Context consistency improves ensemble perception of facial expressions

Humans have developed the capacity to rapidly extract summary statistics from the facial expressions of a crowd, such as computing the average facial expression. Although dual-task paradigms involving memory and ensemble tasks have recently found that this ensemble coding ability is biased by visual working memory, few studies have examined whether the context-dependent nature of memory itself can influence the perceptual averaging process. In two experiments, participants made forced-choice judgments about mean facial expressions that were paired with task-irrelevant background images, and the background images either matched or mismatched across encoding and response phases. When the backgrounds matched, it was at either the perceptual level (uniformly oriented lines with the same orientation in encoding and response phases, in Experiment 1 ), or at the summary statistics level (uniformly oriented lines in the response phase that had the same orientation as the mean of randomly oriented lines that were seen in the encoding phase, in Experiment 2 ). Participants in Experiment 1 showed a higher ensemble precision and better discrimination sensitivity when the backgrounds matched than when they mismatched, which is consistent with the kind of robust contextual memory effect that has been seen in prior research. We further demonstrated that the context-matching facilitation effect occurred at both the perceptual level (Experiment 1 ) and at the summary statistics level (Experiment 2 ). These results demonstrate that the effects of visual working memory on perceptual averaging are obligatory, and they highlight the importance of memory-related context dependency in perceptual averaging.