Data from a sample of severely and persistently mentally ill involuntary patients indicated that differences in violence between males and females in the 4 months prior to hospital admission depended on the measure. In the bivariate analysis, males had a greater prevalence of violence on the two indicators which separated more serious violence from lesser and no violence; but there was no gender difference on the more inclusive measure which incorporated threats and fights not involving weapons or injuries. In multivariate analysis when other relevant predictors were controlled, gender was significant in predicting only the most inclusive indicator of violence and only in interaction with substance abuse problems.