Contrast gain control is drift-rate dependent: an informational analysis

Neurons in the visual cortex code relative changes in illumination (contrast) and adapt their sensitivities to the visual scene by centering the steepest regions of their sigmoidal contrast response functions (CRFs: spike rate as a function of contrast) on the prevailing contrast. The influence of this contrast gain control has not been reported at nonoptimal drift rates. We calculated the Fisher information contained in the CRFs of halothane-anesthetized cats. Fisher information gives a measure of the accuracy of contrast representations based on the ratio of the square of the steepness of the CRF and the spike-rate dependency of the spiking variance. Variance increases with spike rate, so Fisher information is maximal where the CRF is steep and spike rates are low. Here, we show that the contrast at which the maximal Fisher information (C(MFI)) occurs for each adapting drift rate is at a fixed level above the adapting contrast. For adapting contrasts of 0 to 0.32 the relationship between C(MFI) and adapting contrast is well described by a straight line with a slope close to 1. The intercept of this line on the C(MFI)-axis is drift-rate dependent, although the slope is not. At high drift rates relative to each cell's peak the C(MFI) offset is higher than that for low drift rates. The results show that the contrast coding strategy in visual cortex maximizes accuracy for contrasts above the prevailing contrast in the environment for all drift rates. We argue that tuning the system for accuracy at contrasts above the prevailing value is optimal for viewing natural scenes.