The findings indicate that offender characteristics, related to the probability that the offender would be re-arrested after release, was more significantly correlated to parole decisions than the seriousness of the offense itself. Three statistical techniques were used to predict recidivism within three years of release. The simple summation method was most effective in distinguishing between high and low risk inmates. This method was able to identify inmates unlikely to be rearrested for violent offenses, although it was less capable in identifying high risk inmates. The report recommends that guidelines be adopted by the National Parole Board in order to incorporate this predictive capability. Offenders classified as good statistical risks would be granted an operating presumption favoring full parole release at their initial date of eligibility while poor risk inmates would move into a program of graduated conditional releases. 16 tables, 6 figures, and 4 appendixes