Abstract
In the past we have shown that a substantial part of the early ERP responses and responses in early visual cortex up to area LO can be well understood on the basis of statistics derived from the distribution of contrasts in an image. One of these statistics, spatial coherency (Groen et al., 2013), describes images on the basis of a range from structured to Gaussian. Images on the structured side typically have a strong figure/ground segmentation often depict man-made scenes, images on the Gaussian side are fractionated and typically depict nature scenes or contain scrambled content. Here we explore to what degree we can understand responses in area LO and ventral cortex using such statistics. We ask the question whether responses are better understood using statistics related to the spatial structure of an image or whether they are better understood on the basis of a categorical contrast of for instance image vs. scrambled image. We show that activity in the posterior part of LO is better explained by the SC statistic, activity in the more anterior part of LO is better explained by the categorical contrast of image vs. scrambled.
Meeting abstract presented at VSS 2015