At a community-based domestic violence program, 218 men with a history of partner abuse were randomly assigned to either feminist-cognitive-behavioral or process-psychodynamic group treatments. The treatments were not hypothesized to differ in outcome. However, men with particular characteristics were expected to have lower recidivism rates depending on the type of treatment received. Treatment integrity was verified through audio-taped codings of each session. The partners of 79% of the 136 treatment completers gave reports of the men's behavior an average of 2 years post-treatment. These reports were supplemented with arrest records and self-reports. Rates of violence did not differ significantly between the two types of treatment nor did reports from the women of their fear level, general changes perceived in the men, and conflict resolution methods. However, interaction effects were found between some offender traits and the two treatments. As predicted, men with dependent personalities had better outcomes in the process-psychodynamic groups and those with antisocial traits had better outcomes in the cognitive-behavioral groups. The results suggest that more effective treatment may occur if it is tailored to specific characteristics of offenders.