Purchase this article with an account.
DW Dong; Dynamic Temporal Decorrelation: A Theory of Saccadic Effects on the LGN Responses . Invest. Ophthalmol. Vis. Sci. 2002;43(13):3929.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: Our goal is to understand why lateral geniculate nucleus (LGN) has different responses between and during saccades. We explore the hypothesis that LGN improves the efficiency of visual representation by dynamic change of its response properties that allows to maintain the temporal decorrelation of natural time-varying images. Natural time-varying images possess significant temporal correlations which persist even after retinal processing. It has been shown that the LGN improves efficiency of visual representation through temporal decorrelation of the retinal signal (Dong and Atick 1995). But under natural viewing conditions, the temporal correlations are changed significantly by the saccadic eye movements and hence the visual signal sent to the LGN has quite different characteristics for the correlations during, across, and between saccadic eye movements. Methods: Because the statistics of the input signals depends on saccade timing and duration, maintaining temporal decorrelation requires that the LGN response also changes according to saccade timing and duration. Based on the recently measured statistical properties of visual input during free viewing of natural time-varying images (Dong 2001), we derive the dynamic temporal receptive-fields that achieve this decorrelation. Results: During and right after a saccade, the predicted receptive-field behaves as a temporal low-pass filter; whereas between two saccades, the predicted receptive-field behaves as a temporal difference (band-pass) filter. The predicted filter between saccades agrees quantitatively with that measured in experiments. The predicted filter during saccades has a smaller response to a stimulus at certain midrange temporal frequencies, as expected from saccadic suppression. However, it has bigger responses for stimuli at relative low and relative high temporal frequencies. Futhermore, the optimal temporal coding not only changes dynamically according to the relative timing of saccades but also adapts to different levels of noise and depends on the statistics of visual scenes. Conclusion: The theory of efficient coding gives a quantitative account for why LGN responses are different during and between saccades. It also shows that depending on the testing stimulus and on the recent adaptation to the visual scenes, the saccadic effects can be either suppression or enhancement of the LGN response during saccades.
This PDF is available to Subscribers Only