Abstract
The lightness of a target region is influenced not only by the contrast and contrast polarity of that region's border, but also by the contrasts and contrast polarities of other nearby borders. Recent studies have sought to model the process by which the effects of multiple edges that together influence the target lightness are combined spatially. Such models have typically assumed that the relative weights given to the target edge and other nearby edges are determined solely by low-level factors, such as the distance between the target and the inducing edge, and the edge contrast polarities. Here I show that the weights assigned to borders in perceptual edge integration can vary in predictable ways, depending on the particular lightness judgment that an observer is asked to perform, and on the particular information that the observer attends to in the stimulus. Subjects matched two incremental disk-and-ring stimuli in terms of their brightness (perceived luminance), lightness, or brightness contrast. Two lightness conditions were run in which changes in the luminance of the ring surrounding the target were viewed as either reflectance changes or as illumination changes on the target side of the display only. The matches made in the two lightness conditions differed in a manner consistent with a perceptual edge integration model in which the weights given to the various edges are adjusted in sensible ways depending on the observer's assumption about the nature of the illumination. I will argue that an edge integration theory that includes top-down control of the edge weights has the power to form a bridge between previous edge integration models and lightness anchoring theories based on the notion of illumination frameworks; and that a theory in which edges are selectively integrated under the influence of attention also provides novel and important insights into other lightness phenomena, including phenomena involving perceptual transparency.