Abstract
Making accurate decisions requires the brain to maintain evolving representations of accumulated evidence. The population that maintains this evolving representation may change over the course of evidence accumulation. For example, intervening actions like eye movements and navigation can shift the set of neurons that encode subsequent inputs and outputs. A recent study showed that signals representing accumulated evidence are transferred between parietal neurons with different response fields, enabling continuous evidence integration across both smooth pursuit and saccadic eye movements. More generally, changes in the neural population representing accumulated evidence may result from a switch in the set of neurons representing the relevant behavioral output or a switch in the set of neurons receiving task-relevant information. Here, we present a model that achieves this flexible transfer of graded information without changing synaptic connectivity. Graded signals are widely broadcast throughout the population, with a dynamic gating mechanism controlling which neurons are receptive to this information. This mechanism supports a continuous decision process across changing frames of reference, offering a potentially general framework for cognitive continuity in dynamic environments.