ArticlePDF Available

Abstract and Figures

The purpose of this study was to examine the measurement invariance of the Self-Compassion Scale by IRT differential test functioning in ten distinct populations (n = 13623 participants) from ten different countries: Australia (n = 517), China (n = 321), Czech Republic (n = 5081), Germany (n = 2510), Italy (n = 384), Portugal (n = 512), Slovakia (n = 1181), South Korea (n = 1813), Turkey (n = 471), and USA (n = 833). We assessed differential test functioning for the two SCS subscales, Self-compassionate responding and Self-uncompassionate responding separately, because previous bifactor and two-tier analyses of the scale showed the best fit with two separate general factors, and not for the overall score. Only 13 of the 45 comparisons for Self-compassionate responding and 13 of the 45 comparisons for Self-uncompassionate responding (analyses of every pair) demonstrated measurement invariance (no differential test functioning). Generally, our results revealed that the two subscales of Self-compassionate responding and Self-uncompassionate responding were not equivalent among all countries and groups. Therefore, it is impossible to compare overall scores across all countries. Two subscales of the Self-Compassion Scale (Self-compassionate responding and Self-uncompassionate responding) are valid and reliable instruments with substantial potential of use cross-culturally, but results reveal significant cross-cultural differences in the way these two constructs are measured by the subscales of the SCS. Future analyses of the meanings and connotations of this construct across the world are necessary to develop a scale which allows cross-cultural comparisons of various treatment outcomes related to self-compassion.
Content may be subject to copyright.
Mediterranean Journal
of Clinical Psychology
ISSN 2282-1619
1
Volume 9, n 1, 2021
Health Psychology
The multiple group IRT measurement invariance analysis of the
Self-Compassion Scale in ten international samples
Martin Kanovský 1, Júlia Halamová 2 *, Nicola Petrocchi 3, Helena Moreira 4,
Eunjoo Yang 5, Jan Benda 6, Michael Barnett 7, Elmar Brähler 8, Xianlong Zeng 9,
Markus Zenger 10
Abstract
The purpose of this study was to examine the measurement invariance of the Self-Compassion Scale
by IRT differential test functioning in ten distinct populations (n = 13623 participants) from ten
different countries: Australia (n = 517), China (n = 321), Czech Republic (n = 5081), Germany (n =
2510), Italy (n = 384), Portugal (n = 512), Slovakia (n = 1181), South Korea (n = 1813), Turkey (n =
471), and USA (n = 833). We assessed differential test functioning for the two SCS subscales, Self-
compassionate responding and Self-uncompassionate responding separately, because previous
bifactor and two-tier analyses of the scale showed the best fit with two separate general factors, and
not for the overall score. Only 13 of the 45 comparisons for Self-compassionate responding and 13 of
the 45 comparisons for Self-uncompassionate responding (analyses of every pair) demonstrated
measurement invariance (no differential test functioning). Generally, our results revealed that the two
subscales of Self-compassionate responding and Self-uncompassionate responding were not
equivalent among all countries and groups. Therefore, it is impossible to compare overall scores across
all countries. Two subscales of the Self-Compassion Scale (Self-compassionate responding and Self-
uncompassionate responding) are valid and reliable instruments with substantial potential of use cross-
culturally, but results reveal significant cross-cultural differences in the way these two constructs are
measured by the subscales of the SCS. Future analyses of the meanings and connotations of this
construct across the world are necessary to develop a scale which allows cross-cultural comparisons
of various treatment outcomes related to self-compassion.
1 Institute of Social Anthropology, Faculty of Social and Economic Sciences, Comenius University in
Bratislava, Bratislava, Slovakia
2 Institute of Applied Psychology, Faculty of Social and Economic Sciences, Comenius University in
Bratislava, Mlynské luhy 4, 821 05 Bratislava, Slovakia
3 Department of Economics and Social Sciences, John Cabot University, Rome, Italy
4 Cognitive and Behavioural Centre for Research and Intervention, University of Coimbra, Coimbra,
Portugal
5 Department of Psychology, Korea University, Seoul, South Korea
6 Department of Psychology, Faculty of Arts, Charles University, Prague, Czech Republic
7 Department of Psychology and Counseling, University of Texas at Tyler, USA
8 Department of Medical Psychology and Medical Sociology, University of Leipzig, Leipzig, Germany
and Department of Psychosomatic Medicine and Psychotherapy, University of Mainz, Mainz,
Germany
9 Faculty of Psychology, Beijing Normal University, Beijing, China
10 Faculty of Applied Human Studies, University of Applied Sciences Magdeburg and Stendal,
Germany and Integrated Research and Treatment Center AdiposityDiseases - Behavioral Medicine,
Psychosomatic Medicine and Psychotherapy, University of Leipzig Medical Center, Germany
E-mail corresponding author: julia.halamova@gmail.com
MJCP|9, 1, 2021 Kanovský et al.
2
Keywords:
Self-compassion; Measurement invariance; Differential test functioning; Cross-cultural
studies.
Received: 12 October 2020
Accepted: 6 March 2021
Published: 29 April 2021
Citation: Kanovský, M., Halamová, J., Petrocchi, N., Moreira, E., Yang, E.,
Benda, J., Barnett, M., Brähler, E., Zeng, X., Zenger, M. (2021). The multiple
group IRT measurement invariance analysis of the Self-Compassion Scale in ten
international samples. Mediterranean Journal of Clinical Psychology, 9(1).
https://doi.org/10.6092/2282-1619/mjcp-2682
1. Introduction
Inner speech of patients is known to have huge impact on their responsiveness (Shahar et al.,
2015) to medical as well as psychological treatment. Inner speech can be in the form of
uncompassionate or even harsh self-critical inner speech, which is one of the main risk factors
for various types of psychopathology. For example, self-critical inner speech is associated with
social anxiety (Shahar et al., 2012), depression (Greenberg & Watson, 2005), posttraumatic stress
disorder, self-harming behavior, suicidal tendencies (O’Connor & Noyce, 2008), bipolar
affective disorder, schizophrenia, eating disorders, and borderline personality disorder (Meares
et al., 2011). Inner speech in the form of a self-compassionate voice can function as an antidote
to the self-critical voice. According to Neff and Dahm (2014, p. 121), “Self-compassion is simply
compassion directed inward, relating to ourselves as the object of care and concern when faced
with the experience of suffering.” Multiple intervention research studies (Conversano et al.,
2020; Falconer et al., 2014; Neff & Germer, 2013) have shown that learning to have a self-
compassionate inner voice is not only possible but also is an important resilience factor which
improves responsiveness to different kinds of treatment (Merlo et al., 2020; Terry & Leary,
2011; Warren et al., 2016) and it is related to many behavioural factors of happy and healthy life
too. Therefore, it is crucial to provide a reliable, valid and invariant instrument for measuring
levels of self-compassion which can enable comparisons in health treatment across and within
different countries.
Neff (2003) developed a scale to measure self-compassion, the Self-Compassion Scale SCS,
which is the most frequently used scale for measuring self-compassion in both research and
clinical practice. Despite a wide range of translations of the SCS and its increasing popularity
among researchers and clinicians, to the best of our knowledge, there is no evidence that the
SCS is invariant across different countries and language translations. There is currently only one
invariance study with the SCS, in which Montero-Marín et al. (2016) compared Brazilian and
Spanish samples in the dimension of SCS, which consists of negatively-worded items and did
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
3
not find strong construct invariance between them. However, the original long format, English-
language version of the SCS has now been translated into eighteen different languages. These
are Chinese (Chen et al., 2011), Czech (Benda & Reichová, 2016), Dutch (López et al., 2015),
French (Kotsou & Leys, 2016), German (Hupfield & Ruffieux, 2011), Greek (Mantzios et al.,
2013), Hungarian (Tóth-Király et al., 2016), Iranian (Azizi et al., 2013), Italian (Petrocchi et al.,
2014), Japanese (Arimitsu, 2014), Korean (Lee & Lee, 2010), Norwegian (Dundas et al., 2016),
Portuguese (Castilho & Pinto-Gouveia, 2011), Slovak (Halamová et al., 2017), Spanish (Garcia-
Campayo et al., 2014), Thai (Pisitsungkagarn et al., 2014), and Turkish language (Deniz et al.,
2008). In addition, Neff (2016) suggests that research findings obtained with translations should
not be automatically generalized to the original version of the SCS, because of potential
difficulties with the quality of translations and/or potential cultural factors (Behling & Law
2000).
In a previous study of the factor structure of the SCS, in different countries and samples,
Halamová et al. (2021) found that the most reliable and appropriate use of the SCS is not
computing the overall score, but rather calculating two dimensional scores for Self-
compassionate responding and Self-uncompassionate responding separately. We note that this
conclusion was not based on testing simple two-factor model, but rather more general two-tier
model with two general factors and six specific factors. Considering that the SCS (Neff, 2003)
is the most frequently used scale of self-compassion across cultures, it is crucial to inspect
whether the use of the SCS in comparisons between countries and languages is justified.
Therefore, it is vital to scrutinize the psychometric properties and measurement invariance of
the SCS in various languages in addition to English (Neff et al., 2017).
1.1 The Self-compassion scale and culture
Research is scarce when it comes to exploring whether self-compassion levels differ across
different cultures, and only one study has tested measurement invariance of the SCS between
its different translations Montero-Marín et al. (2016). Other studies have compared the levels of
self-compassion between different cultures and societies using the SCS, but they did not
previously test whether the measure was invariant in the different contexts (e.g., Birkett, 2014;
Khramtsova & Chuykova, 2016; Neff et al., 2008), producing various findings which we discuss
below. Nevertheless, all these findings must be considered with caution since measurement
invariance, a prerequisite for meaningful differences, was not previously tested in these studies.
SCS scores have been shown to be significantly and positively associated with well-being in all
cultures, suggesting that people may benefit from self-compassion despite cultural differences
(Neff et al., 2008). However, Neff and Vonk (2009) proposed that various cultures provide
different messages regarding the meaning and value of self-compassion or its counterpart, and
that individual variation in SCS scores within a particular culture may be partially dependent on
MJCP|9, 1, 2021 Kanovský et al.
4
personal acceptance or rejection of these cultural messages. Clearly, there are unique aspects of
every society that must be considered if we want to explore the impact of culture on levels of
Self-compassionate responding and Self-uncompassionate responding. For example,
differences in political and historical backgrounds, religious beliefs and cultural worldviews, or
variation in parenting practices may have an impact. Individuals may develop their way of
relating towards one’s self on the basis of their interactions with other people (mainly
caregivers), so that experiences with them are internalized and expressed either as a self-
compassionate or a self-critical (self-uncompassionate) inner voice (Gilbert & Irons, 2008).
Indeed, results of Neff et al. (2009) indicated that three cultures (USA, Thai, and Taiwanese)
differed significantly in SCS scores from one another, although differences within cultures in
SCS scores were as large as differences between cultures. Furthermore, a cross-cultural study
examined differences in self-compassion between Chinese and American undergraduates and
found a non-significant difference in the overall Self-Compassion score (Birkett, 2014).
However, Chinese students reported significantly higher levels of both positive and negative
aspects of self-compassion than American students, the authors interpreted this as Chinese
students experiencing aspects of self-compassion in both the positive and negative ways.
Additionally, SCS scores from both Chinese and American students in Birkett’s study are similar
to scores of American undergraduates in a cross-cultural study including Thai and Taiwanese
students conducted in previous research (Neff et al., 2008). According to Khramtsova and
Chuykova (2016), American participants were higher on self-compassion than Russian
participants measured by the shorter version of SCS. Kwan et al. (2009) investigated cultural
differences while exploring sources of self-esteem. They suggested that self-compassion was
one of three major sources of self-esteem together with self-efficacy and narcissism. In their
study, they found that the Chinese participants had significantly higher levels of self-compassion
and narcissism than the American participants. However, narcissism was significantly correlated
with levels of self-compassion among the Chinese participants but not the American
participants.
It appears that the types of constructs that correlate with self-compassion vary across cultures.
With regards to semantics, Kitayama and Karasawa (1997) found that in the East Asian culture
of Japan, individuals tend to have positive feelings towards themselves while still being self-
critical. Heine et al. (1999) argue that self-criticism is not a psychological problem for individuals
living in interdependent and collectivist cultures. This suggests that although some people may
display higher levels of self-uncompassionate responding than other people, they may not
necessarily display lower levels of self-compassionate responding. Additionally, Zeng et al.
(2016) reported that SCS cannot be validated among Chinese Buddhists, as Self-kindness and
Common Humanity were neither correlated with their opposite dimensions nor linked to better
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
5
emotional outcomes. They further illustrated that the ideas of Self-kindness and Common
Humanity reflected in SCS were different from Buddhist philosophy and culture, although the
concept of self-compassion originated from Buddhism. In all, empirical studies showed many
cultural issues may impact the understanding and function of self-compassion.
However, there are only a few and very recent extant studies dealing with measurement
invariance of the SCS (Costa et al., 2016; Cunha et al., 2016; Montero-Marín et al., 2016), and
only one of them deals with equivalence across various cultures. According to Costa et al. (2016)
a weak measurement and structural configural invariance of the two-factor model of the SCS
across clinical samples (diagnosed with borderline personality disorder, anxiety disorder, eating
disorder) and general populations showed that both properties and interpretations of scores of
SCS were equivalent. Results of Cunha et al. (2016) confirmed the measurement invariance
across genders for Portuguese adolescents. Montero-Marín and colleagues (2016) found that
configural invariance (i.e., common factors are associated with the same items across groups)
and partial metric invariance (i.e., weak factorial - common factors have the same meanings
across groups) were achieved but not scalar invariance (i.e., strong factorial comparisons of
latent groups means are meaningful). When comparing the negative dimension of the SCS
between Brazilian and Spanish samples, the authors found that the first-order factor loadings
were not the same in the two groups. Although both Brazilian and Spanish samples reported
equivalence in their understanding of the positive factor of the SCS (Self-compassionate
responding), the equivalence was not detected in the negative part of the SCS, and therefore the
authors could not compare the mean levels of the latent variables between the groups. This
means that, generally, the scale could not be considered invariant.
In summary, studies using the SCS around the world indicate that self-related processes captured
by the SCS appear to be useful and meaningful constructs across cultures (Neff & Vonk, 2009).
However, previous research on the links between SCS and culture suggest potential cultural
differences in levels of self-compassion (Self-compassionate responding and Self-
uncompassionate responding). To date, very few differences in SCS scores have been found
according to cultural background, which appears to play an important role in levels of self-
compassion. Certainly, more research on cultural differences in self-compassion and self-
criticism is necessary (Voruz, 2013). To develop this area of research, more cross-cultural
comparisons on measurement invariance of tools assessing self-compassion is needed, which is
the aim of the present study.
1.2 Aim of the Study
Stemming from the existing literature on the SCS, the aim of this study was to investigate
measurement invariance by IRT differential test functioning of the ten different samples and
nine language versions of the two dimensions of the SCS, Self-compassionate responding and
MJCP|9, 1, 2021 Kanovský et al.
6
Self-uncompassionate responding (Halamová et al., 2021), with the main goal of determining
whether comparisons between total scores of the two dimensions of SCS across countries and
languages are justified.
2. Research Methods
2.1 Measuring Instrument
The Self-Compassion Scale (Neff, 2003) measures six aspects which, according to the author of
the scale, constitute a self-compassionate response to difficult circumstances. The scale includes
26 items rated on a 5-point Likert-type format of frequency (1 = almost never; 5 = almost
always). The subscale Self-Kindness represents the ability of taking care and being warm towards
oneself when encountering difficult situations. Self-Judgment reflects how critical the
individuals behave towards themselves. Common Humanity reflects the personal understanding
that suffering is part of the shared human experience. Isolation relates to the sense of loneliness
and isolation after failure. Mindfulness is a non-judgmental state of mind in which individuals
observe their thoughts and feelings as they are, without trying to suppress or deny them. Over-
identification captures responses to challenging situations that involve becoming absorbed with
negative thoughts and feelings. The scale as a whole measures the degree to which individuals
display self-kindness against self-judgment, common humanity versus isolation, and
mindfulness versus over-identification. Neff (2016) perceives these components as distinct and
at the same time influencing each other (see Table 1 for items description). Negative and positive
items are represented in roughly equal proportions. Self-Judgment, Isolation, and Over-
identification are reverse coded. After reverse-coding of negative items, means are calculated
for each subscale, and a grand mean is calculated that is considered to represent a global measure
(Neff, 2016) of self-compassion: higher overall scores indicate greater levels of self-compassion
(Neff, 2003). The reliability and validity of the overall SCS scale and of its six subscales appear
to be quite good in various translations, as well as in different corresponding countries
(Halamová et al., 2021). However, recent cross-cultural findings on the factorial structure of the
SCS have failed to replicate the higher order factor of self-compassion originally proposed by
Neff, suggesting that a total score of the scale may not be appropriate to use as an overall
measure of self-compassion across different translations and cultures (Halamová et al., 2021).
Findings by Halamová and colleagues (2021) justify the use of two separate dimensions of the
SCS, namely Self-compassionate responding and Self-uncompassionate responding. Thus, in
the present paper, the Self-compassionate responding subscale and the Self-uncompassionate
responding subscale will be analysed separately. From now on, in order to avoid potential
confusion with results pertaining the overall Self-compassion score, we will refer to the two
dimensions of the SCS as Self-compassionate responding subscale and Self-uncompassionate
responding subscale.
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
7
Table 1. Dimensions and scale items of The Self-compassion scale
Dimensions
Self-
compassionate
responding
Self-Kindness
5.
12.
19.
23.
26.
Mindfulness
9.
14.
17.
22.
Common
Humanity
3.
7.
10.
15.
Self-
uncompassionate
responding
Self-Judgment
1.
8.
11.
16.
21.
Over-
identification
2.
6.
20.
24.
Isolation
4.
13.
18.
25.
MJCP|9, 1, 2021 Kanovský et al.
8
2.2 Sampling procedure
As we used the same research samples as the study of Halamová et al. (2021) the sampling
procedure was also the same. The first author of this study wrote around 130 emails inviting
cooperation to all authors of already published research studies as found at www-self-
compassion.com and from Google scholar using the following search terms, “self-compassion
scale” and “neff”. Inclusion criteria were at least 300 adult participants in a sample (to allow
convergence of IRT models) and using 26 items of SCS and a non-clinical population. We sent
out 130 number of emails and 14 number of people both agreed to and used the data. After
preliminary analysis, we excluded three of four Turkish data sets because the statistical models
failed to converge. The Netherland sample was also excluded due to use of a 24-item scale while
the rest of the samples used the 26-item scale. After all sampling procedures and established
collaborations, we included 10 data sets for further analyses.
2.3 The research samples and procedures from different countries
Out of currently eighteen different language versions of SCS we obtained data from nine of
them. Our research samples consist of two distinct English language samples from different
countries (Australia (n = 517) and USA (n = 833) and eight different language translations
samples eight different countries namely, China (n = 321), Czech Republic (n = 5081), Germany
(n = 2510), Italy (n = 384), Portugal (n = 512), Slovakia (n = 1181), South Korea (n = 1813),
and Turkey (n = 471). See Table 2 for more information about the samples. In total, we tested
10 distinct samples including 13623 participants. The data collection from all these samples was
in accordance with the ethical standards of the institutional and/or national research committee
and with the 1964 Helsinki declaration and its later amendments or comparable ethical
standards.
Table 2. Sample information and internal consistency coefficients for SCS in ten different
samples and ten various countries
Country
N
Female
%
M Age
SD
Language
Cronbach α
Cronbach α
Correl.
Total
SUCR/SCR
SUCR/SCR
Australia
517
100
53.64
5.49
English
.94
.92/.90
.59***
Czech
Republic
5081
71.80
28.20
11.10
Czech
.91
.86/.87
.64***
Germany
2510
53.70
50.19
17.36
German
.88
.87/.88
.14***
China
321
53.28
25.45
4.10
Chinese
.90
.89/.88
.35***
Italy
384
67.40
33.38
10.55
Italian
.93
.91/.89
.54***
Portugal
512
66.20
42.48
6.11
Portuguese
.87
.87/.85
.26***
Slovakia
1181
66.00
30.30
12.40
Slovak
.86
.86/.84
.27***
South
Korea
1813
48.80
39.28
11.27
Korean
.91
.93/.91
.23***
Turkey
471
83.40
20.65
1.35
Turkish
.88
.78/.79
.78***
USA
833
71.20
21.00
3.86
English
.89
.93/.89
.17***
Note. M = Mean. SD = standard deviation. SUCR = Self-uncompassionate responding. SCR = Self-
compassionate responding.
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
9
Australia
The participants from Australia were originally enrolled from the electoral roll (Murray et al.,
2004) and then selected for the further study (Brown et al., 2014).
China
An online sample service enrolled participants from China (Zeng et al., 2016).
Czech Republic
The participants from the Czech Republic (Benda & Reichová, 2016) were recruited through
Facebook and respondents under 18 were excluded.
Germany
The participants from Germany are representative for the German general population and they
were recruited by the random-route-technique (Körner et al., 2015).
Italy
An online survey company recruited the participants from Italy (Petrocchi et al., 2014).
Portugal
The Portuguese sample was collected in public schools and in the general community (Moreira
et al., 2015, 2016).
Slovakia
Slovak participants were recruited mainly online by convenience sampling (Halamová et al.,
2017).
South Korea
The participants from South Korea were recruited by an online survey company (Hwang et al.,
2016).
Turkey
Turkish participants were undergraduate students enrolled at a university (Şenyuva et al., 2014).
USA
Participants were recruited through the university research website (Barnett & Sharp, 2016).
2.4 Data Analysis
In testing measurement invariance/equivalence, confirmatory factor analysis (CFA) is the
prevalent approach (Vanderberg & Lance, 2000). Despite the obvious advantages of IRT, these
models are not used frequently. In psychometric literature, there is an ongoing debate among
researchers comparing these two approaches (Kankaraš et al., 2011; Kim & Yoon, 2011; Meade
& Lautenschlager, 2004; Raju et al., 2002; Reise et al., 1993). While CFA models assume that
MJCP|9, 1, 2021 Kanovský et al.
10
the item responses are continuous and linear, IRT models assume the item responses are either
nominal or ordinal. Unlike CFA models, IRT models are inherently non-linear. Furthermore,
CFA models typically estimate a single intercept per item because they use a continuous
assumption. On the contrary, IRT models typically compute multiple parameters (thresholds)
analogous to item intercepts per item IRT models the polytomous data as categorical, and as
a consequence IRT models usually result in greater sensitivity to more-nuanced group
differences such as in central tendency or the use of extreme scores. Recent research shows that
IRT models can quite accurately detect inequivalence in the intercept and slope parameters both
at the scale and the item level (Kankaraš et al., 2011). However, CFA performs well only when
inequivalence is located in the slope parameters, but wrongfully indicates inequivalence in the
slope parameters when inequivalence is located in the intercept parameters (Kankaraš et al.,
2011).
IRT non-linear factor analyses were used for models, but we acknowledge that non-linear
confirmatory factor analysis is available (via WLSMV estimator in Mplus software and R
package lavaan). Difference between non-linear IRT and non-linear CFA is that IRT method is
full-information and CFA non-linear method is not (DeMars, 2012). The main advantage of the
full information IRT approach is that item modelling relationships are far more general in that
non-monotonic, non-parametric, or just plain customized item response models are easier to
cope with because we can avoid worries about the sufficiency of using the polychoric correlation
matrix used in the CFA non-linear approach (WLSMV). However, we should note that
performing robust Maximum Likelihood estimation (MLR) (e.g., in Mplus) based on categorical
variables with the integration method in order to get IRT-parameters through CFA could avoid
the limitations of polychoric correlations.
SCS subscales (Self-compassionate responding and Self-uncompassionate responding) are
unidimensional, and one general latent factor explains a substantive part of the variance as it
was shown in previous research (Halamová et al., 2021) by means of bifactor models and non-
parametric IRT Mokken scale analysis. However, we performed our analyses for each
population separately. Our previous results provided no information about whether the test
score was comparable across different populations. IRT models are better suited than CFA
models to explore this issue. The CFA measurement invariance analyses provide insights
regarding the relationship between latent factors, so their use is preferable when the goal is to
answer questions of the invariance of a multifactorial framework and many other aspects, such
as configural invariance, metric invariance, scalar invariance, invariant uniquenesses, invariant
factor variances, invariant factor covariance, equal factor means, etc. (Vandenberg & Lance,
2000). Since we decided to test the invariance of single, unidimensional scales (Self-
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
11
compassionate responding and Self-uncompassionate responding, respectively), IRT analyses
are more desirable.
However, the results from previous research of Halamová et al. (2021) recommending two
separate scores do not argue in favour of the simple two-factor model they conclude that the
two-tier model has better fit than the bifactor model. The two-tier model of SCS scale has two
general factors and six specific factors, and the bifactor model of SCS scale has only single
general factor and six specific factors. The conclusion of the work of Halamová et al. (2021) is
that the use of the single total (sum) score is not justified because the single general factor is not
essentially unidimensional therefore we should use two separate sum scores which sufficiently
capture the variance. This conclusion does not entail that the two-factor model is feasible. We
cannot use the two-factor model to test measurement invariance in the standard confirmatory
way: the fit of such model with data is far below the acceptable values. To repeat, the argument
in the previous manuscript (Halamová et al., 2021) was that only two separate scales of SCS are
unidimensional enough to capture sufficient amount of variance and present analysis is based
on this conclusion assuming the essential unidimensionality of separate scales and using the IRT
DTF procedure to detect differential test functioning. Secondly, we are unsure how to settle the
issue of whether these two general factors are response style / method factors or really different
constructs: some authors (e.g., Montero-Marín et al., 2016) argue that self-criticism (or self-
uncompassionate responding) is in fact a different construct than self-compassion, but we
believe that the possibility of response style / method factors cannot be ruled out. However,
even if two factors are response style / method factors due to the reverse scoring of their items,
this is a further argument in favour of testing them separately. Thirdly, fitting the nested
(second-order / hierarchical) factor models does not allow for the calculation of some indices,
such as Omega hierarchical or ECV (see Mansolf & Reise, 2017 for theoretical justification and
Cucina & Byle, 2017 for meta-analysis).
In the context of IRT models, measurement equivalence is tested by inspecting the differential
item functioning (DIF), and/or the differential test functioning (DTF). Differential item
functioning (DIF) means that an item within the SCS questionnaire measures the construct (self-
compassion) differently for one population than for another. As a consequence, the presence
of DIF compromises test validity. If this item bias accumulates to the extent to produce biased
overall test scores, a test displays differential test functioning (DTF). DTF is present when
respondents who have the same standing on the latent ability, but belong to different groups,
obtain different scores on the test.
DIF is routinely tested and when it is found, some method of purification is usually adopted:
items with DIF are flagged and removed. However, if a test has many items (e.g., SCS has 26
items) and only some of them have DIF, then the impact of these DIFs on the overall test score
MJCP|9, 1, 2021 Kanovský et al.
12
may be negligible (DeMars, 2011). We can consider yet another situation: there could be large
DIF effects in favour of one population for some items, but these effects are simultaneously
cancelled out by DIF for other items in favour of other populations. Therefore, the presence of
DIF for some items does not necessarily imply that the overall test itself is biased. On the other
hand, it is also possible to have DTF in a situation where little or no DIF has been detected.
Nontrivial DTF can occur in cases when the parameters systematically favour one group over
another. Consequently, the aggregate of these small, insignificant differences at the item level
can become substantial at the test level (Chalmers et al., 2016). DTF has greater practical sense
for our purpose than DIF: we do not intend to inspect particular items of SCS subscales, nor
do we intend to improve/purify them. Our purpose is more practical: we intend to test the
assumption that the (expected) total score of the SCS subscales is equivalent across different
populations, therefore only the latent ability and not the belonging to a particular group has
any impact on the (expected) total score.
Testing the DTF involves two statistical measures (Chalmers et al., 2016). The first statistic (the
signed DTF) tests whether there is any systematic test scoring bias, indicating that some group
consistently scored higher across a specified range of latent ability. The second statistic (the
unsigned DTF) assesses whether the test curves have a large degree of overall separation on
average, suggesting that there may be substantial DTF at particular levels of latent ability. The
signed DTF values can range from - TS to TS (TS stands for the highest possible test score).
Negative values of the signed DTF indicate that the reference group scores systematically lower
than the focal group on average, while positive values indicate that the reference group scores
higher. The unsigned DTF ranges from 0 to TS because the area between the curves is zero
when the test scoring functions have exactly the same functional form. The signed DTF values
are always lower than or equal to the unsigned values: when the curves do not cross, the signed
DTF is equal to the unsigned DTF. If there is a small value for the signed DTF and a large value
for the unsigned DTF, test curves intersect at one or more locations to create a balanced overall
score, but there is substantial bias at particular levels of latent ability.
To test for differences among all countries, the multilevel IRT models are appropriate. We did
not use traditional and popular GLMM model (e.g., lmer function in package lme4; see De
Boeck et al., 2011), because it did not allow graded-response polytomous models such models
would require multivariate logits and estimation of different discrimination parameters. Instead,
we used the extended mixed-effects IRT model (Chalmers, 2015).
For all statistical analyses, the “mirt” package (Chalmers, 2012) in statistical software R (R Core
Team, 2017) was used. Our analysis proceeded as follows.
(1) We performed pairwise tests of all samples, separately for Self-compassionate responding
and Self-uncompassionate responding: the total number of tests was 2*((10*9)/2) = 90. We
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
13
reported only the signed DTF values because they are of more practical use than the unsigned
DTF values, as we explained above. The fitted IRT models were Samejima graded response
models, which are in fact the sequences of 2 PL models.
(2) For samples with insignificant signed DTF equivalence obtained, we also reported the latent
means difference (for Self-uncompassionate responding subscale and Self-compassionate
responding subscale). Latent means in the reference group were constrained to zero, and latent
means in the focal group were estimated.
(3) We performed the multilevel IRT analysis of all samples, separately for Self-compassionate
responding and Self-uncompassionate responding, to check for relative differences among
countries. We reported random effects.
3. Results
To justify the use of the two subscales, we report here the fit of two-tier models from previous
research (Halamová et al., 2021), see Table 3. Two-tier models included two general factors and
six specific factors, and previous research demonstrated that single general factor is not
essentially unidimensional to capture sufficient amount of common variance, and that two
general factors are necessary. In Table 4, we present the fit of multiple-group IRT models used
for the DTF analyses.
Table 3. Fit Indices of Two-Tier IRT Models of 26-Item SCS
IRT model
Sample
CFI
TLI
RMSEA
SRMR
AUS
.954
.941
.051
.060
CHI
.910
.887
.053
.098
CZK
.905
.880
.063
.057
GER
.953
.940
.053
.087
ITA
.907
.885
.062
.066
POR
.932
.913
.052
.108
SVK
.939
.922
.044
.090
KOR
.956
.944
.065
.103
TUR
NC
N/A
N/A
N/A
USA
.950
.936
.059
.092
Note. NC = model failed to converge. N/A = no information due to non-convergence of the model.
AUS = Australia (n = 517). CHI = China (n = 321). CZK = Czech Republic (n = 5081). GER =
Germany (n = 2510). ITA = Italy (n = 384). POR = Portugal (n = 512). SVK = Slovakia (n = 1181).
KOR = South Korea (n = 1813). TUR = Turkey (n = 471). USA (n = 833).
MJCP|9, 1, 2021 Kanovský et al.
14
Table 4. Fit Indices of multiple-group IRT Models
Self-uncompassionate
Self-compassionate
Samples
CFI
RMSEA
SRMSR 1
SRMSR 2
CFI
RMSEA
SRMSR 1
SRMSR 2
AUS/CZK
.911
.042
.036
.055
.968
.034
.049
.023
AUS/GER
.956
.025
.036
.066
.945
.028
.045
.021
AUS/CHI
.962
.033
.034
.031
.939
.051
.046
.081
AUS/ITA
.936
.043
.043
.028
.934
.055
.052
.077
AUS/KOR
.960
.055
.036
.066
.944
.041
.049
.060
AUS/POR
.917
.043
.036
.077
.923
.055
.044
.070
AUS/SVK
.934
.048
.036
.071
.958
.032
.048
.071
AUS/TUR
.967
.045
.036
.079
.935
.055
.042
.082
AUS/USA
.901
.052
.043
.057
.923
.046
.044
.079
CZK/GER
.938
.037
.037
.082
.917
.026
.036
.077
CZK/CHI
.900
.042
.037
.028
.903
.032
.036
.060
CZK/ITA
.901
.042
.036
.065
.905
.028
.036
.038
CZK/KOR
.903
.046
.037
.095
.904
.028
.036
.032
CZK/POR
.902
.042
.038
.065
.908
.028
.037
.063
CZK/SVK
.908
.044
.036
.061
.948
.023
.036
.079
CZK/TUR
.928
.041
.036
.077
.928
.030
.036
.066
CZK/USA
.908
.044
.038
.070
.909
.027
.037
.071
GER/CHI
.907
.045
.048
.068
.951
.031
.040
.054
GER/ITA
.907
.039
.047
.072
.951
.033
.040
.057
GER/KOR
.908
.048
.049
.078
.914
.034
.039
.071
GER/POR
.908
.047
.048
.077
.949
.034
.038
.077
GER/SVK
.907
.050
.050
.073
.969
.027
.038
.078
GER/TUR
.924
.046
.047
.078
.952
.041
.037
.075
GER/USA
.906
.052
.048
.074
.939
.035
.040
.082
CHI/ITA
.910
.034
.045
.082
.913
.056
.047
.079
CHI/KOR
.908
.056
.045
.077
.906
.041
.047
.077
CHI/POR
.916
.034
.046
.073
.904
.056
.047
.079
CHI/SVK
.909
.050
.045
.077
.957
.036
.051
.074
CHI/TUR
.945
.051
.045
.041
.926
.068
.048
.072
CHI/USA
.907
.048
.045
.079
.906
.059
.051
.076
ITA/KOR
.902
.057
.037
.071
.907
.036
.049
.075
ITA/POR
.909
.043
.036
.072
.903
.049
.048
.069
ITA/SVK
.910
.050
.040
.075
.964
.032
.051
.058
ITA/TUR
.962
.047
.034
.069
.937
.056
.050
.080
ITA/USA
.905
.045
.035
.077
.925
.044
.046
.076
KOR/POR
.913
.056
.035
.075
.924
.050
.040
.057
KOR/SVK
.917
.056
.036
.071
.910
.043
.040
.078
KOR/TUR
.925
.064
.034
.069
.908
.057
.040
.077
KOR/USA
.900
.052
.035
.079
.904
.060
.038
.078
POR/SVK
.925
.057
.055
.077
.974
.025
.060
.072
POR/TUR
.905
.068
.053
.082
.962
.042
.058
.079
POR/USA
.903
.051
.053
.080
.910
.051
.062
.081
SVK/TUR
.907
.058
.046
.071
.965
.032
.050
.071
SVK/USA
.905
.058
.045
.080
.921
.040
.049
.079
TUR/USA
.909
.060
.071
.083
.914
.057
.070
.080
Note. AUS = Australia (n = 517). CHI = China (n = 321). CZK = Czech Republic (n = 5081). GER =
Germany (n = 2510). ITA = Italy (n = 384). POR = Portugal (n = 512). SVK = Slovakia (n = 1181).
KOR = South Korea (n = 1813). TUR = Turkey (n = 471). USA (n = 833).
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
15
We obtained 13 measurement equivalencies out of 45 comparisons for the Self-
uncompassionate responding subscale (see Table 5) and 13 measurement equivalencies for the
Self-compassionate responding subscale (see Table 6). We have to note that we did not propose
any overall null-hypothesis (e.g., that all samples will be invariant across the world): p-values in
Tables 5 and 6 are based on multiple imputation estimate (1000 draws) of the expected test
scores for each sample and we had the null-hypothesis for each particular sample (i.e., that the
bias of total scores between two particular samples is zero).
For the Self-uncompassionate responding subscale, the Czech sample was equivalent to six of
the other samples; the Turkish and Chinese samples to four other samples; the Slovak, German,
Italian, Korean, and USA samples to two other samples; the Australian sample to one other
sample; and finally Portugal sample was equivalent to no other sample. As for the Self-
compassionate responding subscale, the Czech and Italian samples were equivalent to four other
samples; the Slovak, German, Portugal and USA samples to three other samples; the Australian,
Chinese and Korean samples to two other samples; and the Turkish sample to no other sample.
Table 5. Signed differential test statistics for the Self-uncompassionate responding subscale
sDTF
AUS
CZK
GER
CHI
ITA
KOR
POR
SVK
TUR
CZK
-0.45,
p = .063
-
-
-
-
-
-
-
-
GER
3.62,
p ˂ .001
2.44,
p ˂ .001
-
-
-
-
-
-
-
CHI
-1.17,
p ˂ .007
-0.39,
p = .279
-0.59,
p = .143
-
-
-
-
-
-
ITA
-1.24,
p ˂ .002
-0.27,
p = .432
-1.28,
p ˂ .001
-0.02,
p = .973
-
-
-
-
-
KOR
2.02,
p ˂ .001
1.51,
p ˂ .001
-0.14,
p = .532
4.39,
p ˂ .001
4.12,
p ˂ .001
-
-
-
-
POR
1.11,
p ˂ .006
0.72,
p ˂ .032
0.91,
p ˂ .008
2.72,
p ˂ .001
2.43,
p ˂ .001
1.13,
p ˂ .001
-
-
-
SVK
-0.87,
p ˂ .003
-0.02,
p = .918
-1.73,
p ˂ .001
1.03,
p ˂ .004
0.85,
p ˂ .007
-0.92,
p ˂ .001
-1.74,
p ˂ .001
-
-
TUR
-1.68,
p ˂ .001
-0.20,
p = .541
-1.33,
p ˂ .001
-0.36,
p = .424
-0.51,
p = .209
-0.69,
p = .056
-2.26,
p ˂ .001
-0.72,
p ˂ .032
-
USA
1.18,
p ˂ .002
0.02,
p = .939
-1.31,
p ˂ .001
0.91,
p ˂ .016
0.69,
p ˂ .047
-0.80,
p ˂ .001
-1.57,
p ˂ .001
-0.11,
p = .678
1.17,
p ˂ .002
Note. Insignificant sDTFs (equivalence between expected total scores obtained) are highlighted in bold.
AUS = Australia (n = 517). CHI = China (n = 321). CZK = Czech Republic (n = 5081). GER =
Germany (n = 2510). ITA = Italy (n = 384). POR = Portugal (n = 512). SVK = Slovakia (n = 1181).
KOR = South Korea (n = 1813). TUR = Turkey (n = 471). USA (n = 833).
MJCP|9, 1, 2021 Kanovský et al.
16
Table 6. Signed differential test statistics for the Self-compassionate responding subscale
sDTF
AUS
CZK
GER
CHI
ITA
KOR
POR
SVK
TUR
CAN
2.10,
p ˂ .001
-
-
-
-
-
-
-
-
CZK
-1.87,
p ˂ .001
-
-
-
-
-
-
-
-
GER
-1.00,
p ˂ .001
0.56,
p ˂ .001
-
-
-
-
-
-
-
CHI
2.86,
p ˂ .001
0.97,
p = .052
0.04,
p = .938
-
-
-
-
-
-
ITA
-0.85,
p ˂ .044
0.03,
p = .931
-0.31,
p = .392
-3.98,
p ˂ .001
-
-
-
-
-
KOR
-0.04,
p = .889
0.79,
p ˂ .001
0.51,
p ˂ .005
-3.95,
p ˂ .001
1.55,
p ˂ .001
-
-
-
-
POR
0.77,
p = .069
0.72,
p ˂ .032
0.62,
p = .076
-1.98,
p ˂ .004
1.65,
p ˂ .001
0.73,
p = .051
-
-
-
SVK
-1.59,
p ˂ .001
-0.10,
p = .619
-0.63,
p ˂ .003
-5.61,
p ˂ .001
-0.39,
p = .229
-0.89,
p ˂ .001
-2.15,
p ˂ .001
-
-
TUR
-2.59,
p ˂ .001
-1.02,
p ˂ .002
-1.51,
p ˂ .001
-6.09,
p ˂ .001
-1.90,
p ˂ .001
-1.43,
p ˂ .001
-3.04
p ˂ .001
-1.37,
p ˂ .001
-
USA
1.14,
p ˂ .002
-0.01,
p = .987
-0.54,
p ˂ .025
-5.10,
p ˂ .001
-0.33,
p = .355
-0.62,
p ˂ .013
-1.77,
p ˂ .001
-0.13,
p = .608
1.50,
p ˂ .001
Note. Insignificant sDTFs (equivalence between expected total scores obtained) are highlighted in bold.
AUS = Australia (n = 517). CHI = China (n = 321). CZK = Czech Republic (n = 5081). GER =
Germany (n = 2510). ITA = Italy (n = 384). POR = Portugal (n = 512). SVK = Slovakia (n = 1181).
KOR = South Korea (n = 1813). TUR = Turkey (n = 471). USA (n = 833).
Note that no perfect transitivity was present: for example, as to the Self-uncompassionate
responding subscale, both USA and Australian samples were equivalent to the Czech sample,
but they were not equivalent one to another. Therefore, we could not create a single linear rank
based on the differences in latent means of equivalent samples, but rather clusters of mutually
comparable samples. For example, again in the case of the Self-uncompassionate responding
subscale, we could fully compare Czech, Slovak, and USA samples because all of them were
mutually equivalent. However, adding another sample for instance, Italian was not possible:
it was equivalent with Czech sample, but not with USA and Slovak samples. Another possible
cluster was composed of Czech, Chinese, Turkish, and Italian samples all of them were
mutually equivalent. However, we could create two hierarchical scales, based on differences in
random effects from the multilevel models (Table 7): for the Self-uncompassionate responding
subscale, the highest position belonged to the German sample, followed by Korean, Portugal,
and Australian samples. Lower levels of Self-uncompassionate responding were displayed in
Czech, USA, and Slovak samples. Very low levels of Self-uncompassionate responding were
found in Chinese, and Italian samples, and the lowest level was in the Turkish sample. As regards
to the Self-compassionate responding subscale, the highest level was reported in the Turkish
sample followed by Slovak, Italian, and Czech samples, At the opposite end, we found
Portuguese, Australian, and Chinese samples to have lower levels of Self-compassionate
responding. We did not attempt to provide any systematic interpretation of these differences
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
17
far more detailed research is required to do so. However, we could see that no discernible
pattern emerged from mutually equivalent samples there was no cultural, linguistic, or
geographical continuum able to explain clusters of mutually equivalent countries.
Table 7. Random effects for countries from the multilevel IRT model
Self-compassionate
responding
Self-uncompassionate
responding
Turkey (0.20)
Germany (0.01)
Slovakia (-0.06)
South Korea (-0.47)
Italy (-0.15)
Portugal (-0.68)
Czech Republic (-0.23)
Australia (-0.96)
USA (-0.25)
Czech Republic (-1.07)
Germany (-0.26)
USA (-1.32)
South Korea (-0.78)
Slovakia (-1.34)
Portugal (-0.94)
Italy (-1.62)
Australia (-1.10)
China (-1.66)
China (-2.14)
Turkey (-1.98)
In the Appendix, we present all test score functions for all comparisons. It was clear after
inspection that even very large differences at particular levels of θ might have a negligible effect
on differences in expected total scores if they were compensated after the intersection of test
score functions. If test score functions did not intersect, the unsigned DTF equals to the signed
DTF: it means that the reference group scores were systematically lower (or higher) than the
focal group across all the range of latent ability. In Appendix, we present more detailed
information for each comparison: (1) absolute values of sDTF and uDTF, (2) their standardized
(percentage) values, (3) 95% confidence intervals for (1) and (2) values, (4) the significance test
of the sDTF (based on 1000 simulation samples), and (5) visualization of test score functions.
This database was useful for subsequent interpretation and could serve as a starting point for
further research. It was beyond the scope of this article to speculate either about the lack of
equivalence between countries, or about the differences in Self-uncompassionate responding
and Self-compassionate responding subscales when equivalence was obtained.
An interesting finding is that some countries showed more invariance with other countries than
others. For example, the Turkish sample was equivalent to no other sample for the Self-
compassion responding subscale. In contrast, the Czech sample was equivalent to six other
samples for the Self-uncompassionate responding subscale. Hypothetically, there are some
clusters of similar countries regarding Self-compassionate responding and Self-
uncompassionate responding subscales, and there are some countries which seem to view self-
compassion differently. As no perfect transitivity was present in our samples, we cannot create
a single linear rank based on the differences in latent means of equivalent samples nor compare
countries on levels of Self-uncompassionate responding and Self-compassionate responding
subscales.
MJCP|9, 1, 2021 Kanovský et al.
18
Although it was impossible to compare total score across all countries, and the differences in
latent means are available only for samples displaying the invariance (Tables 8 and 9), we can
check random effects (relative differences across countries) from the multilevel IRT model. We
can see (Table 7) that for the Self-uncompassionate responding subscale, the highest scores were
found in the German sample, followed by Korean, Portuguese, and Australian samples. Lower
levels of the Self-uncompassionate responding are displayed in Czech, USA, and Slovak
samples. Very low levels of the Self-uncompassionate responding are in Chinese and Italian
samples, and the lowest level is in the Turkish sample. For the Self-compassionate responding
subscale, the highest level is in the Turkish sample, followed by the Slovak, Italian, and Czech
samples, and the lower levels we can find in Portuguese, Australian, and Chinese samples.
Table 8. Latent means differences of invariant samples for the Self-uncompassionate
responding subscale
Mean
AUS
CZK
GER
CHI
ITA
KOR
SVK
CZK
0.188
-
-
-
-
-
-
CHI
-
0.461
1.309
-
-
-
-
ITA
-
0.428
-
0.042
-
-
-
KOR
-
-
0.326
-
-
-
-
SVK
-
0.116
-
-
-
-
-
TUR
-
0.584
-
0.589
0.167
1.032
-
USA
-
0.158
-
-
-
-
0.018
Note. Latent means estimations of populations in first row were constrained to zero. All items were
reverse-scored, so the positive values express lower amount of self-criticism. Differences significant at
0.05 level are highlighted in bold. AUS = Australia (n = 517). CHI = China (n = 321). CZK = Czech
Republic (n = 5081). GER = Germany (n = 2510). ITA = Italy (n = 384). POR = Portugal (n = 512).
SVK = Slovakia (n = 1181). KOR = South Korea (n = 1813). TUR = Turkey (n = 471). USA (n = 833).
Table 9. Latent means differences of invariant samples for the Self-compassionate responding
subscale
Mean
AUS
CZK
GER
ITA
KOR
SVK
CHI
-
-1.076
-1.139
-
-
-
ITA
-
-0.111
0.088
-
-
-
KOR
0.087
-
-
-
-
-
POR
-0.079
-
-0.456
-
-0.194
-
SVK
-
-0.008
-
0.085
-
-
USA
-
0.002
-
0.086
-
-0.024
Note. Latent means estimations of populations in first row were constrained to zero. All items were
reverse scored, so the positive values express lower amount of self-criticism. Differences significant at
0.05 level are highlighted in bold. AUS = Australia (n = 517). CHI = China (n = 321). CZK = Czech
Republic (n = 5081). GER = Germany (n = 2510). ITA = Italy (n = 384). POR = Portugal (n = 512).
SVK = Slovakia (n = 1181). KOR = South Korea (n = 1813). TUR = Turkey (n = 471). USA (n = 833).
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
19
To conclude, there was a remarkable amount of measurement equivalence between different
countries which demonstrates that Self-uncompassionate responding and Self-compassionate
responding subscales are valid and reliable instruments with substantial cross-cultural potential.
However, many comparisons resulted in a lack of measurement equivalence and therefore
displayed differential test functioning. This lack of measurement equivalence may be the result
of linguistic variation, real differences in levels self-compassion across countries, or by
peculiarities in sampling procedures (our samples were far from being representative for
respective populations except for the German sample).
4. Discussion
The goal of this study was to examine measurement invariance by IRT differential test
functioning across ten different samples and nine language versions of the two factors of the
SCS (Self-compassionate responding subscale and Self-uncompassionate responding subscale).
Specifically, we were interested in determining whether comparisons between overall scores of
the two SCS subscales across countries and languages were justified.
The main strength of this study was the amount of countries analysed and thoroughness of the
data design and study, which allowed us to assess the cross-cultural extent and implications of
the evaluated constructs of Self-compassionate responding and Self-uncompassionate
responding. We used a large sample size of at least 300 participants in every sample, recruited
from ten different countries with diverse language and cultural backgrounds.
Considering the diversity of cultures, we were not surprised to find that only one third of the
comparisons between two countries proved to be invariant. For the Self-uncompassionate
responding subscale, only 13 of the 45 comparisons (analysis of every pair) demonstrated
measurement invariance (no differential test functioning). Generally, the results revealed that
the observed constructs Self-compassionate responding and Self-uncompassionate responding
from the SCS were not equivalent among all countries and groups. Therefore, it is impossible
to compare their total scores across all countries. Considering the possible culturally and
linguistically different expressions of self-compassion, future testing around the meanings and
connotations of the constructs across the world is necessary. In addition, it would be beneficial
to look carefully at the items of SCS and test whether they are measuring what they should
measure. This may aid in the development of a new scale allowing for cross-cultural
comparisons of various treatment outcomes related to self-compassion.
The primary advantage of this study is that we could assess and compare, without bias, real
differences in raw or latent scores between countries which do not display differential test
MJCP|9, 1, 2021 Kanovský et al.
20
functioning. Therefore, the differences in raw and latent scores are not distorted by different
functioning of tests in specific countries and languages.
Our research findings are consistent with Heine et al. (1999), who suggested that simultaneously
higher levels of Self-uncompassionate responding may not necessarily be associated with lower
levels of Self-compassionate responding. For example, in the current study, the Czech sample
showed high levels of both Self-uncompassionate responding and Self-compassionate
responding, and the Chinese sample showed low levels in both. Moreover, the German sample
showed very high levels of Self-uncompassionate responding, but it did not show particularly
low levels of Self-compassionate responding. Nevertheless, the Slovak, Italian, and Turkish
samples showed the typical pattern (Neff, 2003) of being low in Self-uncompassionate
responding and high in Self-compassionate responding. These results provide additional
support for not using total SCS score but using scores separately for Self-uncompassionate
responding and Self-compassionate responding subscales.
In addition, based on the results of this study we advise against the use of the unweighted total
scores to compare the relative group responses, while ignoring differential weights of group
membership: e.g., after performing independent t-test or Mann-Whitney non-parametric test on
unweighted total scores, one could easily come to the conclusion that the two populations had
essentially equal mean test scores. But DTF analyses could discover that one population had a
lower latent mean compared with another population and at the same time the first population
could be scored more favourably on the test obtaining positively biased test scores. These two
situations jointly explain the observed equivalence in the total scores, but they have very
different theoretical interpretations and completely different consequences for practical
purposes.
Our results show that the differences in the levels of Self-compassionate responding and Self-
uncompassionate responding subscales cannot be attributed to a simple cultural distinction
between Western and Eastern cultures. Of the two Asian samples, the Chinese sample had low
Self-compassionate and Self-uncompassionate responding, but the Korean sample had low Self-
compassionate responding, but high Self-uncompassionate responding. Such results support
that specific cultural values rather than West-East differences are crucial in understanding cross-
cultural differences of overall self-compassion (Neff et al., 2008), and further indicate that these
cultural values may work differently with the two factors of the Self-Compassion Scale.
The main limitation of our study was that values for the considered variables were self-reported,
and therefore they may have been influenced by socially desirable responses. On the other hand,
the sample was recruited mainly online but also in paper-pencil form. Despite studies that
confirm the reliability of the data obtained from online source (Ritter et al., 2004), these samples
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
21
might be more biased than those obtained using traditional methods. Also results could be
influenced by different forms of obtaining data.
As self-compassion is a construct of high clinical significance, improving understanding of
cross-cultural similarities and differences in Self-uncompassionate responding and Self-
compassionate responding, as measured by the two subscales of the SCS, would have great
impact on practice. This is so, because negative relation to oneself in the form of excessive self-
uncompassionate (or even self-critical) inner speech is one of the most important psychological
processes that influence susceptibility to psychopathology, and also its persistence (Falconer et
al., 2015). On the other hand, self-compassion is generally associated with better psychological
health (Gilbert et al., 2004; Neff, 2003). In an applied context, understanding the differences of
Self-uncompassionate responding and Self-compassionate responding subscales across
countries can help to inform more effective practices.
5. Conclusion
The results of this study substantially contribute to the growing body of knowledge about
similarities and differences among cultures with respect to the two factors of the Self-
Compassion Scale, Self-uncompassionate responding and Self-compassionate responding.
While suggesting that Self-uncompassionate responding and Self-compassionate responding
may tentatively be reflected as general, universal constructs, the results of this research reveal
significant cross-cultural differences in the way these constructs are measured by the subscales
of the SCS.
Given that the expression of self-compassion may vary across different cultures and languages
including Self-uncompassionate responding and Self-compassionate responding, future
international testing of the meanings and connotations of the constructs is necessary. Generally,
the results revealed that the observed constructs Self-uncompassionate responding and Self-
compassionate responding from SCS were not equivalent among all countries and groups.
However, Self-uncompassionate responding and Self-compassionate responding subscales are
valid and reliable instruments with substantial potential of use cross-culturally which needs to
be further explored to allow reliable comparisons of subscales scores across different countries.
Funding
Writing this work was supported by the Vedecká grantová agentúra VEGA under Grant
1/0578/15 and 1/0075/19.
Conflict of Interest Statement
The authors declare that the research was conducted in the absence of any potential conflict of interest.
MJCP|9, 1, 2021 Kanovský et al.
22
References
1. Arimitsu, K. (2014). Development and validation of the Japanese version of the Self-Compassion Scale. The
Japanese Journal of Psychology, 85(1), 5059. https://doi.org/10.4992/jjpsy.85.50
2. 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.
3. Barnett, M. D., & Sharp, K. J. (2016). Maladaptive perfectionism, body image satisfaction, and disordered
eating behaviors among US college women: The mediating role of self-compassion. Personality and Individual
Differences, 99, 225-234. https://doi.org/10.1016/j.paid.2016.05.004
4. Behling, O., & Law, K. S. (2000). Translating questionnaires and other research instruments: Problems and
solutions. Thousand Oaks: Sage Publications. https://doi.org/10.4135/9781412986373
5. Benda, J., & Reichová, A. (2016). Psychometrické charakteristiky české verze Self-Compassion Scale (SCS-
CZ). Československá psychologie, 60(2), 120-136.
6. Birkett, M. A. (2014). Self-compassion and empathy across cultures: Comparison of young adults in China
and the United States. International Journal of Research Studies in Psychology, 3(3), 25-34.
https://doi.org/10.5861/ijrsp.2013.551
7. Brown, L., Bryant, C., Brown, V. M., Bei, B., & Judd, F. K. (2014). Self-Compassion weakens the association
between hot flushes and night sweats and daily life functioning and depression. Maturitas. 78(4), 298-303.
https://doi.org/10.1016/j.maturitas.2014.05.012
8. 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. https://doi.org/10.14195/1647-8606_54_8
9. Chalmers, R. P. (2012). mirt: A multidimensional item response theory package for the R environment.
Journal of Statistical Software, 48(6), 1-29. https://doi.org/10.18637/jss.v048.i06
10. Chalmers, R. P. (2015). Extended mixed-effects item response models with the MH-RM algorithm. Journal
of Educational Measurement, 52, 200-222. https://doi.org/10.1111/jedm.12072
11. Chalmers, R. P, Counsell, A., & Flora, D. B. (2016). It might not make a big DIF: improved differential test
functioning statistics that account for sampling variability. Educational and Psychological Measurement, 76(1), 114
140. https://doi.org/10.1177/0013164415584576
12. 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.
13. Conversano, C., Ciacchini, R., Orrù, G., Di Giuseppe, M., Gemignani, A., & Poli, A. (2020). Mindfulness,
Compassion, and Self-Compassion Among Health Care Professionals: What's New? A Systematic Review.
Frontiers in psychology, 11, 1683. https://doi.org/10.3389/fpsyg.2020.01683
14. Costa, J., Marôco, J., PintoGouveia, J., Ferreira, C., & Castilho, P. (2016). 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. 23(5), 460-468. https://doi.org/10.1002/cpp.1974
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
23
15. Cucina, J. & Byle, K. (2017). The bifactor model fits better than the higher-order model in more than 90%
of comparisons for mental abilities test batteries. Journal of Intelligence, 5, 27.
https://doi.org/10.3390/jintelligence5030027
16. Cunha, M., Xavier, A., & Castilho, P. (2016). Understanding self-compassion in adolescents: Validation study
of the Self-Compassion Scale. Personality and Individual Differences, 93(9), 5662.
https://doi.org/10.1016/j.paid.2015.09.023
17. DeMars, C. E. (2011). An analytic comparison of effect sizes for differential item functioning. Applied
Measurement in Education, 24, 189-209. https://doi.org/10.1080/08957347.2011.580255
18. DeMars, C. E. (2012). A comparison of limited-information and full-information methods in Mplus for
estimating Item Response Theory parameters for non-normal populations. Structural Equation Modeling: A
Multidisciplinary Journal, 19, 610-632. https://doi.org/10.1080/10705511.2012.713272
19. Deniz, M. E., Kesici, S., & Sumer, A. S. (2008): The validity and reliability of the Turkish version of the Self-
Compassion Scale. Social Behavior and Personality, 36(9), 1151-1160.
https://doi.org/10.2224/sbp.2008.36.9.1151
20. 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(1), 58-72.
https://doi.org/10.1080/19012276.2015.1071203
21. Falconer, C.J., King, J.A., & Brewin, C.R. (2015). Demonstrating mood repair with a situation-based measure
of self-compassion and self-criticism. Psychology and psychotherapy. 88(4), 351-65.
https://doi.org/10.1111/papt.12056
22. 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. https://doi.org/10.1186/1477-7525-12-
4
23. Gilbert, P., Clark, M., Hempel, S., Miles, J.N.V. & Irons, C. (2004). Criticising and reassuring oneself: An
exploration of forms, styles and reasons in female students. British Journal of Clinical Psychology, 43, 31-50.
https://doi.org/10.1348/014466504772812959
24. Gilbert, P., & Irons, C. (2008). Shame, self-criticism and self-compassion in adolescence. In NB Allen & LB
Sheeber (Eds.), Adolescent Emotional Development and the Emergence of Depressive Disorders. (pp. 195-214).
Cambridge University Press. https://doi.org/10.1017/CBO9780511551963.011
25. Greenberg, L. S. & Watson, J. (2005). Emotion-focused therapy of Depression. Washington, D. C.: American
Psychological Association Press.
26. Halamová, J., Kanovský, M., & Pacúchová, M. (2017). Self-Compassion Scale: IRT psychometric analysis
validation and factor structure Slovak translation. Psychologica Belgica. 57(4), 190209.
https://doi.org/10.5334/pb.398
27. Halamová, J., Kanovský, M., Petrocchi, N., Moreira, H., López Angarita, A. et al. (2021). The factor structure
of the Self-Compassion Scale in eleven distinct populations across the world. Measurement and Evaluation in
Counseling and Development. 54(1), 1-23, https://doi.org/10.1080/07481756.2020.1735203
MJCP|9, 1, 2021 Kanovský et al.
24
28. Heine, S. J., Lehman, D. R., Markus, H. R., & Kitayama, S. (1999). Is there a universal need for positive self-
regard? Psychological Review, 106, 766- 794. https://doi.org/10.1037/0033-295X.106.4.766
29. Heshmati, R., & Pellerone, M. (2018). Relationship of dispositional mindfulness, self-destruction and self-
criticism in academic performance of boy students. Mediterranean Journal of Clinical Psychology, 6(3).
https://doi.org/10.6092/2282-1619/2018.6.1716
30. Hupfield, J. & Ruffieux, N. (2011). Validierung einer deutschen version der Self-Compassion Scale (SCS-
D). Zeitschrift für Klinische Psychologie und Psychotherapie, 40 (2), 115123. https://doi.org/10.1026/1616-
3443/a000088
31. Hwang, S., Kim, G., Yang, J. W., & Yang, E. (2016). The moderating effects of age on the relationships of
self-compassion, self-esteem, and mental health. Japanese Psychological Research. 58(2), 194-205.
https://doi.org/10.1111/jpr.12109
32. Kankaraš, M., Vermunt, J. K., & Moors, G. (2011). Measurement equivalence of ordinal items: a comparison
of factor analytic, item response theory, and latent class approaches. Sociological Methods & Research, 40(2), 279
310. https://doi.org/10.1177/0049124111405301
33. Kim, E. S., & Yoon, M. (2011). Testing measurement invariance: a comparison of multiple-group categorical
CFA and IRT. Structural Equation Modeling: A Multidisciplinary Journal, 18(2), 212-228.
https://doi.org/10.1080/10705511.2011.557337
34. Kitayama, S., & Karasawa, M. (1997). Implicit self-esteem in Japan: Name letters and birthday numbers.
Personality and Social Psychology Bulletin, 23, 736-742. https://doi.org/10.1177/0146167297237006
35. Kotsou, I., & Leys, C. (2016). Self-Compassion Scale (SCS): Psychometric properties of The French
zranslation and its relations with psychological well-being, affect and depression. PloS one, 11(4).
https://doi.org/10.1371/journal.pone.0152880
36. 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). https://doi.org/10.1371/journal.pone.0136598
37. Khramtsova I.I., & Chuykova T.S. (2016). Mindfulness and self-compassion as predictors of humor styles
in US and Russia. Sotsial'naia psikhologiia i obshchestvo = Social Psychology and Society, 7(2), 93108.
https://doi.org/10.17759/sps.2016070207
38. Kwan, V. S. Y., Kuang, L. L., & Hui, N. H. H. (2009). Identifying the sources of self-esteem: The mixed
medley of benevolence, merit, and bias. Self and Identity, 8, 176-195.
https://doi.org/10.1080/15298860802504874
39. Lee, W. K., & Lee, K. (2010). The validation study of the Korean version of Self-Compassion Scale with
adult women in community [Korean]. Journal of Korean Neuropsychiatric Association, 49(2), 193-200.
40. 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). https://doi.org/10.1371/journal.pone.0132940
41. Mansolf, M., & Reise, S. P. (2017). When and why the second-order and bifactor models are distinguishable.
Intelligence, 61, 120-129. https://doi.org/10.1016/j.intell.2017.01.012
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
25
42. Mantzios, M., Wilson, J. C., & Giannou, K. (2013). Psychometric properties of the Greek versions of the
self-compassion and mindful attention and awareness scales. Mindfulness, 6(1), 123-132.
https://doi.org/10.1007/s12671-013-0237-3
43. Meade, A. W. & Lautenschlager, G. J. (2004). A comparison of item response theory and confirmatory factor
analytic methodologies for establishing measurement equivalence/invariance, Organizational Research Methods,
7(4), 361-388. https://doi.org/10.1177/1094428104268027
44. Meares, R., Gerull, F., Stevenson, J., & Korner, A. (2011). Is self-disturbance the core of borderline
personality disorder? An outcome study of borderline personality factors. Australian and New Zealand Journal
of Psychiatry, 45, 214222. https://doi.org/10.3109/00048674.2010.551280
45. Merlo, E. M., Stoian, A. P., Motofei, I. G., & Settineri, S. (2020). Clinical Psychological Figures in Healthcare
Professionals: Resilience and Maladjustment as the “Cost of Care”. Frontiers in Psychology, 11.
https://doi.org/10.3389/fpsyg.2020.607783
46. Montero-Marín, J., Gaete, J., Demarzo, M., Rodero, B., Lopez, L.C., & 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. 30(7), 1281. https://doi.org/10.3389/fpsyg.2016.01281
47. Moreira, H., Carona, C., Silva, N., Nunes, J., & Canavarro, M. C. (2016). Exploring the link between maternal
attachment-related anxiety and avoidance and mindful parenting: The mediating role of self-compassion.
Psychology and Psychotherapy: Theory, Research and Practice, 89(4), 369-384. https://doi.org/10.1111/papt.12082
48. Moreira, H., Gouveia, M. J., Carona, C., Silva, N., & Canavarro, M. C. (2015). Maternal attachment and
children's quality of life: The mediating role of self-compassion and parenting stress. Journal of Child and Family
Studies, 24(8), 2332-2344. https://doi.org/10.1007/s10826-014-0036-z
49. Murray, G., Judd, F., Jackson, H., Fraser, C., Komiti, A., Hodgins, G., Pattison, P., Humphreys, J., & Robins,
G. (2004). Rurality and mental health: the role of accessibility. Australian and New Zealand Journal of Psychiatry.
38(8), 629-634. https://doi.org/10.1111/j.1440-1614.2004.01426.x
50. Neff, K. D. (2003). The development and validation of a scale to measure self-compassion. Self and identity,
2(3), 223-250. https://doi.org/10.1080/15298860390209035
51. Neff, K. D. (2004). Self-compassion and psychological well-being. Constructivism in the Human Sciences, 9, 27-
37. https://doi.org/10.1037/e633942013-240
52. Neff, K. D. (2016). The Self-Compassion Scale is a valid and theoretically coherent measure of self-
compassion. Mindfulness, 7(1), 264-274. https://doi.org/10.1007/s12671-015-0479-3
53. Neff, K. D., & Dahm, K. A. (2014). Self-Compassion: What it is, what it does, and how it relates to
mindfulness (pp. 121-140). In M. Robinson, B. Meier & B. Ostafin (Eds.) Mindfulness and Self-Regulation. New
York: Springer. https://doi.org/10.1007/978-1-4939-2263-5_10
54. Neff, K. D., Pisitsungkagarn, K., & Hseih, Y. (2008). Self-compassion and self-construal in the United States,
Thailand, and Taiwan. Journal of Cross-Cultural Psychology, 39, 267-285.
https://doi.org/10.1177/0022022108314544
55. Neff, K. D. & Vonk, R. (2009). Self-compassion versus global self-esteem: Two different ways of relating to
oneself. Journal of Personality, 77(1), 23-50. https://doi.org/10.1111/j.1467-6494.2008.00537.x
MJCP|9, 1, 2021 Kanovský et al.
26
56. Neff, K. D., Whittakar, 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, 596607. https://doi.org/1080/00223891.2016.1269334
57. O’Connor, R. C. & Noyce, R. (2008). Personality and cognitive processes: Self-criticism and different types
of rumination as predictors of suicidal ideation. Behaviour Research and Therapy, 46(3), 392401.
https://doi.org/10.1016/j.brat.2008.01.007
58. 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.
https://doi.org/10.3109/09638237.2013.841869
59. Pisitsungkagarn, K., Taephant, N., & Attasaranya, P. (2014). Body image satisfaction and self-esteem in Thai
female adolescents: the moderating role of self-compassion. International journal of adolescent medicine and health,
26(3), 333-338. https://doi.org/10.1515/ijamh-2013-0307.
60. Raju, N. S., Laffitte, L. J. & Byrne, B. B. (2002). Measurement equivalence: A comparison of methods based
on confirmatory factor analysis and item response theory. Journal of Applied Psychology, 87(3), 517529.
https://doi.org/10.1037//0021-9010.87.3.517
61. Reise, S. P., Widaman, K. H. & Pugh, R.H. (1993). Confirmatory factor analysis and item response theory:
two approaches for exploring measurement invariance, Psychological Bulletin, 114(3), 552-566.
https://doi.org/10.1037/0033-2909.114.3.552
62. Ritter, P., Lorig, K., Laurent, D., & Matthews, K. (2004). Internet versus mailed questionnaires: a randomized
comparison. Journal of Medical Internet Research, 6(3), e29. https://doi.org/10.2196/jmir.6.3.e29
63. Shahar, B., Carlin, E.R., Engle, D.E., Hegde, J., Szepsenwol, O., & Arkowitz, H. (2012). A pilot investigation
of emotion-focused two-chair dialogue intervention for self-criticism. Clinical Psychology and Psychotherapy, 6,
496507. https://doi.org/10.1002/cpp.762
64. Shahar, B., Szsepsenwol, O., Zilcha-Mano, S., Haim, N., Zamir, O., Levi-Yeshuvi, S., & Levit-Binnun, N.
(2015). A waitlist randomized controlled trial of lovingkindness meditation programme for selfcriticism.
Clinical psychology & psychotherapy, 22(4), 346-356. https://doi.org/10.1002/cpp.1893
65. Şenyuva, E., Kaya, H., Işik, B., & Bodur, G. (2014). Relationship between self-compassion and emotional
intelligence in nursing students. International journal of nursing practice, 20(6), 588-596.
https://doi.org/10.1111/ijn.12204
66. Terry, M. L., & Leary, M. R. (2011). Self-compassion, self-regulation, and health. Self and Identity, 10(3), 352
362. https://doi.org/10.1080/15298868.2011.558404
67. Tóth-Király, I., the, B., & Orosz, G. (2016). Exploratory structural equation modeling analysis of the Self-
Compassion Scale. Mindfulness, 7, 112. https://doi.org/10.1007/s12671-016-0662-1
68. Vandenberg, R. J. & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature:
suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3(1),
4-70. https://doi.org/10.1177/109442810031002
69. Voruz, S. (2013). An examination of the psychometric properties of the Self-Compassion Scale Short Form
among a help-seeking clinical sample (Doctoral dissertation, Pacific University).
Retrieved from: http://commons.pacificu.edu/spp/430
MJCP|9, 1, 2021 The Invariance Analysis of The Self-Compassion Scale
27
70. Warren, R., Smeets, E. & Neff, K. D. (2016). Self-criticism and self-compassion: Risk and resilience for
psychopathology. Current Psychiatry, 15(12), 18-32.
71. 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(6), 1996-2009. https://doi.org/10.1007/s10943-016-0205-z
©2021 by the Author(s); licensee Mediterranean Journal of Clinical
Psychology, Messina, Italy. This article is an open access article, licensed
under a Creative Commons Attribution 4.0 Unported License.
Mediterranean Journal of Clinical Psychology, Vol. 9, No. 1 (2021).
International License (https://creativecommons.org/licenses/by/4.0/).
DOI: 10.6092/2282-1619/mjcp-2682
Thesis
Full-text available
Background. Frequent comorbidity of mental disorders together with the latest findings in neurobiology have been drawing attention of health professionals over the past years to so-called transdiagnostic factors, including, but not limited to difficulties in emotion regulation and self-compassion. It seems that these factors play a significant role in the etiology of many mental disorders as well as in the maintenance of mental health and well-being. Objectives. The aim of this thesis was to theoretically discuss and scientifically verify the expected relationship between parental emotional warmth in childhood, self-compassion and emotion regulation. Sample and procedure. 440 adult respondents (141 men and 299 women) completed the questionnaire “My Memories of Upbringing”- Short Form (s-EMBU), the Difficulties in Emotion Regulation Scale-Short Form (DERS-SF-CZ) and the Self-Compassion Scale (SCS-CZ) together with several demographic questions in an online survey. Statistical analysis. Descriptive statistics and the Pearson's correlation coefficients were calculated using IBM SPSS Statistics-23 software. The effect of gender and education on key variables was tested using two-way multivariate analysis of variance (MANOVA). Gender differences in correlations were compared using Fisher's r-to-z transformation. We also performed mediation analysis according to Baron and Kenny's procedure. Results. The results revealed significant correlations between the observed variables and showed that self-compassion was a significant mediator of the relationship between parental emotional warmth in childhood and difficulties in emotion regulation. The correlations of parental emotional warmth in childhood with self-compassion as well as with difficulties in emotion regulation were significantly higher in men than in women. Study limitations. This study relied exclusively on self-report measures. Respondents evaluated their parents' warmth only retrospectively.
Article
Full-text available
Current interest in the assessment of measurement equivalence emphasizes 2 major methods of analysis. The authors offer a comparison of a linear method (confirmatory factor analysis) and a nonlinear method (differential item and test functioning using item response theory) with an emphasis on their methodological similarities and differences. The 2 approaches test for the equality of true scores (or expected raw scores) across 2 populations when the latent (or factor) score is held constant. Both approaches can provide information about when measurement nonequivalence exists and the extent to which it is a problem. An empirical example is used to illustrate the 2 approaches.
Article
Full-text available
Background: The health professionals are involved in the paths of care for patients with different medical conditions. Their life is frequently characterized by psychopathological outcomes so that it is possible to identify consistent burdens. Besides the possibility to develop pathological outcomes, some protective factors such as resilience play a fundamental role in facilitating the adaptation process and the management of maladaptive patterns. Personal characteristics and specific indexes such as burdens and resilience are essential variables useful to study in-depth ongoing conditions and possible interventions. The study was aimed at highlighting the presence and the relations among factors as personal variables, burdens, and resilience, to understand health professionals' specific structure and functions. Methods: The observation group was composed of 210 participants, 55 males (26.2%), and 155 females (73.8%), aged from 18 to 30 years old with a mean age of 25.92 years old (SD = 3.33). The study considered personal characteristics of the subjects, such as age, gender, years of study, days of work per week, hours of work per week, and years of work. Our study had been conducted with the use of measures related to burdens (Caregivers Burden Inventory) and resilience (Resilience Scale for Adults). Results: The performed analyses consisted of descriptive statistics, correlations, and regressions among the considered variables. Several significant correlations emerged among personal characteristics, CBI, and RSA variables. Specifically, age and work commitment indexes appeared to be significantly related to the development of burdens, differently from the years of study. Significant correlations emerged among personal and RSA variables, indicating precise directions for both domains. Age and gender were identified as predictors to perform multivariate regression analyses concerning CBI factors. Significant dependence relations emerged with reference to all CBI variables. Conclusion: Pathological outcomes and resilience factors represent two sides of the health professionals' experiences, also known as “invisible patients.” Greater knowledge about present conditions and future possibilities is a well-known need in literature so that the current analyses considered fundamental factors. In line with state of the art, future studies are needed in order to deepen elusive phenomena underlying maladjustment.
Article
Full-text available
The factor structure of mental abilities has most often been depicted using a higher-order model. Under this model, general mental ability (g) is placed at the top of a pyramid, with "loading" arrows going from it to the other factors of intelligence, which in turn go to subtest scores. In contrast, under the bifactor model (also known as the nested factors/direct hierarchical model), each subtest score has its own direct loading on g; the non-g factors (e.g., the broad abilities) do not mediate the relationships of the subtest scores with g. Here we summarized past research that compared the fit of higher-order and bifactor models using confirmatory factor analysis (CFA). We also analyzed additional archival datasets to compare the fit of the two models. Using a total database consisting of 31 test batteries, 58 datasets, and 1,712,509 test takers, we found stronger support for a bifactor model of g than for the traditional higher-order model. Across 166 comparisons, the bifactor model had median increases of 0.076 for the Comparative Fit Index (CFI), 0.083 for the Tucker-Lewis Index (TLI), and 0.078 for the Normed Fit Index (NFI) and decreases of 0.028 for the root mean square error of approximation (RMSEA) and 1343 for the Akaike Information Criterion (AIC). Consequently, researchers should consider using bifactor models when conducting CFAs. The bifactor model also makes the unique contributions of g and the broad abilities to subtest scores more salient to test users.
Article
Full-text available
Health care professionals (HCPs) are a population at risk for high levels of burnout and compassion fatigue. The aim of the present systematic review was to give an overview on recent literature about mindfulness and compassion characteristics of HCPs, while exploring the effectiveness of techniques, involving the two aspects, such as MBSR or mindfulness intervention and compassion fatigue-related programs. A search of databases, including PubMed and PsycINFO, was conducted following the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines and the methodological quality for this systematic review was appraised using AMSTAR-2 (A MeaSurement Tool to Assess systematic Reviews-2). The number of articles that met the inclusion criteria was 58 (4 RCTs, 24 studies with pre-post measurements, 12 cross-sectional studies, 11 cohort studies and 7 qualitative studies). MBSR intervention was effective at improving, and maintaining, mindfulness and self-compassion levels and to improve burnout, depression, anxiety, stress. The most frequently employed interventional strategies were mindfulness-related trainings that were effective at improving mindfulness and self-compassion, but not compassion fatigue, levels. Compassion-related interventions have been shown to improve self-compassion, mindfulness and interpersonal conflict levels. Mindfulness was effective at improving negative affect and compassion fatigue, while compassion satisfaction may be related to cultivation of positive affect. This systematic review summarized the evidence regarding mindfulness- and compassion-related qualities of HCPs as well as potential effects of MBSR, mindfulness-related and compassion-related interventions on professionals' psychological variables like mindfulness, self-compassion and quality of life. Combining structured mindfulness and compassion cultivation trainings may enhance the effects of interventions, limit the variability of intervention protocols and improve data comparability of future research.
Article
Full-text available
The present study verifies the psychometric properties of the Slovak version of the Self-Compassion Scale through item response theory, factor-analysis, validity analyses and norm development. The surveyed sample consisted of 1,181 participants (34% men and 66% women) with a mean age of 30.30 years (SD = 12.40). Two general factors (Self-compassionate responding and Self-uncompassionate responding) were identified, whereas there was no support for a single general factor of the scale and six subscales. The results of the factor analysis were supported by an independent sample of 676 participants. Therefore, the use of total score for the whole scale would be inappropriate. In Slovak language the Self-Compassion Scale should be used in the form of two general subscales (Self-compassionate responding and Self-uncompassionate responding). In line with our theoretical assumptions, we obtained relatively high Spearman’s correlation coefficients between the Self-Compassion Scale and related external variables, demonstrating construct validity for the scale. To sum up, the Slovak translation of The Self-Compassion Scale is a reliable and valid instrument that measures Self-compassionate responding and Self-uncompassionate responding.
Article
Full-text available
The present research investigated the construct validity and psychometric properties of the Self-Compassion Scale (SCS) with a new and advanced statistical procedure, exploratory structural equation modeling (ESEM), in order to contribute to the ongoing discussion about its dimensionality by employing a bifactor-ESEM framework. A Hungarian online representative sample (N = 505, Nfemale = 265, Mage = 44.37) filled out the Hungarian version of the SCS. Confirmatory factor analysis (CFA) and ESEM methods were employed, and first-order and bifactor solutions were examined and compared. The bifactor ESEM model demonstrated the best fit to the data with the joint presence of the general self-compassion factor and the specific factors. Internal consistency was adequate in all cases. Reliability indices—omega and omega hierarchical—showed that not all specific factors had unique contributions over and above the general factor. High levels of gender invariance were also achieved with females having lower general self-compassion and self-judgment latent means, while having higher self-kindness scores. The findings shed new light on the underlying theory behind the SCS and proved the usefulness of the bifactor ESEM framework in the investigation of multidimensional constructs.
Article
Full-text available
Background: The use of the Self-Compassion Scale (SCS) as a single measure has been pointed out as problematic by many authors and its originally proposed structure has repeatedly been called into question. The negative facets of this construct are more strongly related to psychopathology than the positive indicators. The aim of this study was to evaluate and compare the different structures proposed for the SCS, including a new measure based only on the negative factors, and to assess the psychometric features of the more plausible solution. Method: The study employed a cross-sectional and cross-cultural design. A sample of Brazilian (n = 406) and Spanish (n = 416) primary care professionals completed the SCS, and other questionnaires to measure psychological health-related variables. The SCS factor structure was estimated using confirmatory factor analysis by the maximum likelihood method. Internal consistency was assessed by squaring the correlation between the latent true variable and the observed variables. The relationships between the SCS and other constructs were analyzed using Spearman's rs. Results: The structure with the best fit was comprised of the three negative first-order factors of “self-judgment”, “isolation” and “over-identification”, and one negative second-order factor, which has been named “self-criticism” [CFI = 0.92; RMSEA = 0.06 (90% CI = 0.05–0.07); SRMR = 0.05]. This solution was supported by both samples, presented partial metric invariance [CFI = 0.91; RMSEA = 0.06 (90% CI = 0.05–0.06); SRMR = 0.06], and showed significant correlations with other health-related psychological constructs. Reliability was adequate for all the dimensions (R ≥ 0.70). Conclusions: The original structure proposed for the SCS was not supported by the data. Self-criticism, comprising only the negative SCS factors, might be a measure of uncompassionate behaviors toward the self, with good psychometric properties and practical implications from a clinical point of view, reaching a stable structure and overcoming possible methodological artifacts.
Article
Full-text available
Objectives. The purpose of the study was to develop and validate a Czech version of the Self-Compassion Scale (SCS-CZ) and to examine its psychometric properties. Sample and setting. Three separate Czech samples were used. In Survey 1, the SCS was translated into Czech with a back-translation procedure and then administered through an on-line survey. 5368 participants (sample 1) completed the SCS-CZ, two subscales of the FFMQ, the RSES, the SWLS and a scale measuring the intensity of non-clinical narcissism. In Survey 2, 415 participants (sample 2) completed the SCS-CZ, the BDI-II, the STAI X-2 and the UP-SRS. In Survey 3, 58 participants (sample 3) completed the SCS-CZ two times with the time interval of three weeks. Statistical analysis. Using IBM SPSS Statistics-22 and Amos-23, calculations of descriptive statistics, Cronbach's alpha, Pearson correlation coefficients, test-retest analysis and confirmatory factor analysis were performed. Results. The confirmatory factor analysis (sample 1) confirmed the six-factor structure of SCS-CZ after elimination of six items (SRMR=0.045, RMSEA=0.059, CFI=0.929, GFI=0.942, TLI=0.913). A single higher-order factor model was then also confirmed. The Cronbach's alpha coefficients for total SCS-CZ and subscales were as follows: SCS-CZ (alpha=0,89), self-kindness (alpha=0,86), self-judgment (alpha=0,67), common humanity (alpha=0,75), isolation (alpha=0,78), mindfulness (alpha=0,65) and over-identification (alpha=0,69). As expected, in Survey 1 and 2, the SCS-CZ exhibited a significant positive correlation to measures of mindfulness, self-esteem, life satisfaction and self-acceptance and a significant negative correlation to measures of non-clinical narcissism, depression and trait anxiety. In Survey 3, the SCS-CZ showed a test-retest reliability of 0.90. Study limitation. This study relied exclusively on self-report measures in order to assess validity. The authors recommend the use of interviews as well as clinical or meditator samples in future studies.
Article
Previous studies (Gignac, 2016; Murray & Johnson, 2013) have explored conditions for which accurate statistical comparison of a second-order and bifactor model is challenging, if not impossible. However, a mathematical basis for this problem has not been offered. We show that a second-order model implies a unique set of tetrad constraints (Bollen & Ting, 1993, 1998, 2000) that the bifactor model does not, and that the two models are distinguishable to the degree that these unique tetrad constraints are violated. Simulated population matrices, and mathematical proofs, are used to demonstrate that: (a) when a second-order model with cross-loadings or correlated residuals is true, fitting a pure (misspecified) second-order model leads to violation of the tetrad constraints, which in turn leads to the chi-square and Bayesian Information Criterion favoring a (misspecified) bifactor model, and (b) a true bifactor model can be identified only when the tetrad constraints of the second-order model are violated, which is mainly a function of the proportionality of loadings. Three model-comparison approaches are offered for applied researchers.
Article
This study examined the factor structure of the Self-Compassion Scale (SCS) using a bifactor model, a higher order model, a 6-factor correlated model, a 2-factor correlated model, and a 1-factor model in 4 distinct populations: college undergraduates (N = 222), community adults (N = 1,394), individuals practicing Buddhist meditation (N = 215), and a clinical sample of individuals with a history of recurrent depression (N = 390). The 6-factor correlated model demonstrated the best fit across samples, whereas the 1- and 2-factor models had poor fit. The higher order model also showed relatively poor fit across samples, suggesting it is not representative of the relationship between subscale factors and a general self-compassion factor. The bifactor model, however, had acceptable fit in the student, community, and meditator samples. Although fit was suboptimal in the clinical sample, results suggested an overall self-compassion factor could still be interpreted with some confidence. Moreover, estimates suggested a general self-compassion factor accounted for at least 90% of the reliable variance in SCS scores across samples, and item factor loadings and intercepts were equivalent across samples. Results suggest that a total SCS score can be used as an overall mesure of self-compassion.