CONSIDERABLE CONFUSION EXISTS ABOUT ACTUAL RECIDIVISM RATES BECAUSE ANY NUMBER OF CHANGES CAN AFFECT THE ACCURACY OF CALCULATING SUCH RATES; E.G., A POLICY OF LONGER SENTENCES MAY MAKE RECIDIVISM RATES APPEAR TO DECLINE, WHILE A POLICY OF PAROLE FOR GOOD BEHAVIOR MAY MAKE THEM APPEAR TO RISE. FURTHERMORE, WHEN PAROLE FILE INFORMATION IS USED AS THE DATA SOURCE, THE FOLLOW-UP PERIOD MAY BE A GIVEN NUMBER OF YEARS FROM RELEASE DATE OR THE PERIOD OF SUPERVISION, SO THAT A SUBSTANTIAL NUMBER OF CASES WILL HAVE BRIEF FOLLOW-UPS. IN THE PRESENT STUDY, DATA FROM A RANDOM SAMPLE OF 1,806 PRISONERS RELEASED IN THE FIRST 6 MONTHS OF 1970 AND FOLLOWED UP FOR 6 YEARS ARE ANALYZED USING 4 DIFFERENT CRITERION MEASURES FOR RENEWED INVOLVEMENT WITH THE LAW FOLLOW-UP PERIODS. THE RATE OF RECIDIVISM REPORTED COULD VARY FROM 8.7 PERCENT (FOR THE CRITERION PRISON COMMITMENT AND A 4-YEAR FOLLOWUP TO 60.4 PERCENT (FOR THE CRITERION ARREST AND A 6-YEAR FOLLOWUP). UNDER THESE CIRCUMSTANCES, RESULTS OF VARIOUS RESEARCH EFFORTS CAN BE COMPARED ONLY IF METHODOLOGICAL CHOICES ARE CLEARLY STATED. MOREOVER, STANDARDIZATION OF BOTH CRITERION MEASURES AND FOLLOWUP PERIODS SHOULD BE ENCOURAGED IN THE RESEARCH COMMUNITY. ALTHOUGH NOT ALL PROBLEMS INHERENT IN RECIDIVISM RESEARCH ARE AVOIDABLE CLEAR DEFINITIONS OF METHODOLOGY CAN MAKE RECIDIVISM STUDIES USEFUL FOR PROGRAM EFFECTIVENESS EVALUATION AND SOCIAL POLICY DETERMINATION. NOTES, TABLES, 10 REFERENCES, AND AN APPENDIX ARE SUPPLIED.