Capacities of mindreading are essential for human social life. It is hypothesized that people selectively use two types of mindreading strategies; when a target person is perceived to be similar to oneself, people project one’s own mental states to the person (projection): when the target is perceived to be dissimilar, category-based stereotype is used (stereotyping). In this study, we tested this hypothesis with the reaction time paradigm (e.g., Tamir & Mitchell, in press). Given that the both projection and stereotyping are computationally modeled as anchoring-and-adjustment processes, the reaction time paradigm can be used to detect a strategy used in mindreading. We found the stereotyping was unanimously employed independent of the similarity to the target person and projection was employed only when the perceived similarity was high. Our results are congruent with Tamir & Mitchell (in press) and confirmed the utility of the reaction time paradigm as a tool for investigating mindreading strategies.