Ten years ago the law simply assumed that psychologists or psychiatrists were accurate enough in their predictions to be useful in forming mental health and criminal justice policy. However, the first generation of research on the prediction of violent behavior seemed to show that psychiatrists and psychologists were vastly overrated as predictors of violence. This research was one of the factors underlying the move to abolish indeterminate prison sentences and to use retribution as the guiding principle for punishment. In recent years American policy regarding the prediction of violent behavior has entered a third phase. Although no major breakthroughs have occurred in prediction research, the view that prediction is here to stay is now accepted. One reason for this change is the reevaluation of the earlier empirical studies. Another is the pragmatic fact that nothing has replaced predictions as the basis for public policy predictions. Thus, prediction of violence is becoming increasingly influential in the sentencing and release of criminal offenders and in tort law relating to the actions of mentally disordered individuals. Figure, tables, and 17 references.