Simulations of signal flow in a functional model of the cortical column

We describe the simulation of a layered cortex model based on the cortical column as a generic local processor. It simulates the signal flow in the layers I-IV of a set of model columns across three hierarchical cortical areas. It demonstrates the fast formation of an initial stimulus hypothesis, and its subsequent refinement by inter-columnar communication. In this prototype simulation, we implement word recognition from a string of characters. The three cortical areas represent letters, syllables, and words, used as a metaphor for visual stimuli. Focusing on the intra- and inter-columnar dynamics, we show how the different processing subsystems interact in order to switch off expected signals and accomplish symbolic recognition of words, and how representations for new words can be constructed based on old representations (self-reference). (c) 2006 Elsevier B.V. All rights reserved.