关键词: 平均表征, 面孔吸引力, 平均面孔

Abstract:

Previous research demonstrated that ensemble perception of groups can be formed rapidly by extraction of the average of high-level complex features. However, it is unclear whether the average percept is the outcome of extraction from the characteristic value of the average stimulus (for example, average face) created from group members, or from calculation of the average value of group members’ characteristic values. The above two values were confused with each other in prior research, since most average value of group members are similar as the characteristic value of the average stimulus. However, the attractiveness rating of the average face created from a group of faces is usually systematically higher than the mean value of attractiveness ratings of this group of faces. Therefore, it is easier to explore how the ensemble coding of crowd face attractiveness (i.e. group attractiveness) is formed by comparing the attractiveness of the average face with the mean value of attractiveness rating of a group of faces. This could provide a useful approach to explore how the average percept is formed. The present study used the average discrimination paradigm (Experiment 1 & 2) and the scoring paradigm (Experiment 3 & 4) to clarify the mechanism of the formation of average percept by comparing the group attractiveness with the attractiveness of average face. To tackle this issue, whether the average face was presented in the group of faces or not was manipulated (conditions: Avg vs. NoAvg). Group size were also manipulated to explore whether group size modulated the formation of average percept.
In the average discrimination paradigm, a group of faces served as group stimuli to be compare with the probe face for attractiveness. Participants were asked to judge which is more attractive between the group stimuli and the probe face. In the scoring paradigm, participants were asked to rate the attractiveness of group stimuli, the average face created from the group, and each face of the group in isolated manner. Each group consisted of twelve (in Experiments 1 and 3) or four faces (in Experiments 2 and 4). There were two kinds of groups: one is that all group members are original faces, without the average face. The other is that an average face morphed from other original faces was included in the group.
In Experiment 1, the proportions for judging probe average face more attractive than group attractiveness in the Avg condition was similar with the NoAvg condition. In Experiment 2, when the set size was four, the proportions for judging probe average face more attractive than group attractiveness were significantly higher in the NoAvg condition. Moreover, in Experiment 3, the ratings for group attractiveness were not significantly different between Avg and NoAvg conditions. This may indicate that the group attractiveness is based on the average face which was created from group members rather than the mean value calculated from group members’ attractiveness. In addition, the diffusion model analysis showed that the coding time was longer for NoAvg condition, which indicated that the formation of average face needed cognitive resource. In Experiment 4, when the set size was four, the attractiveness rating of the average face was significantly higher than group ratings for the two kinds of groups. The different results in different group size may be interpreted as the outcome of weakened average percept caused by the salient individual face representations in small group. This was evident from several analyses: 1) group attractiveness and the attractiveness of morphed average face decreased with smaller set size (Experiment 4); 2) When the probe face was morphed average face, the proportion for judging probe face as more attractive than group attractiveness was greater, comparing with the condition when the probe was a new face whose attractiveness was similar with the morphed average face (Experiment 2); 3) The performance for the hypothesized condition with average percept included in the set is in between the conditions with/without real average face included (Experiment 2-4). In addition, comparing with Experiment 1, the information accumulation speed in Experiment 2 is slower, the processing time of group attractiveness is longer, reflecting the disturbance of the individual face representation. In summary, the findings supported the hypothesis that group attractiveness is based on the morphed average face. Thus, the ensemble percept relies on the extraction from the average stimulus created from the group.

Key words: average representation, face attractiveness, average face