A RAND report based on a 1978 survey of inmates in California, Michigan, and Texas suggested that crime might be reduced if scarce prison space was allocated to offenders most likely to recidivate. To assess the utility of a selective incapacitation policy, this research investigated criminal histories of the 1978 survey participants to determine recidivism rates and also applied the following recognized predictive models: Federal Salient Factor Score, Greenwood, California Base Expectancy Score, PABLO Scale, and IOWA. More than 50 percent of the inmates that were followed for at least 36 months were arrested after their release, and most were subsequently convicted and incarcerated. Some recidivists were arrested more than once. Rates differed among States, but this appeared to stem from differences in incarceration policies. In all three States, prior record was not a good predictor of recidivism. No models were adequate predictors of time-to-first arrest. All but one had very low, but statistically significant correlations with other indexes of recidivism, such as whether the inmate had any postrelease arrests. This is 5 to 10 percent more accurate than what would be obtained by chance. The analysis indicates that selective incapacitation policies are unlikely to substantially reduce crime. Tables, references, and descriptions of the research methodology.