The 207 participants were all referred to one of the authors by the Missouri State Board of Probation and Parole for pre-parole psychiatric evaluations between October 1982 and September 1984. Each evaluation consisted of a thorough review of probation and parole records and a semistructured personal interview. The data were studied using stepwise discriminant function analysis. Results revealed that diagnosis seemed to have a relatively minor role in the decision to release from parole. Only diagnoses of drug abuse were used by decisionmakers, who otherwise relied on traditional criminological variables. A surprising finding was that psychosis was the most powerful predictor of parole revocation. Nevertheless, the study's predictor variables are not suitable for prediction in an actuarial sense. Instead, these results should stimulate further research on parole selection and outcomes. Tables and 36 references