Abstract
The fovea shows several specializations compared to the peripheral retina, such as higher cone density and lower cone-to-bipolar sampling ratio. Despite the importance of foveal information processing to human vision, we know relatively little about it, due to the technical challenges in measuring the spatial and chromatic sensitivity at the necessary resolution: likely at single-cone spacing (one arc-minute). This fundamental challenge has made it difficult to address elementary questions about V1 neurons that represent the center of gaze, such as the size of their receptive fields (RFs), their conjoined selectivity to color and spatial features, and the degree to which they receive matching chromatic input from the two eyes. Here, we recorded from neurons in foveal V1 of macaque monkey with two chronic multielectrode arrays in one hemisphere and an acute laminar array inserted perpendicular to cortex in the other; neural responses were elicited with spatio-chromatic noise designed to enable us to infer how subcortical inputs are integrated within V1 cells using a data-driven nonlinear modeling approach. Furthermore, we used a model-based eye-tracking approach that leverages the fine spatial structure of foveal RFs to refine eye position estimates beyond the limits of camera-based eye trackers. Across the range of recordings spanning cortical lamina and eccentricities up to 2º, we observed V1 spatio-chromatic RFs spanning as little as 5 arcmin with features as fine as 1-2 arcmins. We found a range of chromatic sensitivities, including S-cone input across all eccentricities. Neurons with stronger chromatic selectivity tended to have larger RFs with less spatial structure and were significantly more common in extragranular layers. Together, these measurements provide the first detailed pictures of spatial-chromatic processing in foveal V1, with the goal of filling in the critical gap between the retina and visual perception at the center-of-gaze.