The nature of conflict resolution and the notion of energy are very clearly described. Different perceptual interpretations are represented by different dynamical states of the system, and those states in turn are defined by multiple local forces, like the local coplanarity and orthogonality constraints that are expressed as local field-like interactions. As in the case of a physical soap-bubble, of which this system is a computational analog, the condition of prägnanz is defined by a low energy state of the system, i.e. a state representing a dynamic attractor, which is represented by a compromise whereby the maximal number of local field-like forces e.g. coplanarity and orthogonality constraints, are satisfied in the maximal number of local elements, exactly as occurs in the soap bubble. Not all of the forces corresponding to prgnanz have been described in this presentation, but the few local forces that have been described illustrate the general principles behind this alternative approach to the problem.
I have included the following text in the new draft:
"Whether or not the preceding description offers an accurate account of the spatial percept resulting from these particular stimuli, is again irrelevant to this discussion. The point is that whichever way the spatial percept is observed to occur, that is the way that the model should be designed to operate also. In other words the parameters of the model, such as the mathematical form of the local interaction fields should be adjusted to match the observed properties of the spatial percept, and this is most easily done in a fully spatial context. This may seem trivially obvious, and yet in the conventional approach this immediate spatial aspect of the percept is generally ignored altogether, as if it were irrelevant to perception. In other words, the effort to adhere to accepted neurocomputational principles has led neural modelers to ignore aspects of perception such as spatial interpolation, that cannot readily be explained in neural terms, even though the percept itself can be described with considerable precision. The great benefit of the perceptual modeling approach therefore is that it liberates models of perception from current limitations in our knowledge of neurocomputational principles. Furthermore, this approach demonstrates how the perception of three- dimensional form can be considered as a low-level perceptual interaction rather than a high level cognitive inference, and can be designed to produce a fully spatial reification of the perceived surfaces.