This study aimed to test the factorial structure and gender invariance of the Self-Compassion Scale Short Form (SCS-SF) in the Italian context (Aim 1), underlying relationships among SCS-SF items (Aim 2), questionnaire performance of SCS-SF (Aim 3), and correlations of SCS-SF (and its subscales) with cognitive, affective, and well-being variables (Aim 4).
Method
Six questionnaires including the SCS-SF and different correlates were respectively administered to six Italian convenience samples (total n = 2068). We performed confirmatory (ULS estimator) factor analysis (Aim 1), exploratory graph analysis and network analysis (Aim 2), multidimensional item response theory (IRT) analyses (Aim 3), and correlational analyses (Aim 4). We tested Aims 1–3 on the global sample and Aim 4 on the six samples separately.
Results
We found a two-factor (hierarchical) solution — which also had strict gender invariance — identifying Compassionate and Uncompassionate Self-Responding (CSR and USR), respectively composed of positive and negative items, while three- and six-factor structures did not hold (Aim 1). The SCS-SF network was structured into the CSR and USR clusters (Aim 2). CSR and — especially — USR had a satisfactory performance in terms of IRT discrimination and information (Aim 3). Correlational analyses supported the convergent and discriminant validity of the scales, showing the cognitive and emotion-regulation strategies associated with self-compassion and confirming that CSR and USR have slightly different correlates (Aim 4).
Conclusions
The Italian SCS-SF and its CSR and USR subscales are valid instruments. Future research and interventions can employ CSR and USR to assess different nuances of self-kindness and self-criticism.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Neff (2023) referred to self-compassion as “being present with our own pain, feeling connected to others who are also suffering, and understanding and supporting ourselves through difficult moments” (p. 195). In Neff’s (2003) conceptualization, self-compassion and the 26-item Self-Compassion Scale (SCS; Neff, 2003) included three bipolar dimensions defining six components: (a) Self-Kindness vs. Self-Judgment, that is a kind vs. a harsh attitude toward the self when facing difficulties; (b) Common Humanity vs. Isolation, which is recognizing difficulties and failings as part of the human condition and shared with other human beings vs. feeling alone in those difficulties; (c) Mindfulness vs. Over-Identification, which is reacting to difficulties with a balanced awareness vs. an obsessive, identified attitude.
According to Neff (2016), this framework allows researchers, depending on their aims and interests, either to consider self-compassion as a unitary — although multifaceted — construct or to distinguish between its six constituent components. This conceptualization was reflected in the Self-Compassion Scale (SCS), a 26-item instrument specifically devised to detect the six components proposed by Neff (2003). Validation studies (Neff et al., 2019) supported the use of the SCS to assess both the six components of self-compassion and a global score representing overall self-compassion. Using the global SCS score, self-compassion has been found positively associated with many positive psychological outcomes, from well-being (e.g., Zessin et al., 2015) to cognitive and emotional skills portraying better coping strategies and adaptive ways to deal with ordinary life (e.g., Ewert et al., 2021), as well as negatively related to psychopathology indexes (e.g., MacBeth & Gumley, 2012).
Over the years, however, growing evidence emerged in favor of a new conceptualization of self-compassion, based on the distinction between Compassionate and Uncompassionate Self-Responding (CSR and USR, respectively: Muris & Otgaar, 2020; Muris et al., 2021). According to this viewpoint, the adoption of a global SCS score could be inappropriate, as it risks hiding relevant differences between CSR (assessed by items referring to Self-Kindness, Common Humanity, and Mindfulness) and USR (assessed by items referring to Self-Judgment, Isolation, and Over-Identification). On the one hand, CSR appears grounded on gentle reactions toward the self and — compared to USR — has been found associated with more positive outcomes, such as positive emotions, self-esteem, empathy, need satisfaction, adoption of motivating styles, and prosociality (e.g., Lόpez et al., 2015; Moè, 2022; Moè & Katz, 2020; Sutton et al., 2018). On the other hand, USR — occasionally called self-criticism, self-derogation, or self-coldness — appears more as a maladaptive disposition, characterized by a negative relationship with the self, and by being associated with anxiety, need frustration, depression, burnout, and other psychopathological symptoms (e.g., Hayes et al., 2016; Moè & Katz, 2020; Muris et al., 2021). From a methodological point of view, this conceptualization has been supported by validation studies confirming a two-factor solution of the SCS, separating positive from negative items (e.g., Lόpez et al., 2015).
The ongoing debate on the internal structure of self-compassion also regards the Self-Compassion Scale-Short Form (SCS-SF), a 12-item version of the SCS developed by Raes et al. (2011). In their validation, involving Dutch and English samples, Raes et al. (2011) found the same structure of the SCS, but Cronbach’s alphas for the six subscales were low (as two items for each component were used), thereby supporting the use of a global score. Over the years, the SCS-SF has been validated in several countries, but the obtained factorial solutions were quite inconsistent: in Bangladesh, Rahman et al. (2023) found support for a two-factor solution identifying CSR and USR; in Spanish (Garcia-Campayo et al., 2014) and Brazilian data (Rocha et al., 2022), a six correlated-factor solution emerged, although Rocha et al. (2022) found that a one-factor solution was more satisfactory; in Chinese data a three-factor solution emerged (Meng et al., 2019), while Alfonsson et al. (2023) detected five SCS-SF factors in Swedish data. The SCS-SF has been validated also in specific populations, such as children aged 8–12 years (Sutton et al., 2018) and psychotherapy clients (Hayes et al., 2016): in both these studies, the presence of two factors, CSR and USR, has been supported.
The heterogeneity of SCS-SF factorial solutions has some implications for research and practice. For instance, the distinction among six factors allows researchers to precisely detect which specific component is involved in the investigated processes: as an example, Browne et al. (2022) found that only Self-Judgment — among the six subscales — exacerbated the negative association between experiences of discrimination and mental health. Relying on a three-factor solution, it is possible to examine the role of specific bipolar dimensions, without distinguishing between positive and negative aspects: for instance, Fuochi et al. (2018a) found that only Common Humanity vs. Isolation — among the three subscales — was positively associated with empathic concern and positive intergroup attitudes. Finally, a two-factor solution, distinguishing between CSR and USR, may allow researchers to disentangle the association between self-compassion and psychopathology (e.g., Muris et al., 2021).
Overall, this unclear picture highlights the need for further investigation into the internal structure of the SCS-SF and its correlates. While it is possible that cross-cultural differences or sample characteristics (e.g., gender, age, presence of psychological symptoms; Neff, 2023) affect its factorial structure, it could also be argued that the SCS-SF relies on a different internal structure compared to the full SCS. More evidence is needed to understand the underlying structure and performance of the SCS-SF, and the different meanings and correlates of its subdimensions.
Thus, the present multi-sample study (n = 2068, six Italian samples with different correlates) aimed to (1) validate the Italian version of the SCS-SF, testing its factorial structure and measurement invariance by gender in a large Italian sample (Aim 1); (2) study — to our knowledge for the first time — the topology of SCS-SF, testing the spacing, links, and relevance of SCS-SF items via network analysis (Aim 2); (3) verify the performance of SCS-SF thanks to item response theory (IRT) analyses (Aim 3); (4) assess the nomological net of SCS-SF scores, by testing the correlates, and convergent and discriminant validity, of the scale (Aim 4). A large array of correlates was considered, including dispositions portraying relationships with oneself and with others, cognitive styles and tendencies, dispositions related to awareness, emotion regulation abilities, and well-being dimensions. We explored a wide nomological net of SCS-SF to gain a deeper understanding of the meaning of the construct and its subscales.
Method
Participants
We collected six samples of Italian adults, involving a total of 2068 respondents, who were recruited through social media or word of mouth and participated on a voluntary basis. Their characteristics are described in Table 1.
Table 1
Characteristics of the samples, and descriptive statistics and reliability (Cronbach’s alpha, McDonald’s omega) of SCS-SF
Sample A
n = 336
Sample B
n = 328
Sample C
n = 431
Sample D
n = 433
Sample E
n = 206
Sample F
n = 334
Percent of women
61.90%
56.40%
62.88%
77.37%
51.94%
69.76%
Age, M(SD)
31.50(12.70)
32.27(14.43)
32.01(11.89)
30.60(13.89)
31.10 (12.68)
32.72(13.83)
Occupation
University student
40.48%
35.06%
41.00%
54.04%
37.38%
40.72%
Employed
47.02%
51.52%
47.80%
34.18%
53.88%
49.10%
Education
Middle school
5%
6%
4%
11%
10%
20%
High school
46%
54%
42%
53%
54%
44%
Bachelor’s or higher degree
49%
40%
54%
36%
36%
36%
SCS-SF: M(SD)
Total score
2.85(0.75)
2.86(0.65)
3.10(0.76)
2.89(0.77)
2.84(0.67)
2.93(0.61)
Positive items (CSR)
3.07(0.81)
2.96(0.72)
3.22(0.78)
3.06(0.75)
2.99(0.70)
3.07(0.71)
Negative items (USR)
3.38(0.95)
3.25(0.90)
3.02(0.98)
3.27(1.01)
3.31(0.89)
3.21(0.87)
SCS-SF: α, ω
Total score
0.87, 0.90
0.81, 0.87
0.87, 0.91
0.86, 0.90
0.79, 0.85
0.75, 0.82
Positive items (CSR)
0.81, 0.87
0.76, 0.85
0.80, 0.88
0.76, 0.85
0.71, 0.79
0.72, 0.81
Negative items (USR)
0.86, 0.92
0.84, 0.89
0.87, 0.92
0.87, 0.91
0.84, 0.89
0.80, 0.88
Other gender categories are man and non-binary (3 individuals over the six samples). Other occupation categories are unemployed and retired
Procedure
Following recommendations to have reliable estimates from both confirmatory factor (Comrey & Lee, 1992) and IRT (Dai et al., 2021) analyses, we aimed to collect SCS-SF data from at least 1200 participants, whereas associations with correlates could be assessed with a smaller sample size. Therefore, we built six different online questionnaires, each including the SCS-SF, and a set of self-report correlates of self-compassion that was different for each sample; measures collected in Samples A-F are reported in Table S1 (Supplementary Materials).
Twenty research assistants (recruiters) disseminated the links of the online questionnaires through social media platforms and personal acquaintances; each research assistant had only one link, so on average each data collection was conducted by three or four recruiters. All questionnaires started with the informed consent form (approved by the Ethics Committee of the University of Padua), explaining to respondents the study’s purposes, the privacy, anonymity, and confidentiality of their responses and of the treatment of their data (i.e., respondents could not be identified by their answers), and the possibility of withdrawing at any time. Then, participants individually completed the online questionnaires including the SCS-SF and the other measures, in Italian. If participants had queries regarding the data collection, they were informed that they could contact the corresponding author of this paper, using the contact details reported at the end of the informed consent form. Participants did not receive any compensation, such as academic credit or monetary rewards, for their participation in the study.
Measures
The items of the SCS-SF (Table 2) were taken from the Italian validated version of the SCS (Veneziani et al., 2017), which is the 26-item scale (Neff, 2003). Regarding correlates, in the few cases in which a validated Italian version of a scale was not available, items were translated adopting a forward and backward translation procedure, to preserve their original meaning. Consistent with test translation guidelines (e.g., Gudmundsson, 2009), scales were translated into Italian by the authors, and then a bilingual (English-Italian) professional (native Italian speaker) performed the backward translation, which confirmed the original meaning of the items. Reliability statistics of all scales are reported in Table 1 for SCS-SF (together with means and standard deviations), and in Table S1 (Supplementary Materials) for all the other scales.
Table 2
Instructions and items of the Italian version of the SCS-SF
Per ogni domanda indica quanto spesso ti comporti nella maniera indicata. La scala di risposta va da 1 = Quasi mai a 5 = Quasi sempre
Item
Component
Factor
Italian version
English version (excerpts)
1
OI
USR
Quando non riesco in qualcosa di importante per me, sono logorato/a da sentimenti di inadeguatezza
… consumed by feelings of inadequacy
2
SK
CSR
Cerco di essere comprensivo/a e paziente verso quegli aspetti della mia personalità che non mi piacciono
… understanding and patient…
3
M
CSR
Quando accade qualcosa di doloroso, cerco di tenere una visione equilibrata della situazione
… take a balanced view of the situation
4
I
USR
Quando mi sento giù, ho l’impressione che la maggior parte delle altre persone sia probabilmente più felice di me
… most other people are probably happier…
5
CH
CSR
Cerco di vedere i miei difetti come parte della condizione umana
… my failings as part of the human condition
6
SK
CSR
Quando sto attraversando un momento molto difficile, do a me stesso/a la cura e la tenerezza di cui ho bisogno
… give myself the caring and tenderness…
7
M
CSR
Quando qualcosa mi sconvolge cerco di tenere le mie emozioni in equilibrio
… try to keep my emotions in balance
8
I
USR
Quando non riesco in qualcosa di importante per me, tendo a sentirmi solo/a nel mio fallimento
… tend to feel alone in my failure
9
OI
USR
Quando mi sento giù, tendo a ossessionarmi e fissarmi su tutto ciò che è sbagliato
… tend to obsess and fixate…
10
CH
CSR
Quando mi sento inadeguato/a, in qualche modo cerco di ricordare a me stesso/a che i sentimenti di inadeguatezza sono condivisi dalla maggior parte delle persone
… feelings of inadequacy are shared by most people
11
SJ
USR
Sono critico/a e severo/a nei confronti dei miei difetti e delle mie inadeguatezze
… disapproving and judgmental …
12
SJ
USR
Sono intollerante e impaziente verso quegli aspetti della mia personalità che non mi piacciono
… intolerant and impatient…
Note. USR = Uncompassionate Self-Responding; CSR = Compassionate Self-Responding; OI = Over-Identification; SK = Self-Kindness; M = Mindfulness; I = Isolation; CH = Common Humanity; SJ = Self-Judgment
Self-Compassion
We employed the SCS-SF (Raes et al., 2011; response scale from 1 = almost never to 5 = almost always). Because results of the Aim 1 section of the present paper supported a two-factor solution, we computed both the total score and the subscales’ scores, i.e., Compassionate and Uncompassionate Self-Responding (CSR and USR).
Self-Oriented Dispositions
We measured anxious and avoidant adult attachment with the Experiences in Close Relationships Scale (ECR; Brennan et al., 1998), and entitlement with the Psychological Entitlement Scale (Campbell et al., 2004; Italian version: Boin & Voci, 2019). We measured positive and negative emotional reactions toward the self (e.g., “friendly toward myself,” “angry with myself”) with the Self-Other Four Immeasurables Scale (SOFI; Kraus & Sears, 2009).
Other-Oriented Dispositions
We administered the perspective taking, empathic concern, and personal distress dimensions of dispositional empathy taken from the Interpersonal Reactivity Index (IRI; Davis, 1983; Italian version: Albiero et al., 2006) and the Gratitude Questionnaire (GQ-6; McCullough et al., 2002; Italian version: Fuochi et al., 2018b). We measured positive and negative emotional reactions toward others (e.g., “friendly toward others,” “angry with others”) with the Self-Other Four Immeasurables Scale (SOFI; Kraus & Sears, 2009).
Awareness-Related Dispositions
We assessed dispositional mindfulness (measuring separately Acting with awareness, Nonjudging, Nonreactivity, Observing, and Describing) with the Five Facet Mindfulness Questionnaire (FFMQ; Baer et al., 2008; Italian version: Giovannini et al., 2014). We measured decentering — which is the tendency to disidentify from internal experience — with the Decentering subscale of the Experiences Questionnaire (Fresco et al., 2007), and nonattachment — which is a flexible, balanced way of relating to one’s experiences without clinging to or suppressing them — with the Nonattachment Scale (NAS-7; Sahdra et al., 2016).
Cognitive Tendencies
We measured rumination with the self-rumination subscale of the Rumination-Reflection Questionnaire (Trapnell & Campbell, 1999; Italian version: Vannucci & Chiorri, 2018), thoughts suppression with the White Bear Suppression Inventory (Wegner & Zanakos, 1994; Italian version: Pica et al., 2015), and cognitive flexibility with the Cognitive Flexibility scale (Martin & Rubin, 1995).
Emotion Regulation
We employed the Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004; Italian version: Giromini et al., 2012), the Emotion Regulation Questionnaire (Gross & John, 2003; Italian version: Balzarotti et al., 2010), and the Cognitive Emotion Regulation Questionnaire (Garnefski et al., 2001) to assess many adaptive and maladaptive emotion-regulation strategies. We employed the Affective Control Scale (Williams et al., 1997) to measure fear of specific emotions.
Well-Being Indicators
We measured positive and negative emotions with the Positive and Negative Affect Schedule (Watson et al., 1988; Italian version: Terracciano et al., 2003), life satisfaction with the Satisfaction with Life Scale (Diener et al., 1985), and subjective stress with the Perceived Stress Scale (Cohen et al., 1983). We assessed eudaimonic well-being with the Presence of Meaning subscale of the Meaning in Life Questionnaire (Steger et al., 2006), and two versions, with different lengths, of the Psychological Well-Being scale (PWB-18, Ryff & Keyes, 1995; PWB-54, Ryff, 1989; Italian version of the items: Ruini et al., 2003).
Data Analyses
To address Aim 1, we employed confirmatory factor analysis (CFA) with the unweighted least squares (ULS) estimator for ordinal data. We tested the factorial structures explored in previous studies, i.e., two-factor, three-factor, six-factor, and unidimensional models, starting from the one suggested by the scree plot, which was a two-factor solution. Lastly, we tested measurement invariance by gender of the two-factor SCS-SF model (with correlated factors and ULS estimator), assessing configural invariance (i.e., no equality constraints), metric invariance (i.e., equality of loadings across groups), scalar invariance (i.e., equality of loadings and intercepts across groups), and strict invariance (i.e., equality of loadings, intercepts, and residual variances across groups). As recommended by the literature (e.g., Cheung & Rensvold, 2002), we considered measurement invariance across groups acceptable when changes among types of invariance (i.e., from configural to metric, from metric to scalar, from scalar to residual invariance) did not exceed (in absolute value) 0.01 for CFI and TLI, and 0.015 for RMSEA.
To address Aim 2, we conducted network analysis on SCS-SF items, with the Gaussian-Markov random field estimation (EBIC-glasso method): First we performed an exploratory graph analysis (EGA; Walktrap algorithm) to reliably assess and visualize the clustering structure of items (nodes), and then we plotted the network with the Fruchterman-Reingold algorithm to report edge weights, i.e., the size of the associations among nodes. We also calculated five centrality measures: three indexes of strength, computed as the sum of a node’s edge weights with all other nodes in the network (strength), other nodes within its own cluster (in-strength), and nodes in other clusters (out-strength); betweenness, which measures how often a node is in paths between other nodes; and closeness, which assesses how strongly a node is directly and indirectly connected to all the other nodes. Lastly, we checked the stability of the network with bootstrapping techniques, applied to the EGA network and its clusters, edge weights, and centrality indexes.
To address Aim 3, we employed multidimensional item response theory (MIRT) with the graded response model for polytomous items (Samejima, 1969), setting the model validated by CFA — i.e., a CSR-USR two-factor structure with correlated factors — and evaluating the items’ discrimination and threshold parameters, as well as test information. Discrimination parameters represent the items’ ability to differentiate among individuals with different levels of the latent trait, whereas item thresholds are the crossover points between two response categories along the latent trait. According to Baker (2001), the differentiation ability of the item is moderate when the discrimination parameter is in the range 0.65–1.34, and low (high) when it is under (above) this range. Test information portrays the accuracy of the scale in measuring any value of the latent trait. We also evaluated local independence by employing Yen’s Q3 statistic, which measures the correlation between the residuals of item pairs, and identifying pairs of items showing a positive residual above 0.20 (Christensen et al., 2017). Lastly, we conducted an IRT graded response model on the total set of SCS-SF items (with USR items reverse-coded), to compare the psychometric performance of the unidimensional model with the performance of the two-factor model validated by CFA, focusing on the discrimination ability of the items.
To address Aim 4, we computed Pearson correlations between SCS-SF, CSR, USR scores and the correlates. We evaluated correlations based on sign, statistical significance, and size; for the size of correlation coefficients, we relied on Cohen et al. (1983), according to whom a correlation is small if the coefficient in absolute value is between 0.10 and 0.30, moderate if it is between 0.30 and 0.50, and strong if it is equal to or above 0.50. Analyses addressing Aims 1, 2, and 3 were performed on the aggregate sample, to have the maximum sample size possible (n = 2068), whereas Aim 4 was addressed within the specific subsamples (Table 1), which included different correlates. All analyses were performed with R (R Core Team, 2024).
Results
Aim 1: Italian Validation of the SCS-SF
Mardia’s test on the items showed a deviation from multivariate normality (multivariate skewness = 1177.76, p < 0.001; multivariate kurtosis = 32.86, p < 0.001). Before CFA, we conducted a scree test on SCS-SF, computing the eigenvalues of factors (first, 3.93; second, 1.25; third, 0.40; fourth, 0.23; fifth, 0.05) and plotting them in a scree plot (Figure S1, Supplementary Materials). As only the first two eigenvalues were greater than 1, the scree test suggested retaining two factors.
Therefore, following previous research (e.g., Muris et al., 2021; Sutton et al., 2018) and the scree plot results, we tested a factorial structure with two correlated factors, one for positive items (measuring CSR) and one for negative items (measuring USR), with CFA. We employed the ULS estimator, consistent with Mardia’s test results and with the ordinal nature of the 5-point response scale. In two-factor solutions, estimating the correlation between the two factors is the same as having a general, second-order factor. This solution (Fig. 1) produced satisfactory fit indexes (CFI = 0.98; TLI = 0.97; RMSEA = 0.06; SRMR = 0.06; factors’ correlation = 0.25, p < 0.001; factor loadings in Table 3). Three-factor and six-factor CFA models did not converge.
Fig. 1
Confirmatory factor analysis of the SCS-SF. SK, Self-Kindness; SJ, Self-Judgment; CH, Common Humanity; I, Isolation; M, Mindfulness; OI, Over-Identification; CSR, Compassionate Self-Responding; USR, Uncompassionate Self-Responding. Standardized parameters
Table 3
Standardized factor loadings and MIRT parameters of SCS-SF items (correlated two-factor structure, CSR and USR)
We also tested a unidimensional structure with CFA, but fit indexes were not satisfactory (CFI = 0.89; TLI = 0.86; RMSEA = 0.14; SRMR = 0.12). Considering that CSR and USR — the two factors supported by CFA — respectively included only positively worded items (CSR) and only negatively worded items (USR), as a final option, we tested a factorial structure with a general factor and a method factor with all negative items loading on it. In scales with many negatively worded items — like some self-esteem measures (Motl & DiStefano, 2002) — negatively worded items can generate a method effect due to their different interpretations and response styles compared to positively worded items. This method effect, which is unrelated to the measured construct, can be handled with the method factor. Performing a unidimensional model with a method factor on the SCS-SF also produced satisfactory fit indexes: CFI = 0.98; TLI = 0.97; RMSEA = 0.07; SRMR = 0.05. In summary, the Italian SCS-SF strongly supports the use of two subscales, i.e., CSR and USR, and a general self-compassion score computed as the mean of CSR and (reverse-coded) USR.
Lastly, we tested measurement invariance by gender (women: n = 1339; men: n = 716, including three respondents who described themselves as non-binary), and found that the differences between configural and metric invariance (ΔCFI = 0.977–0.975 = 0.002; ΔTLI = 0.971–0.972 = − 0.001; ΔRMSEA = 0.078–0.077 = 0.001), between metric and scalar invariance (ΔCFI = 0.975–0.971 = 0.004; ΔTLI = 0.972–0.973 = − 0.001; ΔRMSEA = 0.077–0.076 = 0.001), and between scalar and strict invariance (ΔCFI = 0.971–0.971 = 0.000; ΔTLI = 0.973–0.973 = 0.000; ΔRMSEA = 0.076–0.075 = 0.001) were all below the thresholds. Therefore, the two-factor model supported strict measurement invariance (i.e., equality of loadings, intercepts, and residual variances) by gender.
Aim 2: Network Analysis of SCS-SF Items
Exploratory graph analysis (EGA) reliably identified two communities within SCS-SF nodes (see Figure S2 in Supplementary Materials), corresponding to CSR (i.e., nodes representing positive items) and USR (i.e., nodes representing negative items): these two clusters were connected only by some weak, negative links. The network plotted with Fruchterman-Reingold algorithm (Fig. 2; continuous edges: positive relations; dashed edges: negative relations) had the same clustering structure of the EGA network.
Fig. 2
The network of SCS-SF items. Nodes represent single items of the SCS-SF. SK, Self-Kindness; SJ, Self-Judgment; CH, Common Humanity; I, Isolation; M, Mindfulness; OI, Over-Identification. Continuous edges, positive associations; dashed edges, negative associations
×
In this self-compassion network, the two strongest edges were the one linking the two Self-Judgment (SJ) items and the one between the two Mindfulness (M) items. Interestingly, Over-Identification (OI), and Isolation (I) items were all strongly associated between each other; similarly, Self-Kindness (SK) and Common Humanity (CH) items were all associated between each other, thereby suggesting that the inner structure of SCS-SF is not grounded on the six or three (SJ vs. SK, M vs. OI, CH vs. I) aspects of self-compassion originally hypothesized for the longer scale. Therefore, the network analysis also suggests that the two-factor structure (CSR and USR) prevails over the three-factor or six-factor structures, at least in the Italian context.
Regarding centrality indexes (Table 4), results showed that SK1 and SJ2 were the nodes with higher closeness, while OI2 was the node with higher strength and betweenness. In-strength and out-strength results showed that all nodes were strongly connected to nodes within their own (either CSR or USR) cluster. In-strength was maximum for M2 and CH1 in the CSR cluster, and for OI2 and I2 in the USR cluster; out-strength, which identifies nodes that act as bridges between the clusters, was not as high as in-strength, but was relatively higher for SK1 and SJ2. Indeed, the edge between these two nodes was the connection between the CSR cluster and the USR cluster (see Fig. 2 and Figure S2 in Supplementary Materials). All nodes were important to the self-compassion network, in different ways.
Table 4
Scores of centrality indexes from network analysis
Node
Strength
In-strength
Out-strength
Betweenness
Closeness
Self-Kindness 1
− 0.327
0.574
− 0.187
1.097
1.175
Self-Kindness 2
− 0.077
0.653
− 0.147
− 0.115
0.583
Mindfulness 1
− 0.737
0.706
− 0.080
− 0.808
− 0.121
Mindfulness 2
0.511
0.781
− 0.087
0.404
0.583
Common Humanity 1
0.357
0.781
− 0.034
− 0.462
− 0.803
Common Humanity 2
− 1.601
0.570
− 0.023
− 1.155
− 1.675
Self-Judgment 1
0.158
0.839
− 0.029
0.924
0.895
Self-Judgment 2
1.038
0.780
− 0.196
1.097
1.175
Over-Identification 1
− 0.370
0.753
− 0.049
1.155
− 1.044
Over-Identification 2
1.672
0.965
− 0.116
1.617
0.898
Isolation 1
− 1.570
0.567
− 0.119
− 1.155
− 0.899
Isolation 2
0.946
0.878
− 0.049
− 0.289
− 0.766
Lastly, the assessment of the stability of the network proved that it was stable and reliable. First, in 90.6% of the bootstrapped samples, the EGA network was estimated to have two dimensions, 95% CI = [1.28, 2.72], while three and four dimensions were estimated respectively in 7.6% and 1.8% of bootstrapped samples: the strongly higher frequency of two − compared to three and four − dimensions suggests that it represents the true dimensionality of the data. Second, all SCS-SF nodes were assigned to their (CSR or USR) cluster in 100% of bootstrapped samples, with slightly lower values for the Self-Judgment nodes (97%) and the Mindfulness nodes (92%; see Figure S3, Supplementary Materials). Third, edge-weight accuracy and stability of centrality indexes − especially strength − were satisfactory (see Figure S4 and Figure S5, Supplementary Materials).
Aim 3: IRT Analysis of SCS-SF
MIRT analyses were conducted setting as a latent model a two-factor structure with CSR and USR as correlated factors, composed of SCS-SF positive and negative items respectively, consistent with CFA and network analysis results. Item parameters derived from MIRT are reported in Table 3: both CSR and USR items had high discrimination parameters, but higher for USR than for CSR items, especially for the items Over-Identification 2 and Isolation 2. Only the item Common Humanity 2 (“When I feel inadequate in some way, I try to remind myself that feelings of inadequacy are shared by most people”) had moderate discrimination. The scores of the thresholds were quite symmetrical around zero for both CSR and USR, with a narrower distribution for USR items. Overall, both subscales displayed a satisfactory performance, with USR items showing a very strong ability to differentiate among individuals with different levels of the latent trait.
Figure 3 reports the test information curve of SCS-SF, with ϴ1 and ϴ2 being the latent traits for CSR and USR, respectively. The information given by SCS-SF, namely I(ϴ), reached its peak when both CSR and USR latent traits were in their median values; on the contrary, SCS-SF provided less information (see the valleys in Fig. 3) at extreme values of CSR and USR. Keeping the other latent trait constant, the USR subscale provided more information than the CSR subscale, but especially on the median values of its latent trait. Overall, SCS-SF seems to estimate more accurately the USR latent trait, with higher precision on the central points of the distribution.
Fig. 3
Test information function of SCS-SF (CSR = ϴ1; USR = ϴ2)
×
Then, we tested local independence: Yen’s Q3 statistic exceeded 0.20 only for one pair of items (Self-Judgment 1 and Self-Judgment 2, Q3 = 0.235), and the average Q3 value across all item pairs was − 0.062, thereby suggesting that the assumption of local independence was mainly respected (Christensen et al., 2017). Lastly, we performed a unidimensional graded response model on SCS-SF items (with USR items reverse-coded); item parameters are reported in Table S2 (Supplementary Materials). Discrimination results were lower for the unidimensional SCS-SF model compared to the correlated two-factor SCS-SF model reported in Table 3, with CSR items being particularly affected. In fact, all CSR items showed moderate discrimination in the unidimensional model, whereas in the two-factor structure all CSR items except Common Humanity 2 had high discrimination. In summary, the performance of SCS-SF was higher in the two-factor model compared to the unidimensional model.
Aim 4: Correlates of SCS-SF
Table 5 reports the correlations that the SCS-SF total score, and the CSR and USR scores, have with the correlates measured in samples A–F.
Table 5
Correlates of SCS-SF total score, CSR, and USR
Scales
Sample
SCS-SF total
CSR
USR
Self-oriented dispositions
Anxious attachment
C
− 0.48***
− 0.28***
0.52***
Avoidant attachment
C
− 0.27***
− 0.25***
0.21***
Psychological entitlement
A, B
− 0.01, − 0.08
0.08, − 0.04
0.09, 0.08
SOFI-Self positive
B
0.33***
0.33***
− 0.22***
SOFI-Self negative
B
− 0.38***
− 0.27***
0.34***
Other-oriented dispositions
IRI Perspective taking
A, B
0.15**, 0.18***
0.18**, 0.33***
− 0.09, 0.00
IRI Empathic concern
A, B
0.00, − 0.07
0.04, 0.08
0.03, 0.16**
IRI Personal distress
A, B
− 0.44***, − 0.38***
− 0.33***, − 0.28***
0.41***, 0.33***
Gratitude
B
0.27***
0.35***
− 0.11*
SOFI-Other positive
B
0.11*
0.19***
− 0.01
SOFI-Other negative
B
− 0.19***
− 0.15**
0.16**
Awareness dispositions
FFMQ Awareness
C, D, F
0.40***, 0.38***, 0.22***
0.27***, 0.22***, 0.06
− 0.41***, − 0.40***, − 0.26***
FFMQ Nonjudging
C, D, F
0.55***, 0.49***, 0.44***
0.38***, 0.26***, 0.14*
− 0.55***, − 0.56***, − 0.51***
FFMQ Nonreactivity
C, D, F
0.50***, 0.51***, 0.39***
0.49***, 0.53***, 0.43***
− 0.39***, − 0.38***, − 0.20***
FFMQ Observing
C, D, F
0.11*, 0.15**, 0.09
0.20***, 0.25***, 0.25***
− 0.01, − 0.05, 0.08
FFMQ Describing
C, D, F
0.41***, 0.34***, 0.09
0.40***, 0.27***, 0.12*
− 0.32***, − 0.31***, − 0.03
Decentering
C
0.69***
0.65***
− 0.55***
Nonattachment
A, B
0.61***, 0.52***
0.56***, 0.48***
− 0.48***, − 0.37***
Cognitive tendencies
Rumination
E, F
− 0.62***, − 0.43***
− 0.35***, − 0.11*
0.65***, 0.52***
Thoughts suppression
E
− 0.43***
− 0.19**
0.50***
Cognitive flexibility
E
0.38***
0.36***
− 0.28***
Emotion regulation
DERS Nonacceptance
C, D
− 0.58***, − 0.56***
− 0.39***, − 0.37***
0.60***, 0.58***
DERS Goals
C, D
− 0.48***, − 0.47***
− 0.29***, − 0.37***
0.52***, 0.45***
DERS Impulse
C, D
− 0.55***, − 0.49***
− 0.41***, − 0.38***
0.53***, 0.46***
DERS Awareness
C, D
− 0.30***, − 0.29***
− 0.40***, − 0.36***
0.15**, 17***
DERS Strategies
C, D
− 0.65***, − 0.65***
− 0.44***, − 0.48***
0.65***, 0.62***
DERS Clarity
C, D
− 0.49***, − 0.38***
− 0.38***, − 0.30***
0.45***, 0.35***
ERQ Reappraisal
C
0.19***
0.28***
− 0.07
ERQ Suppression
C
− 0.32***
− 0.21***
0.32***
CERQ Self-blame
E
− 0.34***
− 0.10
0.44***
CERQ Acceptance
E
0.11
0.21**
0.00
CERQ Rumination
E
− 0.33***
− 0.08
0.43***
CERQ Refocusing
E
0.30***
0.29***
− 0.23***
CERQ Planning
E
0.26***
0.34***
− 0.13
CERQ Reappraisal
E
0.32***
0.35***
− 0.20**
CERQ Perspective
E
0.33***
0.37***
− 0.21**
CERQ Catastrophizing
E
− 0.35***
− 0.20**
0.37***
CERQ Other-blame
E
− 0.15*
− 0.03
0.21**
ACS Anger
D, F
− 0.37***, − 0.29***
− 0.23***, − 0.09
0.39***, 0.33***
ACS Positive affect
D, F
− 0.41***, − 0.23***
− 0.29***, − 0.06
0.41***, 0.28***
ACS Depressed mood
D, F
− 0.68***, − 0.52***
− 0.47***, − 0.26***
0.68***, 0.51***
ACS Anxiety
D, F
− 0.66***, − 0.52***
− 0.49***, − 0.25***
0.64***, 0.53***
Well-being dimensions
Positive affect
D, E
0.45***, 0.39***
0.38***, 0.29***
− 0.39***, − 0.36***
Negative affect
D, E
− 0.55***, − 0.43***
− 0.42***, − 0.19***
0.53***, 0.49***
Satisfaction with life
D, E, F
0.48***, 0.36***, 0.34***
0.38***, 0.30***, 0.27***
− 0.44***, − 0.30***, − 0.26***
Perceived stress
A, D
− 0.65***, − 0.64***
− 0.46***, − 0.44***
0.63***, 0.65***
Presence of meaning
B
0.35***
0.32***
− 0.25***
PWB-18
B, D, E
0.41***, 0.54***, 0.38***
0.48***, 0.41***, 0.29***
− 0.21***, − 0.51***, − 0.34***
PWB-54 Autonomy
F
0.32***
0.26***
− 0.24***
PWB-54 Self-acceptance
F
0.49***
0.33***
− 0.42***
PWB-54 Environmental mastery
F
0.43***
0.27***
− 0.39***
PWB-54 Personal growth
F
0.11*
0.03
− 0.13*
PWB-54 Positive relations
F
0.33***
0.17**
− 0.32***
PWB-54 Purpose in life
F
0.15**
0.03
− 0.18***
SOFI, Self-Other Four Immeasurables Scale; IRI, Interpersonal Reactivity Index; FFMQ, Five Facet Mindfulness Questionnaire; DERS, Difficulties in Emotion Regulation Scale; ERQ, Emotion Regulation Questionnaire; CERQ, Cognitive Emotion Regulation Questionnaire; ACS, Affective Control Scale; PWB, Psychological Well-Being Scale
*p < 0.05, ** p < 0.01, *** p < 0.001
Individuals scoring higher on SCS-SF and CSR, and lower on USR, reported lower anxious and avoidant attachment, and more positive, less negative emotional reactions toward the self (SOFI-self). All these associations had moderate size. Psychological entitlement was unrelated to SCS-SF, CSR, and USR, thereby suggesting that self-compassion is not self-indulgence.
Correlations with other-oriented dispositions had overall smaller sizes: in this set, the strongest correlates were personal distress, which was negatively associated with SCS-SF and CSR and positively associated with USR, and gratitude, which was positively associated with SCS-SF and CSR and very weakly associated with USR. Empathic concern was unrelated to aspects of self-compassion, whereas perspective taking held small but positive associations with SCS-SF and CSR, but not with USR. Correlations were weak also for SOFI-Other, with SCS-SF and CSR being negatively associated — and USR positively associated — with negative emotional reactions toward others, whereas only CSR was positively associated with positive emotional reactions toward others.
Correlations were stronger for awareness-related dispositions — except the Observing facet of FFMQ, which was weakly and positively associated with CSR and unrelated to SCS-SF and USR. Mindfulness facets, decentering, and nonattachment were strongly and positively associated with SCS-SF and CSR, and negatively associated with USR. As for cognitive tendencies, SCS-SF and CSR were negatively associated with rumination and thoughts suppression, which both had positive and strong associations with USR. SCS-SF and CSR were also positively — and USR negatively — correlated with cognitive flexibility.
SCS-SF and CSR were negatively associated — while USR was positively associated — with difficulties in emotion regulation: correlations were strong, except for the weak positive association between DERS Awareness and USR. Similarly, USR was positively correlated with other maladaptive emotion-regulation strategies, such as CERQ Self-Blame, Rumination, and Catastrophizing, and ERQ suppression. These variables were instead negatively associated with SCS-SF, and only in some cases with CSR. An inverse correlation pattern held for adaptive regulation strategies, i.e., CERQ Refocusing, Planning, Reappraisal, and Perspective, which were positively associated with SCS-SF and CSR, and negatively with USR. It is interesting to notice that some ERQ and CERQ strategies were associated with USR but not CSR, and vice versa: USR — but not CSR — was positively correlated with CERQ Self-Blame, CERQ Other-Blame, and CERQ Rumination, whereas CSR — but not USR — was positively correlated with ERQ Reappraisal, CERQ Acceptance, and CERQ Planning. Consistent with the correlational patterns regarding emotion regulation, SCS-SF and CSR were negatively — and USR positively — associated with fear of emotions (i.e., Affective Control), especially with fear of negative emotions.
Lastly, SCS-SF and CSR were positively — and USR negatively — correlated with positive affect, life satisfaction, presence of meaning, and all dimensions of psychological well-being except personal growth. Correlations had opposite signs for negative affect and perceived stress.
Discussion
The goal of this multi-sample study was to explore in-depth the validity, performance, and nomological net of the SCS-SF, through four specific aims: testing the factorial structure and gender invariance of its Italian version (Aim 1); identifying the network of SCS-SF items (Aim 2); understanding the performance of SCS-SF with IRT tools (Aim 3); assessing the correlates of SCS-SF, considering a large set of variables, spanning from self- and other-oriented dispositions to cognitive and emotion-regulation strategies and well-being (Aim 4).
Regarding Aim 1, CFA strongly supported a correlated two-factor solution identifying Compassionate and Uncompassionate Self-Responding (CSR and USR), portraying positive and negative self-compassion items, respectively. The two-correlated factor solution also proved to have strict measurement invariance by gender, thereby supporting the validity of CSR and USR scores for different gender groups. Having correlated factors in a two-factor solution is equal to having a hierarchical two-factor solution; therefore, our results also supported the use of the total SCS-SF score, as suggested by the first validation by Raes et al. (2011). Our validation results were consistent with several — but not all — previous studies validating the SCS-SF in various samples and countries (Hayes et al., 2016; Rahman et al., 2023; Sutton et al., 2018), and with other studies finding a two-factor solution for the longer SCS (e.g., Lόpez et al., 2015). In our dataset, the six-factor solution found in a couple of SCS-SF validation studies (Garcia-Campayo et al., 2014; Rocha et al., 2022) did not reach convergence, thereby supporting only the computation and use of CSR, USR, and the total SCS-SF scores. The inconsistent results of the various SCS-SF validations could suggest that in different cultural contexts, self-compassion can acquire specific nuances due to culturally shaped meanings. These cultural differences could affect associations among items, and therefore result in certain factorial structures being appropriate for some contexts, but not others.
Regarding Aim 2, network analysis performed with EGA showed that the items of the SCS-SF were grouped into two separate clusters replicating the two-factor structure found in CFA, i.e., the CSR vs. USR structure. Consistent with our CFA results supporting neither the six-factor nor the three-factor structures, in the SCS-SF network only two of the original subscales emerged within the two-factor grouping of nodes: Self-Judgment and Mindfulness, which had strong associations between their respective items (SJ1 and SJ2, M1 and M2). The other four subscales were placed in the network only following the CSR-USR location, and these subscales’ items were overall related to all other items within the respective CSR or USR cluster. The only connection between the two clusters was the edge between one Self-Judgment item and one Self-Kindness item, thereby indicating that the inner structure of associations among SCS-SF items tends to be bidimensional, rather than unidimensional. In summary, the distinction between CSR and USR was supported by both factor analysis and network analysis in our data. These results are consistent with previous validation studies finding this double factorial structure for SCS and SCS-SF (e.g., Hayes et al., 2016; Lόpez et al., 2015) and suggest that it could be useful to start addressing CSR and USR as two separate and relevant components of self-compassion, in both future studies and self-compassion interventions.
Interestingly, in the SCS-SF network, Over-Identification and Isolation items were close to each other and had strong connections among themselves, suggesting that these two subscales may represent similar constructs, and supporting the five-factor structure found by Alfonsson et al. (2023). Indeed, in the Swedish validation of the SCS-SF, the best fitting solution was a five-factor structure in which the Over-Identification and Isolation items loaded on only one factor, whereas the other items loaded on their respective subscales (Alfonsson et al., 2023). Notably, also in the article originally creating and validating the SCS-SF (Raes et al., 2011), the correlation between the Over-Identification and the Isolation subscales was the highest of the correlations among all six subscales, being 0.66 for the Dutch sample and 0.61 for the English sample. The Over-Identification construct is characterized by feelings of inadequacy and recurring thoughts on individual failings (e.g., Neff, 2003); Isolation is logically associated with Over-Identification because fixating on personal difficulties and failing can lead to feeling unworthy and alone in one’s own suffering, and thus isolated.
Regarding Aim 3, multidimensional IRT analyses computed on the bidimensional CSR-USR structure of SCS-SF showed that the two subscales had high performance: all items, with some minor differences among them, were able to discriminate the different levels of the latent trait in individuals and provided a satisfactory amount of information. Interestingly, the performance of USR was slightly better than that of CSR: USR — compared to CSR — items were more able to differentiate among individuals with different values of the latent trait, and provided an overall larger amount of information, even though this additional information was concentrated around the median values of the latent trait. Moreover, the two-factor CSR-USR IRT model showed a better performance than the unidimensional IRT model computed on all SCS-SF items. These IRT results are consistent with our factor analysis and network analysis results, as well as with previous studies finding a bidimensional structure of SCS-SF items (e.g., Hayes et al., 2016). However, IRT analyses add another motivation to consider the two SCS-SF subscales separately: CSR and USR appear as two valid questionnaires to assess compassionate or critical reactions toward the self in front of difficulties and failures, and their validity seems higher when they are considered separately. Lastly, IRT results could be employed in the future to build an even shorter version of the self-compassion scale, by selecting from each of the six SCS-SF dimensions the item with the highest discrimination parameter, thereby creating a six-item self-compassion scale composed of three CSR and three USR items.
Regarding Aim 4, results not only strengthened the convergent and discriminant validity of the SCS-SF, but also expanded the knowledge on the nomological net of SCS-SF, CSR, and USR, by including a large number of correlates referring to various cognitive, emotional, and self-related dispositions. Some of them were expected to be associated with self-compassion, others to be unrelated based on the distinction between self-compassion and other constructs such as self-indulgence or self-esteem (e.g., Neff, 2023). Convergent validity was supported by the correlations with self-oriented dispositions and other abilities related to emotion regulation and awareness: individuals with higher levels of SCS-SF and CSR and lower levels of USR reported more positive, less negative emotional reactions toward the self (as shown by correlations with SOFI). Moreover, individuals scoring higher on SCS-SF and CSR, and lower on USR, were less immersed in and less judgmental toward their thoughts and feelings (as shown by correlations with FFMQ Nonjudging, Decentering, and Nonattachment), and reported a variety of emotion-regulation strategies related to being kind toward oneself in times of difficulties. In particular, scoring higher on SCS-SF and CSR, and lower on USR, was associated with a higher acceptance of negative emotions related to personal experiences (as shown by the correlations with DERS Nonacceptance) and with a lower tendency to blame oneself for the occurrence of negative events or experiences (as shown by the correlations with CERQ Self-Blame), thus fully confirming the original conceptualization of self-compassion (Neff, 2003, 2023) and the distinction between CSR and USR. In particular, these findings support previous research showing that CSR was associated with higher emotional well-being (e.g., Sutton et al., 2018), whereas USR was related to forms of emotional distress and dysregulation (e.g., Hayes et al., 2016). As for divergent validity, SCS-SF, CSR, and USR were unrelated to psychological entitlement, which is the personal belief to deserve more than others: this confirms that self-compassion, measured by the SCS-SF, does not overlap with self-indulgence or other forms of self-serving attitudes and behavior.
Concerning the other correlations, results were consistent with previous literature showing that self-compassionate people report more secure, less anxious, and less avoidant adult attachment (Lathren et al., 2021); stronger inclination toward other people, especially higher perspective taking and lower personal distress (Fuochi et al., 2018a); higher levels of dispositional mindfulness and gratitude (e.g., Garcia-Campayo et al., 2014; Voci et al., 2019); more adaptive emotion-regulation strategies, including lower rumination (Ewert et al., 2021); and higher levels of hedonic and eudaimonic well-being (e.g., Voci et al., 2019; Zessin et al., 2015).
Additionally, we assessed some constructs that were rarely considered in previous self-compassion research, such as thought suppression and cognitive flexibility. Besides rumination, which is known to be negatively associated with self-compassion (e.g., Ewert et al., 2021), we found that scoring higher on SCS-SF and CSR, and lower on USR, was associated with lower suppression of thoughts and higher cognitive flexibility. Therefore, self-compassion appears grounded not only on a kind attitude toward the self, but also on adaptive cognitive styles: in difficult situations or when facing failure, the more people are self-compassionate, the higher their acceptance instead of suppression of negative thoughts, and the greater the searching and finding of alternative and innovative potential solutions to the situations. These results emphasize a few potential mediators of the effects of self-compassion on overall well-being: cognitive flexibility and reduced thought suppression.
Other correlates of self-compassion rarely considered in the past were investigated in our large set of emotion-regulation strategies (i.e., CERQ subscales, ACS). Interestingly, SCS-SF, CSR, and USR had strong correlations with a large array of emotion-regulation strategies and difficulties, suggesting that self-compassion can be viewed as — or is closely related to — an emotion-regulation strategy. In particular, and consistent with previous research (Allen & Leary, 2010; Ewert et al., 2021), we found that, when facing difficulties, the more people are self-compassionate, the less they engage in avoidance and self-criticism, or fall in spirals of negative — self-directed or other-directed — emotions and thoughts. On the contrary, they tend to accept their failings and reappraise the situation, putting things into perspective and finding new strategies to deal with their problems. It is likely that these skills are grounded on the self-caring, encouraging attitude toward oneself in times of trouble. Moreover, we found negative correlations between high self-compassion (high CSR and SCS-SF and low USR) and affective control, especially for negative emotions, suggesting that self-compassionate individuals are less scared by their emotions, especially when these emotions are unpleasant and troubling. Consistent with the previously mentioned correlation between self-compassion and decentering, self-compassionate people are able to let go of negative experiences by accepting unpleasant emotions and finding their inner resources to react. All these results extend previous research results by providing a larger and more complete set of interrelations between self-compassion and emotion regulation skills, emphasizing that self-compassionate people engage more in self-regulation strategies, which mediate the relationships that self-compassion has with well-being and adaptive coping.
Lastly, we observed some differences between CSR and USR that can contribute to the ongoing debate on the bidimensional structure of self-compassion and on the different meaning of the two subscales (e.g., Muris & Otgaar, 2020; Muris et al., 2021; Neff, 2023). Besides emerging as two separate dimensions in factor and network analyses, CSR and USR sometimes showed different correlation patterns: on the one hand, only CSR was correlated with adaptive attitudes toward other people, in particular higher perspective taking and more positive emotional reactions toward others, and more constructive emotion-regulation strategies, such as reappraisal, acceptance, and planning. On the other hand, only USR was correlated with higher use of maladaptive emotion-regulation strategies, i.e., blaming oneself and others and ruminating, and with lower purpose in life and personal growth. This is consistent with previous results showing that CSR was more associated with positive outcomes, such as positive emotions, empathy, need satisfaction, and prosociality (e.g., Lόpez et al., 2015; Moè, 2022; Sutton et al., 2018), whereas USR was more associated with maladaptive dispositions and psychopathological symptoms (e.g., Hayes et al., 2016; Muris et al., 2021). However, we also have to acknowledge that many correlation patterns did not differ between CSR and USR, except for the opposite sign of the association. This would suggest that CSR and USR could also be seen as two complementary constructs, one being not exactly the opposite of the other, but both reflecting the complex phenomenon of self-compassion.
It is possible that CSR and USR capture two different sides of the relationship with the self, and that USR sometimes signals a form of psychological ill-being or dysfunction that is grounded on reduced cognitive skills and emotion-regulation strategies. Probably, such cognitive skills and emotion-regulation strategies can be improved through self-compassion trainings, that is structured interventions such as the Mindful Self-Compassion intervention by Germer and Neff (2019) or Compassion Focused Therapy (Gilbert, 2010). These interventions typically last around 8 weeks and aim at promoting positive attitudes towards oneself, also through mindfulness exercises. Therefore, self-compassion interventions could simultaneously aim at increasing CSR and decreasing USR, on the one hand by favoring a caring attitude toward the self, and on the other hand by reducing excessive self-judgment and self-absorption into personal failures and weaknesses, with a particular focus on dysfunctional emotion-regulation strategies.
Limitations and Future Research
This study has at least three potential limitations that need to be acknowledged. First, relying on convenience samples may limit the external validity of the results, which should be replicated in more representative samples. Second, we only employed self-report measures, which are often sensitive to response biases, such as social desirability. Third, we focused only on the Italian context, without exploring the intercultural validity of SCS-SF; future studies could extend these results to other cultural contexts or adopt a cross-cultural perspective to study the possible different meanings of CSR and USR.
However, this six-sample study provided robust evidence on the factor structure, network topology, questionnaire performance, and correlates of SCS-SF and its subscales, especially shedding light on many aspects of SCS-SF that have been a subject of debate. The SCS-SF is likely to be the combination of two complementary aspects of self-compassion, i.e., compassionate and uncompassionate self-responding, which are not exactly the opposite of one another, but assess different nuances of self-kindness and self-criticism. Future research could start addressing CSR and USR as two separate and relevant components of self-compassion, expanding the exploration on their measurement, as suggested by our factor, network, and IRT analysis, and studying their possible mediators in their relationships with well-being, as suggested by our correlational results. Moreover, future self-compassion interventions could aim at increasing CSR and decreasing USR, focusing on the cognitive and emotional regulation strategies that link these two self-compassion components with well-being. In conclusion, the SCS-SF is a valid, multidimensional instrument, useful for both scholars and practitioners.
Declarations
Conflict of Interest
The authors declare no competing interests.
Ethics Approval
The studies reported in this manuscript involve adult human participants, who provided their informed consent before participating to the data collection. These studies were approved by the Psychological Research Ethics Committee of the University of Padua, protocol #537-a. All procedures of these studies are in accordance with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.
Informed Consent
In the data collection of all samples reported in this manuscript, all individual participants provided their informed consent at the beginning of the online questionnaire.
Use of Artificial Intelligence
AI tools were not used for this research.
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.
Met BSL Psychologie Totaal blijf je als professional steeds op de hoogte van de nieuwste ontwikkelingen binnen jouw vak. Met het online abonnement heb je toegang tot een groot aantal boeken, protocollen, vaktijdschriften en e-learnings op het gebied van psychologie en psychiatrie. Zo kun je op je gemak en wanneer het jou het beste uitkomt verdiepen in jouw vakgebied.
