Abstract
In order for animals with binocular vision to use retinal image disparity to infer depth, they must be able to maintain some alignment between the eyes. Primates and carnivores frequently make coordinated saccades, and when they fixate on an object, their eyes converge or diverge to align their retinal images at that point. Rodents do not have a fovea, saccade less frequently, and their eyes are typically directed at different objects. However, there is binocular disparity tuning in the primary visual cortex of mice. To study how mice might use this disparity tuning to discriminate depth, we examined binocular eye movements while mice were trained to discriminate between relative near and far surfaces rendered in dynamic random dot stereograms. We found that naïve head-fixed mice had very coordinated saccades between the eyes with mostly horizontal eye movements. As mice were able to significantly discriminate surfaces, they substantially reduced their eye movements. Overall, binocular eye movements were still positively correlated, but leading up to stimulus onset and during stimulation, binocular eye movements were negatively correlated. For both near and far surfaces, the mice would start converging their eyes seconds before stimulus onset and then diverge their eyes once making a decision about whether the surface was near or far. This is because the screen was in front of them, which is not their natural gaze location. We analyzed the average vergence angle with respect to their average gaze for correct and incorrect trials and found that when mice failed to detect near surfaces, their eyes did not converge. When mice failed to detect far surfaces, their eyes converged too much. This data demonstrate that mice do align their eyes and that their successful stereoscopic discrimination depends on proper binocular alignment.
Meeting abstract presented at VSS 2016