Describing the evolution of cognition as an informational process, we want to compare the complexity and computational power of connectionist models of some cognitive functions to the maximal information transmitted in brain relevant genes during neocortex evolution.
In this paper we implement propositional logic as Boolean functions in neural networks and investigate what types of functions emerge in nets with specified architecture (feedforward vs. fully backcoupled) and random weights. For N=2,3 arguments the relative portions of Boolean functions are given as results of a Monte Carlo simulation.