THE DATA USED TO ESTIMATE THE MODELS DEVELOPED WAS INFORMATION ON ALL PERSONS RELEASED FROM THE NORTH CAROLINA DEPARTMENT OF CORRECTION DURING THE FIRST 6 MONTHS OF 1975. INFORMATION ON PERSONAL CHARACTERISTICS WAS OBTAINED FROM COMPUTERIZED INMATE HISTORIES, WITH POSTRELEASE CRIMINAL ACTIVITY DETERMINED FROM A SEARCH OF DEPARTMENT OF CORRECTION RECORDS IN FEBRUARY 1977. THE DEPENDENT VARIABLE IN THE MODELS WAS THE LENGTH OF TIME AFTER RELEASE UNTIL REIMPRISONMENT (LTFPCV). FROM AN EXAMINATION OF PAST RESEARCH ON THE DETERMINANTS OF CRIMINAL RECIDIVISM, IT WAS HYPOTHESIZED THAT THE MEAN OF LTFPCV WAS A LINEAR FUNCTION OF 14 INDEPENDENT VARIABLES. TO CHECK ON THE PREDICTIVE ACCURACY OF THE MODELS, THE SAMPLE OF 2,216 WAS RANDOMLY DIVIDED INTO 2 GROUPS. THE FIRST GROUP, CONTAINING 1,616, WAS USED TO ESTIMATE THE MODELS AND WAS THE ESTIMATION SAMPLE. THE SECOND GROUP, COMPOSED OF 600 SUBJECTS, WAS USED TO TEST THE PREDICTIVE ACCURACY OF THE MODELS. RESULTS OF THE FINAL SPECIFICATION INDICATE THAT THE TYPE OF EX-INMATE LIKELY TO RETURN TO PRISON SOONEST IS YOUNG, SINGLE, UNEDUCATED, AND IMPRISONED FOR A CRIME OTHER THAN AGAINST A PERSON, AND WITH MANY PREVIOUS CONVICTIONS AND RULE VIOLATIONS. THE MODEL WAS TESTED BY USING IT FOR PREDICTIONS ABOUT SUBJECTS IN THE VALIDATION SAMPLE. THERE WAS AN OVERPREDICTION OF A STATISTICALLY INSIGNIFICANT 5.7 PERCENT. THE MODELS DEVELOPED MAY BE USED TO PREDICT EXPECTED PRISON POPULATIONS AND IMPROVE CORRECTIONAL PROGRAM EVALUATION. AS AN ILLUSTRATION OF THE USE OF THE MODELS, AN INNOVATIVE VOCATIONAL EVALUATION PROGRAM FOR JUVENILE OFFENDERS WAS EVALUATED. THE MODEL PREDICTED SIGNIFICANTLY HIGHER RATES OF RECIDIVISM THAN ACTUALLY OCCURRED PARTICIPANTS FOR THE SECOND, THIRD, FIFTH, AND SIXTH MONTHS AFTER RELEASE. THERE WERE NO SIGNIFICANT DIFFERENCES BETWEEN ACTUAL AND PREDICTED RATES OF RECIDIVISM BEYOND 6 MONTHS. RESULTS INDICATE THE EFFECTIVENESS OF THE PROGRAM OVER THE SHORT TERM AFTER RELEASE AND POINT UP THE NEED FOR A LONG-TERM FOLLOWUP PROGRAM. TABULAR DATA, MATHEMATICAL EQUATIONS, AND REFERENCES ARE PROVIDED. (RCB)