Mice and humans use ambiguous local motion to stabilize gaze

Animals need to infer properties of the environment that are only ambiguously encoded by neurons in the sensory periphery. In vision, local edge motion seen through small receptive fields does not uniquely specify the direction of global or object motion. Can these ambiguous local motion signals drive behavior directly, or must the ambiguity first be resolved? We addressed this question in reflexive gaze stabilization by tracing motion signals from the retina through the nucleus of the optic tract (NOT) to eye movements, combining retinal and in vivo electrophysiology, computational modelling, and eye tracking in mice and humans. By using stimuli that dissociate local edge motion from global, object motion, we found that NOT cell responses were strongly biased towards local motion and were largely explained by linear pooling of retinal inputs. This effect propagated to behavior: eye movements followed local motion, to the point that changing the edge orientation of vertically moving objects produced horizontal eye movements. Humans showed a qualitatively similar bias, which increased when central vision was masked. Thus, ambiguous local motion signals can drive effective behavior; a conserved sensorimotor strategy in which visually guided behavior relies on a heuristic rather than an accurate reconstruction of the external world.