A Model for Latencies in the Visual System

The visual system is modeled as a multilayer network of coincidence detecting neurons switching to a state of sustained firing after the production of their first spike. Inhibition from the next layer stops the sustained firing only when the information has been used (coincidence detected). Simulations of such a self-timed network show the propagation of a firing probability wave from layer to layer. The latency of the activity onset in each layer is determined by the level of the sustained activity and by the initial jitter of onset times in the first layer. Theoretical analyses confirm the observed relations. Our results indicate that the layer-to-layer latencies in the visual system may be mainly caused by the jitter in onset time in retinal ganglion cells.