DEGREES OF CONTRAST AND THE TOPIC-FOCUS ARTICULATION

The aim of the present contribution is to present s veral hypotheses that may help to characterize and identify contrast as such and degrees of its intens ity, and to point out how the phenomena connected with contrast can be handled in a descriptive frame work that uses a dependency based syntax and includes a description of the topic-focus articulat ion (TFA) as one of the aspects of the underlying sentence structure. We also illustrate how the chos en descriptive framework is tested in a syntactic annotation of a large corpus. First we examine the chosen description of TFA (Section 1), then the means of expression of contrast are discussed (Sect ion 2), as well as issues of contrast in focus (Section 3), and especially in topic (Section 4). 1 Topic-focus articulation 1.1 Topic, focus, and contextual boundness in a linguistic description In the theoretical framework of the Praguian Functi onal Generative Description (FGD, see Sgall et al. 1986, Haji čová et al. 1998), the semantic basis of the articul ation of the sentence into T(opic) and F(ocus) is understood as the relation of aboutness: a prototypical declarative sentence asserts that its F holds (or, with negatio n, d es not hold) about its T. Thus, the core of the semantico-pragmatic interpretation of a decl arative sentence might be based on a formula such as F(T) or, for a negative sentence, a s non-F(T), if for the aim of the present discussion issues such as those of intension, lambd calculus and type theory are put aside. Within both T and F, an opposition of contextually bound (CB) and non-bound (NB) nodes is distinguished, which is understood as a grammati cally patterned opposition, rather than in the literal sense of the term. In the underlying le ft-to-right order, NB dependents follow and CB dependents precede their heads. In unmarked cases, the main verb (V) and those of i ts direct dependents that on the surface follow it belong to F, and the items preceding V ar e parts of T. In marked (non-prototypical) cases, V can be CB, i.e. in T, or (a part of) F may precede V; usually the intonation centre Eva Hajičová, Petr Sgall 2 (sentence stress) then marks F, occupying a marked position. The dependents of nouns primarily are NB. Let us illustrate this view by a typical example (w e understand the intonation center, in the prototypical case expressed by a falling pitch, to be placed at the end of the sentence; in other, marked positions it is denoted by capitals (which w e use also in some other cases, to avoid a possible misunderstanding); let us recall that, in our underlying representations, the counterparts of function words are just indices of n de labels, not occupying independent syntactic positions: (1) My.t brother.t was visiting.t/f one.f of his.t friends.f yesterday.t. focus: (was visiting) one of his friends (intonation center on friends) Here and in the sequel, t denotes a CB item, f deno tes a NB one, and c is used to denote a contrastive CB item. The verb in (1) is ambiguous in that it is NB (and thus a part of F) on one reading and CB (a part of T) on another; while the former is an ap pro riate “full” answer to (2), the latter answers (3). (2) What was your brother DOING yesterday? (3) Whom was your brother VISITING yesterday? This view, the motivation of which has been publish ed several times, makes it possible to analyze similar sentences (with an ambiguous part t h t may contain other words, not only verbs) with a single opposition of T and F. Thus the discrepancy between the singl e relationship of aboutness and two dichotomies assum ed to constitute the information structure (e.g. by Junghanns and Zybatow 1997, ex. (2), p. 29 0) can be avoided and the T-F articulation (TFA) of the sentence can be assigned a specific po sition within the system of language (de Saussure’s langue, Chomsky’s linguistic competence ), namely that of one of the basic aspects of the underlying, tectogrammatical representations of sentences (TRs). No separate le vel of information structure is needed. The TRs contain no nonterminal symbols; each of the ir nodes is labelled by a complex symbol composed of a lexical and a morphological part (values of morphological categories such as number, tense, modalities), and each edge i s labelled by the symbol indicating a syntactic relation (i.e. the type of the dependency relation ). Degrees of Contrast and the Topic-Focus Articulatio n 3 1.2 TFA and contrast in a large corpus The approach of FGD makes it possible to capture TF A and contrast in sentences of most different degrees of complexity. The chosen descrip tive framework, FGD, is being checked with examples taken from the syntactically annotated Prague Dependency Treebank (PDT), in which sentences from running text, from t he Czech National Corpus (CNC, which contains hundreds of millions of word occurrences i n journalistic fiction and other texts) are analyzed by a semi-automatic procedure. In the PDT scenario, three layers of annotation are present, with TFA and contrast being represented (together with underlying dependency re lations) on the underlying syntactic level. The resulting sentence representations have the for m of tectogrammatical tree structures (TGTSs), with the following characteristic properti s: (a) only autosemantic words are represented as sepa rat nodes, with the exception of the coordinating conjunctions (in this point, TGTSs dif fer from the theoretically based TRs), (b) nodes deleted on the surface are restored, (c) the condition of projectivity is met (i.e. no d iscontinuity of sentence parts is allowed), (d) tectogrammatical functions (‘functors’), i.e. k inds of the dependency relation such as (i) arguments: Actor/Bearer, Patient, Addressee, Origin , Effect and (ii) different kinds of adjuncts (temporal, local, condition, manner, etc.) are assi gned as labels of the edges of the tree (or, equivalently, as indices in the labels of the depen dents), (e) basic features of TFA are introduced (f, t, c, see Sect. 1.1 above). Let us note that in the present experimental phase 2000 sentences have been annotated in what concerns their underlying syntactic structure its lf (‘large collection‘), with only 200 sentences having been annotated in full detail (the so-called ‘ model collection‘), and the annotations of 2000 sentences contain a treatment o f TFA). In the sequel, after a more general discussion of t he phenomena of contrast, the checking of our descriptive framework on the material from PDT is illustrated by the Czech examples (30), (31), (33) and (34). Eva Hajičová, Petr Sgall 4 2 The means of expression of contrast Several typical means of expression of contrast ca n be distinguished: (i) E.g. in Czech, strong pronominal forms are used with certain pronouns; the typical cases of oppostion of weak and strong forms are: Czech ho – jeho ‘him-Gen,Acc’, mu – jemu ‘him-Dat‘, t ě – tebe ‘you-Gen,Acc’ , ti – tobě ‘you-Dat’, se – sebe ‘Refl.Gen,Acc’, si – sobě ‘Refl.Dat’. The strong forms are used to express NB pronominal forms, or CB contrastive forms as ona and jeho in (4)(a), respectively; they are also used in pre positional case forms and in coordination, cf. tebe in (5) and tobě in (6). (4)(a) (Petr ji nazval konzervativcem.) Potom.t jeh o.c urazila.t ona.f. (Petr called her a conservative). Then him insulted she. Then he was insulted by HER . (b) (Petr ji nazval konzervativcem.) Potom.t h o.t opustila.f. (Petr called her a conservative.) The sh e LEFT him. (5) Na tebe jsem se celý týden těšil. for you I-have-been Refl (the) whole week lookin g-forward I have been looking FORWARD to you for the whole w ek. (6) Tobě nebo Martinovi to pošlu zítra. . to-you or to-Martin it I-will-send tom orrow I will send it to you or to Martin tomorrow. The corresponding weak (“short”) forms are used onl y as CB, without contrast, cf. ho in (4)(b); it should be noted that Czech, a pro-drop l anguage, has a zero form in the Nominative of all the personal pronouns, which occurs as their weak form (this is the case of the Degrees of Contrast and the Topic-Focus Articulatio n 5 counterpart of she in (4)(b) or of I in (5)and (6)), although the „strong“ forms já, ty, on, my , etc., may also occur without contrastive function, esp. in colloquial speech. In German, English and many other languages (and al so in Czech with pronominal forms such as je ‘them.Acc’, ji ‘ her.Acc’), only an opposition of accented and unacc ented forms is present as expressing that of contrastive (and NB) vs. non-contrastive use. Following up Koktová’s (1999) observation that weak forms of pronouns in Czech cannot be used in certain positions in T, we use the oppos ition of strong and weak personal pronouns as an operational test for the contrastive use in T. Thus, jeho in (4)(a) is contrasted with she; there is no such contrast in (b). However, the appl ication of this test is limited, since not only in coordination or with a preposition, but also whe n used as NB, in focus, the pronominal form is similar to that expressing a contrastive (p art of) topic (marked with c), as is the case of ona in (4)(a), and also of jeho in (7)(b): (7)(a) Jeho.c jsme vid ěli včera.f. – Him we saw yesterday. (b) Včera.c jsme vid ěli jeho.f – Yesterday we saw HIM. (ii) Rising stress (or, perhaps, falling-rising), having the form of L H, falls – perhaps optionally – on a contrastive (part of) topic in e xamples such as jeho in (7)(a) or včera in (7)(b); cf. also jeho in (4)(a). In the sequel we indicate such a “phrasa l” or contrastive stress by italics. It would be interesting to check to wha t extent such examples can be characterized as bearing a hat contour, and under which conditions can the hat contour be tak n as a criterion for contrastive T. Steube (2001) examines similar examples in relation to the concept of I-topic, cf. also Jacobs (1997). Specific pragmatic properties accomp any such accentuation at least in German (bound with specific illocutionary types and scope effects).