The Short Sussex-Oxford Compassion Scales for self, to others, and from others (S-SOCS-S; S-SOCS-O; S-SOCS-FO): Development and Validation of 10-item Forms
- Open Access
- 08-04-2026
- RESEARCH
Abstract
Delen
Compassion is increasingly receiving clinical, educational and scientific attention. Compassion relates to how we can respond to suffering, including the awareness and recognition of suffering, understanding the universality of suffering, feeling for the person suffering, tolerating uncomfortable feelings and motivation to act/acting to alleviate suffering (Strauss et al., 2016). Compassion is widely recognised as an innate human capacity and core virtue (Gu et al., 2020). Hereby, we can distinguish compassion toward oneself, towards others, and from others towards ourselves (Gilbert et al., 2017), with most research focusing on the benefits of self-compassion. Research in the past decades has shown that higher compassion is related to greater well-being, happiness, and lower depression and anxiety. There is also evidence that compassion can be cultivated through training and compassion interventions (Martins et al., 2025).
To examine compassion, we need a precise measurement tool that is reliable, valid and feasible to implement (Gu et al., 2020). Several questionnaires have been developed that aim to measure compassion, grounded on different definitions and conceptual models. For instance, Neff (2003a) focused on self-compassion, conceptualized as an adaptive self-to-self response to suffering, having self-kindness (versus harsh self-criticism), common humanity (versus isolation and separation), and mindfulness (versus over-identification). Empirical research has consistently shown that self-compassion is associated with greater psychological wellbeing and resilience, as well as lower levels of depression, anxiety, and stress (Germer & Neff, 2013; Neff, 2003b). Although, initially developed as a self-related construct, further research has demonstrated conceptual and empirical links between self-compassion and compassion for others, suggesting partially shared underlying mechanisms, such as emotion empathy or prosocial motivation (Neff & Pommier, 2013). Such contributions informed models of compassion that conceptualise it as a multidimensional construct with focus on compassion towards others, or from others (Gilbert et al., 2017; Hwang et al., 2008; Pommier et al., 2020).
Strauss et al. (2016) reviewed existing scales and concluded that the available compassion measures had important limitations, including a lack of consensus on definition, only assessing some but not the full range of facets of compassion, and psychometric weak measures (e.g., low internal consistency, poor factor structure). To address these gaps, (Gu et al., 2020), developed the Sussex-Oxford Compassion Scales (SOCS). The SOCS are two parallel 20-items scales: the SOCS-O for compassion for others, and the SOCS-S for self-compassion. Each SOCS scale is built upon the five-element model of compassion and was validated across large samples. Resulting SOCS-S/O scales demonstrated a stable five-factor structure with general hierarchical compassion factor and good reliability. Recently, building upon the five-element model of compassion, the SOCS-FO was developed to measure compassion from others towards oneself (Halamová et al., 2026).
While the parallel 20-item SOCS-S, SOCS-O, and SOCS-FO questionnaires represent strong advantages, their lengths can be impractical in different settings and populations. Measuring all three constructs of compassion would require participants to answer 60 items, which may be burdensome for participants, especially when the scales are part of a longer questionnaire. Additionally, due to similarity in the content of the parallel items, this might increase respondents’ fatigue. These disadvantages may limit the feasibility of the scales in time-constraint and longitudinal research (e.g., during clinical screenings or intervention studies). While some compassion-related scales were successfully shortened to brief formats – for example, Raes et al. (2011) adapted the 26 items of Self-Compassion Scale to a 12 items scale – it has not been investigated whether it is feasible to shorten the Sussex-Oxford Compassion Scales.
Creating short forms of scales can be done using more traditional methods that pick items based on individual statistics (e.g., item correlations, factor loadings), or using more modern item-response and algorithmic approaches (e.g., Mokken scaling analysis; Van der Ark, 2012). To shorten the SCS, Raes et al. (2011) selected items that had high correlations with their subscale and high correlations with the full SCS, and then retained items based on theoretical domain coverage. One of the most recommended approaches for short scale creation is Ant Colony Optimization (ACO; Raborn & Leite, 2018) – a metaheuristic inspired by how ant colonies find optimal paths. This method was for instance successfully used to create a 15-item version of the Big Five Inventory-2 (Olaru & Danner, 2021). The advantage of ACO is that, unlike conventional methods, it evaluates candidate subsets on overall criteria, such as model fit, measurement invariance, and reliability. Karl et al. (2024) mention that traditional short-form construction often relies on subjective cutoff choices and may overlook the interdependencies of items, subject to researcher bias. In contrast, ACO balances exploration of new items and their exploitation, converging on a solution that is globally optimal across multiple criteria. In practice, it results in ACO-generated short forms showing more accuracy and reliability, while retaining theoretical coverage.
Using ACO, this study aimed to create three brief, parallel short forms of the SOCS-S (self-compassion), SOCS-O (compassion for others), and SOCS-FO (compassion from others) scales. Our goals were threefold. First, we aimed to retain the conceptual five-factor structure and strong reliability of the original scales, while substantially reducing item count to ten (two items for each of the five subscales) across the three SOCS scales. Second, we sought to preserve a second-order general compassion factor, consistent with the theoretical model of compassion (Gu et al., 2020). Finally, we explored whether there was a solution in which the three Short Sussex-Oxford Compassion Scales for self, others, and from others (S-SOCS-S, S-SOCS-O, S-SOCS-FO) would preserve parallel structure with the same items, as in the original SOCS structure (i.e. with only minor differences in the item wording for the three measures: self, others, or from others). Hereby, the overall aim of this research was to create an efficient, short measurement tool of compassion that fits the latest compassion theory, ensuring both empirical rigour and practical applicability.
Method
Participants
The final sample comprised 2009 helping professionals, including healthcare providers (e.g., physicians, nurses, paramedics), mental health specialists (e.g., psychologists, counselors), educators (e.g., teachers, speech therapists), community support workers (e.g., volunteers, clergy), and others. The sample consisted of 77% women, 23% men, and less than 1% identifying as another gender. Age of participants ranged from 18 to 76 years (Full sample n = 2009: M = 42.01, SD = 11.41; subsample n1 = 1005: M = 42.4, SD = 14.05; subsample n2 = 1004: M = 41.91, SD = 13.98). Professionals reported an average of 15 years of practice (SD = 10.51) and a mean weekly client contact duration of 20 h (SD = 12.4).
Procedure
Helping professionals including healthcare providers, educators, and community support workers, were recruited through professional association networks, personal contacts, online social media outreach, and snowball sampling, with participants directed to a survey in REDCap. Electronic informed consent was obtained prior to participation.
Measures
Sussex-Oxford Compassion Scales (SOCS)
The Sussex-Oxford Compassion Scales (SOCS) assess compassion in three contexts: compassion for others (SOCS-O), self-compassion (SOCS-S) (Gu et al., 2020), and compassion experienced from others (SOCS-FO), developed by Halamová et al. (2026). Each version is a 20-item self-report measure based on Strauss et al. (2016) five-element definition of compassion. Items cover five compassion facets: recognizing suffering, understanding its universality, feeling for the sufferer, tolerating discomfort, and motivation to alleviate suffering and are rated on a 5-point Likert scale (1 = "not at all true" to 5 = "always true"). Example items: SOCS-S ("I recognise signs of suffering in myself"), SOCS-O ("I notice when others are feeling distressed"), and SOCS-FO ("Other people are sensitive to my distress"). The reliability of the scales in the current sample, measured by McDonald’s ω, was as follows: SOCS-O ω = 0.94, SOCS-S ω = 0.89 and SOCS-FO ω = 0.96.
The Compassion Fatigue Self-Test (CFST)
CFST is a self-report questionnaire originally developed by Figley and Stamm (1996) to measure compassion fatigue. Revised scale used for our study has 66 items, which is comprised of compassion fatigue, burnout, and compassion satisfaction facets. Respondents rate how often they experience statements about their helping role on a Likert scale (e.g. from 0 = "not at all" to 5 = "very often"). An example item includes "I feel estranged from others" (compassion satisfaction). The CFST reliability for the total scale in this sample was ω = 0.78.
The Generalized Anxiety Disorder Scale (GAD-7)
GAD-7 is a brief 7-item measure of anxiety symptom severity (Spitzer et al., 2006). Each item (e.g. "Trouble relaxing") asks how often the symptom was present over the last two weeks, rated 0 ("not at all") to 3 ("nearly every day"). Total scores range 0–21, with higher scores indicating more severe anxiety. The GAD-7 reliability in this sample was ω = 0.86.
The Patient Health Questionnaire (PHQ-9)
PHQ-9 is a 9-item self-report measure of depressive symptoms (Kroenke et al., 2001). It includes one item for each DSM-IV criterion for depression, scored 0 ("not at all") to 3 ("nearly every day") based on the past two weeks. An example item would be "Poor appetite or overeating". Higher scores indicate more severe depression. The PHQ-9 reliability in our sample was ω = 0.85.
The Forms of Self-Criticising/Attacking & Self-Reassuring Scale (FSCRS)
FSCRS is a 22-item instrument measuring self-criticism and self-reassurance (Gilbert et al., 2004). It has three subscales: inadequate self (reflecting feelings of personal inadequacy), hated self (reflecting self-directed anger), and reassured self (reflecting self-reassurance). An example item for reassured self subscale would be "I still like being me". Items are rated from 0 ("not at all like me") to 4 ("extremely like me"). The FSCRS scale reliability in our sample was ω = 0.77.
Data Analyses
We used the ShortForm package in R for the initial item selection (Raborn & Leite, 2018). ShortForm uses Ant Colony Optimization (ACO), a probabilistic technique inspired by the behaviour of ants that deposit pheromones to mark efficient paths. In the context of the short form scale development, this algorithm simulates multiple "ants" selecting item subsets depositing virtual pheromones to mark more successful paths. These subsets are evaluated based on pre-specified model fit, such as CFI, RMSEA, TLI. Better-fitting solutions reinforce item selection probabilities, and the algorithm continues iterating until it stabilizes on a solution repeatedly selected by different "ants". Two decimal places were used to report most of the statistics; three decimal places were used to report the goodness-of-fit statistics (e.g., CFI, RMSEA), standardised and unstandardised model coefficients.
As ACO finds the optimal model within the specified sample, we divided our sample (n = 2009) randomly into two equally sized subsamples (n1 = 1005, n2 = 1004) to avoid sample-specific overfitting solution. The item selection was carried out in n1 using ACO, with the goal of identifying solutions that worked across three SOCS scales (model 1). Then, in the first phase, we evaluated solutions manually to find a candidate item set that retained consistent and similar parallel structure across all SOCS scales (goal 3; model 2). We then validated model 2 solution in sample n2 using confirmatory factor analysis (CFA).
Based on the configuration options in the ShortForm package manual, we set the ACO algorithm to run with 100 ants per iteration, allowing the model to evaluate a wide range of item combinations in each step. The process was set to stop after 40 consecutive iterations without change in the best solution (step = 40), so the search will continue only while some meaningful improvements were being made. We also limited the total number of possible runs to 1000 (max.run = 1000), which is a safeguard against excessive computation time in case the convergence is slow. Such set parameters are in line with the default priors provided in the package documentation and the authors’ recommended flexibility (Raborn & Leite, 2018). The pheromone evaporation rate was set to .5 to balance exploration of new item combinations with reinforcement of high-performing solutions.
We specified ACO algorithm identifying the first-order model with five subscales and two items in each and the fit criterion as following: CFI > 0.95 & TLI > 0.95 & RMSEA < 0.08. For interpretation of the model fit results for the final S-SOCS structures, we followed similar guidelines as Gu et al. (2020), applying both more liberal and conservative thresholds. Model fit was considered acceptable if the Comparative Fit Index (CFI) and Bentler-Bonett Normed Fit Index (NFI) were at or above 0.90 (liberal criterion) or 0.95 (conservative). We interpreted the Root Mean Square Error of Approximation (RMSEA) as indicating acceptable fit when ≤ 0.10 (liberal) or ≤ 0.06 (conservative), and the Standardized Root Mean Square Residual (SRMR) as acceptable when < 0.10 (liberal) or < 0.05 (conservative) (Gu et al., 2020).
For the internal consistency of the total combined scores across three scales (i.e., with 10 items in each scale) we considered Cronbach’s alpha (α) values ≥ 0.70 to be acceptable and ≥ 0.80 to represent strong reliability. At the same time, we acknowledge alpha's methodological limitations for shorter scales, especially with two items (Eisinga et al., 2013). While in the first place we evaluated total alpha, due to sensitivity of alpha level to the number of items, we considered also alpha level ≥ 0.50 as acceptable for the subscales with two items in each, and < 0.50 as unsatisfactory (Taber, 2018).
Results
Item Selection
Results of ACO item selection for each scale with following first-order model parameters (CFI > 0.95 & TLI > 0.95 & RMSEA < 0.08) and most similar items for a parallel structure across the three SOCS versions, are presented in Table 1. For the facets Feeling and Recognizing, we identified two parallel items for all three SOCS scales (R1, R6; F3, F8). For the Acting facet, we identified two parallel items for two of the three SOCS scales (S-SOCS-O and S-SOCS-FO: A15, A20). Model 1 converged for all scales, showed excellent to acceptable fit, and strong reliability.
Table 1
ACO-derived S-SOCS-S, S-SOCS-O, and S-SOCS-FO items (model 1; sample 1)
Subscale | S-SOCS-O | α | S-SOCS-S | α | S-SOCS-FO | α
| |||
|---|---|---|---|---|---|---|---|---|---|
R | socs_o_R_1 | socs_o_R_6 | 0,77 | socs_s_R_1 | socs_s_R_6 | 0,66 | socs fo_R_1 | socs_fo_R_6 | 0,72 |
U | socs_o_U_7 | socs_o_U_2 | 0,78 | socs_s_U_12 | socs_s_U_17 | 0,75 | socs_fo_U_7 | socs_fo_U_12 | 0,73 |
F | socs_o_F_3 | socs_o_F_8 | 0,59 | socs_s_F_3 | socs_s_F_8 | 0,64 | socs_fo_F_3 | socs_fo_F_8 | 0,67 |
T | socs_o_T_4 | socs_o_T_14 | 0,51 | socs_s_T_9, | socs_s_T_19 | 0,67 | socs_fo_T_14 | socs_fo_T_19 | 0,62 |
A | socs_o_A_15 | socs_o_A_20 | 0,72 | socs_s_A_10 | socs_s_A_20 | 0,74 | socs_fo_A_15 | socs_fo_A_20 | 0,74 |
Total | 0.84 | 0.80 | 0.88 | ||||||
Model fit | 5 factors | ||||||||
CFI = 0.972, TLI = 0.950, RMSEA = 0.059, SRMR = 0.026 | CFI = 0.977, TLI = 0.959, RMSEA = 0.050. SRMR = 0.024 | CFI = 0.978, TLI = 0.960, RMSEA = 0.059, SRMR = 0.021 | |||||||
Second-order | |||||||||
CFI = 0.960, TLI = 0.939, RMSEA = 0.065, SRMR = 0.036 | CFI = 0.940, TLI = 0.911, RMSEA = 0.075, SRMR = 0.058 | CFI = 0.973, TLI = 0.959, RMSEA = 0.060, SRMR = 0.025 | |||||||
While ACO-derived item sets showed strong and statistically significant first-order factor loadings across all three S-SOCS scales, the second-order CFA models based strictly on the ACO-selected items showed at least one non-significant loading on the higher-order compassion factor in each scale (See Online Resource 1). For S-SOCS-S, it was the Acting to alleviate suffering subscale, and for S-SOCS-O/FO scales – the Tolerating uncomfortable feelings. The ACO algorithm optimised item selection with respect to first-order model fit and did not evaluate higher-order structural coherence. Thus, the ACO-derived solutions supported a five-factor representation of compassion, which motivated refinement of item selection to obtain a model compatible also with a second-order structure.
Given the similarity of the two items selected by ACO in model 1, we proceeded testing models with the same sets of parallel items across all three S-SOCS scales for the facets Understanding, Tolerating, and Acting. We iterated this process until a parallel solution with the same items was found across all three S-SOCS scales that accounted for both the first-order and second-order models (model 2). For instance, the Acting subscale had the same items for S-SOCS-O/FO (A15; A20), but not for S-SOCS-S (A10; A20). We evaluated if interchanging A10 and A15 across scales would show similar output. Results indicated a good fit and model convergence.
Following the identification of highly similar items by the ACO algorithm in initial models, we balanced statistical criteria also with theoretical domain coverage and factor loadings evaluation. While ACO frequently selected semantically similar items (e.g., SOCS-O R1: “I recognise when other people are feeling distressed without them having to tell me” together with R6: “I notice when others are feeling distressed”), we prioritised items that minimised content overlap to better capture the breadth of the compassion construct, additionally evaluating factor loadings and comparing them with the reported loadings from original research.
Thus, based on ACO output, theoretical evaluation of item content, and factor loadings, we selected items like R11 “I’m quick to notice early signs of distress in others” (SOCS-O) over R1 “I recognise when other people are feeling distressed without them having to tell me” to pair with R6 for the Recognising facet. Similar theoretical considerations guided choices for such items as U2: “I understand that everyone experiences suffering at some point in their lives” (SOCS-O) and A15: “When I see someone in need, I try to do what’s best for them” (SOCS-O), resulting in the final structure: R = ~ R11 + R6; U = ~ U2 + U7; F = ~ F3 + F8; T = ~ T14 + T19; A = ~ A15 + A20. By selecting items from distinct semantic clusters chosen by ACO algorithm within subscales, we aimed to preserve conceptual breadth of original scales, while maintaining their psychometric performance.
We then checked if this solution converged across three S-SOCS scales both for five factors and second-order factors, with results showing good to acceptable fit (Table 2) as well as strong to acceptable Cronbach’s alpha. We proceeded testing this model on a validation sample n2 with full results described in the Tables 2 and 3.
Table 2
Fit Indices for S-SOCS-S, S-SOCS-O, and S-SOCS-FO scales (final model 2; samples n1 & n2)
Scale | Sample | Model | CFI | NFI | TLI | RMSEA [90% CI] | SRMR |
|---|---|---|---|---|---|---|---|
S-SOCS-O | n1 | 5-factor | 0.973 | 0.964 | 0.951 | 0.056 [0.045–0.068] | 0.025 |
Second-order | 0.953 | 0.943 | 0.929 | 0.067 [0.057–0.077] | 0.037 | ||
n2 | 5-factor | 0.978 | 0.970 | 0.961 | 0.049 [0.038–0.061] | 0.026 | |
Second-order | 0.966 | 0.956 | 0.949 | 0.056 [0.046–0.067] | 0.034 | ||
S-SOCS-S | n1 | 5-factor | 0.964 | 0.955 | 0.936 | 0.058 [0.047–0.069] | 0.031 |
Second-order | 0.930 | 0.919 | 0.895 | 0.074 [0.064–0.084] | 0.053 | ||
n2 | 5-factor | 0.965 | 0.955 | 0.938 | 0.057 [0.046–0.068] | 0.029 | |
Second-order | 0.915 | 0.904 | 0.872 | 0.081 [0.072–0.091] | 0.055 | ||
S-SOCS-FO | n1 | 5-factor | 0.974 | 0.968 | 0.953 | 0.064 [0.053–0.075] | 0.022 |
Second-order | 0.968 | 0.961 | 0.953 | 0.064 [0.054–0.074] | 0.026 | ||
n2 | 5-factor | 0.981 | 0.976 | 0.966 | 0.057 [0.046–0.068] | 0.023 | |
Second-order | 0.970 | 0.963 | 0.955 | 0.066 [0.056–0.076] | 0.031 |
Table 3
Final items and reliability of the S-SOCS-S, S-SOCS-O, and S-SOCS-FO (model 2; validation sample n2)
S-SOCS-O | S-SOCS-S | S-SOCS-FO
| ||||||
|---|---|---|---|---|---|---|---|---|
α | ω | α | ω | α | ω | |||
Subscale | Item | |||||||
R | R_11 | R_6 | 0.71 | 0.71 | 0.65 | 0.65 | 0.77 | 0.77 |
U | U_2 | U_7 | 0.71 | 0.71 | 0.56 | 0.58 | 0.69 | 0.69 |
F | F_3 | F_8 | 0.60 | 0.59 | 0.69 | 0.69 | 0.69 | 0.69 |
T | T_19 | T_14 | 0.54 | 0.54 | 0.54 | 0.55 | 0.71 | 0.71 |
A | A_15 | A_20 | 0.74 | 0.75 | 0.71 | 0.72 | 0.78 | 0.78 |
Total | 0.83 | 0.87 | 0.79 | 0.85 | 0.89 | 0.92 | ||
Convergent and Divergent Validity
To assess convergent and divergent validity we examined correlations between the S-SOCS scales and other measurements (Table 4). All three S-SOCS scales showed significant associations with measures of anxiety (GAD-7), depression (PHQ-9), self-criticism/reassurance (FSCRS), and compassion satisfaction/fatigue and burnout (CFST).
Table 4
Correlation coefficients between S-SOCS-S, S-SOCS-O, and S-SOCS-FO and other constructs, including full SOCS scales, in validation sample n2
GAD-7 | PHQ-9 | FSCRS-IS | FSCRS-RS | FSCRS-HS | CFST-CS | CFST-B | CFST-CF | SOCS-S | SOCS-O | SOCS-FO | |
|---|---|---|---|---|---|---|---|---|---|---|---|
S-SOCS-S | −0.38*** | −0.42*** | −0.45*** | 0.61*** | −0.47*** | 0.55*** | −0.38*** | −0.33*** | 0.96*** | 0.40*** | 0.34*** |
Recognising suffering | −0.17*** | −0.20*** | −0.22*** | 0.33*** | −0.32*** | 0.32*** | −0.20*** | −0.20*** | 0.90*** | 0.37*** | 0.14*** |
Understanding the universality of suffering | −0.14*** | −0.19*** | −0.11*** | 0.27*** | −0.27*** | 0.34*** | −0.16*** | −0.21*** | 0.90*** | 0.63*** | 0.27*** |
Feeling for the person suffering | −0.27*** | −0.30*** | −0.35*** | 0.49*** | −0.32*** | 0.39*** | −0.28*** | −0.20*** | 0.89*** | 0.15*** | 0.31*** |
Tolerating uncomfortable feelings | −0.33*** | −0.35*** | −0.41*** | 0.44*** | −0.30*** | 0.39*** | −0.29*** | −0.25*** | 0.89*** | 0.24*** | 0.29*** |
Acting to alleviate suffering | −0.36*** | −0.39*** | −0.42*** | 0.51*** | −0.40*** | 0.42*** | −0.34*** | −0.28*** | 0.91*** | 0.23*** | 0.27*** |
S-SOCS-O | −0.10*** | −0.18*** | −0.11*** | 0.32*** | −0.22*** | 0.48*** | −0.16*** | −0.16*** | 0.44*** | 0.96*** | 0.29*** |
Recognising suffering | −0.06 | −0.12*** | −0.06 | 0.20*** | −0.15*** | 0.30*** | −0.12*** | −0.09** | 0.37*** | 0.93*** | 0.09*** |
Understanding the universality of suffering | −0.14*** | −0.20*** | −0.06 | 0.31*** | −0.25*** | 0.35*** | −0.17*** | −0.22*** | 0.66*** | 0.92*** | 0.30*** |
Feeling for the person suffering | 0.00 | −0.09** | −0.01 | 0.19*** | −0.14*** | 0.31*** | −0.07* | −0.05 | 0.14*** | 0.84*** | 0.25*** |
Tolerating uncomfortable feelings | −0.18*** | −0.19*** | −0.21*** | 0.29*** | −0.18*** | 0.42*** | −0.19*** | −0.18*** | 0.30*** | 0.87*** | 0.22*** |
Acting to alleviate suffering | −0.03 | −0.11*** | −0.05 | 0.21*** | −0.15*** | 0.39*** | −0.06 | −0.03 | 0.23*** | 0.91*** | 0.24*** |
S-SOCS-FO | −0.16*** | −0.21*** | −0.14*** | 0.31*** | −0.16*** | 0.45*** | −0.26*** | −0.18*** | 0.36*** | 0.31*** | 0.98*** |
Recognising suffering | −0.01 | −0.05 | −0.02 | 0.13*** | −0.02 | 0.22*** | −0.08* | −0.01 | 0.12*** | 0.09*** | 0.95*** |
Understanding the universality of suffering | −0.12*** | −0.14*** | −0.08** | 0.22*** | −0.12*** | 0.32*** | −0.17*** | −0.15*** | 0.26*** | 0.31*** | 0.91*** |
Feeling for the person suffering | −0.14*** | −0.20*** | −0.14*** | 0.29*** | −0.17*** | 0.42*** | −0.24*** | −0.18*** | 0.30*** | 0.29*** | 0.90*** |
Tolerating uncomfortable feelings | −0.18*** | −0.21*** | −0.15*** | 0.27*** | −0.15*** | 0.41*** | −0.26*** | −0.18*** | 0.29*** | 0.23*** | 0.92*** |
Acting to alleviate suffering | −0.22*** | −0.25*** | −0.18*** | 0.34*** | −0.20*** | 0.47*** | −0.29*** | −0.21*** | 0.29*** | 0.28*** | 0.93*** |
Each S-SOCS scale was positively correlated with adaptive constructs: specifically, all three scales correlated moderately to strongly with compassion satisfaction and with FSCRS reassured self subscale (e.g., S-SOCS-S vs. Reassured Self r = 0.61). Overall, S-SOCS-S (self-compassion) showed the strongest and broadest pattern of associations, being negatively correlated with FSCRS Inadequate Self and Hated Self subscales (r = −0.45 to −0.47) and with CFST Burnout/Compassion Fatigue (r = −0.38 to −0.33). This meaning that higher self-compassion was linked to lower self-criticism and burnout. Overall, S-SOCS-S had the largest effect sizes, reflecting expected positive links between self-compassion measure with positive traits and inverse with distress (Gu et al., 2020). The S-SOCS-O (compassion for others) and S-SOCS-FO (from others) scales showed also significant correlations in the predicted directions with a smaller weak to moderate effect size. In sum, such patterns support the convergent and divergent validity of the short scales with correlations being in theoretically meaningful directions, consistent with prior compassion research (Strauss et al., 2016).
Discussion
In the current study, we developed and validated three parallel 10-item short forms of compassion measurement for self, others, and from others (S-SOCS-S, S-SOCS-O, and S-SOCS-FO, respectively), based on the five-facet definition of compassion by Strauss et al. (2016). Our primary motivation was to create short forms to minimise participant burden in compassion assessments, potentially applying them for use in multi-wave interventions and ecological momentary assessments, making the scales more feasible to use in research and practice. In a sample of 2009 helping professionals from Slovakia, we confirmed the conceptual five-factor structure of compassion using these newly developed short forms. The final item selection involved balancing statistical output from the Ant Colony Optimization (ACO; Raborn & Leite, 2018) procedure, examination of factor loadings, and theory-driven item content analysis. Each short scale showed strong total-scale reliability (Cronbach’s α > 0.80) and acceptable to good model fit, along with sound psychometric properties.
First, our results add to the consistent evidence for the five-facet compassion model underlying the SOCS scales (Strauss et al., 2016). The five-facet definition of compassion includes recognition of suffering, understanding its universality, feeling for the person suffering, tolerating uncomfortable feelings, and motivation to alleviate suffering. Conceptually, the S-SOCS scales align with the original SOCS, confirming that even with a reduced set of two items per facet, the distinct components of compassion can be measured separately and meaningfully. This confirmation reinforces the idea that compassion, whether directed towards the self, others, or experienced from others, is multifaceted and not reducible to a single dimension. Each element of compassion (e.g., acknowledging common humanity or taking compassionate actions) remains relevant across the three scales and may have differential relationships with outcomes such as prosocial behaviour, wellbeing, or burnout (Beaumont et al., 2016; Liu et al., 2025). The S-SOCS scales not only mirror the original SOCS content but also offer a unified framework for comparing different compassion domains. By selecting parallel items across the three S-SOCS scales, we provide research way to examine distinct predictors and outcomes of these facets, while minimising measurement bias.
When making item selections for the short forms, we were guided by using parallel items to measure compassion for self, others, and from others, as well as keep in the items that are semantically different so to potentially capture unique aspects of each compassion facet. For example, when choosing the two items for the facet Recognising Suffering, instead of selecting two rather similar statements, such as “I’m good at recognising when I’m feeling distressed” and “I notice when I’m feeling distressed”), we paired the latter statement with “I’m quick to notice early signs of distress in myself”.
The parallel item selection for the three forms of compassion represents a noteworthy methodological innovation with potential theoretical implications. To our knowledge, this is the first measurement approach based on a unified theory of compassion that enables direct comparison across compassion for self, others, and from others using parallel indicators. Existing compassion measures predominantly focus on assessing compassion in a single domain (e.g., self-compassion or compassion for others), and only a minority of instruments cover multiple compassion constructs, highlighting the need for further multidimensional scales development (Jiang et al., 2023).
Evidence of convergent validity further supports the S-SOCS scales as a reliable measure of compassion, with associations in meaningful and expected directions. For example, self-compassion (S-SOCS-S) showed moderate to strong positive associations with adaptive constructs (self-reassurance, compassion satisfaction), and negative associations with maladaptive (self-criticism, compassion fatigue, burnout). Such is in line with the literature on self-compassion: Brophy et al. (2024) mention that self-compassion promotes resilience and serves as a buffer against stress and depression with higher self-compassion strongly linked to lower anxiety. Likewise, in our data participants with higher S-SOCS-S scores reported significantly less burnout and compassion fatigue.
In contrast, compassion for others (S-SOCS-O) and from others (S-SOCS-FO) showed weaker links to depression and anxiety, supporting divergent validity and distinguishing them from personal distress. This pattern is also theoretically driven: compassion for others and from others pertain more to social support and relationships context rather than personal wellbeing (Gilbert et al., 2017). In line with the aforementioned study, S-SOCS-FO had small negative correlations with depressive symptoms and anxiety. At the same time, both S-SOCS-O and S-SOCS-FO had positive moderate correlations with compassion satisfaction, implying that higher rates of compassion for others and from others are connected with positive aspects of carrying and working as a helping professional. This is consistent with the definition by Stamm (2010) of professional quality of life, thus also supporting the convergent validity of the S-SOCS scales.
The S-SOCS scales showed strong associations with their full-length versions. Their scores can be transformed for use in reviews and meta-analyses of compassion studies involving the full scales by multiplying the total compassion score by 2: e.g., original SOCS scales range from 20 to 100, and the S-SOCS from 10 to 50. Multiplying a S-SOCS score by 2 would result in a directly comparable value. Additionally, the S-SOCS Understanding the universality of suffering subscale has identical items in both the S-SOCS-S and S-SOCS-O scales. Thus, if both scales are administered simultaneously for screening, the total length can be reduced from 20 items to 18, as items U2 and U7 are shared across the two scales.
The 10-item S-SOCS scales have several practical implications. Clinically, the short forms might facilitate routine assessment of compassion in education, healthcare, or other social service settings without extensive burden on participants. Due to the link between compassion and wellbeing, such tools can quickly help to identify professionals at risk of burnout or guide compassion-focused trainings. In research, the parallel scales will help compare the three compassion dimensions with each other and enable large-scale screening procedures. In addition, experiences of compassion towards self, others, and from others might be interrelated and can influence each other, highlighting the importance of simultaneous measurement (Kirby et al., 2017). Each compassion dimension shows distinct associations with mental health outcomes, such as shame, self-criticism, depression, stress, wellbeing, and is also subject to individual differences including gender and education, underscoring the need for precise measurement to support prevention of psychological distress and improvement of quality of life (Fuochi et al., 2018; López et al., 2018; Steindl et al., 2021).
The S-SOCS scales are concise measures of the three compassion domains recognised by theory and prior research. The S-SOCS-S (compassion towards self) scale aligns with self-compassion effects on mental health, while the S-SOCS-O/FO (compassion for others/from others) scales capture more social and supportive facets, showing limited overlap with personal distress measures. Our findings suggest that the 10-item S-SOCS-S, S-SOCS-O, and S-SOCS-FO scales are psychometrically sound and theoretically coherent: they retain the multifactorial compassion structure of the originals while efficiently measuring compassion for self, for others, and from others as a general hierarchical factor. The three S-SOCS scales showed good reliability and fit, and should prove useful for research and clinical screening where a rapid assessment of compassion is needed.
Limitations and Future Directions
Our study has several limitations. First, our sample was a convenience sample of Slovak helping professionals. This limits generalizability with possibility for the future studies to explore other cultural settings. Secondly, our study was cross-sectional, with S-SOCS scales require validation across other settings and populations. Finally, the very short subscales (two items per facet) have limitations by their nature: Cronbach’s alpha is sensitive to the number of items (Eisinga et al., 2013), so reliability estimates may be lower (as seen for some of the facets). Although within the acceptable range, they may require other methods of evaluation. Additionally, in the present study measurement invariance across groups and test–retest reliability were not examined; therefore, the equivalence of the S-SOCS scales across groups and their stability over time remain to be established in the future research. Future studies can also evaluate the S-SOCS scales sensitivity to changes across compassion domains in mindfulness, resilience or compassion interventions.
Declarations
Ethical Approval
All procedures performed in studies involving human participants were by the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study’s protocol was approved by the Ethical committee of Faculty of Social and Economic Sciences at Comenius University Bratislava FSEV 1647/-4/2022/SD-CIII/1.
Informed Consent
Informed consent was obtained from all individual participants included in the study.
Conflict of Interest
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, 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 changes were made. 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/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.