Abstract
People are exquisitely sensitive to deviations in bilateral symmetry. In faces, symmetry acts as a marker for physical health, and has been implicated as a cue for attractiveness judgements and mate selection. Portrait photographers claim that illumination can increase the attractiveness of their models, presumably as, even in simple shapes, shape perception is biased in the direction of the illuminant. We investigate explicitly whether asymmetric illumination affects the perceived symmetry of faces, and whether this could mediate changes in facial attractiveness. 16 participants underwent a psychophysical symmetry judgement task, where faces generated in FaceGen were manipulated under three horizontal lighting conditions (45o left, frontal, 45o right), and 19 directional symmetry conditions, from 25% leftwards asymmetric, to 25% rightwards. Psychometric curves fitted to the data showed that the point of subjective equality was biased significantly in the direction of the illuminant (Repeated Measures ANOVA, F(2,30) = 17.224, p < 0.001 η2 = .535). There was a leftwards bias across thresholds which could be consistent with previously reported preferences for top-left illumination or internalised rightwards asymmetry of Caucasian faces. In two further experiments we investigated whether this perceptual shift in the judgement of symmetry generalised to attractiveness judgements, by testing whether peak attractiveness judgements would be awarded to faces shifted along the asymmetry axis in the direction of the illuminant, the 'perceived symmetry', as per the results of experiment 1. There was a small but non-significant effect of lighting direction, which could be attributed to the relatively small effect size of symmetry on attractiveness compared to other facial attributes, the fact that computer generated and natural faces have different results in the asymmetry-attractiveness literature, and is consistent with research showing that directional asymmetries have a smaller effect on attractiveness judgements than fluctuating asymmetries.
Meeting abstract presented at VSS 2018