一貫性の多様な次元 : 日常会話と思考の「自然な流れ」の解明に向けて

Abstract

In contrast to the dominant tendency in cognitive psychology, it seems intuitive that human thinking and natural conversation are dependent on each other. Both are composed of segments connected with each other and both construct flow. This article focuses on so-called 'linguistic coherence' from several different points of view. This feature of discourse seems to relate strongly to these elements common to thinking and conversation (sec.1). Hitherto, coherence has been examined predominantly in the domain of discourse analysis. Most of these analyses, however, have remained static and rather plain, describing the relations between segments that analysts have found eχ post facto in already fixed and determined texts. As a result even some currently widely used taxonomies fail to distinguish between quite different levels of coherence relations and to consider interdependency among these levels. Therefore, this article considers how these difficulties might be addressed. Of central importance is the distinction between semantic and pragmatic relations of coherence. This distinction can be understood as one separating the interaction level from the order of the described world. Most researchers in discourse analysis, however, have examined only the former level of coherence (sec.2). While philosophical consideration of speech acts has drawn our attention to the performative aspect of utterances, most research from this perspective concentrates only on the analysis of individual utterances, not on sequences. This article emphasizes the importance of the interaction level of coherence in constructing discourse using such formulations about conversation structure developed in the domain of conversation analysis as adjacency pairs and story structures. The merit of analytic tools formalized in this academic field is that they treat sequences of utterances primarily as sequences of actions and put emphasis on the orientation of participants toward the on-going conversation. It becomes clear that even apparently purely semantic coherence can be realized only in the frame of the interaction level of coherence (sec.3). Coversation analysis also has its weakness, however. To examine the flow of thought from this interactional point of view in the next stage, not only conversation structure but also the form of inference which has sensitivity to, and enables the flexibility of, interaction must be made explicit in a fashion adequate for cognitive science. Relevance theory has started to consider inference-making during the understanding of utterances, but it remains unclear how listeners construct explicatures from surface utterances and how and what kind of structure of background knowledge they recall and use to derive implicatures. It is argued that the form of inference needed for understanding utterances in interaction must be dynamically and flexibly adjusted to the dimention of the utterance upon which the focus of understanding should be put. Finally, a hypothetical form of inference is advocated (sec.4).