Desync and connect: noninvasive MEG evidence of long-range coupling between locally desynchronized regions

Abstract

A prevailing assumption in systems neuroscience is that long-range phase coupling between cortical regions requires locally synchronized oscillatory activity within each interacting area. Here we ask whether distal phase coupling can be detected from MEG data when both interacting regions undergo concurrent event-related desynchronization (ERD). Using controlled simulations with progressively disrupted local phase coherence, we first establish that the Context-Dependent PSIICOS (CD-PSIICOS) framework -- a projection-based connectivity estimator that separates power-dominated spatial leakage from interaction-specific cross-spectral structure -- can recover distal coupling under increasing local phase dispersion, whereas estimators based on the imaginary part of the cross-spectrum fail at near-zero lags. We then analyzed recordings of a center-out reaching dataset from nine participants. A group decomposition of the coupling time courses identifies four long-range networks in alpha (8-13 Hz) and two in beta (15-25 Hz), each with its own time window. For each network we follow the local connections between its own sources as the power-dominated subspace is projected out, separating leakage-driven from leakage-resistant local coupling. In the earliest alpha network the leakage-resistant pairs stay at or below their pre-stimulus level at every rank while the power of the same sources falls, so this interhemispheric interaction may have no stable local counterpart. In the beta networks such a counterpart does survive, but the coupling it carries is an order of magnitude weaker than the long-range coupling of the same sources. Only in the latest alpha window, by which time power has returned to baseline, do local and long-range coupling rise together. Factorising the coupling into the amplitude of the single-trial cross-terms and the consistency of their phases separates the same cases: in the networks without a local counterpart, phase consistency increases inside the window while the amplitude of the cross-terms declines, whereas in the network whose local coupling rises with it, the increase reflects a growth of the cross-terms themselves rather than any improvement in their phase agreement. Long-range phase coupling can therefore persist, and in places strengthen, between regions undergoing macroscopic ERD, and noninvasive MEG can resolve this dissociation when leakage-aware estimation is employed.