A canonical gated neural circuit model for flexible perceptual decisions
Flexible perceptual decision-making requires rapid, context-dependent adjustments, yet its underlying neural circuit mechanisms remain unclear. Here, we reverse-engineer a canonical neural circuit model that integrates sensory evidence and selects actions via distributed neuronal encoding, guided by data from a task that dissociates perceptual choice from motor response. The model's nonlinear gating of action selective (AS) neurons replicates parietal cortical activity observed during task performance. Critically, recurrent excitation within the evidence integration (EI) population supports sensory evidence accumulation, working memory for sequential sampling, and reward rate optimisation. Moreover, the dynamics of EI and AS neurons respectively mirror parietal activity related to sensory evidence encoding and ramping-to-threshold firing in a separate task, suggesting that decision readout engages both populations. The model also explains decision interference in a two-stage decision task, capturing observed accuracy decrements while predicting slower decisions. Together, these findings propose a foundational circuit-level mechanism unifying perceptual, memory-based, and abstract decision-making.