The classification problem in educational testing is a decision problem. One must assign subjects to one of several available treatments on the basis of test scores, where the success of each treatment is measured by a different criterion. Examples of classification decisions include individualized instruction, counseling, and clinical settings. Classification problems have generally been resolved using linear regression techniques. The author proposes replacing these techniques with an empirical Bayesian decision analysis. The case of classification with a threshold utility function is analyzed, and a method for finding optimal assignment rules is described. The illustration is restricted to assignment on the basis of a single test score without quota restrictions. The purpose of the paper is to contrast the differences between the traditional linear regression and a Bayesian approach to the classification problem. (Author/DWH)