English (pdf)
Article in xml format
Article references
Automatic translation
Permalink
Shame is a transdiagnostic factor associated with symptoms of post-traumatic stress disorder (DeCou et al., 2023), obsessive-compulsive disorder (Laving et al., 2022), social anxiety (Swee et al., 2021), binge eating (O’Loghlen et al., 2022) and other eating disorders (Nechita et al., 2021), as well as psychosis (Davies et al., 2024). Additionally, this construct is also relevant for stigmatized physical conditions. For example, in the case of people living with HIV/AIDS, shame is associated with less healthcare utilization and poorer physical health, which could impact the survival of this population (Bennett et al., 2016). Furthermore, shame is also related to negative attitudes towards seeking help for mental health difficulties (Rüsch et al., 2014), hindering access to appropriate care and possibly aggravating psychological conditions that require intervention before causing more harm. Moreover, according to the review by Di Sarno et al. (2024), empirical findings suggest that shame could negatively affect psychotherapy in terms of therapeutic alliance, especially when the levels of shame avoidance are high in clients and when shame-proneness is high in therapists, making the latter less able to work with ashamed patients.
Shame is classically defined as an experience that involves evaluating the self as inadequate, while scientific research on the construct highlights the differences between state and trait shame, also called shame proneness or dispositional shame (Lear et al., 2022). State shame refers to a temporarily felt and context-dependent emotion, whereas trait shame, also called shame-proneness, refers to a continuous and generalized tendency to experience shame, even in response to minimally activating stimuli (Sullins et al., 2024). Apart from the difference between state and trait shame, the biopsychosocial model of shame by Gilbert (2022) distinguishes between external shame, which occurs when the negative evaluation of the self-derives from others, and internal shame, when the individual experiences a negative self-evaluation.
Until recently, the most widely used instruments assessed shame solely in its external or internal form, lacking items that measured both dimensions within the same scale (Lear et al., 2022). Later on, Ferreira et al. (2022) developed the External and Internal Shame Scale (EISS) in Portugal, an instrument which aims to assess trait shame in a bidimensional manner; one component related to others’ evaluations, and another related to self-evaluation. Although relatively recent, the EISS scale has already shown associations with psychopathological symptoms, such as depression, anxiety, and stress (Cunha et al., 2021; Hiramatsu et al., 2021; Fekih-Romdhane et al., 2023; Ferreira et al., 2022). When creating the scale, Ferreira et al. (2022) found a good fit for a model with one higher order factor, global sense of shame, with two lower order factors, namely, internal and external shame [χ2 = 126.73, df = 19, p < 0.001, CFI = 0.96, GFI = 0.95, TLI = 0.94, RMSEA = 0.09 (90% CI = 0.08 - 0.11), SRMR = 0.04]. This hierarchical model was also tested by a study of measurement invariance among the countries of Portugal, France, Singapore, Japan, and Australia (Matos et al., 2023), revealing a good model fit (χ2 = 342.70, df = 19, p < 0.001, CFI = 0.93, GFI = 0.94, TLI = 0.89, RMSEA = 0.11 (90% CI = 0.10 - 0.12), SRMR = 0.05]. However, in the Arabic adaptation of the scale done in Lebanon, Fekih-Romdhane et al. (2023) found only a one-factor structure for the items, with adequate fit indices [χ2 = 66.71, df = 18, p < 0.001, CFI = 0.958, TLI = 0.935, RMSEA = 0.096 (90% CI = 0.072 - 0.122), SRMR = 0.036].
In addition to describing and explaining the phenomenon of shame, theoretical models have also sought to identify different coping styles when experiencing shame. The Compass of Shame model describes four kinds of reactions to shame, which would function as scripts to manage the emotion: withdrawal, attack self, attack others, and avoidance (Capinha et al., 2021). The poles of the compass are not to be taken as purely independent factors; they are able to occur simultaneously and successively. Withdrawal and attack self are reactions based on internalization of shame, in which the withdrawal involves limiting the exposure to shame by removing oneself from the situation or hiding from others, while attack self is linked to self-criticism over perceived shameful attributes. Attack other is an externalization-based coping style in which an individual responds with anger and rejection toward the source of shame, attempting to silence its message. Lastly, avoidance entails attempting to escape awareness of shame-related feelings through distraction or denial of the emotional significance of the experience (Capinha et al., 2021). Later, an adaptive response to shame was incorporated into the model, characterized by self-reassurance in the face of criticism, openness and flexibility in handling negative feedback or the need for repair, and seeking affection and acceptance in relationships (Martin et al., 2021).
The Compass of Shame Scale (CoSS) was developed by Elison et al. (2006) to assess how these four coping mechanisms are employed in shame-evoking situations, demonstrating an acceptable fit to the data [CFI = 0.94, RMSEA = 0.055 (90% CI = 0.052 - 0.059)]. All subscales of the CoSS have been found to correlate with each other (Capinha et al., 2021; Paulo et al., 2020; Yelsma et al., 2002). Interestingly, withdrawal, attack self, and attack other correlate with trait shame (Akıncı & Atintaş, 2020), albeit not specifically with trait shame measured by the EISS (Ferreira et al., 2022). The shame coping styles of the CoSS have also shown associations with depression and gambling (Rapinda et al., 2022), aggression and self-injury (Martin et al., 2021), and symptoms of psychopathology (Capinha et al., 2021), showing the relevance of this model for mental-health difficulties.
Throughout time, the CoSS has undergone changes, appearing in more than one version. Its most recent form, CoSS-5, aims to measure the four dysfunctional shame coping styles alongside an adaptive one in a five-factor model. Capinha et al. (2021) demonstrated the scale’s validity in a Portuguese sample; [χ2 = 4089.52, df = 1585, p < 0.001, CFI = 0.787, RMSEA = 0.051 (90% CI = 0.049 - 0.053), SRMR = 0.068]. Nonetheless, the CoSS-5’s theoretical structure is not always replicated. In the Portuguese study conducted by Vagos et al. (2019) using a sample of adolescent boys, the five-factor model of the CoSS-5 showed only a partial adjustment, [CFI = 0.802, RMSEA = 0.049 (90% CI 0.048 - 0.050); SRMR = 0.070]. In the case of its German version (Martin et al., 2021), there were incompatible measures for the five-factor model (GFI = 0.88, AGFI = 0.87, NFI = 0.85) that were solved by removing the adaptive subscale, ending with a four-factor structure (GFI = 0.91, AGFI = 0.90, NFI = 0.89). Moreover, a Dutch adaptation of a slightly shortened version of the CoSS-5 (Schalkwijk et al., 2016) identified a model with six factors: Internalization (with the subscales of attack self and withdrawal), Externalization (with the subscales of attack other and avoidance), Adaptive, and the additional “shame-proneness” subscale, with a higher order factor that encompasses named “shame regulation”, revealing a good fit (χ2 = 1754, df = 732, p < 0.001, CFI = 0.903, RMSEA = 0.065).
Since shame is a multifaceted phenomenon with distinct forms and manifestations with transdiagnostic factors that underly many mental health conditions, it is crucial to have robust measures capable of operationalizing and capturing this factor in the Brazilian population. To measure the construct of shame in the Brazilian population, only the following instruments are available (to our knowledge): the Shame and Stigma Scale (SSS) (Pirola et al., 2017), the body shame subscale of the Body Self-Objectification Scale (Loureiro et al., 2022), and the Other as Shamer Scale - 2 (OAS-2), which measures external shame (Santos et al., 2015). These measures, however, present some limitations, both in terms of assessing different constructs within the same measure as seen with the SSS (Pirola et al., 2017), or with evaluating only parts of shame instead of capturing the construct as a whole, as seen in the subscale presented in the Loureiro et al. (2022) study which only captures body shame, or with the OAS-2, which only measures external shame.
Moreover, both the EISS and CoSS-5 have not yet been adapted for use with the Brazilian population, nor have there been psychometric studies of this nature in our context. Since this transdiagnostic is a highly culturally sensitive experience, as posited by the biopsychosocial model of shame (Gilbert, 2022), researchers and clinicians cannot assume that results of trait shame and shame coping styles found elsewhere can be applied to the Brazilian context.
In this scenario, the present study has the following goals: a) investigating the internal structure of the CoSS-5 and EISS scales when adapted for Brazilian Portuguese; b) examining the reliability of CoSS-5 and EISS scores considering internal consistency; c) exploring the relationships between trait shame and shame coping styles in a Brazilian sample; and d) since shame is a transdiagnostic factor, investigating differences in trait shame levels and shame coping strategies between individuals with and without a self-reported psychiatric diagnosis.
The following hypotheses will be investigated:
H1a) internal structure of the EISS scale: considering previous studies (Ferreira et al., 2022; Matos et al., 2023), the Brazilian Portuguese adaptation is expected to maintain the two-factor structure (internal and external shame) with a higher-order global shame factor, as in the original version.
H1b) internal structure of the CoSS-5 scale: considering previous studies (Capinha et al., 2021; Elison et al., 2006), the Brazilian Portuguese adaptation is expected to replicate the original five-factor structure.
H2a) reliability of the EISS scale: the Brazilian Portuguese adaptation is expected to show internal consistency reliability (≥ 0.70), as found in previous studies (Ferreira et al., 2022; Matos et al., 2023).
H2b) reliability of the CoSS-5 scale: the Brazilian Portuguese adaptation is expected to show internal consistency reliability (≥ 0.70), as found in previous studies (Capinha et al., 2021; Vagos et al., 2019).
H3) relationships between trait shame and shame coping styles: trait shame is expected to correlate positively with maladaptive coping, and negatively with adaptive shame management. It also predicts coping styles and influences emotional responses, as shown in previous studies (Akıncı & Atintaş, 2020; Capinha et al., 2021; Ferreira et al., 2022; Martin et al., 2021; Rapinda et al., 2022).
H4) differences between individuals with and without a diagnosis: individuals with a self-reported psychiatric diagnosis are expected to show higher trait shame, maladaptive shame coping styles, and lower adaptive shame coping, as shame levels are linked to poorer mental health outcomes (Capinha et al., 2021; Cunha et al., 2021; Ferreira et al., 2022; Rapinda et al., 2022).
METHOD
PARTICIPANTS
Initially, an online calculator (Soper, 2025) was used to estimate the required sample size to investigate the internal structure of the instruments, considering an anticipated effect size of 0.30, a statistical power of 0.80, and a significance level of 0.05. For the EISS (Ferreira et al., 2022), eight items and two latent variables were considered, resulting in a minimum sample size of 90 participants. For the CoSS-5 (Schalkwijk et al., 2016), 58 items and five factors were considered, which resulted in a minimum sample size of 150 participants. Since this was the smallest possible number, researchers aimed to collect more data to increase sample variability. The predetermined inclusion criteria to participate in the research were to be Brazilian and to be an adult, namely, over 18 years old, given that the objective was to have a wide sample with varied socioeconomic and education levels. A total of 312 responses were collected, with 270 remaining after data cleaning (age range: 18-71 years, M = 36.35, SD = 11.01). In summary, the majority of the sample consisted of women (81.48%), Catholics (29.63%), individuals with a single marital status (44.07%), White participants (64.44%), those earning more than five times the minimum wage (40.37%), individuals with a college degree (54.44%), and those without a self-reported psychiatric diagnosis (62.96%). (see Table 1).
Table 1 Sample Descriptions
| Descriptions | n | % |
|---|---|---|
| Gender | ||
| Woman (cis or trans) | 46 | 17.0 |
| Man (cis or trans) | 220 | 81.5 |
| Non-binary | 3 | 1.1 |
| Other | 1 | 0.4 |
| Religion | ||
| Atheist/agnostic | 56 | 20.7 |
| Buddhist | 4 | 1.5 |
| Catholic | 80 | 29.6 |
| Spiritist | 35 | 13.0 |
| Evangelical | 45 | 16.7 |
| Other | 50 | 18.5 |
| Marital status | ||
| Marien | 89 | 33.0 |
| Divorced/separated | 18 | 6.7 |
| Single | 119 | 44.1 |
| Stable Union/cohabiting | 44 | 16.3 |
| Race | ||
| Yellow | 1 | 0.4 |
| White | 174 | 64.4 |
| “Pardo”* | 70 | 25.9 |
| Black | 25 | 9.3 |
| Income | ||
| 1 to 2 minimum wages | 39 | 14.4 |
| 3 to 5 minimum wages | 106 | 39.3 |
| More than 5 minimum wages | 109 | 40.4 |
| Less than 1 minimum wage | 9 | 3.3 |
| Didn't know or wasn't sure | 7 | 2.6 |
| Education | ||
| Incomplete elementary school | 1 | 0.4 |
| Completed elementary school | 2 | 0.7 |
| Completed technical high school | 10 | 3.7 |
| Incomplete undergraduate degree | 31 | 11.5 |
| Completed undergraduate degree | 147 | 54.4 |
| Incomplete master’s degree | 25 | 9.3 |
| Completed master’s degree | 33 | 12.2 |
| Incomplete doctorate | 13 | 4.8 |
| Completed doctorate | 8 | 3.0 |
| Self-reported psychiatric diagnosis | ||
| Yes | 100 | 37.0 |
| No | 170 | 63.0 |
Note. *a Brazilian term that indicates mixed race or medium complexion
INSTRUMENTS
External and Internal Shame Scale (EISS) (Ferreira et al., 2022): the EISS was developed to assess trait shame through eight items, which are answered on a five-point scale, where 0 = never and 4 = always. In its original version, created by Ferreira et al. (2022), two dimensions were proposed: internal shame (when devaluation of the self comes from the own person) and external shame (when the self is understood as inferior, inadequate, and rejectable by others). For this study, the version in the process of adaptation to the Brazilian context was used. The EISS was adapted from Portuguese from Portugal to Brazilian Portuguese, following the guidelines of the International Test Commission (2017) and the recommendations proposed by Borsa et al. (2012). Two translators carried out the translation of the items and instructions into Brazilian Portuguese, which were then evaluated by three judges and a sample (n = 30) of the target population. The back-translation stage was not conducted for this scale, since there were not two different languages, but two variations of the same language, namely, Brazilian Portuguese and Portuguese of Portugal. Finally, the instrument resulting from the evaluation for the target population was sent directly to the corresponding author of the EISS construction article so that it could be evaluated by the creators of the original instrument. These procedures yielded adequate results.
CoSS - 5 (CoSS-5; Elison et al., 2006): the CoSS-5 is a 58-item self-report instrument that assesses five shame coping styles: four dysfunctional (withdrawal, avoidance, attack self, and attack other) and one adaptive. To assess the dysfunctional styles, twelve shame-inducing situations are presented with one response representing each of the four styles, the participant must choose the alternative that best represents them. The alternatives corresponding to the functional style are presented separately in the last ten items of the scale. Respondents must indicate how often they would react to the item on a five-point scale ranging from 1 = never to 5 = almost always. Similar to the EISS, this study used the version of the CoSS-5 that was being translated into Brazilian Portuguese. Thus, the International Test Commission (2017) guidelines and the recommendations proposed by Borsa et al. (2012) were followed, as well as the steps described for the EISS and an inversed translation of the version generated by the evaluation of the target population from Brazilian Portuguese to English, carried out by two Brazilian translators with knowledge of the foreign language so that the instrument could be presented to the original author. These procedures showed adequate results.
DATA COLLECTION
The project was approved by the Research Ethics Committee of the Federal University of Minas Gerais (CAEE: 82465424.0.0000.5159; Approval number: 7.171.572). To recruit participants, a snowball sampling method was utilized alongside an online outreach strategy for the dissemination of the online protocol. A digital image was developed with an invitation for voluntary participation, also containing key information about the study. This image included a hyperlink that redirected the participant to the online survey, which was hosted on Google Forms. The first page of the survey presented an electronic Informed Consent Form, detailing the study’s purpose, procedures, and ethical considerations. All respondents had to confirm their acceptance to the informed consent form in order to have access to the other parts of the survey. Recruitment materials were disseminated via the researchers’ professional and academic networks using mainly social media platforms, but also through the official communication channels of the University of Brasília (Brasília, Federal District of Brazil) and the Federal University of Minas Gerais (Belo Horizonte, Minas Gerais, Brazil). Data were collected between December 2024 and January 2025, using an online protocol created on Google Forms. In addition, as a strategy to control the number of careless responses, four control items were included in the protocol that required a specific response (e.g., mark the number five and move on). Hence, participants who marked at least one response option other than the required content were excluded from the analysis. As a direct benefit, all participants had access to an explanatory video about shame at the end of the protocol.
DATA ANALYSIS
All analyses were conducted in R software (R Core Team, 2025) version 4.3.3. Initially, careless responses (n = 42) identified from control items with content requiring a specific response were excluded. Then, the internal structure (H1a and H1b) of the EISS and CoSS-5 instruments was investigated through confirmatory factor analysis (CFA) using the lavaan package (Rosseel, 2012) and the Weighted Least Squares Mean and Variance-Adjusted (WLSMV) estimator. Items in both instruments (EVEI and CoSS-5) were treated as ordinal. The adequacy of the CFA models were evaluated using the Comparative Fit Index (CFI), Tucker Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and Standardized Root Mean Square Residual (SRMR) indices. CFI and TLI values above 0.90, SRMR values below 0.08, and RMSEA values below 0.08 and with the upper confidence interval ≤ 0.10 were considered indicative of the adequacy of the models. For the EISS, a model composed of two factors (internal shame and external shame) was initially tested. The results showed high covariance between the two dimensions (0.914). Therefore, a model composed of a single factor was investigated. Regarding the CoSS-5, the model was initially investigated according to the original proposal, consisting of five factors (withdrawal, avoidance, attack self, attack others, and adaptive). Since the model showed unsatisfactory indices, an exploratory factor analysis (EFA) with oblimin rotation and a polychoric correlation matrix was performed, using the EFA.dimensions package (O’Connor, 2023). In addition, the results of the eigenvalues greater than one, the empirical Kaiser criterion, the Hull method test, and parallel analysis were used to select the factor retention. Thus, according to the results of the EFA, a modified structure for the CoSS-5 was investigated through CFA with the exclusion of some items and composed of three factors (Internalization, Attack Others, and Adaptive). The reliability of the instruments (H2a and H2b) was investigated using Cronbach’s alpha (Cronbach, 1951) and Composite Reliability (CR; Valentini & Damásio, 2016).
To investigate the associations between Trait shame and shame coping styles (H3), Spearman’s correlations between the dimensions of the EISS and the CoSS-5 were initially verified using the psych package (Revelle, 2024). Subsequently, a Structural Equation Modeling model (SEM) was carried out using the lavaan (Rosseel, 2012) package and the WLSMV estimator. In the model, the EISS scores (trait shame) were inserted as predictors of the CoSS-5 scores (shame coping styles). The adequacy of the SEM model was investigated using CFI, TLI, and RMSEA indices, with the same cut-off points indicated for the CFA. Finally, comparisons in trait shame levels and shame coping styles between people with and without a self-reported psychiatric diagnostic (H4) were investigated using the Mann-Whitney test with effect size calculation through bootstrapping (95% confidence interval), using the rstatix package (Kassambara, 2021).
RESULTS
The CFA fit indices indicated adequacy for both models proposed for EISS, composed of two factors (External Shame and Internal Shame) [χ2 = 30.76, df = 15, p = 0.009, CFI = 0.992, TLI = 0.986, RMSEA = 0.062 (90% CI = 0.030 - 0.094), SRMR = 0.034], and of a general factor [χ2 = 39.29, df = 16, p < 0.001, CFI = 0.989, TLI = 0.980, RMSEA = 0.074 (90% CI = 0.045 - 0.103), SRMR = 0.039]. However, considering the high covariance between the two dimensions (0.914), we decided to use the model composed of a general factor (trait shame), partially confirming with H1a. For the CoSS-5, the five-factor model as originally proposed showed unsatisfactory indices [χ2 = 3863.90, df = 1585, p < 0.001, CFI = 0.808, TLI = 0.800, RMSEA = 0.073 (90% CI = 0.070 - 0.076), SRMR = 0.107], not confirming H1b. In this model, nine of twelve items belonging to the Avoidance factor had low factor loadings (< 0.30). Additionally, these items exhibited weak correlations with each other. Considering that the exclusion of these items would substantially change the avoidance factor, it was removed from subsequent analyses. An EFA was carried out with the remaining CoSS-5 items. The tests for factor retention indicated the extraction of three or four factors. In the five-factor model, a high level of covariance was observed between the withdrawal and attack self-dimensions. Therefore, these two dimensions were grouped into one internalization factor. Another CFA model for the CoSS was tested, considering three factors: internalization, attack others, and adaptive. Two items with low factor loadings (< 0.30) were removed. Thus, the final CoSS-5 model showed good fit indices [χ2 = 1924.12, df = 896, p < 0.001, CFI = 0.907, TLI = 0.902, RMSEA = 0.065 (90% CI = 0.061 - 0.069), SRMR = 0.082]. Regarding the factor loadings, for the EISS, the values were all above 0.55 (Table 2), while for the CoSS-5 the values were above 0.42 (Table 3). In addition, the instruments showed good reliability indices (see Table 2 and 3).
Table 2 Factor Loadings, Reliability, and Descriptive Statistics of the External and Internal Shame Scale
| Items | Mean | SD | Minimum and maximum | Skewness | Kurtosis | Loadings |
|---|---|---|---|---|---|---|
| Item 1 | 1.37 | 0.95 | 0 - 4 | 0.38 | -0.34 | 0.671 |
| Item 2 | 1.54 | 1.08 | 0 - 4 | 0.20 | -0.85 | 0.686 |
| Item 3 | 1.84 | 0.86 | 0 - 4 | 0.03 | -0.35 | 0.557 |
| Item 4 | 1.28 | 1.12 | 0 - 4 | 0.44 | -0.74 | 0.816 |
| Item 5 | 1.94 | 0.93 | 0 - 4 | 0.47 | -0.19 | 0.648 |
| Item 6 | 1.54 | 1.05 | 0 - 4 | 0.29 | -0.53 | 0.744 |
| Item 7 | 0.78 | 1.01 | 0 - 4 | 1.15 | 0.54 | 0.794 |
| Item 8 | 2.61 | 1.06 | 0 - 4 | -0.19 | -0.82 | 0.690 |
| Cronbach's α | 0.86 | |||||
| Composite Reliability | 0.89 |
Table 3 Factor Loadings, Reliability and Descriptive Statistics of the Compass of Shame Scale - 5
| Items | Mean | SD | Minimum and maximum | Skewness | Kurtosis | Factor Loadings | ||
|---|---|---|---|---|---|---|---|---|
| Internalization | Attack Other | Adaptive | ||||||
| Item 2 | 2.85 | 1.03 | 1 - 5 | 0.14 | -0.44 | 0.525 | ||
| Item 3 | 3.00 | 1.09 | 1 - 5 | -0.07 | -0.57 | 0.578 | ||
| Item 5 | 3.63 | 1.17 | 1 - 5 | -0.53 | -0.52 | 0.768 | ||
| Item 6 | 3.05 | 1.23 | 1 - 5 | 0.07 | -0.91 | 0.512 | ||
| Item 9 | 3.55 | 1.11 | 1 - 5 | -0.28 | -0.82 | 0.774 | ||
| Item 12 | 3.33 | 1.12 | 1 - 5 | -0.17 | -0.57 | 0.764 | ||
| Item 15 | 3.34 | 1.27 | 1 - 5 | -0.39 | -0.85 | 0.740 | ||
| Item 16 | 2.87 | 1.22 | 1 - 5 | -0.01 | -0.94 | 0.710 | ||
| Item 18 | 3.77 | 1.08 | 1 - 5 | -0.62 | -0.24 | 0.639 | ||
| Item 19 | 3.51 | 1.22 | 1 - 5 | -0.40 | -0.79 | 0.658 | ||
| Item 21 | 3.11 | 1.25 | 1 - 5 | -0.03 | -0.92 | 0.734 | ||
| Item 22 | 3.38 | 1.16 | 1 - 5 | -0.32 | -0.65 | 0.774 | ||
| Item 25 | 3.38 | 1.27 | 1 - 5 | -0.37 | -0.82 | 0.781 | ||
| Item 28 | 3.11 | 1.26 | 1 - 5 | 0.03 | -1.00 | 0.760 | ||
| Item 31 | 3.31 | 1.28 | 1 - 5 | -0.29 | -0.93 | 0.775 | ||
| Item 32 | 3.26 | 1.02 | 1 - 5 | -0.12 | -0.43 | 0.562 | ||
| Item 34 | 3.51 | 1.30 | 1 - 5 | -0.44 | -0.96 | 0.784 | ||
| Item 35 | 3.53 | 1.04 | 1 - 5 | -0.21 | -0.63 | 0.422 | ||
| Item 37 | 3.34 | 1.21 | 1 - 5 | -0.16 | -0.84 | 0.798 | ||
| Item 38 | 3.58 | 1.23 | 1 - 5 | -0.45 | -0.75 | 0.810 | ||
| Item 41 | 3.62 | 1.26 | 1 - 5 | -0.60 | -0.65 | 0.758 | ||
| Item 44 | 3.00 | 1.38 | 1 - 5 | 0.10 | -1.21 | 0.805 | ||
| Item 47 | 2.54 | 1.33 | 1 - 5 | 0.39 | -1.10 | 0.770 | ||
| Item 48 | 3.73 | 1.18 | 1 - 5 | -0.56 | -0.63 | 0.747 | ||
| Item 4 | 2.67 | 1.09 | 1 - 5 | 0.18 | -0.68 | 0.653 | ||
| Item 7 | 2.39 | 1.06 | 1 - 5 | 0.56 | -0.19 | 0.710 | ||
| Item 10 | 1.84 | 0.83 | 1 - 5 | 0.85 | 0.65 | 0.589 | ||
| Item 13 | 1.59 | 0.87 | 1 - 5 | 1.45 | 1.58 | 0.680 | ||
| Item 20 | 2.90 | 1.16 | 1 - 5 | 0.21 | -0.71 | 0.709 | ||
| Item 23 | 2.31 | 1.06 | 1 - 5 | 0.54 | -0.30 | 0.640 | ||
| Item 26 | 2.28 | 0.97 | 1 - 5 | 0.31 | -0.56 | 0.666 | ||
| Item 29 | 1.72 | 0.96 | 1 - 5 | 1.25 | 0.94 | 0.507 | ||
| Item 36 | 2.65 | 1.14 | 1 - 5 | 0.30 | -0.56 | 0.669 | ||
| Item 39 | 2.13 | 0.89 | 1 - 5 | 0.58 | 0.28 | 0.569 | ||
| Item 42 | 3.34 | 1.25 | 1 - 5 | -0.29 | -0.91 | 0.670 | ||
| Item 45 | 1.91 | 0.86 | 1 - 5 | 0.66 | -0.11 | 0.591 | ||
| Item 49 | 2.05 | 0.81 | 1 - 5 | 0.57 | 0.58 | 0.538 | ||
| Item 50 | 2.35 | 0.95 | 1 - 5 | 0.20 | -0.65 | 0.677 | ||
| Item 52 | 2.91 | 1.11 | 1 - 5 | -0.08 | -0.87 | 0.638 | ||
| Item 53 | 2.45 | 1.02 | 1 - 5 | 0.30 | -0.59 | 0.474 | ||
| Item 55 | 3.18 | 1.07 | 1 - 5 | -0.24 | -0.50 | 0.487 | ||
| Item 56 | 2.35 | 0.94 | 1 - 5 | 0.41 | 0.09 | 0.960 | ||
| Item 57 | 2.51 | 1.07 | 1 - 5 | 0.35 | -0.43 | 0.958 | ||
| Item 58 | 1.91 | 0.89 | 1 - 5 | 1.05 | 1.45 | 0.792 | ||
| Cronbach's α | 0.95 | 0.86 | 0.75 | |||||
| Composite Reliability | 0.96 | 0.89 | 0.89 | |||||
The factors of the CoSS-5 and EISS scales showed significant associations with each other (all p < 0.001). The trait shame dimension showed a positive correlation with internalization (ρ = 0.72) and attack other (ρ = 0.42), and a negative correlation with adaptive (ρ = -0.53), confirming H3. The adaptive dimension showed a negative correlation with internalization (ρ = -0.55) and attack other (ρ = -0.44). Finally, internalization showed a positive correlation with attack other (ρ = 0.53).
The SEM model, which aimed to investigate the predictive effects of trait shame on the dimensions of shame coping styles (see Figure 1) showed adequate fit indices (χ2 = 2406.19, df = 1265, p < 0.001, CFI = 0.914, TLI = 0.910, RMSEA = 0.058 [90% CI = 0.054 - 0.061], SRMR = 0.078). Trait shame was a significant (p < 0.001) positive predictor of internalization and attack other, as well as a negative predictor of the adaptive domain, confirming H3. The coefficient of determination (R2) indicated that trait shame explained 67.8% of the variance of Internalization, 20.4% of Attack Other, and 33.3% of Adaptive.
Comparisons between individuals with and without a self-reported psychiatric diagnostic indicated significant differences for all the trait shame and shame coping styles dimensions, confirming H4. The group of individuals with a diagnostic had higher scores in trait shame, internalization, and attack other, and lower scores in adaptive. Effect sizes ranged from small (r = 0.24) to moderate (r = 0.39) (see Table 4).
Table 4 Comparisons Between Groups
| No diagnostic | With diagnostic | Mann-Whitney | Effect size | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Variable | M | SD | Rank mean | M | SD | Rank mean | U | Z | p | r | r 95% CI |
| Trait shame | -0.16 | 0.57 | 115.46 | 0.29 | 0.63 | 169.54 | 5096.5 | -5.49 | < 0.001 | 0.33 | 0.21 - 0.44 |
| Internalization | -0.16 | 0.45 | 111.96 | 0.27 | 0.49 | 175.52 | 4498.0 | -6.46 | < 0.001 | 0.39 | 0.29 - 0.49 |
| Attack Other | -0.12 | 0.59 | 120.87 | 0.19 | 0.59 | 160.37 | 6013.0 | -4.01 | < 0.001 | 0.24 | 0.13 - 0.35 |
| Adaptive | 0.11 | 0.46 | 152.77 | -0.18 | 0.48 | 106.14 | 11436.0 | -4.74 | < 0.001 | 0.29 | 0.16 - 0.39 |
DISCUSSION
The objectives of this study were to a) investigate the internal structure of the CoSS-5 and EISS scales when adapted for Brazilian Portuguese (H1a and H1b), b) examine the reliability of CoSS-5 and EISS scores considering internal consistency (H2a and H2b), c) explore the relationships between trait shame and shame coping styles in a Brazilian sample (H3), and d) investigate differences in trait shame levels and shame coping strategies between individuals with and without a self-reported psychiatric diagnosis (H4).
Regarding H1a and H1b, results partially confirmed the replication of the same internal structure of the original scales in the Brazilian population. After the necessary changes were made, both the EISS and CoSS-5 showed good fit indices. In addition, adequate reliability indices were observed for the structural versions of the instruments used here, confirming H2a and H2b, indicating the possibility of using the instrument with Brazilian adults. With respect to the EISS scale (H1a), although the bidimensional model of internal and external shame showed good fit indices, the high covariance between the two factors favored a one-factor model of shame. In this sense, it is important to notice that results somewhat differed from the original study (Ferreira et al., 2022), in which a high-order factor, namely global shame, coexisted with two lower order factors (external and internal shame). In our case, it seems the distinction between the specific dimensions of internal and external shame, although clinically and theoretically relevant (Gilbert, 2022), is rather unnecessary for Brazilians. However, our study is closer to the pattern found in the Arabic adaptation of the EISS, in which a unidimensional model was found (Fekih-Romdhane et al., 2023). More directly, results have shown that in the Brazilian context the presence of one facet of trait shame is almost always accompanied by the other. This aligns with the theories that explain the emergence of internal shame as an outcome of a continuous exposure to episodes of external shame (Gilbert, 2022), in the sense that negative judgment arising from others becomes “internalized” and then is enacted by the individual.
With respect to the internal structure of the CoSS-5 (H1b), our results were not able to replicate the structural validity found in Capinha et al. (2021). Instead of favoring a 5-factor structure (not confirmed by CFA) through EFA, our results have shown a trifactorial model. The attack other and adaptive dimensions were maintained from the original version, but the attack self and withdrawal dimensions were merged into a unique dimension which we named “internalization” based off the Elison et al. (2006) and Schalkwijk et al. (2016) studies. In the present research, the avoidance dimension was excluded from the final version of the scale, since its items had unacceptably low factor loadings. This result may be attributed to the fact that avoidance, as described by Elison et al. (2006), is a coping style characterized by very low awareness of shame. Consequently, participants might not have associated avoidant reactions with their experiences of shame, either because the behavior described in the item reveals a broad tendency towards many emotions (not just shame) or because they were unaware of their shame or unwilling to recognize it.
The result obtained in this research is a clear alteration of the original model of the scale, even though the replication of the five-factor structure was also not found in the studies of Schalkwijk et al. (2016) and Martin et al. (2021). In addition, when Akıncı and Atintaş (2020), adapted the original four-factor Compass of Shame Scale, they initially conducted an EFA and found 11 factors with eigenvalues greater than one before reaching 4 factors through higher eigenvalues and screen plot test, also having to remove items because of low correlations with their own factors. In this sense, the result that was found in the Brazilian context is not necessarily an aberration, but it might demonstrate some limits of the theoretical model on which the CoSS-5 is grounded. Given that the CoSS-5 is a scenario-based instrument designed to assess reactions to the state of shame, it is important to note that states are significantly more influenced by context than traits. Therefore, in the Brazilian cultural context, internalization (withdrawal and attack self), attack others and adaptive might sufficiently describe coping styles for states of shame, whereas avoidance could pertain to a wider relation with many different emotional states beyond shame.
The results of H1a and H1b should not necessarily lead to an interpretation that the models behind the EISS and CoSS-5 are flawed or that the scales themselves are poorly constructed. In fact, what Lear et al. (2022) concluded from a systematic review on measures of generalized shame is that reliability and validity of scales that aim to assess this construct is very limited, particularly in relation to measures of global negative self-concept. In this sense, the problems of the EISS and CoSS-5 scales might not be specific to their theoretical models but may actually be related to the very construct of shame. Additionally, Lear et al. (2022) conclude their review suggesting that the value of the shame construct would reside in measures that assess shame in a contextualized manner and focus on responses to shame. In this sense, the CoSS-5, a measure of shame coping styles, should pose an advantage to EISS for measuring shame. In the case of this study, however, both scales did not reveal a factor structure corresponding to their conceptualizations of shame. This should also not come as a surprise since the experience of shame differs highly between cultures (Bierbrauer, 1992; Casimir & Schnegg, 2002; Wong & Tsai, 2007). Thus, instead of searching for a universal replication of the theoretical models behind shame measures, research could take a culturally sensitive approach that aims to capture the expression of shame in different countries.
Furthermore, with respect to the relations between trait shame and shame coping styles (H3), it was found, as expected, that the global trait shame dimension of the EISS significantly both correlated with and predicted the internalization, attack other and adaptive dimensions of the CoSS-5 version of the scale obtained in our study, confirming a pattern of association between the Compass of Shame poles and trait shame (Akıncı & Atintaş, 2020). One possible explanation for the high predictive power of the EISS factor over the internalization dimension of the CoSS-5 is that a global negative devaluation of the self either from the self or others (trait shame) favors responses of attacking the self and withdrawing when facing a situation in which the self might be perceived as inferior or inadequate (state shame). Also, considering the self to be flawed and worthless (trait shame) could lead to a disposition for the need to be aggressive towards others. This can be especially true towards those who might generate a perception of the self that is defective in state shame evoking situations, maybe using aggression as a mechanism to not increase the distress caused by trait shame by further reinforcing it. Lastly, experiencing the self as globally unworthy could hinder the use of adaptative strategies to regulate shame, potentially through damaging the mental health resources of the individual, since trait shame is highly correlated with psychopathology (Cunha et al., 2021; Ferreira et al., 2022).
On this matter, confirming H4, our results aligned with previous studies by showing that individuals with a self-reported psychiatric diagnosis had significantly higher levels of trait shame, internalization, attack other coping styles and lower levels of the adaptive shame coping style than individuals with no self-reported diagnosis (Capinha et al., 2021; Cunha et al., 2021; Fekih-Romdhane et al., 2023; Ferreira et al., 2022; Hiramatsu et al., 2021; Rapinda et al., 2022). This further confirms the clinical relevance of the construct as a transdiagnostic factor. When clinicians are working with psychopathology, it might be useful to assess and potentially address levels of trait shame and the coping styles of state shame of individuals.
Regarding the implications of the present study for the practice of clinicians, since the internal and external dimensions of the EISS have a high covariance, clinicians should not assume that internal and external shame exist separately and, when faced by either internal presentation of shame by clients, should also address external shame and vice-versa. Furthermore, when working with clients who have high dysfunctional reactions to state shame, such as internalizing (i.e., attacking the self, withdrawing from the situation) and attacking others, more than simply teaching adaptative strategies of how to manage the emotion of shame, clinicians should also address a possible underlying trait shame in such individuals. Also given that shame was higher in individuals with self-reported psychiatric diagnoses, clinicians might target underlying shame through psychological interventions which have been shown to be effective to target such difficulties (Goffnett et al., 2020; Norder et al., 2022), especially cognitive-behavioral (Diekmann et al., 2023) and compassion-based treatments (Matos & Steindl, 2020; Germer, 2025).
Regarding the limitations of this study, generalizability of findings might be restricted by a sample comprised mostly of women, white individuals, and persons with a high income and educational level, a group that is not representative of the majority of the Brazilian population. Additionally, even though the literature on factor analysis does indicate adequacy for some models with smaller samples, depending on construct complexity and instrument quality (Kyriazos, 2018), the relatively small sample size of this study could have affected the unacceptable fit indices of the CoSS-5 scale, since the item-participant ratio was low. Therefore, further studies with larger samples are required to verify the replication of results so that the adequacy of this instrument in the Brazilian population can be validated. Also, the study has a correlational nature, so the associations between the variables of trait shame and shame coping styles might be bidirectional. In this sense, different from what was discussed in the previous paragraphs, shame coping styles might have a causal explanatory power over trait shame, a hypothesis that this study cannot disconfirm. Additionally, psychiatric diagnoses were self-reported, meaning the validity of the presence or absence of mental health symptoms in both the clinical and non-clinical groups cannot be confirmed.
Finally, a few suggestions for future research are provided. First, other validation studies of both the EISS and CoSS-5 are needed in the Brazilian context to further determine that the scale is appropriate, such as discriminant and convergent validity studies. In this sense, it is important that the associations of both the EISS and CoSS-5 scales with other shame measures, such as the Brazilian adaptation of the Other as Shamer Scale - 2 (Santos et al., 2015), are verified, as well as their relation with the construct of self-esteem to verify construct validity in Brazilian samples, given the discourse made by Lear et al. (2022). Additionally, other studies need to be conducted to verify the invariance of these scales in samples of men, non-white, low-income and low-education Brazilians, individuals which were underrepresented in this sample. It would also be especially adequate in these future studies of validation of the scales to include measures of psychopathology to have actual, objective measures of mental health symptoms and then make more valid comparisons with more accurate results. Ultimately, to establish a causal relationship between trait shame coping styles, longitudinal studies are essential.
CONCLUSION
The present study provided evidence of validity for the EISS and CoSS-5 scales when adapted to the Brazilian context. Changes in their factorial structure differing from their original proposals had to be made in order to reach adequate fit indices for the instruments’ internal structures. In this study, a one-factor structure was adopted for EISS, representing general shame, instead of maintaining the original division between external and internal shame given the high covariance between the two dimensions. In the case of CoSS-5, a tri-factor study was found in which two of the original factors were merged into one and another factor was excluded due to unacceptably low factor loadings, leading to a final version of the scale containing three coping styles of shame: internalization, attack others, and adaptive. Regarding reliability, overall, acceptable values of alpha and composite reliability were found. Additionally, relations between the variables of EISS and CoSS-5 were found, with significant predictive effects of trait shame on shame coping styles indicators. Finally, results showed that individuals who reported having a psychiatric diagnosis exhibited higher levels of shame compared to those without a diagnosis, indicating the construct’s clinical relevance. Thus, it is expected that the findings of this study will provide empirical support for researchers and professionals in using the Brazilian versions of the EISS and CoSS-5 scales for assessing shame.
