Temporal Dissociation Between Inter-Area and Local Neural Signals in Perceptual Choice

Perceptual decisions arise from neural activity distributed across multiple brain regions, yet it remains unclear how inter-area interactions and local population activity differentially contribute to choice across time. Here, I analyzed simultaneous multi-region Neuropixels recordings from the Steinmetz et al. (2019) dataset to compare the predictive power of cross-area correlations versus local firing rates for behavioral choice in mice performing a visual discrimination task. I found a clear temporal dissociation: during the pre-stimulus period, both inter-area correlations and local activity predicted choice at modest but significant levels (∼54%, p < 0.001), with no difference between signal types. In contrast, during the post-stimulus period, local population activity substantially outperformed interarea correlations (64.5% vs. 57.2%, p < 0.0001), rising to 69% accuracy by 250-500 ms post-stimulus. This temporal switch—from symmetric contribution before stimulus onset to local dominance afterward—demonstrates that the neural basis of perceptual choice shifts from distributed network states to localized evidence coding as sensory information becomes available. These findings provide direct quantitative evidence for a dual-mechanism model of perceptual decision-making in which pre-stimulus inter-area dynamics set the stage for post-stimulus local computation.