How early visual cortex encodes the binocular statistics of the natural world (360G-Wellcome-205946_Z_17_Z)

Stereoscopic “3D” depth perception relies on sensing small disparities between the images viewed by the left and right eyes. Comparing the two eyes’ images and extracting these disparities is a complex process which begins in the brain’s primary visual cortex (V1). Many neurons in this cortical area are sensitive to binocular disparity, even when no other depth cues are present. I want to understand how they achieve this. Current neural network models capture many aspects of this computation, but real neurons seem better adapted to the statistics of the real world than current models can explain. I plan to record from individual neurons in the primary visual cortex as many thousands of stereoscopic images are viewed. I will exploit new analytical techniques to fit more sophisticated models. I will explore how current models fail, e.g. by failing to capture how real neurons respond to differences in contrast and spatial frequency. I will then modify current models, e.g. by incorporating divisive normalisation. I hope this will produce new insights into how the two eyes’ images are combined in the brain, with benefits to other aspects of perception.