A random sample of all persons released from Federal prisons by parole, mandatory release, or expiration of sentence during 1972 was drawn. The sample (n=2,127) was restricted to Federal prisoners with maximum sentences of more than 1 year and 1 day who were not violators of either Selective Service or Immigration and Naturalization Service laws. For each sample, information was obtained on personal characteristics, previous employment, criminal justice system characteristics, criminal history, and offense characteristics. The study found that increase in the severity of punishment is weakly related to participation in crime, while increases in the certainty of punishment are positively related to recidivism. Another study sample consisted of 432 males released in 1971-72 from Maryland's State prisons to the Baltimore area. All subjects had meager financial resources, were repeat offenders, had no known history of alcohol or drug abuse, and had not been on work release for more than 3 months. Although the Baltimore sample was very different from the Federal sample, the findings are comparable. The effects of punishment severity and certainty on recidivism were insignificant, while the effects of improved employment on recidivism were remarkable; better wages were found to reduce recidivism. Tabular data and six references are provided.