Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Behavioral and neural mechanisms of face identity learning and generalization: a magnetoencephalography (MEG) study

Poster Session C - Sunday, March 8, 2026, 5:00 – 7:00 pm PDT, Fairview/Kitsilano Ballroom

Jieun Cho¹ (), Sung Jun Joo², Sang Chul Chong³, Hee Yeon Im^1,4; ¹University of British Columbia, ²Pusan National University, ³Yonsei University, ⁴BC Children’s Hospital Research Institute

Reliable recognition of a person’s identity requires the visual system to generalize across visual variations, adapting to changes in viewpoint, lighting, and expression. This study investigated how learning from variable exemplars refines perceptual and neural representations of facial identity to support generalization. During learning, participants (N=44) viewed multiple exemplars of unfamiliar faces. Before and after learning, they completed a numerosity judgment task, estimating the number of identities across images. To test whether variability exposure promotes generalization, the post-learning task included previously seen (post-old condition) and novel (post-new condition) exemplars. Neural activity was measured using MEG during all tasks. Behavioral and neural responses were compared across pre- and post-learning sessions to examine how learning reshapes facial identity recognition. Behaviorally, participants reported perceiving fewer identities after learning, reflecting a tendency to group distinct exemplars as the same identities. MEG revealed learning-related changes in neural oscillation within the face network. Repeated exemplars (post-old) elicited reduced alpha/beta power in the posterior fusiform gyrus and posterior superior temporal sulcus during early time window, indicating more efficient encoding in perceptual regions. In contrast, novel exemplars (post-new) elicited increased power in the high-beta to low-gamma range at later time windows in the anterior fusiform gyrus and temporal pole, suggesting integration of novel perceptual inputs into existing identity representations. These findings show that learning from variable exemplars engages the face-sensitive brain regions in a temporally organized hierarchy, from perceptual to abstract identity processing regions, suggesting a neural mechanism for flexible recognition of identities across novel exemplars and contexts.

Topic Area: EMOTION & SOCIAL: Person perception