Postscript: Qualifying and Quantifying Constraints on Perceived Transparency.

factors that modulate the perceived lightness and opacity of transparent surfaces. However, we believe that the core insights that shaped our model will be incorporated into future successful models of the perception of transparency. There appear to be two primary dimensions that capture the psychological experience of transparent surfaces: opacity and lightness (or, more generally, color). Physical models such as Metelli’s (1974a, 1974b, 1985) also appear to embody the same two dimensions, which is probably why the model garnered so much attention as a candidate model of perception. The problem is that the physical decomposition embodied in Metelli’s (1974a, 1974b, 1985) model does not accord with the perceptual experience of these dimensions—which is why we developed a new model to replace it. Albert (2008) claims he has shown our model to be incorrect without providing any understanding of the data for which the model has provided excellent accounts. We believe that the logic that has led him to this conclusion is flawed. Our model was motivated by the observed dependence of transmittance and lightness matches on mean luminance within the filter region. Even if we were to ignore the methodological problems discussed above, Albert’s data merely indicate that mean luminance can influence transmittance matches in a manner beyond that captured in computations of contrast. It is currently impossible to determine whether such dependencies will replace those captured in our contrast ratio models or will merely provide an additional factor that contributes to the perception of transparency. Finally, Albert (2008) offered no theoretical account of how transparent surfaces are perceived, nor does he attempt to model or provide any theoretical insight into his own data. Instead, he merely asserts that, for his task, our model fails to explain his data. All existing models fail to explain his data, so it is unclear why he is content with discussing his data only in the context of our model. Given his evidence, we are inclined to believe that this putative failure is not due to the shortcomings of extant theory, but rather, to the methods used in his experiments.