A computational generative model for long term memory based on an intentional analysis of knowledge structures is discussed together with a representation of information processing as a probabilistic set of sequential and parallel transformations of decision functions which produces a random walk along a propositional state space.