Inhibition in cortical circuits

The observation that neural circuits of the neocortex are adapted to many dif-ferent tasks raises deep questions of how they are organized and operate. Most theories of cortical computation propose that the cortex processes its information in a feedforward manner through a series of hierarchically orga-nized stages and that each of these stages is dominated by the pattern of the input to the local cortical circuit. The most infl uential of these models of the local circuit is Hubel and Wiesel’s ( 1962) proposal for the circuits that under-lie simple and complex cells in the cat’s primary visual cortex. Felleman and Van Essen ( 1991) extended the notion of a processing hierarchy in their com-prehensive summary wiring diagram for the primate visual system. In these models of intra- and interareal cortical circuits, sensory information from the retina is passed through successive stages of cortical processing, each of which increases the feature selectivity of visual receptive fi elds. Thus, from the concentric center-surround receptive fi elds of the retina and dorsal lateral geniculate nucleus, simple cells are created, then complex cells from simple cells, and eventually the face cells, object-specifi c cells, and 3-D motion-specifi c cells of the high levels of the cortical processing hierarchy. This serial processing schema is conceptually simple, which makes it very attractive for theorists (e.g., Riesenhuber and Poggio, 1999). More recent experimental and theoretical considerations of the cortical circuits, however, have suggested a rather different architecture: one in which local circuits of cortical neurons are connected in a series of nested positive and negative feedback loops, called “recurrent circuits ” ( Fig. 2.1; Douglas et al., 1989; Douglas and Martin, 2004, 2007). Excitatory neurons outnumber the inhibi-tory neurons by 5 to 1, so this ratio might be expected to create an unstable positive feedback. However, because the recurrent connections also exis