Several experiments investigating the perceptual, acoustical and neural bases of the ‘voice attractiveness averaging phenomenon’ are briefly summarized. We show that synthetic voice composites generated by averaging multiple (same gender) individual voices (short syllables) are perceived as increasingly attractive with the number of voices averaged. This phenomenon, independent of listener or speaker gender and analogous to a similar effect in the visual domain for face attractiveness, is explained in part by two acoustical correlates of averaging: reduced ‘Distance-to-Mean’, as indexed by the Euclidean distance between a voice and its same-gender population average in f0-F1 space and increased voice ‘texture smoothness’ as indexed by increased harmonics-to-noise ratio (HNR). These two acoustical parameters co-vary with perceived attractiveness and manipulating them independently of one another also affects attractiveness ratings. The neural correlates of implicitly perceived attractiveness consist in a highly significant negative correlation between attractiveness and fMRI signal in large areas of bilateral auditory cortex, largely overlapping with the Temporal Voice Areas, as well as inferior prefrontal cortex: more attractive voices elicit less activity in these regions. While the correlations in auditory areas were largely explained by distance-to-mean and HNR, inferior prefrontal areas bilaterally were observed even after co-varying out variance explained by these acoustical parameters, suggesting a role as abstract voice attractiveness evaluators.
