The reviewers would like to see a model that produces one clearly specified output for any particular input, with the rigid determinism of a digital computer algorithm. But Gestalt theory demonstrates that perception itself does not have this quality, because perception is fundamentally multi-stable, with the stability generally, but not always, limited to a few discrete perceptual states. In fact this is one of the most significant aspects of perception which requires explanation by computational models. For example figure 7 (a) can be perceived as either a rectangle sloping backwards in depth, or a trapezoid in the plane of the page, but it can also be perceived (with some difficulty) as an even steeper trapezoid sloping back towards the observer, among other perceptual interpretations. It is this very multistability and flexibility of the perceptual interpretation which requires explanation by the perceptual model, and the present model attempts exactly such an interpretation. For the gray shading in figure 7 (D) delimits the full range of possible perceptual interpretations of the stimulus of figure 7 (a). The different stabilities of these different interpretations would be determined by the intrinsic constraints of the system.

Figure 7

Again, this is not a complete model that accounts for all of the observed properties of perception, but a paradigmatic hypothesis that suggests the kind of mechanism that would be required to account for the Gestalt properties of perception. If the reviewers missed this essential point, there is really no hope for them!