Since the occurrence of overprediction of college grades for blacks has been found primarily in studies using SAT scores as predictors, a study was initiated to investigate the overprediction phenomenon when the ACT was used in conjunction with high school grades as a predictor of college GPA. College freshman-year GPA was used as the criterion variable. Separate regression equations and multiple R's were determined for males and females. Predicted college GPA for black and white males and females were determined by use of same-sex regression equations. Comparison of actual vs predicted college GPA were made for the races and sexes separately. No strong and consistent overprediction patterns for black males and females were revealed. In the few instances of substantial overprediction, it was primarily a matter of overestimating black female grades by means of regression equations based on high school grades. When same-sex regression equations were used, black female GPAs were most frequently overpredicted, but when total regression equations were used, black and white male GPAs were most frequently overpredicted. The greatest frequency of underprediction occurred with white female college GPAs. (KM)