Laminated Models of Within-Area Integration in the Neocortex

Adrian Robert, Martin I. Sereno,    September 2002


To lay groundwork for better testing theories of cortical function that may depend on interactions between multiple cortical layers, we constructed conductance-based spiking models of cortical circuitry with realistically differentiated cell types and lamination. Complexity, quantity, and inconsistency in anatomical data describing the circuitry were eliminated through application of constraint satisfaction and quantitative averaging techniques, which constitute a major contribution of this paper. In simulations, three dynamical regimes observed in slices were also observed in the model: transient response, wave propagation, and explosion. We found that, first, a `tiered' lateral axonal arborization structure (several classes of arborization width) is in fact required to accurately model the lateral spread of stimulation-evoked activity demonstrated in electrophysiological and optical recording experiments. By excluding medium and longer range projections, activity propagation becomes slower but more robust than is actually observed. The longer connections thus may be sufficiently strong to affect computational processes such as Hebbian learning, in which the scale of lateral interactions is known to be important. Second, the model showed partial independence of activity in superficial and deep cortical layers, which depends on the laminar restriction of pyramidal axonal arbors found in many areas and many species. This partial uncoupling between laminae has also been seen experimentally. Thus, an area is likely to perform two distinct horizontal integrations of information at one x-y point in its superficial and deep layers.
Back to online papers page.