I can hardly believe I am reading this!

Gestalt theory is not experimental data that needs to be explained by models!?!?

What Gestalt theory has shown, through a number of cleverly devised demonstrations, is that perception exhibits a holistic emergent global aspect that defies explanation by any kind of atomistic or local-processor theory of neurocomputation, as popularly conceived in neural network theory. This is a fundamental and very significant observation about the nature of perceptual processes which is very much in need of explanation by models! The fact that Gestalt theory itself never came up with a satisfactory neurophysiological or computational mechanism to account for these enigmatic properties of perception in no way invalidates the observations themselves, which are plainly evident to anyone who looks at a Gestalt figure.

Viewing a Gestalt illusion is exactly equivalent to a one-subject psychophysical experiment, with the viewer as the subject. The difference between psychophysics and viewing a Gestalt illusion is that in psychophysics, the subjective observation must be communicated to the experimenter through the bottleneck of a simple keypress response. In the case of the Gestalt illusion on the other hand, the subject is also the experimenter, and therefore the subjective experience due to the stimulus is plainly evident to the experimenter in the form of a rich spatial structure in conscious experience. The subjective appearence of the figure for a given stimulus is exactly the "result" or "data" generated by the experiment, and that data is exactly what requires explanation by models!

The real power and significance of this manner of investigation is that by bypassing the bottleneck of the simple keypress response, the data or evidence produced by the Gestalt experiment is so rich and complex as to defy the kind of simplistic modeling that is often invoked to account for psychophysical keypress data.

One problem which has consistently plagued models of Gestalt phenomena is the issue of which aspects of the illusion are explicitly represented in the neurophysiological mechanism of the brain, and which are simply encoded implicitly, or maybe not at all in the brain. This issue can never be resolved until an exact mapping has been established between a subjective experience and the corresponding neurophysiological state.

This is exactly why the perceptual modeling approach is so important, because it separates the question of the neurophysiological code of subjective experience from the information content of that experience. If a Gestalt illusion is experienced as a spatial structure, i.e. an image composed of edges and surfaces, then the output of the perceptual model should also be an image composed of edges and surfaces. This functional description of the perceptual transformation offers a concrete specification for the necessary requirements for a model to account for that perceptual function.

the following phrase ... does not really make sense "Gestalt theory also provides phenomenological evidence suggestive of some kind of top-down feedback,..." This treatment of a theory as a type of evidence is inappropriate.

I stand by that statement! Gestalt theory does provide evidence suggestive of some kind of top-down feedback. That evidence is in the form of visual demonstrations, introduced by Gestalt theory, which I find very suggestive of some kind of top-down feedback.