An intervention pre- and posttest design was used to study conditions that facilitate children's self-regulation of problem solving activities. In two studies, children were trained to solve a three disc Towers of Hanoi problem. After training, they were asked to solve the problem, describe how they solved it, and perform a four disc generalization task. In the first study, a total of 125 first- through third-graders were assigned to one control and three training conditions. In training, children: (1) practiced solving the problem; (2) watched an adult model solve the problem in the most efficient way; or (3) viewed a video recording of their own prior performance on the problem. Training conditions had differential effects on performance. Training with a model hindered descriptions of solutions and performance of the generalization task, while the video training condition facilitated the regulation of problem solving activities. In the second study, a total of 60 first- and second-graders were placed in one of three video training conditions or a control condition. Children: (1) viewed a recording of their prior performance on the three disc problem; or (2) viewed a video recording of a child model's inaccurate performance on the problem; or (3) viewed a video recording of a child model solving the problem in the most efficient way. Preliminary analyses revealed the superiority of training in which children viewed their own performance. (Author/RH)