Role of V1 ocular dominance for binocular integration

Abstract

Neurons in primate primary visual cortex (V1) combine left and right-eye information to form a binocular response. Models of binocular combination attempt to account for V1 binocular responses as a function of monocular responses and stimulus contrast energy. What has been largely lacking in such models is information about neuronal preference for one eye over the other, i.e., ocular dominance (OD) in V1. Here, we tested the extent to which leading models of binocular combination, with and without OD channels, can predict V1 binocular responses in awake macaque monkeys. To do this, we presented static gratings to either eye (monocular) or both (binocular) at corresponding retinal points while animals fixated a central cue. Gratings appeared on screen for 200ms and varied in Michelson contrast [0, 0.055, 0.11, 0.225, 0.45, 0.9] between trials. We recorded neurophysiological responses to these stimuli using linear multielectrode arrays placed in the parafoveal region of V1. Our sample of V1 units featured a range of ocular dominance profiles (0 to 1, where 1 is exclusively monocular). In line with previous reports, V1 binocular responses constituted a sublinear combination (~60%) of monocular responses. Divisive normalization best predicted sublinear binocular responses as a function of contrast, outperforming alternative contemporary models. Model performance was improved by introducing an ocular dominance weight term to the stimulus drive (i.e., numerator) of divisive normalization. Together, our results extend the binocular normalization framework to the level of spiking responses and suggest that V1 neurons’ preference for eye plays a significant computational role for binocular integration.