The term 'violent behavior' -preferable to 'dangerousness' -- can be defined as acts characterized by the application or overt threat of force which is likely to result in injury to people. The prediction of violent behavior is currently used to assist in making a wide variety of legal decisions, from civil commitment to the imposition of the death penalty. It is also one of the more ethically and technically difficult tasks the legal system asks of mental health professionals. In clarifying these aspects, the monograph suggests several ways to improve clinical decisionmaking. Pure clinical prediction, often based on a clinician's experience and expertise and determined subjectively or intuitively, is less precise about predictor variables employed and may use different predictors for different cases. Incorporating statistical concepts is seen as one of the most promising avenues for improving the accuracy of clinical predictions. For example, statistical prediction, which uses lower order, often demographic, variables and combines them by means of automatic, mathematical rules, should be employed whenever possible. During evaluation, the clinician should make base rates of violence a prime consideration, obtain information on valid predictive relationships, and not overreact to positive associations. The accuracy of clinical prediction may also be improved through increased attention to situational or environmental predictors of violence. A model of some of these situational and environmental factors is presented and is based on interactions among the following: (1) stressful events (e.g., frustrations, annoyance, insults); (2) behavioral coping responses, including violent ones (e.g., murder, robbery, rape, assault) and nonviolent ones (e.g., withdrawal, avoidance); (3) affective reactions, such as predisposing ones (e.g., anger, hatred) and inhibiting ones (e.g., empathy, guilt, anxiety, fear); and (4) cognitive processes, including predisposing appraisals and expectations (e.g., violent fantasies and self-statements perceived intentionally) and inhibiting appraisals and expectations (e.g., nonviolent self-statements, expectations of punishment). A 14-item checklist provides a summary of this interaction for the clinician's use during evaluations of the threat of future violence. In addition, a case study, tabular data, and over 250 references are included.