STUDY DATA WERE DRAWN FROM THE PAROLE RECORDS OF 455 ADULT MALE OFFENDERS RELEASED FROM FEDERAL PRISON IN WASHINGTON STATE PRIOR TO JANUARY 1, 1950. APPLIED TO THESE DATA, THE DISCRIMINANT FUNCTION METHOD REDUCED THE ERROR OF PREDICTION OVER HALF AGAIN AS MUCH AS DID THE BURGESS METHOD. THE STUDY REPORT OFFERS A CRITICISM AND EVALUATION OF TRADITIONAL PAROLE PREDICTION METHODS (BURGESS, GLUECK, OHLIN). ASSUMPTIONS OF PAROLE PREDICTION -- STABILITY IN TIME, UNBIASED SAMPLE, PROPORTIONAL WEIGHTS OF VARIABLES AND OF SUBCLASSES OF VARIABLES, MUTUAL INDEPENDENCE OF VARIABLES, THAT VARIABLES ARE EXHAUSTIVE, NORMALITY OF DISTRIBUTION OF SUBCLASS VALUES, CONTINUITY OF VARIABLES, LINEARITY OF RELATIONS BETWEEN VARIABLES AND CRITERION (PAROLE VIOLATION) -- ARE REVIEWED. STATISTICAL DETAILS OF THE APPLICATIONS OF THE DISCRIMINANT FUNCTION TECHNIQUE AND THE BURGESS METHOD TO THE WASHINGTON CASE DATA ARE PROVIDED, TOGETHER WITH A TENTATIVE EVALUATION OF PAROLE PREDICTION STATE-OF-THE-ART AND SUGGESTIONS FOR FURTHER RESEARCH. A 12-PAGE BIBLIOGRAPHY IS INCLUDED. (LKM)