Abstract
The response characteristics of retinal ganglion cells are often studied in anesthetized animals with paralyzed eyes. Under natural conditions, however, the input to the retina is also modulated by the self-motion of the observer. Eye movements continually occur, even during visual fixation, when microscopic eye movements incessantly shift the retinal image and modulate visually-evoked responses in ganglion cells. It is, therefore, critical to understand how neuronal response properties determined in the absence of retinal image motion apply to natural viewing conditions. In this study, we describe an “equivalent filter” of retinal ganglion cells, a filter which incorporates the statistics of human fixational eye movements and the cell response characteristics measured with an immobile stimulus in order to estimate the response properties of ganglion cells under the normal instability of visual fixation. Traces of oculomotor activity recorded from human observers while maintaining fixation on a small marker were used to reconstruct the typical spatiotemporal input to the retina during natural fixation. This input signal was then processed by filter models of parvocellular (P) and magnocellular (M) cells designed on the basis of published data from neurophysiological experiments with anesthetized and paralyzed macaques. We show that neuronal sensitivity to time-varying inputs shifts towards higher spatial frequencies during fixational eye movements. That is, the peak sensitivity of the equivalent filter occurred at a spatial frequency which was more than twice that of the contrast sensitivity function measured in the neurophysiological experiments. This effect occurred for both P and M cells at all the considered visual eccentricities. Thus, contrast sensitivity functions measured in the absence of eye movements seriously underestimate neuronal sensitivity to high spatial frequencies under normal viewing conditions.
Supported by NIH R01 EY18363, NSF BCS-0719849, and NSF IOS-0843304.