggests again that {ft, et al. are perhaps not best described as "adnominal" or

"introducing contrastive exposure"; rather, the generalization would be that any argument or adjunct in focus must occupy slot 6, within

the scope of the modal, exclusive and scalar particles listed in 4 and 1 9 5. Put another way, as all numbered slots in (23) are optionally

filled, when 5, 6 and 7 are empty, as in most clauses in the corpus, the modals and other wide-scope particles will appear to be "preverbal". This perception, however, is simply an artifact of certain frequencies of clause-type coupled with a failure to conceptualize optionally empty positions in the abstract structure of the clause. From the perspective taken here, operator morphemes such as l|I,fi, etc. do not "float" from "preverbal" to "pre-exposed nominal" position. Rather, they are confined to "pre-focus"

12 (239) shows co-occurrence of Ш with ^ Щ, in that order, suggesting that, of the morphemes in position (4), modal Щ, at least, can take scope over Щ and its negative forms, ^ Щ. and i\-.

238


position, as we have attempted to define it above, and it is focused phrases, whether arguments, adjuncts or entire VPs, that must "move" from their canonical sites to occupy the position in the immediate linear scope of such operators.13

The scheme proposed in (23) will no doubt require revision and refinement on the basis of further study. (Certain improvements in (23) will in fact be proposed below, in (36) and (38).) Taken alone, it far from exhausts the syntactic mysteries of the left periphery of the LZC clause. We have shown that narrow focused phrases in the language are doubly marked: by position, in site (6), and by preverbal resumption for subjects (Й) and objects (jkl Ž-), and inversion (Ap NP > NP Ap), etc. for "prepositional" phrases. However, totally unmarked, "bare" NPs can also occur somewhere between the subject and the predicate, where they receive various "adverbial" construals (pp. 99-102), making them quite distinct

13 Another morpheme which perhaps should be assigned to position 4 of (23), at least in the language of the Zuozhuan, is the quantifier уи i^'all' (pp. 128-29), usually "adverbial", but "adnominal" before the focused subject in (454), the only "adnominal" example in that text.

(454) g Ш * M > Щ m Ë '»' £ Ж Ш (Zz 348 Zo З.п.8) qi wei gua jun ju qun chen shi shou qi kuang 'Surely it will be not only our ruler but all his ministers who will receive the benefit of your gift.'

This then appears to be another case of a focused subject moving to a position within the scope of a wide-scope quantifier or operator, directly comparable to (20) and (21) above.

More generally on the syntax of quantifiers, Professor Pulleyblank's all- too-brief account of "exceptions" to SVO order (pp. 14-15) should be expanded to include mention of the wide-scope adnominal quantifiers fan Я 'all' (p. 127) and тех % 'every' (p. 130), neither of which is ever attested in oblique positions. The adnominal use of Щ with fronted oblique NPs is not treated in the Outline, but see Ly 5.3.15 and Me 30.4B.2 for examples. Note that preverbal obliques quantified by Л and, in Lunyu, Щ differ formally from both T and F in that they are not resumed elsewhere in the clause. The construction with Щ (and À Л) in Me 30.4B.2 is different. Both Щ À and A À function as objects here, but they are connected to the verb by M and then resumed postverbally by Ž., the resumption pattern described above as characteristic of topics. '

239

A propos de... /CLAO 30(2001) 215-256

from narrow focused phrases both in terms of form and meaning. The following sentence is cited in Chapter XI "Pronouns and related words" as evidence that ren À can mean 'each'.

(352) Ъ~£КШ%% (Zz 283 Xg 15.Ш.З) bu ruo ren you qi bao 'It would be better that each (of us) keep his (own) treasure.1

[translation altered]

The interpretation of the example is not at issue: the corpus includes sentences in which the word À in a preverbal position must be construed as 'each1. The question we can now ask is, which position? Is the distributive meaning 'each' available to NPs in subject position, site (2) in (23)? If not, then (352) belongs in the next chapter of the Outline, "Adverbs". Because the subject ('we') is covert in (352), it is not easy to see that it is a non-subject position which licenses the distributive construal of A. In neither of the following two examples can Л be construed as subject. In (24), a (partial) subject precedes the distributive, while in (24a), reading À as subject leads to incoherence.

(24) f Qi guo zhi zhu gong zi qi ke fu aid SR NC sire PN state GN several sire son 3G AP zhe fei gong zi Jiu ze Xiao bai ye son PN RS PL

yu zi ren shi yi ren yan CM 2p man serve 1 man LP3 The scions of Qi who may be assisted (are), if not Scion Jiu, then Xiao Bai (=only Jiu and Xiao Bai deserve to be assisted). (Let's me) and you each serve one of them.'

(Hf 23.26.4)

240


(24a) ЪЯлАШ—Щ (Hf 35.7.17) bu ru ren fa er jia NA be. like man fine 2 carapace 'It is better to fine each (of them) two pieces of armor ...'

The subject slot in (24) is occupied by a coordinate NP

marked by у и Щ, the underlying NP, wo у и zi Ш H ~F 'I and you',

only partially represented as Щ ~F 'and you'. If this is the correct

analysis, then À, construed as 'each', occupies a position between subject and verb.

In (24a) À can hardly be the subject of fa Щ 'punish; fine', because construing it as such leaves us with an incomplete proposition: ?'It would be better for A (someone?) to fine two pieces of armor'. Fine whom? More likely, À here is the underlying direct object of Ш: Ш A — ¥ 'fine the people involved two pieces

of armor1. With Л in object position, however, we suggest the clause can only mean 'fine those involved (collectively) two pieces of armor', i.e. 'require them as a group to supply two pieces of armor as a fine', because no bare NP can have a distributive construal in an oblique slot.14

In (24) the domain of quantification by À is the subject, [Щ] % ~F, while the target is the entire predicate, Щ — A M, the

quantifier A distributing the target evenly over the domain. In (24a), however, pre verbal A distributes a second object target, — ¥ 'two pieces of armor' over a covert primary object, ('them'). A in

14 Nor can distributive and most other quantifying morphemes occur in oblique positions. Cf. *Ш %x К I *Ш Щ À — ¥, both syntactically ill-formed, because LZC ^ is an "adverbial", not "adnominal" quantifier, and wide-scope [Щ NP] only occurs preverbally. All the evidence we have suggests that the two postverbal oblique positions in the LZC clause are simply closed to distributive and wide scope quantification. See Harbsmeier (1981, pp. 166-75) for valuable observations on what appear to be the only three non-numerical quantifiers, all definite, attested in oblique positions and Harbsmeier (1998, pp. 154-56) for general claims about the severely limited quantificational capacity of the LZC predicate.

241

A propos de... /CLAO 30(2001) 215-256

(24a) might thus be thought of as "fronted for distributive construal" from immediately postverbal object position to the position between the covert subject and the verb Ш. This sort of anaylsis would be in line with Prof. Pulleyblank's explanation of the "adverbial use of nouns" as involving the movement of an NP in a canonically postverbal phrase into immediately preverbal position (pp. 99-100). Thus, A Щ — tP might be thought of as "derived from" ÍA— Ф, just as the author claims that ren H A JL 'stand like a man1 (354) is derived from H ru ren ÎL Ш A, shan si ill Ш 'die in the mountains' (356) from si yu shan fi Ш Ш , and shou yuan ^ Ш (357) from yuan ... у i shou fë... ПК ¥-, etc.

In the case of distributive À, however, such a syntactic derivation must be rejected on empirical grounds. In the first place, the covert object of Щ in (24a) is anaphoric Ž., not indefiniteÀ. Furthermore, the proposed derivation would account only for cases in which the quantifier's domain is an oblique phrase, as there is no obvious postverbal position from which a distributive with a preceding subject domain can be "moved". How could the A of (352), А^ЖШог of (24), ЦтА* — AM be derived by movement out of a postverbal position?

Further evidence of the impossibility of deriving distributively construed preverbal NPs from oblique positions occurs in the same Hanfeizi passage cited in (24).

(25) IttlI^i^^íi (Hf 35.7.2) bai xing li mai niu er jia wei wang dao 100 surname vicinage buy ox CO family BE king pray The commoners by vicinage purchased oxen (one ox per vicinage) and by household made supplication on behalf of the king.'

There is no identifiable position within the LZC predicate from which the distributives H JH 'each/by

vicinage1 and jia Ш 'each/by household' can be preposed. Semantically, the subject bai xing W

Ш names the set of all individuals of a certain social status, while

242


Ж and Ш are the names of (communally organized) subsets of the referents of the subject NP. In a non-subject, prepredicate position, such NPs parcel out the performance of the action named by the predicates into separate, collective instances of activity. Alternatively, we might say that in each clause, the action of the target predicate, eg. Ж 41, is distributed over the entities Mandlí, each of which is subsumed by the domain subject, U Ш.

This sort of data suggests that the broadest generalization is

not simply that À "can also mean 'each'" (p. 98), nor that some NPs interpreted as distributives are the result of movement out of VP while others are not. Instead, one might posit a general "adverbial NP" position between subject and VP in which the precise construal of bare NPs is constrained semantically. Thus, any bare count-noun in that site will receive a distributive construal iff its meaning can be subsumed by either of the sets named by the core argument NPs, as Л lit. 'a person' is subsumed by the covert subject 'we' of (352),

the semi-overt subject of (24) and for the covert object of Щ in

(24a), and as M and Ж are subsumed by the W Ш of (25). If this precondition in terms of subsumption is not fulfilled,

a prepredicate À, etc. will not be construed distributively. Consider the notorious example shi ren H er ti ^ À 3L fin Щ 'The pig stood up like a man and wailed1 (354), where the distributive construal of the same À in (presumably) the same position as in (352) is blocked by the semantic mismatch of i^withÀ. If we make exception for the special domain of metaphor (and its narrative

elaboration into Orwellian allegory), Щ. simply does not subsume À .

5. THE RIGHT PERIPHERY: SECONDARY PREDICATES AND PREDICATE-INTERNAL QUANTIFICATION.

In connection with the "adverbial" quantification we have been considering here, let us turn again to consider the right periphery of the LZC clause. There is an important generalization to be proposed about this site, one not easily seen from Professor

243

A propos de... /CLAO 30(2001) 215-256

Pulleyblank's account of the predicate in terms of verb, "objects"

(pp. 27, 31, passim), and four types of complement-like phrases: "embedded noun predicate" (p. 33), "locative" (p. 54), "descriptive" (p. 57) and "numerical" (p. 58).

We noted above in section I that most locatives in Ш, etc. are distinguished qualitatively as "adjuncts" from "descriptive complements with #P and Щ", adjuncts being a class of optional, typically circumstantial phrase, not demanded by the argument structure of most verbs. The three other complement phrase-types in Prof. Pulleyblank's typology are also optional, but in their semantics, clearly differ from both objects and circumstantial adjuncts. What #П/Ш complements share with "embedded noun predicates" and "numerical complements" is the function of predicating.15 When they follow a verb and its object, if any, t$- phrases simply do not have this function. As also noted above, in their very high-frequency adjunctive function, № -phrases appear to obey movement rules analogous to those for objects, whereas the three types of clause-final complement (NP, #П / Ш and numerical) seem not to be targeted by such rules. One clear indication of this is that complements are interrogated in situ, rather than by fronting of the question word, as in the case of interrogated oblique phrases. (Compare he wei ЩМ 'do what?1 with wei shei МШ 'be who?1, also ru zhi he #P Ž. Щ lit. 'treat it how?'; see pp. 91-97.) Further work may uncover additional distinctions between these three types of complement and oblique phrases, which always precede them.

There is a fourth type of clause-final complement, the adjectival, apparently unnoticed in the Outline. Bare adjectives, of course, have an uncontroversial predicative function in LZC (pp. 24-25).

15 The claim that NP, [Ш NP] and [Ш NP] can function as predicates in LZC is uncontroversial. For numerical predicates, see Outline p. 58.

244


(26) =?Ц%2.Ш& (Hf 13.1.25) zi xi ku zhi bei ye 2p QW weep 30 sad PL 'Why are you weeping for it (so) sadly?1

(27) [ш] ШШШ^^Х (Me 37.5A.6.2) Yi shi ze yu minweijiu PN dispense bounty LP folk AN last.long '(Yi) did not dispense bounty to the people for a long time.1

(28) 1ЙЙЖШ&^ (Zh 45.17.88) tiao-yu chu you song-rong minnow emerge swim leisurely 'Out swim the minnows, so free and easy' (Graham 1981, p. 123)

(29) ^ci Ш ЖШШ^ (Me 10.2A.2.2) fu zi guo Meng Bin yuan y i Sir exceed PN far PL

'You, sir, surpass Meng Bin by far.'

I suggest "secondary predicate" as a cover term for all four types of clause-final complement: NP, adjectival, numerical and ru

Ш/ you Щ. One crucial point to note is that Prof. Pulleyblank's

"complement in Ш", as in (183), is not really a separate type, but rather a subtype of "embedded noun predicate" - one introduced by

a prepredicate particle.16

(Cf. prepredicate particle wei ^ introducing the stative predicate jiu X'last long' in (27) above.) Recognizing that secondary predicates may consist of [particle + predicator] leads directly to an improved understanding of certain aspects of quantification in LZC.

16 On the particle status of Щ, see again Outline p. 18, "Verbless comparison with Щ", where the fact that the morpheme is never found negated is cited as good evidence of its non-verbal status. To such negative evidence one might add the apparently unattested ?*| t, ?*ЯЯ Ш, etc.

245

A propos de... /CLAO 30(2001) 215-256

Let us return to the syntax of distributive À, 'each', discussed above. We argued that this interpretation of the morpheme is licensed only in a position between the subject and the predicate, the precise site to be determined by further study. We also showed that this non-subject À can distribute either an overt object target over a domain subject, as in (352) and (24), or an empty first object domain over an overt second object target, as in (24a), 'fine each (of them) two pieces of armor*. Perhaps prepredicate À can only be construed with a first object as domain if that position is empty, as in (24) ^#ПАМ 0 — Ф.17 This idea appears to receive some support from (30), where À takes a second object (02) as its target, distributing it over an overt first object domain (01).

V 01 02 2ndary predicate (Hf 35.7.8) (30) £ В W Z À ZLf

wang yue "zi" zhi ren er jia king say fine 3O man 2 carapace The king said, 'Fine them each two pieces of armor."18

17 Cf. the analogous syntax of the "pronominal adverb" zi Ě 'self, which cannot refer to an overt DO. If a DO is present (or the verb intransitive), i will refer to the subject with the reading 'personally' or 'of (subject's) own accord'. Cf. Outline p. 136. 1 8 The verb in this example is zi Ш 'mulct', here written Щ ci. Its argument structure is identical to that of Щ 'fine': ©01 (person) O2(amount)\ as suggested by the second clause of the following example from the Qin bamboo strips unearthed at Shuihudi.

Subj VI 2ndarypred V2 01 O2 (a) ill îî * IE 1лШ Й1 Ш Ж MJi— ¥

heng shi bu zheng shi liu liang yi shang "zi" guan se fu yi jia level scale NA correct 16 liang IN up mulct reeve 1 carapace 'If the stones of a lever scale are off by more than 16 liáng (~250g), fine the official reeve one piece of armor.' (Shd 1 13)

In terms of their ditransitive argument structure, the two verbs Ш and Ш are similar, but not identical, to "privative" duo Щ {Outline pp. 32, 108). Cf. also Peyraube's translation of the quoted part of (30) £. В : Ш Z. À — ¥ , "II taxa ces gens d'une amande de deux cuirasses" (1988, p. 91), where Ž. and Л have

246


We have claimed that NPs with distributive interpretation are not

licensed in either object position, 01 or 02. Thus, the only way À can distribute an 02 over an 01, as in (30) to mean '(fine) them each two pieces of armor' is for the 02, ZL Щ , itself to become a

predicate, the sort of phrase distributive Л requires as syntactic host. This in turn implies that — Щ moves out of its original non- predicating 02 site into clause-final position, where, as a predicate, it can accommodate "adverbial" (more accurately, "prepredicate")

À 'each1. If the evidence of rightward movement of an oblique phrase

to create a secondary predicate as a host for quantification is not clear enough in (30), consider a final example with the "adverbial" distributive ge & 'each', syntactically analogous to distributive À.

V 01 locative 2ndary predicate

(31) 7}$7j<££#-A (Gz 57.17.12) nai qu shui zuo you ge yi ren then take river left right each 1 man 'Then take from each river area one person.'19

Here & distributes the object, — À, of qu Ш. 'take' over the adjunct, 7jc ;£ £f, to mean 'take one person from each river area'. But -§-, unlike English 'each', is debarred from a locative phrase like(|&)7jc ti >fr: we never find the sort of structure implied by

*Jb Ж — Л # зК £ %i. Nor, apparently, can prepredicate -§- reach into the predicate to distribute one oblique phrase over another.

According to Harbsmeier's work on #(1981, pp. 82-87), & Щ. — * К (Ш) тК íí ̂ can only mean '(they) each took one person

been construed together as a determiner phrase, "ces gens". The use of Ž. as a determiner is not attested elsewhere in the Hanfeizi, as far as I can tell. (The reading also neglects the initial ï В , missing the illocutionary force of the king's words.) 19 In (31), I follow Yasui (57.12) on the interpretation of ï £" in the phrase 7K ÍL £î, an "unmarked" locative (Outline, p.55), equivalent to № 7jc ;£ £f.

247

A propos de... /CLAO 30(2001) 215-256

from the river area(s)', 'each' distributing 'one person' over the

covert subject NP to constitute a meaning very different from that of (31). The syntactic solution is basically as in (30), but the movement involved is more obvious. Thus, we find — A, the direct object of Ж, not in its canonical position immediately to the right of the verb, but in the right periphery of the clause, where it can assume the role of numerical predicate to accommodate the "adverbial" quantifier &, which can then distribute 'one person' evenly over multiple 'river areas'.20

20 Cf. Cikoski (p. 78) and Harbsmeier (1981, p. 85) for previous analyses of (31). Cikoski treats the — of -A as a causative verb, 'cause-to-be-one the person', because he assumes # must be preverbal, rather than prepredicate. Harbsmeier opts for verb ellipsis, apparently deriving (31) from an underlying ? )b M (№) ?K ti íí, fa Ж ■ ' À on the basis of the same assumption. Note, however, that according to Harbsmeier's own conclusions about the syntactic behavior of fa, the quantifier in a putative fa Ж -■- Л could only have the subject of Ж as its domain, an interpretation that fails to solve the problem presented by this sentence, that of quantifying within the VP. Given the syntactic constraints on the interpretation of fa (namely, it must take the [cjovert NP which precedes it, eg. a subject, as domain), what is needed for the correct interpretation of (31) is for 7jc tc. ÍÍ and — Л to interact alone, as domain and target of ^-quantification, without interference from the (covert) subject of Ш, an NP extraneous to the quantification intended. Both previous analyses fail to appreciate (i) the predicative function of [Number+NP] phrases (see again Outline, p. 58), and the fact that (ii) such predicates can be preceded by particles (as in (189) with ffi and (27) with ^) and (iii) the whole [particle+numerical predicate] phrase embedded in the right periphery of the clause as the language's sole strategy for allowing one oblique phrase to quantify over another.

Note further that the first clause of Harbsmeier's example (41), you feng er zi zhe ge wanjia zhi xian yi X M — ř # fa Ř \ Ж Z. Ш "... '(If I) go on to enfeoff the two barons each with a district of ten thousand households, ...' (1981, p. 85; translation altered), is amenable to the analysis we provided for example (30) above, the 02, 11^1, moving from argument position to secondary predicate position for distribution by fa over 01, Z. Ť #.

248


The following example further illustrates how i§- creates a separate quantificational phrase, [ír + Target], to the right of its domain whenever its quantifies between oblique NPs.

(32) 3Ei Ш. Ai wang she qi king set. up 3G

mu pastor,

0.

Щ zhi tailor PI zhi

Ж) qi 3G

qi

Щк , 0k zhi duty Ml, 01 gong tribute

# VX Ж) ge yi qi each IN 3G

# 1Д Kj ge yi qi

Яг 1;

suo neng OR able

suo you XV

'The king establishes his pastors and regulates their duties, (the duties) each according to their (the pastors') abilities. He regulates their tribute, (the tributes) each according to their resources.1 (Zl 61.29-30)

The above translation reflects the claimed syntactic constraint on OC j~r that it follow its domain. In modern English, however, a domain of quantification phrase such as 'the duties each1 can be realized in situ as a quasi-determiner 'each of -phrase, collapsing the two ОС phrases, a VP, P\ Ж Ш , and a quantificational phrase, & Шч Ж Яг Ш, into a single quantified English VP: 'The king establishes his pastors and regulates each of their duties according to their abilities'. ОС, we claim, lacks this second option for the in situ, VP-internal expression of distributive quantification.

Data like that in (32) suggest a more precise

characterization of predication in LZC. Describing # ])X Ж Яг Ш as 'secondary predicate' in the bipartite rl'J Ж Ш , & VX Ж Яг Ш, implies that certain LZC adpositions can actually head predicates without becoming "full verbs" (the diachronic antecedent implied by the descriptive term 'coverb'). This is the sort of structure found when the newly asserted information in a clause is conveyed by an adpositional phrase alone, the content of the remaining phrases all presupposed. The focused status of the adpositional phrase may be explicitly marked by a focus-sensitive operator, such as modal il& 'must; will surely; always', in (33).

249

A propos de.../ CLAO 30(2001) 215-256

(33) [Л^ДЙЯШ, ШШР¥ ; »#£ ;] AIP # ЕШ Ж (Zz 434. Zo 31.7.3) ru Ying bi yi geng chen enter PN MN IN gercg c/zen '[Six years hence, when this month arrives, Wu will probably enter Ying, but will in the end fail to overcome it.] (That) entry into Ying will surely be on gengchen [day 17].'

Compare the default, predicate-focus alignment of the same phrases in (34) with (35), the relevant part of AP-focused (33).

(34) predicate

'Wu will surely enter Ying on gengchen.'

predicate (35) [££/£] A MP 0 VÀfkJu

'(Wu's) entry into Ying will surely be on gengchen.'

The above claims, hitched to those in (23), imply the following still very preliminary picture of the LZC clause. Once again, all positions are optionally filled, even "predicator" under certain (admittedly unusual) circumstances. (WS = wide scope, QW = question word)

1 2 3 7 PREDICATOR 01 02/Adict 2nd pred (36) T S WS >SL\ Щ/Ъ ¥{f)\4im no WS quant, (particle+) NP/

QW Щ,... 'what/whom?1 no F, no QW Num/Adj/AP 'Why?'

In the examples cited in this review, matrix negator ^, preverbal adpositional phrases and bare distributive NPs do not interact with the operators and focused phrases whose linear order we have tried to capture in (36). Data cited by Wei (p. 267),

250


however, demonstrate that ^ occurs between a constituent WH-

phrase, Щ (NP) 'what (NP)?', and the predicator, as shown in (37).

(37) !Д1Ш1ШЛйШ^З£ (Zz 92 Xi 4.4.-3) yi ci gong cheng he cheng bu ke IN DP attack wall QW wall NA overcome 'If we use this to attack city walls, what walls will we fail to overcome?'

This sort of evidence allows us to revise the preverbal alignment sketched in (36) as (38) below.

12 3 4 5 6 7 8 PREDICATOR...

(38) T S WS #,&£/# F(f) Щ1Ш ^ QW Ш, ... 'what/whom?1 'Why?'

6. CONCLUSION

Needless to say, intermediate students of classical Chinese who meditate on (38) are unlikely to achieve satori; indeed, their quiz marks may even suffer. Far better that they be encouraged to study Prof. Pulleyblank's lucid analyses of virtually all of the most important constructions in LZC. In commenting on some of these analyses, I have tried to suggest how they might be refined by introducing compositional semantic constraints on what are clearly ambiguous markers (eg. M") or positions (eg. a bare NP between subject and verb).

On broader syntactic issues, a linguist working on LZC without a general theory of the nature of the LZC clause will be ill- equipped to "bring the grammar of Classical Chinese into the linguistic mainstream" (p. xiv), a development hoped for by Prof. Pulleyblank and other linguists of Chinese. The theory sketched in (36, 38) is considerably more detailed and hence, I would suggest,

251

A propos de... /CLAO 30(2001) 215-256

appreciably closer to the "whole story", at least about focus and many of the preverbal operators, than any set of claims found in the Outline. At the same time, there are canonically preverbal phrase- types still missing from (38). My suspicion is that any theory of the LZC clause, Prof. Pulleyblank's, my own or any other, can be refined through further investigation of the relation between the possible positions of phrases in the clause and their interpretation. On this basic question, several interrelated points have emerged from the above discussion. Beyond the SVO basic order which LZC unquestionably shares with English, (i) the LZC clause contains not one, but several syntactic positions between S and V. These "preverbal" positions, together with the right periphery of the clause, are exploited by a syntax which, quite unlike that of English, (ii) simply does not permit certain semantic operations (eg. narrow focus, distributive quantification) in either canonical subject or oblique positions because (iii) it is highly sensitive to scope both of the quantificational and focusing kinds, the latter subsuming both narrow NP focus and constituent question formation. Specifying the details of that sensitivity within a framework which allows direct comparison with the grammars of languages similar to LZC, not necessarily in terms of basic word order, but rather in terms of "movement typology" (e.g. Hungarian), appears to be one useful direction in which work on LZC might proceed. Whatever shape it may assume, future progress in the field will surely in large part be built on the sturdy foundations laid by Prof. Pulleyblank in this book.

ABBREVIATIONS

AA agent-oriented ability: x nJJ^ VP 'x can VP' AN aspectual negative: ^ 'not yet, not completely, not entirely' AP patient-oriented ability/permission: x nj V 'x can/may be V-eď BE benefactive: ^4 'for (the sake of)' CM comitative: Й13 '(together)

with'

CO connective: ffn 'and (yet)' etc.

252


DP demonstrative pronoun/adjective: Ц, fck, Ш » ^ etc- ЕР exclamatory particle ц% F focus f focus resumptive GN genitive (NP £ NP) / nominalizer (NP £ VP) LP locative-path J^/ f/ ^ ' at, from, than', etc. LP3 locative-path + 3p pronoun: Щ MN modal particle of necessity: >& 'must; always' NA negative adverbial: ^ NC negative copula: |^ NE negative existential: &E OR object relativizer: Щ PL particle PN proper noun QE question/exclamation particle: W, ^p-, Ш, fi^.etc. QW question word: foj 'what; how; why1, |§ 'who(m)' RS resumptive proform: filj ; resumptive use of Щ , £_ , Sff , etc. S subject SR subject relativizer: ^ T topic t topic resumptive WS wide scope XV existential verb: Щ 3G 3p genitive / nominalizer in subjectless clauses: Ji 30 3p oblique, (in)direct object pronoun: ^

BIBLIOGRAPHICAL REFERENCES

Primary sources (editions used as in the Outline, except where noted):

Gg &%.&

, 1968. Hf П$:?ШЦ\.ЩШЖЩ±т.11яг *ФШШ), 1982.

253

A propos de.../ CLAO 30(2001) 215-256

Ly ШШ Me Ží Shd ШйШШШ^ШМйШШШ^ЩШШ'ЬШ.Ш.1Шг it) Ж

fôtt, 1978. Xz Zh Zl

Zz

Secondary sources:

CHOU Fa-kao ЩШМ (1959-62). Zhongguo gudai yufa. Ф I [Grammar of Ancient Chinese]. Taipei : Zhongyang yanjiuyuan lishi yuyan yanjiusuo.

CIKOSKI John S. (1975). Chinese 2A : Introduction to Classical Chinese. Mimeograph. Department of Oriental Languages, University of California, Berkeley.

DONG Zhiguo lip® (1988). Gudai Hanyu juxing daquan. tyrtljíik [Compendium of sentence-forms in Ancient Chinese]. Tianjin : Tianjin guji chubanshe.

É. KISS Katalin (1995). Introduction. In É. KISS Katalin (éd.). Discourse configurational languages. New York: Oxford University Press.

É. KISS Katalin (1998). Identificational focus versus information focus. Language, 74, pp. 245-73.

FENG Shengli (1996). Prosodically constrained syntactic changes in Early Archaic Chinese. Journal of East Asian linguistics, 5, pp. 323- 371.

GOLDBERG Adèle E. (1995). Constructions : A construction grammar approach to argument structure. Chicago: University of Chicago Press.

254


GRAHAM A.C. (1977). The Chinese particle tzeng •§■. Early China, 3, pp. 31-35.

GRAHAM A.C. (1981). Chuang-tzu : The seven inner chapters and other writings from the book Chuang-tzu. London: George Allen & Unwin.

GRAHAM A.C. (1983). Yun ží and Yueh В as verbs and as particles. Acta Orientalia, 44, pp. 33-71.

HARBSMEIER Christoph (1981). Aspects of classical Chinese syntax. London : Curzon Press.

HARBSMEIER Christoph (1998). Science and Civilisation in China, v. 7, part I : Language and Logic. Ed. by Kenneth Robinson under the general editorship of Joseph Needham. Cambridge : Cambridge University Press.

LAMBRECHT Knud (1994). Information structure and sentence form : topic, focus and the mental representation of discourse referents. Cambridge : Cambridge University Press.

LEVIN Beth (1993). English verb classes and alternations. Chicago : University of Chicago Press.

PEYRAUBE Alain (1988). Syntaxe diachronique du chinois : Evolution des constructions datives du XIVe siècle av. J.-C. au XVIIIe siècle. Paris : Collège de France. (Mémoires de l'Institut des Hautes Etudes Chinoises, v. XXIX).

PULLEYBLANK Edwin G. (1960). Studies in early Chinese grammar. Asia Major, 8, pp. 36-67.

PULLEYBLANK Edwin G. (1991). Some notes on morphology and syntax in classical Chinese. In : ROSEMONT, Henry (éd.). Chinese texts and philosophical contexts: essays dedicated to Angus C. Graham. La Salle IL : Open Court. Pp. 21-45.

255

A propos de.../ CLAO 30(2001) 215-256

PUSTEJOVSKY James (1995). The generative lexicon. Cambridge MA : MIT Press.

RIZZI Luigi (1997). The fine structure of the left periphery. In HAEGEMAN Liliane (éd.). Elements of grammar : Handbook in generative syntax. Dordrecht : Kluwer. Pp. 281-337.

SUN Chaofen (1996). Word-order change and grammaticalization in the history of Chinese. Stanford : Stanford University Press.

TRASK R.L. (1993). A dictionary of grammatical terms in linguistics. London : Routledge.

VALLDUVÍ Enric (1992). The informational component. New York : Garland.

VAN VALIN Robert D., Jr. and Randy J. LAPOLLA (1997). Syntax : structure, meaning and function. Cambridge: Cambridge University Press.

WEI Pei-chuan %% i§ ^ (1999). Lun xian-Qin Hanyu yunfu de weizhi. ШУсМШШШШ № fî Ш [On operator position in pre-Qin Chinese]. In : PEYRAUBE Alain and SUN Chaofen (eds.). Studies on Chinese historical syntax and morphology : Linguistic essays in honor ofMei Tsu-lin. Paris : Ecole des Hautes Etudes en Sciences Sociales, Centre de Recherches Linguistiques sur l'Asie Orientale. (Collection des Cahiers de Linguistique Asie Orientale, 3). Pp. 259-297.

XU Jie and LI Ying-che fê?*, $31 tï (1993). Jiaodian he liangge feixianxing yufa fanchou: fouding, yiwen. ÍŘ 1й f О Щ ÍIHN !Ш4 ШШШ.Щ: ■ЁЯсШЩ [Focus and two non-linear grammatical categories : negation and interrogation]. Zhongguo yuwen, 2, pp. 81-92.

YANG Bojun and HE Leshi ШЙ*ё, ЩШ±(1992). Gu Hanyu yufa ji qi fazhan. Й"Щ In Ш Ш Jk. % ШШ. [ Ancient Chinese and its development]. Beijing : Yuwen chubanshe.

256


YASUI Mamoru [Sokken]. $:#Ш [ Ж$Т ], nd. Guanzi zuangu. Щ~тШп&. [Collected glosses on the Guanzi]. Rpt. Taipei : Heluo tushu, 1976.

Derek HERFORTH SEAMELS, A 18 Sydney University NSW 2006 AUSTRALIA [email protected]

pulleyblank-classical chinese grammar review

Documents