Twenty-five long-term mindfulness meditators (15 females; mean age 40.6 years, SD = 8.64) were recruited through mailing lists and flyers distributed at various meditation centers. Inclusion criteria were at least 5 years of meditation practice, with a practice frequency of at least three times per week during the last 3 months. On average, meditators had 13.1 years (SD = 5.9) of meditation experience and meditated 247.8 min (SD = 104.9) per week. Twenty-five healthy matched controls of the same gender and age (15 females; mean age 40.4 years, SD = 8.80) were recruited through advertisements in the University Medical Center of Freiburg. Participants in the control group had no prior experience of contemplative practice including meditation, Tai Chi, Qi Gong and Yoga. The maximum age was set at 50 years for all participants. Further exclusion criteria were a history of psychiatric conditions, neurological diseases and visual impairment that cannot be corrected by means of visual aid.

One participant of each group who had a high ratio of incorrect responses (greater than three standard deviations from the group mean) and three participants (two meditators and one control), who had high EEG artifacts (see below), were excluded from the analysis. Thus, we compared 22 long-term mindfulness meditators (see Table 1 for the characteristics of the meditation experience) with 23 healthy matched controls. The achieved power (1-β) of the present matched-pairs design (total sample size = 45) is 0.75 when the effect size and α level are 0.4 and 0.05 (two-tailed), respectively.

Most of the participants reported that they were right-handed except for three meditators (two left-handed and one mixed-handed) and four controls (one left-handed and three mixed-handed). Three controls and one meditator did not report their handedness. This study was approved by the ethics committee of the University Medical Center of Freiburg. All participants gave written informed consent in accordance with the Declaration of Helsinki. As listed in Table 1 the long-term meditators were engaged in a range of different meditation traditions. It is crucial to take such differences into account in studies on the effects of specific meditation practices. However, as this study is concerned with generic mechanisms of cognitive control, which are thought to be a basic principle of all types of meditation practices that involve a mindfulness component (Malinowski, 2013), it was deemed appropriate to consider them together. Therefore, any effects found in this study should be considered to be relatively generic.

Participants performed the flanker-type ANT (Figure 1) with concurrent EEG recording. The ANT has been used to examine the efficiency of alerting, orienting and conflict functions independently from each other within a computerized single task that combines a cued detection and flanker-type paradigm. Response conflict was introduced by surrounding the left- or right-pointing target arrow with either congruent or incongruent flankers. Cues presented prior to the appearance of the target arrow provided modulations of alerting and orienting functions (see below).

Participants held a two-button computer mouse with both hands and each thumb was placed on one mouse button. A fixation cross at the center of the screen was always displayed. After a random period varying between 400 ms and 1600 ms, a cue (an asterisk) appeared for 100 ms in one of the following positions: above or below the fixation cross (spatial-cue), in the center (center-cue), or not at all (no-cue). Five-hundred milliseconds after the cue stimulus onset, a target arrow appeared for a maximum duration of 1700 ms either above or below the fixation cross. The target arrow was horizontally surrounded by two flanker arrows on each side, which pointed either in the same direction (←←←←← or →→→→→, congruent target) or in the opposite direction (→→←→→ or ←←→←←, incongruent target) as the central arrow. Participants were asked to press either their left or right thumb as fast and as accurately as possible depending on the direction of the central target arrow. The duration of each trial was 4000 ms. After a 24-trials practice, participants had to perform three blocks of trials. Each block comprised 96 pseudo-randomized trials that consisted of 48 congruent and 48 incongruent trials, one third of which were either no-cue, center-cue, or spatial-cue trials (32 each). Since the present study focused on response conflict effects, the trials were only analyzed according to congruency, which allow the estimation of an individual conflict effect irrespective of the particular cueing condition (Fan et al., 2002). Behavioral data and ERP analyses in respect of cueing conditions can be found in Jo et al. (2016). In brief, whereas cueing conditions revealed no group effect, compared to the control group, mindfulness meditators showed fewer error responses and exhibited higher parietal P3 amplitude during incongruent trials irrespective of response timing. The spectral data reported here have not been included in the previous study. Participants are not exactly the same as in the previous study due to exclusions based on EEG artifacts, but this difference is minor and does not affect the results.

EEGs were recorded with a 64-channel DC-EEG amplifier using active electrodes (Brain Products, Germany) without applying low cut-off and notch filters. An initial reference was placed at FPz and sampling rate was set at 500 Hz. One channel EOG was recorded to monitor ocular movements. Using Matlab (Mathwork, Inc., Natick, MA, USA) and EEGLAB toolbox ver. 13 (Delorme and Makeig, 2004), EEG data were first down-sampled offline to 250 Hz and high-pass filtered at 0.1 Hz. After segmentation into epochs ranging from −2000 ms to 2000 ms from target stimulus onset, ocular artifact correction was performed using independent component analysis (ICA) with visual inspection of EOG recording. Trials containing EMG and other artifacts were manually removed. For further analyses only correct response trials with a maximum RT of 1000 ms were selected. On average, 95.6% trials (meditators, 95.8%; controls, 95.4%; p = 0.818) and 92.4% trials (meditators, 93.3%; controls, 92.4%; p = 0.375) were analyzed for congruent and incongruent conditions, respectively.

All EEG segments were converted to current source density (CSD) using the method described by Kayser and Tenke (2006), which is based on the spherical spline surface Laplacian algorithm (Perrin et al., 1989). CSD transformation highlights the local electrical activities and diminishes the volume conduction effects. Although CSD is considered an estimate of electrical activity in neuronal populations, it should not be generalized to source-based activity. Time-frequency analysis from 2 Hz to 45 Hz were then computed using a Morlet wavelet implemented in EEGLAB with the number of cycles increasing linearly with frequency, from 2 cycles at the lowest frequency to 22.5 at the highest. Time-frequency decomposition was performed for target stimulus-locked and response-locked epochs. For the latter, the EEG segment was re-sorted to be time-locked to the response onset. Time-frequency decomposition produced instantaneous complex number z(f,t), where f is the frequency and t is the time point at 125 Hz. From the resulting complex number, we estimated power as p(f,t) = real[z(f,t)]² + imag[z(f,t)]² for each frequency. In order to minimize sensitivity to noisy trials, each trial was normalized by dividing each frequency by the mean spectral power of the frequency over the full length of the segment for each participant (Grandchamp and Delorme, 2011). After computing the trial average for each condition and each frequency, baseline correction was applied by dividing the mean spectral power for the period ranging from −300 ms to −100 ms before the cue stimulus onset (i.e., from −800 ms to −600 ms before the target stimulus onset). This baseline correction was applied for both target stimulus-locked and response-locked epochs. The outcomes were then converted to a decibel scale (10 × log10 [power/baseline]).

Instantaneous phase angle was defined as φ(f,t) = angle[z(f,t)] and used to compute inter-channel phase synchronization (ICPS) between two channels. ICPS is defined as N−1|∑eiθ (f,t)|, where N is the number of trials and θ is the phase angle difference between two given electrodes, φ₁(f,t) − φ₂(f,t). ICPS ranges from 0 to 1 and values close to 1 indicate higher phase synchrony over trials between two electrodes.

Conflict modulation of MFC theta power was inspected by comparing time-frequency power plots of incongruent and congruent trials for pooled data from both groups (see Figure 3). This comparison showed the strongest theta-band (4–8 Hz) activity at electrode FCz, with a peak activity around 500 ms after target stimulus onset of stimulus-locked epochs and around −200 ms before response onset of response-locked epochs. The average power between 400 ms and 600 ms of the stimulus-locked epochs and between −250 ms and −150 ms of the response-locked epochs were subjected to statistical analyses.

To investigate ICPSs in theta oscillations, we selected the same electrode FCz as the seed electrode. Then, regions of interest were selected based on FCz-seed ICPS topographical maps of the average of all conditions (across groups; Figure 4) and guided by the respective literature (Cavanagh et al., 2009; van de Vijver et al., 2011; Nigbur et al., 2012). We chose electrodes F5/6 to cover the DLPFC, electrodes CP3/4 for the motor cortex, and electrode P2 for the right parietal cortex. Since congruent and incongruent trials showed different distributions of RTs (see left panel in Figure 2) and ICPS activities were more closely aligned to the response onset than target stimulus onset (compare the peak activities in Figure 5), ICPS analyses were performed on response-locked epochs. The averaged ICPS activity within the same time-window as defined above (from −250 ms to −150 ms before the response onset) and a time-window showing the peak activity (from −50 ms to 50 ms around the response onset) were subjected to statistical analyses. In addition, inspection of FCz-seed ICPS topography maps led to analysis of stimulus-related effect on ICPS between the MFC and the parieto-occipital cortex (Figure 4A). Therefore, we selected further electrodes (P5/6, P7/8 and PO7/8 (PO)) to cover this brain area.

Effects of conflict modulation and group on RT, error rate (ER) and MFC theta power were separately tested using a 2 × 2 repeated measures analysis of variance (ANOVA) with congruency (congruent, incongruent) as a within-subject factor and group (controls, meditators) as a between-subject factor. For the analyses of MFC-centered theta network, first, the effect on ICPSs of selected electrodes was tested using a 3 × 2 × 2 repeated measures ANOVA with electrode (FCz–F5/6, FCz–CP3/4, FCz–P2) and congruency (congruent, incongruent) as within-subject factors and group (controls, meditators) as a between-subject factor. Second, to further specify the functional connections, each pair of electrodes was separately subjected to a 2 × 2 repeated measures ANOVA with congruency (congruent, incongruent) as a within-subject factor and group (controls, meditators) as a between-subject factor. These follow up ANOVAs are confirmative analyses of the effects found in the main ANOVA to highlight which connection shows the most prominent effect, rather than determining significant effects per se.

Participants responded faster and made fewer errors during congruent trials (RT: mean = 516.798 ms, SE = 7.598; ER: mean = 1.045%, SE = 0.203) than incongruent trials (RT: mean = 588.300 ms, SE = 10.241; ER: mean = 4.322%, SE = 0.598) as reflected by significant main effects of congruency for both RT (F_(1,43) = 333.41, p < 0.001, η² = 0.886) and ER (F_(1,43) = 41.87, p < 0.001, η² = 0.493). This conflict effect showed no differences between groups in terms of RTs (the left panel in Figure 2), resulting in a non-significant group effect (F_(1,43) = 0.077, p = 0.783, η² = 0.002) and a non-significant congruency × group interaction (F_(1,43) = 1.535, p = 0.222, η² = 0.034). In contrast to RTs, fewer ERs were found among meditators (the right panel in Figure 2) as reflected by a significant main effect of group (F_(1,43) = 4.416, p = 0.041, η² = 0.093) and congruency × group interaction (F_(1,43) = 5.263, p = 0.027, η² = 0.109). Post hoc comparisons showed that the group difference is mainly due to incongruent trials (controls: mean ER = 5.676%, SE = 1.061; meditators: mean = 2.417%, SE = 0.515) than congruent trials (controls: mean ER = 1.238%, SE = 0.315; meditators: mean = 0.852%, SE = 0.254). Therefore, these results indicate that meditators responded more accurately than controls irrespective of RTs, especially after conflicts (see also Figure 3 in Jo et al., 2016).

Time-frequency power plots show increased theta power (4–8 Hz) during incongruent compared to congruent trials within 400–600 ms after target stimulus onset (stimulus-locked epochs) and between −250 ms and −150 ms before response onset (response-locked epochs; see Figures 3A,C). Spatial specificity of this theta power, assessed by topographical maps shows the strongest theta power over the electrode FCz (Figure 3B). As expected, the ANOVAs on FCz theta power revealed enhanced power in incongruent trials compared to congruent ones in stimulus-locked epochs (congruent: mean = 1.763 dB, SE = 0.203; incongruent: mean = 2.185 dB, SE = 0.218; F_(1,43) = 14.192, p < 0.001, η² = 0.248) and in response-locked epochs (congruent: mean = 1.098 dB, SE = 0.173; incongruent: mean = 1.757 dB, SE = 0.188; F_(1,43) = 56.841, p < 0.001, η² = 0.569). Although meditators showed higher power in congruent (mean difference between groups in stimulus-locked epochs = 0.276 dB, response-locked epochs = 0.227 dB) and incongruent trials (stimulus-locked epochs = 0.199 dB, response-locked epochs = 0.056 dB), we found no other significant main effects or interactions indicating group difference for both stimulus- and response-locked epochs (all p ≥ 0.352, all η² ≤ 0.020).

The electrode × congruency × group ANOVA on ICPS before response onset (average between −250 ms and −150 ms of response-locked epochs) revealed a significant main effect of congruency (F_(1,43) = 10.639, p = 0.002, η² = 0.198), showing an overall enhanced synchrony during incongruent (mean = 0.158, SE = 0.008) compared to congruent trials (mean = 0.144, SE = 0.006). Furthermore, we found a significant interaction of congruency × group (F_(1,43) = 5.049, p = 0.030, η² = 0.105). No other main effects or interactions were significant (all p ≥ 0.185, all η² ≤ 0.041). Additionally, lateralization effect was also examined by comparing ICPSs between FCz–F5 and FCz–F6 and between FCz–CP3 and FCz–CP4 for each congruent and incongruent condition and the results showed no significant effects (two-tailed paired t-test, all p ≥ 0.492).

To further specify the significant effects, follow-up congruency × group ANOVAs were performed on each pair of electrodes, i.e., FCz–F5/6, FCz–CP3/4, FCz–P2. A significant main effect of congruency was observed on FCz–CP3/4 (F_(1,43) = 10.263, p = 0.003, η² = 0.193) and FCz–P2 (F_(1,43) = 5.199, p = 0.028, η² = 0.108; Figure 5), while FCz–F5/6 did not reach significance (F_(1,43) = 2.244, p = 0.141, η² = 0.050). The congruency × group effect was observed on FCz–CP3/4 (F_(1,43) = 5.230, p = 0.027, η² = 0.108), indicating that ICPS is comparable between groups during congruent trials (controls: mean = 0.147, SE = 0.10; meditators: mean = 0.143, SE = 0.010) but meditators exhibited considerable enhanced synchrony compared to controls during incongruent trials (controls: mean = 0.152, SE = 0.012; meditators: mean = 0.174, SE = 0.013). No other main effects or interactions were significant (all p ≥ 0.102, all η² ≤ 0.061). These results indicate conflict modulations by FCz–CP3/4 and FCz-P2 synchronies. Furthermore, meditators exhibited an increased FCz–CP3/4 synchrony after conflicts, as compared to controls.

The same analysis was applied for ICPS around response onset (average between −50 ms and 50 ms of response-locked epochs). The electrode × congruency × group ANOVA revealed a significant main effect of electrode (F_(2,86) = 14.940, p < 0.001, η² = 0.258), indicating the highest synchrony between FCz and CP3/4 (mean = 0.304, SE = 0.023) followed by FCz–P2 (mean = 0.231, SE = 0.020) and FCz–F5/6 (mean = 0.205, SE = 0.018). No other main effects or interactions were significant (all p ≥ 0.129, all η² ≤ 0.046).

To test whether the strongest ICPS over the motor cortex (FCz–CP3/4) is dependent on the responding hand, we further conducted a repeated measures ANOVA with congruency (congruent, incongruent) and lateralization (contralateral, ipsilateral) as within-subject factors and group (controls, meditators) as a between-subject factor. It revealed a significant main effect of lateralization (F_(1,43) = 21.470, p < 0.001, η² = 0.333), indicating that ICPS between FCz and CP3/4 is enhanced on contralateral (mean = 0.323, SE = 0.025) compared to ipsilateral areas (mean = 0.281 SE = 0.022; Figure 6). No other main effects or interactions were significant (all p ≥ 0.386, all η² ≤ 0.018). These results indicate that FCz–CP3/4 synchrony around response onset is specific to the responding hand.

Lastly, we quantified the stimulus-related effect on ICPS between FCz and PO. In contrast to the other theta phase synchronizations, FCz–PO is more closely aligned to the stimulus onset than to the response onset (Figure 7), which shows maximal activity around 180 ms after the onset of the target stimulus in stimulus-locked epochs and around 80 ms after response in response-locked epochs. The congruency × group ANOVA around these peak activities (average within 130–230 ms and 30–130 ms of stimulus- and response-locked epochs, respectively), revealed no significant main effects or interactions (all p ≥ 0.094, all η² ≤ 0.064), indicating the absence of conflict effects or group differences.

Having found increased ICPS between FCz and CP3/4 and fewer ERs in meditators, we further examined whether the increased theta phase synchrony over the motor cortex was predictive of low ER. Indeed, correlation analysis revealed that, in comparison to congruent trials, individuals with higher increase of ICPS between FCz and CP3/4 after conflict generally showed lower ER (Spearman’s coefficient r = –0.271; p = 0.036, one-tailed; Figure 8).