Year: 2026
Type: preprint
Abstract: Understanding how the human brain processes dynamic audiovisual events requires datasets that combine naturalistic stimulation, high temporal resolution, and systematic experimental manipulation of sensory information. Here, we introduce AVMoments-EEG, the first large-scale, high-temporal-resolution human EEG dataset designed to characterize neural responses to naturalistic audiovisual events. Across 10 participants, we recorded 64-channel scalp EEG during free viewing of 3-second naturalistic video clips derived from the Audiovisual Moments in Time dataset. The stimulus set comprised 896 training videos (16 event types × 56 videos per type), each presented three times, and 64 held-out test videos (16 types × 4 videos per type), each presented 40 times under three conditions: (1) intact audiovisual input, (2) intact visual input with scrambled audio, and (3) intact auditory input with scrambled visual input. This design yielded 10,368 trials per participant and combines broad stimulus coverage with highly repeated test stimuli. AVMoments-EEG enables temporally resolved investigation of naturalistic event processing and modality-specific audiovisual contributions, while providing a benchmark for video-to-EEG encoding, EEG-based decoding, and brain–model alignment. The dataset offers a resource for studying dynamic event perception, multisensory processing, and the temporal correspondence between human neural responses and artificial neural network representations.
Authors: Shuning Tang, Zitong Lu, Leiting Li, Dongwei Li
Language: English
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
Award: G3199013967
OpenAlex ID: W7214074539