The cell assembly, Mark III: transitions between brain states and the localization and generalization of function

A connectionist model of the cerebral cortex as a Markov chain is presented. The model reconciles localization of cortical functions, apparent in neuropsychological data, with the equipotentiality and diffusion of function that was noted in classical studies of learning. In addition, the model suggests that the specificity of some cortical activity can develop quickly from initial connections that are predominantly chaotic. It is not necessary to assume a high degree of "prewired" connections in the cortex of newborns. While the model does not require initial settings to be chaotic, neither does it require initial ordering stronger than chaos. Specifically, a large and initially randomly connected network of neurons rapidly develops several ergodically distinct subnetworks that operate independently. While the ergodic units operate independently and distinctly, they exhibit the property of equipotentially in that the specific events caused by an initial event are equally likely to be at any point in the ergodic set after very few synaptic events.