To overcome the problems that have so far impeded the scientific study of violence among the mentally disordered, researchers must enrich predictor variables, strengthen criterion variables, broaden subject sampling strategy, and synchronize research efforts. Such measures will rejuvenate the field of risk assessment as it yields results different from those thus far produced. If an actuarially valid array of risk markers for violence could be reliably identified, clinicians could be trained to incorporate these factors into their routine practice, and the accuracy of clinical predictions of violence among the mentally disordered would be increased. Such an increase in predictive accuracy would not obviate the issues of social policy or professional ethics that attend any preventive use of the state's police power. It would mean, however, that relatively fewer people would be erroneously institutionalized as "dangerous" and that relatively fewer people in the community would be victimized by patients erroneously released or left untreated as "nondangerous." 1 figure and 80 references