Abstract
The validation of the short form of the Self-Compassion Scale (SCS-SF) was conducted in the Mexican population by administering the original version (SCS), previously validated in Mexico, to a sample of 1,250 university students. A four-factor solution was identified, explaining 48.7% of the variance. The three items with the highest loadings for each factor were selected, resulting in a 12-item version with the dimensions of Self-Kindness, Common Humanity, Mindfulness, and Self-Criticism. This structure was confirmed with acceptable fit indices and good item loadings per dimension. Convergent and divergent validity analyses indicated adequate correlations, and the test-retest reliability at 3 months was acceptable. Thus, a validated 12-item version of the SCS-SF for university students in Mexico was obtained, demonstrating sound psychometric properties.
Plain Language Summary
This study tested a short version of the Self-Compassion Scale (SCS-SF) in Mexican university students. The goal was to create a brief and reliable tool to measure how kind and understanding people are toward themselves in difficult situations. A total of 1,250 students answered the original version of the scale, and the results showed that 12 key questions worked best for this population. These questions measure four main aspects of self-compassion: being kind to oneself, recognizing that everyone makes mistakes, staying calm and aware of the present moment, and identifying self-critical attitudes. The new version showed good reliability and consistency over time, as well as strong statistical validity. This means that the 12-item Self-Compassion Scale is a useful and trustworthy tool for assessing self-compassion among university students in Mexico.
Introduction
Self-compassion is the ability to acknowledge one’s suffering without avoiding it, while cultivating the desire to heal it and alleviate it with kindness. It understands failures and pain as part of the human experience without engaging in self-judgment (Neff, 2003a). According to the evolutionary model, self-compassion encompasses motivational, emotional, and cognitive-behavioral competencies that promote well-being, sensitivity, and empathy toward stress, as well as tolerance for discomfort without judgment and a warm, supportive attitude toward oneself (Gilbert et al., 2017; Gilbert & Procter, 2006).
Neff conceptualizes self-compassion as comprising three core dimensions related to emotional, cognitive, and attentional responses to suffering and life experiences. These dimensions are: (1) self-kindness, which involves being compassionate and understanding toward oneself instead of engaging in self-criticism or judgment; (2) common humanity, which recognizes that personal difficulties are part of the shared human experience rather than isolated events; and (3) mindfulness, which entails maintaining a balanced view on painful thoughts and emotions without identifying with them. Each dimension has both positive and negative components, resulting in six factors: self-kindness versus self-judgment, common humanity versus isolation, and mindfulness versus over-identification (Neff, 2003a, 2023).
Empirical evidence from various countries highlights the beneficial impact of self-compassion on health, physical and mental, particularly on symptoms of anxiety, depression, and perceived stress. Additionally, self-compassion has been associated with beneficial effects on conditions such as chronic pain, addictions, disruptive behaviors, and eating disorders (Homan & Sirois, 2017; Muris & Petrocchi, 2016; Palmer-Cooper et al., 2023; Ramsey et al., 2023; Serpell et al., 2020).
As a result, self-compassion is considered a protective factor, fostering resilience in affliction and promoting overall well-being (Muris & Otgaar, 2023). Furthermore, self-compassion contributes to a sense of security, reducing feelings of threat or overwhelm triggered by external pressures (Gilbert & Procter, 2006).
A meta-analysis documented an inverse association between self-compassion and psychopathological symptoms (Muris & Petrocchi, 2016). When analyzing its six factors separately, common humanity was the only dimension not identified as an independent predictor of psychopathology, whereas the negative components showed the strongest predictive effects. The general effect size for the self-compassion scale across 18 studies was r = −0.53 (95% CI [−0.70, −0.36]; Z = 36.37, p < .001; Muris & Petrocchi, 2016).
Since the introduction of self-compassion as a psychological construct, at least 36 scales—ranging from long forms to short forms and adaptations for different populations—have been developed to measure both compassion and self-compassion (Jiang et al., 2023). Among these, the most widely used is Neff’s Self-Compassion Scale (SCS; López Tello et al., 2023), which consists of 26 items rated on a 5-point Likert scale.
This scale assesses the six dimensions of self-compassion, incorporating both positive and negative aspects. It has demonstrated strong psychometric properties, including high internal consistency and, in confirmatory factor analyses, good model fit, for the six-factor model and for the higher-order single-factor model, where self-compassion is explained through the interrelation of its six dimensions (Neff, 2003b).
The Spanish adaptation and validation of the full version of the SCS were conducted in Spain in 2014 (Garcia-Campayo et al., 2014). This 26-item version has since been adapted in various Latin American countries, including Chile (Araya et al., 2017), Colombia (Martínez-Ramos et al., 2022), and Mexico (López Tello et al., 2023).
The short form of the SCS (SCS-SF) was introduced in 2010 in Belgium and the United States, with validation performed in both Dutch and English (Raes et al., 2010). The SCS-SF was constructed using a sample of 271 (79% women) first-year Dutch-speaking psychology students at a university in Belgium, based on the validated full version in Dutch, which has two fewer items and a 7-point response scale instead of 5 points (Raes et al., 2010), compared to the original version (Neff & Vonk, 2009), as part of the process of adapting the scale to the language and culture.
Based on the six dimensions originally proposed by Neff, the authors selected two items for each dimension, choosing those with the highest correlations with the general scale and with the specific subscale; at the same time, they assessed whether the selected items reflected the breadth of content of the original subscale (Raes et al., 2010). Subsequently, they replicated and validated the scale in an independent sample of 185 Dutch-speaking adults (70.8% women) and the English version in 415 students (65.5% women) at a university in Texas. In the confirmatory factor analysis, adequate fit indices were obtained for a higher-order model representing a first-order general self-compassion factor and six second-order factors. The total score demonstrated high internal consistency (α = .86); however, the internal consistency of some subscales was low (<0.60), in both versions (Dutch and English), the self-kindness subscale had the lowest internal consistency. Due to this limitation, the authors recommend using the short form only to assess the overall construct, while relying on the full scale for analyzing individual dimensions (Raes et al., 2010).
The short form developed by Raes et al. (2010) has since been validated in multiple countries (Babenko & Guo, 2019; Baggaley et al., 2024; Bratt & Fagerström, 2020; Büyüköksüz et al., 2025; Garcia-Campayo et al., 2014; Maya et al., 2024; Meng et al., 2019; Villalón López et al., 2023). The first validation of both the full and short forms in Spanish was conducted in Spain (Garcia-Campayo et al., 2014), while the only validation of the SCS-SF in Latin America was performed in Chile (Villalón López et al., 2023; Table 1).
Main Characteristics and Results of the Validated Short Versions of the Self-Compassion Scales.
All these validations have preserved the original items from the SCS-SF proposed by Raes et al. (2010), maintaining the dimensional structure of Neff’s (2003b) Self-Compassion Scale. Distinctly, an adaptation conducted among Turkish university students demonstrated measurement invariance when compared with a non-clinical sample from the United States (Büyüköksüz et al., 2025).
However, none of these studies account for potential variations related to sociocultural factors, since individuals are affected by interactions within the society in which they live, which is framed by cultural aspects that arise from the interaction of norms, beliefs, roles, values, and attitudes; this determines behavior and personality development, which, in the case of Mexicans, can be influenced by traditional aspects (Díaz-Guerrero, 2007; Palacios & Martínez, 2017).
Specifically, the concepts of compassion and self-compassion have shown cultural differences. While people from the United States perceive compassion toward others and themselves as a the display of attitudes or expressions focused on positive aspects, such as a kind response with a small smile, in countries such as Germany (Koopmann-Holm et al., 2021) or China (Seow et al., 2025), compassionate behaviors focus on negative aspects, even with expressions that reflect distress or difficulty. On the other hand, compassionate behavior in the Latino population reflects the sadness of others and is composed of positive and negative aspects in similar levels, that is, Latinos value displays of compassion with positive feelings and accept negative expressions (Hernandez et al., 2025; Larco et al., 2024).
Montero-Marin et al. (2018) conclude that there is a different conceptualization of self-compassion that is influenced by the contextual values of the participants. Those whose values are related to uncertainty avoidance, long-term orientation, and restraint are associated with the positive elements of self-compassion, while values focused on individualism are associated with the negative elements. On the other hand, the importance of distinguishing self-compassion from self-criticism has been raised, as it has been observed that self-compassion is mainly determined by the positive items on the Self-Compassion Scale (SCS; Montero-Marin et al., 2018), and that the sum of the total items on this tool is not justified (López et al., 2015).
In Mexico, the only validated self-compassion measures available are the Compassion Scale for the Mexican population (López Tello & Moreno Coutiño, 2019) and the adapted and validated version of Neff’s SCS, which was tested in a non-probabilistic sample of 343 individuals over the age of 16 (López Tello et al., 2023). These remain the only psychometrically validated instruments for measuring self-compassion in Mexico.
Short-form scales are valuable tools in both clinical and research settings where time constraints limit the feasibility of using longer instruments. As long as they maintain validity, short forms offer several advantages, including reduced respondent fatigue, lower time demands, and minimized reactance (American Educational Research Association et al., 2018). A validated short form for assessing self-compassion could streamline data collection, making it more efficient in terms of both time and cost (Raes et al., 2010).
Thus, the objective of this research is to develop and validate the short form of the Self-Compassion Scale (SCS-SF) in Mexican university students.
Method
Ethical Considerations
This study was approved by the Research, Research Ethics and Biosafety Committees of the University Center for Health Sciences of the University of Guadalajara (approval code: CI-01024) and conducted in accordance with the ethical principles of the Declaration of Helsinki (World Medical Association, 2025). Participants received verbal and written information about the purposes of the study and provided electronic informed consent prior to participation. Participation was voluntary, and data confidentiality and anonymity were ensured.
Participants
This was an instrumental study conducted in August 2024. The sample size was calculated based on N = 1,750 students, entering the health sciences campus at a university in Jalisco, Mexico, using a 95% confidence interval and a 5% error limit, resulting in a required sample of n = 315, which meets the recommendation of including at least 10 subjects for each question on the scale or between 200 and 300 observations (Boateng et al., 2018). A non-probabilistic convenience sampling strategy was employed, recruiting participants who were available during the orientation course.
A total of 1,250 participants aged 18 to 58 years (M = 19.3, SD = 3.04) were included, with 66.9% identifying as women (n = 836). The most represented academic programs were Medicine (21.6%) and Nursing (21.3%).
To ensure robust psychometric validation, the total sample was randomly divided into two groups. The exploratory factor analysis (EFA) was conducted with a sample of 560 participants (M = 19.27, SD = 2.71), of whom 339 (65.2%) were women. The confirmatory factor analysis (CFA) was performed on the remaining 690 participants (M = 19.34, SD = 3.26), with 497 (68.1%) identifying as women.
Instruments
Participants completed a sociodemographic questionnaire that collected information on age, sex, and academic program. Additionally, two validated psychological scales were applied.
The Self-Compassion Scale (SCS), in its 26-item version proposed by Neff, adapted and validated in the Mexican population by López Tello et al. (2023), through a process of direct translation and back-translation of the original English instrument, without using the version already validated in Spain by Garcia Campayo et al. (2014), due to language differences between the two countries, was used to measure self-compassion.
The scale employed a 5-point Likert scale ranging from almost never (1) to almost always (5). The omega reliability coefficients for the six components of the scale were 0.48 for self-kindness (e.g., “I try to be loving toward myself when I’m feeling emotional pain”), 0.80 for self-judgment (e.g., “I’m disapproving and judgmental about my own flaws and inadequacies”), 0.63 for common humanity (e.g., “When things are going badly for me, I see the difficulties as part of life that everyone goes through”), 0.73 for isolation (e.g., “When I think about my inadequacies, it tends to make me feel more separate and cut off from the rest of the world”), 0.76 for mindfulness (e.g., “When something upsets me, I try to keep my emotions in balance”), and 0.68 for over-identification (e.g., “When I’m feeling down, I tend to obsess and fixate on everything that’s wrong”). To obtain a total self-compassion score, the items from the negative dimensions were reverse-coded. The Mexican validation of the scale showed adequate model fit indices in an exploratory structural equation model, with RMSEA = 0.048, CFI = 0.97, TLI = 0.95, and WRMR = 0.501. However, the chi (χ2 = 294.669, df = 164, p < .001) was significant (Jordan Muiños, 2021).
The Positive and Negative Affect Scale for the Mexican Population (PANA-M) was also applied to measure affective states. This instrument, developed by Velasco-Matus with a solid conceptual framework, consists of 20 items divided into two 10-item subscales assessing positive and negative affect. Items are rated on a 5-point Likert scale ranging from never to always and include emotions such as “happiness,”“joy,”“suffering,” and “pain.” The scale is considered culturally specific and allows content comparisons with other international measures. It explains 80% of the variance and has a Cronbach’s alpha of .92. The authors recommend presenting items in alphabetical order, a practice followed in this study (Velasco Matus et al., 2021). For validation purposes, the positive affect subscale was used for external convergent validation, while the negative affect subscale was used for divergent validation.
Procedure
The study was part of a project approved by the Research, Research Ethics, and Biosafety Committees of the Institution with report CI-01024. The study objectives were explained to the participants, and informed consent was provided. The questionnaire battery was administered electronically via Google Forms to all new students who agreed to participate during their orientation course, after the project objectives and voluntary participation had been explained. Participants received the survey link, read the informed consent form, and, if they agreed, completed the questionnaire. Responses were reviewed to eliminate duplicate entries and cases that did not meet the inclusion criteria.
As self-compassion is considered a trait-like construct, which has been shown to remain stable for up to 5 months (Raes, 2011), a follow-up assessment was conducted 3 months later. A total of 203 participants completed this second assessment, allowing for test-retest reliability analysis.
Data Analysis
Statistical analyses were performed using IBM SPSS Version 29.0 and RStudio 2023.03.0. Univariate normality was assessed using the Kolmogorov-Smirnov and Shapiro-Wilk tests, in addition to an analysis of skewness and kurtosis. Internal consistency was evaluated using ordinal alpha. The original 26-item SCS and the proposed short form (SCS-SF) were subjected to exploratory factor analysis to assess dimensionality.
The dataset was randomly divided into two groups for EFA and CFA. The CFA tested two structural models: a bifactor model and a second-order hierarchical model, as recommended by Neff et al. (2019). In the bifactor model, each item is assumed to be influenced simultaneously by a general factor and specific orthogonal factors (Flores-Kanter et al., 2018). In contrast, the second-order hierarchical model assumes that the relationship between the general factor and the items is mediated by first-order factors (Flores-Kanter et al., 2018).
Model fit was evaluated using established criteria. An acceptable fit was defined as χ2/df ≤ 5 (Hair et al., 1998), SRMR and RMSEA ≤ 0.08 (Browne & Cudeck, 1992), and CFI, TLI, and GFI values ≥ 0.90 (Hu & Bentler, 1999). Internal consistency was assessed using ordinal alpha (α), as the data were ordinal in nature, with values ≥ .7 considered acceptable (Nunnally, 1994; Viladrich et al., 2017).
Construct validation was performed through confirmatory factor analysis and assessed using convergent, divergent, and discriminant validation. Spearman’s correlation test was used to examine associations between the short form and other psychological constructs. The stability of the scale was evaluated using test-retest reliability. Criterion validity was examined by analyzing the correlation between the original full-scale SCS and the proposed SCS-SF, following the guidelines of Carvajal Valcárcel et al. (2011).
The objective of the analyses was to estimate and validate the construct of the short version of the SCS, seeking the most parsimonious and appropriate factorial structure for the Mexican sample.
The analysis did not start from a fixed structure hypothesis, but rather established expected results that guided the evaluation of the final solution. These included: (1) Replication of the dual nature of self-compassion, with a clear differentiation between positive and negative dimensions. (2) Demonstration of rigorous internal construct validity, verifying convergence (AVE ≥ 0.50) and discrimination between factors (HTMT ≤ 0.85). (3) Compliance with external criterion validity, manifested in the expected significant correlation with external measures of affect (e.g., Positive Affect).
Results
Normality
Tests for normality indicated that none of the distributions followed a normal pattern. Both the Kolmogorov-Smirnov and Shapiro-Wilk tests were significant (p < .05), suggesting that all distributions significantly deviated from normality. Skewness and kurtosis values exceeded 1.0, a threshold which can begin to induce bias in the estimates (Curran et al., 1996). Consequently, the data were treated as non-normally distributed
Exploratory Factor Analysis (EFA)
The Kaiser-Meyer-Olkin (KMO) test for sample adequacy and Bartlett’s test of sphericity were performed to determine whether the data were suitable for factor analysis. The KMO index was 0.943, indicating high sampling adequacy, while Bartlett’s test of sphericity was significant (χ2 = 6,701.357, df = 325, p < .001), confirming that factor analysis was appropriate.
The EFA was conducted on the 26 items of the Self-Compassion Scale (SCS) using the subsample of 560 participants. The EFA was performed using the unweighted least squares extraction method with Promax rotation. The factors retained were determined based on the eigenvalue criterion (eigenvalues greater than 1). Items with factor loadings below 0.40 were eliminated (Table 2).
Structure Matrix.
Note. The items selected by factor for the short version are in bold.
To establish robust retention criteria, Parallel Analysis was also applied using the “fa.parallel” function from the “psych” package in R software. The results of this analysis suggested retaining four factors and three components. Given that the objective of our research is to identify latent psychological constructs (factors) underlying the observed responses on the scale, we opted for the recommendation of four factors (see Figure 1).
Parallel analysis screen plot.
From the original six factors, a four-factor solution was obtained, explaining 48.7% of the variance. Table 2 presents the factor loadings from the structure matrix, showing the degree to which each item saturates in its respective factor and considering the interrelationships among the scale’s dimensions.
For the estimation of reliability and construct validity, psychometric best practices require each factor to contain a minimum of three items, as this allows for the calculation of item covariances and the estimation of internal consistency (Brown, 2015). Therefore, for the short form (SCS-SF), the three items with the highest factor loadings were selected for each factor, ensuring that they did not exhibit higher loadings in another factor.
Factor 1 contained two items from the Over-Identification subscale and one item from Self-Judgment. Due to the common conceptual aspects reflected in these items, this factor was labeled “Self-Criticism” (SC). Factors 2 and 3 contained items from the Self-Kindness (SK) and Common Humanity (CH) subscales, respectively, and thus retained these names. Factor 4 included the three highest-loading items from the Mindfulness (M) subscale: Item 14 (0.747), Item 9 (0.667), and Item 17 (0.625).
Five items were excluded because they had a primary factor loading of less than 0.40. Another nine items were discarded because they were not among the three items with the highest saturation within their corresponding theoretical factor, despite having acceptable loadings. All retained items were verified to maintain theoretical consistency. The only exception was Item 17 (Mindfulness), which was retained in its theoretical dimension (Mindfulness) because its loading exceeded the threshold of 0.40 and was essential for the conceptual representation of the factor, although it loaded slightly more on Self-Criticism.
Following the selection of 12 items for the Self-Compassion Scale Short Form (SCS-SF), the EFA was repeated. The four-factor structure remained stable, with factor loadings greater than 0.40, and adequate reliability coefficients for the total scale, with an ordinal alpha < .70. The accumulated variance explained by the short form was 60.2%, with a balanced distribution across the four factors. The Self-Kindness factor contributed the most to the variance (16.8%), while Common Humanity accounted for the least (12.9%; Table 3).
Final Pattern Matrix for the Short Version of the Self-Compassion Questionnaire.
Note. Promax rotation method. SK = Self-Kindness; M = Mindfulness; SC = Self-Criticism; CH = Common Humanity.
The SCS-SF validated in this study is presented in Appendix 1 in its original Spanish-language version, as it was designed for use with the Mexican population.
The short form (SCS-SF) proposal for the Mexican population coincides with 6 of the 12 items from the short version originally proposed by Raes et al. (2010). The two Mindfulness (M) items were retained. For Self-Kindness (SK), Item 12 was included, despite having the lowest factor loading (0.749) among the three items within this dimension. Consequently, Item 26 was excluded, although it was included in the SCS-SF proposal by Raes et al. (2010). A similar selection process occurred in the Common Humanity (CH) dimension, where Item 15 from the original scale was excluded, as it ranked fourth in factor loading, while Item 10, which aligned with the original SCS-SF, had the highest factor loading for this dimension (0.710).
The factors associated with self-compassion deficits were consolidated into a single dimension: Self-Criticism (SC). Only two items from the Over-Identification subscale (Items 2 and 6) were retained, coinciding with those in the original SCS-SF, along with one item from the Self-Judgment subscale (Item 16), which was not part of the proposal of Raes et al. (2010). The excluded items from the negative dimensions ranked fourth (Item 1), fifth (Item 25), seventh (Item 13), and ninth (Item 11) in factor loading (Table 3).
Construct Validity and Confirmatory Factor Analysis (CFA)
The construct validity of the SCS-SF was assessed through confirmatory factor analysis (CFA) and internal consistency tests in a sample of 730 participants. Two models were tested using the Weighted Least Squares Mean and Variance adjusted (WLSMV) estimator, appropriate for ordinal variables: the second-order hierarchical model, which employed non-orthogonal rotation, and the bifactor model, which used orthogonal rotation. In the bifactor model, the general factor explains the common variance among items, while the group factors remain uncorrelated.
Both models demonstrated acceptable fit indices, with χ2/df values of 3.42 for the second-order hierarchical model and 2.68 for the bifactor model. All values of the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR) were equal or below 0.08. Other model fit indices also indicated good fit: comparative fit index (CFI) = 0.95 and 0.98, Tucker-Lewis index (TLI) = 0.94 and 0.97, and goodness-of-fit index (GFI) = 0.99 and 0.99. The model fit indices for the retest are presented in Table 4.
Fit indices for Hierarchical and Bifactor Models.
Note. The fit indices are reported for both the initial sample (n = 730) and the retest sample (n = 218). χ2 = chi-square; df = degrees of freedom; CFI = comparative fit index; TLI = Tucker-Lewis index; GFI = goodness-of-fit index; RMSEA = root mean square error of approximation; IC 90% = 90% confidence interval.
The ordinal α coefficients for both the second-order hierarchical model (Figure 2) and the bifactor model (Figure 3) of the SCS-SF exceeded the threshold of ≥ 0.7 in both the CFA sample and the retest sample (see Table 5). The convergent validity and discriminant validity were assessed using the Average Variance Extracted (AVE) and the Heterotrait-Monotrait Ratio of Correlations (HTMT), respectively, within the final four-factor hierarchical model. The AVE was not calculated for the bifactor model because the metric lacks clear meaning when item variance is simultaneously explained by two sources (a general and a specific factor). In the chosen hierarchical model, convergent validity was confirmed as all four factors achieved AVE values equal to or exceeding 0.50 (ranging between 0.50 and 0.67). Furthermore, discriminant validity was satisfactory, as all HTMT values were below the strict threshold of 0.85 (ranging between 0.33 and 0.72), confirming that the four dimensions are statistically distinct constructs (Henseler et al., 2015).
Second-order hierarchical model of the short form of the Self-Compassion Scale (SCS-SF) with standardized coefficients (p < .05).
Bifactor model of the short form of the Self-Compassion Scale (SCS-SF) with standardized coefficients (p < .05).
Reliability Coefficients α Ordinal and Mean (Standard Deviation) by Dimension and Model.
Note. The second-order coefficients correspond to the second-order hierarchical model, while the general factor coefficients correspond to the bifactor model. M = mean; SD = standard deviation.
These results propose that the SCS-SF has an adequate factor structure for deriving both subscale scores and an overall score. In the case of the second-order hierarchical model, the subscales are correlated, and the overall score represents the shared variance among the four subscales (Figure 2).
In the bifactor model, the general factor explains the common variance among all items while the four subscales capture the unique variance of each item within its respective factor, with no correlations between the subscales (Figure 3).
Table 6 presents the standardized factor loadings for both models. The second-order hierarchical model shows higher factor loadings for each dimension, whereas in the bifactor model, the factor loadings per dimension are lower, with some cases falling below 0.300 and not reaching statistical significance. The low fit of the Mindfulness dimension in the two-factor model indicates that this is not the most appropriate factor structure for the scale in the Mexican sample. The two-factor model assumes that the items are simultaneously influenced by a general factor and orthogonal specific factors. However, our results, which favor the second-order hierarchical model, suggest that the Self-Compassion construct in this population operates with a mediated structure: the general factor (Global Self-Compassion) is related to the items through the first-order dimensions, rather than these dimensions operating independently. This validates the conceptual hierarchy of the construct.
Factor Loadings in the Models.
Note. Standardized coefficients are presented, with standard errors in parentheses. Bolded coefficients indicate non-significant values, while the remaining coefficients were statistically significant (p < .05).
Reference parameter with no standard error.
External Convergent and Discriminant Validation
External convergent and divergent validation was assessed using Spearman’s correlation test with the positive (PA) and negative affect (NA) subscales of the Positive and Negative Affect Scale for the Mexican Population (PANA-M).
The correlation between the PA subscale and the total Self-Compassion Scale (SCS) score was ρ = .617 (95% CI [.580, .651]), while for the short form (SCS-SF), it was ρ = .580 [.541, .617], both statistically significant, demonstrating adequate convergent validity.
Similarly, adequate divergent validity was observed with the NA subscale, showing an inverse correlation for both versions of the scale. The correlation was ρ = −.572 [−.610, −.533] for the SCS and ρ = −.463 [−.506, −.416] for the SCS-SF, both statistically significant (Table 7).
Spearman’s Correlation Between the Subscales of the Short Form of the Self-Compassion Scale (SCS-SF) and the Total Scores of the Full and Short Versions, as Well as the Positive and Negative Affect Scale (n = 1,250).
Note. **p < .001. SCS = Self-Compassion Scale; SCS-SF = Short Form of the Self-Compassion Scale; SC = Self-Criticism; SK = Self-Kindness; CH = Common Humanity; M = Mindfulness; NA = Negative Affect; PA = Positive Affect.
Discriminant validation was assessed by analyzing sex differences in the total scores of the SCS, the SCS-SF, as well as the PA and NA scales. Levene’s test was used to evaluate variance homogeneity, revealing that all variables except positive affect exhibited significant differences in variance.
For the SCS-SF total score, significant differences were found between women (M = 40.1, SD = 8.9) and men (M = 41.7, SD = 8.9), t (907) = −3.2, p < .001, with men displaying somewhat higher levels of self-compassion.
Stability of the Questionnaire, Test-Retest
To estimate the stability of the scale, a test-retest procedure was conducted, with a second application administered after 3 months to 218 participants. The total scores at both time points were summed, and the corresponding correlation coefficient was calculated based on the results of the normality tests.
For the total score of the SCS in its original version, the correlation between the first and second applications was ρ = .711 (95% CI [.637, .773], p < .01), while for the short form (SCS-SF), it was ρ = .678 ([.597, .746], p < .01). Among the subscales, the self-criticism dimension exhibited the highest test-retest correlation (ρ = .664, [.580, .734], p < .01), whereas the common humanity dimension had the lowest (ρ = .525, [.419, .618], p < .01).
Additionally, the retest data confirmed the maintenance of adequate fit indices (Table 4) and acceptable ordinal α coefficients (Table 5), with values of ≥ .7.
Criterion Validity
To assess criterion validity, the total score of the final version of the SCS-SF was correlated with the total score of the original SCS. A high correlation indicates that both versions measure the same construct. The correlation between the total scores of the original and short versions was ρ = .922 (95% CI [.913, .930], p < .001), confirming that the short form maintains the evaluation of the same concept (Table 8).
Spearman’s Correlation Between the Dimensions of the Short Form of the Self-Compassion Scale (SCS-SF) and the Full Version (n = 1,250).
Note. *p < .001. SCS = Self-Compassion Scale; SCS-SF = Short Form of the Self-Compassion Scale for Mexican University Students; SC = Self-Criticism; SK = Self-Kindness; CH = Common Humanity; M = Mindfulness.
Additionally, correlations between the four dimensions of the short form were examined, all of which were statistically significant. The strongest correlation was observed between self-kindness and mindfulness (ρ = .521, 95% CI [.479, .562], p < .001), while the weakest correlation was found between self-criticism and common humanity (ρ = −.160, [.104, .215], p < .001; Table 8).
Discussion
The aim of this study was to develop and validate a short form of the Self-Compassion Scale (SCS-SF) for the Mexican university student population. The final version consists of 12 items distributed across four dimensions, explaining 48.7% of the variance, with good fit indices in the confirmatory factor analysis (CFA). Although the resulting short form consists of 12 items, as do other validated short versions (Babenko & Guo, 2019; Baggaley et al., 2024; Bratt & Fagerström, 2020; Maya et al., 2024; Villalón López et al., 2023), these versions retain two items per dimension from the six original subscales. In contrast, the present study identified a four-factor structure, with three items per factor, following psychometric recommendations that each factor should include at least three to four items to maintain stability, particularly in samples of fewer than 200 participants (Lloret-Segura et al., 2014).
The original SCS-SF, proposed by Raes et al. (2010), was developed in a sample of 271 university students in Belgium and sought to preserve the theoretical structure of Neff’s (2003b) Self-Compassion Scale. The selection of items was based on those with the highest correlations with both the total scale score and their respective subscales. The total SCS used by Raes et al. (2010) was based on a Dutch version that, due to translation issues, eliminated two questions and applied a 7-point response scale instead of 5.
In contrast, the Mexican version proposed in this study selected items based on their highest factor loadings within the four dimensions identified in the exploratory factor analysis (EFA). This is the first Spanish-language SCS-SF developed from an application of the full 26-item scale, whereas other validations, including those conducted in Canada (Babenko & Guo, 2019), Chile (Villalón López et al., 2023), Spain (Garcia-Campayo et al., 2014; Maya et al., 2024), Poland (Holas et al., 2024), the United Kingdom (Baggaley et al., 2024), and Sweden (Alfonsson et al., 2023; Bratt & Fagerström, 2020), directly adopted the 12-item selection from Raes et al. (2010), focusing primarily on CFA-based validation. However, there are two studies that perform exploratory factor analysis using the principal component method, based on Raes’ original 12 items, for application in older adults (Bratt & Fagerström, 2020) or in populations with dementia (Baggaley et al., 2024).
Of the 12 items in this study’s short form, six coincide with those in the Raes et al. (2010) version, including two items from the mindfulness subscale, one from self-kindness, one from common humanity, and two from the over-identification subscale (Table 3). These differences may be influenced by cultural aspects specific to the Mexican population. Philosophical and psychological analyses suggest that self-perception issues among Mexicans may contribute to an internalized sense of inferiority, affecting attitudes and behaviors (Ramos, 2001). On the other hand, it has been identified that the Mexican personality includes self-sacrifice, flexibility, self-modification, resignation, and sympathy, as well as being lazy, conformist, proud, and ingenious. Mexicans seek to be accepted and obtain opportunities by being sociable, friendly, and easy to get along with, with an ambivalent personality between positive and negative traits (Cruz Martínez et al., 2013; Palacios & Martínez, 2017); these aspects may have influenced the factor structure observed in this validation.
According to Montero Marin et al. (2018), the conceptualization of self-compassion differs based on the context in which participants develop. When analyzing cultural aspects regarding behavior in the responses obtained from the Self-Compassion Scale (SCS), he observes that positive items better explain the factors associated with self-compassion, based on dominant values such as uncertainty avoidance, long-term orientation, and restraint, while indulgence and individualism are more associated with the negative elements of self-compassion.
These findings may be related to the fact that in the SCS-SF proposal, all negative items were incorporated into a single dimension, an aspect already reported by other researchers, in which this aspect is referred to as self-criticism or self-coldness (López et al., 2015; Mills et al., 2007; Montero-Marín et al., 2016; Muris et al., 2018; Muris & Petrocchi, 2016; Zeng et al., 2016). This dimension has shown stronger associations with mental health problems such as mood disorders and anxiety (Muris et al., 2018; Muris & Petrocchi, 2016).
A similar situation was observed in the validation of the SCS-SF in Swedish adults, where adequate fit indices were found for the first-order model. However, due to a high correlation between over-identification and isolation, the authors proposed a five-factor model that merged these dimensions, which led to only marginal improvements in model fit (Alfonsson et al., 2023).
Moreover, as Swain et al. (2008) pointed out, psychological scales are designed to assess the presence of traits, beliefs, and attitudes; thus, negatively worded items that assess their absence may introduce complications in the factorial structure, as participants respond better to the reflection of their experience.
It has already been described that the concept of compassion toward others and toward oneself is not homogeneous across different cultures. It has been clearly identified that the way of showing empathy is different; thus, people in the United States avoid negative emotions because they consider them to be something bad, so they associate compassion with positive aspects, with a kind response or a small smile; while people from Germany and China show compassion with negative attitudes, displaying expressions that show distress or difficulty, closer to what the person they want to show compassion to is experiencing (Hernandez et al., 2025; Koopmann-Holm et al., 2021; Seow et al., 2025).
On the other hand, when it comes to the Latin American population, there is no fear or shame in expressing emotions, as emotional expression is culturally valued. Therefore, showing compassion for the sadness of others is composed of positive and negative aspects in similar proportions; that is, the positive is valued and the negative is accepted (Hernandez et al., 2025; Larco et al., 2024). This may be because Latin American societies are collective and interdependent, where loyalty to the group (family or community) is considered an important value, as is the display of “sympathy” as a cultural value, which leads to deeper relationships where emotions are shared as part of compassionate behavior and an active relationship is shown. In contrast, it has been identified that in the United States there are fewer terms to describe negative emotions. It is a society in which there is greater avoidance of negative emotions, making it less likely that compassion will be expressed through the sharing of emotions, which could lead to less empathy (Hernandez et al., 2025).
Previous validations of the full SCS have yielded varying findings regarding the number and composition of its dimensions (Araya et al., 2017; López et al., 2015; Zeng et al., 2016). Studies supporting the original six-factor structure have relied primarily on CFA-based validation (Coroiu et al., 2018; Garcia-Campayo et al., 2014; López Tello et al., 2023; Martínez-Ramos et al., 2022; Petrocchi et al., 2014), particularly through a six-factor bifactor model, as recommended by Neff et al. (2019). While these models demonstrate adequate fit indices, not all studies report individual item loadings, which are crucial for evaluating the independent relationships of items within both their respective subscales and the overall scale.
In this study, the second-order hierarchical model emerged as the best fit for the Mexican population, as the bifactor model produced non-significant factor loadings for the mindfulness dimension (Table 6), which shows a possible problem with the model, since it has been identified that bifactor models tend to have adequate fit values that are even better than other models, and that this may hides the presence of anomalies (Ondé Pérez & Alvarado Izquierdo, 2022).
Thus, in the validation of the complete scale in Mexico, the bifactor model showed good fit indices, while the factor loadings for the general factor were low for several items in the positive dimensions and adequate for their respective groups. In contrast, the items in the negative dimensions had factor loadings > 0.55 in the general dimension and factor loadings of < 0.32 for most items in their specific dimension, even with values of 0.01 (López Tello et al., 2023).
Regarding the mindfulness dimension, it is an element that enables individuals to remain emotionally open in the present moment, with a clear vision and non-judgment, helping them to maintain balance and perspective in the face of a previous situation of suffering (Gilbert & Procter, 2006; Neff, 2003a). Consequently, mindfulness fosters psychological distance from difficult situations to allow the emergence of self-kindness and the recognition of our shared humanity. This illustrates the interconnection among the components of self-compassion, as a trainable trait encompassing cognitive, affective, and behavioral aspects (Asensio-Martínez et al., 2017). In this sense, the lack of significant loading on the group factor in our bifactor model might be related to the high factor loading observed for item 14 and the low factor loading for item 17.
It is worth noting that of the nine studies reviewed in which the SCF-SF was validated with CFA, only three applied bifactor analyses, one conducted in adults from clinical and non-clinical samples that obtained inadequate fit indices (Alfonsson et al., 2023), another in adolescents, which achieved adequate but weaker fit indices than the six-factor model (Maya et al., 2024) and the last one in a clinical an non clinical sample where the six bifactor model was not identified (Büyüköksüz et al., 2025).
Other models tested in the validation of the SCS-SF are unifactorial (Alfonsson et al., 2023; Babenko & Guo, 2019; Holas et al., 2024; Maya et al., 2024; Villalón López et al., 2023), two-factor (Alfonsson et al., 2023; Babenko & Guo, 2019; Bratt & Fagerström, 2020; Villalón López et al., 2023), three traditional -factor (Büyüköksüz et al., 2025; Villalón López et al., 2023), three-factor (one positive and two negative) with 10 items (Meng et al., 2019) and six-factor (Bratt & Fagerström, 2020; Büyüköksüz et al., 2025; Garcia-Campayo et al., 2014; Maya et al., 2024; Villalón López et al., 2023), as well as the second-order hierarchical model (Alfonsson et al., 2023). The results are varied, as while adequate fit indices were found for the two- and six-factor models in medical students and doctors (Villalón López et al., 2023), these same models are inadequate in older adults (Bratt & Fagerström, 2020).
These findings contribute to the ongoing debate regarding the optimal factor structure of self-compassion, with diverse models proposed (Baggaley et al., 2024; Coroiu et al., 2018; Holas et al., 2024; Villalón López et al., 2023). For instance, studies in Buddhist and non-Buddhist Chinese populations have reported high intercorrelations among negative dimensions, leading to validation of a model that includes only the three positive dimensions and a model that combines the three negative dimensions (Zeng et al., 2016). Other studies propose a two-factor model, in which positive items form one factor and negative items form another (Babenko & Guo, 2019; López et al., 2015). Some researchers even question whether the negative dimensions should be part of self-compassion, given their correlation with other psychological constructs (López et al., 2015). In contrast, a unidimensional approach with residual correlations has been proposed in Poland (Holas et al., 2024). These discrepancies suggest that the optimal factor structure may vary depending on sample characteristics such as cultural background, age, and population-specific traits.
While the proposed Mexican short form differs from previous versions in terms of its included items and four-factor structure, it retains the theoretical foundation of the original scale. The three positive subscales—self-kindness, common humanity, and mindfulness—remain intact, while all negative dimensions were consolidated into a single factor: self-criticism. These differences may reflect cultural variations among Mexican university students, further supporting the decision to rely on EFA in this validation. A similar approach was taken in the Chinese version of the short form (Meng et al., 2019) and in Chile during the validation of the complete version, with 17 remaining items restructured into 4 dimensions: isolation, kindness, self-judgment, and self-understanding (Araya et al., 2017), differing from the factor structure observed in Mexico.
The internal consistency of the Mexican SCS-SF (α = .85) is slightly lower than the values reported by Raes et al. (2010) in Dutch (α = .87) and English (α = .86) versions, with greater internal consistency than that found in older adults with (α = .796; Baggaley et al., 2024) and without dementia (α = .68; Bratt & Fagerström, 2020), Spanish adolescents (α = .723; Maya et al., 2024), and Polish adolescents (α = .82; Holas et al., 2024).
However, this proposal demonstrates adequate internal consistency (>0.70) for all four dimensions, whereas Raes’ version exhibited low internal consistency for self-compassion (α = .55 Dutch, .54 English) and common humanity (α = .60 Dutch, .62 English; Raes et al., 2010), like the version by Bratt & Fagerström (2020) validated in older adults in Sweden, in which only the self-judgment dimension had acceptable values (α = .72). This suggests that the Mexican SCS-SF effectively measures both global self-compassion and its underlying dimensions, having an advantage in its application over other validated versions and supporting its construction process. Many other SCS-SF validations do not report internal consistency data, general, or for its dimensions, as they focus primarily on CFA fit indices.
The intercorrelations among the four dimensions were generally consistent with previous research, with the highest correlation observed between mindfulness and self-kindness. However, weak correlations were found between common humanity and self-criticism, and moderate correlations with self-kindness. These results align with those reported by Raes et al. (2010), which ranged between 0.25 and 0.66, likely due to differences in item selection criteria—Raes’ version prioritized high inter-item correlations, whereas this study selected items with the highest factor loadings per dimension.
Additionally, the correlation between the SCS-SF and the full SCS (ρ = .92, p < .001) closely mirrors results from Spanish populations (r = .92; Garcia-Campayo et al., 2014) and is slightly lower than Raes’ version (r = .98; Raes et al., 2010), confirming that the short form measures the same construct as the full scale.
Divergent validation with the negative affect subscale showed a moderate inverse correlation, consistent with previous research on depression, anxiety, and stress, which reported correlations ranging from −.38 to −.54 (Baggaley et al., 2024; Garcia-Campayo et al., 2014). Convergent validation with positive affect followed expected trends, with higher values than those found in Spanish validations (r = .39 for mindfulness; Garcia-Campayo et al., 2014) but lower than those reported in dementia populations for self-esteem (r = .68; Baggaley et al., 2024).
Regarding discriminant validity, the significant sex differences observed align with findings by Bratt and Fagerström (2020), suggesting that the Mexican SCS-SF can effectively distinguish group differences in self-compassion.
Finally, the test-retest correlation (ρ = .68, p < .01) suggests moderate to high stability (Hernández Lalinde et al., 2018), although slightly lower than values reported in Spanish populations (r = .89; Garcia-Campayo et al., 2014) and Raes’ 5-month assessment (r = .71; Raes, 2011). These results reinforce the notion that self-compassion is a malleable trait, playing a key role in coping with adversity (Muris & Otgaar, 2023).
Implications for Practice
The SCS-SF validated in Mexican students can have different applications in clinical practice, research, and community programs where self-compassion as a global concept needs to be measured quickly. In addition, it shows adequate internal consistency (>.76) for each of the dimensions obtained, allowing for quick and low-cost assessment of each of them. This is a major advantage over the first short version by Raes et al. (2010), who acknowledges that the subscales obtained have low internal consistency both in the Danish population (0.55 to 0.81) as in the English version (0.54 to 0.75). They therefore suggests that, if information from the subscales is required, it is better to use the long scale, aspects that could be maintained in the short versions validated in other countries that use the same questions (Babenko & Guo, 2019; Baggaley et al., 2024; Bratt & Fagerström, 2020; Garcia-Campayo et al., 2014; Maya et al., 2024; Villalón López et al., 2023).
Limitations
The authors admit several limitations of this study. First, the use of non-probabilistic convenience sampling drawn from a university population may restrain the generalizability of the findings. Future research must explore the applicability of this version in other age groups and diverse populations, allowing for the assessment of measurement invariance of the proposed model (Büyüköksüz et al., 2025; Meng et al., 2019). Additional limitations include the reliance on self-report measures and the fact that the proposal for a four-factor solution cannot be compared with previous studies. Furthermore, test-retest reliability was assessed over 3-months intervals; future research should examine temporal stability using shorter intervals.
Despite these limitations, the Mexican SCS-SF allows self-compassion to be assessed by summing the entire scale or its components independently, considering the three positive dimensions of self-compassion and its negative dimension, which could be assessed beyond a lack of compassion, as self-criticism. Also, the proposed scale exhibits adequate internal consistency at both the overall and subscale levels, making it a concise and reliable tool for assessing self-compassion. With a 54% reduction in the number of items, this version reduces administration time, making it more practical for use in various research and applied settings.
Conclusion
A validated version of the Self-Compassion Scale Short Form (SCS-SF) was developed for the Mexican population, providing an effective tool for evaluating self-compassion and its dimensions in research, clinical, and population-based applications. The SCS-SF consists of 12 items across four dimensions, preserving the structure of the positive dimensions from the original scale while consolidating all items related to a lack of self-compassion into a single dimension, termed “self-criticism” (Appendix A).
This version demonstrates strong psychometric properties, including suitable internal consistency, as well as convergent, divergent, and discriminant validity. Additionally, its 3-month stability suggests intermediate reliability over time, and a decrease in the total self-compassion score in the 3 months between measurements.
Footnotes
Appendix A
Acknowledgements
The authors would like to thank all the students who agreed to participate in this study, as well as the staff and interns of the Academic Services Coordination of the University Center of Health Sciences at the University of Guadalajara, for their support in the logistics of the fieldwork.
ORCID iDs
Ethical Considerations
This study is part of the project “Effect of the Body Project program and the promotion of self-compassion and good treatment on the eating behavior of university students” approved by the Research, Research Ethics and Biosafety Committees of the University Center for Health Sciences of the University of Guadalajara (approval code: CI-01024). The study was carried out in accordance with the ethical principles of the Declaration of Helsinki (World Medical Association, 2013), the Belmont Report principles, and Mexican regulations for health research, which consider this project as minimal risk, as it involved only the administration of psychological tests without behavioral interventions.
Consent to Participate
Participants received verbal and written information about the study objectives and procedures and provided inform consent prior to voluntarily participation. Students were explicitly informed that refusal or withdrawal would not result in academic or personal consequences. Confidentiality and anonymity were ensured, and contact information was used exclusively for follow-up purposes related to the study. Potential risks for participating were minimal and limited to possible fatigue during the application of the questionnaire, therefore, battery was designed to required no more than 30 min. Potential benefits included receiving individualized feedback and contributing to scientific knowledge aimed at promoting health and well-being among health university students. In accordance with the Council of International Organizations of Medical Sciences (CIOMS) Guidelines 25, the authors declare no conflict of interest, and report that no commercial funding or external corporate support was received for this study.
Author Contributions
Martha Betzaida Altamirano Martínez: Main idea, field work and writing of the document. Norma Alicia Ruvalcaba Romero: Main idea, revision of analysis and drafting. Saul Ramírez de los Santos: Review of information and drafting of the document. Claudia Unikel Santoncini: Methodology, document drafting and revision. Gabriela Macedo Ojeda: Document drafting and review. Oscar Gómez González: Data analysis, writing of results and review. All authors reviewed and agree with the version submitted for publication.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The authors declare that the database is available for any revision that may be required.