Abstract
Faces of personally familiar individuals are represented in a view invariant manner in the brain. Changes in viewpoint, age, and distance, as well as impoverished viewing conditions do not affect recognition of familiar faces as much as they affect unfamiliar faces. However, fine tuning of the face processing network to familiar faces might not allow similarly invariant behavioral responses to changes identity specific features. We hypothesized that the categorical boundary for identifying images of a familiar face may be narrower than categorical boundaries for unfamiliar face images. We used a continuum of morphs between familiar and unfamiliar faces to find the threshold at which a stimulus was categorized as the familiar face. As a control, we used a continuum of morphs between two unfamiliar faces to find the threshold at which one unfamiliar face was distinguished from another. We analyzed categorical perception curves for unfamiliar-familiar morphs and unfamiliar-unfamiliar morphs, by comparing the frequency of identity judgments for each level of morphing. We also analyzed the reaction times in the experimental and control conditions. Our results indicated slower response times when the morphed stimuli were drawn from unfamiliar-familiar continuum as compared to the unfamiliar-unfamiliar continuum. More interestingly, we found that the morph percentage threshold for reporting a stimulus as familiar was closer to the familiar face end of the morph continuum than the unfamiliar face end, suggesting lower tolerance for deviation from a familiar identity. This indicates that subjects use a more conservative threshold for categorizing the identity of a familiar face than when categorizing the identity of an unfamiliar face. These results suggest that the face identification system is more finely tuned to familiar faces and less tolerant of changes to their features.
Meeting abstract presented at VSS 2017