The purpose of this study is to illustrate how a test may be used more effectively for prediction through the identification of subgroups of individuals for whom the test is especially valid as a predictor. It was hypothesized that scores on the Strong Vocational Interest Blank would be more predictive of success in an engineering school for noncompulsive than for compulsive students. Students were classified as compulsive or noncompulsive in two ways: on the basis of (1) scores on the Accountant Scale of the Strong blank, and (2) reading speed in relation to vocabulary. The compulsive subgroups were composed of students who most resembled accountants in one case, and students whose speed of comprehension was low in relation to vocabulary in the other. The value of the Strong scores as predictors of freshman average grade was determined for compulsive and for noncompulsive groups. The results tended to confirm the hypothesis that Strong scores are more valid predictors of academic success for compulsive than for noncompulsive students. It is concluded that this study demonstrates that the use of tests could be improved by discovering subgroups of people for whom a test is especially appropriate as a predictor.