Abstract
When people see a group of faces, they tend to code the average property, such as the mean identity. In some experiments, participants extracted this ensemble representation rather than the individual identities. This has led to the suggestion that coding an ensemble is the “default mode” when processing identity for groups of faces (de Fockert & Wolfenstein, 2009). These studies used groups of very similar identities, which participants may have found difficult to distinguish as individuals, and which may have facilitated the coding of an ensemble. If ensembles are the default way to represent groups, then they should have a coding advantage over those of individual identities also for groups that consist of easily distinguishable faces. Here, we tested whether an advantage of the ensemble identity varies with the similarity of the group identities. The identities in each group were experimentally manipulated to range from extremely similar to very dissimilar. In Experiment 1, participants viewed groups of four faces, followed by a set average (blends between group identities) or an individual exemplar, and indicated whether this was a member of the group or not. Our results showed that for sets of dissimilar faces, participants more frequently endorsed individual exemplars than set averages as members. For similar faces, which participants tended to see as the same person, participants endorsed set averages more frequently than exemplars. However, this set average advantage was attenuated when the design was modified, such that participants had to select between the set average and the exemplar, to directly test participants’ preferences for ensemble or exemplar representations. In sum, there was no ensemble advantage for easily distinguishable faces, and even for very similar faces, the ensemble advantage was not as robust as expected, suggesting that ensemble representations may not be the default mode to code identity of face groups.
Meeting abstract presented at VSS 2015