Abstract
Neurons in the lateral geniculate nucleus (LGN) and primary visual cortex (V1) show contrast-dependent spatial frequency tuning: at low contrasts, they tend to prefer lower spatial frequencies. Proposed mechanisms include changes in the pooling properties of feed-forward filters, and the action of a recurrent, tuned gain control. It is unclear whether the cortical effects are computed de novo, or simply reflect the properties of thalamic input. We measured LGN and V1 responses to optimally-oriented sinusoidal gratings and grating mixtures presented at a range of contrasts. The mixture stimuli contained 3, 5, or 7 gratings at logarithmically-spaced spatial frequencies. We recorded well-isolated, single neurons in opiate-anesthetized paralyzed macaque monkeys. To test for linearity, we compared the response to grating mixtures with the summed responses to their components presented alone. Some cells’ responses closely matched this linear prediction. Most cortical cells and magnocellular LGN cells were, however, sublinear. As expected from single-grating contrast response functions, the sublinearity was greater for mixtures whose components drove strong responses, and cells with strong saturation in the contrast-response function showed greater overall suppression to mixtures. We wondered whether this sublinearity was spatial-frequency dependent. V1 neurons show more nearly linear summation for mixtures at or above the peak frequency than for lower frequencies, an effect well captured by our previously proposed V1 model of frequency-tuned gain control. Summation in LGN neurons – even those with contrast-dependent tuning shifts – showed little frequency dependence. This suggests that the contrast dependence of V1 neurons is not inherited from their LGN afferents.