Visual space curves before eye movements

In most experiments within the field of cognitive and systems neuroscience, fixation is often required prior to the onset of an experimental trial. However, the term "fixation" is rather misleading since our eyes are constantly moving. One type of transient miniature movement ubiquitously observed during fixation is commonly referred to as fixational eye movements or microsaccades. Perimicrosaccadic compression of visual space -- the ability of retinotopic cells to transiently exhibit predictive spatiotemporal retinotopic compressive shifts toward the target of an impending microsaccade -- is known to dramatically alter visual perception. However, whether perimicrosaccadic compressive shifts can become spatially asymmetric, that is, continuously directed toward a specific foveal region over another (e.g., an upper over a lower region in the fovea) and for what purpose, remains poorly understood. Assuming that these transient shifts are indeed asymmetric under certain conditions, the perceptual and oculomotor consequences such asymmetricity might accompany across visual space is poorly understood. Equally unaccounted for is a mechanistic account of the neural computation and architecture that could support these transient asymmetric shifts while the visual system actively maintains retinotopic organization. Here, we systematically measured visual sensitivity in human subjects to faint probes presented during fixation and around the time of saccadic eye movement at geometrically symmetric retinotopic locations in the foveal, parafoveal, and peripheral regions of visual space. Remarkably, we observed transient local asymmetric visual sensitivity differences between these symmetric retinotopic locations where none should be observed. Equally surprising, we observed the trajectories of saccadic eye movements, which are expected to travel along a linear path, routinely deviate along a curved path toward orthogonal eccentric locations. To provide a mechanistic account of the neural computation and architecture that may explain our results, we proposed a novel neurobiologically inspired phenomenological force field model in which underlying attentional and oculomotor signals are modulated by transient eccentric error signals that manifest as temporally overlapping predictive forces and impinge on the retinotopic visual cortex. These forces, which transiently bias putative population sensitivity toward an orthogonal retinotopic foveal region and, around the time of a saccadic eye movement, along an axis orthogonal to the saccade direction toward a mislocalized peripheral region, succinctly capture the essence of our empirical observations.