Abstract

A long-standing debate in perception concerns the question of whether perceptual organization is guided by internal efficiency or by external veridicality. This article focuses on the simplicity principle which is a modern information-theoretic version of Occam’s razor, but also compares it to the likelihood principle which reflects classical information theory. Both principles can be modelled by Bayes’ rule which combines quantifications of view-independent and view-dependent factors to predict stimulus interpretations. Whereas the likelihood principle relies on hardly quantifiable frequencies of occurrence of things in the world, the simplicity principle relies on better quantifiable structural complexities of individual things. The simplicity principle is further argued to be sufficiently veridical in everyday perception, and neurally realizable via cognitive architecture enabling transparallel processing. This quantum-like form of representational processing relates to connectionist modelling and complements dynamic-systems ideas about neuronal synchronization.