This study aims at revealing the process and mechanism of meta learning in acquiring expertise in the field of insight problem solving. We gave a participant a series of ten geometric insight problems and analyzed the processes microscopically. During solving a series of problems, the participant’s performance was greatly improved. This improvement seems due to three factors. The first factor is increasing variety of combination patterns. The second factor is increasing appropriateness of evaluation, which enables the participant to quickly discriminate feasible trials from unfeasible ones. The third and most important factor is distribution of problem solving load to the environment. Instead of memorizing the goal pattern and mentally evaluating the fitness of a current combination pattern to the goal, the participant recruits external resources on the spot to offload the cognitive burden. These mechanisms are very different from elaborating and sophisticating individual operations. Rather, they contribute to establish the meta level knowledge that enables the participant to improve his performance.
