The publication reports on a statistical analysis of educational level as it affects the prediction of enlisted performance in Army training courses. Samples of the men at each of four educational levels were formed: non-high school graduation, high school graduation, some college, and college graduation. Using the Gulliksen-Wilks analysis of covariance, ACB test scores and final training course grades were analyzed for the various samples to determine whether the predictors are equally effective for the different educational groups. Results of the present analysis suggest that for the less-well-educated, the ACB measures predict higher training performance than the individual is likely to achieve. On the other hand, the likely performance of the college graduate tends to be underestimated. Finds further suggest that effectiveness in predicting training performance could probably be improved by including level of education in the aptitude area composites used in classification. (Author)