Development of an emotion discrimination task to measure how people average the emotional expression of the crowd

Facial expressions serve as a channel for emotion communication. Although much research has focused on facial emotion processing from single faces, little is known about how people average emotions from several faces. To study this phenomenon, we developed a task to measure how people average the emotional expression (fear/surprise) of a crowd of faces. Forty-two individuals participated in the four experiments. The results show that people are relatively fast and accurate averaging the emotional expression of the crowd. Moreover, people tend to average faster fear than surprise without making more errors. Interestingly, computational modeling analyses suggest that this pattern of performance may result from the combination of two biases: (i) a bias towards fear, which is independent of the stimuli seen, and (ii) a bias towards surprise, which increases with the number of faces seen...