To find an acceptable way of reducing testing time without altering the administrative use of admissions tests, a study was conducted to test the prediction that scores on subtests of the General Achievement Tests (GAT) in Social Studies, Natural Sciences, and Mathematics could be used to predict the total test score. All answer sheets for 1000 subjects (250 male and 250 female) who had previously taken the tests were rescored for part scores on the two sections of each test. Pearson product-moment coefficients of correlation were then computed by sex for the two subtests and the total score on each of the three GAT tests. Regression equations were then derived from the correlations and used for the prediction of total scores made by subjects in the cross-validation sample. Correlations were then run between the predicted total scores and obtained scores previously recorded. Results showed that none of the original part-whole correlations exceeded +.95. The study findings resulted in discontinuance of the complete GAT tests at a four-year urban college. Each college applicant took only the 15-minute subtests on the three tests, and a total score in scale-form was predicted thereby for use in the admissions process. This reduced the testing time for the GAT from 120 to 45 minutes. The conversion of subscores into predicted total scores was part of the computer-scoring operations for the admissions test battery. It is concluded that the disadvantages of reduced reliability were offset by the advantages of reduced testing time. (DB)