Abstract
We perceive and represent the overall shape of an evergreen tree despite many deviations and irregularities in branches and needles along its boundaries. This sort of phenomenon is hard to understand if all contour information is combined into a shape percept. We propose a new hypothesis about separate systems for processing local contour information and global information about overall shape. These systems are independent and process information differently. Whereas the global system encodes low frequency contour variations accurately, the local system encodes only a small set of summary statistics to describe typical features of high frequency variations along a contour. In Experiment 1, we tested this hypothesis by sequentially showing two shapes that could differ in local features, global features, or both. Participants showed low sensitivity to changes in local contour features and a lack of additivity for shapes that differed in both local and global features compared to shapes that differed globally. In Experiment 2, we controlled for the amount of physical dissimilarity between local and global shape changes. Sensitivity remained higher for global features with physical similarity equated. In Experiment 3, we compared participants’ sensitivity to new sets of contour features with matched statistical properties with new features that differed in frequency and amplitude. Sensitivity to change was higher when statistical properties changed than when new features were generated from the same statistical distribution. We directly tested our hypothesis of the independence of local and global properties in Experiment 4 using a visual search task. Although local and global shape differences each produced pop-out out on their own, search when they were integrated together required focal attention. Taken together, these findings support the notion that separate mechanisms process local and global contour information and that the kinds of information these mechanisms encode are fundamentally different.