The present study is a follow-up of a sample of 348 men convicted of manslaughter, attempted manslaughter, or arson who were released from incarceration. Multiple factors assessed at the time of incarceration, including demographic, behavioral, family history, and biochemical variables, and psychiatric diagnoses were used in an attempt to discriminate between those who became recidivists during the follow-up period and those who did not. Violent recidivism was most strongly associated (sensitivity of 90%) with impulsivity of the original crime in killers and attempted killers; for arsonists, having made a suicide attempt was the strongest predictor (68% sensitivity). For predictive purposes, both single factor associations and multiple entries into discriminant analysis produced too many false-positives, i.e., the high rate of false designation as recidivist remained a problem.