Rethinking Configurational and Feature-Based Processing in Visual Recognition

Poster Presentation: Monday, May 19, 2025, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Face and Body Perception: Parts and wholes

Yuxuan Zeng1 (), Ren E Hentz1, David E Osher1; 1The Ohio State University

Second-order configuration has long been considered a hallmark of face perception, unique among high-level visual categories. Behavioral phenomena, such as the difficulty in recognizing inverted faces, are frequently attributed to the unique configurational information inherent to faces. However, it remains unclear whether such information is exclusive to faces or if it can be generalized to other visual stimuli. To address this question, we developed novel, abstract stimuli that contained only second-order configural information, with no resemblance to faces, which contain identical features across the stimulus set (i.e. they can only be distinguished though configuration and not by the presence or absence of features). In parallel, we designed control stimuli that included only feature information without distinct configurations (i.e. they can only be distinguished by their features). Twenty participants were trained to categorize these stimuli and were tested on tasks traditionally associated with configurational processing, including inversion, misalignment, part-whole, and composite tasks. Reaction times during categorization were compared across tasks and conditions relative to the final training stage. Results showed significant effects in the configuration condition across all tasks (p = 3.79x10-3 inversion, 1.35x10-4 misalignment, 2.15x10-3 part-whole, 4.04x10-5 composite; Bonferroni corrected). Importantly, the feature condition revealed no significant effects for any of these manipulations. These findings align with prior research on face and object perception but challenge the prevailing notion that second-order configuration is exclusive to faces. By demonstrating that second-order configuration can exist independently of facial features, our results invite a re-evaluation of the interplay between configurational and feature-based information in broader object recognition processes, including face perception.