Abstract
White’s effect is an example of how the brightness of an image region depends both on its local luminance and on its surround. The effect is commonly measured using a matching task, where observers adjust the luminance of a test patch so as to match the brightness of a target. However, a perceptual match might not always be possible, as the contexts of test and target can produce perceptual differences that cannot be compensated by adjusting luminance. Furthermore, to fully characterize the effect, it needs to be measured along the whole range of luminances and not just for some intermediate values. A more comprehensive characterization of the effect will provide better constraints to models of brightness perception. Here we use Maximum Likelihood Conjoint Measurement (Knoblauch & Maloney, 2012), to measure brightness scales for White’s effect, as the method allows us to estimate scales as a function of target luminance and context simultaneously. Stimuli were parametric variations of White’s effect consisting of a black-and-white square-wave carrier grating with two targets placed on the black or white phases. On each trial the targets differed in luminance and phase placement (collinear with the same or with different phase), and participants reported which target appeared brighter. Results confirm the classical White’s effect: for equiluminant targets of intermediate luminance the one collinear with the black grating looks brighter than the one collinear with white. Brightness differences vary across the scale and are maximal for intermediate luminance values. Furthermore, perceptual scales for white and black grating phases are compressive, non-linear functions of luminance. Our findings provide an augmented test bed for evaluating computational models of brightness perception, as accurate models should be able to predict the entire range of a perceptual dimension.