Schedule of Events | Symposia

A computational study of subordinate-level processing in faces and objects utilizing the expertise hypothesis

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

HSUAN-KAI WENG¹ (ba06107064@tmu.edu.tw), Christoph Dahl; ¹Taipei Medical University

The perceptual expertise hypothesis suggests that subordinate-level processing is a hallmark of specialized categorization, often exemplified by face perception.The literature suggests that the mechanisms underlying subordinate-level facial processing extend to non-face objects. The current study investigates whether these mechanisms are, in fact, identical for non-face object categories and extend to other face categories. We trained artificial neural networks on subordinate-level classification for human, monkey, and cartoon faces, as well as object categories. We subsequently tested these networks under specific visual manipulations designed to isolate part-based and configurational processing. Our findings reveal certain inconsistencies with the perceptual expertise hypothesis; subordinate-level processing for objects differs significantly from that for faces. Furthermore, subordinate-level processing across different face types demonstrates moderate modulations in configural processing, while subordinate processing of object categories only relies minimally on configurations of facial parts. This suggests that subordinate-level processing may not generalize across these categories, challenging the assumption that object and face expertise rely on analogous neural or computational mechanisms.

Topic Area: PERCEPTION & ACTION: Vision