OPINIONS DIFFER ABOUT THE BEST METHOD FOR TRAINING JUDGES TO MAKE CLINICAL FORECASTS. SOME EVIDENCE SUGGESTS, HOWEVER, THAT JUDGMENTS ARE MORE LIKELY TO IMPROVE UNDER PREDICTION CONDITIONS THAT ARE PRECISELY DEFINED. THIS STUDY ASSESSED THE EFFECT OF PROVIDING IMMEDIATE FEEDBACK TRAINING TO JUDGES KNOWN FROM A PREVIOUS STUDY TO PREDICT EDUCATIONAL CRITERIA AT RELATIVELY HIGH, MODERATE, OR LOW LEVELS OF ACCURACY. THE CRITERIA PREDICTED WERE FRESHMAN AND OVERALL COLLEGE GRADES. IN COMPARISON WITH JUDGES WHO RECEIVED NO TRAINING, THE FORECASTS OF "LOW" ACCURACY JUDGES SHOWED SUBSTANTIAL IMPROVEMENTS FOR BOTH PREDICTED CRITERIA. HOWEVER, THE TRAINING HAD NO NOTICEABLE AFFECT ON THE JUDGEMENTS OF THE "HIGH" OR "MODERATE" ACCURACY JUDGES. THIS DOCUMENT IS A NATIONAL MERIT SCHOLARSHIP CORPORATION REPORT, VOLUME 3, NUMBER 3, 1967. (AUTHOR)