Reverse Correlation Task

The reverse correlation task is a data-driven method for visualizing the mental representations people hold of social categories and traits, such as what a trustworthy, dominant, or criminal face looks like in the mind's eye. Adapted to social perception by Dotsch and Todorov in 2012, the technique superimposes random visual noise on a base face to create many slightly different images, and asks participants to repeatedly choose, from pairs, the image that best fits a target trait. By averaging the noise patterns from the chosen images, the researcher produces a classification image -- a picture that reveals the visual features the participant's mind associates with the trait, without the experimenter ever specifying those features in advance. Independent raters then judge the classification image to confirm it conveys the intended trait. The method made it possible to render otherwise hidden mental representations and biases as concrete, testable images.