Surface Perception of Lightness in Different Contexts
36.4003, Sunday, 17-May, 2:45 pm - 6:45 pm, Pavilion
Christiane Wiebel1, Manish Singh2, Marianne Maertens1; 1Modelling of Cognitive Processes Group, Technische Universitaet Berlin, 2Department of Psychology & Rutgers Center for Cognitive Science (RuCCS), Rutgers University
We recently showed that a contrast-based model of lightness perception outperformed a number of other models in predicting lightness matches across different viewing conditions (Zeiner & Maertens, 2014; Maertens & Shapley, 2013). By considering regional variations in contrast range, the model successfully predicted lightness constancy across changes in illumination and for different transparent media. Here, we tested the model's predictive power for stimuli that involve a systematic variation of the surface reflectances that surround a target surface. Such changes – in the absence of illumination changes – are expected to produce failures of observers' lightness constancy which are known as type II constancy errors. We tested two exemplars of contrast-based models, the scaling contrast model (Zeiner & Maertens, 2014) and the contrast-ratio model (Singh, 2004). We used rendered images of custom-made checkerboards to test the perceived lightness of twelve surface reflectances as a function of the viewing context (plain view vs. light and dark transparency) and of the reflectances surrounding the target. The reflectances of the surround checks were selected so as to either vary around the mean luminance of all checks (average condition), or so as to vary around a relatively low or high luminance, resulting in extreme contrast values between target and surround. Observers matched the perceived surface lightness of the target check presented in the above conditions by adjusting the intensity of an external test probe. Both models predicted the observed lightness matches well across different viewing contexts (plain view vs. transparency). In contrast to the model predictions, we observed only small behavioral effects of the surround manipulation. Based on inspection of the extreme surround stimuli, we suspected that the local darkening or brightening of the adjacent checks, was itself sufficient to signal a different viewing context. An appropriate model adjustment indeed yielded a much better fit to the data.