People often conduct preprocessing to simplify primary processing. Usually, there is a trade-off between the costs of performing preprocessing and primary processing. Therefore, the utility of preprocessing differs depending on the task complexity. We conducted three experiments to find out whether people could adaptively estimate the utility of preprocessing and then take rational action. The overall result was that in performing a high complexity task, almost all the participants made a rational choice. However, for a low complexity task, the participants gradually learned to conduct preprocessing despite it not being effective. These results were explained based on theoretical perspectives proposed in previous studies.