Abstract
The disparity energy model successfully accounts for many properties of disparity selective neurons in the primary visual cortex. According to this model, the shape of the disparity tuning curve should reflect the shape of the underlying monocular subunits. For complex cells,the Fourier amplitude spectrum of the disparity tuning curve should be closely correlated with that of the monocular RF. Two previous studies found only a weak correlation between preferred monocular spatial frequency and the peak in the Fourier transform of the disparity tuning curve (disparity frequency). The extent to which this was the result of sampling variation was not assessed. A systematic comparison of monocular selectivity to spatial frequency and disparity tuning was therefore undertaken. Circular patches of sinusoidal luminance gratings were presented monocularly to each eye at the preferred orientation. Disparity selectivity was measured with binocular random dot patterns, and the disparity was applied in a direction orthogonal to the preferred orientation (along the axis for which spatial frequency tuning was measured).
For many cells, the data fit well with the predictions of the energy model. However, the majority showed significant differences between the spatial frequency tuning and the Fourier transform of the disparity tuning. Most commonly the disparity tuning curve contained more power at low frequencies than was observed monocularly. In many cases a bandpass tuning response to monocular gratings was associated with a Gaussian (low-pass) disparity tuning curve. Also, many cells had disparity frequencies which were significantly lower than the preferred spatial frequency for gratings.
These data show that the existing form of the energy model cannot account for the disparity tuned responses of all V1 neurons. An initial stage of disparity energy detectors, followed by an appropriate combination of their outputs, could produce the results observed here.