RESULTS OF RESEARCH ON THE DETERMINANTS OF CRIMINAL RECIDIVISM BASED ON DATA PROVIDED BY THE NORTH CAROLINA DEPARTMENT OF CORRECTION IS DESCRIBED. BASICALLY, INFORMATION ON ALL 4881 INDIVIDUALS RELEASED DURING THE FIRST SIX MONTHS OF 1975 WAS USED TO DEVELOP TWO PREDICTIVE MODELS. THESE WERE VALIDATED WITH FOLLOW UP DATA FROM DEPARTMENT OF CORRECTIONS RECORDS IN FEBRUARY 1977. THE PREDICTIVE ABILITY OF THESE MODELS WAS THEN TESTED ON A SECOND SAMPLE OF EX-OFFENDERS AND RESULTS COMPARED WITH ACTUAL FOLLOW UP. THE DEVELOPMENT OF THE TWO MODELS IS DESCRIBED IN DETAIL. THE FIRST USED A VARIABLE CALLED TTSENT, DEFINED AS THE TOTAL TIME (IN MONTHS) OF ALL NORTH CAROLINA PRISON SENTENCES RECEIVED BY AN INDIVIDUAL DURING HIS FOLLOW UP PERIOD. THE SECOND, CALLED LTFPCV, MEASURED THE LENGTH OF TIME (IN MONTHS) FROM RELEASE UNTIL RETURN TO PRISON IN NORTH CAROLINA. RESULTS INDICATE THAT THE INDIVIDUAL MOST LIKELY TO HAVE A LARGE VALUE OF TTSENT IS A YOUNG, BLACK MALE ALCOHOLIC WITH MANY PREVIOUS CONVICTIONS, WHOSE RELEASE WAS UNSUPERVISED, AND WHOSE CRIME AT THE TIME OF THE STUDY WAS AGAINST PROPERTY. THE INDIVIDUAL LIKELY TO RETURN TO PRISON THE SOONEST WAS YOUNG, SINGLE, UNEDUCATED, WITH MANY PREVIOUS CONVICTIONS AND RULE VIOLATIONS, AND WHOSE CRIME AT THE TIME OF THE STUDY WAS NOT AGAINST A PERSON. FURTHER REFINEMENTS FOUND THAT THE BEST PREDICTIVE MODELS FOR FEMALES AND NONBLACK MALE ADULTS WERE OBTAINED BY USING DATA FROM THE PARTICULAR SUBGROUP ONLY, WHILE FOR OTHER GROUPS THE BEST MODEL WAS THE OVERALL ONE. THE MODEL WAS FIRST APPLIED TO RECIDIVIST BEHAVIOR OF INDIVIDUALS CONVICTED OF FELONIOUS BREAKING AND ENTERING, FELONIOUS DRUG OFFENSES, AND FELONIOUS ASSAULTS. PREDICTIVE ACCURACY HAS BEEN WITHIN 10 PERCENT. STUDY WITH THE MODELS SUGGESTS THAT THE OVERALL RATE OF RECIDIVISM IN NORTH CAROLINA MAY INCREASE SINCE THE HIGHEST RATE OF RECIDIVISM IS FOR FELONIOUS BREAKING AND ENTERING (WHOSE NUMBERS ARE INCREASING) WHILE THE LOWEST RATE OF RECIDIVISM IS FOR FELONIOUS ASSAULT (WHOSE NUMBERS ARE DECREASING). ALSO, THE AVERAGE LENGTH OF SENTENCE FOR ENTERING INMATES CAN BE EXPECTED TO RISE, SINCE THE AVERAGE SENTENCE OF THOSE CONVICTED OF FELONIOUS BREAKING AND ENTERING AND FELONIOUS DRUG OFFENSES (THE TWO OFFENSES INCREASING MOST RAPIDLY) IS FAR HIGHER THAN THAT FOR FELONIOUS ASSAULT. THE MODELS CAN ALSO BE USED TO STUDY THE RESULTS OF TREATMENT PROGRAMS. IF SIGNIFICANT CHANGES ARE FOUND IN RECIDIVISM RATES, THE TREATMENT PROGRAM WOULD PROBABLY BE A FACTOR. (GLR)