Once again I get the impression that this reviewer is either being intentionally obtuse, or his mind is incapable of grasping more subtle spatial concepts. His logic exhibits the hard Boolean klunk-klunk character of a digital computer!
But the whole point of the present paper is that the logic built into perception is in some sense the polar opposite of the deterministic or positivist logic characteristic of the digital computer. This was in fact the principle thrust of Gestalt theory.
The information about the illuminant provided by a perceived surface is not of a definite or positivist nature, but of an indefinite or probabilistic nature. I propose that the visual system handles this kind of uncertainty by accumulating evidence from many different surfaces, in the manner of a three-dimensional histogram.
The information from a single surface is indeed in the form of a probability density cosine function, with a peak at the normal to the surface, and zero in the direction normal to the normal, i.e. in the plane of the surface, because no illumination can occur beyond that angle. The illuminant is less likely away from the normal, because it would take a brighter (and therfore more improbably bright) illuminant to account for illumination from far off-normal directions.
But the point is that the illumination profile is not deduced from a single surface, but from all surfaces simultaneously. In the presence of a large number of surfaces all at different orientations, the cumulative evidence projected from all of these surfaces simultaneously would produce an illumination profile with a sharper peak in the direction visible to all of them. I propose that this is the kind of unique spatial computation performed by perceptual processes.
Again I feel that it is a waste of time trying to convince this particular reviewer. Some people just don't want to understand beyond their normal sphere of thought.