Abstract
A number of recent studies have indicated that perceptions of lightness/brightness are determined by the probabilistic relationship between the luminances in the retinal stimulus and the possible physical sources (reviewed in Purves and Lotto, “Why we see what we do”, Sinauer, 2002). To date, these analyses have relied primarily on qualitative paradigms, or limited quantitative arguments to rationalize the lightness/brightness percepts elicited by Cornsweet edges, stimuli that elicit Mach bands, or the reduced cue conditions used in brightness scaling experiments. Considering the linkage between stimulus luminances and lightness/brightness percepts more broadly, however, the visual system must instantiate a more fundamental set of underlying statistical relationships to generate lightness/brightness percepts in any and all circumstances. Given the high dimensionality of real-world effects on the probability distribution of lightness/brightness sources, it seems inevitable that the relevant statistical instantiations entail the conditional probabilities of concatenations of luminance values in retinal images with respect to the underlying natural sources. Accordingly, we have explored whether anomalies of lightness/brightness can be explained in these terms by analyzing the distribution of luminance in 4200 images of natural visual environments in the so-called Netherlands database (hlab.phys.rug.nl). A large number of samples were generated using various templates in which the pattern of light was similar to the basic unit in several well-known lightness/brightness stimuli that generate unusual percepts (e.g., the Hermann grid, the Wertheimer-Benary pattern, White's stimulus and the criss-cross pattern), thus allowing us to compute the probability distribution of light in the ‘target’ area of these stimuli. The relative shifts in the probability distributions of luminance in these stimuli appear to account for the associated perceptions of lightness/brightness.