Models of the dynamics of neural populations.

Three requirements are posed for constructing models to simulate EEG dynamics. The element of the model should be an interactive ensemble of neurons and not single neurons. The observations must be statistical, such as EEG waves and averages of unit activity over time and over local neighborhoods containing neural subsets. The state variables and operations of the model must be clearly related to behavioral functions such as sensory reception and perception. A model is presented that exemplifies these requirements. Its key feature is the dependence of its levels of interaction on the level of its input, so that with each burst of input the model switches from an equilibrium state to a limit cycle state. A mechanism is described for coding sensory input into the spatial modulation of the limit cycle activity viewed as a carrier. It is suggested that sensory recepts and percepts exist at different hierarchical levels in the brain, recepts at the level of single neurones, and percepts at the level of neural ensembles, the latter being possibly manifested in the EEG.