Cell-type-specific computations for depth perception from motion parallax

To guide behaviour, the brain must infer the structure of the external world based on incomplete and unreliable sensory inputs. In the visual system, the signals available to the brain are limited to the two-dimensional images formed on the retinae. To reconstruct the three-dimensional location of objects in the environment, visual circuits must infer the missing depth information. This ability is innate in most mammals, not requiring visual experience, and involves the neocortex. While animals take advantage of both monocular and binocular signals to estimate depth, binocular vision is not necessary for depth perception. Animals’ innate capacity for depth perception is thought to rely on visual motion as an essential depth cue. However, we lack a mechanistic understanding of cortical circuits that process visual motion to estimate depth.The innate nature of depth perception suggests that it is mediated by specialized cell types, but how different cell types that make up the cortical microcircuit respond to three-dimensional visual stimuli has not been investigated. In proposal we will record their activity in virtual reality environments, where motion parallax acts as the only cue of depth, and determine how V1 cell types encode depth information. We will then test the hypothesis that V1 cell types make distinct contributions to depth judgments using a behavioural task that challenges mice to discriminate depth based on motion parallax. Finally, we will characterize the circuits underlying these computations by identifying monosynaptic inputs onto single depth-selective neurons and study how they give rise to their response properties using computational models.This project will advance our understanding of specialized circuits in the visual cortex supporting depth perception and provide broader insight into how the brain infers latent sensory variables and uses this information to guide behaviour in the dynamic three-dimensional world.