Neural bases of accumulator models

Abstract One class of cognitive models, accumulator models (AMs), is a framework to explain cognitive mechanisms of reaction timing. In AMs, an internal signal, which grows linearly with time upon the onset of an external stimulus, is postulated. Reaction time is defined by the time at which the signal reaches a predefined threshold level. Indeed, cortical neurons with activity that grows seemingly linearly with time are largely found, which is consistent with AMs. Here we show that stochastic dynamics of a recurrent network of bistable spiking neurons can produce linear growth of neuronal activity. This suggests possible neural bases of AMs.