Abstract
Lightness contrast and assimilation are both observed in matching experiments performed with disk-and-ring displays, where the two opposite-signed induction effects are part of a larger quantitative pattern in which lightness varies as a parabolic function of surround luminance (Rudd, JOV, in press). Here I show that a parabolic pattern is also obtained with manipulations of surround size; and that both parabolic effects result from a long-range perceptual grouping process. Specifically, target lightness varies as a function of the squared difference between the surround luminances (in log units), or surround widths, on the target and matching sides of the display. These squared differences reflect biological definitions of luminance and size similarity, respectively. The parabolic pattern is enhanced (i.e. grouping strength increases) when an observer is biased to see the target and matching disks as being co-illuminated, either by presenting the target and match configurations against a common white background or by instructing the observer to assume a common illuminant (with a black background). The parabolic pattern disappears when the observer is biased to see the disks as being separately illuminated by both presenting the configurations against a common black background and instructions to assume separate illuminants. With the black background, but no special instructions, the results are mixed. To account for the results, I propose a model in which the observer can adopt either of two strategies for matching the disks. If the two disks are interpreted as being independently illuminated, the observer adopts a ratio-matching strategy. If disks are interpreted as co-illuminated, the observer adopts a strategy involving long-range grouping. In the most general case, the matches are based on a weighted sum of the two strategies. Combining this model with the idea that visual similarity depends of squared differences accounts for the parabolic matching functions and lightness assimilation.