The spatial basis of neural representation

The spatial organization of cortical neural activities is one of the most important elements linking neural network models and natural neural networks. The columnar organization indicated by several works suggests a strong spatial constraint on neural network functions. Recently, attention has grown on the correlation between a vigorous firing of a particular neuron and the "supporting" neural firings in the cortical vicinity. Here, we investigate the relation between a vigorous firing of a particular neuron and the surrounding neural activities using a generic neural network architecture. We analyze the the spatial firing pattern in the vicinity of a particular neuron when the neuron is firing vigorously (preferred cortical state, PCS). We find that the spatial heterogeneity of the PCS is more pronounced in the both extreme ends of activity scale, and not so much in the intermediate activity region. This result, taken together with an analysis of the reaction of the network to a dynamic stimulus, put a strong constraint on the computational capabilities of a neural network. We discuss the implications of our result for the spatial constraints imposed on the neural information processing.