An overview of research focused on the prediction of dangerous behavior among criminals and mental health patients is presented. Studies (N=3) conducted by the Bureau of Special Projects Research, New York Department of Mental Hygiene are described, as well as studies providing evidence on clinical and statistical predictions. The findings of these studies address the following areas of concern: (1) the inability of clinicians to achieve a level of prediction which exceeds the level obtainable strictly by chance; (2) the vast overprediction of dangerous behavior; and (3) the superiority of statistical predictions in determining the potential extent of dangerous behavior. (Author/HLM)