Experienced meditators show greater forward traveling cortical alpha wave strengths

Abstract

Mindfulness meditation involves training attention, commonly toward sensory experiences, with nonjudgmental awareness. Theoretical perspectives propose that meditation increases the precision of sensory processing and reduces the generation/elaboration of top‐down expectations. Research suggests forward traveling cortical alpha waves may reflect bottom‐up inhibition to enhance signal‐to‐noise ratios of sensory processing, while backward traveling alpha waves may reflect top‐down inhibition based on expectations. We used electroencephalography to test whether the strength of forward and backward traveling cortical alpha waves differed between meditators and a matched sample of nonmeditators during eyes‐closed resting (N = 97) and during a visual cognitive (Go/No‐go) task (N = 126). Our results showed meditators produced stronger forward traveling waves compared to nonmeditators while resting with their eyes closed and during task performance. Meditators also exhibited weaker backward traveling waves while resting with their eyes closed. These results may indicate a neural mechanism underpinning enhanced attention associated with meditation, as well as a potential neural marker of reductions in mind‐wandering, suggested to be associated with meditation. The results also support models of brain function that suggest attention modification is achievable through mental training to increase sensory awareness, which might be indexed by the greater strength of forward traveling cortical waves.

Evidence shows mindfulness meditation improves attention and mental health. Neural mechanisms underlying these effects are not fully understood. Cortical travelling waves index inhibitory functions for top‐down expectations (backward waves travelling from frontal to posterior regions) or to enhance sensory processing (forward waves). Meditators showed different travelling wave patterns. Stronger forwards waves in meditators may indicate an emphasis on sensory processing. Weaker backwards waves while resting may indicate weaker top‐down expectations.

INTRODUCTION

Mindfulness meditation involves training attention to present moment experiences, concurrent with an attitude of nonjudgmental awareness. ¹ The practice of mindfulness meditation is associated with improvements in well‐being and some cognitive functions, which are likely driven by changes in neural activity. ^{2 − 7} Research shows meditation is associated with increased gray matter volumes, changes in metabolic activity in specific brain regions, increases in neural oscillatory power, and altered activity during cognitive tasks. ^{8 − 13} However, our understanding of the neural mechanisms underlying the effects of meditation is incomplete, with debate still underway about even the broader aspects of potential explanations. For example, it is still uncertain the extent to which the effects of meditation are due to altered top‐down attentional control, to alterations in bottom‐up processing, or related to both mechanisms. ^{14 − 18}

Recent theoretical perspectives have attempted to advance understanding by explaining how meditation alters the parameters of the predictive processing framework of brain function. This framework suggests that because the brain has only indirect access to information about its environment (through sensory inputs), it acts as a hierarchical Bayesian inference model generator, constructing a predictive model of its environment and place within that environment. ^{19 − 21} Within this framework, sensory perception, decision‐making, and motor functions are all explainable by one overarching driving force—the minimization of the long‐term average prediction error. ²² , ²³ Minimization of prediction error can be achieved through two routes. First, the brain can update prior beliefs in accordance with sensory evidence, making predictions more closely matched to incoming sensory input; this is perceptual inference. ^{21 − 23} Second, minimization of prediction error can be achieved using the brain's motor control functions to manipulate the body's musculature, causing alterations to the body's orientation so the sensory organs sample inputs that conform to prior beliefs; a process known as active inference. ¹⁹ This minimization of long‐term average prediction error allows organisms to act to remain in preferred (or predicted) states for their own preservation and well‐being. ²⁴

Furthermore, prediction errors are weighted by expected precision. To achieve this, increased synaptic gain is allocated to neural activity underlying the processing of prediction errors that are afforded more precision, enabling prior belief updating to be weighted according to the precision assigned to the prediction error. This precision weighting is thought to underpin and enable attentional functions. ²³ , ²⁵ , ²⁶ , ²⁷ Experimental research exploring neural markers of predictive processing indicates that when sensory inputs deviate from expectations encoded in cortical regions, precision‐weighted prediction errors are passed up the neural hierarchy. At each level, excitatory synaptic connections are thought to transmit these errors laterally and to the level above, while inhibitory synaptic connections convey predictions laterally and to the level below, suppressing expected sensory inputs and regulating the precision weighting of bottom‐up signals. ²¹ , ²⁸

A simplification of the broader pattern of neural processes is that prediction error–tied sensory inputs are predominantly processed in posterior brain regions, which are lower in the cortical hierarchy, and pass unresolved prediction errors up the cortical hierarchy. ²⁹ , ³⁰ In contrast, frontal regions are higher in the cortical hierarchy and generate more abstract Bayesian predictions about sensory inputs ²¹ , ³¹ (although we note there are deviations from this broad pattern; e.g., the auditory cortices are located laterally, approximately the middle of the posterior/anterior axis). Recent theoretical developments that attempt to map the predictive processing framework to known neuroanatomy and neurophysiology have also suggested that the predicted posterior distribution over hidden (external) states inferred from processing sensory inputs can also be passed up the cortical hierarchy, with these posteriors still able to elicit prediction errors for expected outcomes in regions higher in the cortical hierarchy. ³² For example, brief sequences of identical tones could match expectations held in auditory processing regions (which form predictions on short time scales only), with posterior estimates for the matching of that expected state passed up the neural hierarchy, while still violating predictions in higher‐order regions, which might expect a tone change based on the integration of posterior estimates over the longer time scale that is integrated in the processing undertaken by higher‐order regions. ³³ Given the different time scales represented across different levels of the cortical hierarchy, ³⁴ top‐down processes instantiated higher in the cortical hierarchy are suggested to relate to the experience of the conceptual‐self, which processes abstractions including thoughts about past events and future events (counterfactual states that reflect mental time‐traveling), while bottom‐up sensory processes reflect the more present‐moment centered experiencing‐self. ³⁵ , ³⁶

Within this predictive processing framework, the attention training aspect of meditation (which is commonly focused on bodily sensations) has been proposed to enhance the brain's capacity to increase precision‐weighting of hierarchically lower perceptual processes, increasing synaptic gain for the weighting of prediction errors. ³⁵ , ³⁷ , ³⁸ Additionally, the nonjudgmental component of meditation has been suggested to reduce the production and elaboration of counterfactual predictive models, ³⁷ , ³⁹ reducing the magnitude and precision of information flow down the neural hierarchy from frontal to sensory regions. ³⁵ This exemplifies how the predictive processing framework provides an elegant model that can explain a wide variety of brain functions and behavior using a limited and parsimonious set of principles. However, although there is empirical evidence that the predictive processing framework can explain a wide range of brain functions, this evidence base is still developing. Additionally, while predictive processing–based explanations fit elegantly with intuitions about how meditation affects the brain, no empirical evidence is currently available to support predictive processing explanations of the mechanisms underlying meditation. An opportunity to address this gap may be provided by cortical traveling waves.

Recent research suggests that cortical traveling alpha waves may subserve both top‐down and bottom‐up inhibitory gating functions. ^{40 − 44} These traveling waves show amplitude peaks and troughs that travel through the cortex periodically. ⁴² , ⁴⁵ Oscillations detected in occipital electrodes that are also present in more anterior electrodes, but with phase delays that progressively increase with distance, can be assumed to be traveling forward. ⁴⁶ In contrast, phase lags progressively increasing from frontal to occipital electrodes indicate a backward traveling wave. ⁴⁶ Figure 1 depicts a backward traveling wave extracted from our data.

FIGURE 1.

A visual representation of the traveling wave computation method using a two‐dimensional fast Fourier transform (2D‐FFT) using real data obtained from an example participant in our study. First, 1‐s epochs were extracted from the continuous electroencephalography (EEG) data (after EEG preprocessing to reduce artifacts). The time series from each electrode were then organized into matrices, with posterior electrodes at the bottom of the matrices. This ordering is depicted in panels A and B, which show the data from a 1‐s epoch after applying a bandpass filter in the alpha frequencies of interest to highlight the traveling waves. Each epoch was then used as the input for a 2D‐FFT. This provided a matrix of power values at each spatial and temporal frequency. Within this matrix, the x‐axis provides power values at each temporal frequency, while the y‐axis provides power values at each spatial frequency (shown in panel C). Within the y‐axis (spatial frequencies), values above 0 reflect backward traveling spatial waves (from frontal to posterior electrodes), values below 0 reflect forward traveling spatial waves (from posterior to frontal electrodes), and values at 0 reflect standing waves (waves that do not travel). Finally, to assess the wave strength in each spatial and temporal frequency, we divided the power in each cell in the matrix by the power obtained after randomly shuffling the order of the electrodes in each epoch 100 times, repeating the 2D‐FFT computations for each shuffle, and averaging the outputs of these null 2D‐FFTs. We then computed a log10 transform on the output of this division and multiplied this value by 10 to obtain a measure of the strength by which the real traveling wave strength exceeded the permutation‐based null traveling wave strength on the decibel scale. Panel D depicts these normalized values at each spatial frequency at the 10 Hz temporal frequency for the 1‐s example epoch depicted in the other frames. However, note that to constrain the number of comparisons, we measured the maximum forward and backward traveling waves across all forward/backward spatial frequencies and temporal frequencies and normalized these against the maximums from the null data as is typical when using 2D‐FFTs to measure traveling waves. ⁴⁵

Studies of visual attention suggest that backward traveling waves index top‐down predictions and relate to the direction of attention. ⁴² , ⁴⁵ , ⁴⁷ Attention‐based modulation of backward traveling waves has been found regardless of whether visual stimuli are presented or not, suggesting backward waves are modulated by the intentional allocation of attention, without visual input being required. ⁴⁷ Evidence shows that backward traveling alpha waves are stronger ipsilateral to visual stimuli presented in the nonattended visual hemisphere (i.e., in the hemisphere that processes the visual information participants are not attending to). ⁴⁴ , ⁴⁷ Furthermore, backward traveling alpha waves have been shown to decrease in the hemisphere contralateral to task‐related visual target stimuli (with the reduced inhibition suggested to enable visual processing of those targets), an effect that is modulated by target load (with larger reductions in wave strength as target number increased). ⁴⁴ Backward traveling alpha waves have also been shown to increase during closed eyes resting and during prolonged presentation of an unchanging stimulus (when visual processing requirements are low). ⁴²

Taken together, these results indicate that backward traveling alpha waves perform a top‐down inhibitory function based on the direction of attention. ⁴⁵ , ⁴⁷ Backward traveling waves have also been shown to be stronger during memory recall than encoding, ⁴⁸ concordant with the view that episodic memory recall relies on top‐down mechanisms of gain‐control over prediction error units. ⁴⁹ Finally, pre‐stimulus backward traveling alpha waves have been shown to increase more strongly when a cue indicates an increased likelihood of an upcoming target stimulus (which required a participant response). ⁴³ The increased strength of the backward traveling waves was associated with increased pre‐stimulus alpha power in visual processing regions, an effect that mediated biases toward responding, particularly in participants who showed behavioral response patterns that were more strongly informed by prior predictions. ⁴³ This observation provides direct evidence that backward traveling waves implement top‐down predictions as proposed by the predictive processing framework.

In contrast, forward traveling alpha waves have been demonstrated to increase in strength with visual stimulation, and to decrease with a prolonged presentation of an unchanging stimulus. ⁴² They have also been found to be stronger in the hemisphere responsible for processing stimuli, irrespective of whether participants attended to the stimuli. ⁴⁷ Forward traveling waves have also been shown to be stronger during memory encoding than recall, suggesting that memory formation is reliant on feedforward neural interactions from sensory to frontal regions. ⁴⁸ These results suggest that forward traveling alpha waves may index the bottom‐up transmission of prediction errors up the cortical hierarchy. However, some recent evidence suggests that the relationship between sensory processing and scalp‐measured forward traveling alpha waves is indirect. Instead of representing the transmission of prediction errors, forward traveling alpha waves may reflect bottom‐up spreading inhibition of non‐task‐relevant brain regions and neural activity. ⁴⁴ , ⁵⁰ This inhibition may function to increase the signal‐to‐noise ratio of the processing of sensory information, which is also being passed up the neural hierarchy. ⁴⁴ , ⁵⁰

Since meditation has been suggested to reduce the generation or elaboration of top‐down predictions, while increasing precision weighting devoted to processing sensory information and bottom‐up prediction errors, we tested whether experienced meditators show weaker backward waves and stronger forward waves compared to nonmeditators. We expected: (1) meditators would show stronger forward waves during a visual attention task, which may reflect increased inhibition to enhance the signal‐to‐noise ratio of sensory information processing; and (2) meditators would show weaker backward wave strength, both during visual attention and during eyes‐closed resting, which may reflect reduced top‐down prediction generation. Additionally, to interrogate the functional relevance of traveling waves, we performed exploratory analyses of forward and backward wave strength from meditators and nonmeditators during a working memory task that separated memory set, delay, and probe presentation periods. Within this task, we expected the period where participants were presented with visual stimuli to remember to be associated with stronger forward waves, while we expected the period where participants had to recall memory set information would be associated with stronger backward waves.

MATERIALS AND METHODS

Participants

Eyes‐closed resting electroencephalography (EEG) and task‐related EEG data were recorded across two studies of experienced meditators, which examined differences in neural activity between meditators and nonmeditators. The first dataset included 34 meditators and 36 healthy nonmeditators, and the second dataset included 39 meditators and 36 healthy nonmeditators. Both datasets included eyes‐closed resting EEG data ¹⁰ , ⁵¹ and a Go/No‐go task that presented Go and No‐go trials with equal probability, with no more than three of each trial type presented consecutively. ² , ⁵² , ⁵³ However, due to time constraints within sessions, not all participants provided both types of recordings, and some recordings were excluded due to artifact contamination. As such, the final analyzed samples for each type of EEG recording differed in size, with 126 participants providing Go/No‐go task recordings (69 meditators and 57 nonmeditators), 97 participants providing eyes‐closed resting recordings (50 meditators and 47 nonmeditators), and 84 participants providing both recordings (47 meditators and 37 nonmeditators).

During eyes‐closed resting, participants were instructed to simply rest with their eyes closed, not to meditate, but to just let their mind do as it pleased. This instruction was provided to enable characterization of trait differences in resting EEG, rather than differences related to meditation states. More details on these datasets can be found in Bailey et al. ⁵¹ for the resting data and Bailey et al. ² , Bailey et al. ⁵³ , and Bailey et al. ⁵² for the Go/No‐go data (note that behavioral results for the Go/No‐go data have been previously reported separately for dataset one, ² and are in preparation for publication for dataset two). Additionally, 29 meditators and 29 nonmeditators completed a modified Sternberg working memory task. ³ Within this task, memory set stimuli were presented simultaneously, followed by a brief delay period and then a probe presentation period, where participants responded as to whether the probe was contained within the memory set. ³ A visualization of the Go/No‐go and modified Sternberg task are presented in Figure 2.

FIGURE 2.

Schematic depictions of the tasks used to analyze forward and backward traveling wave strength. Top: The Go/No‐go task presented emoticon‐style happy and sad faces in a 50:50 ratio, with stimulus response pairings switched between two randomly counterbalanced blocks so that all participants responded to an equal number of happy and sad faces and were required to withhold response to an equal number of happy and sad faces. Bottom: The modified Sternberg working memory task presented all memory set letters simultaneously and always contained a random selection of eight letters. The probe was contained within the memory set on 50% of the trials.

Participants were recruited through community advertising and advertising at meditation centers. Inclusion criteria for meditators required a current mindfulness meditation practice with at least 2 h per week of practice during the preceding 2 months, and a minimum of 6 months since they started practicing meditation (demographic and meditation experience statistics are presented in Table 1). Meditation practices were required to be mindfulness‐based, aligned with the definition provided by Kabat‐Zinn: ¹ “paying attention in a particular way: on purpose, in the present moment, and nonjudgmentally.” Meditation practices were also required to involve focused attention on the breath or bodily sensations. Nonmeditators were required to have less than 2 h of lifetime experience with any type of meditation. Participants who self‐reported any history of mental or neurological illness or current psychoactive medication or recreational drug use were excluded. Participants were excluded if they met the criteria for any DSM‐IV illness or scored in the moderate or higher range on the Beck Depression Inventory or Beck Anxiety Inventory. All participants were between 19 and 65 years of age and had normal or corrected‐to‐normal vision. All participants provided informed written consent prior to participation. The study was approved by the ethics committee of the Alfred Hospital and Monash University and was conducted in accordance with the Declaration of Helsinki.

TABLE 1.

Demographic details for the different datasets.

	Meditators mean (SD)	Nonmeditators mean (SD)	Statistics
Eyes‐closed resting data
Age (years)	37.06 (12.08)	33.085 (12.99)	t(95) = 1.565, p = 0.121
Gender	29 females, 21 males	26 females, 21 males	χ2 = 0.071, p = 0.790
Meditation experience (years)	8.086 (8.861)	0
Meditation practice (hours per week)	5.782 (4.035)	0
Go/No‐go data
Age (years)	36.89 (13.04)	32.74 (13.21)	t(124) = 1.773, p = 0.079
Gender	32 females, 37 males	29 females, 28 males	χ2 = 0.253, p = 0.615
Meditation experience (years)	8.236 (9.271)	0
Meditation practice (hours per week)	6.337 (4.782)	0
Sternberg working memory data
Age (years)	37.31 (11.50)	36.45 (13.92)	t(56) = 0.257, p = 0.798
Gender	19 females, 10 males	17 females, 12 males	χ2 = 0.293, p = 0.588
Meditation experience (years)	9.22 (10.87)	0
Meditation practice (hours per week)	5.41 (4.11)	0

EEG recordings

A Neuroscan 64‐channel Ag/AgCl Quick‐Cap was used to record EEG data using Neuroscan Acquire software through a SynAmps2 amplifier (Compumedics). Online, electrodes were referenced to an electrode between Cz and CPz. Electrode impedances were reduced to below 5 kΩ prior to the start of each recording. EEG data were recorded at 1000 Hz with an online bandpass filter of 0.05−200 Hz. Data were preprocessed offline in MATLAB (The Math Works, 2023a) using EEGLAB, ⁵⁴ with data cleaning implemented using the default settings for the wICA_ICLabel cleaning approach of the RELAX preprocessing pipeline. ⁵⁵ , ⁵⁶ This method uses a wavelet‐enhanced independent component analysis to reduce the influence of eye movement, muscle, and other artifacts. A summary of the RELAX preprocessing details can be found in the Supporting Information, and the full description can be found in the RELAX manuscripts. ⁵⁵ , ⁵⁶

Traveling wave computation

To determine the strength and direction of cortical traveling alpha waves, we applied the method developed by Alamia and VanRullen. ⁴⁵ Within each participant, 1‐s epochs were extracted from each EEG file. Epochs were extracted every 500 ms within the resting data (so contained 500 ms overlaps with neighboring epochs). For Go/No‐go data, epochs were time‐locked to stimulus presentation for all stimuli where participants provided a correct response (or nonresponse for No‐go trials). For the Sternberg task, the 1‐s epochs were extracted from the middle of the memory set presentation, delay period, and probe periods for trials where participants provided the correct response to the probe stimuli. This meant that within the probe period, we measured activity starting 500 ms after the probe was presented. This avoided visual processing‐related activity, a decision implemented to focus analyses on activity related to participants’ attempts to match their recall of the memory set to the probe. A schematic of tasks and the analyzed periods can be viewed in Figure 2.

Within each 1‐s epoch, the EEG signal from seven midline electrodes (Oz, POz, Pz, CPz, Cz, FCz, FPz) were extracted, providing a two‐dimensional matrix (electrode × time). A two‐dimensional fast Fourier transform (2D‐FFT) was applied to each electrode × time matrix within each epoch separately. The output matrix of this 2D‐FFT provides both temporal frequencies and spatial frequencies, with spatial frequencies represented in the vertical axis of the matrix. Waves propagating in the forward direction (from occipital to frontal electrodes) are represented in the upper quadrant of the matrix, while backward propagating waves (from frontal to occipital electrodes) are represented in the lower quadrant. ⁴⁵ Next, we extracted the maximum value within the alpha band (8−13 Hz) from each quadrant within each epoch to measure the forward and backward traveling waves. ⁴⁵ To ensure traveling wave values reflected real signals, we performed the same traveling wave computations on a surrogate null version of the data. These surrogate null versions were obtained by shuffling the order of the electrodes in the electrode × time matrix separately within each epoch prior to the wave computation followed by computation of the 2D‐FFT on the null data. ⁴⁵ This shuffling of the electrode order destroyed the ability of the analysis to detect the spatial pattern of traveling waves provided by progressively increasing phase lags in more frontal electrodes (for forward waves) or more posterior electrodes (for backward waves), while still preserving the temporal oscillatory pattern, providing a distribution that was matched to the real data for oscillatory power but reflecting a null distribution for the spatial structure. ⁴⁵ Finally, to ensure our statistical comparisons focused on the signal strength of the forward and backward traveling waves above the null, we divided the values from the real data within each epoch by the values within the surrogate data for the same epoch, then multiplied the result by 10*log10 and averaged values across all epochs within each participant and condition. This provided a value for each participant and condition that reflected the ratio of the strength by which the real forward and backward waves exceeded the surrogate forward and backward waves, with values on a log scale (providing values in decibel units [dB]). ⁴⁵

Statistical comparisons

Statistical analyses were performed using JASP 0.17.2.1. ⁵⁷ To control for the effect of outliers on our statistical analyses, prior to statistical analysis, we winsorized values that were more than three scaled median absolute deviations from the median across all participants within the forward and backward waves for each condition separately. Winsorization was achieved by replacing the outlying value with the next most outlying value.

For our primary analysis, we conducted repeated measures ANOVAs separately within the resting and Go/No‐go data to test for an interaction between group (meditators and nonmeditators) and traveling wave direction (forward and backward), including all available participants in each ANOVA. To provide both a frequentist p‐value and an indication of the strength of evidence for each analysis, we performed both frequentist ANOVAs and also Bayesian ANOVAs, and have reported the Bayes Factor (BF10) for each main effect and interaction effect compared to models stripped of that effect. Given that our expectation was for meditators to show stronger forward waves and weaker backward waves relative to controls, our primary focus was on the interaction between group and wave direction. As such, we controlled for multiple comparisons experiment‐wise across these interactions for the eyes‐closed resting and Go/No‐go dataset using the false discovery rate method of Benjamini and Hochberg, ⁵⁸ denoting corrected p‐values as FDR‐p. Additionally, we explored the drivers of significant interactions using post‐hoc t‐tests to make comparisons between the groups within specific conditions of interest. These post‐hoc t‐tests were also controlled for multiple comparisons using the false discovery rate.

In addition to these primary analyses, we performed several exploratory analyses to help with the interpretation of the functional implications of the traveling waves. First, within the subset of participants who provided both resting and Go/No‐go data, we tested the correlation between forward and backward wave strength separately within each of these datasets across all participants. Matching the data across both recording types enabled us to assess potential differences in the strength of the relationship between the direction of the traveling waves under these recording conditions. Next, we tested for correlations between forward and backward wave strength and task accuracy in the Go/No‐go dataset (with task accuracy measured by d‐prime). After the visual inspection of the data suggested a potentially interesting pattern, we also conducted exploratory repeated measures ANOVAs within the matched resting and Go/No‐go dataset to test for interactions between the groups and EEG recording conditions (resting or Go/No‐go) in the strength of the backward waves.

Finally, to provide a deeper understanding of the potential functional relevance of the forward and backward waves, we explored the changing strengths of these waves during different periods of a sequential Sternberg working memory task. In particular, we tested wave strength within the memory set presentation period, delay period, and memory probe presentation period. To achieve this, we performed a repeated measures ANOVA with two within‐participant factors: wave direction (forward or backward) and working memory period (memory set presentation, delay period, and probe presentation period), and one between participant factor: Group. Finally, to explore whether these differences were functionally relevant, we inserted forward and backward wave strengths from each period of the working memory task as continuous predictors in a linear regression to predict working memory performance (measured by d‐prime), with group as a between‐participant factor.

RESULTS

Meditators show stronger forward traveling waves and weaker backward traveling waves during eyes‐closed resting

Our analyses of the eyes‐closed resting data showed that meditators exhibit stronger forward waves and weaker backward waves during eyes‐closed resting compared to nonmeditators (see Figure 3). This suggests that meditation experience is associated with a greater strength of neural activity that may reflect the bottom‐up processing of sensory information, as well as less generation of top‐down neural activity when no task demands are present. However, we note that although this interpretation is based on evidence for the functional role of traveling waves from previous research, it is not possible to determine the cognitive processes driving differences in resting neural activity, which lack the experimental manipulation of cognitive processes necessary to determine their function.

FIGURE 3.

Forward and backward traveling cortical alpha wave strength within each group, measured from midline electrodes during the eyes‐closed resting state (values provided in decibels [dB]). Post‐hoc testing showed that the interaction was driven by higher forward wave strength within the meditator group (FDR‐p = 0.022, Cohen's d = 0.475, BF10 = 2.294), and lower backward wave strength within the meditator group compared to the nonmeditators (FDR‐p = 0.022, Cohen's d = 0.498, BF10 = 2.919). * FDR‐p < 0.05.

Specifically, within eyes‐closed resting data, a repeated measures ANOVA showed a significant interaction between group and wave direction (F(1,95) = 6.178, p = 0.015, FDR‐p = 0.015, ${\eta }_p^2$ = 0.061, ${\eta }_G^2$ = 0.056, BFincl = 31.265). Post‐hoc t‐tests showed the interaction was driven by higher forward wave strength within the meditator group (FDR‐p = 0.022, Cohen's d = 0.475, BF10 = 2.294) and lower backward wave strength within the meditator group compared to the nonmeditators (FDR‐p = 0.022, Cohen's d = 0.498, BF10 = 2.919). There was also a significant main effect of wave direction, with both groups showing higher values for backward wave than forward wave (F(1,95) = 50.966, p < 0.001, ${\eta }_p^2$ = 0.349, ${\eta }_G^2$ = 0.326, BFincl = 1.717*10 ¹⁴ ). The main effect of group was not significant (F(1,95) = 0.655, p = 0.420, ${\eta }_p^2$ = 0.007, ${\eta }_G^2$ < 0.001, BFincl = 0.178). Additionally, across participants, a strong negative correlation was present between forward and backward wave strength (r = −0.848, 95% CI [−0.774, −0.899], p < 0.001, BF10 = 1.621*10 ²¹ ). Table 2 presents the means and standard deviations for the forward and backward wave strengths during eyes‐closed resting from each group.

TABLE 2.

Means and standard deviations (SD) for the strength of the forward and backward waves in the eyes‐closed resting and Go/No‐go data for each group.

	Forward waves (dB) mean (SD)	Backward waves (dB) mean (SD)
Eyes‐closed resting
Meditators	0.568 (0.487)	0.922 (0.380)
Nonmeditators	0.358 (0.389)	1.091 (0.291)
Go/No‐go task
Meditators	0.732 (0.210)	1.047 (0.197)
Nonmeditators	0.625 (0.201)	1.093 (0.226)

Meditators show stronger forward traveling waves but no difference in backward wave strength during an attention task

Our analyses of the Go/No‐go task data indicated that forward wave strength increased during the visual attention task compared to eyes‐closed resting. This result may suggest that increased bottom‐up inhibition is implemented to protect the task‐relevant sensory processing required to perform visual attention tasks, in alignment with previous research. ⁴² , ⁴⁴ , ⁵⁰ In alignment with the resting results, meditators also showed stronger forward waves during the task compared to nonmeditators (Figure 4). However, in contrast to resting results, meditators showed a similar backward wave strength during the task compared to nonmeditators, reflecting an increase from meditator's resting backward wave strength. Nonmeditators, on the other hand, showed the same backward wave strength during the task compared to while at rest. These results may suggest that meditators engaged typical top‐down attention processes when required for task completion, as well as continuing to generate stronger bottom‐up inhibition to protect sensory processing during the task. In contrast, and perhaps surprisingly, the results might suggest that nonmeditators generated the same amount of top‐down neural activity when resting with their eyes closed as they did while completing a cognitively demanding task.

FIGURE 4.

Top: Forward and backward wave strength during the Go/No‐go task within each group. Post‐hoc t‐tests showed that meditators had higher forward wave strength than nonmeditators (FDR‐p = 0.008, Cohen's d = 0.518, BF10 = 7.942), but no significant differences were present between groups in backward wave strength (FDR‐p = 0.228, Cohen's d = 0.217, BF10 = 0.371). ** FDR‐p < 0.01. Bottom: Backward waves within each group from the eyes‐closed resting and Go/No‐go datasets. There was a significant interaction between group and resting/task (F(1,82) = 6.381, p = 0.013, ${\eta }_p^2$ = 0.072, ${\eta }_G^2$ = 0.026, BFincl = 3.795). Post‐hoc t‐tests indicated the interaction was driven by the meditators showing an increase in their backward wave strength from resting to the Go/No‐go task (FDR‐p = 0.030, Cohen's d = 0.370, BF10 = 2.783), while nonmeditators showed no difference (FDR‐p = 0.293, Cohen's d = 0.176, BF10 = 0.299). * FDR‐p < 0.05.

With regard to Go/No‐go task accuracy, the meditator group showed higher d‐prime scores than the nonmeditator group, indicating meditators performed the task with higher accuracy across both Go and No‐go trials (t(124) = 2.751, p = 0.007, d = 0.492, BF10 = 5.554) (see Figure S1). Next, repeated measures ANOVA analysis of the task‐related EEG data showed a significant interaction between group and wave direction (F(1,124) = 6.059, p = 0.015, FDR‐p = 0.015, ${\eta }_p^2$ = 0.047, ${\eta }_G^2$ = 0.033, BFincl = 11.068). Post‐hoc t‐tests showed the interaction was driven by higher forward wave strength in meditators than nonmeditators (FDR‐p = 0.008, Cohen's d = 0.518, BF10 = 7.942), but no significant differences in the backward waves (FDR‐p = 0.228, Cohen's d = 0.217, BF10 = 0.371). There was also a significant main effect of wave direction, with both groups showing higher backward wave strength than forward wave strength (F(1,124) = 159.121, p < 0.001, ${\eta }_p^2$ = 0.562, ${\eta }_G^2$ = 0.470, BFincl = 2.217*10 ³² ) The main effect of group was not significant (F(1,124) = 2.173, p = 0.143, ${\eta }_p^2$ = 0.017, ${\eta }_G^2$ = 0.005, BFincl = 0.287). Additionally, while there was a negative correlation across participants between the strength of forward and backward waves in the Go/No‐go data (Figure 5), confidence intervals for the r‐value suggested the relationship was significantly weaker than in the resting data (r = −0.333, 95% CI = [−0.128, −0.511], p = 0.001, BF10 = 15.079) compared to the resting data (r = −0.848, 95% CI = [−0.774, −0.899]). Table 2 presents means and standard deviations for the forward and backward wave strengths during the Go/No‐go task from each group. Table 3 presents the means and standard deviations for the behavioral performance during the Go/No‐go task from each group.

FIGURE 5.

Correlations between forward and backward cortical traveling alpha wave strength in the eyes‐closed resting EEG data and the Go/No‐go task–related EEG data. Note the broader spread of values (and different scale) in the eyes‐closed resting data compared to the Go/No‐go task–related data, as well as the stronger correlation in the resting data (r = −0.848, 95% CI = [−0.774, −0.899], p < 0.001, BF10 = 1.621*10 ²¹ and r = −0.333, 95% CI = [−0.128 to −0.511], p = 0.001, BF10 = 15.079, respectively).

TABLE 3.

Means and standard deviations (SD) for the strength of the forward and backward traveling cortical alpha waves in each period of the Sternberg working memory task for each group.

		Meditators mean (SD)	Nonmeditators mean (SD)
Forward waves (dB)	Memory set presentation period	0.739 (0.331)	0.610 (0.356)
	Delay period	0.478 (0.418)	0.401 (0.463)
	Probe presentation period	0.567 (0.293)	0.530 (0.261)
Backward waves (dB)	Memory set presentation period	0.951 (0.310)	1.000 (0.285)
	Delay period	0.999 (0.351)	0.964 (0.348)
	Probe presentation period	1.151 (0.306)	1.062 (0.265)

Since the visual stimuli presented in the Go/No‐go task also elicit event‐related potentials, an argument could be made that our traveling wave results might be driven by event‐related responses in progressive cortical sources with progressively increasing peak and trough delays (instead of traveling waves per se). To assess whether the traveling wave patterns we observed were still present in periods that are less affected by visual processing event‐related activity, we conducted an exploratory comparison of traveling wave strength after excluding the first 400 ms of the epoch, focusing our analysis on the 400−900 ms period following stimuli presentation. Our analysis of this 400−900 ms time window showed a significant group by wave direction interaction, with a larger effect size than our primary analysis (F(1,124) = 10.160, p = 0.002, ${\eta }_p^2$ = 0.076, ${\eta }_G^2$ = 0.055, BFincl = 163.425). Again, the effect was driven by a difference between groups in forward wave strength (t(124) = 3.814, p < 0.001, Cohen's d = 0.683, BF10 = 346.173). In contrast, when we restricted our analysis to the first 500 ms following the stimulus (the 0−500 ms period), the interaction was no longer significant (F(1,124) = 1.705, p = 0.194, ${\eta }_p^2$ = 0.014, ${\eta }_G^2$ = 0.010, BFincl = 0.136). This result indicates that the difference between groups in forward traveling waves was driven by meditators showing stronger forward traveling alpha waves in the latter half of the Go/No‐go trials rather than the first half. Thus, our results are unlikely to be driven by visual processing event‐related responses in progressive cortical sources with increasing peak and trough delays.

With regard to relationships between the traveling waves and task performance, despite the difference between groups in task accuracy and forward wave strength, no correlations were significant between forward or backward wave strength measured across the entire Go/No‐go epoch and task accuracy across participants (all p > 0.2, all BF10 < 0.3). However, when analyses were restricted to the later period within the epoch (the period showing the larger effect for the comparison between meditators and nonmeditators), a significant positive correlation was found between d‐prime task performance and forward traveling waves, although with only weak Bayesian support for the alternative hypothesis (r = 0.164, p = 0.033, BF10 = 1.148).

Finally, because means from resting and Go/No‐go data suggested an interesting interaction between group and condition for backward waves, we performed an exploratory repeated measures ANOVA restricted to participants who provided data for both conditions. The analysis showed a significant interaction between group and condition (F(1,82) = 6.381, p = 0.013, ${\eta }_p^2$ = 0.072, ${\eta }_G^2$ = 0.026, BFincl = 3.795). Post‐hoc tests showed the interaction was driven by meditators displaying an increase in backward wave strength from resting to the task (FDR‐p = 0.030, Cohen's d = 0.370, BF10 = 2.783), while nonmeditators showed no difference (FDR‐p = 0.293, Cohen's d = 0.176, BF10 = 0.299). In contrast, the interaction between group and condition was not significant for the forward waves (F(1,82) = 1.256, p = 0.266, ${\eta }_p^2$ = 0.015, ${\eta }_G^2$ = 0.006, BFincl = 0.373). Instead, there was a significant main effect of group, with the meditators showing stronger forward waves in both datasets (F(1,82) = 7.388, p = 0.008, ${\eta }_p^2$ = 0.083, ${\eta }_G^2$ = 0.052, BFincl = 4.537), as well as a significant main effect of condition, with both groups showing stronger forward waves in the task compared to the eyes‐closed resting condition (F(1,82) = 27.575, p < 0.001, ${\eta }_p^2$ = 0.252, ${\eta }_G^2$ = 0.117, BFincl = 21761.033).

Traveling wave strengths are modulated by working memory processes

Our analysis of traveling waves during the working memory task showed that forward traveling wave strength was higher during memory set presentation, during which time participants encoded the visual information and attempted to remember it for later recall (Figure 6). In contrast, backward wave strength was higher while the probe stimuli were presented, during which time participants responded as to whether the probe stimulus matched their recollection of any of the memory set. These results provide support for suggestions that forward traveling waves are indicative of an underlying function that inhibits non‐task‐relevant processing to increase signal‐to‐noise ratio, thus facilitating effective encoding of visual information. ⁴⁰ , ⁴⁴ , ⁴⁸ , ⁵⁰ They also provide support for suggestions that backward traveling waves may reflect top‐down modulation of precision to reinstate cortical representations of memory set items required for successful memory retrieval. ⁴⁸ , ⁴⁹

FIGURE 6.

Top: Forward and backward wave strength (in decibels [dB]) for each period of the Sternberg working memory task across both groups. Post‐hoc t‐tests indicated forward wave strength was significantly greater in the memory set period (“Letters”) compared to the delay period (FDR‐p = 0.006, Cohen's d = 0.502, BF10 = 74.415). Forward wave strength was also greater in the memory set period compared to the probe period (FDR‐p = 0.015, Cohen's d = 0.350, BF10 = 3.533), while the delay and probe periods did not differ from each other (FDR‐p = 0.059, Cohen's d = 0.264, BF10 = 0.939). In contrast, backward wave strength was significantly higher in the probe period compared to the memory set period (FDR‐p = 0.006, Cohen's d = 0.422, BF10 = 13.605), and compared to the delay period (FDR‐p = 0.015, Cohen's d = 0.356, BF10 = 3.953), while the memory set period and the delay period did not differ (FDR‐p = 0.903, Cohen's d = 0.016, BF10 = 0.145). ** FDR‐p < 0.01. * FDR‐p < 0.05. Bottom: Regression analysis indicated that, controlling for other factors, performance in the working memory task was predicted by forward wave strength in both the memory set presentation period and the probe presentation period. Results showed forward wave strength in the memory set presentation period predicted better d‐prime scores (t = 2.234, p = 0.030), while forward wave strength in the probe presentation period predicted worse d‐prime scores (t = −2.595, p = 0.012).

Within the working memory task data, meditators showed higher accuracy (t(51) = 2.503, p = 0.008, Cohen's d = 0.688, BF10 = 3.373). Our traveling wave analysis showed there was a significant interaction between task period and wave direction (F(2,112) = 5.041, p = 0.008, ${\eta }_p^2$ = 0.083, ${\eta }_G^2$ = 0.030, BFincl = 140.688). Post‐hoc t‐tests indicated forward wave strength was significantly greater in the memory set period compared to the delay period (FDR‐p = 0.006, Cohen's d = 0.502, BF10 = 74.415) and compared to the probe period (FDR‐p = 0.015, Cohen's d = 0.350, BF10 = 3.533), while the delay and probe periods did not differ from each other (FDR‐p = 0.059, Cohen's d = 0.264, BF10 = 0.939). In contrast, backward wave strength was significantly higher in the probe period compared to the memory set period (FDR‐p = 0.006, Cohen's d = 0.422, BF10 = 13.605), and compared to the delay period (FDR‐p = 0.015, Cohen's d = 0.356, BF10 = 3.953), while the memory set period and the delay period did not differ from each other (FDR‐p = 0.903, Cohen's d = 0.016, BF10 = 0.145).

In contrast to the results for the resting and Go/No‐go data, analysis of the working memory task showed no significant main effects or interactions involving group (all p > 0.09, all BF10 < 0.4, reported in full in the Supporting Information). It is possible that this null result was due to the smaller sample size for these comparisons, with Bayes Factors providing inconsequential evidence for the null hypothesis for some interactions involving group, and between group patterns showing the same directions as the resting and Go/No‐go results (see the Supporting Information). Table 3 presents the means and standard deviations for the behavioral performance during the Sternberg working memory task from each group. Table 4 presents means and standard deviations for the forward and backward wave strengths during the Sternberg working memory task from each group.

TABLE 4.

Means, standard deviations (SD), and statistics for the behavioral performance in the Go/No‐go task and Sternberg working memory task for each group.

	Go/No‐go d‐prime mean (SD)	Sternberg d‐prime mean (SD)
Meditators	4.009 (0.774)	1.943 (0.700)
Nonmeditators	3.626 (0.781)	1.473 (0.670)
	t(124) = 2.751, p = 0.007, Cohen's d = 0.492	t(51) = 2.503, p = 0.008, Cohen's d = 0.688

Results of a regression to test associations between wave strengths and working memory performance (d‐prime) indicated the overall model was significant (F(7,50) = 2.262, p = 0.044). The model including forward wave strength in the memory set presentation period and probe period provided evidence supporting the alternative hypothesis: BF10 = 8.468. Effects were driven by a significant interaction, where forward wave strength in the memory set presentation period related to higher accuracy (t = 2.234, p = 0.030), while forward wave strength in the probe presentation period related to worse accuracy (t = −2.595, p = 0.012). Group was also a significant predictor (t = 2.105, p = 0.040). Figure 6 depicts the marginal effects of forward waves in the memory set presentation and probe periods on d‐prime scores.

DISCUSSION

Our results showed experienced mindfulness meditators generate stronger forward traveling alpha waves compared to nonmeditators, both while resting with their eyes‐closed and while performing a visual attention task. Furthermore, meditators showed weaker backward traveling waves while resting compared to nonmeditators. Meditators also showed an increase in the strength of their backward waves from eyes‐closed resting to the visual attention task, while nonmeditators showed no difference. Finally, forward and backward traveling waves were differently modulated during different periods of a working memory task, and some modulations were associated with working memory performance. Forward wave strength also increased from eyes‐closed resting to the visual attention task, and forward wave strength was weakly related to task performance. These results have important implications for our understanding of the neural effects associated with meditation, providing indications of specific neural mechanisms, and supporting a predictive processing explanation of how meditation alters brain activity to affect subjective experience and cognitive function. Our results also provide the first evidence that forward and backward wave strengths are likely to be modifiable by mental training.

To understand our results in context, we note that research suggests that the brain is broadly organized according to a hierarchy of complexity. Within this hierarchy, the general pattern is that progressively more posterior regions process sensory inputs and are informed by brief time windows. ³⁴ , ⁵⁹ , ⁶⁰ In contrast, more frontal regions integrate information from an increasingly diverse range of sensory inputs and broader time windows and represent the environment in a more complex and abstract manner. ³⁴ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² While posterior regions are involved in higher‐order cognition (e.g., occipital and parietal regions are involved in working memory ⁶³ ), evidence indicates that the cognitive functions provided by more posterior regions are more constrained than those implemented in frontal regions. ⁶² Frontal regions also represent the terminus for processing of sensory information and are the origin of most top‐down influences. ⁵⁹ , ⁶¹ , ⁶² Although there are anatomical deviations from this general posterior–anterior organization (e.g., the auditory cortices are located in lateral regions aligned with the middle of the posterior–anterior axis), phase gradients in EEG data indicate that EEG predominantly captures a broad anterior–posterior pattern of traveling waves. ⁴⁴ , ⁶⁴ , ⁶⁵ , ⁶⁶

In addition to these anatomical considerations, evidence suggests that forward cortical traveling alpha waves measured from EEG data reflect bottom‐up information flow from sensory to frontal regions, while backward traveling waves reflect top‐down information flow from frontal to sensory regions. Forward traveling cortical alpha waves have been shown to be generated in response to visual stimuli, and have been suggested to index an underlying mechanism for processing sensory information. ⁴² Some research has suggested that the strength of forward traveling waves reflects the strength of information flow (and prediction errors) transmitted to progressively higher layers in the cortical hierarchy. ⁴⁵ As such, the stronger forward waves observed in meditators may align with an important component of meditation practice—the continual direction of attention to sensations. However, the exact function of forward traveling alpha waves has not been conclusively determined. Some recent research has suggested that the direct function of forward traveling alpha waves may be to provide bottom‐up spreading inhibition of non‐task–relevant neural activity, which can enhance the signal‐to‐noise ratio of sensory information processing. ⁴⁴ , ⁵⁰ Following this interpretation, our results may suggest that, in meditators, the putative inhibitory function of the forward traveling waves is more pronounced than in nonmeditators. This proposed increase in bottom‐up inhibition may serve to enhance the signal‐to‐noise ratio of task‐relevant sensory processing in the Go/No‐go task, where meditators showed stronger forward waves primarily during the post‐stimulus blank screen (when processing of the visual stimuli was no longer task‐relevant). This may have enabled the improved task performance shown by meditators, with a weak correlation present between task performance and forward traveling waves in the latter half of Go/No‐go trials.

However, a clear functional interpretation of the different pattern of traveling wave strengths in meditators during eyes‐closed resting is not provided by our results. If forward traveling alpha waves do reflect bottom‐up spreading inhibition, then the stronger forward waves in meditators during eyes‐closed resting could reflect increased inhibition to protect other processing (e.g., of attended sensory information). In this context, our results might suggest that meditators may exhibit increased bottom‐up inhibition of visual processing in the eyes‐closed state. This could either enhance the brain's ability to process other sensory signals (e.g., the sound of their breath and bodily sensations experienced even during rest), ³⁵ , ³⁷ , ³⁸ or facilitate a general reduction in processing when the environment does not impose any task demands. In contrast, backward waves are spontaneously generated and have been suggested to index the strength of higher‐order predictions generated by frontal brain regions. ⁴² , ⁴⁵ As such, the lower strength of backward waves during eyes‐closed resting in the meditator group aligns with a component of meditation practice—nonjudgmental present‐moment awareness of current sensations. ¹ However, again, since our resting recordings do not provide information about the cognitive processes engaged, the interpretation of the function of the weaker backward waves is speculative.

Our results, therefore, at least partially align with theoretical explanations related to bottom‐up predictive processing effects of meditation. These explanations suggest that repeated allocation of attention to sensations during meditation practice might produce neuroplastic changes that increases the precision with which bottom‐up prediction errors are processed. ³⁵ , ³⁷ , ³⁸ Our results in the Go/No‐go task indicate increased bottom‐up inhibition of nonrelevant sensory information, suggesting increased precision of the processing of relevant sensory information. This might indicate that, in meditators, the putative inhibitory function of the forward traveling wave acts more strongly than in nonmeditators to increase the signal‐to‐noise ratio of sensory processing in the Go/No‐go task (with meditators showing stronger forward waves during the post‐stimulus blank screen when sensory processing is no longer task‐relevant). Our results showing stronger forward waves in meditators during eyes‐closed resting may also indicate stronger inhibition to protect the processing of sensory information in meditators. However, the results may simply reflect increased inhibition of visual information processing in meditators during eyes‐closed resting. Further research is required to determine the functional relevance of the forward wave effects we observed. We also note that when the stronger forward waves and weaker backward waves in meditators during eyes‐closed resting are considered holistically, a novel explanation of the effects of meditation within the predictive processing framework can be proposed. We explain this later in our discussion.

Our findings regarding backward wave strength also suggest the effects of long‐term meditation are not driven by an increase in top‐down attentional control processes, ¹⁶ which would be reflected by stronger backward waves. Instead, our results suggest the effects of meditation may be more likely driven by increases to bottom‐up processing, ¹⁴ , ¹⁵ although additional work is required to determine the exact functional role of forward traveling waves, as well as their specific anatomical pathways. The lower strength of backward traveling waves in meditators also supports a predictive processing account of meditation, suggesting practicing nonjudgment present‐moment awareness produces a trait reduction in the generation and elaboration of higher‐order predictions. ³⁷ , ³⁹

Meditators also showed an increase in the strength of their backward waves from eyes‐closed resting (where they showed lower backward wave strength) to performance of the visual attention task (where they showed the same backward wave strength as nonmeditators). In contrast, nonmeditators showed no difference between the eyes‐closed resting and task conditions. It is worth noting that in addition to the visual processing requirements of the task, higher‐order cognitive functions were critical for accurate task performance. These higher‐order functions include response inhibition, performance monitoring, working memory to remember task demands, and generation of predictions about upcoming stimuli. In this context, it is perhaps more interesting to explore the finding that nonmeditators generated the same backward wave strength during resting as they did during the task. This result suggests that nonmeditators engaged in similar levels of higher‐order cognitive activity while simply resting as they did while undertaking demanding executive functions. We suggest that their high backward wave strength during resting reflects the continual generation and elaboration of higher‐order predictions even in the absence of task requirements. These higher‐order predictions likely consist of thoughts about oneself, the future, or the past, ⁶⁷ , ⁶⁸ which may arise via the instantiation of counterfactual or fictive prediction errors in the absence of task demands. ³⁷ , ⁴⁹ , ⁶⁹ Viewed from this perspective, the lower strength of resting backward waves shown by meditators may reflect the reductions in rumination and worry suggested as a psychological mechanism by which mindfulness improves well‐being and protects against depression. ⁷⁰ This result might be particularly valuable for translation to clinical applications. In particular, recent research has shown that individuals with schizophrenia display elevated levels of backward traveling waves, ⁷¹ and individuals with post‐traumatic stress disorder show lower posterior to frontal alpha functional connectivity (a pattern that is likely to align with reduced forward traveling waves). ⁷²

Considered together, a possible (but tentative) holistic explanation for the weaker backward waves and stronger forward waves in meditators during eyes‐closed resting is that the pattern may reflect an altered neural strategy of inhibition and feedforward/feedback processing. Evidence suggests that the brain functions via a balance of inhibition and excitation, with the maintenance of both influences being critical to healthy brain function. ⁷³ Evidence also indicates that forward and backward traveling alpha waves are anticorrelated, potentially reflecting a system that maintains balance in part by modulating the relative flow of these inhibitory processes in different directions. ⁴² The typical pattern of neural activity (exhibited by nonmeditators) may be that when visual input is diminished by eye closure, the neural system is compelled to provide cortical inhibition via backward waves. These backward waves may then fulfill their top‐down inhibitory function, acting to direct attention toward other sensory streams or counterfactual processing (thoughts of the past and future). In contrast, the practice of meditation, where attention is progressively directed to bodily sensations (eliciting concurrent increases in forward traveling waves which may function to inhibit nonattended information) may have strengthened the ability of experienced meditators to engage bottom‐up inhibitory mechanisms. Therefore, in contrast to the pattern shown in nonmeditators, experienced meditators may instead implement inhibitory functions in a bottom‐up manner via forward waves even in the eyes‐closed resting state, such that backward traveling alpha waves (and their associated top‐down inhibitory function) are not compulsively engaged. During eyes‐closed resting, these bottom‐up inhibitory mechanisms may then provide the necessary inhibitory function, eliminating the necessity to increase top‐down feedback to maintain a functional excitation/inhibition balance. This may then enable meditators to rest without generating the same strength of top‐down inhibitory engagement. This lower engagement of top‐down processing during rest might then relate to less generation/elaboration of counterfactual processing (which may underpin mindfulness‐related reductions in thoughts about the past and future). Further research is required to examine the accuracy of this speculative explanation.

Our results also have implications for our understanding of brain function more generally, providing evidence for traveling waves as a correlate of large‐scale network coordination which can be adjusted to meet the demands of the current environment. In particular, our results support an association between forward waves and visual attention, ⁴⁰ , ⁴² , ⁴⁷ with both groups showing an increase in forward wave strength from eyes‐closed resting to the visual cognitive task. However, our results extend previous research indicating that visual stimulation decreases backward waves ⁷⁴ to show that backward waves can increase from eyes‐closed resting to a visual attention task that requires higher‐order processes for accurate task performance (in meditators). This provides additional support to the suggestion that backward waves are important when top‐down expectations are involved in a task. ⁴³ , ⁴⁶

Our results also show that forward and backward wave strength is modulated by specific working memory processes. Periods where participants attempted to encode stimuli for later recall were associated with higher forward wave strength, and stronger forward waves during this period predicted task accuracy across participants, supporting suggestions that forward waves index underlying mechanisms for encoding visual information. ⁴⁰ , ⁴⁸ In contrast, the period where participants attempted to recall the memory set was associated with increased backward wave strength. This increase is perhaps indicative of a top‐down increase of precision to reinstate cortical representations of memory set items, enabling working memory recall. ⁴⁸ , ⁴⁹

While our results provide several interesting findings, our study is not without limitations. First, an obvious limitation is the use of a cross‐sectional design, which does not allow for the inference of causality. However, a longitudinal study including participants with the same average degree of experience with meditation practice as the current study (an average of 8 years since participants started practicing meditation) would be an incredibly difficult undertaking. Our results also indicated only small to moderate effect sizes, the detection of which was only made possible with our relatively large sample size. Longitudinal studies of the effects of meditation (which have to date always included much less experience with meditation than the current study) might be expected to produce even smaller effect sizes, reducing the chance that longitudinal studies would detect significant effects. Having noted this point, it may be interesting for future research to explore whether a reduction in resting backward waves could be a marker of reduced rumination or worry following a mindfulness intervention for depression, as reduced rumination or worry has been suggested to reflect clinical mechanisms of action. ⁷⁰ Traveling waves might then have utility in providing feedback about progress with mindfulness practice, or as a predictor of who might benefit most from a mindfulness intervention.

Second, there was only a weak correlation between Go/No‐go task performance and forward traveling wave strength during the later part of the task epochs, querying the functional relevance of traveling waves. However, we note that overall, participants performed highly accurately in the task, so this may reflect a ceiling effect. Additionally, our analysis focused only on the strength of the traveling waves. Previous research suggests that speeds of forward waves relate to reaction times, ⁵⁰ so traveling wave speed might be worth exploring in future research. Our primary analysis of traveling waves also used a 1‐s period after the presentation of each Go and No‐go stimulus, and our correlations tested traveling wave strength averaged across all trials within each participant. Traveling wave strengths have been shown to dynamically shift in response to task demands. ⁴⁷ In support of this suggestion, our analysis indicated correlations only between behavioral performance and forward traveling waves in the later part of the Go/No‐go epochs. This effect may indicate that the forward traveling waves fulfill a bottom‐up inhibition function, reducing the processing of the nonstimulus presentation period, which may have enabled participants to devote more processing to selecting the appropriate behavioral response. Nonetheless, future research may benefit from implementing more fine‐grained analyses.

Third, the exact anatomical pathway and cortical generators of the broad traveling wave pattern have not been conclusively determined. Previous research has argued that traveling waves are a more appropriate way to analyze EEG data than many other methods (including event‐related potential analyses and analyses of frequency band power) due to their ability to capture the relationship between spatial and temporal patterns. ⁷⁵ , ⁷⁶ Indeed, traveling waves have been shown to capture more of the variance in EEG data than event‐related potential methods, ⁷⁵ and evidence suggests that event‐related potentials are produced by the averaging of traveling waves rather than the other way around. ⁵⁰ , ⁶⁴ Nonetheless, it might be suggested that our traveling wave results were driven by event‐related responses in progressive cortical sources with progressively increasing peak and trough delays. However, exploratory analysis of the early and later periods within the Go/No‐go epochs showed that the traveling wave effects were driven by differences later in the epoch, during the period less influenced by visual event‐related potentials. This suggests our results are unlikely to be explained by visual processing event‐related responses in progressive cortical sources with increasing peak and trough delays. Still, future research using alternative methods (e.g., magnetoencephalography) may detect additional traveling wave directions related to meditation (perhaps including lateral waves), or provide more certainty about the origins and functions of the cortical traveling waves we have observed. We note that the combination of volume conduction of electrical signals through the brain and scalp, as well as the sparse array of EEG electrodes, may together act as a low‐pass spatial filter, such that only long spatial wavelengths are captured in traveling wave analyses. ⁷⁷ As such, it may be that lateral traveling waves from regions that are not aligned with the posterior–anterior axis are thus filtered out of EEG data, or that they are weaker in comparison to waves traveling in the posterior–anterior direction (or both). Nonetheless, based on the available evidence, we can conclude that meditators appear to exhibit stronger bottom‐up waves propagating up the cortical hierarchy during both task and resting conditions, while showing weaker top‐down waves traveling from frontal regions down the cortical hierarchy during eyes‐closed resting.

Fourth, volume conduction poses a common issue in analyses of connectivity between brain regions ⁷⁸ and might be considered as a confound to our results. However, we note that the effects of volume conduction are instantaneous, decay quickly over distances more than 2 cm, and are negligible at distances >10 cm. ⁷⁹ , ⁸⁰ Since traveling wave analyses are only sensitive to progressively increasing phase lags, the instantaneous effects of volume conduction on phase correlations between electrodes can produce only a relatively localized spatial blurring influence on the global phase gradient patterns that produce the traveling waves measured at the scalp. ⁶⁴ As such, researchers who examine broad frontal to posterior traveling waves have convincingly argued that local neural dynamics are unlikely to explain the consistent anterior–posterior ordering of these waves. ⁴⁴ , ⁷⁴ , ⁷⁷ Our results could be explained by a large number of sources activated in sequence to give the appearance of a continuous traveling wave, but in this case, our interpretations of the function of the signal that travels up the cortical hierarchy would still apply ⁶⁵ and the patterns we observed would still be of interest in understanding brain activity in experienced meditators. Furthermore, in support of our argument that our results are not produced by the effects of volume conduction of a limited number of sources, our exploratory analysis applying a surface Laplacian transform to the data prior to the traveling wave analysis to constrain the signals recorded at each electrode to more localized cortical regions recapitulated our primary results (see Supporting Information).

Finally, prior to resting recordings, we instructed participants to rest with their eyes closed and simply let their mind do as it pleased, but not to meditate. However, it is impossible to confirm that meditators did not habitually commence a meditation practice. If they were meditating instead of resting, our results might reflect the state effects of meditation rather than trait differences in brain activity. The lack of ability to verify the cognitive processes of participants during these resting recordings also limits our ability to interpret the functional relevance of the traveling waves during resting. As such, while we have interpreted our results through the lens of the functional associations between traveling waves demonstrated by previous research and the lens of predictive processing theories of the effects of meditation (which provide an elegant alignment in our results), our results do not provide certainty that the differences between the groups in traveling wave patterns during resting are related to the functions of traveling waves demonstrated via manipulations of cognitive processes with experimental tasks by previous research. Alternative explanations may be possible and should be explored in future research. However, we note that meditators also showed stronger forward waves during the Go/No‐go task when they would not have been able to meditate, indicating that the greater strength of forward waves is a trait effect associated with meditation experience. Additionally, even if meditators did habitually commence meditating, this would still indicate the natural habits of their brain. As such, our results still reflect differences in brain activity associated with experience in meditation.

CONCLUSION

Our results showed experienced meditators generated stronger forward traveling cortical alpha waves than nonmeditators while resting with their eyes closed and during a visual attention task. Meditators also generated weaker backward waves while resting, but increased their backward wave strength to be equivalent to nonmeditators during the visual task. These results at least partially align with a predictive processing perspective of the effects of meditation. The stronger forward waves that we detected in meditators perhaps reflect increased bottom‐up spreading of inhibition of nonrelevant neural activity, enabling stronger signal‐to‐noise ratios for the processing of attended signals. The stronger forward waves in meditators may have arisen because of meditators’ attention training to sensations, which may have strengthened their ability to engage these inhibitory mechanisms to increase the precision with which sensory information is accorded. In contrast, weaker backward waves might indicate less generation and elaboration of higher‐order beliefs (or less mind‐wandering) when no task demands are present.

AUTHOR CONTRIBUTIONS

All authors contributed to drafting and revising the manuscript and approved its final version. N.W.B. conceived of the study, was responsible for designing the study, collecting data and supervising data collection, preprocessing data and applying data analyses, statistical analysis, and writing the initial draft of the manuscript. A.T.H. supervised data collection, and provided interpretation and manuscript reviews. K.G. provided input into the data analysis, interpretation, and manuscript reviews. M.P.N.P., J.H., A.W.C., N.C.R., and B.M.F. provided interpretation and manuscript reviews. P.B.F. contributed to the design of the study, and provided interpretation and manuscript reviews.

COMPETING INTERESTS

In the last 3 years, P.B.F. has received equipment for research from Neurosoft and Nexstim. He has served on a scientific advisory board for Magstim and received speaker fees from Otsuka. He has also acted as a founder and board member for TMS Clinics Australia and Resonance Therapeutics. The other authors declare that they have no conflicts of interest.

PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1111/nyas.15401.

Bailey, N. W. , Hill, A. T. , Godfrey, K. , Perera, M. P. N. , Hohwy, J. , Corcoran, A. W. , Rogasch, N. C. , Fitzgibbon, B. M. , & Fitzgerald, P. B. (2025). Experienced meditators show greater forward traveling cortical alpha wave strengths. Ann NY Acad Sci., 1550, 173–190. 10.1111/nyas.15401

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author, N.W.B., upon reasonable request.