These characteristics were, in turn, related to the probability that the offender would be rearrested after release. Three statistical techniques were tested to determine if they could predict whether the offender would be rearrested within 3 years of release for any serious (indictable) offense. The simple summation method pioneered by E. W. Burgess proved to be the most effective way for distinguishing between 'high' and 'low' risk inmates. When tested for its ability to predict violent criminal activity after release, the method was unable to isolate 'high risk' inmates with precision although it was able to identify large numbers of persons who were extremely unlikely to be rearrested for violent offenses after release. The report proposes a set of guidelines for the systematic incorporation of this predictive technique into the decisionmaking processes of the National Parole Board. Offenders identified as 'good statistical risks' should be granted an operating 'presumption' in favor of full parole release at their initial date of eligibility, while 'poor risk' inmates should receive a 'presumption' against parole, but would be given priority status for a carefully planned program of graduated conditional releases. Tables, figures, footnotes, and about 90 references are supplied. Appendixes include information on the structure and powers of the National Parole Board, parole rates between 1959 and 1976, and materials related to general recidivism prediction and to violent recidivism prediction. (Author abstract modified)