Item

Abstract

Binding operations, the dynamic establishment of semantic relations between distributed computational results, are key functions of cognitive processes in the brain. A common binding strategy is anatomical convergence of feature-encoding neurons onto conjunction-specific cells that respond only if the respective constellation of features is present. It is argued that this binding mechanism needs to be complemented by a more flexible mechanism that allows the brain to define relations on the fly in order to account for the nearly unbounded ability of evolved brains to represent the ever-changing constellations of elementary features characterizing perceptual objects. Evidence is reviewed which suggests that such dynamic encoding of relations is achieved by synchronization of temporally structured neuronal responses and hence by convergence in the time domain. A hallmark of neuronal responses is their oscillatory patterning that concentrates discharges to a particular phase of the oscillation cycle. By self-organized phase locking of these oscillatory responses the discharges of neurons can be synchronized. This in turn jointly enhances the saliency of the synchronized responses and establishes unequivocal relations among them. Neuronal oscillations cover a broad frequency range but here mainly the high-frequency oscillations in the beta- and gamma-frequency range are considered as these have been investigated most thoroughly in the context of binding functions.