*3.2. Simplicity*

Compared to the likelihood principle, the simplicity principle is less of a modeling principle and more of an explanatory principle. By this, I do not mean to claim that simplicity explains all contrast polarity phenomena. For instance, in Glass patterns, simplicity predicts stronger moiré effects for identical dipoles than for nonidentical ones, which may often be adequate but, as indicated, not in the case of Figure 2c,d. The point is that I consider simplicity to be a fundamental force in perception, which nevertheless—just as gravity in physics, for instance—interacts with other forces, yielding results that now and again may deviate from what simplicity on its own would yield.

In this respect, notice that the contrast polarity reversal in Figure 2b can be said to trigger local groupings which destroy the symmetry. It can also be said, however, to yield antisymmetry, which, on formal theoretical grounds within the simplicity paradigm, is predicted to be not one of the instantaneously detectable visual regularities [25,26]. The earlier mentioned reversal in dot patterns seems an exception to both rules [9]. Furthermore, the reversal in Figure 1 clearly implies that the parallelogram interpretation becomes less complex compared to the other two interpretations—provided one applies, unlike Pinna and Conti did, the simplicity idea correctly. This idea is specified next in some more detail (see Table 1, lower part, for a first impression thereof).

In both mathematics and perception research, the simplicity idea falls within a descriptive framework. It relies on regularity extraction to obtain simplest descriptive codes, which capture hierarchical organizations of given input. There are a few domain-dependent differences (see Table 1). Unlike algorithmic information theory in mathematics does, structural information theory in perception research employs a fixed descriptive coding language extracting theoretically and empirically grounded visual regularities, and it classifies things on the basis of the hierarchical organizations described by simplest descriptive codes (which are taken to reflect mental representations) [27–35].

> **Table 1.** Overview of probabilistic and descriptive frameworks.
