Abstract
The early cortical response to a natural scene is composed of signals from many overlapping neural populations tuned to narrow bands of image content. Psychophysical studies equate these populations with perceptual channels that undergo numerous forms of cross-channel interaction. In particular, psychophysical data suggest that cross-channel suppression is biased toward lower spatial frequencies (SFs). The neural bases of such interactions are largely unknown. In this study, we sought a neural picture of this cross-SF suppression through the lens of visual evoked potentials (VEPs). The stimuli were derived from 64 source natural scene images that were each divided into four SF bands (centered at 8, 16, 32, and 64 cycles per image) and then recombined into all 16 possible permutations (including the ‘blank image’ permutation, four ‘single-band’ permutations, six ‘two-band’ permutations, etc.). While undergoing 128-channel EEG, participants (n = 18) viewed the resulting 1024 unique stimuli (subtending 18.5deg diameter at fixation) for 500ms at their original contrast, and gave a perceived-contrast estimation response for each trial. Within a sliding window, we modeled the multi-band stimulus VEPs by combining the single-band VEPs according to a family of related rules, including some expressed as divisive gain control models with variable input and output weights. According to the gain control rules, responses to higher-SF image content contribute disproportionately more to cross-frequency suppression, while responses to lower-SF content are disproportionately subject to suppression. The SF-dependence of suppression was pronounced during earlier epochs (50–150ms post-stimulus); after 200ms, suppression appeared evenly weighted across SF. In sum, the modeling results suggest a high-to-low-SF flow of cross-frequency response suppression in the early VEP response to natural scene stimuli. The known pattern of perceptual suppression therefore seems to be shaped very early in the visual response to complex scenes, possibly serving to bias perception towards higher SF content.
Acknowledgement: Templeton Foundation grant to the CSC, James S McDonnell Foundation (220020439) grant to BCH