Abstract
Segmentation of second-order texture boundaries defined by regional differences in the properties of their constituent micropatterns is typically explained using Filter-Rectify-Filter (FRF) style models. The FRF model assumes that the rectified outputs of first-order linear filters which analyze local micropatterns is subsequently analyzed by a second stage of filtering which detects global differences in micropattern properties. For a texture boundary defined by two varieties of micropatterns, in the present study ones varying in orientation and contrast polarity, a version of the FRF model in which different variety-specific channels are analyzed separately by different second-stage filters makes the prediction that segmentation thresholds should be identical in two cases: (1) Boundaries with an equal number of micropatterns on each side but different proportions of each variety (feature-defined boundaries) and (2) Boundaries with different numbers of micropatterns on each side, but with each side having an identical number of each variety (density-defined boundaries). We tested the segmentation thresholds for feature-defined and density-defined second-order boundaries for textures comprised of (1) horizontal and vertical odd-phase Gabor functions or (2) positive and negative polarity DC-balanced difference-of-Gaussian (DOG) functions. We find lower segmentation thresholds in both cases for density-defined boundaries than for feature-defined boundaries. This rules out, at least for these stimuli, the simplistic version of the FRF model in which the only information available to the second-stage filters comes from first-stage filters responsive exclusively to each micropattern variety. Control experiments controlling for RMS contrast demonstrated that density-based segmentation is not a simple artifact of contrast sensitivity. We suggest that density boundary segmentation can possibly be explained by the existence of first-stage mechanisms which generalize over micropattern varieties, so that the outputs of these first-stage mechanisms can be subsequently utilized by second-stage filters.