Approximately 450 violent offenders are referred to the program each year. The evaluation used a pretest/posttest nonequivalent group design for Study 1 (n=39) and Study 2 (n=50). Violent offenders who participated in the program were compared with a waiting-list control group on the dependent variables of anger knowledge, trait anger, anger expression, observed aggressive behavior, and prison misconduct. Differential treatment effects according to trait-anger level were examined in Study 2, anticipating greater gains for high trait-anger violent offenders. Data analyses provided little support for the hypothesized participant treatment gains relative to the control-group participants. Hypothesized differential treatment effect by trait-anger level was not supported. Failure to detect a significant treatment effect may reflect the complexity of the program. The program consists of anger management components that have been found effective in previous research with individuals with serious anger problems, including cognitive and relaxation training, assertive skills training, and examination of positive and negative consequences of angry/aggressive behavior. Howells (1996) has discussed the need for interventions with offender populations to be more extensive and intensive. The program evaluated may have been unrealistically brief, given the serious and engrained violent behavior of many participants. Further, participants in the program had been selected on the basis of their violent history rather than anger problems per se. Anger is not a necessary condition for violence. A more precise break-down of the characteristics of an offender's violent behavior may be necessary to determine which offenders are appropriate for anger management programs. 3 tables and 53 references