Summary
The intrinsic uncertainty of sensory information (i.e., evidence) does not necessarily deter an observer from making a reliable decision. It is believed that uncertainty is minimized by integrating (accumulating) this incoming sensory evidence. Traditionally, this integration has been modeled using two different approaches. First, rate-code integrators represent accumulated afferent inputs as monotonically increasing (‘ramping’) spiking activity. Second, location-code integrators represent accumulated inputs as the location of a highly localized ‘bump’ of elevated spiking activity. In general, rate-code integrators are thought to be a natural circuit candidate for perceptual decision-making, but they cannot account fully for recent experimental data suggesting that observers hold accumulated evidence in the temporal gaps in evidence. Here, we propose a novel location-code neural integrator that maintains accumulated evidence during such temporal gaps. Furthermore, our location-code integrator can be read out in two modes that mirror spiking activity patterns in cortical areas associated with decision-making.