Abstract
Abstract: :
Purpose: In humans and other vertebrates, the retinal image is constantly moving due to small fixational eye movements. 'Dynamic theories' of hyperacuity propose that temporal patterns of retinal responses induced by these small movements are used to recover high-resolution spatial information about the visual scene. In theory, the spatial resolution of this strategy is limited only by the temporal precision of retinal responses. We explored the precision of retinal responses induced by fixational eye movements. Are responses sufficiently precise in time to recover high spatial resolution information? Methods: Spike trains from salamander RGCs were recorded extracellularly using a multi-electrode array. Stimuli were natural and artificial scenes moving to simulate the drift-like motion of the salamander eye. Time-varying image position was defined by a 5 Hz low-pass filtered Gaussian distribution with zero mean and standard deviation of 6 µm. We also explored the effect of increasing motion amplitude up to 60 µm. Results: RGCs responded sparsely and in bursts, with low mean firing rates (typically <0.5 Hz) but high peak rates (≷150 Hz). For most cells, mean firing rate increased with increasing drift amplitude. To assess temporal precision, we calculated the standard deviation ('jitter') in the timing of the first spike of each burst across repeated trials. The mean temporal jitter across all cells was 50 ms for a drift amplitude of 6 µm, decreasing to 27 ms at an amplitude of 60 µm. The smallest temporal jitter measured for a single event was 2 ms, for the largest drift amplitude. Using the image velocity, we converted temporal jitter into an equivalent spatial jitter, which corresponds to the trial-to-trial variation in spatial position of the image that triggered each firing event. Mean spatial jitter was 11 µm at a drift amplitude of 6 µm, rising to 54 µm at an amplitude of 60 µm. The smallest spatial jitter recorded was 3 µm, for the smallest drift amplitude. Conclusion: The spatial jitter was typically several times smaller than RGC receptive field centres measured using standard methods (halfwidth 60 µm). In many cases, spatial jitters were smaller than the physical size of salamander rods (10 µm). We conclude that firing events induced by fixational eye movements are sufficiently precise in time to permit high-resolution spatial information to be recovered from temporal patterns of activity across RGCs. Supported by the Wellcome Trust and E. Mathilda Ziegler Foundation.
Keywords: 406 eye movements • 557 retina: proximal(bipolar, amacrine, and ganglion cells) • 620 visual acuity