Conversational Exchange Analysis

Abstract

In recognition of the deficiencies of the speech classification systems available for the analysis of informal conversation, the present article is concerned with the development, and subsequent description of a comprehensive speech analysis system. Developed from a review of the literature and the analysis of tape-recordings and transcripts, Conversational Exchange Analysis (CEA) comprises four sets of rules for the division and subsequent classification of speech. Initially segmenting conversational speech into units representing individual ideas, CEA is then used to classify speech along three separate dimensions representing how information is made salient in the conversation (Activity), the sort of information exchanged (Type), and the referent of the utterance (Focus). Measures of inter-observer agreement, using Cohen's (1960) kappa (Activity, kappa = 0.969 (p<0.001); Type, kappa = 0.957 (p<0.001); Subject focus, kappa = 0.994 (p<0.001); Object focus, kappa = 0.981 (p<0.001) ), indicate a high degree of reliability for category assignment.