Neuronal information coding by oscillation phase prediction error: implications for consciousness and control of voluntary function.

Neuronal oscillatory firing patterns are widely thought to serve as coincidence detectors, so as to synchronize information processing across brain regions. I hypothesize here that, instead, oscillatory function can be better understood by reversing conventional models and regarding oscillations slower than 30 Hz as background activity occurring in the absence of conscious or voluntary function. Action potentials occurring out of phase with the local population oscillation then emerge as information carriers, according to a prediction error mechanism that was first described in the drug abuse literature. Shannon information calculations show that coding by out-of-phase action potentials is far more efficient than are conventional synchronization models, and a simple calculation of the degree to which an action potential is out of phase correctly predicts the amount of information content. This model appears to account for a wide range of existing experimental observations of visual attention, voluntary motor function, and movement disorders. It also suggests an intuitively simple way of understanding consciousness, based upon a "self-observing" feedback mechanism of neocortical neurons firing out of phase. The hypothesis also suggests that the function of slow-wave sleep may be to re-entrain desynchronized oscillations.