Vatanak Lavan¹ (), Masataka Sawayama², Laurie Bayet¹, Bei Xiao¹; ¹American University, ²Hokkaido University

Neural representations of material perception are typically studied as solely stimulus-driven, focusing on the mapping between stimulus features and brain responses. However, people perform different tasks with the same objects rather than simply viewing them passively. Recent works show that task demands systematically constrain how stimuli are encoded in the brain. Here, we investigated whether task context modulates the neural representation of material categories. We recorded scalp EEG using a 128-channel HydroCel Geodesic Sensor Net while six participants (ages 18-25) performed binary judgments on three different material attributes - glossiness, softness, and edibility - of objects from identical images. Our stimulus set included 144 photographs of real-world objects across 8 material categories, with 6 exemplars per category and 3 viewpoints per exemplar, under controlled lighting. Stimuli were presented across two sessions, each with three randomized blocks. Each trial consisted of a task cue (750ms), stimulus viewing (750ms), and a response period. We performed time-resolved binary decoding of event-related scalp potentials using a linear support vector machine with 41ms sliding windows (±20ms) and leave-one-block-out cross-validation. Behavioral results showed significant task effects on reaction time in Session 1 (edibility: 437ms < softness: 477ms < glossiness: 500ms) that disappeared by Session 2. Representational dissimilarity analysis of behavioral judgments revealed that edibility and softness judgments were category-specific, whereas glossiness judgments varied within and across categories. Subject agreement was highest for edibility (κ = 0.651) but lowest for glossiness (κ = 0.366). Neural decoding showed that task and material category were decodable above chance during the task cue and stimulus-viewing periods, respectively (~60-70% peak accuracy). However, cross-task decoding of material categories achieved accuracy similar to within-task decoding throughout stimulus presentation. Our initial results suggest that although task information is present in the brain, it exerts minimal effect on the neural representation of materials.