Abstract
Previous research suggests that learning to categorize objects along a given dimension produces changes in the perceptual representation of that dimension, including an increase in its perceptual separability (i.e., the dimension's representation staying relatively invariant across changes in other dimensions). This suggests that object dimensions representing variation in familiar categories, such as face gender and race, might show higher perceptual separability from other dimensions than completely novel stimulus dimensions, such as unfamiliar identity. Three groups of participants completed different identification tasks involving four faces, which resulted from the combination of two levels of facial expression (neutral and sad) and two levels of a second, target dimension. For group Id, the target dimension was composed of two unfamiliar identities, both caucasian males. For group Gn, one of the unfamiliar identities was replaced by a female, making gender the target dimension. For group Rc, the same unfamiliar identity was replaced instead by an asian male, making race the target dimension. A model-based analysis using General Recognition Theory with Individual Differences (GRT-wIND) showed violations of perceptual separability for all dimensions in all groups, but these were stronger for the unfamiliar identity dimension than for the familiar gender and race dimensions. Violations of perceptual separability for the emotion dimension were also stronger when it was paired with the unfamiliar identity dimension than when it was paired with the familiar gender and race dimensions. These results suggest that perceptual separability of a face dimension correlates with its familiarity, and that categorization in the natural environment (e.g., by gender and race) may have similar influences on dimension representation as categorization training in experimental settings.
Meeting abstract presented at VSS 2018