Gamma activity (γ, >30 Hz) is universally demonstrated across brain regions and species. However, the physiological basis and functional role of γ sub-bands (slow-γ, mid-γ, fast-γ) have been predominantly studied in rodent hippocampus; γ activity in the human neocortex is much less well understood. We use electrophysiology, non-invasive brain stimulation, and several motor tasks to examine the properties of sensorimotor γ activity sub-bands and their relationship with both local GABAergic activity and motor learning. Data from three experimental studies are presented. Experiment 1 (N = 33) comprises magnetoencephalography (MEG), transcranial magnetic stimulation (TMS), and a motor learning paradigm; experiment 2 (N = 19) uses MEG and motor learning; and experiment 3 (N = 18) uses EEG and TMS. We characterised two distinct γ sub-bands (slow-γ, mid-γ) which show a movement-related increase in activity during unilateral index finger movements and are characterised by distinct temporal–spectral–spatial profiles. Bayesian correlation analysis revealed strong evidence for a positive relationship between slow-γ (~30–60 Hz) peak frequency and GABAergic intracortical inhibition (as assessed using the TMS-metric short interval intracortical inhibition). There was also moderate evidence for a relationship between the power of the movement-related mid-γ activity (60–90 Hz) and motor learning. These relationships were neurochemical and frequency specific. These data provide new insights into the neurophysiological basis and functional roles of γ activity in human M1 and allow the development of a new theoretical framework for γ activity in the human neocortex.

The frequency spectrum is a central method for representing the dynamics within electrophysiological data. Some widely used spectrum estimators make use of averaging across time segments to reduce noise in the final spectrum. The core of this approach has not changed substantially since the 1960s, though many advances in the field of regression modelling and statistics have been made during this time. Here, we propose a new approach, the General Linear Model (GLM) Spectrum, which reframes time averaged spectral estimation as multiple regression. This brings several benefits, including the ability to do confound modelling, hierarchical modelling, and significance testing via non-parametric statistics. We apply the approach to a dataset of EEG recordings of participants who alternate between eyes-open and eyes-closed resting state. The GLM-Spectrum can model both conditions, quantify their differences, and perform denoising through confound regression in a single step. This application is scaled up from a single channel to a whole head recording and, finally, applied to quantify age differences across a large group-level dataset. We show that the GLM-Spectrum lends itself to rigorous modelling of within- and between-subject contrasts as well as their interactions, and that the use of model-projected spectra provides an intuitive visualisation. The GLM-Spectrum is a flexible framework for robust multilevel analysis of power spectra, with adaptive covariate and confound modelling.