The pericentral primary sensorimotor cortices generate the ‘mu rhythm’ with a distinct spectral signature exhibiting two peaks, generated predominantly anterior (20 Hz) or posterior (10 Hz) to the central sulcus; it defines a ‘background’ network state upon which somatosensory inputs will impinge. We used the high spatiotemporal resolution of magnetoencephalography to analyze the perturbation dynamics of these cortical rhythms in response to a series of paired electric median nerve stimuli: single trials were sorted off-line according to increasing power of the 10- or 20-Hz rebounds which occurred 300–600 ms after the first stimulus; using subaverages formed from the upper and lower 20% of this distribution, we analyzed somatosensory evoked fields (SEF) and power modifications caused by the second stimulus in the pair. We report three key findings: (1) the power level of rhythm rebounds triggered by the first stimulus predicted the rebound strength after the second stimulus applied 600 ms later; yet, it was uncorrelated across the 2.4-s interval separating subsequent stimulus pairs. (2) Conventional averaging camouflages substantial trial-to-trial variations of rhythm dynamics including, for example, even non-occurrences of rhythm rebounds. (3) For six of the seven subjects, the background rhythm power did not affect any SEF component; for the subject with the strongest rhythms only intracortically generated deflections (peaking after the thalamocortical input component N20m) varied as function of pre-stimulus 10- or 20-Hz power. Thus, the perturbation dynamics of the pericentral mu rhythm exhibits a significant intertrial variance, which becomes effective mainly at a time scale larger than 600 ms.