Abstract
Perceiving and evaluating the overall mood of a crowd is important for social interaction. To evaluate the mood of a crowd, the visual system should consider the average of both the facial expressions and gaze directions of a crowd, because the judgment of facial expression is influenced by its gaze direction (Adams & Kleck, 2003). Previous studies have shown that people can average facial expressions (Haberman et al., 2009) and gaze directions (Sweeny & Whitney, 2014), but they did not show the relationship between them. The current study investigated whether the ability to compute the average of facial expressions was related to that to compute the average of gaze directions. We presented four faces with different facial expressions and gaze directions. Participants (N = 125) were asked to estimate the average of facial expressions, that of gaze directions, or both of them sequentially. We calculated the error of each task by calculating the difference between the presented and reported averages and normalizing it by the possible reporting range of each feature. We found that the error autocorrelations of facial expressions and gaze directions were high (.63 < rs < .94), indicating that participants were able to compute summary statistics reliably. More importantly, participants’ abilities to average facial expressions were correlated with those to average gaze directions (.37 < rs < .51). Thus, when facial features are related to each other in the processing of a single face, the ensemble processing of those facial features is related to each other. These results suggest that no relationship between the ensemble processing of facial features found in previous studies (Haberman et al., 2015; Sama et al., 2019) is likely because the facial features examined in the previous studies were not related in the processing of a single face.