Application of stochastic process analysis to initial counseling interviews

The purpose of this study was to investigate whether a a finite, first-order Markov chain, provides an adequate model of how discourse is sequentially organized in twelve counseling interviews. Six counselors, two at each - of three experience levels were randomly assigned a male and a female client from 12 student volunteers. The clients met with a leader for a guided fantasy experience which was followed by a session with the assigned counselor. Each speaker turn from each session was categorized according to the Discourse Activity Analysis System (DAAS). DAAS is a system of categories which describes the function of a speaking turn in initiating and sustaining a topic and in managing the flow of talk. The categories of turn-taking behavior of each client/counselor dyad were used to generate transitional probability matrices in order to test for the Markovian assumptions of stationarity and first-order dependence. All matrices met the Markovian assumption of stationarity. None of the matrices met the assumption of first-order dependence. The discussion focuses upon limitations and implications related to (a) use of the DAAS, (b) the nature of the specific interviews used in the study, and (c) the effects that different ways of sampling discourse may have upon classifying speaker turns. Suggestions for future research are made.