Neural dynamics become significantly faster and more nonlinear at the peak of information processing

Optimal temporal integration windows for decoding mouse visual population spike trains contract after stimulus onset and then re-expand, reaching finest resolution when stimulus identity is most accurately decoded.

• Why it matters: Neural coding is often assumed to use a fixed timescale, but it is unclear how the information-optimal temporal resolution evolves within a single stimulus-evoked episode and what shapes that evolution.
• What they did: Using Integration-Split, a decoding framework that estimates time-varying temporal resolution, the authors analyzed spike trains across four visual paradigms and the mouse visual hierarchy, and compared results to a hippocampal control.
• The result: The optimal scale tracked an inverse relationship with stimulus decodability and closely followed an autocorrelation-based signal-to-noise optimum, while stronger nonlinear predictive structure aligned with finer resolution and was recapitulated in Transformer attention.

AI summary of the abstract · may contain errors · not medical advice · read the paper.