Abstract

Freeman's pioneering work -- and neurodynamics in general -- has largely ignored specification of an anatomical framework within which features of coherent objects are represented, associated, deleted, and manipulated in computations. Recent theoretical work suggests such a framework can emerge during embryogenesis by selection of neuron ensembles and synaptic connections that maximize the magnitude of synchrony while approaching ultra-small-world connectivity. The emergent structures correspond to those of both columnar and non-columnar cortex. With initial connections thus organized, spatio-temporal information in sensory inputs can generate systematic and specific patterns of synchronous oscillation, with consequent synaptic storage. The theoretical assemblies of connections resemble experimentally observed 'lego sets', while facilitation and interference among synchronous patterns, particularly when executed by fast synapses under metabolic entanglement, imply powerful parallel computation.