ABSTRACT
Perceptual organization, broadly defined, is a set of visual processes that parses retinal images into their constituent components, organizing them into coherent, condensed, and simplified forms so that they can be readily interpreted and recognized. It generally includes many computational processes before object recognition, such as filtering, edge detection, grouping, segmentation, and figure-ground segregation. These processes are considered to be preattentive, parallel, and automatic (Treisman & Gelade, 1980) and mediated by feedforward and intra-areal mechanisms (Palmer, 1999). Attention, on the other hand, is thought to be driven by figure-ground organization, but not vice versa, even though some psychological evidence does suggest that later processes such as recognition and experience could influence earlier perceptual organization (Palmer, Neff, & Beck, 1996; Peterson & Gibson, 1991). The nature and the extent of top-down influence on perceptual organization thus remains murky and controversial. In this article, I first sketch a theoretical framework to reason about the computational and neural processes underlying perceptual organization. This framework attempts to unify the bottom-up organizational processes and the topdown attentional processes into an integrated inference system. I then discuss some neurophysiological experimental findings that lend strong support to these ideas.
