In this article we describe some functional properties of the model of a somatosensory cortical macrocolumn—the segregate—described in the preceding companion article. These functional properties emerged in the model network in the course of stimulus-driven self-organization of its afferent connections under control of short-range inhibitory and longer-range excitatory lateral interactions among its minicolumns. In general, self-organization leads the model network to develop complex, nonlinear functional properties, and makes its neurons sensitive to the shape and temporal features of peripheral stimuli. The properties acquired reproduce some of the known properties of somatosensory and visual cortical networks. In particular, it is shown that, even though the network is exposed only to stationary point stimuli during self-organization, neurons in the model still acquire the ability to discriminate the direction of a moving stimulus, as well as the orientation of a stationary bar stimulus. Different stimulus directions and orientations are represented by different neurons in the model network, and the maps of neurons having these preferences have many properties in common with real cortical maps. In addition, we demonstrate the model network's ability to discriminate among spatially complex stimuli, such as letters of the alphabet. The Parallels between the emergent structural and functional properties of the model network and the properties of sensory neocortex suggest that the model captures some of the basic mechanisms by which sensory cortical modules develop and maintain their elegantly detailed and appreciable information-processing capabilities.

You do not currently have access to this article.