A previous study showed that observing others' trials had a positive effect on performance in insight problem solving, whereas observing one's own past trials had a negative effect. We can assume that these effects are caused by the following two factors: one is that the amount and variety of information may increase by observing others' trials, which in turn enhances the possibility of adopting a new perspective or gaining an insight. The second factor is that, regardless of the type of information that a person gains through observation, the fact that this information is obtained from himself⁄herself may disrupt constraint relaxation and consequently, insight problem solving. In this study, we tested whether or not a person's attribution of the observed actions to self disrupts his⁄her performance on the task. For this purpose, we compared the participants' performances across the following four conditions: (1) the solo condition, in which participants were asked to solve a T-puzzle alone; (2) the self-observation condition, in which each participant was asked to alternate between solving the puzzle and observing each of his⁄her own past trials for 30 seconds; (3) the fake other-observation condition, in which each participant was asked to follow the same procedure as in the self-observation condition, but was instructed that the trials he⁄she observed were those undertaken by another person; and (4) the other-observation condition, in which each participant was asked to alternate between solving the puzzle and observing each of another person's past trials for 30 seconds. The results revealed that the participants' performances in the self-observation condition were inferior to those in the other three conditions. The results indicate that observation may disrupt insight problem solving if one attributes the observed actions to oneself, but not if one attributes them to another person.