THREE ALTERNATIVE PROCEDURES FOR MEASURING ASSOCIATION WERE EACH USED IN PARALLEL BASE EXPECTANCY ANALYSIS ON THE SAME DATA BASE. THE STUDY GROUP CONSISTS OF 929 NARCOTIC LAW VIOLATORS WHO WERE PAROLED IN 50 JURISDICTIONS. PAROLE PERFORMANCE WAS DERIVED AFTER A TWO-YEAR FOLLOW-UP PERIOD. THE 1968 PAROLEES WERE USED TO DEVELOP OFFENDER RISK RATINGS. THESE RATINGS WERE APPLIED TO 1969 PAROLEES TO DETERMINE WHETHER THEY WOULD BE ASSOCIATED WITH PAROLE OUTCOME IN A NEW SAMPLE. THE THREE METHODS WORKED ABOUT THE SAME FOR DEVELOPING OFFENDER RISK RATINGS. ONCE PERSONAL CHARACTERISTICS OF OFFENDERS WERE TAKEN INTO CONSIDERATION, THE NUMBER OF MONTHS SERVED HAD NO CONSISTENT RELATIONSHIP TO PAROLE OUTCOME. THIS REMAINED TRUE FOR ALL THREE PARALLEL TECHNQIUES USED. (AUTHOR ABSTRACT)