Using Markov Chain Analyses in Counselor Education Research
Duys, David K., Headrick, Todd C., Counselor Education and Supervision
This study examined the efficacy of an infrequently used statistical analysis in counselor education research. A Markov chain analysis was used to examine hypothesized differences between students' use of counseling skills in an introductory course. Thirty graduate students participated in the study. Independent raters identified the microskills students used and the effectiveness of participants' counseling sessions. Significant differences were found in the counseling processes of students who were rated as effective and ineffective using a Markov chain analysis. Implications for future research using Markovian analyses in counselor education research are explored.
Counselor educators have studied the processes associated with effective and ineffective counseling exchanges for many years (Chwalisz, 2001; Hubble, Duncan, & Miller, 1999; Sexton, 1996). Efforts have been made to identify how counselors use verbal skills (Sipps, Sugden, & Faiver, 1988), conceptualization skills (Ladany, Marotta, & Muse-Burke, 2001), interactive behaviors (Roll, Crowley, & Rappl, 1985), and power in counseling dyads (Cummings, 2000). Findings from related studies have influenced the way that counseling courses are taught (Woodard & Lin, 1999), methods used in supervision (Granello, 2002), and the development of theoretical models that attempt to explain and predict counselor development (Blocher, 1983; Stoltenberg, McNeill, & Delworth, 1998).
Researchers have attempted to examine dynamics in counseling relationships using different research designs (Paulson, Truscott, & Stuart, 1999); by far, the most common designs have been quantitative in nature. However, some quantitative designs have been limited to the use of frequency counts of desirable counseling behaviors without being able to simultaneously measure the processes of counseling. Although observing frequencies provides helpful information about ratios and distributions of variables, information about processes or transitions between variables has not been addressed in these designs.
Markov Chain Analysis
A Markov chain analysis has been an underutilized method of quantifying the counseling process. A primary advantage of this methodology is that it allows researchers to compare and contrast counseling processes by analyzing transitions between events. The identification of transition patterns provides a theoretical model for, and a predictor of, the sequences that occur in counseling. In other words, this analysis can reveal the probabilities that a counseling session will shift between different states of activity. The states of interest can be any measurable event, such as narrative content, nonverbal responses, interventions, or shifts in power dynamics.
Previous studies (e.g., Lichtenberg & Heck, 1986; Tracey, 1985) usually credit Hertel (1972) for stimulating researchers to examine Markovian models in the context of counseling relationships. In each case, researchers identified different states of interest to describe events in the counseling process. For example, Lichtenberg and Heck examined the way that counselors and clients establish their respective roles over time. In another study, Tracey used a Markov chain to detect measurable shifts between beginning and working stages of a counseling relationship in the context of a single-subject design. Examining gestalt dynamics in counseling sessions, Benjamin (1979) measured transitions between states associated with interpersonal and intrapersonal power dynamics.
A Markov analysis of a counseling process begins by identifying events of interest (states) that are of concern to the researcher or researchers. Probabilities are calculated that one state will follow another by dividing the number of times a transition occurs between two of the states (e.g., a counselor follows a question with a reflection) by the total number of transitions between any of the states identified (e. …