Abstract
Gillam and colleagues (e.g. Gillam and Grove, JEP 2011) found that (monocular) subjective contours along linearly aligned edges are much stronger when the inducers vary, for example in size, separation and orientation, than for regular inducers. A smooth linear alignment for otherwise unrelated inducers ("entropy contrast") is powerful evidence for occlusion. However the methods used, such as paired comparison and ratings, may allow a possible subjective influence. Here we use similar figures, each comprised of 5 aligned rectilinear shapes, in a stereoscopic occlusion detection task, which provides an entirely non-subjective evaluation of the power of entropy in inducing subjective contours along linear alignments. Thresholds for detecting the presence of a subjective occluder in depth were obtained using a Quest procedure with two alternative forced choice. One of two successive stimuli on each trial had the depth step. The planar inducing figures were tilted 45 degrees with a 45-degree alignment. Depth steps were created by a small expansion of the inducers in one eye in a 45-degree direction producing both vertical and horizontal disparities at the alignment. Thresholds were obtained for three sets of inducers (a) orthogonal to the alignment and completely regular in size and separation (b) orthogonal but irregular in height, width and separation (c) irregular in these properties as well as orientation. Support ratio and size/orientation properties were controlled for. For inducers varying in orientation, both disparities varied across inducers, adding depth entropy to other entropies. Mean depth thresholds across the 9 observers were 10.5, 8.3 and 3.6 arcmin respectively for the 3 inducer types with an ANOVA showing all differences to be highly significant. Thus depth detection thresholds decreased with increasing inducer entropy. In a second experiment we showed that the entropy effect could not be accounted for by the relative magnitudes of vertical and horizontal disparity
Meeting abstract presented at VSS 2014