Abstract
A fundamental problem in vision science is determining the nature of the computations and representations underlying lightness perception. One view is that lightness perception involves the explicit decomposition (or scission) of images into causal factors, with separate estimates of illumination, reflectance, and transparency. An alternative view asserts that the image is separated into 2-D frameworks, and principles of anchoring and grouping are used to determine relative lightness. According to the frameworks view, there is no need to posit distinct layers in order to estimate lightness. We have recently shown that layered image representations (scission) can play a decisive role in lightness perception in conditions of inhomogeneous transparency (Anderson & Winawer, 2005). We argued that the contrast relationships along borders play a key role in causing scission, which in turn was responsible for the lightness illusions in these displays. To further assess the causal role of scission in these illusions, we manipulated the contrast relationships along target/surround borders by simply introducing a thin grey outline around the targets. These outlines cause polarity reversals along the boundaries of the target/surround regions, and are therefore inconsistent with transparency. We find that these manipulations completely abolish the lightness illusions and percepts of transparency in these displays. We argue that these phenomena, in conjunction with demonstrations reported by Hering and Mach, and more recently by Kennedy & Bai (2000), provide strong evidence that the visual system computes lightness by decomposing images into layered representations.