Our brains can extract a summary representation of the facial characteristics provided by a group of faces. To date, there has been a lack of clarity as to what calculations the brain is actually performing during this ensemble perception. For example, does ensemble processing average the fiducial points (e.g., distance between the eyes, width of the mouth) and surface characteristics (e.g., skin tone) of a set of faces in a fashion that produces what we call a ‘morph average’ face from the group? Or does ensemble perception extract a general ‘gist average’ of the face set (e.g., these faces are unattractive)? Here, we take advantage of the fact that the ‘morph average’ face derived from a group of faces is more attractive than the ‘gist average’. If ensemble perception is performing morph averaging, then the adaptation aftereffects elicited by a morphed average face from a group should be equivalent to those elicited by the group. By contrast, if ensemble perception reflects gist averaging, then the aftereffects produced by the group should be distinct from those elicited by the more attractive morphed average face. In support of the morph averaging hypothesis, we show that the adaptation aftereffects derived via temporal ensemble perception of a group of faces are equal to those produced by the group’s morphed average face. Moreover, these effects increase as a linear function of increasing attractiveness in the underlying group. We also reveal that spatial ensemble processing is not equal to temporal ensemble processing, but instead reflects the ‘gist’ attractiveness of the group of faces; e.g., these faces are unattractive. Finally, we show that gist averaging of a spatially presented group of faces is abolished when a temporal manipulation is additionally employed; under these circumstances, morph averaging becomes apparent again. In summary, we have shown for the first time that temporal and spatial ensemble statistics reflect qualitatively different perceptual calculations.
- Rapid serial visual presentation
- Ensemble statistics