Stress is increasingly recognized as a problem among university students worldwide. Following the onset of the COVID-19 pandemic, the prevalence of depression among US undergraduate students increased from 19.2 to 26.3%, and anxiety increased from 23.6 to 33.9% (American College of Health Association, 2022). The pandemic elevated mental and physical health concerns, especially for students with lower incomes, minority students, and those with other marginalized status, who face higher risk for health disparities (Huckins et al., 2020; Wang et al., 2020). Marginalized and vulnerable students suffer from elevated risk of stress and poor mental health compared to others (Diaz et al., 2021). Additional factors that commonly pose challenges to US university students include academic and financial demands, novel living arrangements, health behavior adjustments, and social pressures (American College of Health Association, 2019; Lisnyj et al., 2021). University counseling centers operate with insufficient resources (Dunbar et al., 2018), and students are often ill-equipped to cope effectively during difficult times, which contributes to poor academic outcomes and subsequent disease risk (Pascoe et al., 2020). Moreover, rising academic competition and social pressures lead many students to develop perfectionistic tendencies accompanied by excessive self-criticism (Fernández-García et al., 2022) marked by guilt and feelings of failure, unworthiness, self-scrutiny, and fear of disapproval (Dinger et al., 2021). The prevalence of self-criticism in undergraduates is concerning, as it is a transdiagnostic risk factor for psychopathology including major depressive disorder (Allen, 2021; Firestone, 2021). In contrast to a self-critical and punitive approach to perceived inadequacies, self-compassion may function as a potential resilience factor against negative reactions to stress and failure.

Self-compassion is gaining attention as a protective factor against stress-related mental and physical pathology (Bluth & Neff, 2018). Self-compassion is conceptualized using three pillars: self-kindness, being kind and understanding toward oneself in instances of pain or failure rather than being self-critical; common humanity, perceiving one’s experiences as part of the larger human experience rather than seeing them as separating and isolating; and mindfulness, holding painful thoughts and feelings in balanced nonjudgemental awareness rather than over-identifying with them (Neff, 2003). It is proposed that self-compassion may ease pain and feelings of failure by allowing individuals to relate to themselves in a kind, accepting manner when faced with personal suffering. Self-compassion is a positive self-attitude that may protect against negative outcomes of self-judgment, isolation, and rumination; and thus protect against distress and mood disorders. In a meta-analysis of the association between self-compassion and suicidal ideation, Suh and Jeong (2021) found self-compassion to have a significantly negative association with suicidal thoughts and non-suicidal self-injury. Additionally, in a 2022 meta-analysis, Tran and colleagues found that self-compassion and psychological well-being mediated the relationship of mindfulness with stress, anxiety, and depression. Taken together, these meta-analyses demonstrate a robust association between self-compassion and greater psychological well-being, as well as fewer symptoms of psychopathology.

Although a growing body of literature supports the association of self-compassion with mental health outcomes, less is known about associations with physical health outcomes. Recent evidence suggests that undergraduates with greater self-compassion report fewer physical health symptoms and fewer psychosomatic complaints (Sirois, 2023). In another study, Brady (2023) found that self-compassionate people engage in more health-promoting behaviors, particularly stress management behaviors. Therefore, self-compassionate individuals experience less perceived stress and engage in more health-promoting behaviors which, in turn, promote greater overall physical health. This finding is supported by research on self-compassion that has demonstrated associations with reduced physiological stress responses and arousal (Amiri Pijakelai et al., 2023). A plausible explanation is that practicing self-compassion allows individuals to experience acceptance, kindness, and warmth, which promotes a self-soothing effect and a calmer physiological state (Gilbert, 2020). Thus, self-compassion may effectively downregulate the sympathetic nervous system and reduce the physiological stress response. Such effects have been demonstrated in intervention studies examining various forms of self-compassion training. For example, Hatton-Bowers et al. (2023) found that mindfulness and self-compassion interventions improved heart rate variability in a controlled trial among early childhood teachers. There were significant differences in emotion dysregulation between groups, such that intervention participants showed better emotion regulation and mental well-being. In sum, the research on self-compassion and physiological stress suggests that individuals who either practice or possess trait-level self-compassion may experience moderated psychological and physiological stress responses.

The hypothalamic-pituitary-adrenal (HPA) axis plays a pivotal role in the mediation of stress responses by altering the secretion patterns of its end product, cortisol (Eddy et al., 2023). Sterling and Eyer’s groundbreaking 1981 paper, “Allostasis: A New Paradigm to Explain Arousal Pathology,” introduced the concept of allostasis and highlighted the HPA axis’ regulatory function in maintaining stability through adaptive stress responses (Sterling, 2012). Under normal conditions, cortisol responses provide a frontline defense against threats to homeostasis (i.e., stress). On the other hand, chronic HPA activation and aberrations in cortisol secretion patterns have been implicated in the pathogenesis of several forms of systemic, neurodegenerative, and affective disorders (Viau, 2002). Research has significantly refined this concept. Chronic stress, characterized by frequent repetition, substantially increases vulnerability to alterations in overall cortisol levels, rhythms, loss of flexibility in responses, and altered day-to-day stability of cortisol secretion. Elevated or dysregulated cortisol poses a risk of damage, both to the brain and body systems (Degering et al., 2023). Studies consistently report high correlations between serum and salivary cortisol, indicating that salivary cortisol levels reliably estimate serum cortisol levels (Daniel et al., 2006; Eatough et al., 2009).

Diurnal cortisol secretion patterns can be measured in saliva, and are characterized by an evening nadir, gradually increasing levels in the hours prior to awakening, a 50–60% increase during the 30–45 min after awakening, and a gradual decline throughout the day with some mild postprandial elevations. Saliva has the advantage of reflecting only the unbound, active proportion of cortisol in the systemic circulation. The diurnal cortisol slope is a prognostic factor in both cancer and cardiovascular illness and is often measured using waking and evening levels assayed in saliva samples collected over numerous consecutive days (Kraemer et al., 2006). Previous studies that have estimated the reliability of diurnal salivary cortisol slope calculations have determined that at least 2–3 days (Kraemer et al., 2006), and ideally 5–8 days of cortisol collection (Segerstrom et al., 2014) are necessary for reliable estimation of the slope. Chronic stress is often marked by high cortisol levels and dysregulated daily rhythms with flattened diurnal cortisol slopes. Cortisol dysregulation predicts a spectrum of adverse mental and physical health outcomes including depression, anxiety, metabolic syndrome, obesity, fatigue, inflammatory disorders, cancer, cardiovascular disease, and elevated overall mortality (Hakamata et al., 2023). Consequently, cortisol serves not only as a marker of health disparities but also as a contributor to such disparities (Ortiz et al., 2022; Sephton et al., 2000). Recognizing the multifaceted impact of cortisol, it becomes evident that promoting university student success and well-being necessitates lifestyle choices and interventions geared towards helping students effectively cope with stress and regulate their cortisol secretion patterns.

Salivary cortisol has emerged as a crucial biomarker for stress assessment, as underscored by Hellhammer et al. (2009). Research on mindfulness interventions, particularly those emphasizing self-compassion, has explored the potential to regulate cortisol profiles, with implications for undergraduate students. Rosenkranz et al. (2016) demonstrated reduced stress and inflammatory responsiveness among experienced meditators compared to a control group. In tandem with this, Jaiswal et al. (2019) advocated for a holistic examination of anxiety and mindfulness, a perspective reinforced by Leach et al.’s study (2020. This study found that mindfulness was associated with healthier cortisol profiles and steeper diurnal slopes among stressed undergraduate students.

Examining the impact of mindfulness training on police officers, Grupe et al. (2021) reported positive outcomes, including improvements in psychological distress, mental health symptoms, and sleep quality. However, the study found no significant differences in hair cortisol, diurnal cortisol slope, and inflammatory markers when compared to a control group. Brown et al. (2012) established the role of trait mindfulness in modulating neuroendocrine and affective responses to social evaluative threats, urging further exploration of the neural pathways involved. The study by Pires et al. (2018) introduced a unique perspective, revealing that women with higher self-compassion exhibit greater activation of the precuneus—a brain region associated with self-referential processing—and lower stress levels in response to affective stimuli.

Despite these promising findings, the precise mechanisms linking self-compassion and mindfulness to physiological stress biomarkers, and the neurobiological impact of self-compassion per se, remain areas requiring continued investigation, particularly among undergraduate students whose cortisol secretion is elevated during examination periods. A study by Shunta Maeda (2022) conducted among male Japanese students at Tohoku University provides preliminary support that cortisol levels of participants with low trait self-compassion are likely to benefit from self-compassion induction after experiencing acute stress. Moreover, a study by Bluth et al. (2016) examined the buffering effect of self-compassion among adolescents and found that a high trait self-compassion group exhibited lower cortisol output during the Trier Social Stress Test, suggesting a positive association between self-compassion and a lower acute physiological stress responses. Another recent study by Ketay et al. (2023) compared the effectiveness of trait self-compassion and self-compassion training in lowering subjective stress and physiological stress responses. They found that trait self-compassion is associated with lower subjective stress and lower cortisol levels following social stress exposure, but self-compassion training did not affect the cortisol response to acute laboratory stress (Ketay et al., 2023). Therefore, research suggests that trait self-compassion may be more strongly associated with stress responses than are intervention-based changes in self-compassion. More research is needed to investigate these relationships, especially among undergraduate students.

Regarding the relationship between self-compassion and diurnal cortisol slope, recent research highlights the role of positive affect and psychological distress as potential mediators. A study by Ho et al. (2022) evaluated these associations among cancer patients, finding that the direct effect of self-compassion on diurnal cortisol slope and mean cortisol concentration was not significant. However, this longitudinal study supports an indirect association between self-compassion and steeper diurnal cortisol slope via positive affect (Ho et al., 2022). Therefore, holding a caring and positive attitude towards oneself may predict positive thinking and emotions, which might facilitate healthier neuroendocrine functioning.

Given the known relationships between cortisol and stress in undergraduates, and those between stress and self-compassion—a relationship between cortisol and self-compassion within this population is anticipated, although it was not previously found in an intervention study (Arch et al., 2014). Indeed, limited preliminary research has examined self-compassion training as it relates to cortisol secretion (Ketay et al., 2023; Torbati et al., 2020), and the findings are mixed.

The present study sought to expand our understanding of self-compassion within the context of undergraduate mental and physical health. Our work is theoretically grounded by the Mindfulness-Stress-Health model published by Salmon et al. (2012). They theorized that mindfulness has both direct and indirect effects on outcomes such as psychological well-being and stress physiology. The current study examines how self-compassion may function within a similar theoretical framework. The first aim of the study was to replicate existing evidence regarding the relationships of self-compassion with affective and emotional well-being. The second aim was to explore the relationship between self-compassion and salivary cortisol profiles collected over multiple days in a naturalistic home-based setting. We hypothesized that greater levels of trait self-compassion would be associated with greater positive affect, less negative affect, lower perceived stress, and fewer symptoms of anxiety and depression. Our second hypothesis was that students who reported greater self-compassion would also have healthier cortisol profiles, such as steeper diurnal cortisol slopes and greater day-to-day diurnal cortisol slope stability.

Method

Participants

Participants consisted of 59 undergraduates aged 18–47 (M = 21.74, SD = 3.89) recruited through an online system that draws student participants primarily from introductory psychology courses for which they typically receive course credit or extra credit. The sample consisted of predominantly non-Hispanic White (71.2%; 42 participants) female (86.4%; 50 participants) college students from a large United States Midwestern university who participated in the study during one academic semester. The university from which the sample was drawn is engaged in a corporate partnership: A substantial proportion of students receive a full tuition waiver in return for working full-time second and third-shift positions at a local package-sorting facility. Shift work is defined as shifts beginning between 8:00 p.m. and midnight and ending between 4:00 a.m. and 7:00 a.m. Among the current sample, seven students (11.86%; 7 participants) were shift workers. All participants completed written informed consent and were provided debriefing and referral resources. Demographic variables are presented in Table 1. Ten-dollar gift cards were given to participants after the intervention as an incentive to participate.

Table 1 Sample characteristics (n=59)
Full size table

Procedure

During a 1-hr laboratory visit, students provided informed consent and completed self-report questionnaires measuring psychosocial variables including perceived stress, affect, anxiety, and depressive symptoms. They received kits and instructions for 4 days of home-based salivary cortisol collection. They were also provided an iPod loaded with mindfulness audio tracks (see the “Mindfulness-Based Intervention” section). Research personnel arranged to pick up saliva collection kits and bring them back to the lab after the 4-day collection. Participants themselves returned to the lab 2 weeks after their initial visit. At this time, they provided follow-up self-report data using the same measures described above. They were required to bring their iPods so that research staff could download data on the use of the mindfulness audio tracks. After the download, students were allowed to keep the iPods loaded with mindfulness tracks. They also received a US$10 gift card for participation in the study.

Measures

Demographic Characteristics

Demographic information was collected prior to the completion of the self-report measures. Participants responded to questions regarding demographics (e.g., age, sex, ethnicity), academic year, grade point average, marital status, current living situation, and hours and shifts of paid employment. Participants were also asked questions regarding tobacco and oral contraceptive use, as these substances are known to covary with some measures of HPA-axis activity.

Self-compassion

The Self-Compassion Short-Form (SCS-SF) was used to assess trait levels of self-compassion, or compassion turned inward, especially during perceived failure or suffering (Raes et al., 2011). Items from this measure correspond with six facets of self-compassion: self-kindness, self-judgment, common humanity, isolation, mindfulness, and over-identification. Items are rated on a Likert-type scale ranging from 1 (almost never) to 5 (almost always). A summary score is calculated, with higher scores indicating greater levels of trait self-compassion. The SCS-SF is theoretically sound and is reliable and valid in student samples in both Belgium and the USA as well as in a community sample (Raes et al., 2011). In this study, the McDonald’s ω, an optimal measure of reliability (Hayes & Coutts, 2020), was 0.90.

Perceived Stress

The Perceived Stress Scale (PSS; (Cohen et al., 1991) assesses the perception of stress using a Likert scale ranging from 0 (never) to 4 (very often). A summary score is calculated by reverse coding responses to 4 positively stated items and then summing all 10 items. The PSS is a widely used and well-validated instrument. It demonstrates high internal consistency and construct validity among an undergraduate sample (Roberti et al., 2006). In this study, McDonald’s ω was 0.90.

Affect

Affect was measured using the Positive and Negative Affect Schedule (PANAS; Watson et al., 1988). The PANAS is a 20-item self-report measure of positive and negative affect. Each item is an affective state adjective (i.e., “distressed”) and is rated on a scale of 1 (very slightly or not at all) to 5 (extremely). The 10 negative items are summed to provide a negative affect summary score, and the 10 positive items are summed to provide a positive affect summary score. This instrument is widely used and has demonstrated acceptable internal consistency and validity (Crawford & Henry, 2004). In this study, McDonald’s ω was 0.89.

Anxiety Symptoms

The Beck Anxiety Inventory (BAI; Beck et al., 1988) assesses the severity of anxiety symptoms over the past week using a Likert scale ranging from 0 (not at all) to 3 (severely). A summary score is calculated, with higher scores indicating greater anxiety. The BAI is a widely used and well-validated instrument that demonstrates high internal consistency and high discriminant validity among an undergraduate sample (Creamer et al., 1995). In this study, McDonald’s ω was 0.95.

Depressive Symptoms

The Beck Depression Inventory-II (BDI-II; Beck et al., 1996) was used to assess the severity of depression symptoms over the past week using a Likert scale ranging from 0 to 3. A summary score was calculated by summing all 21 items with the total score ranging between 0 (no depressive symptoms) and 63 (more depressive symptoms). The BDI-II has been used extensively in clinical research with diverse populations. It has demonstrated high internal consistency among undergraduates (Storch et al., 2004). In this study, McDonald’s ω was 0.91.

Mindfulness-Based Intervention

Data were collected as part of an iPod-based mindfulness intervention study, with the objective of measuring effects of listening to Mindfulness-Based Stress Reduction (MBSR) tracks. The iPod-based mindfulness tracks used in this study were recorded by a clinical psychologist with several decades of mindfulness teaching and personal practice. The listening tracks were based on the MBSR program. The MBSR program was initiated by Jon Kabat-Zinn at the University of Massachusetts in 1979 and is centered around the concept of mindfulness meditation by purposefully focusing attention on the body and breath. Participants were not provided any face-to-face introduction to concepts of mindfulness but were simply given an iPod Nano device loaded with audio MBSR tracks that included this introduction as well as other didactic and meditation practice tracks. Participants had the opportunity to listen to MBSR didactic and practice tracks over a 2-week period. Didactic tracks provided instructional narratives about the foundations of mindfulness and guided mindfulness practices including meditations on eating, breathing, listening, thinking, feeling, and the body as a whole. Because of the remote nature of the intervention, no yoga component was included. Table 2 provides the titles of the audio tracks, their categorization as didactic or practice tracks, and listening times for each. Participants were asked to listen for 30 min a day, five times a week for 2 weeks. They recorded when and what they listened to in a paper log. The use of the intervention was measured objectively by data downloaded from the iPod Nano devices. Downloadable data included the title of each audio track completed and the number of times each audio track was used. Summary scores were calculated depicting the total number of minutes the participants listened to the mindfulness tracks: scores were calculated for didactic, practice, and total listening time.

Table 2 Listening tracks provided in the iPod-based mindfulness intervention. Tracks were adapted from MBSR. Notation depicts coding as didactic or meditation practice tracks, and minutes in length.
Full size table

Salivary Cortisol Collection

Four days of saliva samples were collected after the completion of baseline psychosocial data collection; during the first 4 days, participants had access to the iPod-based mindfulness intervention. Although this design has multiple inherent limitations (absence of a control group, collection of cortisol data during, rather than before/after the intervention), it was chosen for three reasons: (1) to accommodate the limited budget available from the undergraduate research grant that funded the study, (2) to favor reliable estimation of diurnal cortisol slope using four consecutive days of saliva collection (Kraemer et al., 2006; i.e., rather than 2 days pre- and 2 days post-intervention), and (3) to leverage the use of iPod objective intervention use data with concurrent physiological data collection.

HPA-axis activity was analyzed using eight salivary cortisol samples per participant: Subjects collected samples immediately upon waking and at bedtime for four consecutive days. We deemed 4 days adequate for diurnal slope estimation given that the current study assessed cortisol in a primarily young sample of undergraduate college students, among whom we did not expect cortisol to be influenced by age-related diseases that affect HPA axis rhythmicity (Otte et al., 2005). Participants used salivette devices with cotton swabs. Samples were refrigerated until their return to the lab. Cortisol samples were processed and frozen at −80°C until assay using kits and reagents from Salimetrics (Salimetrics, Inc., State College, PA). All assays were completed by trained technicians in the senior investigator’s lab. The assay sensitivity was 0.007 μg/dL. Inter-assay CVs were 7.01% (low control) and 5.66% (high control). Intra-assay CVs were 5.47% (low control) and 3.98% (high control). Waking samples were excluded for saliva collections that occurred more than 15 min following the reported waking time (n = 9). Due to data exhibiting asymmetry towards higher values, raw cortisol values were log-transformed in an attempt to correct for typical positive skew and to reduce the impact of extreme values to make the distribution more symmetrical. The diurnal cortisol slope was calculated using the unstandardized beta of natural log-transformed cortisol regressed on collection time (Kraemer et al., 2006; Sephton et al., 2000). Cortisol slope stability was calculated using the mean absolute value of the difference between pairs of slopes over the 4-day collection period. Waking and bedtime cortisol means were calculated using the log-transformed waking and bedtime samples. Due to dysregulated diurnal cortisol profiles expected (and observed here), shift workers (n = 7) were excluded from the analyses that incorporated cortisol measures.

Data Analyses

Based on the recommendation of Kraemer and Blasey (2004), ordinal predictor variables were mean-centered, and binary predictor variables were effects-coded to reduce the potential risk of multicollinearity and statistical inference errors. Due to the proportionally low number of ethnically diverse students in the sample, ethnicity was dichotomized into majority (non-Hispanic White) and minority status, which included students who endorsed being Asian, African American, Hispanic, Native American, or “other.” Use of the intervention was assessed by the total number of minutes participants listened to the mindfulness tracks, and separately by total minutes recorded use of didactic versus practice tracks. Bivariate correlations tested the effects of intervention use on slopes of change from pre- to post-intervention for self-compassion, perceived stress, affect, anxiety, and depressive symptoms. Separate hierarchical regressions further evaluated intervention use on follow-up levels of self-reported perceived stress, mood, anxiety, and depressive symptoms while adjusting for baseline levels of each of these variables.

Preliminary bivariate correlations were run to examine associations between self-compassion and potential empirically derived control variables including grade point average, hours of paid employment, living situation, sex, minority status, and shift-work status. Bivariate correlations and multiple regressions were used to test associations between baseline trait self-compassion, psychosocial model factors, diurnal cortisol slope, diurnal mean, 4-day diurnal cortisol slope stability, and intervention use. To examine the associations between trait self-compassion and HPA-axis activity, hierarchical regressions were used. Separate hierarchical regressions tested associations of intervention use with concurrent HPA axis activity. For all tests involving cortisol, we entered theoretically derived control variables in the first block, including tobacco use (i.e., cigarettes, cigars, and smokeless tobacco) and oral contraceptives (Allen et al., 2019). SPSS software version 29 was used for data analysis.

Results

The demographic features of the sample are provided in Table 1. Statistical descriptions of self-report metrics are provided in Table 2. All scales used in this study demonstrated adequate reliability, indicated by McDonald’s ω values higher than 0.70 (Dunn et al., 2014). Objective data downloaded from the iPods revealed that 18 of the 59 subjects (>30% of the sample) did not use the iPod MBSR tracks at all during the 2-week intervention period. Among those who used the iPod, the mean total listening time over the intervention was M = 138.2 min, SD = 184.9 min. Time spent listening to didactic tracks was M = 86.7 min, SD = 103.9 min. Time spent on practice audio tracks was M = 51.6 min, SD = 91.6 min.

Bivariate correlations revealed no effect of intervention use on slopes of change from pre- to post-intervention for self-compassion, perceived stress, affect, anxiety, or depressive symptoms. Assumptions of regressions were met for all analyses. Subsequent multiple hierarchical regressions evaluating the effects of intervention use on self-reported perceived stress, mood, anxiety, and depressive symptoms produced no significant effects.

There were no significant associations of baseline self-compassion with any of the demographic variables tested. However, preliminary analyses comparing shift-workers to non-shift-workers showed significant differences in self-compassion, represented in Figure 1. Shift-workers demonstrated significantly less self-compassion (M = 31.28) than non-shift-working students (M = 36.71; U = 94.5, p < 0.05). Further, shift workers demonstrated a pattern of dysregulated cortisol profiles. For these reasons, shift workers (n = 7) were excluded from the analyses investigating associations between self-compassion and HPA axis activity.

Fig. 1
figure 1

Group difference in self-compassion. Note: Regression analyses show that non-shift workers had higher levels of self-compassion compared with night shift workers. Shift workers demonstrated significantly less self-compassion (M = 31.28) than non-shift-working students (M = 36.71; U = 94.5, p < 0.05)

Full size image

Regression analyses demonstrated that individuals with higher baseline trait self-compassion had significantly lower perceived stress (β = −0.57, p < 0.001, R2 = 0.34), lower negative affect (β = −0.46, p < 0.001, R2 = 0.26), greater positive affect (β = 0.48, p < 0.001, R2 = 0.24), less anxiety (β = −0.35, p < 0.01, R2 = 0.16), and lower depressive symptoms (β = −0.51, p < 0.001, R2 = 0.26; Table 3 and Figure 2). Regression analyses also showed that undergraduates with greater self-compassion had steeper diurnal cortisol slopes (β = −0.43, p < 0.01, R2 = 0.26) and greater cortisol slope stability (β = −0.42, p < 0.05, R2 = 0.20; Tables 4 and 5 and Figure 3). Trait self-compassion at baseline was not significantly correlated with use of the intervention (practice listening time: r = −0.17; didactic listening time: r = −0.12; total listening time: r = −0.17, all p > 0.05). However, use of the intervention was associated with cortisol slope stability, for both total listening time (r = 0.41, p < 0.01) and meditation practice time (r = 0.40, p < 0.05). In contrast, the time spent listening to didactic tracks was not significantly associated with cortisol slope stability. No measure of intervention use was significantly associated with diurnal cortisol slope.

Table 3 Self-report questionnaire descriptive statistics (n = 59)
Full size table
Table 4 Regression analyses of baseline self-compassion on psychosocial measures
Full size table
Table 5 Regression analyses of baseline self-compassion on diurnal cortisol slope and mean cortisol slope stability
Full size table
Fig. 2
figure 2

Scatter plots depicting associations of self-compassion with measures of affect, perceived stress, anxiety, depressive symptoms, and cortisol slope. The R and p -values shown on each plot represent the results of regression analyses

Full size image
Fig. 3
figure 3

Mean and 95% CI waking and evening salivary cortisol, shown for descriptive purposes using a median split on self-compassion scores. For descriptive, not statistical, purposes, raw cortisol values are depicted in the figure. Log-transformed values were used in statistical tests. Error bars indicate 95% confidence interval

Full size image

Discussion

Prior research has found that trait levels of self-compassion predicted depressive symptoms and emotional resilience in student samples. Those with higher levels of self-compassion had fewer symptoms of depression (Cai et al., 2023). In another study with undergraduate participants, self-compassion was shown to predict depression, anxiety, stress, and well-being (Soysa & Wilcomb, 2015). Luo et al. (2018) studied the impact of self-compassion among undergraduate students in China and found that individuals with high trait self-compassion exhibited less negative affect in response to stress. Likewise, Guan et al. (2021) studied the impact of self-compassion among Chinese individuals during the COVID-19 lockdown and found that a self-compassion intervention alleviated negative affect. Furthermore, Lou et al. (2022) provided a meta-analysis that pointed out the negative correlations between self-compassion and depression, anxiety, and stress. Consistent with these findings, our study in an undergraduate student sample concurs, showing that self-compassion is associated with perceived stress, affect, anxiety, and depressive symptoms. As hypothesized, higher self-compassion was associated with decreased perceived stress, greater positive affect, lower negative affect, and fewer symptoms of anxiety and depression. Neff (2003) suggests that self-compassion allows one to acknowledge one’s challenges with care, support, and openness rather than resisting or attempting to change the situation. In this way, self-compassion may function as an adaptive emotion regulation strategy that uses positive cognitive reappraisals and acceptance (Munroe et al., 2022), which are associated with increased positive emotions (Troy et al., 2018) and decreased perceived stress (Doorley et al., 2022). Therefore, future studies of undergraduate students could test potential moderators of the relationship between self-compassion and psychological well-being, such as age, gender, minority status, academic stress, work, and financial demands. Future research could also explore the potential moderating role of self-compassion in the relationship between undergraduates’ affective/emotional well-being and neuroendocrine functioning.

We found that self-compassion levels varied based on shift worker status, and no other group differences were detected. We note that the subsample of shift workers is small. Students in the night shift work-for-tuition program may be most financially and socially at risk to begin with, further reducing their ability to navigate and cope with the stress of having to balance full-time education with circadian and sleep disruption. It may be helpful to view this finding in light of Kurebayashi’s (2020) study among Japanese workers, which found those on a night shift tend to have sleep disturbances predicted by self-coldness (the negative aspect of self-compassion). Furthermore, Kemper et al. (2015) investigated the relationship between self-compassion and sleep disturbances among healthcare professionals: controlling for stress rendered the relationship between self-compassion and sleep disturbances nonsignificant. It is possible that stress caused by socioeconomic status, or by shift work itself, could contribute to lower self-compassion (Grosser et al., 2022). Further research is needed to clarify the relationships between night shift work, economic and job stress, and self-compassion.

Previous research has examined the relationship between self-compassion training and salivary cortisol outcomes, with mixed findings. Arch et al. (2014) found that after undergoing a brief loving-kindness meditation, which is akin to self-compassion training, participants demonstrated a reduced stress response marked by salivary alpha-amylase and parasympathetic responses following a stressful task, but there was no main effect on salivary cortisol. A separate study conducted by Ketay et al. (2023) demonstrated a reduction in salivary cortisol responses to a stress task among participants with higher trait self-compassion. Together, these findings point to the need for more multimodal research to determine the mechanisms by which self-compassion exerts health benefits and/or buffers the effects of psychological stress responses on downstream physiology. Although the current study’s mindfulness intervention did not produce a significant effect on any of the key psychosocial variables, the data contributes to science by providing evidence for a cross-sectional association between trait-level self-compassion, psychosocial variables, and two important measures of HPA-axis activity: diurnal cortisol slope and cortisol slope stability. Both variables are known predictors of physical health. This suggests a potential protective role of self-compassion in day-to-day HPA-axis regulation.

Undergraduates who reported greater use of the mindfulness intervention, as measured by meditation practice listening time and total listening time, had more stable cortisol slopes across the 4-day collection period. It should be noted that this association was not explained by increases in self-compassion during the intervention. A host of other factors could explain the association between meditation practice and daily cortisol regulation, such as the imposition of regularity into a student’s schedule, the beneficial effects of meditation on sleep, and the reduction of depressive or anxious symptoms. In contrast, it is important to note that the circadian function of the HPA axis as measured by the diurnal cortisol slope is unlikely to substantially change during a short (2-week) intervention and is more likely to be modified only over periods of months (Adam, 2012; Smyth et al., 2020). Other studies have shown that even after relatively short mindfulness interventions (e.g., 5-min practice, four times within 2 weeks), psychosocial benefits can accrue quickly and are manifested in measures of trait mindfulness (Strohmaier et al., 2020). The measures of practice listening time, didactic listening time, and total listening time were not associated with self-compassion here; therefore, the use of the intervention did not explain the relationship between self-compassion and cortisol. It is recommended that future research explore the underlying mechanisms of mindfulness interventions in altering stress hormone patterns. This could be done by considering psychosocial constructs such as perceived stress or lifestyle factors and extending the duration and/or intensity of the mindfulness program.

Regression analyses showed that although both mindfulness intervention use and trait self-compassion were associated with greater slope stability, only trait self-compassion was associated with steeper diurnal slope. This finding contrasts with results from a recent meta-analysis (Phillips & Hine, 2021) which found only a small correlation between self-compassion and stress hormones. Studies suggest that self-compassion may influence cortisol patterns through positive affect. For example, Ho et al. (2022) found that the direct effect of self-compassion on diurnal cortisol slopes was not significant among cancer survivors. However, an indirect association was found via positive affect 6 months later. The indirect linkage was non-existent via emotional distress. The authors point out that the emotional distress level of the sample was in the non-clinical range (Ho et al., 2022). Herriot et al. (2018) found that the association between self-compassion and diurnal cortisol slopes of older adults was statistically significant only when their stressors were more chronic and intractable, compared with daily stressors. Zimmaro et al. (2016) suggest that perceived stress is directly associated with diurnal cortisol mean. The current finding that trait self-compassion was associated with a healthier, steeper diurnal slope may be explained by the context of college-related stressors characterized by high academic demands, social pressures, financial worries, and future uncertainties, all of which contribute to high levels of stress when they are excessive and prolonged.

Overall, these findings provide evidence that trait self-compassion shares associations with positive affective, psychological, and physiological outcomes. Undergraduates who relate kindly to themselves may have some protection against negative emotional reactions (i.e., negative affect, anxiety, and depression) as well as physiological reactions to high academic stress.

Our findings on the mental health benefits of this brief iPod-based intervention contrast with those from other studies that have found eHealth interventions beneficial for participants’ psychological well-being, among both undergraduate students (Walsh et al., 2019) and people with medical conditions (Mikolasek et al., 2018).

Limitations and Future Research

One limitation of this study is the homogenous sample. The current study consisted of predominantly non-Hispanic White female undergraduate students. Replication is warranted in a more diverse undergraduate sample. The use of more diverse sampling could provide insight into different underlying responses in times of failure or suffering across different populations.

We examined associations of trait self-compassion with cortisol levels. As noted in the methods section, cortisol was collected during a mindfulness intervention, which is not a self-compassion intervention per se but might be expected to enhance self-compassion. Shiftwork was not an intended focus of the study. However, because we measured diurnal cortisol, we deemed it important to remove shift workers from cortisol analyses due to the known effects of sleep, light at night, and circadian activity rhythms on cortisol secretion patterns.

The association between trait self-compassion and healthier HPA-axis functioning in a naturalistic setting may capture the experience of day-to-day stress among undergraduates. However, the nonsignificant effect of listening to mindfulness tracks on self-compassion could be explained by the brevity of the intervention: possibly, it was not of sufficient duration or intensity to have a significant impact. A more extensive intervention may be recommended for future research. Moreover, over 30% of the sample did not listen to the MBSR iPod tracks at all, possibly making it difficult to detect any intervention effect.

Although some studies suggest measurement periods longer than 4 days are ideal for diurnal salivary cortisol, the ability to examine cortisol concurrently during the intervention was prioritized in this study. To properly examine intervention effects on cortisol, it would have been much preferred to collect 4 days of saliva samples prior to the provision of the intervention and again after the conclusion. Unfortunately, financial resources did not permit a more extensive collection. No placebo effects were accounted for in this study, as previous research has established no placebo effect, nor any stress effect, of saliva sampling (Giacomello et al., 2020).

A strength of the study was utilizing wake-time measures in the cleaning of cortisol data, which ensures a more accurate waking measurement, as many subjects exhibit a 50–60% increase in cortisol 30–45 min after waking. However, it is recommended that future research utilize objective sleep measurement devices, such as wrist actigraphy, to increase the accuracy of wake-time cortisol sampling while still allowing data collection in a home-based naturalistic environment.

Findings from the current study shed light on the extent to which the trait self-compassion protects undergraduates from the psychological and physiological manifestations of stress. As demand for mental health services increases on college campuses, additional research is warranted in exploring factors, such as self-compassion, that may offset the rising academic and social stressors of undergraduate education. Furthermore, additional attention and resources should be devoted to shift-working students, who reported significantly less self-compassion, and therefore may be more vulnerable to negative responses to stress. The interesting finding of a significant correlation between the measures of trait self-compassion and shift work merit future research that would benefit from a larger sample size. Additionally, the association between shift worker status and lower trait self-compassion raises the question of whether there are effects of circadian disruption and/or sleep deprivation on self-compassion.

Literature suggests that subscales within the global self-compassion framework may help explore people’s relationship with their suffering, such as the ways they respond emotionally, understand cognitively, and pay attention to, their suffering (Neff et al., 2019). Recent examination of the factor structure of the self-compassion scale supported the use of a total score and scores from the six subscale domains (i.e., self-kindness, self-judgment, common humanity, isolation, mindfulness, and over-identification) representing the overall self-compassion framework and its constituent parts. Although the current study utilized the SCS-SF, the examination of the subscales is less reliable with the short form (Raes et al., 2011). Examining the validity and reliability of the six self-compassion subscales using the long form would further refine tools that researchers need to explore how self-compassion impacts well-being. Future investigations targeting corollary measurements of neuroendocrine stress, such as cellular aging/telomere length, are encouraged as they will broaden the discussion of stress-health implications among the college undergraduate population.

A growing body of research demonstrates that self-compassion is linked with steeper, healthier, and more stable cortisol slopes and lower cortisol responses to stress (Herriot et al., 2018; Ketay et al., 2023; Maeda, 2022). Our research further highlights a need among undergraduates for healthier coping strategies given the negative implications of stress on both psychological and physiological outcomes. Given the rising academic, financial, and social demands faced by undergraduate students, more research investigating emotional regulation strategies such as self-compassion is warranted. Such strategies could buffer against stress and its associated deleterious psychological and physiological consequences. Further, the insights gained from the current research, in addition to the growing research on the benefits of self-compassion, warrant programmatic research on the efficacy of self-compassion training on college campuses.