next up previous
Next: CohenFalk, Hopp, Tang, Up: No Title Previous: Bialek: How do

Carandin: Cortical gain control and spike encoding

New York University

A large body of data obtained from extracellular recordings in the monkey's primary visual cortex indicates the existence of a divisive gain control mechanism which determines both the size and the latency of the responses. While the simplest biophysical substrate for this mechanism would be shunting inhibition, the membrane conductance increases that would be required may well be unrealistic. An alternative substrate for gain control is the spike encoding mechanism, which occupies a strategic position as a bottleneck for the outflow of information from the cells.

To obtain a simple and robust model of the spike encoding mechanism, we performed intracellular in vitro experiments on regular-spiking cells in slices of the guinea pig visual cortex. We found that the transformation of injected current into output spike trains could be accounted for quantitatively by a model consisting of a static nonlinearity sandwiched between two linear filters. The first linear filter is a passive single-compartment model of the membrane. The static nonlinearity is a rectification stage, and is followed by a high-pass linear filter. The putative biophysical substrate of rectification lies in the threshold for Na activation; that of the high-pass filter lies in the various afterhyperpolarization K currents.

We speculate that the rectification stage is not a good candidate for the site of gain control because it does not affect the latency of the responses. By contrast, the final high-pass linear filter is the ideal site for gain control. Controlling this stage yields the desired effects on the size and latency of the responses.

Tony Zador
Tue Oct 22 16:34:57 PDT 1996