A mechanism for cognitive dynamics: neuronal communication through neuronal coherenceIntroductionFeedforward models of neuronal communication and their neurophysiological testsNeuronal communication through firing rate modulationNeuronal communication through modulations in the degree of spike synchronizationNeuronal communication through neuronal coherenceThe putative role of different frequency bandsSelective coherence for selective communicationCortico-spinal coherence subserves cortico-spinal communicationCoherence, competition and bindingAcknowledgementsReferencesA mechanism for cognitive dynamics:neuronal communication throughneuronal coherencePascal Fries1,21F.C. Donders Centre for Cognitive Neuroimaging, Radboud University Nijmegen, 6525 EN Nijmegen, The Netherlands2Department of Biophysics, Radboud University Nijmegen, 6525 EZ Nijmegen, The NetherlandsAt any one moment, many neuronal groups in our brainare active. Microelectrode recordings have characterizedthe activation of single neurons and fMRI has unveiledbrain-wide activation patterns. Now it is time to under-stand how the many active neuronal groups interactwith each other and how their communication is flexiblymodulated to bring about our cognitive dynamics. Ihypothesize that neuronal communication is mechan-istically subserved by neuronal coherence. Activatedneuronal groups oscillate and thereby undergo rhythmicexcitability fluctuations that produce temporal windowsfor communication. Only coherently oscillating neuronalgroups can interact effectively, because their communi-cation windows for input and for output are open at thesame times. Thus, a flexible pattern of coherence definesa flexible communication structure, which subservesour cognitive flexibility.IntroductionBecause we are equipped with mechanisms of selectiveattention, we can do tasks such as the following: we canfixate on a central cross and press a button only when agreen dot is flashed to the right while ignoring the samedot anywhere else in the visual field. And we can switchattention to do this task at any other spatial position, nowignoring the formerly relevant position. Although in bothconditions, the same physical stimuli are given and thesame behavioral responses are issued, there is obviously astrong cognitive control over the routing of informationfrom sensory to motor areas. Conceptually, the effect ofcognitive top-down control is a modification in thecommunication structure between brain areas.But how do groups of neurons communicate? And howdo top-down influences modify the communication struc-ture within a few hundred milliseconds when anatomicalconnections stay unchanged on that timescale? Althoughwe still know very little about neuronal communicationmechanisms, we often have an implicit concept or modelabout it. In very general terms, the dominant model ofneuronal communication is that a neuron sends itsmessage (encoded in e.g. action potential rate or in thedegree of action potential synchronization) down its axonsto all neurons to which it is anatomically connected. Thosereceiving neurons combine (e.g. sum and threshold) all thedifferent inputs that they receive from all neurons towhich they have connections. An important aspect of thismodel is that both the distribution and the reception ofneuronal signals is governed solely by the structure ofthe anatomical connections, that is, there is no furthercommunication structure beyond the one imposed byanatomical connectedness. However, cognitive functionsrequire flexibility in the routing of signals through thebrain. They require a flexible effective communicationstructure on top of the anatomical communic ationstructure that is fixed, at least on the timescale at whichcognitive demands change.In this article, I hypothesize that this effectivecommunication structure is mechanistically implementedby the pattern of coherence among neuronal groups, thatis, the pattern of phase-locking among oscillations in thecommunicating neuronal groups. Specifically, I hypoth-esize that neuronal communication between two neuronalgroups mechanistically depends on coherence betweenthem and the absence of neuronal coherence preventscommunication. I will address this hypothesis as the‘communication-through-coherence’ (CTC) hypothesis. Itis based on two realizations: first, activated neuronalgroups have th e intri nsic property to oscillate [1,2].Second, those oscillations constitute rhythmic modu-lations in neuronal excitab ility that affect both thelikelihood of spike output and the sensitivity to synapticinput. Thus, rhythmic excitability peaks constituterhythmically reoccurring temp oral windows f or com-munication. Only coherently oscillating (or phase-locked)neuronal groups can communicate effectively, becausetheir communication windows for input and for output areopen at the same times.Previous work has hypothesized that neuronal coher-ence (or phase-locking or synchronization) could provide atag that binds those neurons that represent the sameperceptual object [3–7]. This binding tag would be a flex-ible code for linking neurons into assemblies and therebygreatly enlarging the representational capacity of a givenpool of neurons. This hypothesis is known as the binding-by-synchronization (BBS) hypothesis. The CTC and theBBS hypotheses are fully compatible with each other, butthey are also clearly distinct. Whereas the BBS hypothesisCorresponding author: Fries, P. ([email protected]).Opinion TRENDS in Cognitive Sciences Vol.9 No.10 October 2005www.sciencedirect.com 1364-6613/$ - see front matter Q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2005.08.011is primarily suggesting a representational code, the CTChypothesis considers the mechanistic consequences ofneuronal oscillations for neuronal communication. Itsuggests that at the heart of our cognitive dynamic is adynamic communication structure and that the neuronalsubstrate is the flexible neuronal coherence pattern.In the following, I will first review neurophysiologicaldata that suggest an important role of synch ronousneuronal oscillations for neuronal communication. I willthen present some evidence that directly suggests thatneuronal coherence can serve neuronal communicationand can be dynamically modulated by cognitive demands.Finally, I will review neurophysiological data about theattentional modulation of synchronous neuronal oscil-lations and speculate about the detailed implementationof a flexible communication