Effects of brief mindfulness meditation on compassion and personal distress using the socio-affective video task

Abstract

Recent findings indicate that meditation may enhance compassion and foster prosocial actions aimed at helping others. In the present preregistered research, we aimed to test hypotheses regarding the positive impact of a single short, guided mindfulness meditation on compassionate and personal distress responses, measured with the Socio-affective Video Task (SoVT), and the possible mechanisms of this effect. We hypothesized that individuals in the mindfulness group, compared to the active control group, would exhibit higher levels of compassion and reduced personal distress while viewing videos of people in distress. We also proposed that decentration and self-compassion would mediate this effect. Eighty individuals, novice to meditation, were randomly assigned to either the experimental (meditation) or control condition groups. We did not find a significant overall group effect on the presented videos. However, we demonstrated a serial partial mediation process, in which mindfulness state—measured through manipulation check questions assessing attention focused on breathing and awareness of bodily sensations—serves as the first mediator, and decentration as the second, mediating the relationship between group and SoVT measures. These findings suggest that even brief mindfulness meditation can enhance compassionate/empathic responses through increased mindfulness precursors and improved decentration abilities. More studies with longitudinal or experimental designs are needed to better understand the causal relationship between mindfulness meditation and compassionate responses.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-16416-4.

Introduction

Empathy and compassion serve as pivotal socio-affective states guiding individuals through their social interactions. It is difficult to overstate their importance and their significance is further underlined by robust evidence of their impact on prosocial behavior^1–3. Both empathy and compassion, however intertwined they are, exhibit nuanced conceptual and biological distinctions^4,5. Therefore, it is essential to establish clear definitions that present the shared interdependencies and discernible differences between these two constructs.

Empathy is often reflected by two dimensions - cognitive (also known as perspective taking or theory of mind⁶ which is an ability to imagine the perspective of others and affective. The latter, the affective dimension, is manifested by mirroring the emotions of a sufferer within the observer^4,7. It is usually described as an unpleasant and impairing aspect of empathy and as such, may block potential acts of prosociality^1–3 and is referred to herein as ‘personal distress’.

Compassion is often suspected to enhance helping others⁸. It is defined as a feeling of warmth and concern towards sufferers accompanied with the willingness to alleviate their state^4,9. This definition is supported by numerous studies proving its ability to predict prosocial intentions and behaviors^10–12. It is important however to note that personal distress and compassion are not the opposite sides of motivation to help and subsequently any helping behavior; instead, they are thought to complement each other. While distress and empathy in general allow individuals to recognize others’ suffering, compassion neutralizes the impairing effects of excessive distress and may prompt individuals to take action^12–14. Therefore, in our study, we focus specifically on these two states—personal distress and compassion—as we consider them the primary conditions conducive to prosocial behavior.

While learning about the role of personal distress and compassion in motivating prosocial behavior, scientists became interested in altering those and consequently, aimed to improve social relationships^2,15. Yet, the question of what can trigger an augmentation of these states leaves space for speculation. Among others, some recent studies indicate mindfulness meditation as a catalyst for empathy, compassion, and subsequent prosocial qualities (e.g^16–18. , see also a meta-analysis of effects of mindfulness meditation on prosociality¹⁹. Mindfulness meditation is a practice of non-judgmental awareness of the present state²⁰. Its effects on personal distress and compassion have been widely researched²¹. For example, mindfulness training may diminish personal distress (e.g^22–25), and this could be explained by different facets of mindfulness, e.g. trait nonreactivity could predict lower effects of trait personal distress²⁴. Nevertheless, the effects of mindfulness interventions appear to vary depending on their duration, ranging from months to a single short session^19,26. The 8-week Mindfulness-Based Stress Reduction program²⁰ has demonstrated improvements in various aspects of empathy (by heightening perspective taking but lowering personal distress) and compassion^22,26–28. Recent meta-analyses have explored the relationship between mindfulness training duration and its impact on prosociality, revealing larger effect sizes for brief training compared to longitudinal interventions^29,30. In addition, a more critical stance over mindfulness as a prosociality promoter starts to appear. Although often linked with ethics and higher empathy and compassion³¹, mindfulness does not improve prosociality unconditionally. For example, authors of a meta-analysis on mindfulness and prosocial behavior conclude that mindfulness may not operate simply in a twofold manner—prosocial or antisocial—but may instead require more complex mechanisms, such as prosocial intrinsic motivation, to enhance helping behavior³². Further on, others underline the importance of guilt—an adaptive and socially beneficial emotion—in undertaking prosocial reparative behaviors. While mindfulness may foster certain aspects of prosociality, it might simultaneously inhibit reparative behaviors by reducing feelings of guilt³³. In support of this nuanced view, another study explored the relationship between multifaceted trait mindfulness and actual helping behavior, taking into account experimentally induced guilt. The findings were mixed, leaving the question of how mindfulness relates to prosociality—and through which mechanisms—open for further investigation³⁴.

As the majority of research in the field of prosocial emotions and states along with actions employs correlation design, clearly more experiments using various methods are needed. Emerging paradigms, such as video-based tasks, may offer new insights into empathic and compassionate responses in a more ecologically valid manner, as they mitigate the effects of social desirability bias and self-narratives, and do not rely on an individual’s memory³⁵. For example, Socio-Affective Video Task (SoVT) is a tool developed to measure empathy and compassion using short videos^13,36. In the original study, participants watched a set of 12s videos depicting some people in either neutral (talking, walking etc.) or suffering (emotionally charged) condition. Recently, the validity of SoVT was tested and compared to standard empathy-measuring questionnaires — the Interpersonal Reactivity Index⁷ with promising results demonstrating its usefulness for empathy and compassion assessment³⁵.

In this study, we aim to investigate the effects of a brief mindfulness training on personal distress and compassion using SoVT. We theorize that a brief mindfulness training would cause lower personal distress and higher compassion of participants as a reaction to suffering, compared to an active control group. We also look for possible mediators of these effects, focusing specifically on decentration and self-compassion, independently. Decentration is a process of observing one’s subjective thoughts and emotions using the third person perspective as if they belonged to another person³⁷. Described as a meta-mechanism in early theoretical models of mindfulness (referred to as reperceiving or detachment), it allows the adoption of non-judgmental and compassionate stance towards objects of consciousness³⁸. In a study of the 8-week mindfulness training, the rise of mindfulness facets measured with the Five Facets Mindfulness Questionnaire³⁹ was accompanied with a significant rise of decentering⁴⁰. Interestingly, the change in mindfulness (measured with three facets of the Five Facets Mindfulness Questionnaire: observe, nonjudge, nonreact) correlated highly with the change in decentering (r = .73, p < .01), indicating a close relationship between these constructs. Even brief mindfulness trainings have demonstrated the ability to enhance a decentered approach to emotions, as evidenced by pilot studies⁴¹. Decentration may also influence specifically socially oriented emotional responses by acting as a significant mediator between dispositional mindfulness and personal distress⁴².

Similarly to decentration, self-compassion has gathered increasing attention within the context of mindfulness training^22,43. It is defined as a feeling of kindness and healing towards one’s self^44,45. Interestingly, self-compassion is not explicitly taught within mindfulness training. It is suggested though that within such a 8-week training, as mindfulness facets improve and well-being does too, self-compassion may serve as a mediator of this influence⁴³. Others propose that although mindfulness training does not explicitly address self-compassion, it is one of the mindfulness “invisible heroes” enhancing meditators’ well-being by fostering kindness towards oneself and others⁴⁶. Moreover, personal distress, as a self-focused affective response, often leads to withdrawal or avoidance rather than prosocial behavior^7,47. Excessive exposure to others’ suffering, without adequate emotional regulation, may hinder the ability to respond compassionately. Thus, we hypothesized that an increase in self-compassion might enhance one’s capacity to respond compassionately to others’ distress. However, self-compassion can be perceived as a more complex psychological ability, requiring top-down strategies, and therefore less prompt than decentering. Although studies suggest it can be rapidly taught, for example, through modeling^48,49, its spontaneous emergence warrants further examination, particularly in short-term experimental designs such as the one presented here. As a recent meta-analysis shows⁵⁰, self-compassion interventions hold promise for reducing stress and depressive symptoms, with medium effect sizes observed. However, the average length of interventions in this meta-analysis ranged from 1 to 16 weeks (5.1 on average), highlighting the need for further research into the efficacy of shorter interventions in fostering self-compassion. Numerous studies discussed above have demonstrated significant correlations between dispositional mindfulness, decentering, self-compassion, personal distress and compassion to others (e.g^22,51,52. Some of these studies have indicated or speculated on the potential role of decentering and self-compassion as mediators of the influence that mindfulness meditation may have on empathy and compassion^{42,53–56, 57}. Much less is known about whether inducing a mindful state through meditation practice in novices affects prosocial emotions and the underlying mechanisms of this effect. With the current study, we aim to address this gap. However, we treat our proposed mediators—decentering and self-compassion—as independent potential mechanisms for enhancing prosocial emotions. While some evidence suggests a hierarchical relationship between these constructs, more detailed causal rather than correlational studies are needed [Biehler and Naragon-Gainey 2022].

Present study

In the present study, we aimed to verify if brief mindfulness meditation practice conducted online decreases personal distress and increases compassion towards other people and to investigate if decentration and self-compassion act as mediators of this relationship. In the preregistration of the current research project, we formulate hypotheses regarding the influence of mindfulness practice on empathy and compassion towards others, both ingroup and outgroup members. We decided, however, to refine our hypotheses and for the sake of the clarity of the scientific argument, we describe this preregistered project’s results in two separate manuscripts. The current one is focused on the impact of brief mindfulness training on prosocial emotions towards other people in general, with evaluation of the potential mediators of these effects. Whereas, in the separate manuscript we will describe results regarding the influence of brief mindfulness training on the emotions towards in-group vs. out-group. To see all the hypotheses preregistered under one project, please follow the link: https://osf.io/g42js/. Importantly, the hypotheses, sample size, and methods remain unchanged. We have taken great care to ensure transparency in presenting all results. Within this article, the following hypotheses were tested:

H1: Brief mindfulness meditation training will cause greater compassion and lower personal distress for suffering others than the control.

H1a: There will be a general improvement in compassion for the sufferers in the experimental group compared to the control.

H1b: There will be lower personal distress for sufferers in the experimental group than the control.

H2: State decentration and self-compassion will be mediators of the general influence of mindfulness meditation on prosocial emotions:

H2a: State decentration and self-compassion will be mediators of the general influence of mindfulness meditation on compassion.

H2b: State decentration and self-compassion will be mediators of the general influence of mindfulness meditation on personal distress.

Results

Manipulation check was calculated as a mean of five statements regarding the listening experience and independently, seeing the effect of each statement on the dependent variables. Since the assumption of homogeneity of the variances was not met (F(1,78) = 7.05, p = .009), we used a non-parametric test to verify the manipulation check questions (separately) across the groups: question 1 (“while listening to the audio, my attention was focused on breathing”): W(1, 80) = 294, p < .001; question 2 (“while listening to the audio, I was aware of my bodily sensations caused by breathing”): W(1, 80) = 453.5, p = .001; question 3 (reversed, “I was running away from the content of the audio”): W(1, 80) = 751.5, p = .771; question 4 (“Listening to the audio was easy to me”): W(1, 80) = 742, p = .573; question 5 (reversed, “Listening to the audio was not comfortable to me”): W(1, 80) = 769.5, p = .77. The last three questions presented non-significant results of the manipulation check effect and therefore did not differentiate between the experimental vs. the control group, proving a similar quality of both audio recordings.

Hypotheses testing

Using Hypothesis 1, we aimed to verify the effects of the experiment by comparing the state compassion and personal distress between the experimental group listening to mindfulness audio guidance and the control. We established three SoVT ratings as a base for the comparisons — (H1a) Compassion, Relief, and (H1b) Distress. Our hypotheses predicted a higher response on the Compassion and Relief ratings and lower on the Distress rating for the experimental group.

First, we verified the t-test assumptions and confirmed the normality of distributions for the SoVT ratings using the Shapiro-Wilk normality test as well as their variances’ homogeneity, using the Levene’s test. We received the following results for the SoVT Compassion rating of the normality test: W = 0.96, p = .19 and for the control: W = 0.96, p = .19, proving normal distributions of the two variables. Levene’s test confirmed the homogeneity of the two variances: F(1,78) = 1.05, p = .31. Finally, we compared the groups using the t-test and confirmed no significant differences between them: t(74) = 0.27, p = .79. We observed normal distributions for The SoVT Relief ratings: W = 0.97, p = .30 (for the experimental group) and W = 0.96, p = .19 (for the control), with Levene’s test confirming the homogeneity of the variances: F(1,78) = 1.69, p = .20. The t-test result proved no significant differences between the groups: t(72) = -0.11, p = .91. Finally, the Distress rating proved normal distributions of the variables: W = 0.985, p = .83 (for the experimental group) and W = 0,95, p = .08 (for the control), and homogeneity of the variances: F(1,78) = 3.24, p = .08. The t-test showed no significant differences between the two groups: t(70) = 0.04, p = .97. In summary, Hypothesis 1 was not confirmed, there was no main effect of the experimental vs. control groups on none of the SoVT ratings.

We hypothesized that decentration and self-compassion will mediate the effect of training on compassionate and distressed responses as stated in Hypothesis 2. However, due to low reliability of the self-compassion scale, we focused solely on the effects of decentration. First, we verified the effects of decentration as a mediator on compassion-related measures (Compassion and Relief) using Structural Equation Modeling. We tested the data for the assumptions and established a model presented in Fig. 1 with one predictor (experimental vs. control audio), one mediator (Decentering) and a combination of two outcomes (SoVT, Compassion and Relief ratings). The model did not show satisfactory results: χ²(131, 80) = 757.11, p < .001, TLI = 0.50, RMSEA = 0.24.

Fig. 1.

Hypothetical model for Hypothesis 2.

Secondly, we measured the same model with only the SoVT Distress rating as the only outcome. The second model with Distress showed results close to significance: χ²(34, 80) = 60.04, p = .004, TLI = 0.91, RMSEA = 0.098. According to the convention, the indices show an almost significant effect (0.90 < TLI < 0.95 and RMSEA > 0.08⁵⁸. The SEM models did not show a consistent proof for mediation analyses.

Our hypotheses were designed to measure the effects of a short mindfulness meditation on state compassion and state personal distress. The second aim was to verify the effects of two suggested mediators — self-compassion and decentration. We could not confirm our preregistered hypotheses. There was no significant main effect of audio recordings on any of the outcome variables (Hypothesis 1). There was no significant mediation effect visible in the analyses, except for the model predicting Distress through Decentration which was close to significance (Hypothesis 2). Since the study showed mixed results, we decided to run several exploratory analyses in order to better understand the data and the following outcome.

Exploratory analyses and results

Based on an encouragement of persuading the exploratory section in the preregistered studies, we explored our data⁵⁹. More importantly, we extended the model suggested in Hypothesis 2 by adding a new first-step mediator (the two mindfulness state-related questions of the manipulation check).

Manipulation check questions — inspired by various studies^30,60 initially served as assurance of an effective manipulation that differentiates the experimental meditating group from the control. It drew our attention that the group differences proved at the beginning of the Results section were especially significant for the questions asking about precisely what constitutes the essence of mindfulness meditation— a focus on body and breathing (mindfulness state question 1: “while listening to the audio, my attention was focused on breathing” and mindfulness state question 2: “while listening to the audio, I was aware of my bodily sensations caused by breathing”). We formed a statement deriving directly from the tested hypotheses, however improved with the two mindfulness questions. We suggested that it is not purely the effect of manipulation that plays a key role in differentiating the effects of mindfulness from the control, but the quality of the performance while meditating. To check that statement we tested a serial mediation model in which the audio condition influenced SoVT variables through mindfulness state question 1 or 2 and Decentering.

Exploratory analysis using manipulation check question 1

Distress as an Outcome

Using the PROCESS package⁶¹, the model predicting SoVT Distress as an outcome was significant, F(1, 79) = 39.29, p < .001. The indirect effect, including both mediators (mindfulness question 1 → Decentering → Distress), was also significant, B = 0.293, 95% CI [0.072, 0.617]. However, when Decentering was omitted, the effect was stronger, B = 0.867, 95% CI [0.297, 1.594]. The updated model from Hypothesis 2 with standardized coefficients is shown in Fig. 2. To verify these results, we used Structural Equation Modeling (SEM), which confirmed a significant model fit, χ²(42, 80) = 66.36, p = .01, TLI = 0.93, RMSEA = 0.09.

Fig. 2.

Mediation model with standardized regression coefficients. Standardized regression coefficients for the relationship between the experimental condition and distress as mediated by mindfulness state question 1 and decentering. The standardized regression coefficient between the experimental condition and distress, controlling for the mediators, is in parentheses. **p < .01.

Compassion and relief as outcomes

For Compassion and Relief, only the full mediation model (audio condition → mindfulness question 1 → Decentering → Compassion/Relief) was significant, B = 0.260, 95% CI [0.058, 0.532]. However, SEM did not confirm the model when both Compassion and Relief were included as outcomes, χ²(246, 80) = 781.73, p < .001, TLI = 0.51, RMSEA = 0.23.

Exploratory analysis using manipulation check question 2

We then tested a similar mediation model using manipulation check question 2 to predict Distress. Here, the indirect effect including question 2 → Decentering → Distress was significant, B = 0.114, 95% CI [0.018, 0.269]. SEM confirmed the model, χ²(42, 80) = 68.56, p = .006, TLI = 0.91, RMSEA = 0.09.

For SoVT Compassion, the indirect effect (question 2 → Decentering → Compassion) was significant, B = 0.106, 95% CI [0.015, 0.252], as well as for SoVT Relief, B = 0.101, 95% CI [0.016, 0.242]. However, the full model including Compassion and Relief as outcomes was not supported by SEM, χ²(146, 80) = 784.53, p < .001, TLI = 0.50, RMSEA = 0.23.

More details on SEM analyses verifying the updated model are provided in the Supplementary Materials.

In conclusion, adding the manipulation check questions 1 and 2 allowed some of the SEM models to be significant. Depending on the outcome variables, either serial mediation could significantly predict the outcome or including only the manipulation check questions allowed the mediation to present a significant indirect effect.

Discussion

The current study aimed to verify whether a brief mindfulness training can enhance compassionate response to suffering and inhibit the reaction of personal distress. The second goal was to examine the importance of two assumed mediators linking the impact of mindfulness meditation with personal distress and compassion towards others. We divided participants novice to mindfulness meditation randomly into two groups—experimental one that listened to a short meditation guidance and control which listened to a calmly read text. We asked them mindfulness state questions which partially constituted our manipulation check. Using SoVT, we collected ratings of personal distress and two different aspects of compassion — an emotional one and a motivational one.

First, in Hypothesis 1 we posited that a brief mindfulness meditation would by itself cause higher compassion and lower personal distress to observed human suffering. The findings showed no significant differences between the groups on our SoVT ratings. We couldn’t support the notion that a short mindfulness meditation, without any direct clues regarding compassion or empathy towards others, can improve prosocial feelings in a reaction to suffering. This observation aligns with previous studies that have raised skepticism about positive effects of a brief mindfulness-based training^30,62,63. The incongruent conclusions to whether a brief mindfulness training may impact responses to suffering shows the methodological challenges of empathy and compassion measurement. For decades a paper-pencil questionnaire based on self-assessment has become one of the most common tools measuring empathy. For example, the Interpersonal Reactivity Index, developed by Davis in 1980, has been widely used across the globe to measure empathy with its components, however, with mixed results^64–68. A recent meta-analysis of this questionnaire proved it yet to be inconclusive whether this classically four-dimensional scale is still the most efficient, reliable and valid method for the purposes it claims to serve⁶⁹. Meanwhile the SoVT captures actual momentary states, omitting the effects of social desirability bias and relying on self-narratives and memory³⁵. We suggest the video-based tools can serve as a litmus paper verifying the effects of empathy and compassion on prosociality measured with the classic questionnaires.

The study results raise a question whether a single mindfulness training done by a novice could result in a significant effect on empathy and compassion, strong enough to be detected by a simple mean comparison. There is some evidence for an influence of a short mindfulness training on, f.e. stress reactions⁷⁰ or coping with chronic pain⁷¹, and these improvements are also supported by meta-analyses, such as the one of⁷²—where even one 5-minutes long training can improve current state, relax, diminish anxiety and ruminations. However, as another study states⁷³, it could be that the quality of a meditation practice can be the key factor in the influence of mindfulness. In addition, the impact of mindfulness training on socio-affective processes is not as evident—some articles show benefits of a very short (5-minutes long or longer) mindfulness intervention on prosocial emotions, although these results are not warranted^26,29,74,75, and do not point to specific mediators of such positive influence.

In Hypothesis 2, we sought more detailed information on our experimental manipulation. We hypothesized that the influence of listening to the audio on empathetic and compassionate responses to suffering will be mediated by self-compassion and decentering. Due to the low reliability of the Short Self-Compassion Scale, we could not fully verify the hypothesis, leaving the question of self-compassion’s role in compassion and empathetic distress to suffering open. Ultimately, we decided to test only the potential effects of decentering, as measured by the Decentering subscale of the Toronto Mindfulness Scale. The SEM analysis revealed a close to significant model with audio condition as a predictor, decentration as a mediator and SoVT Distress as the only outcome (other SoVT ratings only lowered the goodness-of-fit measures). For this analysis, the Distress rating was treated as a latent variable, deriving from the eight distress ratings of each video. Usually, each SoVT rating is calculated as a simple mean of responses to all the stimuli, which is a common approach⁷⁶. We see our methodological change and its effects as an interesting implication for further discussion of how to analyze such video-based quasi-behavioral measurement tools.

One notable outcome of our study is the significant difference between experimental and control groups on the manipulation check questions: ‘While listening to the audio recording, my attention was focused on breathing’ and ‘While listening to the audio recording, I was aware of the physical (bodily) sensations of my breathing.’ As expected, participants in the mindfulness condition reported greater focus on breath and bodily sensations than those in the control group. This suggests that our manipulation functioned as intended.

However, when testing decentration as the sole mediator, most effects did not reach significance, suggesting that another factor was at play. By introducing an additional mediation step—our two manipulation check questions separately—we aimed to refine our model. Our findings suggest that merely listening to a guided mindfulness meditation was insufficient to elicit prosocial responses. Instead, the extent to which participants engaged in attentional focus on breath and awareness of bodily sensations—meaning their ability to evoke a mindful state—appeared to be a key determinant. This aligns with prior research indicating that awareness of breathing and bodily sensations are foundational elements of mindfulness, possibly influencing emotional and social outcomes^77,78.

Our results are significant in the context of previous studies, where mindfulness states were frequently assessed through manipulation check using measures such as the Toronto Mindfulness Scale^{29,79,80,81 ,82}. Our findings contribute to this debate by suggesting that attentional focus and awareness of bodily sensations during practice—overlapping prerequisites of mindfulness meditation—may serve as a meaningful mediator of mindfulness-based effects. In the current study, the audio guidance was intentionally designed to exclude explicit prosociality cues, allowing us to test the minimal paradigm that can boost compassionate reactions. Surprisingly, this effect did not rely on experimental manipulation and the ability to focus attention and being aware of bodily sensations played a key role. However, the effect we received does not fully provide a direct proof for the influence of mindfulness meditation on prosocial emotional states. It only partially supports the hypotheses driven from the literature review: a brief mindfulness training does impact positively compassionate responses to suffering through attention focused on breathing and bodily sensation awareness as predicted; it does also enhance distressful reactions to suffering, which should be researched further.

The intricacies of investigations into the effects of mindfulness meditation on prosocial behavior pose typical challenges encountered in this research domain. For instance, a recent meta-analysis³⁰ showed evidence for publication bias in this kind of research. In the current project we decided to use a preregistration form, which includes among others the preregistration analysis plan in order to preclude occurrence of such bias. There is evidence that studies including preregistration have a bigger chance to avoid publication bias as well as p-hacking^83,84.

A variety of meditation techniques are researched altogether in meta-analyses, making it difficult to verify which kind of meditation is specifically effective in promoting prosociality. The explicit ethics-based language typical for Loving-Kindness or compassion meditation is suspected to promote prosociality by simply priming a prosocial language³⁰. In the case of mindful breathing meditation we have employed in the current study, there are no words directly related to prosociality, therefore the mechanism of impacting prosociality must be different. Our research indicates that mindfulness meditation can affect prosocial emotions, even without a direct focus on prosocial behavior. It seems, however, that this effect relies on the ability of the practitioner to develop a mindfulness state, marked by the ability to focus on bodily sensations while breathing.

Previous work showed that the effects of meditation can be stronger when compared to inactive control groups^30,85. Therefore, we included active control that was designed in such a way to mitigate the experimental group. As the aforementioned meta-analyses pointed out a stronger effect of compassion improvement when a meditation teacher was a co-author of a study, we invited a third party experienced mindfulness meditation teacher to record both audio conditions. Notably, both referred meta-analyses indicated that only the effect sizes of single mindfulness session studies with prosociality measurement following directly the intervention are reliable. Therefore we believe that our study creates a strong case for supporting the link between mindfulness and prosocial emotions.

Limitations and future directions

The current study is not free from limitations. The obtained results of influencing prosociality depend strongly on the adopted methodology which includes the aspects of prosociality to be measured and the tools chosen for the measurement. For example, self-reports can be applied with social desirability tests to avoid social biases³⁰. In the current study, we opted for a different class of methods—a quasi-behavioral video-based one which is suspected to be less susceptible to a social desirability than the classical tools⁸⁶. We also encourage a cautious interpretation of the exploratory findings as they warrant a further investigation with a focus on key mechanisms of mindfulness such as the ability to focus attention on breathing and bodily sensations. These mechanisms should be examined using more robust methodological designs, with particular attention to reliability and validity, as the current results were based on single-item measures, which may limit generalizability. Next, our study consisted of one active control group while some recent meta-analyses underline the importance of a variety of groups for comparisons in mindfulness-caused prosociality studies^30,85. A presence of a second and inactive control group is considered a valid method to rule out factors not specific to mindfulness that imply a change in prosocial emotions^30,87. Therefore, we consider such a group design fruitful in distilling the specific factors of mindfulness that may influence empathy and compassion.

We hypothesized self-compassion to be one of the two possible mediators of the influence of mindfulness training on prosocial emotions but we could not verify this effect due to the lack of satisfying reliability level in the self-compassion questionnaire. A further development of this tool supported with some other measures for self-oriented kindness could enrich understanding of mindfulness-based influences. For the future directions, bigger group sizes would allow more fine grained analyses, such as adding other factors to the SEM analyses. For example, more covariates such as personality traits could be considered when analyzing prosociality enhancement⁶³.

In the current study, we did not aim to verify the longitudinal effects of a single mindfulness meditation on prosocial emotions. Yet, as we proved a single effect of such an intervention, a verification of stability of these results is a logical next step for further explorations. A longitudinal study would benefit with another advantage一such designs with multiple points of measurement are more reliable when testing mediators⁸⁸. In the current cross-sectional study, we wanted to omit any biases in empathetic and compassionate responses (the outcome variables). Therefore, we positioned the mediators’ questionnaires that could have biased our participants’ responses only after the measurement of dependent variables. This order of questions allowed us to avoid the influence of self-compassion and decentering questions on reaction to suffering of others, however, it is not the advised order. The results we obtained need to be taken with caution and preferably verified in future studies with different designs.

Conclusions

To conclude, by using a brief mindfulness training and measuring direct immediate outcomes of it, we aimed to verify whether such a short intervention would benefit with bigger compassion and smaller personal distress towards people depicted in short videos. Our study showed no such effects. However, when we included mindfulness-state questions and decentering as mediators between mindfulness training and compassionate and distressed reactions to suffering, we found that the indirect paths were significant. In other words, mindfulness could predict higher compassion, willingness to help the sufferers and personal distress but only within those who maintained attention focused on breathing.

Methods

This study was preregistered on OSF (link: https://osf.io/g42js/).

Participants

Eighty-three people participated in the study, 38 women and 43 men (age: M = 37, range: 20–54 yrs.). In order to exclude any potential outliers, we checked our variables for any data point exceeding three standard deviations from the mean. We found one outlier who performed three standard deviations below the mean on the SoVT compassion rating and consequently, we continued further analyses without him/her. The descriptive statistics for our variables are presented in Table 1.

Table 1.

Sample characteristics.

variable	group	Mean	SD	Median	Min.	Max.
SoVT(c)	experimental	7.476	1.374	7.562	4.875	9.875
	control	7.487	1.625	7.375	3.875	10
SoVT(r)	experimental	7.509	1.382	7.50	4.50	10
	control	7.362	1.772	7.625	3.250	10
SoVT(d)	experimental	6.833	1.523	6.750	3.50	10
	control	6.766	1.984	6.750	3	10
SCS	experimental	3.480	0.508	3.50	2	4.50
	control	3.410	0.51	3.333	2.167	4.667
TMS(d)	experimental	3.439	0.631	3.357	2.429	5
	control	3.212	0.642	3.143	2	5

Descriptive statistics of the manipulated variables. Note. SoVT(c) all: Socio-Affective Video Task, Compassion rating for all sufferers; SoVT(r) all: Socio-Affective Video Task, Relief rating for all sufferers; SoVT(d) all: Socio-Affective Video Task, Distress rating for all sufferers; SCS - Self-Compassion Scale; TMS(d) - Toronto Mindfulness Scale (only Decentering subscale).

Procedure

We conducted a mixed-design online experiment involving two distinct groups: the experimental group, engaged in meditation, and the control group, participating in an active control task. The within-subject design was employed to assess the impact of meditation on compassion and personal distress concerning ingroup and outgroup dynamics within individual participants.

The study was conducted via an online platform by an independent third party (a research agency), utilizing its internal recruitment database for participant enrollment. None of the study authors was directly contacted with any of the participants. The external body ensured adherence to informed consent procedures and anonymization protocols. A total of 416 adult participants, aged 18 and above, were initially recruited for the online screening study. Specific eligibility criteria were established for enrollment in the main study: (1) Only participants who had none or limited experience in mindfulness meditation (either had no experience in mindfulness meditation or have tried it only once) were eligible for the study (n = 317); (2) Participants responded on the Depression Anxiety Stress Scale⁸⁹. Those who answered within two standard deviations from the mean were accepted further (n = 121) to control for the ethical rules of conducting a study; (3) We controlled the remaining participants for the levels of empathy (measured as a feature, by Interpersonal Reactivity Scale⁷, excluding the Fantasy scale), dispositional compassion (measured with the Santa Clara Brief Compassion Scale^90,91 and in mindfulness as a trait, measured with Five Facet Mindfulness Questionnaire⁹². Once again, only participants who answered within two standard deviations from the mean on the levels of empathy, compassion and mindfulness trait were accepted further; 4) The final 83 participants completed the main study. Random assignment into two groups was ensured through a lottery procedure algorithm. These groups were tested again for their levels of empathy (trait perspective taking, personal distress and empathic concern, separately), dispositional compassion and mindfulness as a trait to ensure no significant differences that could cause any artifacts. For the study purposes— based on the literature where a similar methodology was used for studying effects of brief mindfulness meditation, a group of 30–40 participants per group was adequate^16,17,93,94.

For the main study, first participants listened to audio recording of meditation or control text. A trained mindfulness instructor recorded both scripts. In the mindfulness condition, participants were guided to focus on their breathing and be aware of their bodily sensations in a nonjudgmental manner, without explicit prosocial cues. The control condition featured a neutral narrative about a national park, matched for duration, tone and verbal complexity. Directly after the recording, they responded to the manipulation check questions. Then, they watched the SoVT videos (randomized) and rated their compassion, willingness to bring relief to the sufferers and personal distress after each video. Subsequently they responded to the two questionnaires evaluating decentration state (Toronto Mindfulness Scale⁹⁵ and self-compassion state⁹⁶, presented in a randomized order. Figure 3 provides an overview of the study procedure.

Fig. 3.

Study procedure.

The study was conducted according to the ethical standards of the Helsinki Declaration and was approved by the relevant Ethical Committee of [BLINDED for the anonymization purposes]. Participant responses underwent a rigorous process of double anonymization to safeguard their privacy and integrity rights. All participants (within the screening and the main study) confirmed their informed consent of participation in the study.

Measures

The between-group manipulation centered on two distinct conditions: the experimental group receiving meditation training and the active control group. In the mindfulness condition, participants listened to a 10-minute audio recording featuring meditation instructions based on the widely employed mindfulness meditation approach²⁰. The meditation encourages listeners to focus on themselves, their bodies, breathing, and self-acceptance, without directly referencing empathy or compassion towards others. Meanwhile, the active control group’s task was to listen to an audio recording with the very same person as in the meditation guidance reading slowly a text about Polish Magurski National Park. The reading was designed to match the experimental audio in terms of word count, tone, and voice dynamics (referencing⁹⁴. Furthermore, the two audio recordings were verified in a pilot study, showing us a high credibility of the manipulation (for more details, see the Supplementary Materials A).

We developed the mindfulness manipulation check to track the distinct effects of mindfulness versus active control on state mindfulness and to ensure the quality of the recordings. Precisely, our manipulation check comprises two questions related to the experience of mindfulness state (such as “I was conscious of my body sensations during the exercise”, evaluated on a 5-point Likert scale (1 - “does not describe my experience at all”, 5 - “describes my experience very well”) as suggested in³⁰. As the authors of this meta-analysis claim, asking these typical mindfulness state questions allows for better control over factors that influence prosociality through the meditation, compared to active controls. Next, based on⁶⁰, we asked three questions asking the participants about their ability to focus on the recording and follow the reading comfortably with the same 5-point Likert scale as with the mindfulness state questions. The manipulation check questions were not evaluated together as they focused on different aspects of participants’ experiences with the audio recordings.

The SoVT served as personal distress and compassion measurement while being exposed to visual stimuli depicting human suffering. Similarly to the original tool, we asked participants to answer questions on their emotional state using axes ranging from 0 (“no such feeling occurred”) to 10 (“I experienced this feeling a lot”) after each video. The current study follows numerous research proving the validity of video-based empathy and compassion measurement tools such as SoVT and a closely related method EmpaToM when comparing the neutral condition to affective one^{5,13,36,76,97}. We added some adjustments to the original tool. Our participants watched only eight videos with suffering humans (with no neutral condition) instead of the original 24 videos¹³. We asked our participants to rate (a) their emotional resonance with the sufferer which we call hereafter Distress rating, (b) their willingness to relieve the sufferers’ pain, called the Relief rating, and (c) their own compassion for the sufferers called the Compassion rating. These questions were also different from the original questionnaire where after each video participants were asked to rate their positive and negative feelings¹³. To avoid confusion with understanding of the word “compassion”, we encountered our participants with a brief definition of the term (“a feeling of warmth and care accompanied by a need to relieve a suffering person’s pain”), which is a common practice when using similar tools^31,97. Three factors (means) were extracted, two of them reflecting compassion rating and one aiming at measuring personal distress. Cronbach’s Alphas for the three subscales for both ingroup and outgroup sufferers were as follows (respectively): (a) α = 0.912; (b) α = 0.898; (c) α = 0.894. Our study goals required measures to be short so they immediately reflect the influence of the post-audio state of the participants. Since we aimed to verify only the emotional reaction to suffering, we found it unnecessary to use the neutral video condition.

Short Self-Compassion Scale⁸⁴ is a measure containing 6 items, such as “I feel intolerant and impatient toward myself”, with answers ranging from 1 (“Not at all true for me”) to 5 (“Very true for me”), with the last three items reversed. We calculated the mean of all the responses. Our Polish translation was investigated with a pilot study proving a tendency for a higher mean for meditating group (3.38) than active control (3.26), with n = 28. Cronbach’s Alpha showed not a satisfactory level: α = 0.125. Due to a low Alpha measure, we ran a factor analysis, receiving factor loadings visible in Table A, Supplementary Materials B. The Short Self-Compassion Scale has been challenging with its reversed items being pointed in previous studies^98,99. We observed the same issue in the contradicting factor loadings of reversed compared to non-reversed items. For the sake of the study and in order to protect overall simplicity, we decided not to use the scale in further analyses.

We used the Toronto Mindfulness Scale - State (excluding the curiosity subscale)⁹⁵ to measure decentering state. The decentering subscale contains 7 items, such as “I was aware of my thoughts and feelings without overidentifying with them” with answers ranging from 0 to 4. We calculated a mean with a Cronbach’s Alpha acceptable result: α = 0.763.

Analysis plan

We planned group comparisons and mediation testing using R Studio (version 2022.12.0). For Hypothesis 1 we used simple t-test comparisons to compare the experimental group with control. For Hypothesis 2 we used PROCESS indirect effects modeling⁶¹ for mediation analyses and the structural equation modeling (SEM) in order to verify the model suggested in the Hypotheses (see: Fig. 1). To estimate the constructed SEM models we used the maximum likelihood method, using the Lavaan package¹⁰⁰. We verified the models using SEM with some relative fix indexes (e.g., Root Mean Square Error of Approximation) and relatively sample size-independent absolute fit indexes (e.g., Tucker-Lewis Index¹⁰¹.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

The authors confirm the following contributions to the manuscript: study conception and design: GG and PH; data collection and analyses: GG; interpretation of results: GG and PH; draft manuscript preparation: GG and PH. All authors reviewed the results and approved the final version of the manuscript.

Data availibility

The data is available within the Open Science Framework under the link: https://osf.io/dvfpz.

Declarations

Competing interests

The authors declare no competing interests.

Ethics declaration

The authors declare the research was performed in accordance with the Declaration of Helsinki. The study was approved by the Ethics Comittee of National Information Processing Institute in Warsaw, Poland on 18/08/2021. All the methods were carried out by the guidelines and regulations required by Scientific Reports. The informed consent was obtained from each subject within the screening and the main study.