Predictability of Academic Grades for Students of High and Low Academic Promise

THE conjunction of Ghiselli's (1963) research on moderator vari- ables and Humphreys' (1968) work on the predictability of aca- demic performance over the undergraduate years led to the hypothesis tested in this study: namely, that high and low academic promise groups might show differing levels of predictability of their college grades from standard entrance measures.In particular, if unmeas- ured motivational variables, or other noncognitive traits, can compensate for low academic promise, the predictive validities of high school rank in class and measures of academic aptitude should be lower for the low promise group.Procedure.The median high school rank in class and the median ACT composite were computed for Humphreys' sample, and groups representing each of the four quadrants so defined were obtained.Intercorrelations were then computed for the same predictor and criterion variables represented in the earlier study: namely, high school rank in class, the subtests of the ACT and their composite, and the eight semesters of college grades with each semester's average being computed independently of the rest.As in the earlier study, in order to control for mean differences between sexes and among colleges, correlations were computed on a within group basis and weighted averages obtained.The high academic promise group is thus defined as being above the median on both high school rank in class and the ACT com- posite while the low academic promise group is below the median