Penn Arts & Sciences Logo

Home

Friday, April 22, 2016 - 12:00pm

IRCS Conference Room

Bruce Cumming
Intramural Research Program
NIH

Disparity coding in striate cortex and natural statistics

An important cortical function is to generate sensory signals that represent the outside world. To do this effectively, the representation should exploit knowledge of the statistical properties of the world. Binocular vision is a useful system to explore this specialization, because the projection geometry imposes some very strong statistical regularities. Given the image seen by one eye, the set of possible images seen by the other eye comes from a very restricted set. It has been known for some time that the responses of disparity selective neurons in the primate striate cortex reflect some of these constraints. When one eye is presented with an image which is a photographic negative of the other eyes image (anticorrelation), these neurons still respond to disparity (as expected from current models of the underlying mechanism), but these responses are attenuated relative to those elicited by correlated stereograms (which is not expected from the underlying mechanism). This attenuation represents an adaptation to the statistics of the outside world, because anticorrelation is not possible during natural viewing. What mechanisms produce this attenuation remains unclear.

One possibility is that the attenuation is produced by linear summation of several complex-cell-like elements. We recently explored this possibility with a very general method, exploiting an analysis of noise stimuli (spike-triggered covariance). Although this analysis did indicate multiple elements, they did not have the spatial structure necessary to explain the observed responses to anticorrelated stereograms.

This suggests that a more complex mechanism is responsible for the response attenuation. One possibility is that it involves the operation of recurrent circuits. If recurrent processing is responsible for response attenuation, then the earliest part or the response should not show attenuation to anticorrelation. In order to test this possibility, we examined the dynamics of the disparity response with subspace mapping. The results suggest that recurrent processing is not the explanation, but demonstrate a disparity-specific nonlinearity in temporal summation.

This complexity in the representation in striate cortex has important implications for how we think about the representation of a binocular image in the brain. Striate cortex is typically thought to perform a simple binocular cross-correlation, and subsequent processing in extrastriate cortex then elaborates more complex signals. Our observations in striate cortex imply that extrastriate cortex faces a simpler problem.