Abstract
The perceptual constancies, of brightness, shape, color, and size, are thought to be critical for object recognition; objects should in some sense ‘look the same’, even when the conditions of lighting and viewing are altered. Yet the constancies are imperfect, their extent differing over individuals, tasks, and situations. In the case of color constancy, variations are marked, from near zero to near 80% of the ideal, even when the observers are tasked to match object colors rather than lights at the eye (Arend & Reeves, 1986). Is this imperfection just due to random variation, or is there a principled reason? Reeves (2018) argued that as the color of a sample becomes purer, the extent of color constancy should decline, whereas that of brightness constancy should improve, based on an analysis of the underlying physics. We have re-analyzed extensive raw data from color and brightness matches made across differently illuminated ‘Mondrian’ displays by the 20 observers of Foster, Amano & Nascimento (2001), and have found support for both theoretical predictions. Accounts of constancy need to consider sample saturation as well as instructions, even given the inevitable individual differences.