The subjects were 149 current and former male patients in the forensics division of a mental hospital who had committed at least 2 violent crimes (79 patients) or no violent crimes (70 patients). Data were obtained from patient files; univariate analyses showed that the factors of education, occupation, age at first institutionalization, juvenile delinquency, absence of father, and the length of time institutionalized before the age of 18 distinguished significantly between the violent and nonviolent groups. Factor analysis was used to reduce the number of predictor variables; those which emerged were juvenile criminal history, parental absence, socioeconomic status, and family experience. A stepwise discriminant analysis using these reduced factors was then performed on each of two samples which were formed by rndomly splitting the violent and nonviolent groups. All four factors emerged in both discriminant functions. Analyses demonstrated the significance of each factor in assigning the subjects within the same sample and subjects in the other sample to the proper classificaiton. The findings should provide clues as to what variables would be worthwhile to examine in etiology studies using a followup design. The results from such studies could help predict future violent behavior. Related research is reviewed. Data tables and 15 references are included. (Author abstract modified.)