The study hypothesized that offenders serving longer prison sentences would tend to commit subsequent offenses less severe than their previous offenses. Data were obtained from the 1974 and 1979 surveys of inmates of State correctional facilities. The 1974 survey was based on a national sample of 1,670 male offenders last released from a State prison in 1970-74. The 1979 survey was based on a national sample of 934 male offenders last released from a State prison in 1975-79. Recidivism was defined as return to a State institution; the analysis focused on recidivists incarcerated at the time of each survey. Differences were noted for the time served for similar offenses between the two samples. These are attributed to population pressures and sentencing policies during the survey periods. Prior incarceration offenses were categorized as violent, drug trafficking, property, drug possession, and public order. Current incarceration offense severity was categorized as violent or nonviolent. For each prior incarceration offense category, the mean time that had been served by those returned for a violent offense was compared to the mean time served by those returned for a nonviolent offense. No evidence suggests that time served had any influence on subsequent offense severity. Thus, the analysis does not support the theory that varying sentence severity according to crime severity achieves specific deterrence. Tabular data and 34 references.