Issues associated with the ACT (American College Testing) Assessment Program illustrate the need to assess college entrance examinations in terms of the variety of common predictive uses such as estimation of first year grade point average (GPA), or chances of obtaining a C average after certain semesters or in specific courses. Test developers should also be concerned with the diversity of groups and colleges to be tested. The problems professional schools face illustrate the need to examine the relationship between the predictor (test score) and the success factor (usually college GPA). Although tests are developed to spread students over a wide scale, professional school candidates score within a narrow range, thus affecting the predictive ability of the tests; the same narrow range problem applies to grades. Other prediction issues should be considered--hidden variables in college admissions, use of first-year GPA as the sole criterion, specification of the measure used to predict relationships; consequences of misclassification; test bias; and inability to measure nonacademic skills. In summary, a predictive decision is practical only when its various possible uses, misuses, and relationships to ultimate goals are considered. (CP)