Modelling Audio-Visual Reaction Time with Recurrent Mean-Field Networks

The SUM audio-visual integration model shows optimal multisensory drift rates, while the OR model yields suboptimal integration, across simulated 121 participants with varied unisensory evidence accumulation.

• Why it matters: Audio-visual benefits may reflect true evidence integration or alternative effects like independent processing and modality repetition, which remains unclear from behavior alone.
• What they did: Three mean-field recurrent models (OR, SUM, REPEAT) built on Wong and Wang were simulated and fit to reaction times and accuracy using a drift diffusion model, then compared to unisensory optimal predictions and empirical detection data.
• The result: OR and SUM reproduced established unisensory response patterns across children, children with autism, adults, older adults, and people with Parkinson’s disease, but SUM also produced supra-optimal responses in some conditions, while REPEAT attributed repetition priming rather than true integration.