Nonlinear spatial integration underlies the diversity of retinal ganglion cell responses to natural stimuli

Quantitative models of sensory encoding help reveal how neural systems process naturalistic stimuli. In the early visual system, standard models assume that a neuron linearly filters incoming stimuli through its receptive field, but artificial stimuli, such as reversing gratings, often reveal nonlinear spatial processing that is not compatible with linear receptive fields. To investigate whether nonlinear processing is relevant for natural stimulus encoding, we recorded spiking activity of ganglion cells in mouse retina under natural images. We found that linear receptive field models failed to fully capture activity for cells sensitive to fine spatial contrast and that nonlinearities acting in the receptive field center govern this spatial contrast sensitivity. We furthermore identified a ganglion cell type whose spatial nonlinearity makes it particularly sensitive to spatially homogeneous stimuli. Our work highlights receptive field nonlinearities as crucial components for understanding early sensory encoding in the context of natural stimuli.