Use of Gabor elementary functions to probe receptive field substructure of posterior inferotemporal neurons in the owl monkey.

The large receptive fields of inferotemporal neurons in the owl monkey were studied with visual stimuli whose luminance profiles were one-dimensional Gabor functions, i.e. sinusoidal gratings within Gaussian envelopes. The members of one set of such patterns all had a full bandwidth at half-amplitude of 0.8 octaves, but different center frequencies and spatial extents. These spatially restricted stimuli were ideal for determining whether one or more than one spatial frequency band projected onto discrete subsections of the neuron's receptive field. The other set of Gabor stimuli comprised sine waves within Gaussian envelopes of constant size, but with different center frequencies and hence different bandwidths. These stimuli allowed assessment of the neuron's spatial frequency selectivity across the full breadth of its receptive field. Results suggest that only one orientation band and one spatial frequency band provide an input onto each inferotemporal neuron under our experimental conditions. The preferred spatial frequencies found (0.2-0.6 c/deg) were all in the very low spatial frequency range for this animal. Calculations show that about 3.5-7.0 full cycles of the optimal grating usually cover the full width of the receptive field, but the observed spatial frequency tuning is not nearly as sharp as that which would be predicted according to phase coherent linear summation. Moreover, at the preferred spatial frequency, the peak response to gratings in the constant aperture series was generally less than the response to the same preferred spatial frequency in the constant relative bandwidth series. These results suggest either incomplete phase coherent summation from contributing subgroups, non-linear processing, or both.