Modeling the observed center-surround summation in macaque visual area V1

The mechanisms of center-surround summation, the process by which visual cortical neurons integrate the inputs from the classical receptive field and the non-classical surround, are poorly understood. We constructed a set of 32 representative center-surround stimuli using a repertoire of four bar types, and recorded the responses of 83 neurons from visual area V1 in two awake, fixating monkeys to each of the stimuli. We then studied, for each cell individually, the extent to which the observed responses of the cell to center-surround stimuli could be accounted for by a linear regression model of center-surround summation. The model hypothesized that the response of a given cell to a given center-surround stimulus is a weighted linear sum of its responses to the four bar types. This model accurately predicted the observed responses to the center-surround stimuli for about two-thirds of V1 cells (56/83, 67%). The ability of the model to predict the observed responses of the cells was not attributable to overfitting or other modeling artifacts, a lack of surround modulation, or a lack of response modulation across different center-surround stimuli. Furthermore, for many cells, the model was able to predict the cell's responses to novel stimuli, indicating that the model captured the center-surround summation behavior of these cells adequately. Together, our results indicate that this simple bottom-up summation mechanism can account for many important center-surround phenomena in V1, including surround inhibition or facilitation, and selectivity for