Of the 12 intervention sessions, 5 focused on cognitive restructuring and self-instructional training, 5 sessions emphasized conflict resolution and appropriate assertive communications skills, and 2 emphasized the development of active-coping relaxation and time-out skills. All of those participating in the study completed all the sessions. Following program completion, the participating couples were contacted four times over 1 year to determine any recurrence of violence. Any report of recurrence from either partner at a minimal level of push, shove, or grab was considered recidivism. Chi-square analyses on categorical data and analyses of variance on personality test data identified differences between recidivists and nonrecidivists. Recidivists reported higher levels of substance abuse both before and after treatment, and they also evidenced higher narcissism. Referral source (self or court) did not differentiate the two groups nor did record of criminal activity. Subsequent discriminant-function analysis, entering all predicted variables, correctly identified 65.4 percent of the recidivists and 73.1 percent of the nonrecidivists. Clinical and research implications of the findings are discussed. 9 tables and 21 references