Abstract
Visual processing in the fovea (center-of-gaze) is central to human perception, yet most neurophysiological studies of neural processing focus on neurons in the parafovea (>2 degrees eccentricity). This is because the small receptive fields in the fovea have a higher resolution than conventionally used eye-tracking hardware accounts for, making detailed assessment of receptive field properties unreliable. Here, we recorded V1 neurons at eccentricities varying from 0 to 16 degrees of visual angle in awake, fixating macaque monkeys during presentation of a temporally varying random bar stimulus. We used a model-based eye-tracking algorithm to accurately correct for fixational eye movements, allowing for the determination of detailed receptive field properties. To measure these, we fit nonlinear cascade models to the recorded responses, which provided detailed information about their spatiotemporal processing, as well has how multiple "subunits" – each selective to a different spatiotemporal feature – combine nonlinearly to best predict the observed response. Our preliminary analysis of these model-based results shows that as eccentricity increases, models required a larger number of subunits that are more spatially dispersed, consistent with observations that the average overall receptive field size also increased. Surprisingly, we found that the spatial and temporal frequency tuning properties of individual subunits comprising the models had a much weaker dependence on eccentricity. This study thus provides the first detailed comparisons of visual processing between foveal versus parafoveal neurons.
Meeting abstract presented at VSS 2018