The present study was to investigate the knowledge of problem structure acquired by solving a problem. Materials were the modified Duncker's radiation problems. Subjects solved the basic problem first and then solved the target problem. Whether or not the structural features of the problem were the same between base and target problem was manipulated as the experimental factors. The manipulated experimental factors were as follows; (1) goal-causality: the reason why goal state must be realized; (2) constraint-causality: the reason why the constraint occurs in the problem state; (3) object relation in the problem. The following results were obtained; the sameness of the constraint-causality between two problems had an effect on target problem solving only when either the goal-causality or the object relation was the same between base and target problem. These results suggested that the knowledge of problem structure involved the goal-causality and the object relation as well as the constraint-causality and solution plan.