Mindfulness-based neuroscience studies have generally adopted a longitudinal approach instead of targeting situational practice effects. Objective measures of brain functions, such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI), have generally been adopted to test the efficacy of mindfulness interventions. For example, in a study of the EEG power of the experienced Rinpoche, with meditation experience of >10,000h, gamma power enhancement was evident even during a resting state (Lutz et al., 2004), and this effect was sustained even during non-rapid-eye-movement (NREM) sleep (Ferrarelli et al., 2013). For meditation novices, EEG measures following an 8-week MBSR program have been widely associated with convergent and consistent outcomes. Researchers have found MBSR practitioners tend to exhibit stronger beta power in the frontal lobe during mindfulness practice than during the resting state (Gao et al., 2016). Similarly, MBSR practitioners exhibited elevated alpha power in the occipital and right temporal lobes (Ahani et al., 2014). Theta band power was reported to increase in the central, parietal, occipital, and left and right temporal lobes after the MBSR intervention by Ahani et al. (2014). Furthermore, MBSR practitioners exhibited lower delta power in the central–parietal area after MBSR intervention (Gao et al., 2016), and patients with chronic insomnia were also found to have lower delta power in the central lobe during NREM sleep after MBSR intervention (Goldstein et al., 2019). Overall, MBSR intervention is generally believed to enhance high-frequency EEG power (i.e., beta and gamma); however, its effect on low-frequency EEG power (i.e., theta and delta) remains uncertain.

Such neurophysiological evidence concerning MBSR is consistent with the evidence on stress reduction and cognitive improvement (Davidson et al., 2003). For example, the beta power in the frontal and temporal lobes of participants without stress stimuli were higher than that of the participants with stress stimuli (Hayashi et al., 2009), suggesting a negative relationship between beta power and stress level. As for the low-frequency bands, healthy adult participants with a high perceived stress level had higher delta and theta activity in the frontal, central, and parietal lobes, compared with those who had a low perceived stress level (Luijcks et al., 2015). Another study on stress revealed that participants exhibited lower theta power under acute stressful conditions (Gärtner et al., 2014). Furthermore, another study highlighted how cognitive tasks elevated gamma power in comparison with the control conditions (Fitzgibbon et al., 2004), and a study on vigilance suggested that highly vigilant states corresponded to delta power suppression (Smallwood and Schooler, 2015). Overall, EEG evidence generally reveals that stress reduction is positively correlated to low-frequency-band power and negatively correlated to high-frequency-band power, whereas cognitive performance and vigilance state are positively correlated to high-frequency-band power and negatively correlated to low-frequency-band power. EEG spectral powers can serve as objective functional markers of cognitive enhancement and stress reduction.

In addition to the post-intervention effect of an 8-week MBSR program, the situational practice effect of such programs has recently attracted the attention of mindfulness researchers seeking to identify the variations between distinct mindfulness practices. The MBSR program involves a series of mindfulness practices (Kabat-Zinn, 1994; Davidson et al., 2003), such as mindful breathing, body-scan, compassion meditation, and open-monitoring. Davidson described the distinct practices associated with various cognitive effects in his book (Goleman and Davidson, 2017). Among the practices, mindful breathing and body-scan were most frequently used in previous studies (Ahani et al., 2014; Wahbeh et al., 2014; Isbel et al., 2019), and both of them are associated with interceptive perceptions. However, the distinction between mindful breathing and body-scan can lead to different interoceptive effects on behavior and brain mechanisms. The fixed attention and relaxation in mindful breathing may differ from the attentional shifts to and from various body parts during body-scan. Recent studies have assessed the diverse effects of mindfulness practices using questionnaires and behavioral measures. One study demonstrated that people who practiced breath-focused meditation had a more nonjudgmental attitude toward themselves, showed more self-compassion, and experienced less emotional regulation difficulty. In contrast, those who practiced body-scan showed increased capabilities to describe their feelings and reduced rumination tendencies (Sauer-Zavala et al., 2013), suggesting that mindful breathing and body-scan affect different brain functions. Some studies have indicated that body-scan yields more positive outcomes for practitioners than does breathing. For example, body-scan practice leads to a major increase in body awareness and a decrease in thought contents, whereas breathing practice engenders a comparatively less intense change (Kok and Singer, 2017). Another study on veterans with post-traumatic stress disorder disclosed that participants who practiced body-scan exhibited greater mindfulness improvement than their breathing group counterparts (Colgan et al., 2016). Studies with self-rating scales have established that body-scan seems to provide more promising benefits than mindful breathing. On the other hand, neuroimaging studies on mindfulness generally examined mindfulness as a self-focused attention process, usually achieved through mindful breathing (Ferrarelli et al., 2013; Matiz et al., 2019). Shifted attention process, or body-scan practice, was rarely addressed before. The brain mechanisms targeted by the two mindfulness practices remain elusive at the current stage of research. Therefore, we adopted EEG measurement to examine the functional distinctions between mindful breathing and body-scan practices.

Collectively, the EEG spectral alterations associated with various mindfulness practices remain elusive, and whether the mindful breathing and body-scan practices take effect are yet to be tested. We first proposed that the functional distinction between these two mindfulness practices lies in the spatio-spectral disparity of the EEG, meaning that frontal power elevation follows body-scan, and parietal power reduction follows mindful breathing after the 8-week MBSR intervention. Second, we proposed that the functional distinction between practices is amplified by the post-intervention effect (not observed in novices). Accordingly, we designed an EEG experiment to assess the neurophysiological changes in terms of both situational mindfulness practice and post-intervention effects. In addition, we conducted the same protocol with a waitlist control group without MBSR training to enable cross-group comparison.

Forty-three volunteers were invited to participate in the MBSR training and register for an 8-week MBSR course in the fall of 2019. Before the course started, all volunteers were required to undergo one of the three orientation sessions to understand the details of the procedure, compensation, potential risks, and the contributions of this study. Potential participants were screened by applying the following exclusion criteria: being outside the age range of 20 to 80years old; having prior experience of mindfulness meditation; having a metabolic illness, any history of mental illness, neural illness, or epilepsy; being a smoker or drug addict; and having any bodily metallic implant, claustrophobia, or pregnancy. Ten people who did not meet the criteria were excluded, and eventually 33 (4 men and 29 women) people aged between 29 and 68years participated in the study (mean=47.46, SD=8.79; 18 participants in the MBSR group and 15 participants in the waitlist control group). The participants were new to mindfulness training before the study, and they received a free 8-week MBSR course for electing to participate. All of the participants provided written informed consent prior to participation. This study was reviewed and approved by the Taipei Medical University Joint Institute Review Board (TMU-JIRB, project number N201905049).

Each participant completed the EEG–fMRI experiments twice. For the MBSR group, two experiments were scheduled before and after the 8-week MBSR training (pre-test and post-test). For the waitlist control group, two scans were performed 8weeks apart (same pre-test and post-test); however, they were prohibited from accessing and receiving information regarding mindfulness practice during this period. The MBSR group participated in an 8-week standardized MBSR course, as proposed by Santorelli et al. (2017). The MBSR course was instructed by a licensed MBSR instructor. MBSR classes were conducted weekly for 2.5h for 8weeks in addition to a one-day mindfulness workshop. These weekly meetings involved the development of various mindfulness skills, dialogue and reflection on mindfulness home practice, and practice segments. The participants performed types of mindfulness practice, such as a sitting meditation entailing breathing, mindful listening, body-scan, and mindful yoga. Participants were also assigned daily homework comprising both formal and informal meditation activities. Formal activities, including body-scan practice, sitting meditation, mindful yoga, mountain/lake meditation, or loving kindness meditation, required 45min to complete each day. The participants were asked to complete a practice and information sheet. The participants in the waitlist control group were instructed to maintain their usual life activities but not engage with any mindfulness-related information during the 8weeks. After the post-test experiment, the waitlist control group participants underwent an MBSR program for compensation.

The study procedure is presented in Figure 1. In the orientation sessions, the study was explained to the participants; in particular, the use of the fMRI machine was detailed, as this study employed the EEG–fMRI simultaneous scanning technique (however, only EEG and behavioral information are included in this report). Thereafter, the participants finished the informed consent form and a questionnaire battery. The questionnaires were the Five Facet Mindfulness Questionnaire (FFMQ; Baer et al., 2008)—Taiwanese version, the Difficulties in Emotion Regulation Scale (DERS; Gratz and Roemer, 2004), and demographic questionnaires. The questionnaire battery took approximately 30min to complete.

After the orientation sessions and within 2months before the MBSR course, the researchers made an appointment for the participants to come into the laboratory for the pre-test experiment. The participants were introduced to the tasks that they would perform during the scanning while an EEG cap was being set up. In the scanning section, participants laid down in the MRI machine to perform four tasks. A mirror setting was installed with projectors for visual presentations. The first task was the resting state (resting 1): Participants were instructed not to think of anything specific with their eyes closed, not to move their heads, and not to fall asleep. The second task was the mindful breathing (breathing): Participants were instructed to pay attention to their natural breath and focus on the sensation near their noses during respiration. Whenever they realized that they were getting distracted, they were instructed to press a button on the right hand. The third task was the body-scan (body-scan): Participants were instructed to perceive the most salient body sensations and shift focus between various body parts during the session. Similar to the breathing task, they were instructed to press a button when they were distracted. Finally, an additional resting state (resting 2), identical to resting 1, was conducted to evaluate whether the brain status was returned to baseline after the mindful practices. Every task lasted 5min, and if participants opened their eyes during a task, they would see a white fixation cross. The fixation image was black with a white cross in the center (see Supplementary Material). We used E-Prime 2.0.10 (Psychology Software Tools, Pittsburgh, PA, United States) for instruction presentation and response recording. After the pre-test experiment, the MBSR group participated in an 8-week MBSR course while the waitlist control group went about their usual life. Within 3months after the completion of the MBSR course, all participants from both groups participated in the post-test experiments that were identical to the pre-test sessions. From pre-test sections to post-test sections, the mean time duration of all participants was 92.56days (SD=20.77days).

It is noted that our task sequence had no counter-balance design. The consideration behind will be discussed in the limitation section.

In each experiment, simultaneous EEG–fMRI signals were recorded for each functional scan using a 3T PRISMA MRI scanner (Siemens, Erlangen, Germany). The 32-channel EEG data were recorded using an MR-compatible system (Brain Products GmbH, Gilching, Germany) that was positioned according to the international 10/20 system. The built-in impedance in each electrode was 5 kΩ, and abrasive electrode paste (Abralyt HiCl) was used to reduce the electrode–skin impedance. The EEG signal was recorded synchronously with the MR trigger using Brain Vision Recorder (Brain Products) with a 5k-Hz sampling rate and a 0.5μv voltage resolution (reference at FCz). A band-pass filter was set with cutoff frequencies of 250Hz and 0.0159Hz, and an additional 60-Hz notch filter was employed. Here, we have reported only the EEG outcomes, as the MRI results were designed to be reported separately.

Recorded EEG data were re-sampled to 50,000Hz and then corrected for gradient-induced artifacts as Becker et al. (2005) suggested. Ballistocardiographic artifacts were corrected using the adaptive average subtraction method, and the R-peak intervals were estimated from the electrocardiogram electrode through Analyzer 2.1 (Brain Products) after the data were down-sampled to 250Hz. The EEG data were then filtered with a 0.2–40Hz band-pass FIR filter. Thereafter, an independent component analysis (ICA, with Infomax method) to eliminate artifacts caused by electrooculogram (EOG), electromyogram (EMG), and electrocardiogram (ECG) artifacts and the remaining MRI-induced artifacts. The artifact-free data were referenced to an average electrode across whole scalps, as recommended by Allen et al. (2004) and Davidson (1998). The processed data were converted into frequency domain representations using short-time Fourier transformation and Welch’s periodogram method. Specifically, each 5-min EEG signal (75,000 points) was divided into 256-point segments using the Hanning window with 128 points overlapped. Each segment was zero-padded to 512 points, followed by a 512-point fast Fourier transformation. All the resultant spectra were subsequently log-transformed (10log10, results in dB; Allen et al., 2004) and averaged over segments. Finally, average EEG band powers were calculated in the frequency ranges of 1–4Hz, 4–8Hz, 8–13Hz, 13–20Hz, 20–30Hz, and 30–40Hz, representing delta, theta, alpha, low-beta, high-beta, and low-gamma band powers, respectively (Deuschl and Eisen, 1999; Rangaswamy et al., 2002; Teplan, 2002; Hauswald et al., 2015; Abhang et al., 2016; Grent-'t-Jong et al., 2018). These signal processes were conducted using the EEGLAB toolbox 13.6.5b (Swartz Center for Computational Neuroscience, University of California San Diego; Delorme and Makeig, 2004) with Matlab R2019a. The calculation of power spectrum density (PSD; in “dB”) also followed the suggestion of Delorme and Makeig (2004). This study filtered out frequencies above 40Hz since MR gradient-induced artifacts would severely contaminate the spectra over 40Hz (Solana et al., 2014). Therefore, the EEG band of highest frequency in this study was low-gamma band (30–40Hz).

To ensure that all parameters (EEG band power and FFMQ and DERS scores) were fitted for parametric analyses, Kolmogorov–Smirnov tests were performed to examine normality. The chi-square test and t-test were conducted to examine whether demographic features, FFMQ, and DERS levels were different between the MBSR group and the waitlist control group before the MBSR intervention. Given that our working hypothesis is particularly to examine the effects of MBSR intervention, multiple paired t-tests were performed to examine whether a significant difference existed in terms of FFMQ and DERS in both groups before and after the MBSR intervention.

To examine whether EEG power spectra changed before and after the MBSR intervention, a paired-sample t-test was performed for every 0.5-Hz frequency bin in all four tasks (resting 1, breathing and body-scan, resting 2) in both groups and for both channels, Fz and Pz. In order to examine individual variations between EEG activities and behavioral measures, correlation analysis was then performed to reveal whether a correlation existed between the change of EEG band power before and after the MBSR intervention and the change of FFMQ and DERS scores in both groups. The change of EEG power was calculated by subtracting the pre-test EEG band power from the post-test EEG band power (post-test–pre-test), and the change in FFMQ and DERS scores was calculated similarly.

Table 1 presents the mean FFMQ and DERS scores rated by both the MBSR and control groups during the pre-test and post-test sessions. The MBSR group (N=18) scored 114.94±16.99 on the FFMQ_pre, 142.56±23.24 on the FFMQ_post, 97.78±20.72 on the DERS_pre, and 84.72±23.21 on the DERS_post. The control group (N=15) scored 116.80±16.45 on the FFMQ_pre, 115.87±14.65 on the FFMQ_post, 84.87±22.53 on the DERS_pre, and 98.80±16.12 on the DERS_post. All these scores validated the assumption of normality (Kolmogorov–Smirnov test: ps>0.05). In terms of resting-state EEG activity, the logarithmic powers of the delta, theta, alpha, low-beta, high-beta, and low-gamma bands over channels Fz and Pz recorded in the two sessions for both groups also validated the assumption of normality (Kolmogorov–Smirnov test: ps>0.05). The only exception was the post-test low-gamma score for channel Pz in the waitlist control group (p=0.032). Because our analysis focused on the main effects of the MBSR intervention group, parametric tests were employed in the subsequent analyses.

Thirty-three participants (29 women and 4 men, aged 47.67±8.79years) were recruited in this study, of whom 15 (12 women and 3 men, aged 46.67±8.03years) served as the control group and 18 (17 women and 1 man, aged 48.50±9.52years) served as the MBSR group. As evident in the demographic data and questionnaire results presented in Table 1, no significant difference was apparent in terms of gender (χ² [1]=1.60, p=0.206), age (t [31]=0.59, p=0.559), and educational level (t [31]=−0.90, p=0.377) between the waitlist control and MBSR groups.

A between-group analysis showed that the differences between the two groups in both questionnaire responses were significant in the post-test session (FFMQ: t [31]=5.32, p<0.001; DERS: t [31]=−2.08, p=0.046) but not in the pre-test session (FFMQ: t [31]=−0.32, p=0.753; DERS: t [31]=0.39, p=0.702). A within-group analysis showed that after the MBSR intervention, the MBSR group exhibited a significant increase in the FFMQ score (t [17]=5.32, p<0.001) and a decrease in the DERS score (t [17]=−2.14, p=0.047) from the scores before the intervention, whereas the FFMQ (t [14]=−0.27, p=0.796) and DERS (t [14]=0.76, p=0.458) scores of the waitlist control group remained unchanged.

After an 8-week mindfulness training, the effect of mindfulness practices on resting-state EEG activity was examined. Figures 2, 3 illustrate the comparisons of PSD between the pre-test and post-test for Fz and Pz electrodes, respectively; the black asterisks indicate the significant differences in EEG powers identified at certain frequency bins. Student’s t-test was applied for every 0.5-Hz frequency bin ranging from 0Hz to 40Hz (0–0.2Hz was filtered out during preprocessing), followed by false discovery rate correction for conducting multiple comparisons across frequency bins and channels (Keselman et al., 2002).

For the MBSR group, compared with the pre-test session, the EEG activity of Fz and Pz sites measured in the post-test session revealed the following: (1) in resting 1, high-beta and low-gamma powers significantly increased (Fz: 24.5–40Hz except 26Hz, ts≥2.56, ps<0.05; Pz: 20–20.5Hz and 24.5–40Hz, except 25.5, 26, and 35.5Hz, ts≥2.41, ps<0.05); (2) in breathing, high-beta and low-gamma powers significantly increased (Fz: 24.5–40Hz except 25.5–26Hz, ts≥2.31, ps<0.05; Pz: 27–40Hz, ts≥2.70, ps<0.05) and delta powers significantly decreased (Fz: 0.2–2Hz, ts≥−2.72, ps<0.05; Pz: 0.2–3Hz, ts≥−2.61, ps<0.05); (3) in body-scan, high-beta and low-gamma powers significantly increased (Fz: 27–31.5Hz, ts≥2.83, ps<0.05; 34.5–40Hz, ts≥2.74, ps<0.05; Pz: 27–31.5Hz, ts≥3.05, ps<0.05; 34.5–40Hz, ts≥2.94, ps<0.05) and delta power significantly decreased only at Fz (0.2–1.5Hz, ts≥−2.76, ps<0.05); and (4) in resting 2, high-beta and low-gamma powers significantly increased (Fz: 27–31.5Hz, ts≥2.80, ps<0.05; 34.5–40Hz, ts≥2.85, ps<0.05; Pz: 25–31.5Hz except 25.5 and 26Hz, ts≥2.99, ps<0.05; 35–40Hz, ts≥3.56, ps<0.05) and delta powers significantly decreased only at Pz (0.2–0.5Hz, ts≥−2.80, ps<0.05). For the waitlist control group, no significant difference was observed in EEG power between the post-test and pre-test sessions in all four tasks (lower panel of Figures 2, 3, ps>0.05).

Figure 4 presents the spatial distribution of the EEG spectral power differences between the post-test and pre-test of the MBSR group in the four tasks. Paired-sample t-tests were applied to all channels to compare the difference between the post-test and pre-test conditions. The channel locations that do not exhibit significant differences (ps>0.05) were marked as zero difference and depicted in green in the topoplots. The results showed that the high-beta and low-gamma bands had significant spectral differences across the whole scalp in resting 1 and resting 2. There were also some small-scale differences in the delta and low-beta bands in the lateral frontal, left parietal, and right temporal areas in resting 1 and resting 2. The delta band had significant EEG power differences in the frontal and parietal areas, and the high-beta and low-gamma bands had significant power differences across the whole scalp in the breathing and body-scan tasks. There were also significant differences in the theta band in the frontal and occipital areas in the breathing and body-scan tasks. For the breathing task, there are small-scale differences in the alpha and low-beta bands in the right temporal area. For the body-scan task, there is a small-scale difference in the alpha in the right temporal area, and a frontal power asymmetry (right>left) in the low-beta band.

This study further investigated the momentary state effect of acquired mindfulness skills by examining the EEG activity. To this end, the EEG powers during breathing and body-scan were referenced to that during resting 1. Figure 5 shows the spectral comparisons of breathing, body-scan, and resting 2 in the post-test session. Compared with resting 1, the powers of delta, low-beta, high-beta, and low-gamma bands significantly decreased during body-scan at both Fz (0.2–2.5Hz, ts≥−2.55, ps<0.05; 15–17.5Hz, ts≥−2.92, ps<0.05; 19.5–20.5Hz, ts≥−2.59, ps<0.05; 23.5–24.5Hz, ts≥−2.54, ps<0.05; 28.5–37Hz, ts≥−2.59, ps<0.05) and Pz (0.2–4Hz, ts≥−3.08, ps<0.05; 13–16.5Hz, ts≥−3.00, ps<0.05; 29.5–37.5Hz, ts≥−2.44, ps<0.05). However, no significant power change was found in breathing and resting 2 (ps>0.05). The EEG power during body-scan was also referenced to the power during breathing, and no significant difference was found (ps>0.05). For the waitlist control group that did not engage in any mindfulness practice, no significant differences were found in the EEG power between every task pair (ps>0.05).

Correlation analysis was performed to examine the potential correlations between EEG activities and behavioral outcomes. Table 2 presents the correlation coefficients between the post-test–pre-test differences of EEG power in the four tasks and those in the behavioral measures. For the MBSR group, Δlow-beta power in resting 1 for the Pz site was found to have a negative correlation with ΔDERS (b=−0.50, p=0.040), Δtheta in breathing for the Pz site had a marginally significant negative correlation with ΔFFMQ (b=−0.48, p=0.053), Δlow-beta in body-scan for the Fz site had a marginally significant negative correlation with ΔFFMQ (b=−0.48, p=0.054), Δalpha in resting 2 for both Fz (b=−0.64, p=0.006) and Pz sites (b=−0.51, p=0.036) showed a negative correlation with ΔFFMQ, and Δtheta in resting 2 for the Pz site had a marginally significant negative correlation with ΔFFMQ (b=−0.47, p=0.058).

The correlation coefficients between the changes in the EEG band power and behavioral subscales in the MBSR group are shown in Supplementary Material (Supplementary Tables 2, 3). With regard to FFMQ subscales, alpha and theta spectral changes were significantly correlated with observing, describing during resting 1, and nonreactivity subscales during resting 2 in both Fz and Pz (ps<0.05). In the DERS subscales, delta spectral changes in Fz were significantly associated with goals (p=0.037) during resting 2, whereas low-beta and high-beta spectral changes in Pz were significantly associated with awareness and clarity during resting 1 (ps<0.05), respectively.

Table 1 presents the MBSR intervention effects on the FFMQ and DERS scores, indicating that participants demonstrated augmented levels of trait mindfulness and emotional regulation. The EEG spectral results (Figures 2, 3) showed that the 8-week intervention led to an increase in high-frequency EEG activities across all conditions and a decrease in low-frequency EEG activities during the breathing and body-scan practices. As observed in the MBSR group at the follow-up visit in week 8 (Figure 5), body-scan suppressed EEG power across all frequencies. Additionally, the mindfulness-induced changes in theta, alpha, and low-beta band powers significantly correlated with the changes in FFMQ or DERS scores.

The differences between the post-test and pre-test results in resting 1 indicated the post-intervention effect of the MBSR practice on EEG activity. Previous studies have reported that relaxation exercises induced decreased beta and gamma band power (Stinson and Arthur, 2013) and increased theta band power (Field et al., 2010). The current study’s EEG results and previous research on MBSR (Lutz et al., 2004; Cahn et al., 2010) have demonstrated that, unlike relaxation exercises, mindfulness practices yield increased high-frequency EEG activity. This post-intervention neuro-electrical change was sustained in our study irrespective of which mindfulness task the participants engaged in. In the MBSR group, the post–pre-test EEG power comparisons of resting 1, breathing, body-scan, and resting 2 revealed the same pattern, namely, elevated high-frequency (high-beta and low-gamma) EEG power in the post-test EEG scan. The results further supported our argument that the increase in high-frequency EEG power is a post-intervention effect of mindfulness practice. Although mindfulness practice is known to provide the same relaxation effect as other relaxation exercises (Davidson et al., 2003), this study suggested that mindfulness practice differs from relaxation exercises.

The results of mindfulness practice were rather similar to the EEG findings on beta and gamma neurofeedback training (a training aimed at improving beta or gamma power) that showed that 10days of beta and gamma neurofeedback training led to improved episodic memory among healthy adults (Keizer et al., 2010). A similar association between long-term cognitive training and gamma power elevation was also found among the elderly population (Staufenbiel et al., 2014) and patients with Alzheimer’s disease (Van Deursen et al., 2008), schizophrenia (Molina et al., 2020), and attention-deficit/hyperactivity disorder (Yordanova et al., 2001). Furthermore, the results were consistent with the neuroimaging findings that highly focused participants with 6weeks of mindfulness training showed better cognitive performance in the Stroop test, in addition to exhibiting higher dorsolateral prefrontal cortex (dlPFC) activation (Allen et al., 2012). With results showing elevated high-frequency EEG power, mindfulness practice is likely similar in effect to cognitive training.

After the MBSR intervention, the low-frequency (delta) activity at Fz decreased during breathing and body-scan (Figure 2) and that at Pz decreased during breathing (Figure 3). Previous studies on different mindfulness styles (i.e., Vipassana and Qigong) have suggested that compared with non-meditators, practitioners’ delta activity increased in the prefrontal area during the resting state (Tei et al., 2009; Cahn et al., 2010). However, another study on MBSR reported that practitioners’ frontal delta power decreased (Gao et al., 2016), indicating that different mindfulness practices may lead to different EEG results. Our data supported the results of Gao et al. (2016) that MBSR practitioners’ low-frequency EEG power decreases and high-frequency EEG power increases. The current study further suggested that such EEG power changes could be introduced with 8weeks of training.

Some have argued that the delta oscillation of EEG is related to mental task performance because delta oscillation represents people’s attention to their internal process (Harmony et al., 1996; Harmony, 2013). Knyazev (2007) further suggested that delta and alpha oscillations in the prefrontal area may contribute to a reciprocal inhibitory mechanism that can manipulate people’s motivation and attention and moderate people’s mental task performance. Harmony (2013) suggested that the increased delta oscillation in mindfulness practitioners represents the inhibition of the prefrontal cortex in addition to the reduction of emotional and cognitive engagement. MBSR practitioners, however, are instructed to not inhibit their emotions or cognition but to accept and merely observe their inner thoughts just as they are (Davidson et al., 2003). This practice of acceptance may explain the current result that delta power decreased during mindfulness practice after 8weeks of training. This argument is further legitimized by the result that only the MBSR-trained participants, and not the waitlist control group, demonstrated the delta power drop. This is because attentional yet nonjudgmental acceptance of oneself requires continuous and effort-intensive practices to master.

Additionally, this delta power suppression suggested that the practices might involve a stage of high vigilance (Smallwood and Schooler, 2015). A previous study on mind-wandering (Braboszcz and Delorme, 2011) showed that when people were distracted from a task and started mind-wandering, their delta and theta power increased, whereas their alpha and beta power dropped. The EEG results of mind-wandering were exactly the opposite of our mindfulness results of decreased delta power. The contradictory EEG results support the argument that mindfulness is a process of disciplining the mind and stopping mind-wandering (mindlessness; Davidson et al., 2003; Mrazek et al., 2012). Furthermore, another study suggested that mindfulness programs can reduce mind-wandering episodes (Schooler et al., 2014). This claim is supported by the results that participating in 8weeks of mindfulness training promotes working memory capacity, which is the key to maintaining focus in a cognitively demanding and vigilance-requiring situation (Jha et al., 2010). Even 8min of mindful breathing was found to reduce people’s attentional error in a vigilance task (Mrazek et al., 2012). Therefore, EEG spectral whitening (low-frequency power drop and high-frequency power elevation) represents a mental state with less mind-wandering and higher mindfulness.

Previous neuroimaging studies on mindfulness have suggested frontal and parietal areas are mostly related to mindfulness-induced neuro changes. For example, an fMRI study on long-term mindfulness practitioners reported the increased functional connectivity among the (PCC; Brewer et al., 2011). A similar finding was also reported in the post-MBSR intervention participants (Kral et al., 2019). Meanwhile, the fMRI data collected in this study showed that the activation difference between breathing and body-scan in the pre-test was notable in frontal and parietal spatial distribution (Guu et al., 2020). The current EEG data agreed with the previous fMRI results that the changes in mindfulness-induced neuro-activation were evident in the frontal and parietal areas (Figure 4). The results in Figure 4 further suggested that some other brain areas, including the occipital and temporal areas, also exhibit comparable spectral changes. Also, there was a frontal power asymmetry in the low-beta band in body-scan. We expect that further studies on the frontal EEG power asymmetry will be legitimate in the exploration of mindfulness (also see Isbel et al., 2019).

It is noted that four participants in the MBSR group did not show EEG power difference between the post-test and pre-test, even though the general group power elevation was significant (see Figures 2, 3). The four participants reported higher FFMQ (mean difference=24.25, SD=12.37; group mean difference=27.61, SD=22.02) and half reported lower DERS (mean difference=−5.25, SD=26.54; group mean difference=−13.06, SD=25.87). Although both the FFMQ and DERS levels of the four participants changed in the manner intended by the MBSR intervention, the magnitude of these changes was lower than that of the MBSR group average. A potential explanation for this phenomenon was that the MBSR intervention was not effective in some of the participants, and therefore, no neurophysiological change was introduced. The same phenomenon was also found in a previous mindfulness study on the bereaved individuals (Huang et al., 2021) that some of them had their mindfulness levels decreased after the 8-week mindfulness intervention. The phenomenon suggested that mindfulness training did not guarantee success for everyone. Further studies are required to figure out the reason underneath to improve the efficacy of the mindfulness training courses.

Another focus of this study was to examine whether different mindfulness practices contribute to participants’ neuro-electrical responses differently. Compared with resting 1, the MBSR-trained participants showed significantly decreased EEG activities across delta, low-beta, and parts of high-beta and low-gamma bands at both Fz and Pz when practicing body-scan. During the body-scan practice, practitioners were instructed to be aware of their introspective body sensations. As suggested in a previous study (Mirams et al., 2013), body-scan exercise can improve people’s somatic signal detection clarity. The EEG changes associated with the body-scan task found in this study reflected how the body-scan exercise is specific and concrete for practitioners. By contrast, no significant power difference was found when practicing mindful breathing (Figure 5), suggesting that body-scan leads to more significant EEG change among the participants than breathing.

Notably, the preliminary data collected from five mindfulness practitioners with more than 2years of MBSR experience (i.e., the expert group; Figure 6) further revealed that experts’ EEG power was consistently lower than that of the MBSR practitioners with 8weeks of experience, irrespective of whether they were resting, practicing mindful breathing, or body-scan practice. In general, the high-frequency power of the MBSR practitioners elevated and the low-frequency power decreased after the 8-week MBSR training, but all EEG power decreased along with the expertise of mindfulness. Although the expert group had few participants, a general drop in EEG power across low-, middle-, and high-frequency bands could be a neuro-electrical feature of long-term MBSR practice.

The differences in the theta, alpha, and low-beta band powers between the post-test and pre-test were correlated with the change in the FFMQ or DERS scores (as shown in Table 2). In our major findings concerning behavior and EEG biomarkers, we found significant correlations between behavioral indexes and theta, alpha, and low-beta power. These results were consistent with Ahani et al.’ (2014) finding that improvement in mindfulness levels was characterized by increases in the theta, alpha, and beta activities. Another study on emotional regulation also highlighted that frontal bilateral alpha activity and the parietal delta/beta ratio predicted people’s spontaneous emotion regulation level (Tortella-Feliu et al., 2014). This leads to the following question: Why was a major EEG power difference observed in high- and low-frequency bands in our data despite the clear correlation between EEG power and behavioral indexes observed in the theta, alpha, and beta bands? A possible reason could be that the major differences in neurophysiological changes after 8-weeks of mindfulness practice can be observed through delta, high-beta, and low-gamma spectral power and that the biomarker of mindfulness is found in alpha. A previous study on experienced Zen meditators showed that the hours of meditation practice and weekly frequency were negatively correlated with alpha power but not gamma power (Pasquini et al., 2015). Our results further suggest that mindfulness and its positive outcomes involve a complex mechanism. The EEG signals represent the mixtures of these contributing factors. Future EEG biomarker studies on mindfulness may also include the factor of cognitive ability to clarify such a complex mechanism.

It is noted that the task sequence was suggested by licensed and experienced MBSR instructors. Even though we are well aware that counter-balance is valuable in terms of causal interpretation, we treasure the ecological validity of mindfulness practice. In MBSR training, practitioners start learning mindfulness from breathing, then advance to body-scan afterward. It is unlikely that practitioners can practice body-scan without settling their mind with breathing first. In order to ensure that the experimental conditions were faithful to natural MBSR practice as much as possible, the participants were always asked to practice breathing and then body-scan in both scanning sessions.

There may also be some concern about the relatively small sample size of this study. We admitted that this can be a shortcoming of this study, and the major difficulty of data collection came from the limited MRI scanning sessions we can enrolled. We argued that our results are convincing because the results are still significant with small sample size, and the trends are consistent with our predictions based on the previous works on mindfulness.

Another limitation of this study is the frequency range of EEG data. Due to the contamination of MR gradient-induced artifacts, the current study examined the EEG data only up to 40Hz, leading not to make a fully compatible interpretation between the current results and the other studies involving high-gamma activity. The following studies on this topic may examine mindfulness practitioners’ EEG signals outside of an MRI scanner to address this limitation.

We were aware that there is a psychometric discussion on the diverse nature of mindfulness assessment (Grossman, 2011). Although FFMQ is not the one and only one mindfulness assessment developed, FFMQ is one of the most used and sophisticated mindfulness assessments (Carpenter et al., 2019). Especially, the validity of FFMQ on the assessment of MBSR intervention outcomes has wildly been supported (Lamothe et al., 2016; Vibe et al., 2017). Still, further studies may examine the validity of other mindfulness questionnaires on MSBR outcomes and the correlation differences between the questionnaires and EEG band powers.

The current result is only a glimpse of the neuro-evidence of mindfulness practitioners. Other perspectives and evidence, including EEG connectivity analysis and MRI joint analysis, can be used and may reveal the bigger picture of mindfulness neuro-mechanism in the following studies.