Face-inversion effects flex with perceptual learning

Perceptual learning is often highly specific to the trained stimulus. For example, practice effects are abolished for stimuli orthogonal to the trained orientation in tasks such as vernier discrimination (Fahle, 2005). We examined the extent to which such stimulus-specific learning occurs for faces: complex stimuli that are notorious in the literature for their orientation-specific disruptions to identification. On Day 1, observers were trained to identify one of two sets of faces using a 10AFC task; for each set, separate groups were trained with upright or inverted stimuli. On Day 2, all observers performed the task with Same Upright, Different Upright, Same Inverted, and Different Inverted faces at several contrast and external noise levels. Proportion correct was measured on both days. Performance improvements on Day 2 were consistently specific to the trained stimulus set. Practice with upright faces substantially enhanced performance within the same set for the upright orientation, and partially benefited identification in the inverted orientation. Practice with inverted faces also improved performance in the trained orientation for the trained set; however, there was no benefit for upright faces from the same set. Overall, the face-inversion effect for the trained set of faces respectively increased and decreased after practice with upright and inverted faces. A control experiment with phase-scrambled faces showed no transfer of learning across either trained orientation or stimulus set. Thus, the face-inversion effect can be attributed to viewpoint-specific experience with oriented stimulus structure.