Abstract
Faces are perceived to be more attractive when seen in a group than alone, a phenomenon known as 'the cheerleader effect'. The visual system automatically creates ensemble representations of groups, and the recalled characteristics of individual items from the group are biased toward the ensemble average. Across three experiments, we investigated whether the cheerleader effect occurs because the ensemble average of a group of faces has average facial characteristics, which are perceived to be highly attractive. Observers gave attractiveness ratings for target faces shown in a group of faces or objects, and alone. A cheerleader effect measure was created by subtracting the attractiveness of each face when seen alone, from the attractiveness rating when in a group. The largest cheerleader effects, a 1.5% increase, were observed when target faces were presented in groups that could be summarised to give the ensemble representation average facial characteristics; groups with two unique distractor faces (Experiments 1-3), and groups that contained different photographs of the target identity (Experiment 2). Interestingly, the cheerleader effect was significantly reduced, but not eliminated, when the ensemble representation could not have average facial characteristics: when the group contained identical images of the target face (Experiment 1), or when the distractors in the group were houses (Experiment 3a). Interestingly, houses were also perceived to be more attractive in a group of houses than alone (Experiment 3b). Our results demonstrate that the cheerleader effect occurs because individuals are recalled as being similar to a highly attractive ensemble average. We also find, however, that a second mechanism contributes to the cheerleader effect, which is not specific to human faces. We suggest that sample size bias contributes to the cheerleader effect, whereby observers are primed by the numerosity of group size to give higher attractiveness ratings to faces presented in a group.
Meeting abstract presented at VSS 2018