Eight fifth-and sixth-grade children with behavior problems performed in a classroom under three conditions: (1) unreinforced baseline, (2) reinforcement for being on task, and (3) reinforcement for the accuracy and rate of math problems solved. The teacher was absent for a portion of the class session under each of these conditions. In the teacher's absence, on-task behavior declined markedly and disruption markedly increased, regardless of the reinforcement condition in operation. In addition, the teacher's absence resulted in fewer problems attempted and decreased accuracy. However, the extent to which the children became disruptive was reduced and the number of problems attempted increased when reinforcement was contingent on academic accuracy and rate, instead of being contingent on being on task. The results suggest that by providing contingencies for the products of a child's classroom activities, rather than for being on task, the child will become more independent of the teacher's presence, and more under the control of the academic materials.