TWO PROPORTIONAL SAMPLES OF 113 RESIDENTS (68 SUCCESSES AND 45 FAILURES) WERE SELECTED RANDOMLY FROM THE STUDY POPULATION OF 838 SUCCESSES AND 523 FAILURES. CROSS-VALIDATED MULTIPLE REGRESSION ANALYSES SHOW EMPLOYMENT STATUS WHILE AT THE CENTERS TO BE THE MOST POWERFUL PREDICTOR OF SUCCESS FOR THE SAMPLE GROUP. SUBJECTS MOST LIKELY TO BE EMPLOYED WHILE AT THE CENTERS WERE OLDER, WHITE OFFENDERS WHO HAD BEEN CONFINED FOR ASSAULTIVE CRIMES AND WHO HAD PRIOR MISDEMEANANT AND EMPLOYMENT RECORDS. CANONICAL ANALYSIS, A METHOD FOR SHOWING THE DEGREE TO WHICH TWO SETS OF LINEAR VARIABLES CORRELATE, YIELDS PROFILES OF FOUR GROUPS: SUBJECTS WHO WERE EMPLOYED AND SUCCESSFUL AT THE CENTER; SUBJECTS WHO WERE UNEMPLOYED AND UNSUCCESSFUL; SUBJECTS WHO WERE UNEMPLOYED BUT SUCCESSFUL; AND SUBJECTS WHO WERE EMPLOYED BUT UNSUCCESSFUL. THE PROFILES SUGGEST THAT NOT EVERYONE BENEFITS EQUALLY FROM THE PREPAROLE PROGRAM. THE POSSIBILITY OF ESTABLISHING PROGRAMS AT DIFFERENT CENTERS TO DEAL WITH GROUPS WITH DIFFERENT PROBLEMS IS RAISED. DIRECTIONS FOR FURTHER RESEARCH ARE SUGGESTED. SUPPORTING DATA AND A LIST OF REFERENCES ARE INCLUDED. (AUTHOR ABSTRACT MODIFIED --LKM)