In order to select the best items for the initial brief version of the CSS, Exploratory Factor Analyses (EFA) were performed in a database of participants who responded to the CSS (Noronha & Barbosa, 2016). In the first EFA, three items (3, 11 and 12) had factor loadings with values below 0.30. The three items were removed and a new EFA was performed. The results (RMSEA = 0.070; χ²/df = 51558.770/2210; CFI = 0.717; TLI = 0.708) indicated unsatisfactory indices and factorial loads with values between 0.30 and 0.64.

Subsequently, the 48 items of the first version of the CSS-Brief were selected based on the factor loadings and theoretical relevance, in light of the precepts suggested by Peterson and Seligman (2004), so that each strength had two items that represented it. The selected items should be those with higher factorial loads, with the exception of items from the strengths wisdom (4 and 6), spirituality (28 and 51) and hope (27 and 49), which presented equal or close factorial loads.

Initially, CFA were performed with nine factorial structures for the character strengths previously identified in the literature, as shown in Table 1, in order to verify whether they would present adequate fit indices in the CSS-Brief (48 items). The data found are exhibited in Table 2.

According to the exposition made in Table 2, the factorial structures tested by the CFA presented satisfactory fit indices. In this connection, it was decided to identify whether the CSS-Brief factorial structure was factorable and which retention methods were suggested. The KMO (0.954) and Bartlett’s sphericity test (22023.4; df = 1128; p = 0.00001) presented indices considered very good, indicating the possibility of factoring the scale. Hull’s method suggested retention of one factor, while MAP and parallel analysis suggested retention of up to three factors.

Then, the ESEM were conducted in order to reduce the number of CSS-Brief items and identify the structure that would present the best fit indices. Solutions with first and second order factors were tested. Only two factorial structures, with one and two first-order factors, showed theoretical relevance after excluding items that loaded below 0.50. Although the structure with three factors showed good psychometric indices (χ²/df = 70.049/25; p = 0.0001; RMSEA = 0.04; CFI = 0.99; TLI = 0.98), one of the factors remained with only two items and one of the correlations between the factors was low (r = 0.09). After carrying out the analyses and taking into account the suggestions of the Model Modification Indices, the fit indices found in the structure of two first order factors (χ²/df = 669.748/118; p = 0.0001; RMSEA = 0.07; CFI = 0.953; TLI = 0.939) were better than those of the unifactorial structure (χ²/df = 2149.549/209; p = 0.0001; RMSEA = 0.10; CFI = 0.86; TLI = 0.85), with loads greater than 0.50.

In the first factor (α = 0.83), Intrapersonal strengths, six items were grouped (hope, gratitude, spirituality, appreciation of beauty, love of learning and vitality) that relate to experiences that translate a positive orientation towards the future, with energy and appreciation for beauty and learning. The second factor (α = 0.87), intellectual and interpersonal strengths, grouped the strength items emotional intelligence, critical thinking, creativity, authenticity, teamwork, bravery, modesty, leadership, common sense, humor, prudence and impartiality. The 12 items refer to the cognitive strengths that help coping with difficulties, solving problems and establishing interpersonal relationships, in order to create healthy social coexistence settings. The strengths love, forgiveness, self-regulation, kindness, persistence and curiosity had no items that represented them in the structure of 18 items and two factors. Then, the DIF analyses were performed considering the structure of two first-order factors to identify the need to delete more items. The education variable was coded from 1 to 10, in a scalar fashion, that is, 1 corresponded to the lowest level of education found in the sample (incomplete elementary I grade) and 10 corresponded to the highest (complete postgraduate degree). Similarly, age was also coded in a scalar basis (18 to 73). As for gender, 1 corresponded to the male and 2 to the female. The factorial loads and the DIFs values for this structure are shown in Table 3.

The factor loadings varied between 0.55 and 0.87 and the DIFs had low values for gender, age and education (between 0.001 and 0.200). The gender and education variables showed differences in scores for 12 items; men and participants with higher education level were more favored by the items. In the age variable, 18 items had differences in scores, with the youngest participants being favored by the items.

In the present study, the objective was to select items and identify the psychometric properties of the CSS-Brief; in this case, we searched for evidence of validity based on the internal structure of the CSS-Brief, aiming at reducing the number of items in the original version (EFC; Noronha & Barbosa, 2016), since there are advantages in the application of short versions of psychological instruments. In addition, it would allow us to take into account the limitations related to the methodologies used in factor analysis of character strengths instruments. In the first study, most of the items that had higher factor loadings also had better theoretical representations (Peterson & Seligman, 2004). The exception was the strength items hope (27 and 49), spirituality (28 and 51) and wisdom (4 and 6), which had equal loads and highlighted the importance of the theoretical aspect as a criterion for selecting items to avoid decisions that would culminate in erroneous or unreliable results (Damásio, 2012; Patil et al., 2008). As for hope (hoping for the best and working to achieve it), item 49 was chosen (“I know that things will work out”) to the detriment of 27 (“I believe that tomorrow will be better than today”). The latter could cause problems when the respondent would be on a pleasant day, for example, which could lead him to reflect as follows when reading that statement: “Today was a very good day. There’s no way that tomorrow can be better than today.” In this connection, item 49 seems to better represent the expectation that something in the future will be better (Noronha & Batista, 2020b; Peterson & Seligman, 2004).

On the other hand, with regard to spirituality (having strong and coherent beliefs about the higher purpose and meaning of the universe) and common sense (being able to give wise advice to others), items were selected that had statements that best contemplated the definition suggested by Peterson and Seligman (2004). For spirituality, item 28 was chosen (“I believe in a sacred force that connects us to each other) because it refers to a coherent belief with a greater purpose (Peterson & Seligman, 2004). Item 51 (“Believing in a supreme being gives meaning to my life”) was excluded because item 8 (“I feel that my life has a greater meaning”) had already been selected, and have similar content. Similarly, item 6 (“I make good judgments, even in difficult situations”) of common sense was selected, since it indicated an ability that was not associated only with giving advice, characteristic of people known as being wise (Peterson & Seligman, 2004; Seligman, 2019). Item 4 (“I am competent to give advice”) had a content similar to item 63 (“I can help people understand each other when there is a discussion), which is also why it was excluded.

In study 2, it was hypothesized that the previously identified factorial structures would present unsatisfactory fit rates in the CSS-Brief (Hypothesis 1), which was confirmed. One justification refers to the culture of each country being different. Possibly, the VIA classification by Peterson and Seligman (2004), made after a wide review, considering different texts, cultures and religions, encompasses a wide range of content that can make it difficult to identify the same structure in different cultures, causing theoretical divergences (Allan, 2014; Giuliani et al., 2020; Seligman, 2019) and different empirical structures (Ng et al., 2016; Noronha & Batista, 2020a; Solano & Cosentino, 2018). In this connection, some authors have sought to better understand the structure character strengths (Allan, 2014; Ng et al., 2016; Noronha & Batista, 2020a). As there is a need to make more contextualized assessments, in this case, considering the strengths of character in the Brazilian scenario, in the present work other analyses were carried out (EFA, MIMIC, ESEM) to identify the construct structure in Brazil. Thus, potential biases not considered before in the assessment of the construct could be minimized (Heintz et al., 2017).

In hypothesis 2, it was expected that the DIF values for the gender, age and education variables would be less than 0.30, which would indicate little interference of the variables in the endorsement of the CSS-Brief items. The hypothesis was confirmed, indicating that the CSS-Brief has good items to assess character strengths, with levels of bias considered adequate (Linacre, 2012); this would reduce potential errors arising from the specific characteristics of the participants (Heintz et al., 2017), in this case, age, gender and education. Finally, in previous studies, gender and age did not show statistically significant differences for strengths in Brazil and in other countries (Heintz et al., 2017; Noronha & Batista, 2017; Noronha & Martins, 2016), which seems to indicate that such variables are not intervening for the understanding of the construct.

Even with the exclusion of six items, the structure with two first order factors was the one that presented the best results. However, it is considered that there were no losses with the exclusion of the items of curiosity, persistence, kindness, self-regulation, forgiveness and love, since a single strength could correspond to several virtues, besides having similar content among them (Giuliani et al., 2020; Peterson & Seligman, 2004; Ruch & Proyer, 2015). For example, love of learning (mastering new skills and knowledge) seems to encompass, to some extent, curiosity (being interested in the whole experience). Likewise, bravery (not hiding from threat, challenge or pain) includes persistence (finish what you started) and self-regulation (keep your spirits up even in the face of difficult situations) (Peterson & Seligman, 2004; Noronha & Batista, 2020a; Noronha et al., 2015).

Possibly, the methodologies used in the present study were more adequate than those used in the studies reported in Table 1. Some of these studies used the principal component analysis, which tend to inflate explained variances and present higher factor loadings and commonalities when compared to exploratory factor analysis (Costello & Osbourne, 2005; Damásio, 2012). The use of principal component analysis is not recommended when the objective is to observe the existing interrelationship in a latent construct’s set of items (Costello & Osbourne, 2005), in the case of the present study, character strengths. As for the Kaiser-Guttman criterion, its results tend to lack precision, with factors overestimation; hence it is not recommended (Costello & Osbourne, 2005; Patil et al., 2008). Finally, orthogonal rotations were also used, which assumes the lack of correlation (r = 0) among the factors. This assumption tends to leave the data obtained in psychology research incoherent (Costello & Osbourne, 2005), since behaviors, feelings, symptoms and other human realities work interdependently, having relationships among them (Damásio, 2012; Goretzko et al., 2021).

Thus, in addition to there being theoretical approximations in the clusters found for the factors intrapersonal strengths and intellectual and interpersonal strengths (Azañedo et al., 2014; Littman-Ovadia & Lavy, 2012; McGrath, 2014; Noronha & Batista, 2020a; Peterson & Seligman, 2004; Solano & Cosentino, 2018), some limitations related to methodologies, to sociodemographic characteristics and to culture were minimized in the present study. In this connection, it is considered that there were advances in the study of character strengths in the Brazilian scenario. As limitations of this study, we consider the fact that items to control acquiescence and social desirability were not developed. However, it is noteworthy that in the present work the initial psychometric studies of the CSS-Brief were carried out, and the objectives were attained. The structure identified for the CSS-Brief was also different from the VIA classification (Peterson & Seligman, 2004), although it showed some similarities. The factors found have theoretical cohesion, with the structure of two first order factors being the most adequate to evaluate the strengths in the Brazilian scenario. Furthermore, we consider a limitation the fact that most of the sample is composed of participants who had higher education, which does not include an important portion of the general Brazilian sample. In future studies, it is suggested that bias controls, in addition to more in-depth methods on the Item Response Theory, be also performed so that the CSS-Brief presents less impact of scores from different sources and that participants with lower educational levels be also included in the sample.