Brightness estimation supports a two-channel model of achromatic brightness processing

Abstract

Brightness magnitude estimation functions for 3400° white circular increments on a 2700° white 38 cd/m2 background were measured on a modified Tubingen perimeter. The stimuli ranged from 1.7′ to 66′ in diameter, 0.07-1 s in duration, and 0° to 20° in eccentricity. Brightness estimates were obtained at 0.2 log unit luminance intervals from the increment threshold to 1600 cd/m2. A constant-luminance fixation target also served as a reference for the brightness estimates. Although the overall averages can be acceptably fit with power functions, most single-session functions are double-branched, in agreement with Hood and Finkelstein.1 A low-contrast branch first rises steeply with a slope ⪰1, but then asymptotes within 1–2 log units of threshold. A relatively straight high-contrast branch rises out of the low-branch asymptote, with a slope between 0.3 and 0.5. Changes of target size, duration, or eccentricity have little effect on the slope of the high-contrast branch. However, parameter changes that increase threshold can markedly shorten the range and increase the initial slope of the low-contrast branch. The properties of the low-contrast and high-contrast brightness branches are similar to those of the magnocellular and parvocellular retinogeniculate pathways and are consistent with existing brightness matching evidence for a two-channel model of achromatic brightness.2