Locating attractiveness in the face space: Faces are more attractive when closer to their group prototype

TIMOTHY POTTER Universit Catholique de Louvain Louvain Belgium Fonds de la Recherche Scientifique Brussels Belgium Face attractiveness relates positively to the mathematical averageness of a face, but how close attractive faces of varying groups are to their own and to other-group prototypes in the face space remains unclear. In two studies, we modeled the locations of attractive and unattractive Caucasian, Asian, and African faces in participants' face space using multidimensional scaling analysis. In all three sets of faces, facial attractiveness significantly increased with the absolute proximity of a face to its group prototype. In the case of Caucasian and African faces (Study 1), facial attractiveness also tended to increase with the absolute proximity of a face to the other-group prototype. However, this association was at best marginal, and it became clearly nonsignificant when distance to the own-group prototype was controlled for. Thus, the present research provides original evidence that average features of faces contribute to increasing their attractiveness, but only when these features are average to the group to which a face belongs. The present research also offers further support to face space models of people's mental representations of faces. - Previous research on faces has shown a relation between facial attractiveness and the mathematical averageness of a face (e.g., Langlois & Roggman, 1990). Rhodes, Harwood, Yoshikawa, Nishitani, and McLean (2002) also showed that people from different social groups (e.g., Caucasian and Asian) show high intergroup correlations in attractiveness ratings of the same mathematically average faces, regardless of the group memberships of the raters or of the faces being rated. Yet, relatively little is known about how peoples mental representations of faces relate to the faces attractiveness. In particular, no research to date has examined how faces that vary in attractiveness are represented relative to mathematical averages (i.e., to prototypes) for various ethnic groups, a gap that the present research was designed to address. One popular framework used to understand how faces are perceived, encoded, and retrieved is the face space (Lewis, 2004; Valentine, 1991). This space is akin to a mental map in which each dimension is normally distributed and holds information necessary to remembering and recognizing a face. Busey (1998) showed that the face space can be modeled adequately using multidimensional scaling (MDS) analysis. He used faces of bald men and their morphs in a similarity judgment task involving all possible combinations of faces. The morphs appeared more typical than their parent faces and were found closer to the center, in line with predictions made by the face space model. The morphs were also found to be less typical than predicted, consistent with recent research on attractor field models that has shown greater perception of dissimilarity between two faces in more populated, highdensity zones (Corneille, Hugenberg, & Potter, 2007; Tanaka & Corneille, 2007; Tanaka, Giles, Kremen, & Simon, 1998). Concerning race in the face space, Byatt and Rhodes (2004) showed using MDS that, for Caucasian participants, Asian faces cluster in the face space more than Caucasian faces, and this clustering explained why Asian faces were less well recognized. As for attractive faces in the face space, Potter, Corneille, Ruys, and Rhodes (2007), using real faces, recently showed that, because of their perceived similarity, attractive faces cluster in the face space more than unattractive faces, regardless of group. This clustering makes them lie in higher-density zones that require more competition for activation, which accounts for their decreased recognition in comparison with unattractive faces (Lewis, 2004; see also Light, Hollander, & Kayra-Stuart, 1981). However, Potter et al.s study had a heterogeneous, non-Caucasian group of faces (involving Asians, Africans, and North Africans) and did not include prototypes from the various groups, so it was not possible for the authors to determine whether attractive faces were closer to the average face of an attractive faces own specific group. Thus, it remains to be known where faces that vary in attractiveness and group would be located in the face space relative to own- and other-group prototypes. In the present research, we used software from Singular Inversions called FaceGen 3.1 (www.facegen.com) that has recently been used in research on face perception (Russell, Sinha, Biederman, & Nederhouser, 2006; SchulteRther, Markowitsch, Fink, & Piefke, 2007; Shimojo, Simion, Shimojo, & Scheier, 2003). This has enabled us to generate materials suitable to answering our questions. In two studies, we implemented a similarity judgment task involving the most attractive and unattractive faces of two groups, as well as their respective prototypes. We ran an MDS analysis in order to investigate our research questions. Below, we introduce the main questions we were interested in for the present research. Next, we explain why we chose to address these questions using computergenerated faces. Finally, we report and discuss the two studies designed to fit our present research interests. The Present Research Questions 1. Are attractive faces closer to their group prototypes than are unattractive faces (Hypothesis 1A), and do attractiveness ratings of faces increase with proximity to the own-group prototype, irrespective of distance to the other-group prototype (Hypothesis 1B)? If supported, these hypotheses would extend previous research that has shown that category prototypes are cognitively pleasing (e.g., Halberstadt & Rhodes, 2000). Indeed, being a prototypical, good exemplar of its specific group (e.g., of cars or clocksand perhaps, in the present research, of an ethnic group) makes a face easier to process and more attractive (Winkielman, Halberstadt, Fazendeiro, & Catty, 2006). 2. Are attractive faces closer to the other-group proto type than are unattractive faces (Hypothesis 2A), and do attractiveness ratings of faces increase with proximity to the other-group prototype, irrespective of distance to the own-group prototype (Hypothesis 2B)? If supported, these hypotheses would suggest that facial features that transcend group differences play a role in attractiveness. This would be consistent with previous research that has shown that morphs between faces of different races (Asian and Caucasian) are judged as more attractive than the parent faces used to create them (Rhodes et al., 2005). For example, the most attractive Caucasian faces could lie closer not only to the Caucasian, but also to the African, prototype. The corollary is that, for several faces that are equidistant to their group prototype, the most attractive of the faces would be the ones closest to the other-group prototype. Why Did We Use Computer-Generated Faces? First, in order to model a face space, (...truncated)