Spatial-frequency-specific inhibition in cat striate cortex cells.

Responses to single and multiple spatial frequency gratings were recorded from eighty-eight cat striate cortex cells. A cell's response to a grating of its optimum spatial frequency (f) was examined both alone and in the presence of gratings of 1/4, 1/3, 1/2, 2, 3 and 4f, respectively. Some 97% (thirty-seven of thirty-eight) of all simple cells showed significant inhibition of f by one or more of the other frequencies. This inhibition was usually fairly narrowly tuned, with only one or two spatial frequencies producing significant inhibition. Thirty-four simple cells were maximally inhibited by a higher frequency, three by a lower spatial frequency. By far the most common interaction was a considerable inhibition of f by 2f and/or 3f. Of the thirty-seven simple cells showing inhibition to a complex grating, seventeen responded in a manner dependent on the relative phases of the two components. Some showed only inhibition of f; in others, the response to f was either increased or decreased, depending on the relative phase of the two frequencies. The other twenty simple cells showed phase-independent inhibition: the inhibition was of approximately equal amplitude regardless of the relative phase angle of the two grating components. Such phase-independent inhibition cannot be accounted for by linear summation within classical cortical receptive fields. Only eighteen of forty-eight (38%) of the complex cells showed significant inhibition of f by one or more other spatial frequencies. Fourteen of these (29%) were maximally inhibited by a higher spatial frequency, four (8%) by a lower spatial frequency. Inhibitory interactions in complex cells were never dependent on the relative phase of the two component gratings. Six simple cells (16%) and fourteen complex cells (29%) showed significant facilitation of the response to f by one or more (most often lower) spatial frequencies. This enhanced response was greater than the sum of the responses to each component alone, was usually broadly tuned for spatial frequency, and did not depend on the relative phase of the two components. It thus differs from the increased response sometimes seen in a phase-dependent interaction. Some of the observed spatial-frequency-specific interactions are incompatible with either a strictly hierarchical model of cortical architecture or a simple linear filter model of visual cortical processing. The asymmetry of inhibition suggests that it subserves some function other than (or in addition to) the narrowing of spatial frequency tuning functions.