| In order to examine the response-tuned variability of cortical neurons, a reduced model of the upper layers of somatosensory cortex was developed. This model consists of a network of three segregates with 61 minicolumns each, that, in turn, have one excitatory and one inhibitory cell apiece. Each cell in the network was modeled as a single compartment electrical circuit with parallel conductances for channels typical to the upper layers of the cortex. The parameters of the neurons were adjusted so that the network displayed complex nonlinear dynamics similar to cortical networks. Further, lateral connections among the cells were included and allowed to self-organize based on Hebb's rule. Afferent input to the model was provided by a previously established model of layer 4 that incorporated realistic receptive field properties (Favorov and Kelly, Cerebral Cortex, 1994,4:408--427). Next, simulated moving stimuli were applied to the center of these segregate's RFs, and the activity of cells in the upper layers was monitored. Several experimental results were verified, including: (1) the characteristic shape of the coefficient of variation (CV) vs. velocity curve for optimally stimulated cells, (2) the wide diversity of tunings in individual cells' CV, (3) the diversity of tunings in firing rates to different velocities, (4) the correspondence between D'e and CV for directional sensitivity, (5) the velocity dependent decrease in CV with NMDA block, and (6) the firing rates for individual cells may increase with NMDA block. These results are shown to be highly dependent on the network operating in a complex (or to use the term loosely, chaotic) regime---even minute changes in network parameters that shift the system out of chaos drastically impair system performance. This result supports the theory of homeokinesis, which emphasizes the importance of a system maintaining a given dynamic range, and not with attaining a steady state (homeostasis). |
