THE MODEL WAS DERIVED AS PART OF A SURVEY OF 36 STUDIES ON PROBATION/PAROLE AGENCIES. IT WAS CONSTRUCTED BY COMBINING PAROLE/PROBATION REVOCATION WITH OFFENDER AND OFFENSE CHARACTERISTICS. SIX CATEGORIES FOR DESCRIBING CHARACTERISTICS OF OFFENDERS WHO HAVE OR HAVE NOT HAD THEIR PROBATION/PAROLE REVOKED WERE IDENTIFIED: NO OFFENSE (D), NONSEVERE OFFENSE (E), SEVERE OFFENSE (F), NO PRIOR NONSEVERE OFFENSE (G), NO PRIOR SEVERE OFFENSE (H), AND PRIOR SEVERE OR NONSEVERE OFFENSE (I). USING THESE CATEGORIES, SEVEN DIMENSIONS OF AGENCY FIELD SUPERVISION CAN BE DEFINED. THUS SURVEILLANCE IS DEFINED IN TERMS OF THE PROPORTION OF SUBJECTS WHO ARE DEEMED TO HAVE COMMITTED AN OFFENSE, SUCH THAT SURVEILLANCE = E+F+G+H+I/TOTAL SUBJECT POPULATION. TOLERANCE IS DEFINED AS E+F/E+F+G+H+I, AND TREATMENT MAY BE DEFINED AS D/D+E+F. GENERAL DETERRENCE IS COMPUTED BY THE PROPORTION OF THE TOTAL SUBJECT POPULATION IN CATEGORY D (E.G., D/N). COST OF DETERRENCE IS DEFINED AS G+H+I/E+F+G+H+I. THE COST OF PUNISHING OFFENDERS IS DEFINED AS (G+H+I)-D/TOTAL SUBJECT POPULATION, AND THE COST OF REWARDING NONOFFENDERS IS DEFINED AS (E+F)-D/TOTAL SUBJECT POPULATION. APPLICATION OF THIS MODEL FOR EVALUATING THE DECISION OUTCOMES FOR CONTROL AND EXPERIMENTAL GROUPS OF THE CALIFORNIA COMMUNITY TREATMENT PROJECT IS ILLUSTRATED. IN THE CONTROL SAMPLE, (ADOLESCENT OFFENDERS IN CASELOADS OF 50 TO 70), A HIGH SURVEILLANCE RATE IS ASSOCIATED WITH LOW TOLERANCE, WHICH IS IN TURN ASSOCIATED WITH HIGH TREATMENT. THIS PATTERN IS FURTHER ASSOCIATED WITH LOW GENERAL DETERRENCE, HIGH COST OF DETERRENCE, HIGH COST OF PUNISHING OFFENDERS, AND LOW COST OF REWARDING OFFENDERS. IN THE EXPERIMENTAL GROUP (ADOLESCENT OFFENDERS), HIGH SURVEILLANCE IS ASSOCIATED WITH HIGH TOLERANCE AND LOW TREATMENT. THUS, WHEN AN OVERALL ASSESSMENT IS MADE OF THE DIFFERENTIAL EFFECTIVENESS OF THE EXPERIMENTAL AND CONTROL SUPERVISION PRACTICES, IT IS CLEAR THAT THE CONTROL PROGRAM (INSTITUTIONALIZATION PLUS STANDARD PAROLE) ENDS UP WITH MORE SUBJECTS IN THE NONOFFENDER CATEGORY, WHILE THE EXPERIMENTAL PROGRAM ENDS UP WITH MORE SUBJECTS IN THE OFFENDER/NONREVOKED CATEGORY. THIS SUGGESTS THAT IN TERMS OF COMMUNITY COSTS AND VICTIMIZATION, THE EXPERIMENTAL PROGRAM IS MORE COSTLY THAN THE CONTROL PROGRAM. IF EACH OF THE COMPONENTS OF THE MODEL IS MEASURED INDEPENDENTLY, THE MODEL SHOULD BE USEFUL IN PREDICTING THE OUTCOMES OF VARIOUS CHANGES IN THE SYSTEMS OF PROBATION, PAROLE, AND COMMUNITY SUPERVISION AND TREATMENT. REFERENCES ARE PROVIDED. (JAP)