Type I Error Is Inflated in the Two-Phase Reverse Correlation Procedure

Mental images of social categories are highly consequential: They can reveal biases and help elucidate the factors that contribute to those biases. One strategy frequently used to evaluate the properties of mental images is reverse correlation, which is a data-driven method that allows researchers to visualize a person’s mental representation of individuals or groups. In social psychology, this technique often employs a unique two-phase structure. This approach, however, has not yet been carefully validated, and its structure may alter the properties of the statistical tests used to evaluate differences between conditions. Using computer simulations to evaluate the Type I error rate in a typical two-phase reverse correlation procedure, we find that it is inflated in a nontrivial set of circumstances.