A weighted-average model of achromatic transparency.

A model of achromatic transparency based on the idea that neural representations of transparency are activated by proximal contours is described. It is proposed that the weighted average of the magnitudes of the representations of transparency relative to a perceived continuous transparent surface corresponds to the judgement of the overall degree of transparency of the same surface. Tests of this weighted-average model were carried out with bistable patterns formed by two overlapping surfaces that appeared opaque where they were superimposed on the background and transparent where they were superimposed on each other (partial transparency). In agreement with predictions from the weighted-average model, the rated degrees of transparency of these two surfaces were noncomplementary and independent of background reflectance. Two experiments confirmed the contention of this model that the relevant proximal contours for the judgement of partial transparency of the two overlapping transparent surfaces in a bistable pattern correspond to the part where these surfaces are superimposed.