Abstract

How accurately do human observers perceive the properties of an achromatic transparent filter with both reflective and transmissive components? To address this question, a novel six-luminance stimulus was employed, consisting of three transparent layer luminances set against three background luminances, which satisfied the conventional constraints of perceptual transparency. In one experiment, subjects adjusted one of the three layer luminances to complete the impression of a uniform transparent disk. It was found that the luminance-based formulation of Metelli's episcotister model and a model based on ratios of Michelson contrasts best predicted the subjects' settings, which were both accurate and precise. In another experiment, pairs of stimuli selected from a range with various values of the adjustable layer luminance were presented in a series of forced-choice trials. A modified implementation of the pair comparisons method was employed to recover the distribution that describes each subject's preference pattern. Results showed that there exists a reasonably wide range of stimuli that give rise to at least some degree of perceived transparency.