Visual completion and intermediate representations in object formation

Abstract

Abstract Visual perception produces representations of complete objects despite fragmentary inputs. This chapter describes recent advances in understanding contour interpolation in object formation. Persisting controversies are resolved by distinguishing two processes. An automatic contour-linking process based on well-defined geometric relationships produces an intermediate representation. A subsequent scene description process combines various scene constraints to sustain, weaken, or delete contour linkages, producing scene descriptions and perceptual experience. The chapter presents data showing that this approach explains path detection, a poorly understood perceptual phenomenon, and connects it to modal and amodal completion. It also describes recent research showing that most scene constraints combine in a simple additive fashion. This research approach exemplifies the challenges of discovering intermediate representations in perception and illustrates the kinds of evidence that can reveal them. Both theories of perception emphasizing automated perceptual mechanisms and those invoking more open-ended interactions of constraints may be correct in describing different aspects of object formation.