Abstract
Several models of contextual effects on apparent contrast (contrast-contrast) have been proposed over the last 15 years, ranging from lightness induction explanations to detailed quantitative models of pooled contrast gain control processes. We tested the generality of several models on data collected in apparent contrast, contrast discrimination, orientation discrimination, and spatial frequency discrimination experiments. Apparent contrast data were gathered in a staircase matching procedure for a center patch of sinusoidal grating viewed alone or in the presence of surrounds that varied in contrast or phase. PSE values were determined for each condition, from which attenuation effects on the apparent contrast of the center patch could be determined. Discrimination data were gathered under similar stimulus conditions in two-alternative signal-detection rating tasks, from which d' values were calculated for the different surround conditions and judgment tasks. Model predictions for the discrimination tasks were generated by adapting models to predict relative d' values by the addition of some assumptions. The results suggest that a) no one model fits all of the data, even qualitatively, and 2) models relying on concepts such as lightness induction or single pooled gain control processes are too simple.