Modification of scale-free electrophysiological activity induced by changes in excitatory-inhibitory balance is governed by the dynamics of multiple oscillatory relaxation processes

Neurophysiological recordings are dominated by arhythmical activity whose spectra can be characterised by power-law functions, and on this basis are often referred to as reflecting scale-free brain dynamics (1/f {beta}). Relatively little is known regarding the neural generators and temporal dynamics of this arhythmical behaviour compared to rhythmical behaviour. Here we used Irregularly Resampled AutoSpectral Analysis (IRASA) to quantify {beta}, in both the high (5-100 Hz, {beta}hf) and low frequency bands (0.1-2.5 Hz, {beta}lf in EEG/MEG/ECoG recordings and to separate arhythmical from oscillatory modes of activity, such as, alpha rhythms. In MEG/EEG/ECoG data, we demonstrate that oscillatory alpha power dynamically correlates over time with {beta}hf and similarly, participants with higher rhythmical alpha power have higher {beta}hf). In a series of MEG investigations using the GABA reuptake inhibitor tiagabine, the glutamatergic AMPA receptor antagonist perampanel, the NMDA receptor antagonist ketamine and the mixed partial serotonergic agonist LSD we reveal systematic effects of excitation-inhibition balance on both {beta}hf and {beta}lf. Additionally, strong modulations of {beta}hf are seen in monkey ECoG data during general anaesthesia using propofol and ketamine. Surrogate data analysis demonstrates that arhythmical activity is generated by both linear and non-linear schemes, with non-linear effects emerging at critical boundaries. We develop and test a unifying model which can explain, the 1/f nature of electrophysiological spectra, their dynamic interaction with oscillatory rhythms as well as the sensitivity of 1/f activity to excitation-inhibition balance by considering electrophysiological spectra as being generated by a collection of stochastically perturbed damped oscillators having a distribution of relaxation rates.