Gamma Oscillation Model Predicts Intensity Coding by Phase Rather than Frequency

Abstract

Gamma-frequency electroencephalogram oscillations may be important for cognitive processes such as feature binding. Gamma oscillations occur in hippocampus in vivo during the theta state, following physiological sharp waves, and after seizures, and they can be evoked in vitro by tetanic stimulation. In neocortex, gamma oscillations occur under conditions of sensory stimulation as well as during sleep. After tetanic or sensory stimulation, oscillations in regions separated by several millimeters or more occur at the same frequency, but with phase lags ranging from less than 1 ms to 10 ms, depending on the conditions of stimulation. We have constructed a distributed network model of pyramidal cells and interneurons, based on a variety of experiments, that accounts for near-zero phase lag synchrony of oscillations over long distances (with axon conduction delays totaling 16 ms or more). Here we show that this same model can also account for fixed positive phase lags between nearby cell groups coexisting with near-zero phase lags between separated cell groups, a phenomenon known to occur in visual cortex. The model achieves this because interneurons fire spike doublets and triplets that have average zero phase difference throughout the network; this provides a temporal framework on which pyramidal cell phase lags can be superimposed, the lag depending on how strongly the pyramidal cells are excited.