Mindfulness-meditation is a practice of non-judgemental awareness centred on the present moment, with the goal of harnessing a sense of openness and acceptance in attitude toward one’s experiences (Kabat-Zinn, 2013). Regular mindfulness practice has been associated with both neurological and physiological changes. For example, neuroimaging studies of meditators have revealed altered brain structure and activation patterns in regions associated with exteroceptive and interoceptive awareness, emotion, memory, and attention (Fox et al., 2014, 2016). Meditation-related changes have also been observed in heart rate variability (HRV), and these HRV changes have been associated with improvements in attention (Burg et al., 2012), pain perception (Adler-Neal et al., 2020), and emotion regulation (Mankus et al., 2013). Although evidence has pointed to the effects of mindfulness on the brain and heart independently, less is known about the impact of mindfulness on the relationship between the brain and heart. The nervous and cardiovascular systems are inextricably linked in a synergistic and interdependent relationship (Ardell et al., 2016; Pereira et al., 2013). Modulation of the brain-heart interaction can improve cardiovascular functioning, reduce stress, and enhance pain regulation (Alshami, 2019; Samuels, 2007; Silvani et al., 2016). Thus, a greater understanding of the influence mindfulness has on brain-heart interaction could offer insights into the unique and multidimensional health benefits of mindfulness-meditation.
One method to investigate the effects of mindfulness on the brain-heart relationship is to examine the association between electrophysiological signals from the brain using electroencephalography (EEG) and from the heart using electrocardiography (ECG). Analysis of EEG and ECG signals often involves an approach focused on only a single measure, whereby unique features from each signal that are known markers of neurological and/or physiological functioning are analysed in isolation (Biel et al., 2001; Subha et al., 2010). However, a range of perspectives have proposed several possible mechanisms of action related to mindfulness practice, and some of these theoretical frameworks suggest the potential for distinct effects of mindfulness on the relationship between the brain and heart. Although analysis of either ECG or EEG modalities independently can extract specific measures of interest that are associated with potential mechanisms of change related to mindfulness practice, the single analysis approach only offers a limited understanding of underlying mechanisms. Moving beyond this conventional approach, the present study sought to investigate the brain-heart relationship through the lens of three unique conceptual frameworks using three different analysis methods, all of which examined the relationship between the ECG and EEG signals. This approach allows for a comprehensive understanding of the impact mindfulness has on the brain-heart relationship and mitigates the constraints of a single methodological framework. In the following sections, we briefly introduce the three conceptual frameworks that we examine in this study.
1.1 The Bayesian Brain: Heartbeat Evoked Potentials and Interoceptive Precision
The ‘Bayesian brain’ hypothesis proposes that the brain infers probabilistic beliefs about the world in accordance with Bayes’ theorem (Doya, 2011; Knill & Pouget, 2004). In short, Bayes’ theorem prescribes a method of probabilistic reasoning that specifies how much one’s beliefs should change based on new information (Puga et al., 2015). Bayesian brain theories thus suggest that the brain functions by integrating a priori knowledge (prior beliefs) with sensory evidence in a way that mimics or approximates Bayesian inference. From this perspective, subjective perception is not necessarily a reflection of the world as it is, but the ‘best guess’ of a predictive model based on the functional integration of incoming stimuli, past experience, and contextual evidence (Friston, 2012; Ongaro & Kaptchuk, 2019). In the Bayesian view, the brain generates hypotheses about the world and our place in it, and these hypotheses – based on beliefs formed through prior experiences – are met by sensory inputs which confirm or disconfirm predicted inputs (Manjaly & Iglesias, 2020). If the hypothesis does not align with the incoming stimuli, a prediction error occurs, creating an opportunity for the brain to update its beliefs by integrating new information (Friston, 2012; Kwee, 1995; Manjaly & Iglesias, 2020; Ongaro & Kaptchuk, 2019; O’Reilly et al., 2012).
However, prediction errors do not guarantee belief updates. In some situations, existing beliefs can be prioritised and outweigh the importance of sensory signals, thereby reducing the degree of belief updating prompted by prediction errors (Ongaro & Kaptchuk, 2019). For example, chronic conditions involving pain can lead to hypersensitivity, increased anxiety, threat detection, and catastrophic beliefs about pain (Latremoliere & Woolf, 2009; Linton & Shaw, 2011; Vlaeyen & Linton, 2000). Within the Bayesian brain framework, these factors are thought to undermine the reliability of sensory input, placing more weight (i.e., greater precision) on existing beliefs and contextual cues and thus diminishing the extent of belief updating in response to sensory signals (Ongaro & Kaptchuk, 2019). Non-judgmental awareness – a principle of mindfulness-meditation that involves directing attention to sensations while holding a neutral attitude (Bishop et al., 2004) - may influence the magnitude of belief updates following a prediction error. Non-judgemental awareness has been proposed to lessen the influence of existing beliefs while promoting the salience of sensory signals, thus increasing the amount of belief updating that occurs in response to a prediction error (Manjaly & Iglesias, 2020).
One method to explore the influence of mindfulness practice on the brain using the Bayesian brain perspective was proposed by Manjaly and Iglesias (2020), who suggested that heartbeat evoked potentials (HEP) could be utilised to examine the effects of meditation on precision-weighted prediction error and belief updates. The HEP is an event-related cortical response synchronised to the heartbeat that occurs between 200ms to 600ms after the R-peak (highest amplitude of the R wave in the QRS complex; Raj et al., 2018). Greater HEP amplitudes have been associated with increased interoceptive accuracy (Coll et al,.2021; Mai et al., 2018). For example, in two separate studies (Montoya et al., 1993; Petzschner et al., 2019), HEP amplitudes were measured while participants directed attention toward an external stimulus compared to an internal stimulus (heartbeat). Both studies reported greater HEP amplitudes when attention was directed internally compared to externally. Based on these results, the authors reasoned that HEP amplitude provides a quantitative measure of neural sensitivity to interoceptive feedback (Petzschner et al., 2019). Similarly, if mindfulness shifts the balance of precision towards sensory signals (and away from prior expectations), one would expect stronger HEPs on average as a consequence of the increased weighting of interoceptive inputs. Further, as mindfulness-related improvements in interoceptive awareness have been shown outside of active mindfulness practice (Mehling et al., 2018) enhanced interoceptive accuracy as reflected by stronger HEPs are thus likely to also be apparent while meditators are at rest, indicating enhanced interoception outside of active meditation practice periods as a result of prolonged interoceptive training.
1.2 Parasympathetic Regulation: Relationship between Frontal-Midline Theta and Heart Rate Variability
The parasympathetic nervous system is a subdivision of the autonomic nervous system (ANS) that regulates bodily functions (e.g., heart rate; Mankus et al., 2013; Wu & Lo, 2008) and is associated with top-down self-regulation (Silvani et al., 2016). Mindfulness practice has been shown to modulate brain and heart correlates of parasympathetic functioning (Jinich-Diamant et al., 2020; Mankus et al., 2013). For example, frontal midline theta activity (fm-theta), a neural oscillation within the 4Hz to 8Hz range detected predominantly over the medial prefrontal areas, has been observed to increase during meditation compared to rest (Baijal & Srinivasan, 2010; Brandmeyer & Delorme, 2018; Clayton et al., 2015). Fm-theta activity is thought to be associated with sustained and internalised attention and has been found to be increased by long-term meditation practice (Lee et al., 2018; Mitchell et al., 2008). In a study investigating the relationship between meditation, fm-theta, and cardiac dynamics, fm-theta activity during the mindfulness meditative state was anti-correlated with sympathetic activity, suggesting that participants were perhaps less distracted and more relaxed (Kubota et al., 2001). Fm-theta is also generated by the anterior cingulate cortex (ACC), which is linked with emotion and cognitive control processes, as well as autonomic nervous system regulation, providing further evidence that fm-theta is associated with parasympathetic function (Matthews et al., 2004).
Meditation practice has also been shown to enhance HRV metrics of vagally-mediated parasympathetic activity, such as the root mean square of successive differences (RMSSD; Joo et al., 2010; Kirk & Axelsen, 2020) and high frequency heart rate variability (HF-HRV; Mankus et al., 2013; Nagendra & Sasidharan, 2017; Wu & Lo, 2008). RMSSD is a measure of heart beat activity in the time-domain, while HF-HRV reflects activity in the frequency-domain; both measures are highly correlated, reflect parasympathetic reactivity, and both measures can be modulated by changes in breathing patterns (Minarini, 2020; Thomas et al., 2019).
While extensive research has been conducted on the effects of mindfulness on brain and heart metrics of parasympathetic functioning independently, few studies have investigated the relationship between fm-theta and HRV. One such study by Tang et al. (2009), investigated the relationship between the percentage of change between fm-theta and nuHF - a normalised form of HF-HRV that compares the ratio of low and high frequency HRV (Burr, 2007) - before and after meditation training. After five days of short-term meditation training, novice meditators showed a correlation between the percentage change in fm-theta power and nuHF, while no correlation was found in the control group. Tang et al. (2009) argued that these results indicated greater interaction and coupling between the autonomic and central nervous systems following meditation training, suggesting that mindfulness practice may improve self-regulation by enhancing ACC control of parasympathetic activity. However, Tang et al. (2009) measured fm-theta without accounting for the contribution of non-oscillatory EEG activity (commonly referred to as 1/f activity because the distribution of power values across different frequencies shows a 1/f slope). Recent research has demonstrated that 1/f non-oscillatory activity can contribute more power to power-frequency measures than oscillatory activity and thus may confound analyses of oscillatory activity (Donoghue et al., 2020). Hence, unless the 1/f non-oscillatory confound is removed prior to analysis, conclusions might not relate to a relationship between fm-theta and HRV but could instead reflect a relationship between 1/f non-oscillatory activity and HRV. Furthermore, HF-HRV can be more susceptible to differences in respiratory rates than RMSSD, and RMSSD is a more accurate marker of vagal activity with changes in natural breathing patterns (Penttilä et al., 2001; Schmid & Thomas, 2021). Changes in respiratory rates have been associated with long term meditation (Peressutti et al., 2012; Steinhubl et al., 2015) and may confound the observed effects of mindfulness on HF-HRV. Thus, further investigation of the relationship between fm-theta activity (after the subtraction of 1/f non-oscillatory activity) and RMSSD may expand the existing understanding of mindfulness-related practice effects on parasympathetic functioning – a reflection of the coupling between brain and heart.
1.3 Wavelet Entropy and the Brain-Heart Synchronisation Measured via Signal Complexity
In biomedical research, power-frequency spectrum analyses are typically used to examine bodily signals such as EEG and ECG activity (Li et al., 2019; Lomas et al., 2015; Shaffer & Ginsberg, 2017). Spectral analyses are linear models based in the time-frequency domain which can be used to isolate specific frequency features of biomedical signals (such as EEG waveforms) (Gao et al., 2016; Rosso et al., 2001). However, biomedical signals are complex, with non-linear properties, and while linear models provide some insights to neural or bodily functioning, they are unable to capture non-linear features (Bachmann et al., 2018; He et al., 2014). To better understand the nonlinear properties of biomedical signals, nonlinear methods such as entropy analyses have been adopted (Borowska, 2015). Entropy is a measurement of the degree of uncertainty or information content within a system (Quiroga et al., 2001), where higher entropy corresponds to a less predictable, more informative signal (Deolindo et al., 2020; Gao et al., 2016; Rosso et al., 2006). For example, cardiovascular control via sympathetic and vagal regulation is a dynamic and nonlinear process influenced by a multitude of factors, and although linear method are typically used to analyse HRV, nonlinear methods such as entropy analysis have been argued to better represent the dynamic and complex nature of heart rate control (Byun et al., 2019).
In mindfulness research, reduced (permutation) entropy has been found in the EEG activity of experienced meditators compared to novices during both rest and meditation practice (Kakumanu et al., 2018; Vyšata et al., 2014). This result suggests greater synchronisation of the EEG signal in meditators, perhaps indicating reduced information processing associated with an increase in single-pointed attention focus (Kakumanu et al., 2018; Vyšata et al., 2014; Young et al., 2021). In an extension of this research designed to examine how the brain-heart relationship is affected by mindfulness, Gao et al. (2016) and Sik et al. (2017) used the discrete wavelet transform (DWT) to calculate the wavelet entropy of both EEG and heart rate data. DWT decomposes a time series into sets of average or low pass components, where each set of components reflects the evolution of each frequency component across time (Bajaj, 2020; Jacob et al., 2021). The DWT method thus incorporates temporal information with spectral-frequency analysis, thereby taking into account transient fluctuations of electrophysiological signals (Bajaj, 2020; Jacob et al., 2021). Hence, wavelet entropy measures the uncertainty, information content, and complexity of signals in both the time and frequency domain, providing additional insights into the dynamics of EEG and heart rate activity (see Rosso et al., 2006 for a detailed review).
Gao et al. (2016) and Sik et al. (2017) reported reduced Wavelet entropy in EEG and heart rate data during meditation compared to rest in novice meditators. Moreover, both studies also reported a stronger correlation between EEG and heart rate entropy during mindfulness breathing compared to rest. These authors speculated that this result might indicate a greater detachment from visual sensory input and may reflect mindfulness-related effects on the synchronisation of body and mind, suggesting improved coherence of the parasympathetic nervous system.
1.4 Contrasting the Three Theoretical Frameworks
While empirical support for any of the three theoretical frameworks described above would provide evidence to substantiate an increase in mindfulness-related connectivity between the brain and heart activity, the implication of the results from each framework has discernible differences. For instance, under the Bayesian brain framework, significant differences in HEP amplitude between meditators and non-meditators could suggest enhanced bottom-up sensory processing. A significant correlation between fm-theta and RMSSD among meditators would indicate practice related enhancements in top-down parasympathetic regulation. Lastly, a stronger relationship for meditators between EEG and heart rate entropy could provide evidence for greater detachment from external stimuli in meditators than controls.
1.5 Aims and Hypotheses
Results from the studies outlined above (Gao et al., 2016; Manjaly & Iglesias, 2020; Sik et al., 2017; Tang et al., 2009) show evidence of mindfulness-related effects on brain-heart interaction; however, two constraints limit the generalisability of the findings. First, participants primarily involved novice meditators with short-term training. Mindfulness-meditation can affect individuals differently over time, with stronger effects found in meditators with long-term experience (Falcone & Jerram, 2018; Marchand, 2014; Wang et al., 2021). Secondly, comparison conditions within previous research tended to focus on state differences between meditation and rest, rather than a comparison between experienced meditators and non-meditators. While the effects detected during meditation are informative, it is unclear from studies focused only on the meditation state whether effects found during mindfulness practice continue outside of practice sessions. Thus, the present study sought to determine whether mindfulness related changes in brain-heart interaction are evident in experienced meditators while participants are at rest (rather than during active meditation practice). In particular, it was hypothesised that: 1) experienced meditators would show greater neural sensitivity to interoceptive information (as reflected by larger HEP amplitudes) at rest compared to controls; 2) experienced meditators would demonstrate higher levels of autonomic self-regulation, reflected by a significant positive relationship between fm-theta minus 1/f non-oscillatory activity and RMSSD, while no relationship was expected in controls; and 3) experienced meditators would show a significant positive correlation between EEG and ECG entropy while no-significant correlation would be found for controls.