Neurons, glia and the borderline between subsymbolic and symbolic processing

Abstract

Our goal is the exploration of the nature and development of the functional borderline between non or subsymbolic processing and symbolic processing. Newell attributes the existence of a physical symbol system to the fundamental restriction imposed on the amount of information represented and processed in a local neural area by the limited energy available. Symbol tokens overcome the restriction by providing distal access to another local area and further energy while maintaining the linkage necessary for integrated information processing. Critical to further specification of this account of the origins of symbolic processing is clarification of the nature and capabilities of subsymbolic processing proceeding within a local neural area. Contributing to this clarification is our current objective.

We focus on the local cortical neural mechanisms at the millimetric level underlying the acquisition and operation of specific cognitive performance. Our account of the nature of local neural processing significantly deviates from the current norm in assigning a critical information processing role to glia, another type of brain cell, and their interaction with neurons.

Glia appear to possess a type of intracellular and intercellular calcium dynamics which provide a basis of excitability for signaling between them. This raises the possibility that glial networks engage in information processing with very different temporal and spatial characteristics from neuronal signaling. The potential performance capabilities of subsymbolic processing in local neural areas are clarified in a description of mechanisms involved in glia-neuron interaction drawing upon a wide range of neurochemical research. A computational version of the glia-neuron (GN) model has been implemented to permit assessment of local performance capabilities. The results are discussed in terms of their potential significance for cognitive science and artificial intelligence. The complexity of neural-glial interaction in local areas suggests that much specific cognitive processing can occur without the need for symbol tokens to provide distal access to other local areas. The principle that the units in the physical symbol system refer to relatively elaborate local subsymbolic processing areas may assist in explaining the effectiveness of symbolic processing in spite of the constraints of slowness, seriality and limited memory size imposed by its dependence on neural networks. Clarification of the nature of local glial-neural processing and its implications for the relationship between subsymbolic and symbolic processing will assist in the construction of hybrid systems. Our work, also, has the potential to contribute to extension of the biological metaphor underlying ANN.