Retinal motion signals and limits on speed discrimination

Abstract

Sensing motion requires reading out spatiotemporal patterns of activity from ensembles of neurons in the early visual pathways. Most models of this process do not incorporate noise, spatial and temporal sampling, or nonlinearities present in retinal ganglion cell (RGC) light responses. To probe whether retinal or downstream processing places the ultimate limit on motion sensing, we compared psychophysical speed discrimination to the limits imposed by the ensemble responses of primate RGCs. 512-electrode recordings were used to measure ensemble responses of parasol (magnocellular-projecting) RGCs in isolated primate retinas stimulated with moving bars. Speed estimates were obtained from the spatiotemporal orientation of evoked RGC spike trains. Several estimation procedures were tested to ensure that speed estimates reflected the limits imposed by retinal processing. Retinal limits on speed discrimination were determined from the variance of these speed estimates. Psychophysical experiments were then performed with stimuli matched for intensity, contrast, speed, retinal eccentricity and spatial extent. Subjects compared the speed of sequentially presented bars in a 2AFC task. Trial duration was eliminated as a speed cue through randomization. Psychophysical speed discrimination accuracy varied with speed, as did the variance of retinal speed estimates. Psychophysical speed discrimination thresholds were several-fold higher than expected from the variance of retinal speed estimates; this ratio was highest at low speeds. The variance of retinal speed estimates did not depend on the estimation procedure, suggesting that it reflected the limits imposed by the retina. These results suggest that central visual processing or eye movement — rather than retinal processing — limits speed estimation.

Footnotes

 We thank N Bezayiff, W Cunningham, W Dabrowski, JL French, AA Grillo, P Grybos, P Hottowy, S Kachiguine, K Mathieson, M Rahman, and VJ Uzzell; La Jolla Interfaces in Science, NIH MSTP GM07198, Merck Fellowship (ESF); McKnight Foundation, Sloan Foundation, NSF award PHY-0245104 (AML)