For the past two decades, mindfulness, commonly defined as the adoption of a nonelaborative, nonjudgmental awareness to present-moment experience [1
], has garnered increasing interest for its seemingly innumerable benefits, permeating into the broader social discourse and influencing areas including public health, academia, corporations, and even politics [3
]. Despite mounting caution from various academic disciplines that enthusiasm for mindfulness may be outpacing scientific progress [5
], the accelerating proliferation and public embracement of mindfulness appear relatively uninterrupted. As with any growing scientific discipline, balancing optimism with rigor represents a formidable and persistent challenge.
In considering the specific influence of science, widespread co-option of mindfulness may be driven by the disproportionate number of studies examining and reporting the effects of mindfulness (i.e., what it does) relative to studies aimed at discerning its underlying mechanisms (i.e., how it works). Research aimed at exploring the salutary effects of mindfulness may be derivative of broader sociocultural interests in self-improvement and social flourishing—motivations that may maintain a collective predilection toward “discovering benefits” [8
]. Indeed, this appears reflected in the large and continuously expanding number of clinical, academic, social, and professional interventions from which mindfulness serves as a basis and inspiration.
Critically, the proliferation of mindfulness-based applications arguably impedes mechanistic investigation—namely, that rapid change in the dissemination and implementation of mindfulness erodes its definition and uniquely challenges methodical operationalization [7
]. Furthermore, the ever-expanding number of mindfulness-related “benefits” contributes to the intractability in pinpointing general mechanisms that undergird its purported myriad effects. With that said, one potential way to navigate these challenges is to systematically elucidate how mindfulness influences specific well-studied neurocognitive functions that underlie an array of human behaviors.
One such function is error monitoring (also referred to as performance monitoring), a foundational feature of human cognition that facilitates the ability to continuously detect and adjust to errors [10
]. Importantly, error monitoring is crucial in enabling goal-directed action and promoting behavioral adaptation—core abilities that underlie academic achievement, workplace productivity, mental health, and other outcome variables that are commonly associated with mindfulness. To the extent that the adoption and sustainment of mindfulness constitutes a goal-oriented action [15
], the very act of being mindful itself—whether through intentional application of state mindfulness toward daily activities (see [17
]) or engagement in more formal avenues of mindfulness training such as meditation—is likely to recruit, and possibly modulate, the error monitoring system and its downstream behavioral consequences (e.g., detection of mind wandering and subsequent remedial redirection of attention). Indeed, investigating the nature of the relationship between mindfulness and error monitoring may be promising in understanding the means and extent to which mindfulness exerts its broader influence on contemporary life.
In contrast to the relative nascency of mindfulness research, error monitoring has been studied extensively for over 50 years (e.g., [18
]). Importantly, decades of research in cognitive neuroscience have yielded considerable insights into the putative neural substrates of error monitoring—linking error processing systems to a medial frontal network comprising the anterior cingulate cortex (ACC), lateral prefrontal cortex (PFC), supplemental motor areas (SMA), and insula (see [11
] for reviews). Furthermore, this neural network is consistently implicated in the generation of a systematic sequence of event-related potentials (ERPs) after error commission on speeded-choice tasks (e.g., Eriksen flanker tasks). Two of the most reliable and well-studied neural indices of error monitoring are ERPs: the error-related negativity (ERN; [20
]) and the error positivity (Pe; [15
The ERN is a frontal central negative deflection that occurs within 100 ms after error commission and has been source localized to the ACC and SMA (see [22
] for a review). Although the functional significance of the ERN is still debated, two prominent theories grounded in computational modeling have linked the ERN to early detection of: (1) competing response representations (error vs. correct; i.e., conflict monitoring theory, [14
]) and; (2) mismatch between predicted and actual performance outcomes (i.e., reinforcement learning theory, [24
]). Despite their differences, both theories imply that larger ERN amplitudes are associated with higher acuity in detecting performance-related discrepancies. Additionally, it has been posited that the ERN indexes emotional processing of errors [25
] on the basis of its association to brain regions implicated in pain and negative affect (e.g., ACC, [27
]), psychological disorders characterized by affective dysregulation [28
], and affective physiological responses (e.g., skin conductance and startle response, [29
]). Though the exact role of affect in ERN modulation remains unclear, this line of research raises the possibility that interventions that alter affective processing (e.g., mindfulness meditation) may be liable to modulate the ERN.
Following the ERN, the Pe is a central parietal positive deflection peaking approximately between 200 and 400 ms post response. Evidence for the localization of the Pe is not as conclusive as the ERN, with studies pointing to the rostral ACC, posterior cingulate, and insula [31
]. Similar to the ERN, three major theories on the functional significance of the Pe have been proposed, including: (1) conscious error recognition [34
]; (2) responsivity to the motivational significance of the error [21
] and; (3) affective processing of conscious errors [36
]. Although few studies have pitted these theories against each other, accumulating evidence continues to support the Pe as a neural correlate of conscious error awareness whereas comparatively less evidence has been found in favor of the affective processing hypothesis [12
In addition to the abundance of basic research on the ERN and Pe, a wide literature base has implicated error monitoring in various regulatory and functional domains such as stress regulation [40
], impulse control [42
], attention regulation [44
], and academic performance [45
]. Despite the central role of error monitoring in maintaining healthy functioning, surprisingly few studies have examined the intersection between mindfulness and error monitoring.
In one of the first investigations, Teper and Inzlicht [48
] employed a cross-sectional design comparing the ERN, Pe, and behavioral performance between experienced meditators and novice controls. Interestingly, meditators exhibited larger ERN amplitudes and superior accuracy relative to controls. Replicating these findings, Andreu et al. [49
] reported enhanced ERN amplitudes and higher accuracy in experienced Vipassana meditators compared to novices. However, a more recent study by Bailey and colleagues [50
] utilizing advanced whole-scalp analysis reported no differences in behavioral or ERP indices of error monitoring between experienced meditators and novices. Surprisingly, none of the studies found group differences in Pe amplitude despite the conceptual overlap between the Pe and mindfulness as constructs involving conscious awareness.
Experimental designs have produced even more divergent outcomes. Using a single-session experimental manipulation, Larson and colleagues [51
] found diminished Pe amplitudes but no change in the ERN or behavioral performance after novice non-meditators completed a brief guided mindfulness meditation relative to controls. Contradictorily, a clinical longitudinal study examined the effects of mindfulness-based cognitive therapy (MBCT) on adult ADHD patients, finding that MBCT patients exhibited increased
Pe amplitudes, but no changes in ERN or behavioral performance [52
]. Yet another study compared brief single-session inductions of thought-focused relative to emotion-focused mindfulness practice, reporting increased ERN but no change in Pe in only the emotion-focused group [53
]. The relative sparsity of studies combined with the equivocality of the findings signal a need for further clarification into the nature of the mindfulness–error monitoring relationship.
Toward this end, recent critical reviews of mindfulness research highlight several prescriptive factors that appear prudent to consider [7
]. First, mindfulness is a polylithic construct that can reflect a dispositional trait, state of mind, mental training modality (e.g., meditation), or psychological intervention. Importantly, such construct heterogeneity challenges standardized operationalizations of mindfulness and may partially explain the different outcomes in the studies reviewed above. For instance, Larson et al. [51
] examined mindfulness as a brief guided meditation, whereas Andreu et al. [49
] and Teper and Inzlicht’s [48
] cross-sectional design operationalized mindfulness as a derivative of meditative experience. Consequently, it is likely that the “acute” effects of mindfulness training in novices differ from the oft-posited “trait-like” changes associated with cumulative meditative experience [15
]. Moreover, Schoenberg and colleagues [52
] investigated mindfulness in the context of a psychological intervention for ADHD, introducing interpretive complications arising from uncontrolled components of the intervention (e.g., parsing effects of psychoeducation vs. social support vs. mindfulness training) and idiographic factors unique to an ADHD clinical sample. Lastly, Saunders et al. [53
] bisected their mindfulness induction to exclusively direct awareness toward either emotions or thoughts, thereby narrowing the scope of inquiry to mindfulness of specific internal states. Such differences in operationalization and sample characteristics (e.g., novice vs. experienced vs. clinical) represent unique methodological challenges extending from construct heterogeneity that, without proper contextualization, can obfuscate understanding of how different aspects
of mindfulness influence error monitoring.
Second, there is substantial variation among mindfulness practices. This is perhaps best exemplified by the empirically supported distinction between focused attention (FA) and open monitoring (OM) meditation [65
]—two separate meditative practices that are often unwittingly subsumed under the umbrella term “mindfulness meditation”. FA meditation is conceptualized as the voluntary direction of sustained attentional awareness to a target object (e.g., the breath), whereas OM meditation involves non-judgmental monitoring of momentary experience without explicit direction to attend to a preselected target. Importantly, studies comparing FA and OM meditation have shown unique patterns of neural activation [66
] and different effects on cognitive and affective processes ([68
]; see [70
] for a review). Taken together, evidence supports the possibility that functional differences between OM and FA meditation may extend to the domain of error monitoring.
To date, however, studies of mindfulness and error monitoring have given little consideration for technical variation within mindfulness practice—whether it be in the context of cross-sectional designs involving experienced meditators and novices, brief mindfulness inductions (e.g., one session guided meditation), or multi-week mindfulness training programs. For example, Teper and Inzlicht [48
] included participants from a variety of meditative traditions including Vipassana and broadly defined “concentrative traditions”. Vipassana meditation is often considered an OM meditation [66
], whereas “concentrative” appears to suggest some form of FA meditation. Similar considerations apply to Schoenberg et al. [52
] given that standard protocols for MBCT involve FA- and OM-based practices in addition to experiential exercises that draw from both meditations (see [72
] for a systematic dismantling study; [73
]). Importantly, such mixing of FA and OM techniques impedes the ability to parse the extent to which distinguishing features of each respective practice relate to error monitoring. For example, Andreu et al. [49
] and Larson et al. [51
] appeared to homogenize meditative technique, with recruitment of strictly experienced Vipassana meditators in the former, and the employment of a guided breath-oriented FA meditation in the latter. Interestingly, however, Saunders and colleagues’ [53
] novel induction seemingly mixed properties of both FA and OM meditation, instructing participants to direct awareness toward a specific category
of internal experience (thoughts vs. emotions) rather than a fixed target object (as in FA) or any momentary experience (as in OM). Although their study yielded illuminating insights into the specific influence of mindfulness of emotion on error monitoring, the unique nature of the induction challenges whether the conclusions can generalize to FA or OM meditation, two of the most common and standard forms of mindfulness practice. Reviewing these studies through the purview of the FA/OM dichotomy reveals a distinct gap in the literature—namely, that no prospective study has examined the effects of OM meditation on error monitoring.
In addition to supplementing the literature, there are complementary incentives to an experimental investigation of OM meditation, particularly in a novice non-meditating sample. The points reviewed above represent some of the most pressing challenges in mindfulness research—challenges that may be surmounted through active incorporation of the prescriptive recommendations identified by the field (e.g., [7
]). Extrapolating this to the relatively unexplored topic of mindfulness and error monitoring, prudent first steps may be to: (1) fill clear gaps in the literature; (2) address extant issues associated with construct heterogeneity, meditation experience, and technical variation; (3) begin development of a standardized, replicable, and generalizable methodology through incremental testing and refinement of measures that are sensitive to various operationalizations of mindfulness.
Consonant with these steps, the current study sought to examine the effects of a brief guided OM meditation on neural (i.e., ERN, Pe) and behavioral measures of error monitoring in meditation-naïve participants. Measures of trait mindfulness were collected to account for potential group differences in dispositional mindfulness and explore the extent to which individual differences in trait mindfulness relate to error monitoring. Heeding the recommendations of Van Dam and colleagues [7
], this approach succinctly circumscribes mindfulness training to a brief guided OM meditation (as opposed to FA or broader training modality involving mixed meditative techniques), minimizes confounds associated with meditative experience, standardizes training duration, and leverages natural variability in trait mindfulness to extend analysis across multiple aspects of mindfulness (i.e., meditative practice and dispositional trait).
Given the mixed findings from the studies reviewed above in addition to the absence of research investigating the effects of OM meditation on error monitoring, we established our predictions using the best available evidence. Regarding the ERN, both Andreu et al. [49
] and Teper and Inzlicht’s [48
] sample included experienced OM meditators (e.g., Vipassana practitioners) and reported larger ERN amplitudes relative to novices. Furthermore, Saunders and colleagues [53
] reported increased ERN amplitudes as a function of directing mindfulness toward emotions relative to thoughts, positing a link between affective awareness and ERN modulation. In this light, that Larson et al. [51
] did not observe changes in the ERN may be explained by their employment of a FA as opposed to OM induction. Again, FA meditation involves sustained attentional awareness to a fixed target object and demands redirection of attention away from non-target phenomena—put more directly, breath-oriented FA meditation inherently prioritizes awareness of breath over affective experience. On the other hand, OM meditation emphasizes the fostering of momentary awareness which may include arising emotional states among other forms of internal experience (e.g., physical sensations) [65
]. Consequently, it stands to reason that if mindfulness of emotion is central to ERN modulation as suggested by Saunders and colleagues [53
]—a unique property of OM relative to breath-oriented FA meditation—then assuming sufficient mindful awareness of emotion is cultivated during practice, a brief OM meditation induction was predicted to increase ERN amplitude.
With respect to the Pe, the same rationale undergirded our prediction that Pe amplitude would not change—none of the aforementioned studies involving OM meditators [48
] or unique components of OM meditation [53
] reported change in the Pe. Although Larson et al. [51
] reported a decrease in the Pe, it seemed unreasonable to expect replication given the previous reflections on the differences between FA and OM meditation, in addition to Schoenberg and colleagues’ [52
] inconsistent finding that Pe increased with MBCT training. Lastly, behavioral performance was not expected to differ given the predominance of null findings reported in similar studies employing brief mindfulness inductions on novice samples [51
Secondary exploratory analysis examined the relation between trait mindfulness and error monitoring. Although measurement of trait mindfulness remains a topic of considerable debate [55
], there appears to be consensus that trait mindfulness contains multiple subfacets. Indeed, the Five Facet Mindfulness Questionnaire [78
] is an empirically validated measure that captures five factor-derived facets of trait mindfulness: Observing (FFMQ-O), Describing (FFMQ-D), Acting with Awareness (FFMQ-AA), Nonjudging (FFMQ-NJ), and Nonreactivity (FFMQ-NR). Among these facets, FFMQ-AA measures the propensity to attend to the present moment (e.g., ‘It seems I am “running on automatic” without much awareness of what I’m doing’). Given that on-task attention has been implicated in conceptual models of both the ERN and Pe [14
], FFMQ-AA exhibits strong theoretical relevance to error monitoring and may be related to the ERN and Pe. This possibility is further supported by previously reported relationships between FFMQ-AA and attention-related ERPs [79
]. Lastly, exploration of FFMQ-NR and FFMQ-NJ seemed to be a natural follow-up on past suggestions implicating nonjudgment in ERN modulation (i.e., increased nonjudgmental awareness of affective error salience; see [48
]), and nonreactivity in Pe modulation (i.e., reduced error orientation; see [51
3.1. Baseline Mindfulness and Manipulation Check
Descriptive statistics of all measures by group are presented in Table 1
. As expected, there were no group differences in any facet of trait mindfulness, or in overall mindfulness (t
s < |1.49|, p
s > 0.14).
Participant responses on the manipulation check revealed group differences in interest (t
(1, 204) = 4.32, p
< 0.01), learning (t
(1, 204) = 6.02, p
< 0.01), and sleepiness (t
(1, 204) = −2.76, p
< 0.01), such that relative to the meditation group, participants in the control group rated the control audio as more interesting (control: M
= 4.54, SD
= 1.62, meditation: M
= 3.55, SD
= 1.67), indicated learning more (control: M
= 4.68, SD
= 1.36, meditation: M
= 3.51, SD
= 1.42), and endorsed less sleepiness (control: M
= 3.81, SD
= 1.40, meditation: M
= 4.36, SD
= 1.48). Importantly, there were no differences in engagement, arousal, emotional reactivity, or understanding (t
s < |1.92|, p
s > 0.06), suggesting that although groups differed in their experiential appraisal of the audio inductions, participants nonetheless approached the task with equal levels of engagement and comprehension. Notably, with the exception of sleepiness which was not previously measured, this constellation of group differences fully replicated Lin et al. [80
]. To determine whether the unexpected group difference in self-reported interest, learning, and sleepiness confounded the results of the study, all analyses were re-run with the three variables entered as continuous covariates. All output remained the same with respect to statistical significance and effect size. Therefore, the results are henceforth presented in accordance to what was originally described in the methods.
3.2. Behavioral Data
Descriptive statistics for behavioral and ERP data are presented in Table 2
. Overall flanker task accuracy was relatively high (M
percent correct = 82.87%, SD
= 8.73%). Participants made an average of 80.15 errors (SD
= 41.38), with more errors on incongruent trials (M
= 56.26, SD
= 29.97) than congruent trials (M
= 23.89, SD
= 19.18, t
(205) = 16.22, p
< 0.01). Importantly, there were no group differences in overall errors or errors by trial congruency (t
s < 0.90, p
s > 0.37).
The analysis of RTs revealed main effects of Response Type and Congruency, such that RTs on error trials (M = 331.15, SD = 48.22) and congruent trials (M = 379.47, SD = 44.56) were faster than on correct (M = 410.55, SD = 46.52, F(1, 203) = 1236.88, p < 0.01, = 0.86) and incongruent trials (M = 418.95, SD = 52.16, F(1, 203) = 529.00 p < 0.01, = 0.72), respectively—consistent with known speed-response type and speed-congruency trade-offs. These main effects were qualified by a significant Response Type X Congruency interaction (F(1, 203) = 107.39, p < 0.01, = 0.35), such that RT differences between incongruent and congruent trials were larger on correct trials (M = 54.29, SD = 25.12) relative to error trials (M = 23.24, SD = 38.23, t(204) = 10.31, p < 0.01). Notably, there were no significant interactions involving Group (Fs < 2.97, ps > 0.09), indicating that there were no group differences in RTs.
In keeping with the typical post-error slowing (PES) effect, analyses revealed faster RTs following correct responses (M = 394.85, SD = 47.06) than following errors (M = 421.04, SD = 62.16, F(1, 204) = 134.99, p < 0.01, = 0.40). Critically, there was no Response Type X Group interaction (F(1, 204) < 0.01, p = 0.99, < 0.01), indicating that PES did not differ by group. The analysis of post-error accuracy (PEA) revealed a main effect of Response Type, such that accuracy following correct responses (M = 84.54%, SD = 7.07) was slightly higher than accuracy following errors (M = 82.82%, SD = 13.75, F(1, 204) = 4.80, p = 0.03, = 0.02). Again, there was no Response Type X Group interaction (F(1, 204) = 0.96, p = 0.33, < 0.01), suggesting no group differences in PEA.
The analyses involving ERN amplitude revealed an expected main effect of Response Type (F(1, 204) = 416.50, p < 0.01, = 0.67), reflecting larger negativity on error trials (M = –5.16, SD = 3.59) relative to correct trials (M = 0.01, SD = 2.72). There was, however, no significant Response Type X Group interaction (F(1, 204) = 0.06, p = 0.81, < 0.01), indicating that ERN amplitude did not differ by group.
Similarly, the main effect of Response Type on Pe amplitude was significant (F(1, 204) = 574.74, p < 0.01, = 0.74), revealing increased positivity on error trials (M = 4.00, SD = 4.14) relative to correct trials (M = –3.43, SD = 2.85). Critically, there was a significant Response Type X Group interaction (F(1, 204) = 4.62, p = 0.03, = 0.02), such that the Pe was larger in the meditation group (M = 8.10, SD = 4.19) relative to controls (M = 6.77, SD = 4.70; t(204) = 2.15, p = 0.03). For full transparency, the magnitude of this interaction was reduced after re-running the model with interest, learning, and sleepiness as continuous covariates (F(1, 204) = 4.50, p = 0.04, = 0.02). However, the effect size and associated interpretive significance remained unchanged.
3.4. Relationships between ERPs, Behavioral Performance, and Trait Mindfulness
Given that the meditation group exhibited larger Pe amplitude, relationships between ERPs and behavioral performance measures were examined across groups. Correlations among the ERN, Pe, error rate, error RT, correct RT, PES, and PEA separated by group are presented in Table 3
. For both groups, larger (more negative) ERN amplitudes were associated with fewer errors (controls: r
= 0.31, p
< 0.01; meditation: r
= 0.26, p
< 0.01), faster RTs on error (controls: r
= 0.33, p
< 0.01; meditation: r
= 0.30, p
< 0.01) and correct trials (controls: r
= 0.23, p
= 0.02; meditation: r
= 0.31, p
< 0.01), greater PEA (controls: r
= −0.36, p
< 0.01; meditation: r
= −0.31, p
< 0.01), but was unrelated to PES (controls: r
= 0.03, p
= 0.75; meditation: r
= −0.02, p
= 0.86). Similarly, larger Pe amplitudes were associated with fewer errors (controls: r
= −0.42, p
< 0.01; meditation: r
= −0.37, p
< 0.01), faster error RT (controls: r
= −0.26, p
< 0.01; meditation: r
= −0.29, p
< 0.01), greater PEA (controls: r
= 0.41, p
< 0.01; meditation: r
= 0.40, p
< 0.01), but were unrelated to correct RT or PES (controls: rs
< 0.12, ps
> 0.24; meditation: rs
< |0.1|, ps
> 0.33). Notably, all listed correlations between ERPs and behavioral measures did not differ by group (z
s < |0.61|, p
s > 0.52).
In keeping with the secondary exploratory analysis, ERPs were examined in relation to the five facets of trait mindfulness as a function of group. Relationships among the Pe, ERN, and FFMQ are presented in Table 4
. Surprisingly, across both groups, none of the FFMQ subfacets related to the ERN (controls: rs < |0.05|, ps > 0.60; meditation: rs < 0.15, ps > 0.13) or Pe (controls: rs < |0.11|, ps > 0.26; meditation: rs < 0.14, ps > 0.16).