As part of a project formulating optimal rules for decision making in computer assisted instructional systems in which the computer is used as a decision support tool, an approach that simultaneously optimizes classification of students into two treatments, each followed by a mastery decision, is presented using the framework of Bayesian decision theory. The main advantages of handling the three decision points simultaneously compared with separate optimization of such decisions are more efficient use of data and the use of more realistic utility structures. Both optimal weak monotone and strong monotone rules are considered. The results are empirically illustrated using data for 17,259 students for the problem, well-known in The Netherlands, of selecting optimum continuation schools at the end of elementary school on the basis of achievement test scores. (Contains 2 tables, 1 figure, and 6 references.) (SLD)