Page 215 - Encyclopedia of Philosophy of Language

(e.g., Hornby 1971) identify the notions 'topic' and 'comment' operationally with the notions 'given/old information' and 'new information,' respectively. Topic-constituents are considered to be the rep- resentation of given, contextually bound information that is already stored in discourse. Because of their 'given' status they are, other than constituents which represent new information, no candidates for accent assignment.
Though an explication of meaning differences as in (1) is mostly given in terms of the 'topic' or 'aboutness' notion, no general consensus exists about the expla- nation of these phenomena. Some authors refrain from an explanation in terms of the topic-comment distinction. They exploit either the distinction between 'given/old' and 'new information' in the sentence (Halliday 1967; Kuno 1972) or a formally and oper- ationally similar distinction between 'presupposition' and 'focus' (Chomsky 1971; Jackendoff 1972). When, besides a given-new (presupposition-focus) distinc- tion, also a topic notion is assumed, this is not, as in the previous case, operationally identified with the notion 'given' or 'presupposed information.' Halliday (1967), for example, explains meaning differences like those between the sentences in (1) in terms of the given-new distinction, although according to his defi- nition of topic ('theme') all these sentences are about John. Thus Halliday observes: 'Basically, the theme is what comes first in the clause — The theme is what is being talked about, the point of departure for the clause as a message' (Halliday 1967:212).
A similar view also implying a categorical dis- tinction between what Halliday calls theme (topic) and given information is present in various, mainly so- called 'structured meaning' approaches that provide a syntactically oriented (formal) semantic rep- resentation of the focus structures (e.g., Jackendoff 1972; Jacobs 1983; Krifka 1991; Von Stechow 1981, 1989). A similar view is also present in some of the pragmatic accounts (e.g., Vallduvi 1992, 1993). Con- trary to these views, Steedman (1991) proposes an isomorphism between syntactic, informational and intonational structure based on Combinatory Cat- egorial Grammar in which he adopts the same idea of two independent notional pairs but does not define the topic ('theme') of a sentence in terms of word order, that is, as the sentence-initial element.
From the beginning, the notion 'discourse topic' plays an important role in many, in particular, com- putational and (psycho)linguistic, theories and views about discourse coherence, either explicitly or implicitly (e.g., Grimes 1975; Hobbs 1982; Johnson- Laird 1983; Reichman 1978; Schank 1977). Especially in discourse (text) grammars (e.g., Van Dijk 1977) the notion often functions as the explanatory principle for the structural coherence underlying a well-formed discourse. In general it is assumed that a coherent discourse is composed of a set of hierarchically organ-
ized discourse segments under one discourse topic. The overall discourse topic associated with the dis- course as a whole and comprising all smaller discourse segments constitutes the superordinate discourse topic. Under this common superordinate discourse topic, the discourse topics of the subsegments are ordered paratactically or hypotactically, depending on whether there are inclusion relations between the subsegments. Although it is often claimed that struc- tural coherence phenomena in discourse are 'ex- plained' in terms of the notion 'topic,' these explan- ations are generally intuitive and fail to achieve for- mal precision due, mainly, to the absence of an empirically and operationally adequate definition of discourse topic and an unclarity with respect to the relation between discourse and sentence topics or between discourse topics themselves.
3. Topic Identification
In the topic-comment literature several tests and also several operational characterizations are presented to identify sentence topics. The tests and the charac- terizations will be dealt with below in separate sections. In the most satisfactory cases the proposed tests are meant to take as input a topic-bearing sen- tence and to give as output a specification of the topic of that sentence. In many cases, however, the test is only meant to determine whether some sentence element has topic function, indicating that this element is part of the topic constituent of the sentence. A characteristic of all tests, as opposed to charac- terizations, is the fact that the actual discourse context in which the topic-bearing sentence occurs is not a part of the test itself. The operational characterizations proposed in the literature can be classified into con- text-dependent and context-independent character- izations, as is clarified in Sect. 3.2.
3.1 Tests for Topic-hood
Well-known tests for topic identificationare the front- ing test (e.g., Kuno 1972; Lakoff 1971), the 'about'- context test (Reinhart 1981), the 'about'-question test (Gundel 1977), and a test which is commonly known as the question test (e.g., Sgall, et al. 1973, 1986). The first three tests are fundamentally restricted to the identification of noun phrase (NP) topics.
The fronting test is based on the assumption, not commonly accepted, that NPs which are fronted by, for example, a left-dislocation operation structurally mark topic-hood. This test implies that if a struc- turally unmarked sentence containing an NP , (S<NP,>) c a n be acceptably paraphrased according to the scheme Asfor/Concerning/About NP,, S<NPiy, NP, represents the topic of the original sentence. In Rein- hart (1981:64-65) it is pointed out that the application of this test is restricted to sentences which introduce a new topic or lack a specific or generic indefinite topic-NP .
Topic and Comment
193