Hypercomplex cells in the cat's striate cortex.

Abstract

.here has already been some discussion concerning hierarchical versus parallel processing of information in the striate cortex and I would like to add a few comments. Hubel and Wiesel,' in their classical papers, differentiated three categories of cells in the cat's striate cortex. They further suggested a hierarchical model which would explain the organization of the receptive fields of each type of cell. Thus only simple cells would receive a direct excitatory input from lateral geniculate neurons (LGN), and a number of simple cells with a common optimal orientation and slightly offset receptive-field positions would then provide the excitatory input to the complex cells. Finally, different complex cells with a common preferred orientation but again with offset receptive-field positions would provide excitatory and inhibitory inputs to the hypercomplex cells. The inhibitory input to the hypercomplex cell would explain that the sharp reduction in the discharge, as an optimally oriented stimulus, is elongated beyond the excitatory region in the receptive field. If the Hubel and Wiesel model is correct, one would expect a basic similarity between the responses of complex and hypercomplex cells. However, on the basis of the organization of their receptive field, we have been able to differentiate four classes of cells in the cat's striate cortex: simple, complex, and two distinct classes of hypercomplex cells. The majority of hypercomplex cells (Fig. 1, Type 1) have responses very similar to those of simple cells and their receptive fields can be subdivided into