Conditions for Text Coherence

Abstract

The first question in attempting to define text-coherence is what it is exactly we want to define. As in many other areas of linguistics or textual analysis there are two stands on this issue: The first is that we want to define the set of all texts which have been produced or are producible by speakers and are generally comprehensible or functional in their context of utterances, and the second is that we want to define a set of norms or conditions which apply to a (perhaps idealized) subset of the set of comprehensible actual or producible texts those texts which observe maximally, or ideally, the requirements of rationality and cooperation, or that reflect the speaker's knowledge of what counts as a fully coherent text. I am taking the second stand. The conflict between the two approaches is an aspect of the familiar debate concerning the notions of norm and deviance. The aspect of this problem that has been widely discussed is the debate concerning grammaticality and semantic deviance, particularly in the treatment of metaphor. The central belief of those holding the first stand is that the distinction between grammatical and ungrammatical parallels, and is determined by, the distinction between meaningful or comprehensible and meaningless or incomprehensible. The second stand, taken by Chomsky (e.g., 1965), is that these are two distinct phenomena. Grammaticality is defined by the rules of the grammar, which are independent of meaning. Ungrammatical constructions can be meaningful and the converse is also possible grammatical constructions can be meaningless. Within this framework the difference between grammatically or semantically well-formed expressions and deviant linguistic expressions lies not in their comprehensibility, but in the types of operation required in the procedure of assigning meaning to them. While a grammatical expression is interpreted directly by a function mapping surface structures into semantic representations,