Peri-Saccadic Remapping Accounts for Visual Stability

Our brain makes saccadic eye movements up to 3–5 times per second, which disrupt visual images on the retina, yet we perceive the world to be stable. It has been suggested that this perceptual stability is achieved via peri-saccadic remapping of neuronal receptive fields (RFs), that is, before a saccade, neurons start to respond to visual stimuli at their future RFs to be brought up by the impending eye movement. The recent experimental study further unveiled that the remapping of neuronal RF in Lateral Intraparietal Area (LIP) is an elongation process along the saccadic trajectory. Despite that peri-saccadic remapping is known to be closely associated to visual stability, a deep understanding of how it works remains largely unknown. Here, to address this question, we formulate visual stability as an optimization problem, and explore, from a computational point of view, how the response property of a LIP neuron should vary during a saccade, in order to predict the locations of visual images disrupted by the saccade. Through building a biologically plausible network and training it for the corresponding prediction task, we find that the optimal recurrent connections between LIP neurons are asymmetric, biased to the opposite direction of the saccade, such that when combined with the corollary discharge signal responsible for the saccade, a visual stimulus triggers a cortical wave in LIP propagating in the opposite direction of the saccade, manifesting the elongated remapping of neuronal RF. Our model reproduces the experimental data on peri-saccadic remapping, and demonstrates that peri-saccadic remapping accounts for visual stability.