Numerous studies have focused on the ways to cultivate self-compassion (SC). Most of the SC interventions were typically lengthy and did not capture momentary fluctuations in SC after each exercise in the intervention. Therefore, we explored the impact of repeated self-compassionate writing as an ecological momentary intervention (EMI).
Method
Seventy-five participants completed a 3-week study with signals sent twice daily, resulting in 1,980 observations collected. During the 1-week control period, participants were asked to report their experiences of negative events, momentary SC, and momentary stress. In the subsequent 1-week intervention period, they engaged in SC writing whenever they experienced negative events. Trait SC and weekly stress were also evaluated before and after the intervention, including a 1-week follow-up period.
Results
Multilevel modeling showed a slight significant increase in momentary SC and a reduction in momentary stress. However, the slope did not differ significantly between the intervention and control periods. Moreover, there were no significant changes in trait SC nor in weekly stress, although the weekly stress levels showed a decreasing trend with a small effect size from pre to post intervention.
Conclusions
This study provided preliminary evidence of the feasibility and utility of EMI in cultivating SC using repeated self-compassionate writing. Although the effect sizes were small, the study highlighted possible approaches to increase the effectiveness of the intervention.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Self-compassion (SC) is the awareness of pain or suffering that is present within an individual accompanied by the intention and desire to alleviate that pain (Neff, 2003). One of the key concepts of SC involves responding in the same understanding and supportive way one would take with a best friend when one is struggling. Three central aspects of self-compassion include: (1) self-kindness, which refers to treating oneself with compassion instead of being judgmental in difficult situations; (2) common humanity, which is the understanding that other human beings have similar experiences and that one is not alone in the pain one may feel at any given moment; and (3) mindfulness, the ability to observe and notice one’s difficult situation without overidentifying and overanalyzing. These important attributes of self-compassion are in contrast to self-judgment, isolation, and overidentification, respectively.
Since it has been reported that high SC is associated with greater well-being (Zessin et al., 2015) and that low SC is associated with psychopathology (Brown et al., 2021; MacBeth & Gumley, 2012; Suh & Jeong, 2021), numerous studies have placed much emphasis on understanding how to cultivate self-compassionate attitudes. Systematic interventions, such as Compassion-Focused Therapy (Gilbert, 2010) and Mindful Self-Compassion (Germer & Neff, 2019), have been developed and applied in several contexts. Typically, the frequency and length of SC interventions is weekly for 2–8 weeks, with each session lasting several hours (in the case of Mindful Self-Compassion, 2.5-hr sessions weekly for 8 weeks). Recent meta-analyses of SC interventions (Ferrari et al., 2019; Han & Kim, 2023) revealed that the effects were significantly large for eating behavior and rumination and moderate for trait SC, stress, depression, and other variables. Kirby et al. (2017) also conducted a meta-analysis of 21 randomized controlled trials, including longer-term interventions, and demonstrated medium-to-large effect sizes for decreasing several symptoms (e.g., depression, anxiety, and psychological distress).
Although these interventions produced impactful changes afterwards, most studies did not capture momentary fluctuations in SC and other symptoms after each exercise. Taking such micro perspectives enables researchers to understand the underlying mechanisms of why and how SC interventions work to improve mental health. To capture such momentary changes, there are two important approaches to consider, as follows.
The first approach involves short and frequent measurements in daily life. This sampling method is often called Ecological Momentary Assessment (EMA) or experience sampling. In SC research, several studies have attempted such a momentary assessment to measure within-person-level SC (Beuchat et al., 2022; Bicaker et al., 2022; Mey et al., 2023; Thøgersen-Ntoumani et al., 2017, 2021). Recently, a reliable and valid scale to capture the momentary level of SC has been developed and translated (Chishima et al., 2022; Miyagawa et al., 2021; Neff et al., 2021); using the scale is recommended for researchers (Neff, 2023).
The second approach to focusing on momentary changes involves short and frequent interventions. Recently, some SC studies have examined frequent (mainly daily) interventions (Li et al., 2021; Ong & Sündermann, 2022; Rodgers et al., 2018; Schnepper et al., 2020; Swee et al., 2023; Thabrew et al., 2022). For instance, Li et al. (2021) investigated the effect of daily SC interventions for 2 weeks using online guided meditation. They found that daily SC and stress did not show large changes in the 7-day control period, started to improve with time during the 14-day intervention period, and remained stable during the 7-day follow-up period. Similarly, Swee et al. (2023) conducted a 16-day SC writing intervention for high shame individuals and demonstrated its effectiveness on global shame, self-criticism, and anxiety.
With recent improvements in mobile technology, much attention has been paid to interventions that occur more frequently than once a day, called ecological momentary interventions (EMIs), which allow participants to receive brief interventions when they need them (real-time) in natural settings (real-world) (Heron & Smyth, 2010). Literature on EMIs targeting not only problematic behaviors (e.g., smoking) but also mental health (e.g., depression) has been growing (Gee et al., 2016; Marciniak et al., 2020; Versluis et al., 2016). For example, a meta-analysis of 27 studies with EMIs revealed a medium effect size for improving depression, anxiety, and stress (g = 0.57; Versluis et al., 2016). Although these interventions are typically conducted with additional support from mental health professionals, it has also been reported that standalone interventions without professional support can still improve mental health outcomes, such as depression (g = 0.45; Versluis et al., 2016). A review of smartphone standalone EMIs (Marciniak et al., 2020) reported that most of them successfully reduced mental health symptoms and that over 70% of participants rated them as helpful and satisfying.
So far, one EMI including SC called “EMIcompass” has been reported (Paetzold et al., 2022; Reininghaus et al., 2023). The research team incorporated the EMI into the usual face-to-face treatment for young patients with mental health problems. The EMIcompass comprised 6 weeks, combining four sessions of daily and momentary training through a smartphone app. To assess the participants’ real-time stress and conduct momentary interventions, EMA questionnaires (3 days a week and 6 signals a day) were administered. If the participants reported high stress or negative affect, interactive tasks involving compassion-focused techniques (breathing and soothing exercises) were offered. The results showed evidence that the EMIcompass increased the momentary SC and reduced the stress reactivity. However, to our knowledge, no studies to date have examined the effects of EMI in the form of self-compassionate letter writing.
The current study aimed to examine the effect of EMI using self-compassionate writing. Since the writing had to be done repeatedly during the intervention period, we created a short version of the writing task based on the previously existing one (Neff et al., 2021). The current study examined momentary and trait levels of SC and stress as the primary intervention outcomes. We used a within-participant design with two periods: control and intervention, based on a previous study (Li et al., 2021). We also compared the trait SC and weekly stress in the pre-survey (1 day before the intervention) with both post (1 day after the intervention) and follow-up (1 week after the intervention) surveys.
We hypothesized that momentary SC would be higher during the intervention period than during the control period (Hypothesis 1a), and that momentary stress would be lower (Hypothesis 1b). Additionally, we assumed that the slopes of the variables would be steeper in the intervention period than in the control period (Hypotheses 2a for SC and 2b for stress), indicating that the longer the participants experienced the intervention, the more the effects were confirmed. These hypotheses were consistent with those of previous studies (Li et al., 2021). When focusing on trait-level changes, we postulated that trait SC would significantly increase both from pre- to post-surveys and from pre- to follow-up surveys (Hypothesis 3a). Weekly stress was predicted to decrease in the same manner (Hypothesis 3b).
Method
Participants
Our sample size was determined based on the research on daily SC writing interventions (Li et al., 2021; Swee et al., 2023), demonstrating small to medium changes in SC and stress in sample sizes of 60–70. Participants (n = 78) from several universities in Japan were recruited via email from the authors and other researchers using a registration URL link. Participants were undergraduate and graduate students who are native speakers of Japanese. Individuals who responded only to the pre-survey (n = 2) and those who did not experience any negative events during the intervention period (n = 1) were excluded. The final sample consisted of 75 participants (Mage = 19.93 years, SD = 1.52, range = 18–27 years; 69.3% were women). Two participants did not respond to the follow-up survey alone; therefore, we omitted them when using repeated-measures ANOVA. The post hoc power analyses for repeated ANOVA were conducted using G*Power, suggesting that 73 participants were well-powered (99% for trait SC, 100% for weekly stress) to detect the obtained small effect sizes (ηp2 = 0.017, ηp2 = 0.038).
Procedure
All procedures were conducted online in November and December 2021. When recruiting participants, we used the cover research title “Research on Individual Feelings About Everyday Events,” to avoid any speculation on the true purpose of our study.
As shown in Fig. 1, three periods were set: control, intervention, and waiting periods, which continued for 7 days for each period. Orientation and pre-surveys were conducted every day at 6 p.m. between November 15 and 21, before the control period. The participants were allowed to participate whenever they wanted. During the orientation session, the experimenters explained the entire study process and the participants consented to participate. They were instructed to register with an official LINE (the most popular mobile messenger app in Japan) group in our study through their smartphones so that they could receive signals for an intensive survey that started the day after the orientation. Post and follow-up surveys were conducted after the intervention and waiting periods, respectively. The purpose of this study was shown on the last page of the follow-up survey. Each participant received 2,000 JPY (approximately 20 USD) as compensation after completing the follow-up survey.
Fig. 1
A survey flow and measurement
×
For both the control and intervention surveys, we adopted time-based sampling (Shiffman et al., 2008). Signals for the intensive survey were sent to the participants’ LINE app twice daily. The total number of signals across the two periods was 28. To send the signals, a tool for experience sampling called “exkuma” was used. Although the LINE app can be utilized on a PC or tablet, we asked the participants to respond to every signal on their smartphone as much as possible, so that they could respond anytime and anywhere. Two signals per day were sent from 9 am to 3 pm and from 3 pm to 9 pm at approximately 5-hr intervals. If the participants did not respond to the signals, reminder signals were sent one hour after the regular signals. Four hours after the regular signals were sent, the URL links automatically became invalid, indicating that the participants could not respond to the two signals simultaneously.
As shown in Fig. 1, the content of the questionnaire differed between the control and intervention periods. During the control period, we asked the participants whether they experienced any positive or negative events within one hour (Tables S1 and S2 in the Supplementary Information). Although experiencing any negative event was the main target of our study, we also included a question about positive events because having participants imagine only their unpleasant events each time from the beginning of the survey could be stressful, possibly leading to worsening mental health as well as decreasing response rates. The instruction for negative events was as follows: “Between one hour ago and now, did you have any difficult or unpleasant events? If so, how difficult or unpleasant were they for you?” We used a 6-point Likert scale ranging from 0 (I had no such event) to 5 (I had an extremely difficult or unpleasant event). Ten examples were shown as examples of negative events (e.g., “I didn’t receive a reply from my friends or romantic partner.”) to simplify this imagination. When calculating whether they experienced any negative events, we used a binary score that aggregated one or more negative events into one. After completion, the participants were instructed to rate their momentary SC and perceived stress levels.
However, during the intervention period, a branch was introduced to the question on negative events. If participants experienced any negative event within one hour, we asked them to write a self-compassionate message toward themselves using the following instruction: “Please write words of support, encouragement, and kindness to yourself that would be helpful to hear right now. This doesn’t need to be long. If you are not sure what to say, imagine what you would say to a close friend struggling in a similar difficult situation. Try to validate your experience with an attitude of acceptance and non-judgment. You can also remind yourself that you are not the only one to experience this kind of thing.” The full instruction prompts are presented in Table S3. This instruction was developed based on an existing prompt to induce SC state (Neff et al., 2021). We carefully included three SC components in the instructions: self-kindness, common humanity, and mindfulness. We also presented example messages to make the writing easier. If participants had no negative events, we asked them to write about their recent activity using the following instruction: “What have you been doing for the past hour?” It is important to include this writing section to maintain balance when there are no negative events. If this writing section had not been included, some participants might have pretended that they did not experience any negative events to avoid writing.
Measures (Pre, Post, and Follow-Up Survey)
Trait Self-Compassion
The short form of the Japanese Trait SC Scale (Arimitsu et al., 2016; Raes et al., 2011) was used. This scale comprised 12 items selected from a full-item scale. Table S4 in the Supplementary Information presents the English and Japanese items. In the current study, the reliability coefficients were α = 0.842/0.885/0.903 and ω = 0.847/0.890/0.905 in the pre/post/follow-up tests. These items are rated on a 5-point Likert scale ranging from 1 (almost never) to 5 (almost always).
Weekly Stress
Weekly stress was assessed using seven items from the stress subscale of the Japanese Depression Anxiety Stress Scale (Lovibond & Lovibond, 1995). Table S5 in the Supplementary Information presents the English and Japanese items. Participants were asked to indicate the extent to which the items had been applied over the past week. In the current study, the reliability coefficients were α = 0.866/0.829/0.835 and ω = 0.868/0.833/0.839 in the pre/post/follow-up tests. These items were rated on a 4-point Likert scale ranging from 0 (did not apply to me at all) to 3 (applied to me very much or most of the time).
Measures (Intensive Survey)
Momentary Self-Compassion
We used the 6-item form of the momentary SC scale (Chishima et al., 2022; Neff et al., 2021). The reliability and validity of the scale were confirmed by our primary supplemental study (see Supplemental Materials). Table S6 displays the English and Japanese items. Participants answered on this scale even when they did not experience negative events. The reliability scores for within- and between-levels were acceptable (ωwithin = 0.657, ωbetween = 0.750). The intra-class correlation coefficient (ICC) was 0.56.
Momentary Stress
We assessed momentary perceived stress using one item: “I am feeling stressed right now.” This item is rated on a 5-point Likert scale ranging from 1 (not at all) to 5 (extremely). ICC was 0.30.
Data Analyses
To evaluate the momentary effect of the intervention, multilevel modeling was applied given that the data were nested by each participant (Lischetzke et al., 2015), and the ICCs for both momentary SC and stress were sufficiently high. Thus, it is reasonable to conduct analyses to distinguish the variances of momentary variables to which each participant repeatedly responded (within-level) from the variances that differed between participants (between-level). Statistical details and equations for the multilevel modeling are shown in the Supplemental Materials. To examine the long-term effects of the intervention, we adopted repeated measures ANOVAs, setting two outcomes (trait SC and weekly stress) and one predictor of survey timing (pre, post, and follow-up surveys). P-values for the multiple comparisons in the follow-up analyses were adjusted using Shaffer’s method.
For the analyses, we used R (version 4.3.0) and “anovakun” (version 487). We utilized several packages: For example, “lme4” and “multilevelTools” for multilevel model, “psych” for reliability score calculation, “effsize” for effect size calculation, “mice” for data imputation, and “apaTables”, “ggplot2,” and “flexplot” for data visualization.
Results
The participants received signals twice daily for 14 days, excluding reminder signals, theoretically resulting in 2,100 observations. Owing to missing data, the final data comprised 1,980 observations, indicating that each participant responded to 26.4 out of the 28 signals, on average. The rate of negative events per signal was found to be 49.7%. The average length of the SC messages during the intervention period was 20.02 Japanese characters (approximately 10–15 words in English). An example of the messages is: “I understand that you have so much to think about, it’s exhausting. Just think about your class assignment for now. You will be fine.”
Descriptive statistics and correlations with 95% confidence intervals (CIs) are presented in Table 1. We used the person mean (average of 28 measurements) for momentary SC and stress, except when calculating between- and within-correlations. The correlations among pre, post, and follow-up were high (r = 0.81–0.84 for trait SC; r = 0.60–0.78 for weekly stress), although the means of weekly stress remained low over the time (Mpre = 0.86, Mpost = 0.72 at post survey, and Mfollow−up = 0.74), as all were below the midpoint of the scale (1.5). The corresponding correlations between the momentary- and trait-level variables in the pre-test survey were medium to high (r = 0.64 [0.49, 0.76] for SC; r = 0.40 [0.19, 0.57] for stress). The between- and within-correlations between the momentary SC and stress were rbetween = −0.69 [−0.79, −0.55] and rwithin = −0.57 [−0.60, −0.54].
Table 1
Means, standard deviations, and correlations with confidence intervals
Variable
M
SD
1
2
3
4
5
6
7
1. Trait SC (pre)
2.76
0.73
2. Trait SC (post)
2.8
0.80
0.81**
[0.72, 0.88]
3. Trait SC (follow-up)
2.73
0.81
0.82**
[0.73, 0.88]
0.84**
[0.76, 0.90]
4. Weekly Stress (pre)
0.86
0.65
–0.46**
[–0.62,–0.25]
–0.37**
[–0.55,–0.16]
–0.32**
[–0.51,–0.10]
5. Weekly Stress (post)
0.72
0.58
–0.43**
[–0.60,–0.22]
–0.38**
[–0.56,–0.17]
–0.42**
[–0.60,–0.22]
0.71**
[0.57,0.81]
6. Weekly Stress (follow-up
0.74
0.57
–0.39**
[–0.57,–0.17]
–0.35**
[–0.54,–0.13]
–0.43**
[–0.60,–0.22]
0.60**
[0.44,0.73]
0.78**
[0.68,0.86]
7. Momentary SC (person mean)
3.43
0.53
0.64**
[0.49,0.76]
0.55**
[0.37,0.69]
0.63**
[0.46,0.75]
–0.36**
[–0.54,–0.15]
–0.50**
[-0.66,–0.31]
-0.37**
[-0.56,-0.16]
8. Momentary Stress (person mean)
2.66
0.76
–0.53**
[–0.67,–0.34]
–0.42**
[–0.59,–0.21]
–0.44**
[–0.61,–0.24]
0.40**
[0.19,0.57]
0.49**
[0.29,0.64]
0.39**
[0.18,0.57]
-0.69**
[-0.80,-0.55]
Note. SC = self-compassion. M and SD are used to represent mean and standard deviation, respectively. Values in square brackets indicate the 95% confidence interval for each correlation. * indicates p < 0.05. ** indicates p < 0.01.
Analyses of Momentary Outcomes
Considering missing data, alternative analyses with imputed data were conducted (2,100 observations for each momentary outcome). We used multiple imputation methods for the multilevel models by combining 10 imputed datasets from 1,000 simulations. Given that these results did not show large differences from the original results, we decided to continue using the original data (Table S7 in the Supplementary Information).
The momentary SC and stress for each participant and the signal time are shown in Figures S1 and S2 (in the Supplementary Information), respectively. The results of the multilevel model are presented in Table 2; Fig. 2. In terms of momentary SC, the main effect of the period was significant (γ10 = 0.05, p = 0.002), revealing that momentary SC was higher in the intervention period than in the control period. However, the main effect of time and the interaction were not significant (γ20 = 0.01, p = 0.105; γ30 = 0.00, p = 0.253, respectively). For additional analyses to explore more detailed intervention effects, we focused on each of the six components of the SC (Tables S8, S9, and S10; Figures S3, S4, and S5, in the Supplementary Information). The main effects of the period were significant for momentary self-kindness, momentary self-judgment, and momentary common humanity (γ10 = 0.08, p < 0.001; γ10 = −0.07, p = 0.008; γ10 = 0.10, p = 0.001, respectively). Regarding momentary stress, the main effects of the period and time were significant (γ10 = −0.08, p = 0.016; γ20 = −0.02, p = 0.004, respectively), whereas the interaction was not significant (γ30 = 0.01, p = 0.211). These results indicated that momentary stress declined with time in each period and was lower during the intervention period than during the control period.
Table 2
Multilevel analyses on momentary self-compassion and momentary stress
Momentary SC
Momentary Stress
Estimates
p
Standardized Estimates
Estimates
p
Standardized Estimates
Intercept (γ00)
3.43 (3.31 – 3.55)
<0.001
0.00 (–0.17 – 0.17)
2.65 (2.48 – 2.82)
<0.001
0.00 ( –0.13 – 0.13)
Period (γ10)
0.05 (0.02 – 0.08)
0.002
0.07 (0.03 – 0.11)
–0.08 (–0.14 – –0.01)
0.016
–0.06 ( –0.11 – –0.01)
Time (γ20)
0.01 (–0.00 – 0.01)
0.105
0.03 (–0.00 – 0.06)
–0.02 (–0.04 – -0.01)
0.004
–0.07 ( –0.11 – –0.02)
Period * Time (γ30)
0.00 (–0.01 – 0.00)
0.253
–0.03 (–0.07 – 0.02)
0.01 (–0.01 – 0.03)
0.211
0.03 ( –0.02 – 0.09)
Random Effects
Residual Variance(σ2)
0.18
1.07
Intercept Variance (τ00)
0.28
0.53
Time Variance(τ11)
0.01
0.04
Intercept Variance(τ22)
0.00
0.00
Period * Time Variance (τ33)
0.00
0.00
Marginal R2/ Conditional R2
0.006 / 0.625
0.009 / 0.379
Note. SC = self-compassion. Bold p values indicate statistically significant.
Fig. 2
Momentary self-compassion and momentary stress at each signal by period. Note. Panels A and B show the result of momentary self-compassion and momentary stress, respectively. Black lines indicate mean scores for each signal. Blue lines indicate linear regression by period. Gray areas indicate the 95% confidence bands by period
×
As part of our exploratory analyses, following Muris and Otgaar’s (2022) recommendation, we examined a bipolar continuum of SC. This approach divides the total SC score into two components: compassionate self-responding (CS) and uncompassionate self-responding (UCS). CS is calculated by summing self-kindness, common humanity, and mindfulness scores, while UCS is derived from the sum of self-judgment, isolation, and overidentification scores. For CS, we found a significant main effect of period (γ10 = 0.06, p = 0.003), indicating that the positive aspect of momentary SC was higher during the intervention period compared to the control period. However, neither the main effect of time nor the interaction effect were significant (γ20 = 0.00, p = 0.576; γ30 = 0.00, p = 0.836, respectively). This pattern was consistent with our findings for the total SC score. Regarding USC, we observed no significant main effects for period or time, and no significant interaction effect (Table S11; Figure S6; in the Supplementary Information).
Analyses of Trait-Level Outcomes
A repeated measures ANOVA examining the trait-level change over the three periods showed that trait SC did not significantly change (F(2, 72) = 1.22, p = 0.298, ηp2 = 0.017). Similarly, weekly stress did not significantly decrease (F(2, 72) = 2.81, p = 0.063, ηp2 = 0.038), as shown in Fig. 3. As an exploratory analysis, we examined the simple main effects between each period. The results showed that all differences were not significant (t(72) = 2.27, p = 0.078, d = 0.21 for the pre- and post-surveys; t(72) = 1.37, p = 0.174, d = 0.14 for the pre- and follow-up surveys; and t(72) = 0.94, p = 0.352, d = 0.07 for the post- and follow-up surveys), although a decreasing trend with a small effect size between the pre- and post-surveys was observed. Additional analyses using a trait-level CS and UCS showed that both CS and UCS did not significantly change as shown in Figure S7 (F(2, 72) = 2.13, p = 0.123, ηp2 = 0.029 for CS; F(2, 72) = 1.49, p = 0.229, ηp2 = 0.020 for UCS).
Fig. 3
Trait self-compassion and weekly stress at pre, post, and follow-up surveys. Note. Panel A shows the results for trait self-compassion, whereas Panel B shows the results for weekly stress. Error bars indicate 95% confidence intervals and black lines indicate mean-level changes. Yellow points indicate scores by each participant
×
Discussion
This intervention study aimed to examine the effect of repeating short self-compassionate writing on momentary- and trait-level SC and stress. Overall, the results demonstrated that momentary SC and stress were improved compared to those before the intervention, indicating that the momentary intervention was effective in reducing momentary stress.
We hypothesized that, compared to the control period, momentary SC would be higher (Hypothesis 1a) and momentary stress would be lower (Hypothesis 1b) in the intervention period. Although both hypotheses were supported because the effects of the period were significant, which is consistent with previous intensive intervention studies (e.g., Li et al., 2021), the magnitude of the coefficients was small, indicating that writing SC messages only slightly contributed to improving the SC mind state and reducing perceived stress each time. One of the reasons for this result may be that participants’ engagement for our intervention was weak. For example, the number of written characters was very short (an average of 20.02 Japanese characters each time). We allowed the participants to write short messages so that writing twice a day would not become a burden that would increase their momentary stress and lead to dropouts from the intervention. However, a previous study that obtained a large effect size for increasing momentary SC asked participants to write a minimum of 300 Japanese characters in total (Chishima et al., 2022). Therefore, setting a minimum number (e.g., 100) of characters/words may be required to obtain a larger effect.
In terms of the interactions between period and time, no significant effects were observed; therefore, Hypotheses 2a and 2b were not supported. Although these results were not as predicted, some previous studies have also reported that the accumulation effect is not prevalent, especially when the intervention time is short. Swee et al. (2023) examined the effect of 16-day SC writing and found that daily SC and depression did not increase linearly. Participants experienced different negative events of different severities. Therefore, they were influenced by several negative events and needed to write the SC message in different ways each time. Variations in daily negative events may lead to fewer accumulated effects. The finding that SC and stress had a slight overall improvement by the momentary intervention, but the effects did not accumulate linearly, reinforces the importance of focusing on slopes through a multilevel analysis design.
Although the interactions were not significant, the slopes in the control period appeared steeper than those in the intervention period, especially for momentary stress. Similar results were also confirmed in a previous study on daily SC intervention (Li et al., 2021), which showed that daily emotional exhaustion dropped substantially during the 7-day control period. There are two possible explanations for this finding. First, the number of holidays differs between the two periods. November 23 was a Japanese national holiday and came into the control period for most participants, as the control period started from November 16 to 22. Additionally, a university where a large number of participants had irregular holidays on November 29 and 30 due to entrance exams, resulting in four consecutive holidays from November 27. Experience or expectation of holidays could influence a decrease in stress. However, the start timing and university belonging substantially differed among participants; therefore, it is difficult to determine whether holidays are a critical factor.
The second interpretation is that the task of repeatedly reporting the stress level acted as self-monitoring, resulting in successful stress control. Some researchers perceive ecological momentary assessment in itself as a useful self-monitoring tool to maintain/improve mental health. Indeed, some studies have indicated that self-monitoring through momentary assessment has a positive impact on mental health (Bos et al., 2020; Folkersma et al., 2021; Melbye et al., 2020; Piot et al., 2022; Widdershoven et al., 2019). Although our participants could not look at the trajectory of their stress levels, they may have remembered their past responses, which could have helped them identify stressors and develop strategies to manage them effectively.
Focusing on trait-level changes, SC and weekly stress did not change after the intervention period, which does not support Hypotheses 3a and 3b. A possible reason for not fostering trait-level outcomes is the short intervention duration (1 week), given that similar results were reported in the some previous studies (Johnson & O’Brien, 2013; Swee et al., 2023), showing that repeated SC writing during 1 or 2 weeks was not sufficient to increase trait-level SC. Trait SC, compared to momentary SC, is a more stable component of how people typically react after experiencing a painful situation, which is more difficult to modify. A meta-analysis of the effect of SC intervention on self-criticism (Wakelin et al., 2022), which included a very short intervention (e.g., a single session), reported that the shorter the intervention, the weaker the impact on inhibiting self-criticism. Even when focusing on standalone smartphone EMIs for mental health, the median intervention duration was 30 days (Marciniak et al., 2020). Therefore, a larger amount of SC writing, longer intervention duration, or different types of interventions may be required to change trait-level SC and stress.
Limitations and Future Directions for Research
To our knowledge, this study is the first attempt to examine EMI only for SC. Even though the obtained effects were not large, our intervention produced insightful information for future research, shedding light on advanced methods of SC interventions. In addition, we illustrated momentary changes through the ecological assessment, which is a relatively new approach in this field. Online interventions using smartphone applications have attracted a great deal of attention. However, most applications incorporate tasks to improve mental health without evidence-based assessments (Kertz et al., 2017; Larsen et al., 2019). Therefore, an increased accumulation of scientific evidence is required.
The current study has several limitations that deserve mentioning. First, as mentioned above, the effect sizes of the intervention were small at both the momentary and trait levels; thus, the methodology of momentary intervention has the potential for refinement. Even though increasing the number of words in each SC writing may be a promising approach, it is not hard to imagine that participants face difficulty in coming up with long SC messages immediately after experiencing a negative event. Thus, combining other short SC exercises may be another method for improvement (Dong et al., 2022; Paetzold et al., 2022; Reininghaus et al., 2023). For example, reading compassionate messages from others, engaging in short meditations, and conducting soothing touch exercises could be effective momentary interventions. This combination would be beneficial specifically for those with difficulty in writing SC messages. Not only the tasks during the intervention, but also the participants’ readiness for the intervention may lead to limited effects. For example, we intentionally avoided explaining what SC is or why it is important before initiating the intervention, to detect the pure effect of SC writing. However, it is possible that participants’ readiness to undertake our intervention was insufficient. Therefore, in future studies, it would be worthwhile to provide psychoeducation on self-compassion prior to the intervention.
Second, a control condition was not included because of the limited number of students available online during the pandemic. If we could set the control condition that continued to receive the control period signals after the control period, it would enable us to compare the effects of the intervention directly with those who did not write SC messages.
Third, although we sent two random signals per day from 9 am to 9 pm, future researchers may need to increase the number of signals and expand the time span to include morning and night signals to detect real-time negative events more accurately. Combining time-based sampling with event-based sampling, in which participants self-report when they experienced negative events, would work better. Otherwise, passive sensors, such as a smartwatch that measures heart rate variability, can detect participants’ stressful situations without them having to report them.
Fourth, the sample comprises only students in some Japanese universities. The intervention may have had different effects on different age groups, socioeconomic status, and cultures. Although there is value in focusing on non-Western samples, since most SC interventions have been conducted in Western countries (Ferrari et al., 2019), the cross-cultural adaptability of interventions and their downstream effects should be examined. Previous studies have demonstrated that self-criticism in Japan is not substantially harmful (Yamaguchi et al., 2014), and that SC in Japan has a weaker association with positive affect than in the U.S. (Arimitsu et al., 2019). Therefore, intervention effectiveness may be more substantial in Western cultures than in Japan. Additionally, most participants in our study did not report severe stress levels. Future studies should focus on those who are extremely stressed or those who have depressive symptoms.
Fifth, the intervention study was conducted during a pandemic that was stressful for most students (November–December 2021). For example, they are forced to attend online lectures and communicate with friends online. Therefore, future research should examine how the same intervention plays a role during non-pandemic periods.
Finally, we acknowledge that the lack of preregistration for this study is a serious limitation. In future research, it is essential not only to address the aforementioned limitations but also to preregister all procedures and hypotheses. This will be crucial in examining whether the results obtained in this study are robust and can be replicated.
Declarations
Conflict of interest
No conflicts of interest were declared.
Ethics Approval
The Research Ethics Committee of the University of Tsukuba approved all studies.
Informed Consent
We explained the entire study process, and the participants consented to participate before the data collection began.
Use of Artificial Intelligence
ChatGPT and Claude were used for grammatical corrections.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
share
DELEN
Deel dit onderdeel of sectie (kopieer de link)
Optie A:
Klik op de rechtermuisknop op de link en selecteer de optie “linkadres kopiëren”
Met BSL Psychologie Totaal blijf je als professional steeds op de hoogte van de nieuwste ontwikkelingen binnen jouw vak. Met het online abonnement heb je toegang tot een groot aantal boeken, protocollen, vaktijdschriften en e-learnings op het gebied van psychologie en psychiatrie. Zo kun je op je gemak en wanneer het jou het beste uitkomt verdiepen in jouw vakgebied.