THE MODEL, WHICH EMPLOYS MULTIPLE DISCRIMINANT ANALYSIS, TAKES INTO ACCOUNT THE FOLLOWING VARIABLES: INCARCERATION DURATION; AGE UPON RELEASE; SEX; ADMISSION TYPE (PROBATION OR PAROLE VIOLATOR, OR NEW COURT COMMITMENT); OFFENSE TYPE (PROPERTY OR PERSONAL); SENTENCE TYPE (SIMPLE OR MULTIPLE); NUMBER OF PRIOR PRISON COMMITMENTS; NUMBER OF PRIOR NONPRISON SENTENCES; DRUG USE; AND ALCOHOL USE. RESULTS OF BOTH THE TEST AND VALIDATION APPLICATIONS SUGGEST THAT THE DISCRIMINANT MODEL IS USEFUL FOR CLASSIFYING PAROLEES AS GOOD OR POOR RISKS. THE MULTIVARIATE RESULTS PROVE SUPERIOR TO THE RESULTS OF UNIVARIATE ANALYSIS AND DEMONSTRATE THAT THE LATTER CAN BE MISLEADING. THE BEST UNIVARIATE DISCRIMINATOR -- OFFENSE TYPE -- RANKS SIXTH IN THE MULTIVARIATE MODEL. THE BEST MULTIVARIATE DISCRIMINATORS -- AGE UPON RELEASE AND NUMBER OF PRIOR PRISON COMMITMENTS -- RANK SEVENTH AND FIFTH, RESPECTIVELY, IN A UNIVARIATE FRAMEWORK. SUPPORTING DATA AND A LIST OF REFERENCES ARE INCLUDED. (LKM)