IN EXAMINING METHODOLOGICAL ISSUES, THE TWO BASIC APPROACHES FOR PREDICTING PROBATION OUTCOMES, CLINICAL AND STATISTICAL ANALYSES, ARE COMPARED, AND EVIDENCE IS CITED SHOWING THAT PREDICTIONS BASED ON STATISTICAL ANALYSES ARE EQUAL OR SUPERIOR TO CLINICIAN JUDGMENTS. PREDICTION METHODS BLENDING THE VALUES OF BOTH CLINICAL AND STATISTICAL FINDINGS ARE DISCUSSED. NOTING THE WEAKNESSES IN EVALUATING PREDICTION METHODS, IT IS INDICATED THAT IN THE STUDIES REVIEWED SAMPLES WERE NOT REPRESENTATIVE OF THE POPULATION ABOUT WHICH INFORMATION WAS BEING SOUGHT. THE SIZE OF SAMPLES IS ALSO DEEMED TOO SMALL IN MANY STUDIES. BASE RATES, OR THE NUMBER OF 'SUCCESSFUL' PROBATIONERS RELATIVE TO THE TOTAL NUMBER UNDER STUDY, IS NOT REPORTED IN MOST STUDIES, MAKING IT DIFFICULT TO ASSESS IF THE NUMBER OF SUCCESSES IS LARGE ENOUGH TO EVALUATE THE USEFULNESS OF THE PREDICTION METHOD. MANY STUDIES REVIEWED ALSO IGNORED THAT THE UTILITY OF A PROBATION PREDICTION DEVICE IS A FUNCTION OF THE SELECTION RATIO, OR THE PROPORTION OF PERSONS CHOSEN FOR PROBATION TO THE TOTAL NUMBER ADJUDICATED, AS WELL AS THE PREDICTIVE VALIDITY OF THE INSTRUMENT. THREE TYPES OF METHODS FOR COMBINING PREDICTORS IN THE EMPLOYMENT OF PREDICTIVE INSTRUMENTS ARE DISCUSSED: THOSE WHICH USE ALL THE PREDICTORS EQUALLY, THOSE WHICH EMPLOY A DIFFERENTIAL WEIGHTING SYSTEM, AND SO-CALLED 'CONFIGURAL' METHODS. A SHORTAGE OF EMPIRICAL COMPARISONS OF THESE THREE APPROACHES IS INDICATED. CROSS-VALIDATION IS EMPHASIZED AS ESSENTIAL FOR IDENTIFYING BIAS RESULTING FROM CHANCE VARIATIONS IN THE ORIGINAL SAMPLE. THE WORD TO PROBATION MANAGEMENT IS THAT, WHILE PREDICTION INSTRUMENTS FOR PROBATION SELECTION HAVE YET TO BE ESTABLISHED AS RELIABLE, THERE IS EVIDENCE THAT SUCH INSTRUMENTS CAN BE DEVELOPED AND USEFULLY APPLIED, ESPECIALLY IN THE AREA OF PAROLE DECISIONMAKING. NOTES ARE PROVIDED. (RCB)