Content uploaded by Ross B Wilkinson
Author content
All content in this area was uploaded by Ross B Wilkinson on Jul 19, 2018
Content may be subject to copyright.
SCS Factor Structure in 20 Diverse Samples
1
Running Head: SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
Examining the Factor Structure of the Self-Compassion Scale in 20 Diverse Samples: Support for
Use of a Total Score and Six Subscale Scores
Kristin D. Neff,1 István Tóth-Király,2 Lisa M. Yarnell,3 Kohki Arimitsu,4 Paula Castilho,5 Nima
Ghorbani,6 Hailan Xiaoxia Guo,7 Jameson K. Hirsch,8 Jörg Hupfeld,9 Claudio S. Hutz,10 Ilios
Kotsou,11 Woo Kyeong Lee,12 Jesus Montero-Marin,13 Fuschia M. Sirois,14 Luciana K. de
Souza,10 Julie L. Svendsen,15 Ross B. Wilkinson,16 Michail Mantzios17
In press, Psychological Assessment
(Uncorrected version)
1Department of Educational Psychology, University of Texas at Austin, United States; 2Doctoral
School of Psychology, Department of Personality and Health Psychology, ELTE Eötvös Loránd
University, Hungary; 3American Institutes for Research, United States; 4Department of
Psychological Sciences, Kwansei Gakuin University, Japan; 5Faculty of Psychology and
Educational Sciences, University of Coimbra, Portugal; 6Department of Psychology, University
of Tehran, Iran; 7Beijing Hailan Peer Education & Consultation Co., China; 8Department of
Psychology, East Tennessee State University, United States; 9Department of Psychology,
University of Bern, Switzerland; 10Post-Graduate Program in Psychology, Universidade Federal
do Rio Grande do Sul, Brazil; 11Department of Psychology, Université Libre de Bruxelles,
Belgium; 12Department of Counseling Psychology, Seoul Cyber University, Korea; 13 Primary
Care Prevention and Health Promotion Research Network (RedIAPP), Zaragoza, Spain;
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
2
14Department of Psychology, University of Sheffield, United Kingdom; 15Department of
Biological and Medical Psychology, University of Bergen, Norway; 16School of Psychology,
University of Newcastle, Australia; 17Department of Psychology, Birmingham City University,
United Kingdom
Correspondence concerning this article should be addressed to Kristin D. Neff, Department of
Educational Psychology, University of Texas at Austin, 1912 Speedway, Stop D5800, Austin,
TX, USA, 78712-1289. E-mail: kristin.neff@mail.utexas.edu
Author note: The clinical sample from the United Kingdom used in this study was drawn
from the PREVENT trial, a project funded by the National Institute for Health Research Health
Technology Assessment (NIHR HTA) Programme (project number 08/56/01). This trial is
reported in full in the Lancet, DOI: http://dx.doi.org/10.1016/S0140-6736(14)62222-4. We are
grateful to the trial team for allowing us to use the data.
Author contribution statement: KN conceived the paper and wrote the first draft. ITK
conducted all statistical analyses. LY coordinated data collection and contributed her statistical
expertise. The first three authors did the bulk of the writing. MM, the last author, coordinated
putting together the translation information table in the supplementary materials with input from
other authors. All other authors are listed in alphabetical order and contributed data as well as
making comments on earlier drafts of the manuscript.
Public Significance Statement: This study examined the factor structure of the SCS in 20
diverse samples (N = 11,685), and excellent fit was found in every sample for an ESEM single-
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
3
bifactor model (with 95% of item variance explained by a general factor) and an ESEM six-factor
correlated model. Results support use of a total SCS score or six subscale scores, but not two
separate scores representing compassionate and uncompassionate self-responding.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
4
Abstract
This study examined the factor structure of the Self-Compassion Scale (SCS) using
secondary data drawn from 20 samples (N = 11,685) — 7 English and 13 non-English —
including 10 community, 6 student, 1 mixed community/student, 1 meditator, and 2 clinical
samples. Self-compassion is theorized to represent a system with six constituent components -
self-kindness, common humanity, mindfulness and reduced self-judgment, isolation and over-
identification. There has been controversy as to whether a total score on the SCS or if separate
scores representing compassionate versus uncompassionate self-responding should be used. The
current study examined the factor structure of the SCS using confirmatory factor analyses (CFA)
and exploratory structural equation modeling (ESEM) to examine five distinct models: one-
factor, two-factor correlated, six-factor correlated, single-bifactor (one general self-compassion
factor and six group factors), and two-bifactor models (two correlated general factors each with
three group factors representing compassionate or uncompassionate self-responding). Results
indicated that a one- and two-factor solution to the SCS had inadequate fit in every sample
examined using both CFA and ESEM, whereas fit was excellent using ESEM for the six-factor
correlated, single-bifactor and correlated two-bifactor models. However, factor loadings for the
correlated two-bifactor models indicated that two separate factors were not well specified. A
general factor explained 95% of the reliable item variance in the single-bifactor model. Results
support use of the SCS to examine six subscale scores (representing the constituent components
of self-compassion) or a total score (representing overall self-compassion), but not separate
scores representing compassionate and uncompassionate self-responding.
KEYWORDS: Self-Compassion Scale, self-compassion factor structure, bifactor analyses,
Exploratory Structural Equation Modeling (ESEM)
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
5
Examining the Factor Structure of the Self-Compassion Scale in 20 Diverse Samples:
Support for Use of a Total Score and Six Subscale Scores
The construct of self-compassion was first operationally defined and introduced into the
psychological literature a decade and a half ago (Neff, 2003b). Theoretically, self-compassion is
comprised of six components that combine and mutually interact to create a self-compassionate
frame of mind when faced with personal inadequacy or life difficulties: self-kindness versus self-
judgment, a sense of common humanity versus isolation, and mindfulness versus over-
identification. Self-kindness entails being gentle, supportive, and understanding towards oneself.
Rather than harshly judging oneself for shortcomings, the self is offered warmth and acceptance.
Common humanity involves recognizing the shared human experience, understanding that all
humans fail, make mistakes, and lead imperfect lives. Rather than feeling isolated by one's
imperfection - egocentrically feeling as if "I" am the only one who has failed or am suffering -
one takes a broader and more connected perspective with regard to personal shortcomings and
individual difficulties. Mindfulness involves being aware of one’s present moment experience of
suffering with clarity and balance, without running away with a dramatic storyline about negative
aspects of oneself or one’s life experience - a process that is termed "over-identification." As Neff
(2016a) writes, the various components of self-compassion are conceptually distinct and tap into
different ways that individuals emotionally respond to pain and failure (with more kindness and
less judgment), cognitively understand their predicament (as part of the human experience rather
than as isolating), and pay attention to suffering (in a more mindful and less over-identified
manner). The six elements of self-compassion are separable and do not co-vary in a lockstep
manner, but they do mutually impact one another. Put another way, Neff (2016a, 2016b) proposes
that self-compassion represents a dynamic system in which the various elements of self-
compassion are in a state of synergistic interaction.
Over the last few years, research on self-compassion has grown at an exponential rate.
There have been almost 1500 articles or dissertations written about self-compassion since 2003
(based on a Google Scholar search of entries with "self-compassion" in the title, May 2018), over
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
6
half of which have been published in the last two years. The majority of research studies have
utilized the Self-Compassion Scale (SCS; Neff, 2003a) to examine the construct of self-
compassion. The SCS is intended to be used as a total score to measure self-compassion, or else
as six subscale scores to assess its constituent elements: Neff (2016a, 2016b) proposes that the
state of self-compassion entails more compassionate fewer uncompassionate responses to
personal suffering, which is why the SCS measures both.
Neff’s operationalization of the SCS was based on compassion for others as broadly
conceptualized in Buddhist philosophy (2003b), although scores on the SCS have a relatively
weak correlation with compassion for others (Neff & Pommier, 2013). This appears to be because
most people have significantly more compassion for others than for themselves (Neff, 2003a;
Neff & Pommier, 2013), meaning the two do not necessarily go hand in hand.
Research using the SCS suggests that self-compassion is a key indicator of wellbeing. For
instance, cross-sectional research using the SCS shows that self-compassion has moderate to
strong positive associations with outcomes such as happiness, optimism, life satisfaction, body
appreciation and motivation and negative associations with outcomes such as depression, anxiety,
maladaptive perfectionism and fear of failure – findings that are replicated using experimental
methods such as interventions or mood manipulations (see Neff & Germer, 2017, for a review).
While research suggests that self-compassion yields similar mental health benefits as other
positive self-attitude constructs such as self-esteem (Neff, 2011), it does not appear to have the
same pitfalls (Crocker & Park, 2004). For instance, Neff and Vonk (2009) found that while self-
compassion and self-esteem were strongly correlated, simultaneous regressions indicated that
self-compassion was associated with more stable and less contingent feelings of self-worth over
time, and was associated with less social comparison, public self-consciousness, self-rumination,
anger, closed-mindedness and narcissism than self-esteem. Similarly, an experience sampling
study conducted by Krieger, Hermann, Zimmermann and Grosse Holtforth (2015) found that
levels of self-compassion, but not self-esteem, predicted less negative affect when encountering
stressful situations over a 14-day period.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
7
The incremental predictive validity of SCS scores have been demonstrated with constructs
such as neuroticism (Neff, Tóth-Király & Colosimo, in press; Stutts, Leary, Zeveney & Hufnagle,
in press) and self-criticism (Neff, 2003a). Although a key feature of self-compassion is the lack of
self-judgment, overall SCS scores still negatively predict anxiety and depression when
controlling for self-criticism and negative affect (Neff, Kirkpatrick & Rude, 2007).
It should be mentioned that there are other models and measures of self-compassion, and
that there is a lack of consensus in the field on how to define or measure compassion for self or
others (Gilbert et al., 2017; Gilbert, Clarke, Hempel, Miles & Irons, 2004; Strauss et al., 2016).
Given that the SCS is the most commonly used measure of self-compassion, however, the current
study is aimed at examining the psychometric properties of the SCS in a way that is theoretically
consistent with Neff’s (2003b) operationalization of the construct.
The SCS was developed in a sample of U.S. college undergraduates (Neff, 2003a).
Confirmatory factor analyses (CFA) were used to provide support that scale items fit as intended
with the a priori theoretical model (Furr & Bacharach, 2008). An initial CFA found an adequate
fit for a six-factor inter-correlated model and a higher-order factor model. Since that time at least
30 published studies have examined the factor structure of the SCS (see Table S1 in the
supplemental materials for a summary). Multiple translations of the SCS have been published,
most have which have replicated the six-factor structure of the SCS using CFA. While not all
examined the higher-order model, those that did yielded inconsistent findings. For example, a
higher-order factor was supported with a Czech (Benda & Reichová, 2016, Norwegian (Dundas
et al., 2016), and two Portuguese samples (Castilho, Pinto-Gouveia, & Duarte, 2015; Cunha,
Xavier & Castilho, 2016), but not with German (Hupfeld & Ruffieux, 2011), Italian (Petrocchi et
al., 2013) or a third Portuguese sample (Costa et al., 2015).
Recently, there has been controversy over whether or not self-compassion should be
measured as an overall construct, or if compassionate versus uncompassionate self-responding
should be measured separately. Some have found that use of a total score is not justified through
higher-order factor analyses, and have argued that two separate positive and negative factors
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
8
demonstrate better fit (e.g., Costa et al., 2015; López et al., 2015; Montero-Marín et al., 2016).
These researchers tend to use the term “self-compassion” to describe the positive factor and "self-
criticism" or "self-coldness" to describe the negative factor (Costa et al., 2015; Gilbert et al.,
2011; López et al., 2015). However, self-criticism and self-coldness primarily describe self-
judgment, or how people emotionally respond to suffering, and do not describe isolation (a way
of cognitively understanding suffering) or over-identification (a way of paying attention to
suffering). Moreover, these terms may obscure the fact that items representing negative self-
responding are reverse-coded to indicate their relative absence. Therefore, we prefer the terms
compassionate self-responding (CS) to represent the three components of self-kindness, common
humanity and mindfulness and reduced uncompassionate self-responding (RUS) to represent
lessened self-judgment, isolation and over-identification measured by the SCS.
The question of whether the SCS can be used to measure self-compassion as a holistic
state of being or if it should be used to measure two distinct states of being has important
implications for our understanding of what self-compassion is. If self-compassion does not
include RUS, the implication would be that how self-critical, isolated, and over-identified
individuals are in times of struggle have little bearing on how self-compassionate they are. This,
in turn, would have implications for researchers’ attempts to examine the link between self-
compassion and well-being. For instance, Muris and Petrocchi (2017) conducted a meta-analysis
of the link of the SCS subscales with psychopathology across 18 studies and found the three
components representing RUS had a stronger association with psychopathology (e.g., depression,
anxiety and stress) than the CS components. They argue that negative items “may inflate the
relationship with psychopathology” (p. 734) and should therefore be excluded from the SCS. If,
however, RUS is an integral part of self-compassion, then logically speaking it cannot “inflate”
its own association with psychopathology. Rather, RUS could be interpreted to “explain” the link
between self-compassion and psychopathology. Support for this point of view can be found in
studies designed to examine self-compassion through mood induction (i.e., using writing
prompts) or through intervention, which show that increasing self-compassion experimentally
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
9
also leads to reduced negative outcomes such as depression, anxiety, shame, etc. (see Neff &
Germer, 2017). Not including RUS subscales in the measurement of self-compassion, therefore,
could potentially underestimate its relationship to psychopathology.
Some have argued that that the CS and RUS subscales should not be combined into a total
self-compassion score because compassionate responding is associated with parasympathetic
nervous system activity and uncompassionate responding with sympathetic activity (Gilbert,
McEwan, Matos, & Rivis, 2011). However, research with the SCS shows that the CS and RUS
subscales do not substantially differ in terms of their association with markers of sympathetic
response (e.g., alpha-amylase, interleukin-6) after a stressful situation (Neff et al., in press), or
vagally mediated heart-rate variability, a marker of parasympathetic response (Svendson et al,
2016). As Porges (2001) makes clear, the two types of autonomic nervous system responding
themselves interact and co-vary as a system. The issue of whether self-compassion is best
measured as a total score or if CS and RUS should be measured separately is largely a
psychometric question, however, which has yet to be definitively established.
Alternative Models for Examining the Factor Structure of the SCS
It is important that the psychometric analyses used to examine psychological measures be
consistent with the psychological theory underlying those measures (Morin, Arens, Tran & Caci,
2016b). Higher-order models are commonly employed to validate the simultaneous use of a total
score and sub-scale scores in measures of multidimensional psychological constructs (e.g., Chen,
West, & Sousa, 2006; Gignac, 2016). A higher-order model represents several first-order factors
(representing sub-scale scores) and a higher-order factor (representing a total score) that explains
their inter-correlation, but makes the strong assumption that the higher-order factor only
influences individual item responses through the pathway of the first-order factors (appropriate
for certain constructs like IQ). The original SCS publication (Neff, 2003a) used a higher-order
model to justify use of a total and six subscale scores, but as mentioned above, support for a
higher-order model has been inconsistent.
Williams, Dalgleish, Karl, and Kuyken (2014) did not find support for a higher-order
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
10
factor in four different English samples (student, community, meditator, and clinical), but did find
support for a six-factor correlated model. They suggested that the six subscales but not a total
score be used. López et al. (2015) examined a Dutch community sample and did not find support
for a higher-order factor, so conducted Exploratory Factor Analysis (EFA) and found that the
positive items loaded on one factor and the negative items loaded on a second factor. No CFA
was conducted to confirm this two-factor model, however. Costa et al. (2015) examined a
Portuguese clinical sample and compared a higher-order model, a six-factor uncorrelated model,
a two-factor uncorrelated model that separated positive and negative items, and a two-factor
model that included correlated errors designed to improve model fit, and found that the two-factor
model with correlated errors had the best fit. These latter two sets of researchers suggested that
separate positive and negative scores be used rather than a total score.
The bifactor model is an increasingly popular way to model multidimensional constructs
(Reise, 2012; Rodriguez, Reise, & Haviland, 2016a). A bifactor model does not assume that the
general or group factors are higher or lower than the other but rather co-exist, and models the
direct association of the general factor and group factors on individual item responses. This has
the added benefit of enabling the calculation of omega values that represent the amount of
reliable variance in item responding explained by the general factor. Note that with a bifactor
model the group factors are not allowed to correlate. Although perhaps counter-intuitive, this
improves interpretability. For instance, it models those aspects of an item (e.g., When something
upsets me I try to keep my emotions in balance) that are shared by all items in the general factor
(e.g., self-compassion), as well as those aspects that are only shared by other items in its group
factor (e.g., mindfulness). Neff (2016a) argued that a bifactor model provides a better theoretical
fit with her conceptualization of self-compassion than a higher-order model given that behaviors
assessed by individual items are directly representative of self-compassion as a general construct
in addition to its constituent group components.
Neff, Whittaker, and Karl (2017) examined the SCS using bifactor CFA analysis in four
different U.S. populations: undergraduates, community adults, meditators, and a clinical
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
11
population. While the one-factor, two-factor correlated, and higher-order models had poor fit
across samples, the six-factor correlated and bifactor models had acceptable fit using liberal fit
criteria in the undergraduate, community and meditator samples. Fit was inadequate in the
clinical sample. Nonetheless, omega values revealed that over 90% of the reliable variance in
scores could be explained by a general self-compassion factor in all four populations (including
the clinical sample). Findings were interpreted as providing support for use of a total score as
well as six subscale scores, but not as two positive and negative scores. Cleare, Gumley, Cleare
and O’Conner (2018) independently replicated these findings in a Scottish sample: support was
not found for a one-factor, two-factor, or higher-order model, but was found for a six-factor
correlated and bifactor model, with 94% of the variance in item responding explained by a
general factor.
Three additional studies on translations of the SCS have provided evidence for a model
with six group factors and one general factor using a bifactor CFA approach: French (Kotsou &
Leys, 2016) Brazilian Portuguese (Souza & Hutz, 2016), and Italian (Veneziani, Fuochi & Voci,
2017). However, Montero-Marín et al. (2016) did not find support for a CFA bifactor model in
two Spanish and Brazilian-Portuguese samples of doctors, but did find support for two higher-
order factors (CS and RUS) and six first-order factors. Moreover, Brenner, Health, Vogel and
Credé (2017) found that a two-bifactor CFA model with six group factors and two uncorrelated
general (CS and RUS) factors had better fit than a single-bifactor model in a sample of U.S.
undergraduates, though findings for some indicators were poor and the choice of examining two
uncorrelated general factors is not theoretically consistent with the construct of self-compassion.
Thus, the dimensionality of the SCS is still in question. Also, the above-mentioned results
suggest that the assumptions of CFA might be overly restrictive for the SCS, given the
inconsistency of findings.
CFA makes the strict assumption that items can only load on their respective factors, and
may fail to account for two main sources of construct-relevant dimensionality in complex scales
like the SCS, potentially resulting in biased parameters (Morin, Arens & Marsh, 2016a; Morin,
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
12
Arens et al., 2016b). These sources do not refer to random measurement error, but are related to
the idea that items often present more than one source of true score variance and subsequently
belong to more than one construct. The first source refers to the fact that individual items are
expected to be associated with a global factor (e.g., self-compassion), in which specific factors
are not differentiated, as well as specific group factors (e.g., self-kindness or reduced self-
judgment), in which they are differentiated. As mentioned, the relation between specific and
global factors can be modeled in a hierarchical or in a bifactor manner with the latter generally
being preferred unless there are strong theoretical reasons for the application of the former.
The second source of dimensionality comes from the fact that the six components of the
scale are conceptually close and interrelated as a system, meaning items within each subscale
should be expected to have significant associations with items in other subscales. Indeed, a recent
review of simulation studies (Asparouhov, Muthén & Morin, 2015) have shown that when cross-
loadings between items and non-target factors are not expressed (i.e., cross-loadings are
constrained to be zero), parameters are likely to be biased. Exploratory Structural Equation
Modeling (ESEM) is specifically designed to model system level interactions (Marsh, Morin,
Parker & Kaur, 2014; Morin, Marsh & Nagengast, 2013). In CFA, items are strictly allowed to
load on one factor, and these additional associations between items and non-target factors are
reflected in the form of modification indices and/or inflated inter-factor correlations, which are
the only ways overlap can be expressed. In ESEM, these associations are expressed in the form of
item cross-loadings. Unlike Exploratory Factor Analyses (EFA), in which no a priori hypotheses
about models are advanced, ESEM with target rotation (Browne, 2001) can model a priori
hypotheses and therefore be directly compared to CFA models (Marsh et al., 2014). ESEM has
been suggested to result in substantially better fit and less strongly correlated factors than
corresponding CFA solutions (Marsh, Liem, Martin, Morin, & Nagengast, 2011; Morin &
Maïano, 2011; Tóth-Király, Orosz, et al., 2017).
ESEM has rarely been used to examine the SCS. However, Hupfeld and Ruffieux (2011)
as well as Tóth-Király, Bőthe and Orosz (2017) applied ESEM to analyze the factor structure of
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
13
the SCS and found that, compared to CFA, ESEM provided a better fit to the data. Moreover, to
account for the two sources of construct-relevant dimensionality, Tóth-Király, Bőthe and Orosz
(2017) also used the integrative bifactor ESEM framework (Morin, Arens, et al., 2016a, 2016b;
Morin, Boudrias, Marsh, Madore, & Desrumeaux, 2016), and results strongly supported the
presence of a global self-compassion factor as well as the six-specific factors. The overarching
bifactor ESEM framework appears to be especially appropriate for the SCS because it can
simultaneously model both the specific and overall relationship of items using a bifactor analytic
approach as well as their interaction as a system with an ESEM approach.
The Current Study
In the current study, we examined the factor structure of the SCS using both CFA and
ESEM analyses for five distinct models: a single factor, two-factor correlated, six-factor
correlated, single-bifactor model (one general factor and six group factors), and a correlated two-
bifactor model (a general factor representing CS with three group factors representing higher
levels of self-kindness, common humanity and mindfulness, and a general factor representing
RUS with three group factors representing lower levels of self-judgment, isolation and over-
identification). Based on the existing literature, we expected that the one factor and two-factor
correlated models would have poor fit, and the six factor-correlated, single-bifactor and two-
bifactor models would have better fit. We also expected fit indices to be better in ESEM rather
than CFA analyses given that it is more appropriate for modeling system-level interactions. Our
overall goal was to determine the best-fitting solution that is also well-aligned with Neff’s
underlying model of self-compassion (2003b), given that this is the theoretical model that the
SCS was created to measure.
We examined the SCS in 20 different samples. Because the SCS was developed in
English we included 7 English samples, but also 13 samples from non-English speaking
countries. We included student, community, meditator and clinical samples. The meditator and
one of the clinical samples were the same as examined in Neff et al. (2017), and a second
Portuguese clinical sample was also included (Castilho, Pinto-Gouveia, & Duarte, 2015). Given
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
14
that the SCS is commonly used to assess outcomes of meditation-based and clinical interventions
(e.g., Birnie, Speca, & Carlson, 2010; Kelly, Wisniewski, Martin-Wagar, & Hoffman, 2017), we
felt it was important to include these populations. The comprehensiveness of this study was
designed to try to find more definitive answers to questions regarding the factor structure of the
SCS: Should a total score be used, or two separate scores representing CS and RUS?
Method
Procedure
This study was organized by the first three authors, who wanted to examine the factor
structure of the SCS in a variety of international samples. SCS data for three samples from the
United States (US) were contributed by the first and third authors, who originally collected the
data for other research purposes. Appropriate Institutional Review Board (IRB) approval was
received before collecting these data, which were de-identified for the current study before being
statistically analyzed by the second author. To gather SCS data from samples outside of the US,
researchers were contacted in other English and non-English-speaking countries. These
researchers contributed SCS data for 17 additional samples, which had also been collected
previously for other research purposes. (Information about the data source of each sample as well
as participant recruitment procedures can be found in the supplementary materials). SCS
data contributed from sources outside the US were received as secondary data and included no
potential participant identifiers. IRB approval was not required for analyses of these de-identified
secondary data, although researchers from outside the US also obtained local ethics committee
approval before collecting their original data as appropriate.
Participants
The initial number of participants was 11,990 from 20 international samples drawn from
the following counties: Australia, Brazil, Canada, China, France, Germany, Greece, Iran, Italy,
Japan, South Korea, Norway, Portugal, Spain, United Kingdom, and United States. In total, we
included 10 community, 6 student, 1 mixed community/student, 1 meditator, and 2 clinical
samples. Participants were excluded if they were under age 18 or had more than 50% of their
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
15
responses missing. Thus, the final sample included 11,685 respondents (3,296 males, 8,367
females, 22 unspecified), aged between 18 and 83 (M = 32.29, SD = 8.28). Specific sample
characteristics can be seen in Table 1.
Measures
The SCS (Neff, 2003a) is a 26-item self-report questionnaire measuring the six
components of self-compassion: Self-Kindness (5 items; e.g., “I try to be loving towards myself
when I’m feeling emotional pain”), reduced Self-Judgment (5 items; e.g., “I’m disapproving and
judgmental about my own flaws and inadequacies”), Common Humanity (4 items, e.g., “When
things are going badly for me, I see the difficulties as part of life that everyone goes through”),
reduced Isolation (4 items, 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”), Mindfulness 4 items, e.g., (“When I’m
feeling down I try to approach my feelings with curiosity and openness”), and reduced Over-
Identification (4 items, e.g., "When something upsets me I get carried away with my feelings”).
Responses are given on a scale from 1 (almost never) to 5 (almost always). Note that all items in
the Self-Kindness, Common Humanity and Mindfulness subscales are positively-valenced, while
all items in the Self-Judgment, Isolation and Over-Identification subscales are negatively
valenced. Items representing uncompassionate self-responding are reverse-coded before
calculating a total score to indicate their relative absence in a self-compassionate mindset. Means
are calculated for each subscale, and a grand mean is calculated for a total self-compassion score.
Neff (2003a) found that items forming a total SCS score evidenced good internal reliability
(Cronbach's α =.92), as did the six subscales (Cronbach's α ranging from .75 to .81). Test-retest
reliability over a three-week interval was good (total score, Cronbach's α =.93; six subscales,
Cronbach's α ranging from .80 to .88). The current study also employed 12 SCS translations (out
of 16 published): Brazilian Portuguese, Chinese, French, German, Greek, Persian, Italian,
Japanese, Korean, Norwegian, Portuguese, and Spanish. A description of the psychometric
properties of each SCS translation can be found in the supplementary materials.
Analyses
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
16
All statistical analyses were conducted with Mplus 7.4 (Muthén & Muthén, 1995-2015)
with the weighted least squares mean- and variance-adjusted estimator (WLSMV) as it is more
suitable for ordered-categorical items with five or less answer options than estimators based on
maximum-likelihood (e.g., Bandalos, 2014; Finney & DiStefano, 2006). Prior to the main
analyses, negative items were reverse-coded. In order to systematically investigate the potential
sources of construct-relevant dimensionality of the SCS, five corresponding CFA and ESEM
models were tested and subsequently contrasted: (1a, 1b) a one-factor model with a unitary self-
compassion dimension; (2a, 2b) a two-factor correlated model with two unitary factors
representing CS and RUS; (3a, 3b) a six-factor correlated model with six components of self-
compassion; (4a, 4b) a single-bifactor model with a general self-compassion factor and six group
factors; and (5a, 5b) a two-bifactor model including two correlated general factors representing
CS and RUS, each with three group factors. As per typical model specifications, in the CFA-
based models (1a-5a), items were only allowed to load on their a priori target factors with cross-
loadings being constrained to zero. In the ESEM-based models (1b-5b), items were allowed to
load on the non-target factors as well. ESEM was also estimated in a confirmatory manner with
target rotation (Browne, 2001) as per prior suggestions (Asparouhov & Muthén, 2009) and
applications (Tóth-Király, Bőthe, Rigó, & Orosz, 2017). In the correlated models (2a, 2b, 3a, and
3b), factors were allowed to correlate freely. In the case of the bifactor models (4a, 4b, 5a, and
5b), group factors were specified as orthogonal to the general factor, as is standard (e.g., Reise,
2012; Reise, Moore, & Haviland, 2010) but the two general factors were specified as correlated
1
(see also Tóth-Király, Morin, Bőthe, Orosz, & Rigó, 2018 for a similar application or Morin,
Myers, & Lee, in press, for an overview). These models were tested in the total sample and
individual samples.
In model assessment, instead of only relying on the chi-square test which is sensitive to
sample-size (Marsh, Hau, & Grayson, 2005), commonly applied goodness-of-fit indices were
1
In the two-bifactor ESEM model, the general factors were specified as CFA factors (i.e., no cross-loadings between
them), while the six specific factors were specified as ESEM factors (i.e., cross-loadings between them were
allowed).!
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
17
examined with their respective thresholds (Hu & Bentler, 1999; Marsh et al., 2005): the
Comparative Fit Index (CFI; ≥ .95 for good, ≥ .90 for acceptable), the Tucker–Lewis index (TLI;
≥.95 for good, ≥ .90 for acceptable), the Root-Mean-Square Error of Approximation (RMSEA; ≤
.06 for good, ≤ .08 for acceptable) with its 90% confidence interval, and the weighted root mean
square residual (WRMR; ≤ 1.00 for acceptable). Note that we did not compare fit using AIC or
BIC because these information criteria are only available for maximum likelihood-based
estimations, which are less accurate for ordered categorical data. However, the primary purpose
of these indices is to determine which models would be most likely to cross-validate in
subsequent samples, and this study determines cross-validation directly by examining model fit in
20 different samples.
Analyses of data should not be based solely on fit indices, however. The close inspection
of parameter estimates (e.g., factor loadings, cross-loadings and inter-factor correlations) and the
theoretical conformity of the models may also reveal valuable information about measurement
models (as proposed by Hu & Bentler, 1998; Marsh, Hau, & Wen, 2004; Marsh et al., 2011;
Morin, Arens, et al., 2016a, 2016b; Morin, Boudrias, et al., 2016). When comparing first-order
CFA and ESEM models, the emphasis should be on comparison of factor correlations and on the
need to incorporate cross-loadings, assuming that both solutions have well-defined factors with
strong target loadings. If there is a substantial difference in the size of correlations between CFA
and ESEM, the latter results are preferred as they provide more exact estimates (Asparouhov et
al., 2015). If differences are negligible, then CFA is preferred due to its greater parsimony.
Relatively large cross-loadings in the ESEM model may suggest an unmodeled general factor,
which can be tested with a bifactor model. The general factor should also be well-defined by
strong and theoretically meaningful factor loadings. Additionally, reduced cross-loadings and
some well-defined specific factors would also provide support for the bifactor representation
2
. A
particularly important question relates to the inclusion of one or two general factors where, once
2
Naturally, not all specific factors are well-defined in the bifactor model relative to the first-order model, given that
the item-level covariance is disaggregated to two sources (general and specific factors) instead of one (e.g., Morin et
al., 2016a; Tóth-Király et al., 2018).
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
18
again, the close examination of factor loadings is highly informative.
We also assessed the reliability of items in the models. In the case of the six-factor model
we calculated composite reliability (Raykov, 1997) as opposed to Cronbach’s alpha, which has
been criticized as being less useful for determining the reliability of factors (Rodriguez et al.,
2016a). It has the advantage of being model-based, taking into account factor loadings and item-
specific measurement errors as well. Based on Bagozzi and Yi (1988), values above .60 are
considered acceptable, whereas values above .70 are good. As bifactor models allow the
partitioning of the different sources of variance into the global and specific factors, omega (ω)
and omega hierarchical (ωH) were also calculated for the best fitting models based on
standardized estimates (Brunner, Nagy, & Wilhelm, 2012; Rodriguez, Reise, & Haviland,
2016b). Omega estimates the proportion of the variance in item responding that is attributed to
both the global and specific factors. OmegaH estimates the proportion of variance that is
attributed to the general factor only. Finally, we also compared the omega and omegaHs on the
basis of Rodriguez et al. (2016b) to investigate the degree of reliable variance in item responding.
For the variance attributed to the general factor, one should divide the value of omegaH by omega
(i.e., ωH / ω); for the remaining reliable variance attributable to the specific factors, one should
subtract omegaH from omega (i.e., ω - ωH). Reise, Bonifay, and Haviland (2013) suggest 75% or
higher accounted for by the general factor as the ideal amount of variance to justify use of a total
score in spite of the presence of multidimensionality of the data.
Results
Structural analyses
Because results were generally similar for the total sample and the individual samples, we
mainly refer to results for the total sample for the sake of simplicity. We first examined the fit of
the one-factor model for all samples (see the supplementary materials). The one-factor ESEM
solution is fundamentally a one-factor CFA (using only different estimation routines in Mplus) as
there are no cross-loadings in this model. In accord with our hypotheses, results clearly
demonstrate the inadequacy of the unidimensional model (Total sample: CFI = .74, TLI = .73,
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
19
RMSEA = .15 [90% CI .15-.15], WRMR = 14.44). Tables 2, 3, 4, and 5 present model fit indices
for CFA and ESEM analyses for the two-factor, six-factor, single-bifactor and correlated two-
bifactor models, respectively In the case of the two-factor correlated models (see Table 2), both
the CFA (Total sample: CFI = .90, TLI = .89, RMSEA = .10 [90% CI .09-.10], WRMR = 7.48)
and ESEM (Total sample: CFI = .88, TLI = .86, RMSEA = .11 [90% CI .11-.11], WRMR = 6.31)
versions showed marginally acceptable fit indices in some samples, but the majority were not
acceptable by commonly applied standards, hence we rejected these solutions. In the case of the
six-factor correlated CFA and ESEM models (see Table 3), most CFA models had acceptable fit
(Total sample: CFI = .95, TLI = .94, RMSEA = .07 [90% CI .07-.07], WRMR = 5.15). However,
ESEM systematically outperformed these solutions as apparent with excellent fit indices (Total
sample: CFI = .99, TLI = .97, RMSEA = .05 [90% CI .05-.05], WRMR = 1.75).
Following Morin et al. (2016a, 2016b), we also examined standardized item factor
loadings for the corresponding CFA and ESEM solutions for the total sample to select the final
models, presented in Tables 6, 7, and 8. When examining the six-factor correlated models (Table
6), all six factors were well-defined by their respective factor loadings (λ = .65 to .84, Mλ = .76)
in CFA, but this solution also resulted in relatively high factor correlations (r = .38 to .91, Mr =
.64), undermining the discriminant validity of interpretations of items in the six factors. In the
ESEM model, factor loadings (λ = .26 to .97, Mλ = .56) as well as factor correlations (r = .16 to
.66, Mr = .42) were systematically lower. These results are in line with previous studies (Morin et
al., 2016a) showing that ESEM often provides a better representation of the inter-factor
correlations. As expected, there were some cross-loadings (|λ| = .00 to .42, Mλ = .10) between
conceptually similar items (≥ .32; Worthington & Whittaker, 2006). For example, the self-
kindness item "I’m tolerant of my own flaws and inadequacies" cross-loaded on reduced self-
judgment. Overall, cross-loadings were found for two self-kindness items on reduced self-
judgment and one on mindfulness, one reduced self-judgment item on self-kindness, one
mindfulness item on self-kindness and one on reduced over-identification, and two reduced over-
identification items on reduced self-judgment.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
20
The next question that we addressed is whether the single-bifactor model with one general
factor (representing self-compassion) or the correlated two-bifactor model with two general
factors (representing CS and RUS) was able to provide an improved representation of the data.
For the single-bifactor models (Table 4), CFA models were generally inadequate (Total sample:
CFI = .85, TLI = .82, RMSEA = .12 [90% CI .12-.12], WRMR = 10.55), whereas ESEM models
generally had much better fit (Total sample: CFI = .99, TLI = .98, RMSEA = .04 [90% CI .04-
.04], WRMR = 1.42). Results for CFA and ESEM were less differentiated for the correlated two-
bifactor models including two general factors (see Table 5), with generally adequate fit for the
CFA models (Total sample: CFI = .96, TLI = .95, RMSEA = .06 [90% CI .06-.06], WRMR =
4.49), as well as for the ESEM models (Total sample: CFI = .99, TLI = .99, RMSEA = .04 [90%
CI .03-.04], WRMR = 1.20). However, it should be noted that the correlated two-bifactor CFA
model for the total sample had misspecifications, and almost half (9 out of 20) of the individual
samples had negative residual variances, suggesting that the data did not support the hypothesized
models. Therefore, we only compared the parameter estimates of the competing single- and
correlated two-bifactor ESEM models.
The parameter estimates for the single-bifactor model (Table 7) revealed a well-defined
general factor (|λ| = .36 to .75, M = .62) reflecting a global level of self-compassion. As for the
specific factors, common humanity retained a higher degree of specificity (|λ| = .35 to .73, M =
.53) once the general factor was extracted. By the same token, isolation (|λ| = .24 to .58, M = .41)
and mindfulness (|λ| = .28 to .52, M = .41) had moderate degree of specificity, self-kindness (|λ| =
.06 to .56, M = .34) and overidentification (|λ| = .19 to .50, M = .34) had a smaller degree of
specificity, whereas self-judgment (|λ| = .07 to .44, M = .22) retained almost no meaningful
specificity. Finally, cross-loadings also slightly decreased in magnitude (|r| = .01 to .34, M = .09)
relative to the six-factor ESEM model. In the case of the correlated two-bifactor-ESEM model
(see Table 8), while the correlation between the two factors were reduced (r = .09, p = .086), the
two general factors were weakly defined by their respective factor loadings (Positive: |λ| = .01 to
.48, M = .22; Negative: |λ| = .04 to .35, M = .17), arguing against the incorporation of a second
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
21
general factor and supporting the superiority of the single-bifactor ESEM model with one general
factor. Taking these results together, it appears that a six-factor correlated model (representing the
six components of self-compassion) and a single-bifactor model (representing a general self-
compassion factor and six specific factors) are supported, but a correlated two-bifactor model
(representing CS and RUS) is not supported once parameter estimates are taken into account.
Reliability analyses
Finally, we estimated composite reliability indices for items in the six-factor model and
the omega and omegaH indices for items the single-bifactor ESEM model in order to examine
reliability. For the six-factor model (examining the sample as a whole), items in all factors had
acceptable levels of composite reliability using Bagozzi and Yi’s (1988) criteria of > .60 as
adequate and > .70 as good: (self-kindness = .84, reduced self-judgment = .73, common humanity
= .81, reduced isolation = .83, mindfulness = .67, and reduced over-identification = .70).
(Composite reliability for items in the individual samples are available upon request from the first
author). Reliability results for the single-bifactor model for all samples are presented in Table 9,
although again we only discuss results for the total sample here. The single-bifactor model
displayed high omega (.96) and omegaH (.91) values, demonstrating that a large majority of the
variance in item responding can be attributed to the general factor. As per Rodriguez et al.
(2016b), we compared the ratio of omega and omegaH to establish the amount of reliable
variance of items attributable to the general factor (omegaH divided by omega) and that
attributable to the multidimensionality caused by the specific factors (omegaH subtracted from
omega). For the single-bifactor model, 95% of the reliable variance in item responding was
attributed to the general self-compassion factor, whereas 5% was attributed to the group factors.
Discussion
Our analyses, which were designed to determine the best factor structure for the SCS,
found that a one- and two-factor solution to the SCS had an inadequate fit using both CFA and
ESEM. In contrast, a six-factor correlated solution had good fit using ESEM (CFA results for the
six-factor solution were also acceptable) in every sample examined. The single-bifactor ESEM
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
22
model (with one group and six specific factors) also had good fit in every sample. Moreover,
inspection of factor loadings suggested good parameter estimates for a single general factor in
ESEM. While the correlated two-bifactor ESEM model with two correlated general factors also
had good model fit, factor loadings indicated poor specification of separate factors representing
CS and RUS, so this model was rejected. Note that the single-bifactor ESEM model also had the
highest level of theoretical conformity with Neff’s (2003b) view that self-compassion is
comprised of six components that interact as a global system. Results for our final selected
models were remarkably similar across the 20 diverse populations examined - including student,
community, clinical, and meditator samples in 13 different languages - providing strong support
for the generalizability of the SCS to measure self-compassion.
Findings regarding cross-loadings in the ESEM models are also informative. In the six-
factor model all factors were well defined, but eight cross-loadings were found (cross loadings
were found equally within and across the CS and RUS dimensions). These cross-loadings
highlight the importance of using models such as ESEM that can uncover this particular source of
construct-relevant dimensionality. Use of a total SCS score was supported by the finding that
95% of the reliable variance in SCS item responding could be explained by a general factor for
the total sample, ranging from 86% to 96% for the individual samples. This is well over the 75%
or higher suggested by Reise et al. (2013) to justify use of a total score. All of the factors in the
six-factor solution had adequate to good levels of composite reliability based on conventional
thresholds (Bagozzi & Yi, 1988). In the single-bifactor model the general factor was well-defined
and the specific factors were moderately well-defined. These observations give support for the
idea that the specific factors assess relevant components over and above the general factors. They
also support the system level interaction of components. We interpret these results as supporting
use of a global score (representing self-compassion) or six subscale scores (representing self-
kindness, common humanity, mindfulness and reduced self-judgment, isolation and over-
identification), but not two separate CS and RUS scores.
The fact that the one- and two-factor solution had poor fit but a six-factor solution had
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
23
good fit makes sense theoretically. It is potentially problematic to argue that self-compassion is a
unitary construct (no theorists we are aware of have made this argument), or to argue that the
three subscales representing CS versus RUS each form unitary constructs, as proposed by some
(e.g., Costa et al., 2015; López et al., 2015). The three subscales within each of these dimensions
are distinct, and tap into the way that people emotionally respond to suffering (with self-kindness
or reduced self-judgment), cognitively understand their suffering (with common humanity or
reduced isolation), and pay attention to their suffering (with mindfulness or reduced over-
identification). Thus, within the dimensions of CS and RUS the three components are not thought
to be identical.
Given that support was found for use of a total score and also six separate subscale scores,
the question arises - when is use of a total score versus subscale scores warranted? The
nomological network observed between the six subscales and important aspects of functioning
indicates that there are areas of overlap but also difference between the subscales. For instance,
Körner et al. (2015) found that it was mainly isolation that predicted depression, while Alda et al.
(2016) found that common humanity had the strongest association with telomere length.
Moreover, evidence from neuroimaging studies suggest the various components of self-
compassion have distinct brain signatures. Longe et al. (2010) found that self-critical thinking
(similar to self-judgment) and self-reassurance (similar to self-kindness) were associated with
different regions of brain activity. Self-criticism was associated with activity in lateral prefrontal
cortex (PFC) regions and dorsal anterior cingulate (dAC), linked to error processing and
resolution, and also behavioral inhibition. Self-reassurance was associated with left temporal pole
and insula activation, related to empathy. Mindfulness, on the other hand is linked to increased
neural activation in the prefrontal cortex (PFC) and dorsal anterior cingulate cortex (dACC),
associated with attentional control and emotion regulation (Young et al., 2017). These results
suggest that the six components of self-compassion are not one unitary thing, nor are they two
unitary things, but are six distinct but interrelated things.
Use of the subscales may have relevance for understanding the mechanisms by which
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
24
self-compassion engenders well-being. Neff, Long, et al. (in press) recently explored the link of
self-compassion and its components to psychological functioning in seven domains –
psychopathology, positive psychological health, emotional intelligence, self-concept, body image,
motivation, and interpersonal functioning. When examining the zero-order correlations between
observed subscale scores and outcomes, they found that reduced self-judgment, isolation, and
over-identification tended to have a stronger link to negative emotionality and self-evaluation
than self-kindness, common humanity and mindfulness, while the latter tended to have a stronger
association with outcomes like emotional awareness, goal re-engagement, compassion for others
and perspective-taking. For many aspects of psychological functioning, however, such as
happiness, wisdom, contingent self-esteem based on approval, body appreciation, or grit, all six
subscales appeared to make an equal contribution to well-being. They interpreted findings to
mean that although different elements of self-compassion may differentially explain its link with
wellbeing, all are essential to the construct of self-compassion as a whole.
For most researchers, use of the SCS as a total score will be most appropriate given that
self-compassion operates as a system. This view is supported by findings from intervention
research indicating that self-compassion training changes all six components at the same time.
The vast majority of intervention studies using a wide variety of methodologies that examined
change in self-compassion have documented a simultaneous change in all six subscales of
roughly the same magnitude: e.g. self-compassion meditation training (e. g., Toole & Craighead,
2016); online psycho-education (e.g., Krieger, Martig, van den Brink, & Berger, 2016);
Compassion Focused Therapy (e.g., Beaumont, Irons, Rayner, & Dagnall, 2016); Compassionate
Mind Training (e.g., Arimitsu, 2016) and the Mindful Self-Compassion program (e.g., Neff,
2016a). Not only do self-compassion interventions impact CS and RUS to the same degree,
changes in both impact outcomes similarly. Krieger, Berger, and Holtforth (2016) used cross-
lagged analyses to explore whether changes in self-compassion over the course of cognitive-
behavioral psychotherapy led to changes in depression, and findings were the same whether a
total score or two scores representing compassionate or uncompassionate responding were
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
25
examined. They interpreted findings as evidence that self-compassion should be considered an
overall construct rather than two separate constructs. Similarly, Neff (2016a) found that changes
in SCS subscales representing CS and RUS after eight weeks of self-compassion training tended
to be equally predictive of changes in happiness, life satisfaction, anxiety, depression and stress.
These findings suggest that self-compassion is experienced holistically. They also buttress
current study findings supporting the use of a total SCS score to represent self-compassion as
defined by Neff (2003b). Perhaps most importantly, they highlight why there is so much
excitement about the construct of self-compassion in the field of psychology: It is a skill that can
be learned (Neff & Germer, 2013). For researchers who are primarily interested in self-
compassion as a trainable mind-state, therefore, use of a total score is probably most appropriate.
For those more interested in unpacking the mechanisms of how self-compassion enhances well-
being, however, it may be useful to examine the six constituent components themselves.
An important contribution of the present investigation is the finding that self-compassion
is better represented with a single continuum rather than two distinct dimensions of CS and RUS.
This notion was supported by the fact that the positively and negatively valenced items loaded on
the general factor in a similar magnitude in the model including one general factor, whereas these
loadings were weak in the model with two correlated general factors. It should be noted that the
separation of positive and negative items sometimes results from a clustering effect where items
with a similar valence load onto separate factors, basically forming method factors that mostly
originate from the positive versus negative wording of the items (Crego & Widiger, 2014). This
has been shown in research on self-esteem (Greenberger, Chen, Dmitrieva, & Farruggia, 2003;
Marsh, 1996), for instance, where method factors emerged as a results of item wording.
Generally, wording effects would be interpreted as substantively irrelevant artifacts, but in the
case of the SCS, we do not believe that the separation of positively and negatively-valenced items
are a result of item wording only. Rather, the distinction between compassionate and reduced
uncompassionate responding toward oneself is conceptually meaningful and substantially
contributes to the global self-compassion factor. Self-compassion can be conceptualized as a
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
26
holistic state of mind representing the balance of CS and RUS along the three basic dimensions of
emotional responding, cognitive understanding, and paying attention to personal distress.
Limitations and future directions
While this is one of the most comprehensive examinations of the factor structure of the
SCS conducted to date, there were some limitations. For instance, the populations included were
majority female and mainly community and student samples: only one meditator and two clinical
samples were included. Fit in these latter samples was excellent, providing some confidence in
use of the SCS with these populations. Still, it will be important to make sure the factor structure
replicates in specific types of populations (anxious, eating disordered, etc.). Also, although
findings support use of the SCS in different cultures, reliability coefficients and model fit did
vary somewhat across samples (less so for our chosen models). Also, in some countries (e.g.,
China and Japan) multiple measurement models presented identification issues, and it should be
investigated whether these issues relate to model misspecification or sampling-specific errors.
Potential differences in the SCS structure should also be addressed with analyses of invariance
across culture, population type, age and sex, as these may be additional sources of meaningful
variation in the SCS that should be understood. (These analyses are being conducted for the
current dataset and will be presented in a separate paper; Neff et al., 2018).
Given the superiority demonstrated by the ESEM models over CFA models, results
suggest that future attempts to validate translations of the SCS or to examine the properties of the
SCS in specific populations should use this approach (syntax files are available for interested
readers in Appendix Five of the supplementary materials). Additional studies are also needed to
examine the criterion-validity of test score interpretations using this improved representation in
order to better capture the meaning of the subscales once the global level of self-compassion is
accounted for. Use of the bifactor ESEM framework aligns with the proposition of Marsh and
Hau (2007) who emphasized the need for the use of latent variable models which, compared to
observed variables, more accurately define constructs with the explicit inclusion of measurement
errors related to the imperfect items. Bifactor ESEM models are rather complex and sometimes
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
27
difficult to incorporate into predictive models due to the relatively high number of estimated
parameters, but one can construct separate measurement models and “translate” these
measurement models into factor scores saved from these preliminary measurement models that
are better at preserving the a priori nature of the constructs compared to observed variables
(Morin, Meyer, Creusier, & Biétry, 2016; Gillet, Morin, Cougot, & Gagné, 2017). Neff, Tóth-
Király, and Colosimo (2018) successfully used this approach to examine the incremental validity
of self-compassion and neuroticism in predicting wellbeing. (Syntax for saving factor scores is
also included in Appendix Five of the supplementary materials.)
Note that while the current study was aimed at examining the validity of test score
interpretations on the SCS as a measure of Neff’s (2003b) conceptualization of self-compassion,
in no way can it speak to the issue of whether this definition or measurement of self-compassion
is superior to others. For example, Social Mentality Theory (SMT; Gilbert, 1989, 2005) posits
that self-compassion is a state of mind that emerges from mammalian bio-social roles involving
care-giving and care-seeking, while self-criticism emerges from evolved social roles that protect
us from social threats. To this end Gilbert and colleagues developed the Forms of Self-Criticism
and Self-Reassurance Scales (Gilbert, Clarke, Hempel, Miles & Irons, 2004) to measure these
two ways of relating to oneself. More recently, Gilbert and colleagues (Gilbert et al., 2017) have
developed a model of compassion for self, for others, and from others, based on the broadly used
definition of compassion as sensitivity to suffering with a commitment to try to alleviate
it (Goertz, Keltner, & Simon-Thomas, 2010). They developed the Compassion Engagement and
Action Scales, including self-compassion and other compassion scales with items tapping into
engagement with distress (e.g. tolerating and being sensitive to distress) and the motivation to
alleviate that distress (e.g., thinking about and taking actions to help). Notably, the scales do not
include kindness/concern or shared humanity as a feature of compassion. As with the SCS (Neff
& Pommier, 2013), scale scores measuring compassion for self and others are only weakly
correlated, with higher levels of compassion being reported for others than the self. It is unclear if
the desire to alleviate distress operates the same way for self and others, however, given that the
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
28
desire to alleviate personal distress overlaps with resistance to distress. Resistance can exacerbate
psychopathology, which is why mindfulness-based clinical approaches such as Acceptance and
Commitment Therapy (Hayes, Strosahl, & Wilson, 1999) and Mindfulness-Based Cognitive
Therapy (Segal, Williams & Teasdale, 2012) are aimed at reducing resistance to personal distress.
Strauss et al. (2016) propose that measures of compassion should include five key
elements: 1) Recognizing suffering; 2) Understanding the universality of suffering in human
experience; 3) Feeling concern for the person suffering 4) Tolerating uncomfortable feelings in
response to suffering, so remaining open to and accepting of the person suffering: and 5)
Motivation to alleviate suffering. While the SCS taps into most of these elements, no items
explicitly address the motivation to alleviate suffering. This is because the motivation to alleviate
the self’s suffering is easily conflated with resistance to personal distress (undermining the fourth
element) in a way that is less problematic in measures of compassion for others. Still, future
research might fruitfully explore whether adding items to the SCS that are focused on the
motivation to help and support oneself in times of distress could strengthen the measurement of
self-compassion.
To summarize, in the 20 diverse samples we examined, the excellent fit of single-bifactor
ESEM and six-factor correlated ESEM models found across samples strongly supports the
conclusion that self-compassion as measured by the SCS can be viewed as a general construct
(explaining 95% of the reliable variance in item responding), comprised of six separate
components. While the constituent elements of self-compassion are distinct and can be measured
separately, they operate in tandem, as suggested by the large body of research examining self-
compassion interventions. Hopefully these findings can help put some of the controversy over the
factor structure of the SCS to rest: A total score rather than two separate scores should be used.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
29
References
Alda, M., Puebla-Guedea, M., Rodero, B., Demarzo, M., Montero-Marin, J., Roca, M., & Garcia-
Campayo, J. (2016). Zen meditation, length of telomeres, and the role of experiential
avoidance and compassion. Mindfulness, 7(3), 651-659.
Arimitsu, K. (2014). Development and validation of the Japanese version of the Self-Compassion
Scale. The Japanese Journal of Psychology, 85 (1), 50–59.
Arimitsu, K. (2016). The effects of a program to enhance self-compassion in Japanese
individuals: A randomized controlled pilot study. The Journal of Positive Psychology,
11(6), 559-571.
Arimitsu, K., Aoki, Y., Furukita, M., Tada, A., & Togashi, R. (2016). Construction and
Validation of a Short Form of the Japanese version of the Self-Compassion Scale.
Komazawa Annual Reports of Psychology, 18, 1-8.
Asparouhov, T., & Muthén, B. (2009). Exploratory structural equation modeling. Structural
Equation Modeling: A Multidisciplinary Journal, 16(3), 397–438.
Asparouhov, T., Muthén, B., & Morin, A. J. S. (2015). Bayesian structural equation modeling
with cross-loadings and residual covariances: Comments on Stromeyer et al. Journal of
Management, 41(6), 1561-1577.
Azizi, A., Mohammadkhani, P., Lotfi, S., & Bahramkhani, M. (2013). The Validity and
Reliability of the Iranian Version of the Self-Compassion Scale. Iranian Journal of
Clinical Psychology, 2(3), 17-23.
Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural equation models. Journal of the
Academy of Marketing Science, 16(1), 74–94.
Bandalos, D. L. (2014). Relative performance of categorical diagonally weighted least squares
and robust maximum likelihood estimation. Structural Equation Modeling: A
Multidisciplinary Journal, 21(1), 102–116.
Beaumont, E., Irons, C., Rayner, G., & Dagnall, N. (2016). Does Compassion-Focused Therapy
training for health care educators and providers increase self-compassion and reduce self-
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
30
persecution and self-criticism? Journal of Continuing Education in the Health
Professions, 36(1), 4-10.
Benda, J., & Reichová, A. (2016). [Psychometric characteristics of the Czech version of the Self-
Compassion Scale] Psychometrické charakteristiky české verze Self-Compassion Scale
(SCS-CZ). Československá psychologie. 60(2), 20-36.
Birnie, K., Speca, M., & Carlson, L. E. (2010). Exploring self-compassion and empathy in the
context of Mindfulness-based Stress Reduction (MBSR). Stress and Health, 26(5), 359-
371.
Bento, E., Xavier, S., Azevedo, J., Marques, M., Freitas, V., Soares, M. J., et al. (2016).
Validation of the self-compassion scale in a community sample of Portuguese pregnant
women. European Psychiatry, 33, S238. doi: 10.1016/j.eurpsy.2016.01.598
Brenner, R. E., Heath, P. J., Vogel, D. L., & Credé, M. (2017). Two is more valid than one:
Examining the factor structure of the Self-Compassion Scale (SCS). Journal of
Counseling Psychology, 64(6), 696-707.
Browne, M. (2001). An overview of analytic rotation in exploratory factor analysis. Multivariate
Behavioral Research, 36, 111-150.
Brunner, M., Nagy, G., & Wilhelm, O. (2012). A tutorial on hierarchically structured constructs.
Journal of Personality, 80(4), 796–846.
Castilho, P., & Pinto-Gouveia, J. (2011). Self-Compassion: Validation of the Portuguese version
of the Self-Compassion Scale and its relation with early negative experiences, social
comparison and psychopathology. Psychologica, 54, 203-231.
Castilho, P., Pinto-Gouveia, J., & Duarte, J. (2015). Evaluating the multifactor structure of the
long and short versions of the self- compassion scale in a clinical sample. Journal of
Clinical Psychology, 71(9), 856–870.
Chang, E. C., Yu, T., Jilani, Z., Fowler, E. E., Yu, E. A., Lin, J., & Hirsch, J. K. (2015). Under
Assault: Understanding The Impact Of Sexual Assault On The Relation Between Hope
And Suicidal Risk In College Students. Journal of Social and Clinical Psychology, 34(3),
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
31
221-238.
Chen, F. F., West, S. G., & Sousa, K. H. (2006). A comparison of bifactor and second-order
models of quality of life. Multivariate Behavioral Research, 41(2), 189-225.
Chen, J., Yan, L., & Zhou, L. (2011). Reliability and validity of Chinese version of Self-
Compassion Scale. Chinese Journal of Clinical Psychology, 19(6), 734-736.
Cleare, S., Gumley, A., Cleare, C. J., & O’Conner, J. C. (2018). An investigation of the factor
structure of the Self-Compassion Scale. Mindfulness, 9(2), 618-628.
Coroiu, A., Kwakkenbos, L., Moran, C., Thombs, B., Albani, C., Bourkas, S., ... & Körner, A.
(2018). Structural validation of the Self-Compassion Scale with a German general
population sample. PloS one, 13(2), e0190771.
de Souza, L. K., & Hutz, C. S. (2016). Adaptation of the Self-Compassion Scale for use in Brazil:
Evidences of construct validity. Trends in Psychology, 24(1), 159-172.
Deniz, M., Kesici, Ş., & Sümer, A. S. (2008). The validity and reliability of the Turkish version
of the Self-Compassion Scale. Social Behavior and Personality, 36(9), 1151-1160.
Costa, J., Marôco, J., Pinto-Gouveia, J., Ferreira, C., & Castilho, P. (2015). Validation of the
psychometric properties of the Self-Compassion Scale. Clinical Psychology &
Psychotherapy, 23, 460-468.
Crego, C. & Widiger, T. A. (2014). Psychopathy, DSM-5, and a caution. Personality Disorders:
Theory, Research, and Treatment, 5(4), 335–347.
Crocker, J., & Park, L. E. (2004). The costly pursuit of self-esteem. Psychological Bulletin, 130,
392-414.
Cunha, M., Xavier, A., & Castilho, P. (2016). Understanding self-compassion in adolescents:
Validation study of the Self-Compassion Scale. Personality and Individual Differences,
93, 56-62.
Dundas, I., Svendsen, J. L., Wiker, A. S., Granli, K. V., & Schanche, E. (2016). Self-compassion
and depressive symptoms in a Norwegian student sample. Nordic Psychology, 68, 58–72.
Finney, S. J., & DiStefano, C. (2006). Non-normal and categorical data in structural equation
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
32
modeling. In G. R. Hancock & R. D. Mueller (Eds.), Structural equation modeling: a
second course (pp. 269– 314). Charlotte: Information Age Publishing.
Furr, R. M., & Bacharach, V. R. (2008). Psychometrics: An Introduction. Thousand Oaks: Sage.
Garcia-Campayo, J., Navarro-Gil, M., Andrés, E., Montero-Marin, J., López-Artal, L., &
Demarzo, M. M. (2014). Validation of the Spanish versions of the long (26 items) and
short (12 items) forms of the Self-Compassion Scale (SCS). Health and quality of life
outcomes, 12(1), 4.
Ghorbani, N., Chen, Z., Saeedi, Z., Behjati, Z.,Watson,P.J. (2013). Sakhtare Ameli Meghyase
Shafeghat e Khod dar Iran (Factorial Structure of Self-Compassion Scale in
Iran) Pazhohesh Haye Karbordi Ravanshenakhti, 4(3), 29-41. (In Persian)
Gignac, G. E. (2016). The higher-order model imposes a proportionality constraint: That is why
the bifactor model tends to fit better. Intelligence, 55, 57-68.
Gilbert, P. (1989). Human nature and suffering. London: Routledge.
Gilbert, P. (2005). Compassion: Conceptualizations, research and use in psychotherapy. London:
Brunner-Routledge.
Gilbert, P., Catarino, F., Duarte, C., Matos, M., Kolts, R., Stubbs, J…..& Basran, J. (2017). The
development of compassionate engagement and action scales for self and others. Journal
of Compassionate Health Care, DOI 10.1186/s40639-017-0033-3
Gilbert, P., Clarke, M., Hempel, S., Miles, J. N., & Irons, C. (2004). Criticizing and reassuring
oneself: An exploration of forms, styles and reasons in female students. British Journal of
Clinical Psychology, 43(1), 31-50.
Gilbert, P., McEwan, K., Matos, M., & Rivis, A. (2011). Fears of compassion: Development of
three self-report measures. Psychology and Psychotherapy, 84, 239–255.
Gillet, N., Morin, A. J. S., Cougot, B., & Gagné, M. (2017). Workaholism profiles: Associations
with determinants, correlates, and outcomes. Journal of Occupational and Organizational
Psychology, 90(4), 559-586.
Goertz, J. L., Keltner, D., & Simon-Thomas, E. (2010). Compassion: An evolutionary analysis
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
33
and empirical review. Psychological Bulletin, 136, 351–374.
Greenberger, E., Chen, C., Dmitrieva, J., & Farruggia, S. P. (2003). Item-wording and the
dimensionality of the Rosenberg Self-Esteem Scale: do they matter? Personality and
Individual Differences, 35(6), 1241-1254.
Halamová, J., Kanovský, M., & Pacúchová, M. (2018). Self-Compassion Scale: IRT
Psychometric Analysis, Validation, and Factor Structure–Slovak Translation.
Psychologica Belgica, 57(4), 190-209.
Hayes, S. C., Strosahl, K. D., & Wilson, K. G. (1999). Acceptance and commitment therapy (p.
6). New York: Guilford Press.
Hu, L. T., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity to
underparameterized model misspecification. Psychological Methods, 3(4), 424-453.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.
Hupfeld, J., & Ruffieux, N. (2011). Validation of a German version of the self-compassion scale
(SCS-D). Zeitschrift für Klinische Psychologie und Psychotherapie, 40(2), 115–123.
Karakasidou, E., Pezirkianidis, C., Galanakis, M., & Stalikas, A. (2017). Validity, Reliability and
Factorial Structure of the Self Compassion Scale in the Greek Population. Journal of
Psychology and Psychotherapy, 7, 313.
Kelly, A. C., Wisniewski, L., Martin-Wagar, C., & Hoffman, E. (2017). Group-based
Compassion-Focused Therapy as an adjunct to outpatient treatment for eating disorders: A
pilot randomized controlled trial. Clinical Psychology & Psychotherapy, 24(2), 475-487.
Kim, K. E., Yi, G. D., Cho, Y. R., Chai, S. H., & Lee, W. K. (2008). The validation study of the
Korean version of the self-compassion scale. Korean Journal of Health Psychology, 13,
1023–1044.
Körner, A., Coroiu, A., Copeland, L., Gomez-Garibello, C., Albani, C., Zenger, M., & Brähler, E.
(2015). The role of self-compassion in buffering symptoms of depression in the general
population. PloS One, 10(10). doi: 10.1371/journal.pone.0136598
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
34
Kotsou, I., & Leys, C. (2016). Self-Compassion Scale (SCS): Psychometric properties of the
French translation and its relations with psychological well-being, affect and depression.
PloS One, 11(4). doi: 10.1371/journal.pone.0152880
Krieger, T., Berger, T., & grosse Holtforth, M. (2016). The relationship of self-compassion and
depression: Cross-lagged panel analyses in depressed patients after outpatient therapy.
Journal of affective disorders, 202, 39-45.
Krieger, T., Hermann, H., Zimmermann, J., & Grosse Holtforth, M. (2015). Associations of self-
compassion and global self-esteem with positive and negative affect and stress reactivity
in daily life: Findings from a smart phone study. Personality and individual
differences, 87, 288-292.
Krieger, T., Martig, D. S., van den Brink, E., & Berger, T. (2016). Working on self-compassion
online: A proof of concept and feasibility study. Internet Interventions, 6, 64-70.
Lee, W. K., & Lee, K. (2010). The validation study of the Korean version of the Self-Compassion
Scale with adult women in the community. Journal of the Korean Neuropsychiatric
Association, 49, 193–200.
Longe, O., Maratos, F. A., Gilbert, P. Evans, G., Volker, F., Rockliff, H., et al. (2009). Having a
word with yourself: Neural correlates of self-criticism and self-reassurance. Neuroimage,
49, 1849–1856.
López, A., Sanderman, R., Smink, A., Zhang, Y., van Sonderen, E., Ranchor, A., & Schroevers,
M. J. (2015). A reconsideration of the Self-Compassion Scale’s total score: self-
compassion versus self-criticism. PloS One, 10(7). doi: 10.1371/journal.pone.0132940
Mantzios, M., Wilson, J. C. & Giannou, K. (2015). Psychometric properties of the Greek versions
of the self-compassion and mindful attention and awareness scales, Mindfulness, 6, 123-
132.
Marsh, H. W. (1996). Positive and negative global self-esteem: A substantively meaningful
distinction or artifactors?. Journal of Personality and Social Psychology, 70(4), 810-819.
Marsh, H. W., Hau, K.-T., & Grayson, D. (2005). Goodness of fit evaluation. In A. Maydeu-
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
35
Olivares & J. McArdle (Eds.), Contemporary Psychometrics (pp. 275–340). NJ: Erlbaum.
Marsh, H. W., Hau, K. T., & Wen, Z. (2004). In search of golden rules: Comment on hypothesis-
testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing
Hu and Bentler's (1999) findings. Structural Equation Modeling, 11(3), 320-341.
Marsh, H. W., Liem, G. A. D., Martin, A. J., Morin, A. J. S., & Nagengast, B. (2011).
Methodological measurement fruitfulness of exploratory structural equation modeling
(ESEM): New approaches to key substantive issues in motivation and engagement.
Journal of Psychoeducational Assessment, 29(4), 322–346.
Marsh, H. W., Morin, A. J. S., Parker, P. D., & Kaur, G. (2014). Exploratory structural equation
modeling: an integration of the best features of exploratory and confirmatory factor
analysis. Annual Review of Clinical Psychology, 10, 85–110.
Montero-Marín, J., Gaete, J., Demarzo, M., Rodero, B., Lopez, L. C. S., & García-Campayo, J.
(2016). Self-criticism: A measure of uncompassionate behaviors toward the self, based on
the negative components of the self-compassion scale. Frontiers in Psychology, 7:1281.
doi: 10.3389/fpsyg.2016.01281
Morin, A. J. S., Arens, A. K., & Marsh, H. W. (2016a). A bifactor exploratory structural equation
modeling framework for the identification of distinct sources of construct-relevant
psychometric multidimensionality. Structural Equation Modeling: A Multidisciplinary
Journal, 23(1), 116–139.
Morin, A. J. S., Arens, A. K., Tran, A., & Caci, H. (2016b). Exploring sources of construct-
relevant multidimensionality in psychiatric measurement: A tutorial and illustration using
the Composite Scale of Morningness. International Journal of Methods in Psychiatric
Research, 25(4), 277-288.
Morin, A. J. S., Boudrias, J. S., Marsh, H.W., Madore, I., & Desrumeaux, P. (2016). Further
reflections on disentangling shape and level effects in person-centered analyses: An
illustration exploring the dimensionality of psychological health. Structural Equation
Modeling, 23, 438-454.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
36
Morin, A. J. S., & Maïano, C. (2011). Cross-validation of the short form of the physical self-
inventory (PSI-S) using exploratory structural equation modeling (ESEM). Psychology of
Sport and Exercise, 12(5), 540–554.
Morin, A. J. S., Marsh, H. W., & Nagengast, B. (2013). Exploratory structural equation
modeling. In G. R. Hancock & R. O. Mueller (Eds.), Structural equation modeling: a
second course (pp. 395–436). Charlotte: Information Age Publishing, Inc.
Morin, A. J. S., Meyer, J. P., Creusier, J., & Biétry, F. (2016). Multiple-group analysis of
similarity in latent profile solutions. Organizational Research Methods, 19(2), 231-254.
Morin, A. J. S., Myers, N. D., & Lee, S. (in press). Modern factor analytic techniques: Bifactor
models, exploratory structural equation modeling (ESEM) and bifactor-ESEM. In G.
Tenenbaum, & Eklund, R.C. (Eds.), Handbook of Sport Psychology, 4th Edition. Wiley.
Muris, P., & Petrocchi, N. (2017). Protection or vulnerability? A meta-analysis of the relations
between the positive and negative components of self-compassion and psychopathology.
Clinical Psychology & Psychotherapy, 24(2), 373-383.
Muthén, L. K., & Muthén, B. O. (1998-2017). Mplus users guide (8th ed.). Muthén & Muthén.
Neff, K. D. (2003a). Development and validation of a scale to measure self-compassion. Self and
Identity, 2, 223-250.
Neff, K. D. (2003b). Self-compassion: An alternative conceptualization of a healthy attitude
toward oneself. Self and Identity, 2, 85-102.
Neff, K. D. (2011). Self-compassion, self-esteem, and well-being. Social and Personality
Compass, 5, 1-12.
Neff, K. D. (2016a). The Self-Compassion Scale is a valid and theoretically coherent measure of
self-compassion. Mindfulness, 7(1), 264-274.
Neff, K. D. (2016b). Does self-compassion entail reduced self-judgment, isolation, and over-
identification? Mindfulness, 7, 791-797.
Neff, K, & Germer, C. (2013). A pilot study and randomized controlled trial of the mindful self-
compassion program. Journal of Clinical Psychology, 69(1), 28-44.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
37
Neff, K. D. & Germer, C. (2017). Self-compassion and psychological wellbeing. In J. Doty (Ed.)
Oxford Handbook of Compassion Science (pp. 371-386). Oxford University Press.
Neff, K. D., Kirkpatrick, K. & Rude, S. S. (2007). Self-compassion and its link to adaptive
psychological functioning. Journal of Research in Personality, 41, 139-154.
Neff, K. D., Long, P., Knox, M. C., Davidson, O., Kuchar, A., Costigan, A…., & Breines (in
press). The forest and the trees: Examining the association of self-compassion and its
positive and negative components with psychological functioning. Self and Identity.
Neff, K. D., Pommier, E. (2013). The relationship between self-compassion and other- focused
concern among college undergraduates, community adults, and practicing meditators. Self
and Identity, 12(2),160-176.
Neff, K. D., Tóth-Király, I., & Colisomo, K. (in press). Self-compassion is best measured as a
global construct and is overlapping with but distinct from neuroticism: A response to
Pfattheicher, Geiger, Hartung, Weiss, and Schindler (2017). European Journal of
Personality.
Neff, K. D., Tóth-Király, I., Yarnell, L. M., Arimitsu, K., Castilho, P. Ghorbani, N., . . .
Mantzios, M. (2018). Examining invariance in the factor structure of the Self-Compassion
Scale in 20 international samples. Manuscript in preparation.
Neff, K. D., Whittaker, T., & Karl, A. (2017). Evaluating the factor structure of the Self-
Compassion Scale in four distinct populations: Is the use of a total self-compassion score
justified? Journal of Personality Assessment, 99, 596-607.
Neff, K. D. & Vonk, R. (2009). Self-compassion versus global self-esteem: Two different ways
of relating to oneself. Journal of Personality, 77, 23-50.
Petrocchi, N., Ottaviani, C., & Couyoumdjian, A. (2013). Dimensionality of self-compassion:
translation and construct validation of the self-compassion scale in an Italian sample.
Journal of Mental Health, (0), 1-6.
Porges, S. W. (2001). The polyvagal theory: Phylogenetic substrates of a social nervous system.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
38
International Journal of Psychophysiology, 42(2), 123-146.
Raykov, T. (1997). Estimation of composite reliability for congeneric measures. Applied
Psychological Measurement, 21(2), 173–184.
Reise, S. P. (2012). The rediscovery of bifactor measurement models. Multivariate Behavioral
Research, 47(5), 667-696.
Reise, S. P., Bonifay, W. E., & Haviland, M. G. (2013). Scoring and modeling psychological
measures in the presence of multidimensionality. Journal of Personality Assessment,
95(2), 129-140.
Reise, S. P., Moore, T. M., & Haviland, M. G. (2010). Bifactor models and rotations: Exploring
the extent to which multidimensional data yield univocal scale scores. Journal of
Personality Assessment, 92(6), 544-559.
Rodriguez, A., Reise, S. P., & Haviland, M. G. (2016a). Applying bifactor statistical indices in
the evaluation of psychological measures. Journal of Personality Assessment, 98, 223-237
Rodriguez, A., Reise, S. P., & Haviland, M. G. (2016b). Evaluating bifactor models: Calculating
and interpreting statistical indices. Psychological Methods, 21(2), 137-150.
Segal, Z. V., Williams, J. M. G., & Teasdale, J. D. (2012). Mindfulness-based cognitive therapy
for depression. Guilford Press.
Souza, L. K. d., & Hutz, C. S. (2016). Adaptation of the self-compassion scale for use in Brazil:
evidences of construct validity. Trends in Psychology, 24(1), 159-172.
Strauss, C., Taylor, B. L., Gu, J., Kuyken, W., Baer, R., Jones, F., & Cavanagh, K. (2016). What
is compassion and how can we measure it? A review of definitions and measures. Clinical
psychology review, 47, 15-27.
Stutts, L., Leary, M., Zeveney, A., & Hufnagle, A. (in press). A longitudinal analysis of the
relationship between self-compassion and the psychological effects of perceived stress. Self
and Identity.
Svendsen, J. L., Osnes, B., Binder, P. E., Dundas, I., Visted, E., Nordby, H., ... & Sørensen, L.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
39
(2016). Trait Self-Compassion Reflects Emotional Flexibility Through an Association
with High Vagally Mediated Heart Rate Variability. Mindfulness, 1-11.
Toole, A. M., & Craighead, L. W. (2016). Brief self-compassion meditation training for body
image distress in young adult women. Body Image, 19, 104-112.
Tóth-Király, I., Bőthe, B., & Orosz, G. (2017). Exploratory structural equation modeling analysis
of the Self-Compassion Scale. Mindfulness, 8(4), 881-892.
Tóth-Király, I., Bőthe, B., Rigó, A., & Orosz, G. (2017). An illustration of the exploratory
structural equation modeling (ESEM) framework on the Passion Scale. Frontiers in
Psychology, 8:1968. doi: 10.3389/fpsyg.2017.01968
Tóth-Király, I., Morin, A. J. S., Bőthe, B., Orosz, G., & Rigó, A. (2018). Investigating the
multidimensionality of need fulfillment: A bifactor exploratory structural equation
modeling representation. Structural Equation Modeling, 25(2), 267-286.
Tóth-Király, I., Orosz, G., Dombi, E., Jagodics, B., Farkas, D., & Amoura, C. (2017). Cross-
cultural comparative examination of the Academic Motivation Scale using exploratory
structural equation modeling. Personality and Individual Differences, 106, 130-135.
Veneziani, C. A., Fuochi, G., & Voci, A. (2017). Self-compassion as a healthy attitude toward the
self: Factorial and construct validity in an Italian sample. Personality and Individual
Differences, 119, 60-68.
Williams, M. J., Dalgleish, T., Karl, A., & Kuyken, W. (2014). Examining the factor structures of
the Five Facet Mindfulness Questionnaire and the Self-Compassion Scale. Psychological
Assessment, 26(2), 407-418.
Worthington, R. L., & Whittaker, T. A. (2006). Scale development research: A content analysis
and recommendations for best practices. The Counseling Psychologist, 34(6), 806-838.
Young, K. S., van der Velden, A. M., Craske, M. G., Pallesen, K. J., Fjorback, L., Roepstorff, A.,
& Parsons, C. (2017). The impact of mindfulness-based interventions on brain activity: A
systematic review of functional magnetic resonance imaging studies. Neuroscience &
Biobehavioral Reviews.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
40
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
41
Table 1
Characteristics of the Total and Individual Samples
Country
Language
Type
Initial N
Final N
Females
M Age (SD)
Total
Combined
11990
11685
8367 (71.6%)
32.29 (8.28)
AUS
English
Community
316
316
240 (75.9%)
37.20 (14.67)
BRA
Brazilian Portuguese
Community
312
312
241 (77.2%)
30.36 (10.76)
CAN
English
Student
395
362
308 (85.1%)
21.23 (4.02)
CHI
Chinese
Community
262
261
255 (97.7%)
37.02 (7.68)
FRA
French
Community
1554
1545
1362 (88.2%)
43.07 (12.48)
GER
German
Community
396
380
303 (79.7%)
29.43 (10.15)
GRE
Greek
Community
981
974
612 (62.8%)
21.99 (6.09)
IRA
Persian
Student
575
448
239 (53.3%)
25.33 (7.38)
ITA
Italian
Community
384
380
257 (67.6%)
33.56 (10.46)
JAP
Japanese
Student
718
718
291 (40.5%)
19.42 (1.16)
KOR
Korean
Student
353
343
180 (52.5%)
38.80 (9.22)
NOR
Norwegian
Student
327
318
189 (59.4%)
23.03 (3.40)
POR 1
Portuguese
Mixed
1128
1101
824 (74.8%)
24.71 (8.01)
POR 2
Portuguese
Clinical
314
297
236 (79.5%)
29.37 (8.43)
SPA
Spanish
Community
434
434
306 (70.5%)
49.71 (10.83)
UK 1
English
Community
1108
1085
969 (89.3%)
21.38 (5.69)
UK 2
English
Clinical
390
390
300 (76.9%)
50.16 (11.08)
US 1
English
Community
984
974
619 (63.6%)
38.17 (12.88)
US 2
English
Student
844
833
486 (58.3%)
21.22 (3.53)
US 3
English
Meditator
215
214
150 (70.1%)
47.36 (11.62)
Note. AUS = Australia; BRA = Brazil; CAN = Canada; CHI = China; FRA = France; GER =
Germany; GRE = Greece; IRA = Iran; ITA = Italy; JAP = Japan; KOR = South Korea; NOR =
Norway; POR = Portugal; SPA = Spain; UK = United Kingdom; US = United States
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
42
Table 2
Goodness-of-Fit indices for the Total and Individual Samples: Two-Factor Correlated Models
Two-Factor Correlated CFA
Two-Factor Correlated ESEM
CFI
TLI
RMSEA
90% CI
WRMR
CFI
TLI
RMSEA
90% CI
WRMR
Total
.90
.89
.10
.09-.10
7.48
.88
.86
.11
.11-.11
6.31
AUS
.93
.92
.10
.09-.10
1.51
.92
.91
.10
.10-.11
1.25
BRA
.94
.93
.08
.08-.09
1.36
.94
.93
.08
.07-.09
1.13
CAN
.89
.88
.09
.09-.10
1.70
.89
.87
.10
.09-.10
1.38
CHI
.96
.96
.10
.10-.11
1.41
.96
.95
.11
.10-.12
1.20
FRA
.89
.88
.11
.11-.11
3.18
.89
.87
.12
.11-.12
2.68
GER
.81
.80
.11
.10-.12
1.89
.84
.81
.10
.10-.11
1.51
GRE
.92
.92
.08
.08-.09
2.19
.90
.88
.10
.10-.10
1.91
IRA
.81
.79
.09
.08-.09
1.88
.90
.88
.07
.06-.07
1.19
ITA
.85
.84
.12
.12-.13
2.15
.87
.84
.12
.11-.12
1.62
JAP
.86
.84
.10
.09-.10
2.64
.78
.73
.13
.12-.13
2.52
KOR
.91
.90
.09
.08-.10
1.79
.94
.93
.08
.07-.09
1.08
NOR
.89
.88
.10
.09-.10
1.68
.88
.86
.10
.10-.11
1.47
POR 1
.92
.92
.10
.10-.10
2.75
.90
.89
.12
.11-.12
2.22
POR 2
.89
.88
.10
.09-.10
1.57
.89
.87
.10
.09-.11
1.37
SPA
.82
.80
.11
.11-.11
2.25
.88
.85
.09
.09-.10
1.43
UK 1
.88
.87
.11
.10-.11
2.68
.88
.85
.11
.11-.11
2.27
UK 2
.89
.88
.09
.08-.09
1.68
.88
.85
.10
.09-.10
1.40
US 1
.91
.90
.10
.09-.10
2.29
.92
.90
.10
.09-.10
1.86
US 2
.83
.81
.11
.10-.11
2.61
.86
.83
.10
.10-.10
1.93
US 3
.92
.91
.10
.09-.11
1.37
.92
.91
.10
.10-.11
1.16
Note. AUS = Australia; BRA = Brazil; CAN = Canada; CHI = China; FRA = France; GER =
Germany; GRE = Greece; IRA = Iran; ITA = Italy; JAP = Japan; KOR = South Korea; NOR =
Norway; POR = Portugal; SPA = Spain; UK = United Kingdom; US = United States
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
43
Table 3
Goodness-of-Fit Indices for the Total and Individual Samples: Six-Factor Correlated Models
Six-Factor Correlated CFA
Six-Factor Correlated ESEM
CFI
TLI
RMSEA
90% CI
WRMR
CFI
TLI
RMSEA
90% CI
WRMR
Total
.95
.94
.07
.07-.07
5.15
.99
.97
.05
.05-.05
1.75
AUS
.94†
.93
.09
.08-.10
1.27
.98
.97
.06
.06-.07
0.52
BRA
.96
.96
.06
.06-.07
1.05
.99
.98
.05
.04-.06
0.51
CAN
.93
.92
.08
.07-.08
1.30
.97
.94
.06
.06-.07
0.60
CHI
.97
.97
.09
.08-.10
1.17
.99†
.99
.06
.05-.07
0.46
FRA
.92
.91
.09
.09-.10
2.54
.98
.96
.06
.06-.06
0.89
GER
.87†
.85
.09
.09-.10
1.53
.98
.97
.05
.04-.05
0.50
GRE
.97
.96
.06
.05-.06
1.40
.98
.97
.05
.05-.06
0.68
IRA
.85†
.83
.08
.07-.08
1.62
.96
.93
.05
.04-.06
0.66
ITA
.91
.90
.10
.09-.10
1.60
.98
.97
.06
.05-.06
0.53
JAP
.93
.92
.07
.07-.07
1.75
.96
.93
.06
.06-.07
0.84
KOR
.92
.91
.09
.08-.09
1.60
.98
.96
.06
.05-.06
0.53
NOR
.93
.92
.08
.07-.08
1.26
.98
.97
.05
.04-.06
0.48
POR 1
.94†
.94
.09
.08-.09
2.20
.99
.97
.06
.05-.06
0.71
POR 2
.92†
.91
.08
.08-.09
1.30
.97
.95
.06
.05-.07
0.56
SPA
.86†
.84
.10
.09-.10
1.90
.97
.95
.05
.05-.06
0.58
UK 1
.94
.93
.08
.07-.08
1.80
.98
.97
.05
.04-.05
0.67
UK 2
.92
.90
.08
.07-.08
1.41
.98
.96
.05
.04-.06
0.55
US 1
.96
.95
.07
.07-.07
1.51
.99
.98
.04
.04-.05
0.57
US 2
.92
.91
.07
.07-.08
1.73
.98
.96
.05
.05-.06
0.67
US 3
.95
.95
.08
.07-.09
1.05
.99
.98
.05
.04-.06
0.43
Note. AUS = Australia; BRA = Brazil; CAN = Canada; CHI = China; FRA = France; GER =
Germany; GRE = Greece; IRA = Iran; ITA = Italy; JAP = Japan; KOR = South Korea; NOR =
Norway; POR = Portugal; SPA = Spain; UK = United Kingdom; US = United States; † These
solutions had model identification issues, suggesting overparameterization.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
44
Table 4
Goodness-of-Fit Indices for the Total and Individual Samples: Bifactor Models
Bifactor CFA
Bifactor ESEM
CFI
TLI
RMSEA
90% CI
WRMR
CFI
TLI
RMSEA
90% CI
WRMR
Total
.85
.82
.12
.12-.12
10.55
.99
.98
.04
.04-.04
1.42
AUS
.92
.90
.11
.10-.11
1.53
.99
.98
.05
.04-.06
0.40
BRA
.93
.92
.09
.08-.09
1.34
.99
.98
.05
.04-.06
0.45
CAN
.85
.82
.12
.11-.12
2.31
.97
.95
.06
.05-.07
0.51
CHI
.95†
.94
.12
.11-.13
1.57
.99†
.99
.05
.04-.06
0.39
FRA
.89
.87
.11
.11-.12
3.23
.99
.98
.05
.05-.05
0.69
GER
.88
.85
.09
.09-.10
1.53
.99
.97
.04
.03-.05
0.43
GRE
.83
.80
.13
.13-.13
3.51
.99
.98
.04
.04-.05
0.53
IRA
.67
.61
.12
.12-.13
2.34
.97
.94
.05
.04-.06
0.57
ITA
.89
.87
.11
.10-.11
1.88
.99
.97
.05
.04-.06
0.45
JAP
no identification
.97†
.95
.06
.05-.06
0.68
KOR
.63
.56
.19
.19-.20
3.96
.98
.97
.05
.04-.06
0.45
NOR
.87
.85
.11
.10-.11
1.70
.99
.97
.05
.04-.06
0.43
POR 1
.83
.80
.15
.15-.15
4.61
.99
.98
.05
.05-.06
0.62
POR 2
.85
.82
.12
.11-.12
1.79
.98
.95
.06
.05-.07
0.48
SPA
.74
.69
.14
.13-.14
2.84
.98
.96
.05
.04-.06
0.48
UK 1
.89
.87
.11
.10-.11
2.73
.99
.98
.04
.04-.05
0.55
UK 2
.82†
.79
.11
.11-.12
2.04
.98
.97
.05
.04-.05
0.48
US 1
.90
.88
.11
.10-.11
2.54
.99
.99
.04
.03-.04
0.49
US 2
.80
.76
.12
.11-.12
3.20
.98
.96
.05
.04-.05
0.57
US 3
.92
.90
.11
.10-.11
1.32
.99
.99
.04
.02-.05
0.36
Note. AUS = Australia; BRA = Brazil; CAN = Canada; CHI = China; FRA = France; GER =
Germany; GRE = Greece; IRA = Iran; ITA = Italy; JAP = Japan; KOR = South Korea; NOR =
Norway; POR = Portugal; SPA = Spain; UK = United Kingdom; US = United States; † These
solutions had model identification issues, suggesting overparameterization.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
45
Table 5
Goodness-of-Fit Indices for the Total and Individual Samples: Correlated Two-Bifactor Models
Correlated Two-Bifactor CFA
Correlated Two-Bifactor ESEM
CFI
TLI
RMSEA
90% CI
WRMR
CFI
TLI
RMSEA
90% CI
WRMR
Total
.96†
.95
.06
.06-.06
4.49
.99
.99
.04
.03-.04
1.20
AUS
.96
.95
.08
.07-.08
1.09
.99
.98
.04
.03-.05
0.36
BRA
.97†
.96
.06
.06-.07
1.02
.99
.98
.04
.03-.05
0.41
CAN
no identification
.98
.97
.05
.04-.06
0.44
CHI
no identification
.99†
.99
.05
.04-.06
0.36
FRA
.95†
.94
.08
.07-.08
2.06
.99
.98
.04
.04-.05
0.59
GER
.90
.88
.08
.08-.09
1.37
.99†
.97
.04
.03-.05
0.42
GRE
no identification
.99
.99
.04
.03-.04
0.46
IRA
no identification
.98
.96
.04
.03-.05
0.51
ITA
.94
.93
.08
.07-.08
1.31
.99
.98
.04
.03-.05
0.40
JAP
no identification
.99†
.97
.04
.04-.05
0.55
KOR
no identification
.99
.97
.05
.04-.06
0.40
NOR
.93
.91
.08
.08-.09
1.31
.99
.98
.04
.03-.05
0.40
POR 1
.96
.96
.07
.07-.08
1.88
.99
.98
.05
.04-.05
0.54
POR 2
no identification
.98
.96
.06
.05-.06
0.44
SPA
no identification
.99
.97
.04
.03-.05
0.44
UK 1
.94
.93
.08
.07-.08
1.74
.99
.99
.03
.03-.04
0.48
UK 2
.93†
.92
.07
.06-.08
1.29
.99
.97
.04
.03-.05
0.44
US 1
.96
.95
.07
.07-.08
1.56
.99
.99
.03
.02-.04
0.42
US 2
.91
.90
.08
.08-.08
1.83
.99
.97
.04
.04-.05
0.52
US 3
no identification
.99
.99
.04
.02-.05
0.33
Note. AUS = Australia; BRA = Brazil; CAN = Canada; CHI = China; FRA = France; GER =
Germany; GRE = Greece; IRA = Iran; ITA = Italy; JAP = Japan; KOR = South Korea; NOR =
Norway; POR = Portugal; SPA = Spain; UK = United Kingdom; US = United States; † These
solutions had model identification issues, suggesting overparameterization.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
46
Table 6
Standardized Factor Loadings for the Six-Factor Correlated CFA and ESEM Solutions of the
Self-Compassion Scale for the Total Sample
CFA
ESEM
SF (λ)1
SK (λ)
SJ (λ)
CH (λ)
IS (λ)
MI (λ)
OI (λ)
Self-kindness
sk5
.74
.69
.04
.12
.01
.02
.01
sk12
.82
.84
.03
.02
.01
.01
.04
sk19
.84
.80
.03
.00
.00
.08
.01
sk23
.75
.26
.42
.08
.05
.31
.16
sk26
.79
.36
.34
.11
.00
.34
.18
Self-judgment
sj1
.76
.09
.61
.05
.06
.05
.18
sj8
.74
.25
.43
.00
.13
.16
.22
sj11
.74
.09
.51
.02
.13
.05
.13
sj16
.83
.05
.49
.02
.23
.03
.20
sj21
.73
.33
.33
.01
.14
.14
.21
Common humanity
ch3
.70
.04
.11
.45
.15
.19
.01
ch7
.65
.08
.06
.97
.04
.15
.04
ch10
.73
.00
.02
.87
.08
.06
.03
ch15
.84
.09
.05
.43
.07
.29
.06
Isolation
is4
.79
.01
.28
.08
.43
.01
.13
is13
.79
.01
.10
.03
.97
.02
.06
is18
.72
.04
.10
.02
.90
.00
.03
is25
.79
.00
.20
.13
.37
.04
.26
Mindfulness
mi9
.66
.13
.14
.09
.08
.53
.33
mi14
.79
.16
.15
.12
.09
.58
.17
mi17
.77
.14
.01
.16
.10
.49
.05
mi22
.72
.38
.02
.13
.08
.29
.06
Over-identification
oi2
.82
.02
.34
.05
.20
.04
.38
oi6
.78
.03
.40
.08
.16
.01
.31
oi20
.68
.06
.07
.01
.01
.20
.69
oi24
.69
.05
.03
.02
.14
.21
.58
Note. CFA = confirmatory factor analysis; ESEM = exploratory structural equation modeling; SF
= specific factor; 1 = Each item loaded on its respective specific factor, while cross-loadings were
constrained to zero; SK = self-kindness; SJ = self-judgment; CH = common humanity; IS =
isolation; MI = mindfulness; OI = over-identification; λ = standardized factor loadings. Target
factor loadings are in bold. Non-significant parameters (p ≥ .05) are italicized.
SCS Factor Structure in 20 Diverse Samples
47
Table 7
Standardized Factor Loadings for the Bifactor CFA and ESEM Solutions of the Self-Compassion Scale for the Total Sample
Bifactor-CFA
Bifactor-ESEM
GF (λ)
SF (λ)1
GF (λ)
SK (λ)
SJ (λ)
CH (λ)
IS (λ)
MI (λ)
OI (λ)
Self-kindness
sk5
.59
.50
.58
.47
.04
.17
.05
.12
.06
sk12
.66
.54
.64
.56
.01
.11
.03
.12
.03
sk19
.67
.53
.68
.50
.03
.08
.07
.12
.07
sk23
.66
.27
.72
.06
.04
.01
.15
.08
.24
sk26
.67
.33
.73
.13
.13
.06
.19
.12
.26
Self-judgment
sj1
.65
.42
.67
.06
.44
.12
.02
.12
.05
sj8
.62
.44
.66
.04
.20
.13
.06
.25
.13
sj11
.64
.39
.70
.09
.15
.12
.01
.14
.03
sj16
.72
.39
.75
.10
.23
.13
.11
.12
.09
sj21
.63
.36
.67
.11
.07
.10
.05
.25
.13
Common humanity
ch3
.51
.39
.46
.09
.11
.38
.05
.24
.03
ch7
.38
.72
.36
.08
.07
.73
.05
.04
.02
ch10
.48
.64
.44
.11
.03
.65
.07
.10
.01
ch15
.63
.36
.58
.08
.07
.35
.05
.27
.12
Isolation
is4
.69
.26
.66
.08
.20
.05
.26
.06
.10
is13
.64
.58
.64
.06
.02
.06
.58
.00
.05
is18
.56
.57
.57
.07
.03
.06
.55
.02
.07
is25
.69
.25
.67
.07
.11
.00
.24
.10
.20
Mindfulness
mi9
.53
.45
.50
.09
.15
.15
.07
.43
.16
mi14
.65
.55
.59
.12
.10
.19
.00
.52
.04
mi17
.64
.39
.61
.08
.07
.17
.03
.40
.06
mi22
.61
.27
.55
.25
.07
.17
.04
.28
.12
Over-identification
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
48
oi2
.75
.23
.69
.05
.34
.07
.17
.04
.27
oi6
.71
.17
.68
.07
.25
.07
.11
.08
.19
oi20
.58
.57
.59
.12
.06
.05
.05
.03
.50
oi24
.60
.42
.60
.11
.08
.03
.12
.05
.41
Note. CFA = confirmatory factor analysis; ESEM = exploratory structural equation modeling; GF = general factor of self-compassion;
SF = specific factor; 1 = Each item loaded on its respective specific factor, while cross-loadings were constrained to zero; SK = self-
kindness; SJ = self-judgment; CH = common humanity; IS = isolation; MI = mindfulness; OI = over-identification; λ = standardized
factor loadings. Target factor loadings are in bold. Non-significant parameters (p ≥ .05) are italicized.
SCS Factor Structure in 20 Diverse Samples
49
Table 8
Standardized Factor Loadings for the Two-Bifactor CFA and ESEM Solutions of the Self-Compassion Scale for the Total Sample
Two-bifactor-CFA
Two-bifactor-ESEM
CS (λ)
RUS (λ)
SF (λ)1
CS (λ)
RUS (λ)
SK (λ)
SJ (λ)
CH (λ)
IS (λ)
MI (λ)
OI (λ)
Self-kindness
sk5
.70
.32
.43
.24
.31
.34
.10
.37
.06
sk12
.78
.39
.48
.32
.39
.30
.13
.40
.09
sk19
.80
.31
.46
.25
.42
.28
.11
.44
.07
sk23
.77
.21
.31
.22
.49
.26
.14
.40
.05
sk26
.79
.08
.36
.17
.45
.32
.09
.47
.03
Self-judgment
sj1
.72
.28
.16
.00
.71
.07
.30
.06
.26
sj8
.69
.33
.15
.11
.67
.06
.21
.07
.23
sj11
.70
.24
.09
.08
.64
.09
.22
.17
.22
sj16
.79
.22
.06
.04
.66
.08
.34
.15
.31
sj21
.69
.23
.35
.18
.67
.09
.09
.14
.19
Common humanity
ch3
.56
.31
.07
.06
.09
.48
.18
.39
.09
ch7
.44
.70
.03
.07
.04
.79
.04
.16
.06
ch10
.54
.59
.08
.07
.10
.75
.05
.23
.11
ch15
.68
.27
.09
.00
.22
.50
.13
.49
.05
Isolation
is4
.74
.14
.10
.01
.50
.11
.45
.13
.28
is13
.69
.52
.32
.03
.31
.09
.68
.20
.21
is18
.62
.52
.32
.02
.26
.07
.63
.15
.21
is25
.74
.12
.18
.02
.46
.17
.39
.11
.37
Mindfulness
mi9
.59
.36
.09
.06
.08
.26
.07
.58
.29
mi14
.71
.48
.01
.13
.15
.29
.16
.76
.15
mi17
.70
.26
.12
.01
.21
.32
.17
.60
.13
mi22
.67
.12
.27
.07
.22
.32
.14
.49
.04
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
50
Over-identification
oi2
.80
.05
.07
.10
.52
.08
.41
.15
.47
oi6
.76
.02
.04
.02
.56
.11
.31
.11
.37
oi20
.63
.86
.19
.01
.28
.08
.16
.20
.68
oi24
.66
.21
.21
.01
.28
.11
.23
.24
.57
Note. CFA = confirmatory factor analysis; ESEM = exploratory structural equation modeling; CS = general factor representing
Compassionate Self-Responding; RUS = general factor representing Reduced Uncompassionate Self-responding; SF = specific factor;
1 = Each item loaded on its respective specific factor, while cross-loadings were constrained to zero; SK = self-kindness; SJ = self-
judgment; CH = common humanity; IS = isolation; MI = mindfulness; OI = over-identification; λ = standardized factor loadings.
Target factor loadings are in bold. Non-significant parameters (p ≥ .05) are italicized.
SCS Factor Structure in 20 Diverse Samples
51
Table 9
Reliability Estimates for the Bifactor ESEM Model for the Total and Individual Samples
Bifactor
ω
ωH
GF
SF
Total
.96
.91
.95
.05
AUS
.98
.93
.95
.05
BRA
.97
.91
.94
.06
CAN
.96
.88
.92
.08
CHI
Negative Residual Variance
FRA
.97
.92
.95
.05
GER
.96
.88
.92
.08
GRE
.97
.91
.94
.06
IRA
.93
.85
.91
.08
ITA
.96
.89
.93
.07
JAP
Negative Residual Variance
KOR
.95
.82
.86
.13
NOR
.96
.89
.93
.07
POR 1
.97
.90
.93
.07
POR 2
.96
.90
.94
.06
SPA
.94
.83
.88
.11
UK 1
.97
.92
.95
.05
UK 2
.96
.89
.93
.07
US 1
.97
.93
.96
.04
US 2
.95
.87
.92
.08
US 3
.98
.93
.95
.05
Note. ω = Omega; ωH = Omega Hierarchical; GF = Reliable variance explained by the general
factor; SF = Reliable variance explained by the specific factors; AUS = Australia; BRA = Brazil;
CAN = Canada; CHI = China; FRA = France; GER = Germany; GRE = Greece; IRA = Iran; ITA
= Italy; JAP = Japan; KOR = South Korea; NOR = Norway; POR = Portugal; SPA = Spain; UK
= United Kingdom; US = United States
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
52
Supplementary Materials
Examining the Factor Structure of the Self-Compassion Scale in 20 Diverse Samples: Support for
Use of a Total Score and Six Subscale Scores
Table of Contents:
Appendix 1 Table S1: Previous studies examining the factor structure of the Self-Compassion
Scale
Appendix 2: Sample recruitment information
Appendix 3: SCS translation information
Appendix 4 Table S2: Goodness-of-Fit Indices for the Total and Individual Samples: One-Factor
Models
Appendix 5: Syntax files for the examined models
Supplementary Material References
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
53
Appendix 1 Table S1
Previous studies examining the factor structure of the Self-Compassion Scale†
Original paper
Nation (Sample
type)
Sample
Size
Examined alternative models
Final chosen
model
Arimitsu (2014)
Japan
N = 366
Mage = 19.6
(1) one-factor first-order CFA
(2) six-factor first-order CFA
(3) six-factor higher-order CFA (with
one G-factor)
(4) six-factor higher-order CFA (with
two correlated G-factors)
(2)
Azizi et al.
(2013)
Iran (student)
N = 265
Mage = 22.1
(1) six-factor first-order CFA
(1)
Benda &
Reichová
(2016)
Czech Republic
(community)
N = 5368
Mage = —
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
both
Bento et al.
(2016)
Portugal (pregnant
women)
N = 417
Mage = 33
(1) six-factor first-order CFA
(1)
Brenner et al.
(2017)
USA (student)
N = 1115
Mage = 19.4
(1) one-factor first-order CFA
(2) two-factor first-order CFA
(3) three-factor first-order CFA
(4) six-factor first-order CFA
(5) six-factor higher-order CFA (with
one G-factor)
(6) six-factor higher-order CFA (with
two G-factors)
(7) bifactor CFA (with one G-factor and
six S-factors)
(8) bifactor CFA (with two uncorrelated
G-factors and six S-factors)
(8)
Castilho et al.
(2015)
Portugal
(community)
N = 1128
Mage = 24.5
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
both
Portugal (clinical)
N = 316
Mage =
28.69
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
both
Chen et al.
(2011)
China (student)
N = 660
Mage = —
(1) EFA (six factors were extracted)
(2) six-factor first-order CFA
both
Cleare et al.
(2018)
United Kingdom
(community)
N = 526
Mage = 23
(1) EFA (five factors were extracted)
(2) six-factor first-order CFA
(3) six-factor higher-order CFA (with
one G-factor)
(4) one-factor CFA
(5) two-factor first-order CFA
(6) bifactor CFA (with one G-factor and
six S-factors)
(7) five-factor CFA (based on EFA)
(6)
Coroiu et al.
(2018)
Germany
(community)
N = 2510
Mage =
50.23
(1) one-factor first-order CFA
(2) two-factor first-order CFA
(3) three-factor first-order CFA
(4) six-factor first-order CFA
(5) six-factor higher-order CFA (with
two correlated G-factors)
(6) bifactor CFA (with one G-factor and
two S-factors)
(7) bifactor CFA (with two uncorrelated
G-factors and two S-factors)
(7)
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
54
(8) bifactor CFA (with one G-factor and
6 S-factors)
Costa et al.
(2015)
Portugal (clinical)
N = 361
Mage =
25.19
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
(3) two-factor first-order CFA
(3)
Cunha et al.
(2016)
Portugal
(adolescent)
N = 3165
Mage =
15.49
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
both
de Souza &
Hutz (2016)
Brazil (community)
N = 432
Mage = 32.5
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
(3) bifactor CFA (with one G-factor and
six S-factors)
(3)
Deniz et al.
(2008)
Turkey (student)
N = 341
Mage =
19.81
(1) six-factor first-order CFA
(2) EFA (one G-factor being extracted)
(2)
Dundas et al.
(2016)
Norway (student)
N = 277
Mage = 22.9
(1) one-factor first-order CFA
(2) three-factor first-order CFA
(3) six-factor higher-order CFA (with
one G-factor)
(3)
Garcia-
Campayo et al.
(2014)
Spain (student)
N = 268
Mage =
20.54
(1) six-factor first-order CFA
(1)
Halamová et al.
(2018)
Slovakia
(community)
N = 1181
Mage =
30.30
(1) six-factor first-order IRT CFA
(2) bifactor IRT CFA (with one G-factor
and six S-factors)
(3) two-bifactor IRT CFA (with two
correlated G-factors and six S-factors)
(3)
N = 676
Mage =
29.90
Hupfeld &
Ruffieux (2011)
Germany
(community)
N = 561
Mage =
26.04
(1) six-factor first-order ESEM
(1)
Karakasidou et
al. (2017)
Greek (community)
N = 642
Mage =
36.83
(1) six-factor first-order CFA
(2) two-factor first-order CFA
(3) six-factor higher-order CFA (with
one G-factor)
(1)
Kotsou & Leys
(2016)
France (community)
N = 1554
Mage =
42.92
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
(3) bifactor CFA (with one G-factor and
six S-factors)
(3)
Lee & Lee
(2010)
Korea (community
females)
N = 405
Mage = —
(1) six-factor first-order CFA
(1)
López et al.
(2015)
Netherlands†
(community)
N = 1643
Mage = 54.9
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
(3) two-factor EFA
(3)
Mantzios et al.
(2015)
Greece (community)
N = 556
Mage =
24.43
(1) six-factor EFA
(1)
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
55
Montero-Marin
et al. (2016)
Brazil (doctors)
N = 406
Mage =
41.09
(1) one-factor first-order CFA
(2) two-factor first-order CFA
(3) six-factor first-order CFA
(4) six-factor higher-order CFA (with
one G-factor)
(5) six-factor higher-order CFA (with
two correlated G-factors)
(6) six-factor higher-order CFA (with
three correlated G-factors)
(7) six-factor third-order CFA (with one
third order G-factor, three uncorrelated
second order G-factors)
(8) bifactor CFA (with one G-factor and
six S-factors)
(9) three-factor first-order CFA (positive
items only)
(10) three-factor higher-order CFA (with
one G-factor, positive items only)
(11) three-factor first-order CFA
(negative items only)
(12) three-factor higher-order CFA (with
one G-factor, negative items only)
(12)
Spain (doctors)
N = 416
Mage =
49.71
Neff (2003)
USA (student)
N = 391
Mage =
20.91
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
(2)
Neff et al.
(2008)
Taiwan†
(student)
N = 164
Mage = 20.5
(1) six-factor first-order CFA
(1)
Thailand†
(student)
N = 223
Mage = 19.8
(1) six-factor first-order CFA
(1)
Neff et al.
(2017)
USA (student)
N = 222
Mage =
20.94
(1) one-factor first-order CFA
(2) two-factor first-order CFA
(3) six-factor first-order CFA
(4) six-factor higher-order CFA (with
one G-factor)
(5) bifactor CFA (with one G-factor and
six S-factors)
(5)
USA (community)
N = 1394
Mage =
36.01
USA (mediator)
N = 215
Mage =
47.40
USA (clinical)
N = 390
Mage =
50.16
Neff et al.
(2018)
USA (community)
N = 576
Mage =
37.21
(1) one-factor first-order CFA
(2) one-factor first-order ESEM
(3) two-factor first-order CFA
(4) two-factor first-order ESEM
(5) six-factor first-order CFA
(6) six-factor first-order ESEM
(7) bifactor CFA (with one G-factor and
six S-factors)
(8) bifactor ESEM (with one G-factor
and six S-factors)
(9) two-bifactor CFA (with two
correlated G-factors and six S-factors)
(8)
USA (community)
N = 581
Mage =
36.40
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
56
(10) two-bifactor ESEM (with two
correlated G-factors and six S-factors)
Petrocchi et al.
(2014)
Italy (community)
N = 424
Mage =
36.53
(1) six-factor first-order CFA
(2) six-factor higher-order CFA (with
one G-factor)
(3) one-factor first-order CFA
(4) two-factor first-order CFA
(1)
Pfattheicher et
al. (2017)
USA (community)
N = 576
Mage =
37.21
(1) one-factor first-order CFA
(2) two-factor first-order CFA
(3) six-factor higher-order CFA (with
one G-factor)
(4) six-factor higher-order CFA (with
two G-factors)
(5) six-factor first-order CFA
(4)
Tóth-Király et
al. (2017)
Hungary
(community)
N = 505
Mage =
44.37
(1) six-factor first-order CFA
(2) six-factor first-order ESEM
(3) bifactor CFA (with one G-factor and
six S-factors)
(4) bifactor ESEM (with one G-factor
and six S-factors)
(4)
Veneziani et al.
(2017)
Italy (community)
N = 522
Mage =
30.05
(1) one-factor first-order CFA
(2) two-factor first-order CFA
(3) six-factor first-order CFA
(4) six-factor higher-order CFA (with
one G-factor)
(5) bifactor CFA (with one G-factor and
two S-factors)
(6) bifactor CFA (with one G-factor and
six S-factors)
(6)
Williams et al.
(2014)
UK (community)
N = 821
Mage = —
(1) one-factor first-order CFA
(2) six-factor first-order CFA
(3) six-factor higher-order CFA (with
one G-factor)
(2)
UK (meditator)
N = 211
Mage = —
UK (clinical)
N = 390
Mage = —
Zeng et al.
(2016)
China (Buddhist
community)
N = 179
Mage = 35.5
(1) six-factor first-order CFA
(2) two-factor first-order CFA
(3) three-factor first-order CFA (separate
for positive and negative items)
(3)
China (non-Buddhist
community)
N = 232
Mage = 30.1
Note. Literature search was performed on April 10, 2018.; † = Examined an unpublished/unvalidated translation of
the SCS; Mage = mean age; N = number of participants; CFA = confirmatory factor analysis; EFA = exploratory
factor analysis; ESEM = exploratory structural equation modeling; IRT = item response theory; G-factor = general
factor; S-factor = specific factor.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
57
Appendix 2: Sample recruitment information
Australia
This sample was not part of a previously published study. Participants were recruited for
this community sample (N = 316) through announcements posted online and surveys were
completed online. Participants could choose to enter a lottery draw for the chance to win one of
four $50 gift card prizes. Participants were 75.9% females, M age = 37.20 (SD = 14.67).
Education information is not available.
Brazil
See Souza, Ávila-Souza and Gauer (2016) for a full description of recruitment procedures.
Participants for this community sample (N = 312) were recruited through online announcements,
along with the link for a widely used research platform. Participants were 77.2% females, M age
= 30.36 (SD = 10.76). 70% of participants had a university degree.
Canada
See Sirois, Kitner and Hirsch (2015) for a full description of recruitment procedures.
Participants for this student sample (N = 395) were randomly selected from a psychology subject
pool at a large Canadian university. Participants were 85.1% females, M age = 21.23 (SD = 4.02).
China
This sample was not part of a previously published study. Participants in this community
sample (N = 262) signed up for a Mindful Self-Compassion course, and took the SCS at pre-test.
Participants were 97.7% females, M age = 37.02 (SD = 7.68). 8% had a high school education
only, 64% had a college degree, and 23% had a graduate degree.
France
See Kotsou and Leys (2016) for a full description of recruitment procedures. Participants
for this community sample (N = 1554) were recruited through an announcement posted online,
and surveys were completed online. Participants were 88.2% females, M age = 43.07 (SD =
12.48). 64% of participants had at least an undergraduate level of education.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
58
Germany
See Hupfield and Ruffieux (2011) for a full description of recruitment procedures.
Participants for this community sample (N = 396) were recruited via online portals for German-
speaking people with interest in psychological research... Participants were 79.7% females, M age
= 29.43 (SD = 10.15). 71% of participants had at least an undergraduate level of education.
Greece
This sample was not part of a previously published study. Participants were recruited for
this community sample (N = 981) by sending an email to employees of four Greek Universities
inviting them to participate in an online survey, with instructions to pass the link on to other
individuals or groups who might be interested in taking part, but avoid sharing with students.
Following the same procedure, the survey link was also disseminated to employees of two Greek
Hospitals - with instructions to avoid sending the link to patients. The aim was to obtain a
nonclinical and a nonstudent sample, which would be broadly representative of the general
population in Greece. Participants were 62.8% females, M age = 21.99 (SD = 6.09). Education
information is not available.
Iran
This sample was not part of a previously published study. Undergraduate and graduate
seminary students were recruited for this student sample (N = 575). Participants were 53.3%
females, M age = 25.33 (SD = 7.38). 24% were graduate students holding a Bachelor's degree or
higher.
Italy
See Petrocchi et al. (2014) for a full description of recruitment procedures. Participants for
this community sample (N = 384) were recruited via several professional mailing lists and
completed on online survey. Participants were 67.6% females, M age = 33.56 (SD = 10.46). Most
respondents had finished high school (38.2%), 19.6% had a Bachelor’s degree, 42.2% had a
Master’s degree or higher.
Japan
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
59
See Arimitsu, Aoki, Furukita, Tada & Togashi (2016) for a full description of recruitment
procedures. Participants for this student sample (N = 718) were recruited from two metropolitan
Japanese universities. Participants were 40.5% females, M age = 19.42 (SD = 1.16).
Korea
See Woo Kyeong (2013) for a full description of recruitment procedures. Participants for
this student sample (N = 315) were randomly recruited from a counseling psychology subject
pool at an online university. Participants were 52.5% females, M age = 38.80 (SD = 9.22).
Norway
See Dundas et al. (2016) for a full description of recruitment procedures. Participants for
this undergraduate student sample (N = 327) were recruited from a Norwegian university
(medical and psychology students) and a university college (engineering students). Participants
were 59.4% females, M age = 23.03 (SD = 3.40).
Portugal 1
See Castilho, Pinto-Gouveia, and Duarte (2015) for a full description of recruitment
procedures. Participants for this mixed student-community sample (N = 1128) were recruited
from two large universities in Portugal and also from community groups in Portugal using
nonrandom methods. Students were informed of the study by announcements made at the end of
lectures, and participants from the community sample were recruited in several Portuguese
institutions. Participants were 74.8% females, M age = 24.71 (SD = 8.01). 78% of participants
were students.
Portugal 2
See Castilho, Pinto-Gouveia, and Duarte (2015) for a full description of recruitment
procedures. Participants for this clinical sample (N = 316) were recruited from the outpatient
psychiatric services of different public hospitals in Portugal and were referred by the
psychologists and psychiatrists in charge. A trained therapist clinically assessed all participants
using diagnostic structured interviews: Structured Clinical Interview for DSM-IV Axis I
Disorders, Anxiety Disorders Interview Schedule for DSM-IV, Structured Clinical Interview for
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
60
DSM-IV Axis II Personality Disorders, and Borderline Personality Disorder Severity Index. Only
patients with Axis I and II disorders participated in the study. Participants were 80% females. M
age = 28.69 (SD = 8.74). 40% of participants were students.
Spain
See Montero-Marín, Zubiaga, et al. (2016) for a full description of recruitment
procedures. Participants for this community sample (N = 434) were health care professionals who
were randomly recruited from the mailing list of the Aragon Health Service. Participants were
70.5% females, M age = 49.71 (SD = 10.83). 49% subjects were physicians, 42% were nurses,
and 10% were residents.
United Kingdom 1
This sample was not part of a previously published study. Participants for this community
sample (N = 1108) were recruited by sending an invitation email to employees of two British
Universities to take an online survey, with instructions to pass the invitation on to other
individuals or groups who might be interested in taking part, but to avoid sharing the online link
with students. The aim was to obtain a nonstudent sample, which would be broadly representative
of a community sample in the United Kingdom. Participants were 89.3% females, M age = 21.38
(SD = 5.69). Education information is not available.
United Kingdom 2
See Williams et al. (2014) for a full description of recruitment procedures. Participants for
this clinical sample (N = 405) were recruited through primary care settings in the United
Kingdom. Criteria for this group included having a diagnosis of recurrent major depressive
disorder in full or partial remission according to the Diagnostic and Statistical Manual of Mental
Disorders (4th ed.; DSM–IV), having three or more previous major depressive episodes, and
being 18 or older. Participants were 76.6% females, M age = 50.16 (SD = 11.8). 22% of
participants had some education, 41% of participants had a high school or vocational education,
and 32% had a university degree or other professional qualification.
United States 1
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
61
This sample was not part of a previously published study. Participants were recruited for
this community sample (N = 984) from Mechanical Turk. Participants were directed to Survey
Monkey in order to take the study, and were paid 30 cents for completing it. Participants were
63.6% females, M age = 38.17 (SD = 12.88). 37% of participants reported having a 4-year college
degree, 21% completed some college, 14% had a two-year degree, 18% had pursued graduate
school, and 9% had a high school degree or less.
United States 2
See Chang et al. (2015) and Yarnell & Neff (2013) for a full description of recruitment
procedures for this student sample. Data were combined from two studies in order to increase
sample size (N = 844), given that the SCS was developed in a US student sample. Participants
were randomly selected from subject pools at two large southern American universities.
Participants were 58.3% females, M age = 21.22 (SD = 3.53).
United States 3
See Neff et al. (2017) for a full description of recruitment procedures. Participants for this
sample of meditators (N = 215) were recruited via an e-mail that invited them to complete an
online questionnaire via Survey Monkey. E-mails were sent to individuals affiliated with Seattle
Insight Meditation Society, Spirit Rock, the Insight Meditation Society, and similar groups.
Participants reported a wide range in meditation experience from beginner to advanced (1 to 20
years of meditation practice). The average length of meditation practice for the sample was 6.67
years (SD = 3.86). Participants were 70% females, M age = 47.40 (SD = 12.88). Education
information not available.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
62
Appendix 3: SCS translation information
Brazilian Portuguese (Souza & Hutz, 2016)
Translation procedure: Two forward translations, two focus groups (research team and licensed
psychologists), community and major undergraduate samples, two bilingual experts in
psychometrics, back-translation by a bilingual Buddhist, followed by a final check by K. Neff,
were the steps taken to ensure cultural validation.
Internal Consistency: Cronbach's α = .92 for the total score, ranging from .66 to .81 for the
subscales.
Test-retest reliability: N/A
Factor structures tested: Both single- and six-factor correlated CFA model displayed a good fit.
Chinese (Chen, Yan & Zhou, 2011)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .84
Test-retest reliability: r = .89
Factor structures tested: A six-factor EFA and CFA model displayed a good fit.
French (Kotsou & Leys, 2016)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .94
Test-retest reliability: r = .85
Factor structures tested: A higher-order one-factor, bi-factor, and a six-factor correlated model
were calculated. The six-factor correlated CFA model displayed a good fit, while the higher-order
one-factor model displayed a weaker fit and the bi-factor showing an acceptable fit.
German (Hupfeld & Ruffieux, 2011)
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
63
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .91
Test-retest reliability: r = .92
Factor structures tested: A six-factor correlated ESEM model displayed a good fit.
Greek (Mantzios, Wilxon & Giannou, 2015)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .87
Test-retest reliability: r = .89
Factor structures tested: N/A
Italian (Petrocchi, Ottaviani & Couyoumdjian, 2013)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .90
Test-retest reliability: r = .85
Factor structures tested: A six-factor correlated CFA model displayed an adequate fit after
removing two items. A one-factor, higher order one-factor model, and two-factor models were
displayed a poor fit.
Japanese (Arimitsu, 2014)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .84
Test-retest reliability: r = .83
Factor structures tested: A one-factor, higher order one-factor model, higher two-factor, and a six-
factor correlated model were calculated. In comparison, the six-factor correlated CFA model
displayed the best fit.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
64
Korean (Kim, Yi, Cho, Chai & Lee, 2008)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .90
Test-retest reliability: r = .85
Factor structures tested: A six-factor correlated CFA model displayed a good fit.
Norwegian (Dundas, Svendsen, Wiker, Granli & Schanche, 2016)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .89
Test-retest reliability: N/A
Factor structures tested: A one-factor, two-factor, three-factor, and a six-factor correlated model
were calculated. In comparison, the six-factor CFA model displayed the best fit.
Persian (Ghorbani, Chen, Saeedi, Behjati, Watson, 2013)
Translation procedure: A standard forward-backward translation method was utilized, but
psychometric analyses were not conducted. However, information for a similar Persian
translation (Azizi, Mohammadkhani, Lotfi & Bahramkhani, 2013), reported below.
Internal Consistency: Cronbach's α = .78
Test-retest reliability: N/A
Factor structures tested: CFA found a marginally good fit for a six-factor correlated model.
Portuguese (Castilho & Pinto-Gouveia, 2011)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .89
Test-retest reliability: r = .78
Factor structures tested: Both single and six-factor correlated CFA model displayed a good fit.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
65
Spanish (Garcia-Campayo, Navarro-Gil, Andrés, Montero-Marin, López-Artal, &
Demarzo, 2014)
Translation procedure: A standard forward-backward translation method was utilized.
Internal Consistency: Cronbach's α = .87
Test-retest reliability: ICC = .92
Factor structures tested: A six-factor CFA model displayed a good fit.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
66
Appendix 4 Table S2
Goodness-of-Fit Indices for the Total and Individual Samples: One-Factor Models
One-Factor CFA
One-Factor ESEM
CFI
TLI
RMSEA
90% CI
WRMR
CFI
TLI
RMSEA
90% CI
WRMR
Total
.74
.73
.15
.15-.15
14.44
N/A
AUS
.86
.85
.13
.13-.14
2.11
N/A
BRA
.88
.87
.11
.11-.12
1.83
N/A
CAN
.76
.75
.14
.13-.14
2.97
N/A
CHI
.93
.92
.14
.14-.15
2.10
N/A
FRA
.80
.79
.15
.14-.15
4.67
N/A
GER
.75
.73
.13
.12-.13
2.23
N/A
GRE
.74
.72
.16
.15-.16
4.68
N/A
IRA
.60
.56
.13
.12-.13
2.69
N/A
ITA
.76
.74
.15
.15-.16
2.93
N/A
JAP
.49
.45
.18
.18-.18
5.56
N/A
KOR
.55
.52
.20
.20-.21
4.51
N/A
NOR
.78
.77
.13
.13-.14
2.30
N/A
POR 1
.76
.74
.17
.17-.17
5.89
N/A
POR 2
.77
.75
.14
.13-.14
2.34
N/A
SPA
.64
.61
.15
.15-.16
3.49
N/A
UK 1
.78
.77
.14
.14-.14
3.98
N/A
UK 2
.73
.70
.14
.13-.14
2.61
N/A
US 1
.83
.81
.13
.13-.14
3.54
N/A
US 2
.69
.66
.14
.14-.15
4.28
N/A
US 3
.84
.83
.14.
.14-.15
1.94
N/A
Note. AUS = Australia; BRA = Brazil; CAN = Canada; CHI = China; FRA = France; GER =
Germany; GRE = Greece; IRA = Iran; ITA = Italy; JAP = Japan; KOR = South Korea; NOR =
Norway; POR = Portugal; SPA = Spain; UK = United Kingdom; US = United States
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
67
Appendix 5: Syntax files for the examined models
Model 1a: One-factor CFA
! Commands preceded by ! sign are comments that Mplus ignores.
DATA:
FILE IS C:\Users\scs.dat;
! Path to and name of data file changes per study.
VARIABLE:
MISSING ARE ALL (9999);
NAMES ARE
scsj1 scoi2 scch3 scis4 scsk5 scoi6 scch7 scsj8 scmi9 scch10 scsj11
scsk12 scis13 scmi14 scch15 scsj16 scmi17 scis18 scsk19 scoi20
scsj21 scmi22 scsk23 scoi24 scis25 scsk26;
USEVARIABLES ARE
scsj1 scoi2 scch3 scis4 scsk5 scoi6 scch7 scsj8 scmi9
scch10 scsj11 scsk12 scis13 scmi14 scch15 scsj16 scmi17
scis18 scsk19 scoi20 scsj21 scmi22 scsk23 scoi24 scis25
scsk26;
! Specifying that we’re treating the variables as categorical.
CATEGORICAL ARE all;
ANALYSIS:
! Requesting the weighted least squares mean- and variance-adjusted estimator
estimator = wlsmv;
MODEL:
! Specifying the latent self-compassion factor with the ‘BY’ statement
sc BY
scsk5* scsk12 scsk19 scsk23 scsk26
scsj1 scsj8 scsj11 scsj16 scsj21
scch3 scch7 scch10 scch15
scis4 scis13 scis18 scis25
scmi9 scmi14 scmi17 scmi22
scoi2 scoi6 scoi20 scoi24;
sc@1;
! Requesting standardized parameter estimates
OUTPUT: stdyx;
Model 1b: One-factor ESEM
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
68
MODEL:
sc BY
scsk5 scsk12 scsk19 scsk23 scsk26
scsj1 scsj8 scsj11 scsj16 scsj21
scch3 scch7 scch10 scch15
scis4 scis13 scis18 scis25
scmi9 scmi14 scmi17 scmi22
scoi2 scoi6 scoi20 scoi24 (*1);
Model 2a: Two-factor CFA
! All syntax preceding this model are the same as in the previous CFA models unless indicated
! otherwise.
MODEL:
pos BY scsk5* scsk12 scsk19 scsk23 scsk26
scch3 scch7 scch10 scch15
scmi9 scmi14 scmi17 scmi22;
neg BY scsj1* scsj8 scsj11 scsj16 scsj21
scis4 scis13 scis18 scis25
scoi2 scoi6 scoi20 scoi24;
pos@1; neg@1;
Model 2b: Two-factor ESEM
ANALYSIS:
estimator = wlsmv;
rotation = target;
! Target rotation was used in all models in conjunction with the (~) sign for all ESEM models.
! Cross-loadings are targeted to be as close to zero as possible
MODEL:
pos BY scsk5 scsk12 scsk19 scsk23 scsk26
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3 scch7 scch10 scch15
scis4~0 scis13~0 scis18~0 scis25~0
scmi9 scmi14 scmi17 scmi22
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
neg BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1 scsj8 scsj11 scsj16 scsj21
scch3~0 scch7~0 scch10~0 scch15~0
scis4 scis13 scis18 scis25
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2 scoi6 scoi20 scoi24 (*1);
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
69
Model 3a: Six-factor CFA
! All syntax preceding this model are the same as in the previous CFA models unless indicated
! otherwise.
MODEL:
! self-kindness
sk BY scsk5* scsk12 scsk19 scsk23 scsk26;
! self-judgment
sj BY scsj1* scsj8 scsj11 scsj16 scsj21;
! common humanity
ch BY scch3* scch7 scch10 scch15;
! isolation
is BY scis4* scis13 scis18 scis25;
! mindfulness
mi BY scmi9* scmi14 scmi17 scmi22;
! overidentification
oi BY scoi2* scoi6 scoi20 scoi24;
sj@1; oi@1; ch@1; sk@1; mi@1; is@1;
Model 3b: Six-factor ESEM
! All syntax preceding this model are the same as in the previous ESEM models unless indicated
! otherwise.
MODEL:
sk BY scsk5 scsk12 scsk19 scsk23 scsk26
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
sj BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1 scsj8 scsj11 scsj16 scsj21
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
ch BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3 scch7 scch10 scch15
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
is BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4 scis13 scis18 scis25
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
70
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
mi BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9 scmi14 scmi17 scmi22
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
oi BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2 scoi6 scoi20 scoi24 (*1);
Model 4a: Bifactor-CFA (1 G- and 6 S-factors)
! All syntax preceding this model are the same as in the previous CFA models unless indicated
! otherwise.
MODEL:
sc BY scsk5* scsk12 scsk19 scsk23 scsk26 scsj1 scsj8 scsj11 scsj16
scsj21 scch3 scch7 scch10 scch15 scis4 scis13 scis18 scis25
scmi9 scmi14 scmi17 scmi22 scoi2 scoi6 scoi20 scoi24;
sk BY scsk5* scsk12 scsk19 scsk23 scsk26;
sj BY scsj1* scsj8 scsj11 scsj16 scsj21;
ch BY scch3* scch7 scch10 scch15;
is BY scis4* scis13 scis18 scis25;
mi BY scmi9* scmi14 scmi17 scmi22;
oi BY scoi2* scoi6 scoi20 scoi24;
sc@1; sj@1; oi@1; ch@1; sk@1; mi@1; is@1;
sc WITH sk-oi@0;
sk WITH sj-oi@0;
sj WITH ch-oi@0;
ch WITH is-oi@0;
is WITH mi-oi@0;
mi WITH oi@0;
Model 4b: Bifactor-ESEM (1 G- and 6 S-factors)
! All syntax preceding this model are the same as in the previous ESEM models unless indicated
! otherwise.
ANALYSIS:
estimator = wlsmv;
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
71
rotation = target (orthogonal);
! Factors are specified as orthogonal to each other.
MODEL:
sc BY scsk5 scsk12 scsk19 scsk23 scsk26 scsj1 scsj8 scsj11 scsj16
scsj21 scch3 scch7 scch10 scch15 scis4 scis13 scis18 scis25
scmi9 scmi14 scmi17 scmi22 scoi2 scoi6 scoi20 scoi24 (*1);
MODEL:
sk BY scsk5 scsk12 scsk19 scsk23 scsk26
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
sj BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1 scsj8 scsj11 scsj16 scsj21
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
ch BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3 scch7 scch10 scch15
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
is BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4 scis13 scis18 scis25
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
mi BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9 scmi14 scmi17 scmi22
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
oi BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2 scoi6 scoi20 scoi24 (*1);
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
72
Model 5a: Two-bifactor (two-tier) CFA model (2 G- and 6 S-factors)
! All syntax preceding this model are the same as in the previous CFA models unless indicated
! otherwise.
! positive items
po BY scsk5* scsk12 scsk19 scsk23 scsk26
scch3 scch7 scch10 scch15
scmi9 scmi14 scmi17 scmi22;
! negative items
ne BY scsj1* scsj8 scsj11 scsj16 scsj21
scis4 scis13 scis18 scis25
scoi2 scoi6 scoi20 scoi24;
sk BY scsk5* scsk12 scsk19 scsk23 scsk26;
sj BY scsj1* scsj8 scsj11 scsj16 scsj21;
ch BY scch3* scch7 scch10 scch15;
is BY scis4* scis13 scis18 scis25;
mi BY scmi9* scmi14 scmi17 scmi22;
oi BY scoi2* scoi6 scoi20 scoi24;
po@1; ne@1; sj@1; oi@1; ch@1; sk@1; mi@1; is@1;
! general factors are allowed to correlate with each other
po WITH sk-oi@0;
ne WITH sk-oi@0;
sk WITH sj-oi@0;
sj WITH ch-oi@0;
ch WITH is-oi@0;
is WITH mi-oi@0;
mi WITH oi@0;
Model 5b: Two-bifactor (two-tier) ESEM model (2 G- and 6 S-factors)
! All syntax preceding this model are the same as in the previous ESEM models unless indicated
! otherwise.
ANALYSIS:
estimator = wlsmv;
rotation = target (orthogonal);
MODEL:
sk BY scsk5 scsk12 scsk19 scsk23 scsk26
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
sj BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1 scsj8 scsj11 scsj16 scsj21
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
73
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
ch BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3 scch7 scch10 scch15
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
is BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4 scis13 scis18 scis25
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
mi BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9 scmi14 scmi17 scmi22
scoi2~0 scoi6~0 scoi20~0 scoi24~0 (*1);
oi BY scsk5~0 scsk12~0 scsk19~0 scsk23~0 scsk26~0
scsj1~0 scsj8~0 scsj11~0 scsj16~0 scsj21~0
scch3~0 scch7~0 scch10~0 scch15~0
scis4~0 scis13~0 scis18~0 scis25~0
scmi9~0 scmi14~0 scmi17~0 scmi22~0
scoi2 scoi6 scoi20 scoi24 (*1);
po BY scsk5* scsk12 scsk19 scsk23 scsk26
scch3 scch7 scch10 scch15
scmi9 scmi14 scmi17 scmi22;
ne BY scsj1* scsj8 scsj11 scsj16 scsj21
scis4 scis13 scis18 scis25
scoi2 scoi6 scoi20 scoi24;
po@1; ne@1;
po WITH sk-oi@0;
ne WITH sk-oi@0;
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
74
Supplementary Material References
Arimitsu, K. (2014). Development and validation of the Japanese version of the Self-Compassion
Scale. The Japanese Journal of Psychology, 85 (1), 50–59.
Arimitsu, K., Aoki, Y., Furukita, M., Tada, A., & Togashi, R. (2016). Construction and
Validation of a Short Form of the Japanese version of the Self-Compassion Scale.
Komazawa Annual Reports of Psychology, 18, 1-8.
Azizi, A., Mohammadkhani, P., Lotfi, S., & Bahramkhani, M. (2013). The Validity and
Reliability of the Iranian Version of the Self-Compassion Scale. Iranian Journal of
Clinical Psychology, 2(3), 17-23.
Benda, J., & Reichová, A. (2016). [Psychometric characteritics of the Czech version of the Self-
Compassion Scale] Psychometrické charakteristiky české verze Self-Compassion Scale
(SCS-CZ). Československá psychologie. 60(2), 20-36.
Bento, E., Xavier, S., Azevedo, J., Marques, M., Freitas, V., Soares, M. J., et al. (2016).
Validation of the self-compassion scale in a community sample of Portuguese pregnant
women. European Psychiatry, 33, S238. doi: 10.1016/j.eurpsy.2016.01.598
Brenner, R. E., Heath, P. J., Vogel, D. L., & Credé, M. (2017). Two is more valid than one:
Examining the factor structure of the Self-Compassion Scale (SCS). Journal of counseling
psychology, 64(6), 696-707.
Castilho, P., & Pinto-Gouveia, J. (2011). Self-Compassion: Validation of the Portuguese version
of the Self-Compassion Scale and its relation with early negative experiences, social
comparison and psychopathology. Psychologica, 54, 203-231.
Castilho, P., Pinto-Gouveia, J., & Duarte, J. (2015). Evaluating the multifactor structure of the
long and short versions of the self- compassion scale in a clinical sample. Journal of
Clinical Psychology, 71(9), 856–870.
Chang, E. C., Yu, T., Jilani, Z., Fowler, E. E., Yu, E. A., Lin, J., & Hirsch, J. K. (2015). Under
Assault: Understanding The Impact Of Sexual Assault On The Relation Between Hope
And Suicidal Risk In College Students. Journal of Social and Clinical Psychology, 34(3),
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
75
221-238.
Chen, J., Yan, L., & Zhou, L. (2011). Reliability and validity of Chinese version of Self-
Compassion Scale. Chinese Journal of Clinical Psychology, 19(6), 734-736.
Cleare, S., Gumley, A., Cleare, C. J., & O’Connor, R. C. (2018). An Investigation of the Factor
Structure of the Self-Compassion Scale. Mindfulness, 9(2), 618-628.
Coroiu, A., Kwakkenbos, L., Moran, C., Thombs, B., Albani, C., Bourkas, S., ... & Körner, A.
(2018). Structural validation of the Self-Compassion Scale with a German general
population sample. PloS one, 13(2), e0190771.
Costa, J., Marôco, J., Pinto-Gouveia, J., Ferreira, C., & Castilho, P. (2015). Validation of the
psychometric properties of the self-compassion scale. Testing the factorial validity and
factorial invariance of the measure among borderline personality disorder, anxiety
disorder, eating disorder and general populations. Clinical Psychology & Psychotherapy.
doi: 10.1002/cpp.1974
Cunha, M., Xavier, A., & Castilho, P. (2016). Understanding self-compassion in adolescents:
Validation study of the Self-Compassion Scale. Personality and Individual Differences,
93, 56-62.
de Souza, L. K., & Hutz, C. S. (2016). Adaptation of the Self-Compassion Scale for use in Brazil:
Evidences of construct validity. Trends in Psychology, 24(1), 159-172.
Deniz, M., Kesici, Ş., & Sümer, A. S. (2008). The validity and reliability of the Turkish version
of the Self-Compassion Scale. Social Behavior and Personality, 36(9), 1151-1160.
Dundas, I., Svendsen, J. L., Wiker, A. S., Granli, K. V., & Schanche, E. (2016). Self-compassion
and depressive symptoms in a Norwegian student sample. Nordic Psychology, 68, 58–72.
Garcia-Campayo, J., Navarro-Gil, M., Andrés, E., Montero-Marin, J., López-Artal, L., &
Demarzo, M. M. (2014). Validation of the Spanish versions of the long (26 items) and
short (12 items) forms of the Self-Compassion Scale (SCS). Health and quality of life
outcomes, 12(1), 4.
Ghorbani, N., Chen, Z., Saeedi, Z., Behjati, Z.,Watson,P.J. (2013). Sakhtare Ameli Meghyase
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
76
Shafeghat e Khod dar Iran (Factorial Structure of Self-Compassion Scale in
Iran) Pazhohesh Haye Karbordi Ravanshenakhti, 4(3), 29-41. (In Persian)
Halamová, J., Kanovský, M., & Pacúchová, M. (2018). Self-Compassion Scale: IRT
Psychometric Analysis, Validation, and Factor Structure–Slovak Translation.
Psychologica Belgica, 57(4), 190-209.
Hupfeld, J., & Ruffieux, N. (2011). Validation of a German version of the self-compassion scale
(SCS-D). Zeitschrift für Klinische Psychologie und Psychotherapie, 40(2), 115–123.
Karakasidou, E., Pezirkianidis, C., Galanakis, M., & Stalikas, A. (2017). Validity, Reliability and
Factorial Structure of the Self Compassion Scale in the Greek Population. Journal of
Psychology and Psychotherapy, 7, 313.
Kim, K. E., Yi, G. D., Cho, Y. R., Chai, S. H., & Lee, W. K. (2008). The validation study of the
Korean version of the self-compassion scale. Korean Journal of Health Psychology, 13,
1023–1044.
Kotsou, I., & Leys, C. (2016). Self-Compassion Scale (SCS): Psychometric Properties of The
French Translation and Its Relations with Psychological Well-Being, Affect and
Depression. PloS One, 11(4), e0152880
Lee, W. K., & Lee, K. (2010). The validation study of the Korean version of the Self-Compassion
Scale with adult women in the community. Journal of the Korean Neuropsychiatric
Association, 49, 193–200.
López, A., Sanderman, R., Smink, A., Zhang, Y., van Sonderen, E., Ranchor, A., et al. (2015). A
reconsideration of the Self-Compassion Scale’s total score: self-compassion versus self-
criticism. PloS One, 10(7): e0132940. doi: 10.1371/journal.pone.0132940
Mantzios, M., Wilson, J. C. & Giannou, K. (2015). Psychometric properties of the Greek versions
of the self-compassion and mindful attention and awareness scales, Mindfulness, 6, 123-
132.
Montero-Marin, J., Zubiaga, F., Cereceda, M., Demarzo, M. M. P., Trenc, P., & Garcia-
Campayo, J. (2016). Burnout subtypes and absence of self-compassion in primary
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
77
healthcare professionals: A cross-sectional study. PLoS One, 11(6), e0157499
Neff, K. D. (2003). The development and validation of a scale to measure self-compassion. Self
and Identity, 2(3), 223-250.
Neff, K. D., Pisitsungkagarn, K., & Hsieh, Y. P. (2008). Self-compassion and self-construal in
the United States, Thailand, and Taiwan. Journal of Cross-Cultural Psychology, 39(3),
267-285.
Neff, K., Tóth-Király, I., & Colosimo, K. (in press). Self-compassion is best measured as a global
construct and is overlapping with but distinct from neuroticism: A response to
Pfattheicher, Geiger, Hartung, Weiss, and Schindler (2017). European Journal of
Psychology. Early view doi: 10.1002/per.2148
Neff, K. D., Whittaker, T. & Karl, A. (2017). Evaluating the factor structure of the Self-
Compassion Scale in four distinct populations: Is the use of a total self-compassion score
justified? Journal of Personality Assessment, 1, 1-12.
Petrocchi, N., Ottaviani, C., & Couyoumdjian, A. (2014). Dimensionality of self-compassion:
translation and construct validation of the self- compassion scale in an Italian sample.
Journal of Mental Health, 23(2), 72–77.
Pfattheicher, S., Geiger, M., Hartung, J., Weiss, S., & Schindler, S. (2017). Old Wine in New
Bottles? The Case of Self-compassion and Neuroticism. European Journal of Personality,
31, 160-169.
Sirois, F. M., Kitner, R., & Hirsch, J. K. (2015). Self-compassion, affect, and health-promoting
behaviors. Health Psychology, 34, 661-669.
Souza, L. K. de, Ávila-Souza, J., & Gauer, G. (2016). Escala de Autocompaixão. In: C. S. Hutz
(Ed.), Avaliação em Psicologia Positiva (pp. 169-177). São Paulo: Hogrefe/CETEPP.
Souza, L. K. D., & Hutz, C. S. (2016). Adaptation of the self-compassion scale for use in Brazil:
evidences of construct validity. Trends in Psychology, 24(1), 159-172.
Tóth-Király, I., Bőthe, B., & Orosz, G. (2017). Exploratory Structural Equation Modeling
Analysis of the Self-Compassion Scale. Mindfulness, 8, 881-892.
SCS FACTOR STRUCTURE IN 20 DIVERSE SAMPLES
78
Veneziani, C. A., Fuochi, G., & Voci, A. (2017). Self-compassion as a healthy attitude toward the
self: Factorial and construct validity in an Italian sample. Personality and Individual
Differences, 119, 60-68.
Williams, M. J., Dalgleish, T., Karl, A., & Kuyken, W. (2014). Examining the factor structures of
the Five Facet Mindfulness Questionnaire and the Self-Compassion Scale. Psychological
Assessment, 26(2), 407-418.
Woo Kyeong, L. (2013). Self-compassion as a moderator of the relationship between academic
burn-out and psychological health in Korean cyber university students. Personality and
Individual Differences, 54(8), 899-902.
Yarnell, L. M., Neff, K. D. (2013). Self-compassion, interpersonal conflict resolutions, and well-
being. Self and Identity. 2:2, 146-159.
Zeng, X., Wei, J., Oei, T. P., & Liu, X. (2016). The Self-Compassion Scale is Not Validated in a
Buddhist Sample. Journal of Religion and Health, 55(1996), 1-14.
