Researchers have argued that the standard measurement of recidivism usually forces a dichotomous choice in evaluating an offender's behavior -- success or failure. Others feel that the value of the recidivism measure is limited by the variety of definitions and followup time periods used by researchers. Despite the volume of recidivism studies available, few conclusions can be reached that aid the contemporary correctional administrator. To examine recidivism patterns, 200 juveniles under supervision of a large metropolitan probation department were tracked for 2 years from the time of entry into the caseload during the mid 1970's. The probationers were all middle- or upper middle-class males, but all had at least one prior arrest and many were multiple offenders who had been referred previously to diversion programs. A wide variety of delinquent acts was represented, including chronic school truancy, drunkenness, robbery, and manslaughter. The probation officers were experienced with excellent professional credentials and carried an average monthly caseload of 60 delinquents. Case methods employed included family and individual counseling and liberal use of a volunteer big brother program. The probation officer met the probationer each month, and most boys were visited twice a month. Followup data were collected from adult probation records, county jail bookings, and the prosecutor's office. Recidivism was defined as being charged with new criminal offenses or delinquent acts, violating probation conditions, or absconding while on probation. On this basis, 35 percent of the delinquents were recidivists. No juvenile recidivated more than once, and 86 percent repeated the same offense that caused the probationary period. Most recidivists were unemployed, had an unstable home life, and experienced problems in school. These findings suggest the need for earlier intensive intervention in the correctional process which could be achieved by smaller, specialized caseloads. The relationship of recidivism measures to an agency's goals are also discussed. Tables and 32 footnotes are provided.