Modeling neocortical areas with a modular neural network.

The modular organization of neocortex has been speculated to have a role in the operation of memory retrieval. By introducing two modifications in the model considered by O'Kane and Treves (1992a, J. Phys. A: Math. Gen.), we have been able to increase the storage load limit and to destabilize the memory-glass states that marred the memory operation of the original model. The two modifications, a sparse rather than complete activation of cortical modules and a correlation between the patterns of activation and the underlying connectivity, are both consistent with available evidence.