Abstract
Visual images in the real world are spatiotemporally broadband. Yet neurons at the earliest stages of visual processing respond to stimuli over a limited range of spatial frequencies (SFs), implying that the SF bandwidth of early visual motion detectors should also be expected to be rather narrow. To estimate the SF bandwidth, we used a 2-frame 1D vertical pink noise kinematogram, in which both frames were bandpass filtered, with the central SF of the filter manipulated independently for each frame. This study was inspired by Tim Ledgeway’s (1996) paper. To avoid spatial aliasing, there was no actual leftward-rightward shift of the image: instead, the phases of all Fourier components of the 2nd image were shifted by ±¼ wavelength with respect to those of the 1st (Quaia et al. 2017). We recorded the OFRs and perceptual direction-discrimination in human subjects. The OFRs showed a smooth decline in amplitude as the difference between the central SFs of the 1st and 2nd images increased. These dependencies were very well fit by (log)Gaussian functions. The standard deviations of the fits stayed the same as the central SF of the 1st image was changed 11-fold. On the contrary, the perceptual direction-discrimination performance was close to 100% correct when the central SF difference was small, deteriorating sharply to chance level when the difference was increased further. Allowing the subjects to grade the saliency of perceived motion moved perceptual dependencies closer to the OFR ones. The OFR data were well fit by a model incorporating power law summation of Fourier components’ contributions constrained by a finite bandwidth—0.51 and 0.65 octaves in 2 subjects—of early motion detectors. Thus, in addition to traditional studies relying on perceptual reports, the OFRs represent a valuable behavioral tool for studying early motion processing on a very fine scale.