Against popular belief, the most attractive faces are not average face: Attractive features are at the outskirts of the natural distribution of face variations, suggesting a selection pressure away from the average

Zhan, Jiayu, Meng Liu, Oliver G. B. Garrod, Christoph Daube, Robin A. A. Ince, Rachael Jack, and Philippe schyns. 2020. “Beauty Is the Eye of the Cultural Beholder.” PsyArXiv. September 8. doi:10.31234/osf.io/34vsj

Abstract: Is face beauty universally perceived from a common basis of objectively definable face features, or is it irreducibly subjective and in the idiosyncratic eye of the cultural, or even individual beholder? We addressed this longstanding debate by objectively modelling the face beauty preferences of 80 individual male participants across Western European (WE) and East Asian (EA) cultures. With state-of-the-art 3D face capture technology, we derived a generative model that synthesized on each trial a random WE or EA female face whose shape and complexion is constrained by natural face variations. Each participant rated the attractiveness of the face on a Likert scale. We then reverse correlated these subjective ratings with the synthesized shape and complexion face parameters to reconstruct individual face models of attractiveness for same and other ethnicity faces. By analyzing the resulting 80 individual models and reconstructing the representation space of face beauty, we addressed several key questions. Against popular belief, we show that the most attractive faces are not average face. Instead, attractive features are at the outskirts of the natural distribution of face variations, suggesting a selection pressure away from the average. Such features also form their own subspace that is separate from cues of sexual dimorphism (i.e. masculine vs. feminine). Finally, we reveal the global preferences of face features across cultures, and specific cultural and individual participant idiosyncrasies. Our results therefore represent face attractiveness in its diversity to inform and impact fundamental theories of human social perception and signalling and the design of globalized digital avatars.