Abstract
Humans are highly experienced at detecting small variations in face configurations for individual recognition. If visual processing is adapted to efficiently process standard face configurations, such naturally encountered configurations might yield more precise discrimination than the atypical configurations associated with highly distorted faces. A purely geometrical perspective, however, could suggest an opposite prediction: in a frontally viewed undistorted face, the eyes lack a nearby landmark to define their position-they are neither very close to the nose nor very close to the side of the head. If an eye's distances from these two nearby landmarks are perceptually judged with the variability prescribed by Weber's Law, discrimination of eye separation should be relatively inefficient with natural faces.
In two experiments, discrimination of eye separation was assessed using a continuum of faces ranging from center-compressed to center-expanded. The faces were morphed from a single original face using horizontal variations of magnification centered at eye level. Experiment one allowed subjects to adjust a reference face to make it just noticeably more center-compressed, or center-expanded, than the comparison face. Results indicated better discrimination among the more compressed and more expanded faces, than among the more natural ones. In experiment two, subjects made comparative judgments of degree of compression/expansion between a briefly presented compressed face and comparison face while adapted to either a normal face or a similarly compressed face. Results demonstrated no significant effect of adaptation on discriminability.
Supported by NIH grant EY01711.