The receptive field construction of midget ganglion cells in primate retina.

Abstract

The midget pathway of the primate retina provides the visual system with the foundations for high spatial resolution and colour perception. An essential contributor to these properties is centre-surround organisation, in which responses from the central area of a cell's receptive field are antagonised by responses from a surrounding area. Two key questions about centre-surround organisation are unresolved. First, the surround is largely or completely due to negative feedback from horizontal cells to cones: how can this feedback be reconciled with the popular difference of Gaussians (DOG) model, which implies feedforward inhibition? Second, can the spatial extent of centre and surround be predicted from the components - optics, horizontal cell receptive field, ganglion cell dendrites - that give rise to them? We address these questions with a computational model of midget pathway signal processing in macaque retina; model parameters are derived from published literature. We show that, contrary to the DOG model, the surround's effect is better treated as divisive. A simplified version of our model - a ratio of Gaussians (ROG) model - has practical advantages over the DOG, such as accounting for spatiotemporal interactions and pulse responses. The ROG model also shows that both centre and surround radii can be calculated from a sum of squared radii of their components. Finally, chromatic antagonism between centre and surround in the full model predicts cone opponency as a function of eccentricity. We suggest that a signal-processing model gives new insight into retinal function.