At the time of admission, subjects completed extensive social history interviews exploring their family background, educational and work history, prior criminal contacts, and situation immediately prior to incarceration. Psychometric data were collected on the Minnesota Multiphasic Personality Inventory (MMPI) and 14 commonly used experimental scales. The criterion used for recidivism was a reconviction at 2 years after release from the facility. Recidivism predictions were made on the basis of (1) social histories, (2) psychometric variables, and (3) social histories and psychometric variables combined, and compared to the actual recidivism data. Statistical analysis (step-wise multiple regression using a forward selection technique) indicated that social history variables predicted recidivism better than psychometric variables. Moreover, upon cross-validation, less shrinkage in prediction was found for the social history variables as compared to the psychometric data, indicating superior reliability and stability in the social history data. Furthermore, when the psychological test data and social history data were combined, the regression correlation with recidivism was even higher. Thus, any correctional administration interested in obtaining rough predictive estimates of recidivism and having limited resources should utilize social history information. The study includes statistical tables and 34 references. A summary in French is provided.