Influences of viewing geometry on neural computations of motion and depth

University Of Rochester, Rochester NY

Investigators

Linked publications & trials

Abstract

Vision is active—we move our eyes to acquire information and to track objects of interest. However, eye movements have visual consequences; for example, when we smoothly move our eyes to pursue an object of interest, this adds components of motion to the retinal image. These visual consequences of eye movements must be compensated to judge how objects move in the world. The conventional view is that compensation for smooth pursuit eye movements can be achieved by subtracting a vector (equal and opposite to the eye movement) from retinal image velocity. We demonstrate that this conventional view has limited applicability, and that even a simple combination of eye translation and rotation has visual consequences that cannot be compensated by a vector subtraction. Crucially, this means that the brain needs to infer the viewing geometry (defined here as how the eye translates and rotates relative to the scene) and then perform different computations to estimate the motion and depth of objects during pursuit eye movements, depending on the viewing geometry. We develop a theoretical framework that predicts how perception of object motion and depth should depend on viewing geometry. In Aim #1, we use human psychophysics to test our theoretical predictions. Preliminary data show the perceived motion and depth have perceptual biases that depend strongly on viewing geometry, as predicted by our theory, even while the retinal velocity of the judged object remains fixed. These biases occur automatically, without training or feedback. In Aim #2, we test whether neural activity in passively fixating monkeys is influenced by viewing geometry (without training) in ways that can account for perceptual biases. In Aim #3, we measure both behavioral and neural effects of manipulating viewing geometry to assess whether neural modulations in different visual areas (MT, MSTl, and VPS) can account for the observed perceptual biases. Our strong preliminary data suggest that we will make important advances in understanding how inferred viewing geometry interacts with neural computations of motion and depth. The proposed research is directly relevant to the research priorities of the Strabismus, Amplyopia, and Visual Processing program at the National Eye Institute.