Predictions can be stated in several ways. Dichotomous statements that a person is or is not dangerous are clearly ethically questionable. However, predictive testimony in the form of probabilistic or comparative risk statements may be based on reliable research findings. Although clinicians have sometimes failed to abide by accepted principles for applying an empirical base in individual assessments, these failures do not render all violence predictions unethical. Furthermore, the potential for a significant number of false-positive legal decisions does not necessarily make predictive testimony ethically improper, because society does not require certainty about future violence to restrict certain liberties. Proponents of a ban on all predictive testimony by mental health professionals may have identified circumstances in which predictive statements cannot be considered ethical, but this conclusion cannot be generalized to all other contexts. Therefore, psychology, psychiatry, and the courts should not conclude that all predictions of dangerousness are unethical. Footnotes and 37 references (Author abstract modified)