There is growing evidence that focusing on the cultivation of compassion-based emotions has important effects on mental states and well-being (Hofmann et al. 2011
; Maratos et al. 2019
). Research demonstrates that compassion-focused interventions are effective in both clinical and non-clinical populations (Kirby et al. 2017b
). The effectiveness of such interventions is often argued to reflect their empirical bases, including that compassion-focused interventions are derived from an understanding of evolutionary and functional analysis of basic motivational systems. To expand, in the compassion approach originated by Gilbert (see Gilbert 2014
, for review), it is argued that the brain is highly evolved for social processes. Central to this, and supported by neurophysiological research (e.g. Longe et al. 2010
; but see also Kim et al. 2020a
), is the idea of three affect systems within the brain. These are the threat, drive and soothing systems.
The threat-protection system is related to fight-flight tendencies and emotions of anger and anxiety related to sympathetic nervous system activity (see Maratos and Pessoa 2019
, for review). The drive system is related to high arousal positive ‘approach’ emotions such as pleasure and excitement (Depue and Morrone-Strupinsky 2005
). The soothing system is related to those emotions not associated with ‘threat’ or ‘doing’ (Gilbert 2017
); that is, those emotions associated with a state of quiescence, including feeling safe, peaceful and contented. The soothing system also includes emotions related to caring and affiliative behaviours that have been key to mammalian evolution. Importantly, compassion-focused interventions are hypothesised to be linked to the soothing/affiliative system. It is opined that ‘training’ the soothing system (using compassion-focused interventions) allows for better regulation of threat and drive systems linked to sympathetic nervous system activation (Duarte et al. 2015
). Therefore, it is argued that compassion-based interventions do not only impact upon psychological health, but also physiological health (Kim et al. 2020b
; Singer and Bolz 2013
In accord with this, research has demonstrated that compassion-focused interventions stimulate oxytocin, endorphin and regulate sympathetic/parasympathetic activity (for review see Kirby et al. 2017a
). For example, Arch et al. (2014
) investigated if brief compassion training could diminish physiological stress. Once individuals had been trained in self-compassion or a control intervention, they took measures of both cortisol and salivary alpha-amylase (sAA) pre and post a social stress test. Cortisol was their measure of hypothalamic-pituitary-adrenal (HPA) axis activity (often associated with chronic stress responses; see Sapolsky et al. 1986
); and sAA their measure of acute sympathetic nervous system activity (see van Stegeren et al. 2006
). They found that brief self-compassion training dampened/flattened sAA level increases—but not cortisol—and thus concluded that this brief training impacts the fast-acting sympathetic system stress response, implicated in defensive reactions. Further research has demonstrated that receiving oxytocin can increase the ease at which individuals can engage in compassion-focused imagery (CFI). CFI is a brief visualisation technique often used within compassion-focused interventions (see Rockliff et al. 2011
). Thus, compassion-based interventions appear to allow for a state of quiescence, or sympathetic/parasympathetic activity regulation, achieved via (i) dampening of the threat-focused affect system and/or (ii) stimulation of the affiliative affect system.
Interestingly, activation of the affiliative affect system may also be related to pain coping and pain responses. That is, psychological and behavioural research has demonstrated that social support, whether real or imagined, can increase pain tolerance and psychological pain coping (Brown et al. 2003
; Cano and Williams 2010
; Jackson et al. 2005
; Master et al. 2009
). Brown et al. (2003
) found that participants who received support had greater pain tolerance to cold pressor than those who completed the task alone or with no support. Further, Master et al. (2009
) reported that viewing a loved one’s photograph lowered pain responses to thermal pain stimuli compared with viewing objects, suggesting enacting mental representations of loved ones may be sufficient to alter pain responses. Thus, it may be that compassion-based interventions can be used to address pain coping.
Taken as a whole, such research demonstrates the potential buffering effects of affiliative/compassion-based practices when coping with pain. Therefore, the aim of the present study was to explore if brief CFI could improve responses to acute painful stimuli. Specifically, whether engaging in CFI could increase pain tolerance and/or decrease accompanying stress physiology, i.e. salivary alpha-amylase (Nater and Rohleder 2009
) as well as impact upon psychological indicators of positive affect, associated with threat, drive and soothing emotions. We used sAA as our measure of physiological stress, because sAA (i) has been implicated in rapid/acute sympathetic nervous system activation (Arch et al. 2014
; van Stegeren et al. 2006
) and (ii) correlates positively with ratings of acute pain intensity and unpleasantness (Wittwer et al. 2016
). As CFIs have also been found to be effective in reducing fear and anticipatory anxiety (Harwood and Kocovski 2017
), a second aim was to explore the effects of CFI on pain anticipation. To this end, participants practiced and then engaged in control (nature) imagery and CFI whilst subjected to painful stimulation (cold pressor task). To control for order-effects, crossover effects and diurnal variability in sAA, participants took part in the two imagery conditions 7 days apart, but at the same time of day, and order of imagery conditions was counter-balanced.
The purpose of the present study was to investigate the impact of CFI on pain coping, including when subjected to actual pain and when in anticipation of this pain. In respect to actual pain, we found that CFI, as compared with control imagery, attenuated sAA levels to pain. However, whilst pain tolerance did not differ by imagery condition (CFI, control), sAA levels to actual pain predicted decreased pain tolerance in the CFI condition. We further found that anticipatory sAA responses were predictive of pain tolerance, but that this effect was not moderated by condition. To expand, participants with larger anticipatory sAA responses were more likely to remove their hands from the water earlier. Of note, none of the emotional measures of well-being differed by imagery condition, nor by condition over time. Thus, our results indicate effects of CFI on physiological and behavioural measures of pain coping, but not psychological measures of pain coping.
A main result of the present study was that CFI attenuated the stress response to acute pain. That is, participants demonstrated no increase in sAA responses whilst subjected to cold pressor pain, when engaging with CFI, but when the same participants engaged with a control (nature) imagery whilst subjected to cold pressor pain, sAA increases were observed. This result is similar to the findings of Arch et al. (2014
) who also observed blunting of sAA levels when participants were subjected to psychological stress, following compassion training. They argue that lower sAA responses can be interpreted as dampened sympathetic nervous system activity (see also Bauer et al. 2002
) and concluded that self-compassion training can diminish psychobiological responses to threat (in their case the threat of social evaluation). As sAA levels are a reliable marker of psychological stress and fast-acting sympathetic nervous system responding (Nater and Rohleder 2009
), our results demonstrate that engaging with CFI provided downregulation of rapid autonomic nervous system responses to stress when experiencing acute pain. As we only observed these effects when participants engaged in the CFI, and not the control imagery, this demonstrates that our CFI intervention was effective in dampening physiological pain responses. That is, if simply distraction or attention were key, then we would not have observed time-by-condition interaction effects. To expand, we found no difference in our participants’ ability to engage in the CFI and control imagery, including ease of experiencing the imagery, difficulty creating an image, its visualisation aspects and/or resistance towards the imagery. This indicates that it was the compassionate nature of the task, rather than cognitive load, that resulted in the dampening of the physiological sAA stress response. Relatedly, some recent research has found no effect of non-judgmental body scanning on pain responding (Sharpe et al. 2013
). This is consistent with our argument that it is the affiliative emotional aspects of the imagery that is a critical factor in mediating pain responding.
Compassion-focused interventions and pain coping can both be linked to activation of the affiliative affect system, with previous psychological and behavioural research demonstrating that social support increases pain tolerance and psychological pain coping (Brown et al. 2003
; Cano and Williams 2010
; Gregory et al. 2017
; Jackson et al. 2005
). Additionally, Gooding et al. (2020
) have argued that in individuals who experience persistent pain, the threat system may become dysregulated and compassion-based interventions address this dysregulation. In explaining our findings, it can be hypothesised that engaging in CFI activated the affiliative/self-soothing system linked to parasympathetic nervous system activation, allowing for sympathetic nervous system regulation. Recently, oxytocin has been found to increase affiliative processes such as compassion, and intranasal administration of oxytocin increases pain tolerance to experimental noxious stimuli (Boll et al. 2018
). We hence speculate that the blunted sAA responses observed reflect CFI: (i) providing a buffering effect in response to pain/pain perception and appraisal and/or (ii) downregulation of the stress response as a consequence of affiliative affect system activation. However, to further investigate mechanisms underlying exactly how CFI impacted upon the physiological stress response to acute pain, in future research, it is recommended that measures of oxytocin are also taken.
Contrary to predictions, we did not find an effect of condition (CFI vs. control) on pain tolerance; engaging in CFI did not significantly impact upon how long participants were able to keep their hand emerged in the cold water, as compared with control imagery. Whilst the study was powered to detect medium effects, the effects we observed were smaller. Therefore, larger studies are needed to detect similar effects (if they exist). This stated, in prior research, when effects of compassion-based approaches on pain tolerance have been found, they occur only in participants with a history of self-injury (Gregory et al. 2017
). Hooley and St. Germain (2014
) suggest that those without a history of self-injury are likely to have a reasonable level of pain sensitivity, in which case one would not expect changes in pain tolerance, even if changes in (state) compassion are observed. Our results accord with this by demonstrating that in our non-selected participants, CFI did not affect pain tolerance, but did affect the stress response to pain. In future research it may therefore be useful to obtain information regarding history of self-injury in potential populations. Alternatively, it could be that what is key to pain tolerance is distraction or cognitive load. For example, research has demonstrated that distraction via virtual reality reduces reports of pain (e.g. Hoffman et al. 2000
; Kipping et al. 2012
). Consistent with this, we did find that ease of engaging with the imagery tasks was positively correlated with pain tolerance in both
conditions. Added to this, as analysis of our imagery fidelity metrics revealed no difference in the ability of individuals to engage with the two different imagery types, it could be that both were equally effective (or otherwise) in altering pain tolerance. To explore this further, a control ‘no intervention’ group would be necessary. Assuming a suitable sample size, this would allow one to establish if simple distraction is, or is not, key to behavioural indices of pain tolerance.
We did not find a direct effect of CFI on anticipatory sAA responses. However, we did observe that participants with larger anticipatory sAA responses per se, were more likely to remove their hands from the water earlier. Thus, a greater physiological sympathetic nervous system response to the anticipation of pain was associated with decreased pain tolerance (Kyle and McNeil, 2014
). Pain anticipation and expectation is frequently related to pain tolerance (Burton et al. 2016
; Peerdeman et al. 2016
). Whilst we did not find an effect of CFI on anticipatory pain, it should be noted that in previous research where differences have been found (e.g. Harwood and Kocovski, 2017
), the anxiety provoking scenario (a speech task), as well as the measure of anticipatory anxiety (Spielberger State-Trait Anxiety Inventory–State Version) were quite different. Additionally, Harwood and Kocovski only observed effects for high anxious participants. This suggests that a range of psychological and physiological measures be taken in future research to better explore the effects of CFI on anticipatory responses.
sAA responses were predictive of pain tolerance in the CFI condition, but not the neutral condition. To expand, participants who demonstrated larger sAA responses to painful stimuli when engaged in CFI were more likely to withdraw their hand from the cold water earlier. Previously, sAA levels have not been shown to impact upon behavioural indices of pain tolerance. That is, although Wittwer et al. (2016
) found sAA levels to increase with subjective ratings of acute pain intensity and unpleasantness, they found no relationship between sAA and objective ratings of pain tolerance (in their research, this was increases in heat stimuli that an individual could tolerate). As we found sAA responses were predictive of behavioural pain tolerance, but only in the CFI condition, this suggests that there is a relationship between pain sensitivity, the physiological stress responses and compassion. Of note, it has been observed that individuals who demonstrate pain insensitivity are more likely to be self-critical. As self-criticism is diametrically opposed to compassion (Gilbert et al. 2004
; Neff et al. 2019
), it can be speculated that in our study individuals who did not show dampening of the sAA responses did not demonstrate the benefits of the CFI intervention and hence withdrew their hand from the water given the associated physiological stress response and ‘painful’ experience (Peckerman et al. 1994
). However, participants in the current study were selected based on low self-criticism scores; post hoc, we examined relationships between FRCRS subscales, pain tolerance and sAA responses, but they were not related (these data are available from the authors). Gregory et al. (2017
) have found that individuals with a history of self-injury, as compared with those without, demonstrated greater pain sensitivity (to cold pressor) following a task designed to increase self-compassion. As such, they suggest that increased self-compassion (in their study induced by a values affirming task) helped correct the pain insensitivity that is typically found in those who self-harm. In this context, an elevated physiological stress response and ‘painful’ experience would lead to greater sensitivity to pain through the CFI and consequently also explain the reduced pain tolerance for those whom demonstrated larger sAA responses. Future intervention studies should examine these possibilities with a sample with a range of self-criticism and self-compassion scores, and assess changes in state measures, to clarify these relationships.
Finally, whilst we observed a number of changes concerning the effects of CFI on our physiological indices of well-being, our emotional measures of well-being were not influenced by the different imagery conditions, nor by condition over time. Whilst Penlington (2019
) observed changes in psychological indices of well-being (e.g. anxiety), the parameters of that research were considerably different; that is, the intervention lasted a period of 12 months. This highlights the importance of choosing brief measures that are sensitive to change alongside physiological measures that can be obtained easily (e.g. non-invasively). Additionally, whilst it is possible that participants were fatigued by the completion of the same measures, three times over the course of an hour, changes over time were observed in the expected directions. To expand, immediately before pain induction (i.e. anticipation of the painful cold pressor task), participants reported less feelings of safety and relaxation, and greater feelings of negative affect, as compared with baseline and post-pain induction. This suggests that participants were not fatigued by completion of the measures (e.g. acquiescence bias), but neither intervention influenced emotional responding either.
Limitations and Future Research Directions
It is important to note that despite our promising results, before clinical efficacy can be argued, our findings need replication in both larger, and clinical, samples. In those replications, individual differences and potential confounds, such as smoking and hormone status that we did not assess due to the repeated measures design, should be investigated or controlled. A second limitation of the present research was our failure to measure intensity of pain perception which, in previous studies of pain (i.e. self-injury) and compassion, has been found to be an important outcome variable. That is, in such research, compassion-based interventions have been found to robustly impact upon ratings of pain intensity if not pain tolerance (Gregory et al. 2017
). In future research, therefore, a measure of pain intensity (and unpleasantness) would be a significant improvement to the present design. Thirdly, if
CFI is attenuating physiological stress responses to pain via upregulation of the soothing/affiliative system, then we would also expect behavioural reports of self-compassion to increase following the specific CFI. Thus, it is of paramount importance that any future research investigating effects of CFI upon physiological, behavioural and psychological pain coping also includes a state measure of self-compassion. Additionally, our control imagery task may not have been as neutral as intended; recent work has highlighted that imagery and other ways of connecting to nature are stress reducing. For example, proximal nature buffers the effects of low levels of social connection on well-being (Cartwright et al. 2018
). This stated, we did find our hypothesised physiological dampening for CFI relative to the nature imagery task, and participants reported similar imagery fidelity.
Our intervention further promoted self-compassion, which is distinct from compassion for others (López et al. 2018
). Interestingly, empathy to others’ pain has been found to increase
pain perception due to its relation to attachment processes including social rejection (Eisenberger 2012
; MacDonald and Leary 2005
). Thus, studies that investigate the interplay between self-compassion and compassion to others (Gilbert 2014
) in relation to pain are warranted. Of relevance here, Carson et al. (2005
) demonstrated that an 8-week loving-kindness programme for chronic lower back pain patients significantly improved psychological coping. Their programme included elements of loving-kindness (including compassion) towards the self, as well as others—including loved ones, a neutral person and those they had experienced difficulties with, or who had previously harmed them. Results revealed that the loving-kindness meditation (LKM) reduced ratings of pain, anger, and psychological distress in the patient group. Carson et al. hence argued that LKM may be a useful therapy that clinicians can integrate into their pain treatment programmes, whilst also noting that the mechanisms via which it operates needs further exploration. Certainly, teasing apart the relative importance of different aspects of compassion (and LKM) is a valid future direction. This includes investigation of any mediating effects of whom the imagery is directed towards and/or from; and whether the CFI is imagined or embodied.
One strength of the present study was the use of a crossover repeated measures design. This allowed for substantial control over extraneous variables, such as time of day, which, at least in part, minimised baseline differences between variables of interest; and there were no apparent crossover effects. Indeed, the approach is often described as the gold standard of trials (Mills et al. 2009
). However, it does limit the possibility of examining individual differences. Of rising importance here, is the inclusion of aphantasia (i.e. imagery visualisation ability) as a moderating variable. With the efficacy of imagery style interventions becoming increasingly investigated and reported, factoring in an individual’s ability to engage with imagery per se is an important next step in the design of future studies. Certainly, in reviewing the fidelity data, that participant generally reported moderate ratings of their ability to engage with the imagery, including creating an image, indicates that for some engaging in the imagery tasks might not have been that easy. This of course could have also influenced findings (e.g. implementation efficacy).
To conclude, our data demonstrate that CFI provides a dampening of the physiological stress response to actual painful stimuli. To more comprehensively evaluate the effectiveness of CFI on pain coping and pain tolerance, it is important that previous history of self-injury, anxiety etc. is investigated alongside the actual intervention as well as self-reports of compassion and pain intensity/unpleasantness. Additionally, to more fully understand the physiological mechanisms via which CFI is exerting its pain coping effects, a measure of affiliative responding (e.g. oxytocin) would be useful. Other compassion-based approaches, for example compassion for others, could also be evaluated in the context of pain to ascertain if it is the affiliative emotional content of the imagery that is key to effects observed. Indeed, as CFI appears to offer a means by which to dampen physiological stress responses to pain, it may in future offer new avenues for therapeutic interventions, including those for chronic pain patients (Purdie and Morley 2016
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.