Abstract
Recent studies showed that the visual system can accurately extract mean emotion, gender (Haberman & Whitney, 2007), and identity from groups of faces (de Fockert & Wolfenstein, 2009). Meanwhile, in the studies of facial attractiveness, it is reported that averageness positively affected facial attractiveness ratings (Masashi et al., 2009) and that facial attractiveness can be assessed in a brief time (Olson & Marshuetz, 2005). Based on these findings, we hypothesized that the ability to judge mean face identity might play a role in the assessments of mean attractiveness of sets of faces. Through a preliminary experiment, we measured individual attractiveness ratings of 240 faces. In Experiment 1, we investigated how the difficulty levels of mean-identity judgments influenced mean-attractiveness judgments. The difficulty level of mean-identity judgments was defined as the average distance between the individual faces and mean face identity of them. When the average distance was far, it was difficult to judge mean face identity. A set of four faces was presented for 2 seconds and participants sequentially performed two tasks on them (mean attractiveness ratings and mean-identity judgments). The order of the two tasks was randomly varied across trials. We found that ratings of mean attractiveness did not significantly differ from the arithmetic mean of individual attractiveness ratings measured in the preliminary experiment, when the mean-identity task was easy. This result suggests that mean attractiveness judgments follows the arithmetic mean of individual ratings when one can easily judge mean face identity. However, mean attractiveness ratings were significantly lower than the arithmetic mean of individual attractiveness ratings, when the mean-identity task was difficult. We replicated these findings with exposure duration of 500 ms, followed by a phase-scrambled mask for 500 ms in Experiment 2. These results suggest that overall attractiveness judgments depend on our ability to extract mean face identity.
Meeting abstract presented at VSS 2012