The predictive validity of the Scholastic Aptitude Test (SAT) for Mexican-Americans is investigated. Forty-two Mexican-American freshmen students who entered the University of California, Riverside, in the Fall 1971 participated in the study. Analyses of variance concerning ethnic groups on GPA (grade point average) and SAT verbal (SATV) and math scores (SATM) were performed indicating higher Anglo-American scores on all measures. The hypothesis that both groups can be described by the same regression plane can be rejected. When the Anglo-American regression equation was used to predict the grades of Mexican-American students, substantial overprediction resulted. If separate equations are calculated for each subgroup, the SAT is almost as predictively valid for Mexican-American students as it is for Anglo-American students. It would appear that the pattern of standardized regression rates for GPA and SATM (since Mexican-American students score higher on math probably due to bilingualism, in which vocabulary might be more affected than mathematical reasoning) does not maximize grade success for the ability profiles of Mexican-American students as a group. (RC)