Spike timing precision and visual signaling in primate retinal ganglion cells

Valerie J. Uzzell and E.J. Chichilnisky

University of California, San Diego
The Salk Institute

Recent findings indicate that retinal ganglion cell (RGC) spike trains in salamander and rabbit can exhibit striking timing reproducibility (a few ms) in response to repeated stimuli. How important is the precise timing of RGC spikes for conveying the temporal structure of visual stimuli to the brain? Using multi-electrode recordings from isolated retinas of macaque monkeys, we have begun to determine how accurately time-varying stimuli can be (a) classified or (b) reconstructed based on the timing of spikes. In (a), spike trains elicited by a brief flash and by an identical delayed flash were represented as vectors in a multi-dimensional space and classified using linear discriminant analysis. Classification performance was examined as a function of flash delay to determine what delay was sufficient to reliably distinguish the flashes. In (b), RGC spike trains were used to linearly reconstruct the time course of a rapidly fluctuating white noise stimulus. Spike times were then artificially jittered by a variable amount to determine how reconstruction accuracy declined with reduced spike timing precision. Temporal precision assessed using (a) tended to increase with contrast, and tended to decrease with the integration time of the cell's light response measured with reverse correlation, as would be predicted from simple models of the light response. Preliminary results indicate that some primate RGCs can reliably distinguish two flashes occurring 20-30 ms apart. By this measure, the temporal precision of visual signaling by RGCs may be coarser than has been proposed, at least in primate retina. Preliminary results suggest that temporal precision assessed with method (b) was comparable.