Certain tasks in programed instruction can be performed only by computer. One such area is the arrangement of differential reinforcement for sophisticated reinforcement contingencies. That is, the capacity of the computer is required to determine whether the student has met the criterion for reinforcement. With this in mind, a computer-controlled program was designed to teach young children to classify a single array of objects many different ways. The child was presented with many objects and asked to touch all the things that were alike. As he classified the objects by color, shape, size, and the like, the computer presented appropriate reinforcements. Testing of this program showed it was successful in reaching a low error rate and satisfactory improvement in the skill of multiple classification. The computer successfully managed reinforcement contingencies for this task which would be difficult or impossible to arrange with simpler devices or procedures. (JK)