Abstract
How does the visual system assign lightness to surface patches seen through partially-transmissive surfaces? We employed asymmetric lightness matching using stereoscopic transparency displays with textured backgrounds. Observers adjusted the luminance of a comparison patch seen through a transparent surface, in order to match the lightness of a standard patch seen in plain view. Plots of matched-to-standard luminance were linear, and the pattern of slopes—obtained via matches through different transparent surfaces—was consistent with Metelli's alpha. Moreover, the matches revealed a high degree of constancy (80 – 96%). A control experiment confirmed that these matches were indeed transparency based, and could not be explained by simultaneous contrast or anchoring.
The consistency with Metelli's alpha is surprising given that when observers directly match the transmittance of transparent surfaces, their matches deviate strongly and systematically from Metelli's alpha, and are consistent instead with the ratio of contrasts (Singh & Anderson, 2002). Formal analysis reveals, however, that there is a deeper mutual consistency between the two sets of results. In particular, we demonstrate that the success of Metelli's model in predicting lightness through transparency, as well as its failure to predict perceived transmittance, can both be understood in terms of contrast-based mechanisms.
Researchers have often expressed surprise (and/or skepticism) concerning why a rather simplistic physical model (i.e., the episcotister setup) should yield perceptually valid equations. Our analysis provides a principled reason—based on known properties of visual mechanisms rather than a physical model—for the perceptual validity of Metelli's equations in capturing lightness through transparency. The analysis also suggests a solution to the related problem of lightness constancy through veiling luminance (Gilchrist & Jacobsen, 1983).