Explorations of Face Space

According to a norm-based "face space" neural representation, faces are arranged by how far they deviate from a norm or average face, with faces of increasing distinctiveness represented by their distance from the norm. This is supported by evidence from adaptation, development, binocular rivalry, and neuroimaging. A series of experiments examines the norm-based structure of face space. First, adapting to young or old faces caused perceived age aftereffects. When middle age faces were adapted, no shift is shown. Both of these results are consistent with representation relative to a norm based on comparisons of pooled neuronal activity. Next, rivalrous neutral and distorted faces were presented binocularly, creating a constantly changing percept. Distorted faces were clearly dominant in initial experiments, but this discrepancy seems to be partially controlled by the distinct face having a greater pull in categorization. Increasing the types of response choices or limiting fusion decreased the bias, with evidence that observers were often seeing a partially distorted face. Finally, other experiments examined our ability to recognize angular and radial changes in a face space representation. Our ability to consciously access our own face space appears lacking, as well as our ability to equate perceived intensities between faces, especially with faces distant in face space. The evidence presented demonstrates that face space is consistent with a norm-based layout, and that this model might accurately be represented within a relative coordinate system similar to color and other models