Abstract
A face conveys an abundance of information about a person, such as his or her gender, current emotional state, and identity. Such information can be efficiently extracted in a glance, and apparently without much effort, when viewing a single face. However, humans are often confronted with multiple faces at once. In this situation, ensemble representations can be derived from groups of faces via a cognitive mechanism that promotes extraction of average information from the group, often at the expense of accurate information about individual faces. There is some evidence that such ensemble representations can even be generated for facial identity. In two experiments, we tested the assumption that ensemble representation could serve as a quick and efficient way to extract identity information from groups of faces. We systematically varied two parameters that affected the extent to which participants were able to encode the identity from individual faces presented as a group: presentation duration and number identities presented. After viewing ensembles of different identities, probe faces were judged according to whether or not they had been presented. We found that ensemble representations, as measured by incorrect "present" responses given to morphed averages of a face group, were closely linked to memory for individual faces. There was little evidence for ensemble representations being coded either at short presentation durations or when groups comprised many different identities, suggesting that observers do not encode the mean identity of any group of faces at a glance. Instead, ensemble representations for facial identity were found only when representations for individual faces were also prominent. This finding suggests that ensemble representations might be formed during a later stage of face processing, which requires accurate individual representations of each face in a group.
Meeting abstract presented at VSS 2014