Ongoing Dynamics of Peak Alpha Frequency Characterize Hypnotic Induction in Highly Hypnotic-Susceptible Individuals

Abstract

Hypnotic phenomena exhibit significant inter-individual variability, with some individuals consistently demonstrating efficient responses to hypnotic suggestions, while others show limited susceptibility. Recent neurophysiological studies have added to a growing body of research that shows variability in hypnotic susceptibility is linked to distinct neural characteristics. Building on this foundation, our previous work identified that individuals with high and low hypnotic susceptibility can be differentiated based on the arrhythmic activity observed in resting-state electrophysiology (rs-EEG) outside of hypnosis. However, because previous work has largely focused on mean spectral characteristics, our understanding of the variability over time of these features, and how they relate to hypnotic susceptibility, is still limited. Here we address this gap using a time-resolved assessment of rhythmic alpha peaks and arrhythmic components of the EEG spectrum both prior to and following hypnotic induction. Using multivariate pattern classification, we investigated whether these neural features differ between individuals with high and low susceptibility to hypnosis. Specifically, we used multivariate pattern classification to investigate whether these non-stationary neural features could distinguish between individuals with high and low susceptibility to hypnosis before and after a hypnotic induction. Our analytical approach focused on time-resolved spectral decomposition to capture the intricate dynamics of neural oscillations and their non-oscillatory counterpart, as well as Lempel-Ziv complexity. Our results show that variations in the alpha center frequency are indicative of hypnotic susceptibility, but this discrimination is only evident during hypnosis. Highly hypnotic-susceptible individuals exhibit higher variability in alpha peak center frequency. These findings underscore how dynamic changes in neural states related to alpha peak frequency represent a central neurophysiological feature of hypnosis and hypnotic susceptibility.

Keywords: alpha frequency; altered state of consciousness; aperiodic activity; arrhythmic activity; classification; hypnosis; machine learning; resting-state EEG; spectral analysis.

Figure 1

(A) Our analysis pipeline employed the SPRiNT algorithm to extract the variability in rhythmic component parameters in alpha-band activity (8 to 13 Hz) and the parameters of the arrhythmic component using 5 min of eye-closed resting-state EEG (rs-EEG) data, both pre- and post-hypnotic induction. SPRiNT, a time-resolved implementation of the FOOOF algorithm via specparam in MATLAB, was used to analyze each one-second time window, extracting the relative amplitude, width, and center frequency of the peak within the alpha-band range for each rhythmic component, as well as the offset and exponent of the arrhythmic component for each individual and channel i. The standard deviation across the time series was computed to estimate the variability in each parameter to evaluate the dynamics of spectral features (B). We extracted these estimates separately for the pre- and post-induction periods and evaluated the effects for the pre-induction period compared to Δinduction (values from post-induction minus values from pre-induction). Individuals with high and low hypnotic susceptibility were classified based on these metrics using multivariate pattern analysis with leave-one-out cross-validation (LOOCV) employing a linear support vector machine (SVM). EEG channels served as features for this linear model. Each training iteration included balanced classes of individuals with low (light grey) and high (dark grey) hypnotic susceptibility, with one participant left out for validation.

Figure 2

Panels (A,B) show analysis for pre-induction EEG, whereas panels (C,D) show analysis for Δinduction EEG. Panels (A–D) show AUC values for assessing classification performances of SVM linear models for accurately classifying high and low hypnotic-susceptible individuals in LOOVC based on arrhythmic and alpha-band related features of the power spectrum during the EEG pre-induction period (A) and the Δinduction period (C), as well as Lempel–Ziv complexity coefficients calculated during the EEG pre-induction period (B) and for Δinduction (D). Red dots show observed AUC values. Boxplot shows surrogate null distributions based on random permutations where we shuffled the labels during training. The bottom and top whiskers indicate the first and third quartiles, respectively. Three asterisks indicate statistical significance at p < 0.001. Topographies show the averaged coefficient values of the SVM models from all iterations and LOOVC approach for the neural features that can discriminate between high and low hypnotic-susceptible individuals better than chance-level.