Page 217 - Encyclopedia of Philosophy of Language

1985, 1992); and characterizations in terms of ques- tions (e.g., Bartsch 1976; Keenan and Schieffelin 1976; Klein and V on Stutterheim 1987; V an Kuppevelt 1991; Stout 1896; Strawson 1971; V ennemann 1975).
Regarding the first type, the topic of a sentence is identified with the given/old or contextually bound information in the sentence. The identification of sen- tence topics is thus reduced to the identification of given information. A direct consequence of this approach is that sentences that only represent new information are topicless. Therefore, the criterion for sentence topics is that a given-new modulation is defined for it. Apart from the problem of how asen- tence's given information can be identified, a more fundamental problem is whether givenness is a necess- ary and/or sufficient condition for topic-hood. In, for example, Reinhart (1981) it is argued that topics can also have new status and that given information need not be part of the topic. (On the given-newdistinction see, for example, Chafe 1976;Clark and Haviland
1977; Halliday 1967; Prince 1981.)
In Centering Theory the notion of sentence topic
is expressed by what is called the Backward-looking Center Cb of an utterance U, (Cb(Uj)). Cb(Uj) is a discourse entity evoked by Uj that is both contextually given and contextually preferred, implying that this entity was already introduced in the preceding utter- ance Uj_! and predicted to be the one Uj would be 'about.' As is the case with every utterance Uis the set of discourse entities associated with Uj_,, called the set of Forward-looking Centers of that utterance (Cf(Ui_,)), is a (partially) ordered set the ordering of which is language-specific, determined by various formal (syntactic, prosodic, etc.) characteristics of Uj_,. The highest-ranked element of this set is called the Preferred Center Cp of Uj_, (Cp(Ui_,)), which expresses a preference with respect to the topic of the next utterance Uj. As mentioned, the full set of factors responsible for such an ordering is still to be deter- mined. Backward-looking center and Forward-look- ing center correspond to Sidner's (1979) notion of current discourse focus and potential focus, respec- tively.
Although not explicitly part of the theory, the Alter- native Semantics approach suggests that the topic of a sentence is the set of alternatives induced by the focus part of that sentence. The alternatives are defined as propositions obtained by that which in the given context can be inserted into the associated focus frame. As others have noted too (see, in particular, Partee 1991 and Rooth 1992), there is a non-trivial (and probably fruitful) relation between the alter- native set associated with a sentence and Hamblin's (1973) notion of question meaning, formally analyzed as the set of propositions expressed by possible, direct answers to the question.
Characterizations in terms of questions identify the topic which is related to a question-answering sen-
tence with a variety of things. In one approach (Bart- sch 1976; Collingwood 1940; V ennemann 1975), the topic is identified with (one of) the presupposition(s) defined by the question. Others directly define the notion of topic in terms of questions and the set of possible ('alternative') answers they give rise to. This view is already central in the works of the British philosopher and (theoretical) psychologist G. F. Stout (e.g., Stout 1896, 1932): 'Questioning involves the thought of a set of incompatible alternatives. In asking a question we know what it is that we want to know in knowing that one or other of these alternatives is the right answer. But we do not know and have not decided, rightly or wrongly, which it is' (1932: 301). The set of incompatible alternatives is taken to be the topic ('psychological subject') of the question-answer- ing sentence. More recently, in the mid-1980s, Carlson (1985), in the tradition of Hamblin (1973), identifies a topic with question meaning, whereas Klein and Von Stutterheim (1987: 164) take as topic what they call the 'alternative,' which they define as 'the choice between two or more possibilities' as an answer to the question posed. Still others (Van Kuppevelt 1991) identify the topic of a sentence with that which is questioned, i.e. an underdetermined singular or plural discourse entity that needs further specification. This underdeterminedness is then expressed in terms of the corresponding (actual) topic range specifying the (remaining) set of possible extensional counterparts. For all the approaches in terms of questions, the cri- terion for a sentence to have a topic is that it answers a question. In this respect it is assumed that sentences in discourse can also answer 'implicit,' that is not explicitly formulated, questions. However, when a sentence answers an implicit question, topic identi- fication requires a reconstruction of the implicit ques- tion. To date, no fully satisfactory algorithm has been proposed that yields an unambiguous identification of implicit questions in discourse.
4. The Relation Between Sentence Topics and Dis- course Topics
Finally, attention is paid to the relation between sen- tence topics and discourse topics, thereby refraining from other relevant subjects in this area of research, such as the important question of whether the topic- comment distinction is a syntactic, semantic and/or pragmatic phenomenon and the discussion on focus- sensitive operators.
Those (relatively few)authors who distinguish sen- tence topics from discourse topics do not agree with regard to the question of the distinctness versus the continuity of the notions 'sentence topic' and 'dis- course topic.' In the context of discourse grammar, Van Dijk (1977), for example, assumes two notions which he defines in such a way that they are con- ceptually unrelated. A sentence topic is identified with an individual entity (or a set of entities or an ordered
Topic and Comment
195