A justice with considerable experience in presiding over criminal trials commented that a good judge, through keen observation, develops an ability to sense dangerousness. A forensic psychiatrist emphasized that his role was usually to advise about the degree of disorder rather than predict future behavior. The parole decisionmaking process was discussed by a parole board member who said that both statistical and clinical predictions were considered. The moderator asked participants to comment on a recently published report which concluded that predictions about future offenses were hazardous, uncertain, and largely intuitive. A criminologist commented that previous convictions for violent offenses appeared to be a sound predictor, but the parole board member felt that possibly two-thirds of the supposedly dangerous criminals were in fact safe to be released. He also believed that this was the best the parole board could do, given existing research. A sociologist and a legal director of a mental health association used this statement and other research findings to demonstrate the unreliability of predictions. Finally, participants explored ways to improve prediction methods and whether preventive imprisonment went beyond appropriate retribution. Arguments centered on whether protection of the public justifies a system which makes a few mistakes by imprisoning individuals who are not dangerous; offenders' rights; and current public concern with violent crime which ignores injuries committed by drunk drivers, environmental polluters, and white-collar criminals.