The study used data for two North Carolina cohorts: all individuals released between July 1977 and June 1978 and those released between July 1979 and June 1980. Recidivism was determined by a search of State correctional records in April 1984, providing followup times ranging from 46 to 81 months. The analysis compared parametric and nonparametric statistical models, considered the importance of including individual characteristics as explanatory variables in such models, and examined the usefulness of 'split' models, in which not every individual would eventually return to prison. The dependent variable in the analysis was the time until return to prison in North Carolina. Results showed the usefulness of parametric statistical models. Purely nonparametrical models were of little use, however. Efforts to show the usefulness of individual characteristics as explanatory variables showed mixed results. Including a splitting parameter into a model greatly improved the ability to predict the timing of recidivism. Footnotes, tables, author index, subject index, and 84 references.