Introduction

Body image is a multifaceted construct that integrates components related to physical appearance, such as the mental representation of one’s own size, shape, and facial features, and attitudes about appearance [1]. The attitudinal component of body image comprises two dimensions: one that assesses body satisfaction and beliefs about the body—namely, Evaluation—and another regarding the cognitive, behavioral, and emotional importance that one places on his/her own body—namely, Cognitive-Behavioral Orientation. Most studies worldwide have used body satisfaction as the sole measure of assessment, but other attitudinal components are often overlooked [2]. To minimize errors, the use of multiple measures is recommended, with special attention on achieving representation of a wide range of dimensions [3].

In this regard, one of the most comprehensive body image measures is the Multidimensional Body-Self Relations Questionnaire (MBSRQ) [4, 5]. The original instrument, the Body Self-Relations Questionnaire (BSRQ), was composed of 294 items. Later, versions eliminated/replaced some items, and a 54-item version was tested by Brown et al. [4]. The principal component analysis showed the instrument has seven factors, reflecting the two attitudinal dimensions—Evaluation and Orientation—in relation to each of three somatic domains: appearance, health, and fitness. The BSRQ subscales are Appearance Evaluation, Appearance Orientation, Fitness Evaluation, Fitness Orientation, Health Evaluation, Health Orientation, and Illness Orientation. In addition to its seven factors, three appearance-related scales were added later to create the Multidimensional Body-Self Relations Questionnaire (MBSRQ) [5]: the Body Areas Satisfaction Scale, the Overweight Preoccupation Scale, and the Self-Classified Weight Scale. The MBSRQ is a 10-factor scale, with 69 items, constituting the 54-item BSRQ plus the three additional scales.

The MBSRQ has been validated in Spain [6], Mexico [7, 8], Poland [9] and Chile [10]. The Spanish MBSRQ [6] was tested in 131 university students without body image problems (Mage = 23.6, SD = 2.9 years) and 130 cosmetic surgery patients (Mage = 40.3, SD = 10.2 years). The questionnaire was reduced to 45 items distributed across 4 factors instead of 10. The first Mexican MBSRQ [7] was tested in 232 college men ranging in age from 17 to 32 years. The measure was reduced to 61 items distributed across 10 factors. The second Mexican version [8], was tested in 1539 university students with a mean age of 20.56 years (SD = 1.88). The measure was reduced to 14 items distributed in 2 factors (subjective importance of the physical appearance and subjective importance of the physical shape). The Polish MBSRQ [9] was tested in 341 females ranging in age from 18 to 35 years. The measure was reduced to 67 items and consisted of 8 subscales. Finally, the Chilean MBSRQ [10] was tested in a nonclinical sample of 254 women and 197 men aged between 15 and 25 years. The Castilian version of the questionnaire was used [6], and the measure was reduced to 66 items distributed across 7 factors.

The MBSRQ has a short version, the Multidimensional Body-Self Relations Questionnaire-Appearance Scales (MBSRQ-AS) [5]. The instrument has 34 items and assesses body image evaluation and orientation, but with an emphasis on the assessment of appearance-focused body image. It consists of two factors from the original MBSRQ and its three additional subscales. The MBSRQ-AS subscales are: Appearance Evaluation (feelings of physical attractiveness and satisfaction with one’s looks), Appearance Orientation (the extent of investment in one’s appearance), Body Areas Satisfaction Scale (satisfaction with distinct aspects of one’s appearance), Overweight Preoccupation (fat anxiety, weight vigilance, dieting, and eating restraint), and Self-Classified Weight (how one perceives and labels one’s own weight) [5]. The MBSRQ-AS has been validated in Spain [11, 12], Germany [13], Pakistan [14], Malaysia [15], France [16] and Greece [17].

The Spanish MBSRQ-AS [11] was tested in 1041 nonclinical individuals ranging in age from 15 to 46 years. The confirmatory factor analysis showed satisfactory fit indices, confirming that the Spanish version of the MBSRQ-AS has the same five factors as those reported by Cash [5]. This version of the MBSRQ-AS was also tested in early adolescents from 12 to 14 years to confirm the factorial structure [12]. The sample included 355 participants, 189 girls and 166 boys, with ages ranging from 12 to 14 years, and the original MBSRQ-AS 5-factor structure was confirmed. The German MBSRQ-AS [13] was tested in 230 female patients with eating disorders (Mage = 27.36, SD = 6.99 years) and 293 female healthy controls (Mage = 25.69, SD = 6.28 years). The authors could not estimate the model fit for all five subscales due to multicollinearity. Therefore, a confirmatory factor analysis without the subscale Self-Classified Weight was conducted. The analysis demonstrated convincing goodness-of-fit indices, confirming the four-factor structure of the measure. The Pakistani MBSRQ-AS [14] was tested in 850 adolescents and adults. Exploratory and confirmatory factor analyses demonstrated a solution with 27 items retained in 4 factors: Body Area Satisfaction, Appearance Orientation, Appearance Evaluation, and Overweight Preoccupation. The Malaysian MBSRQ-AS [15] was tested in 629 Malaysian Malays (315 women) ranging in age from 18 to 64 years (M = 32.81, SD = 8.65). Exploratory factor analysis indicated that the MBSRQ-AS items reduced to four dimensions, although one factor had less-than-adequate internal consistency. Omitting this factor resulted in a 23-item 3-factor solution, which we tested for fit using confirmatory factor analysis (CFA) alongside the parent 5-factor model. According to the authors, CFA indicated that both models had good fit on some indices, but less-than-ideal fit on other indices, with the 3-factor model showing comparatively better fit.

The French MBSRQ-AS [16] was tested in 772 individuals ranging in age from 18 to 61 years, and only 2 subscales were considered for the factor analysis: Appearance Evaluation and Appearance Orientation. The analyses yielded the same two factors composed of the same items as the original measure. Similarly, the Greek MBSRQ-AS [17] was tested in 1312 high school students ranging in age from 15 to 19 years. An exploratory factor analysis was also conducted only with the Appearance Evaluation and the Appearance Orientation subscales, and results demonstrated that the 2-factor structure of the 19 items was maintained in accordance with the original scale.

Even though not all versions of the MBSRQ and the MBSRQ-AS have shown the same factor structure as the original instrument, all these adaptations presented satisfactory evidence of reliability, stability and several types of validity [9, 11, 13, 16, 17]. Moreover, the MBSRQ and the MBRSQ-AS have been successfully applied to several types of clinical samples (e.g., eating disorders, body dysmorphic disorder, obsessive–compulsive disorder, social anxiety disorder, cancer, polycystic ovary syndrome, and amputees) [18,19,20,21,22,23] and nonclinical samples (e.g., college students, adolescents, midlife individuals and elderly people) [24,25,26,27,28,29].

There is a gap in the Brazilian literature with regard to validated body image tools. A detailed review undertaken by our group about the Brazilian literature on body image [30] demonstrated that the majority of measures available in the country are intended to be used in college students, with half of them evaluating satisfaction/dissatisfaction with the body. The authors noted that females and adolescents of both sexes are studied extensively, whereas other groups are overlooked (e.g., adults, community samples and men), and concluded that the tools that have been developed or adapted in the country do not focus on these ignored groups or other components of body image. This scenario was also noted by other authors [e.g., 31].

Therefore, we think a multidimensional instrument could be of great contribution to Brazilian researchers. Brazil is a useful site to study body image for a number of reasons. For example, a very recent study [32] conducted with 253 girls from first to third grades of high school evaluated the effects of a 6-month school-based intervention, which emphasized body dissatisfaction, unhealthy weight control behaviors and social cognitive-related diet, and physical activity. Results demonstrated an adverse effect of the intervention on unhealthy weight control behaviors and, according to the authors, this negative impact might be because in Brazil there is a high concern for an ideal beauty standard. Additionally, Brazil is the world’s largest consumer of weight loss medications per capita and occupies the second place on the ranking of the world’s top eight countries for cosmetic procedures [33]. For all that, one might argue that the body is as an important form of (physical, symbolic, and social) capital in Brazilian culture [34].

Importantly, the adaptation of an instrument instead of the development of one is advantageous, because cross-cultural studies using the same scale may provide some insight into how body image is perceived in different populations/cultures. Moreover, it also allows for broader discussion of the construct and enables comparison of populations with different sociocultural and economic characteristics [31].

Therefore, the aim of the present study was to translate the MBSRQ-AS into Brazilian Portuguese, to evaluate its factor structure, and to examine the reliability and validity of its scores in adults. The MBSRQ-AS was chosen rather than the MBSRQ, because most researchers are interested only in the appearance-related subscales of the MBSRQ and prefer a shorter questionnaire, without the fitness and health items [5]. Moreover, the questionnaire is widely used in body image studies and it is also able to differentiate between the evaluation of appearance-related aspects and the person’s orientation toward these aspects. Last, the MBSRQ-AS is intended for use with individuals older than 15 years old, and had its norms tested on participants who were 18 years of age or older [5], with no upper age limit. Therefore, the Brazilian version was tested in adults older than 18 years old, respecting the intention of the original measure. The local ethics committees approved the study, and informed consent was obtained from subjects.

First, we translated the MBSRQ-AS into Brazilian Portuguese and investigated its semantic equivalence, content validity and level of verbal comprehension (i.e., to assess the clarity and understanding of each question and instrument in its entirety). We predicted that the Brazilian version should be similar to the original questionnaire in meaning and that the indicators of content validity and verbal comprehension would be satisfactory.

Second, we investigated the factor structure of the Brazilian MBSRQ-AS using confirmatory and exploratory factor analysis, estimated its internal consistency and test–retest reliability using Cronbach’s α coefficient and intraclass correlation tests, respectively, evaluated the association between its factors using Pearson’s correlation test, and analyzed differences between sex and weight categories in a nonclinical sample using separate multivariate analyses of variance (MANOVA), followed by pairwise comparisons with Bonferroni correction. We hypothesized that the MBSRQ-AS items would adhere to a five-factor solution, similar to the original version, via confirmatory factor analysis (CFA). We also expected the reliability indices should be satisfactory as found with other versions worldwide [6, 7, 9,10,11,12,13]. Due to previous findings [11], we predicted that some of the factors would be correlated. Fourth, MBSRQ-AS scores were expected to differ between men and women, as women report more negative body image evaluations and stronger body image investment than men [e.g., 26]. Finally, due to the well-documented link between body image and weight [35,36,37], we hypothesized that MBSRQ-AS scores would differ between overweight/obese individuals and their normal weight counterparts.

Lastly, we estimated the convergent validity of the Brazilian MBSRQ-AS using Pearson’s correlation test. A few hypotheses were raised based on the well-known relations between the subscales from the MBSRQ-AS and several measures of body image and psychosocial functioning [11,12,13, 16, 17]. First, we hypothesized that Appearance Orientation would be related in a positive direction to both subscales from the Appearance Schemas Inventory-Revised (ASI-R), the General Internalization subscale from the Sociocultural Attitudes Toward Appearance Questionnaire-3 (SATAQ-3) and the Eating Attitude Test (EAT-26). Second, the Appearance Evaluation subscale and the Body Areas Satisfaction Scale were expected to relate in a negative direction to both subscales from the ASI-R, the General Internalization subscale from the SATAQ-3 and the EAT-26, and in a positive direction to the Rosenberg Self-Esteem Scale. Third, we expected that the Overweight Preoccupation subscale would be positively related to the General Internalization subscale from the SATAQ-3 and the EAT-26.

Methods

Participants

The semantic equivalence, content validity and level of verbal comprehension of the Brazilian MBSRQ-AS were investigated with a community sample of 89 men (age: M = 30.05; SD = 12.47) and 73 women (age: M = 29.38; SD = 12.25). After receiving Institutional Review Board (IRB) approval from the University of São Paulo, participants were invited to voluntarily participate in the study and not given any type of incentive (material or financial). The mean body mass index (BMI) for male participants was 26.11 kg/m2 (SD = 3.84), whereas that of women was 24.49 kg/m2 (SD = 4.45). The majority of participants were white (men = 66.3%, women = 63%). In terms of highest educational qualification, the majority of participants were educated to a secondary level (i.e., 9–11 years of education) (men = 75.3%, women = 69.6%). No additional demographic information was gathered.

Factor structure, internal consistency, test–retest reliability, association between factors and differences between sex and weight categories were investigated with an online sample of 298 men (age: M = 34.49 years, SD = 12.03) and 308 women (age: M = 34.01 years, SD = 11.44), totaling 606 individuals. After receiving IRB approval from the University of São Paulo, participants were recruited from the SurveyMonkey Contribute member site and were pre-screened based on their sex and age (older than 18 years old). The exclusion criteria used for potential subjects were being younger than 18 years old, having any medical condition that may have a direct or indirect influence on physical appearance (e.g., AIDS, cancer, rheumatic or autoimmune diseases, severe burns) and/or seeking or having had weight loss surgery. Participants did not receive any direct compensation from the researchers, but SurveyMonkey made a US$.50 donation to the charity of their choice along with a chance to win US$100. The mean body mass index (BMI) for male participants was 26.24 kg/m2 (SD = 5.59), whereas that of women was 25.09 kg/m2 (SD = 6.29). Participants’ weight status was computed using standard BMI cutoffs [38] as follows: underweight (4.0% of men and 5.8% of women), normal weight (45.0% of men and 52.6% of women), overweight (30.2% of men and 26.9% of women) and obese (20.8% of men and 14.6% of women). The majority of participants were white (men = 68.1%, women = 68.2%; black or “brown” = 28.1% of men, 28.3% of women). In terms of highest educational qualification, 46.3% of men and 44.5% of women were educated to a secondary level, whereas 48.0% of men and 49.0% of women had a graduate education. No additional demographic information was gathered.

Convergent validity of the Brazilian MBSRQ-AS was investigated with 104 men (age: M = 22.66 years, SD = 4.95) and 133 women (age: M = 20.38, SD = 2.27) recruited from a regional campus, after receiving IRB approval from the University of São Paulo. The exclusion criteria for potential subjects are the same from above. The mean body mass index (BMI) for male participants was 24.99 kg/m2 (SD = 4.00), whereas that of women was 22.10 kg/m2 (SD = 3.86). Participants’ weight status was computed using standard BMI cutoffs [38] as follows: underweight (1.0% of men and 9.0% of women), normal weight (51.9% of men and 74.4% of women), overweight (37.5% of men and 14.3% of women) and obesity (9.7% of men and 2.3% of women). The majority of participants were white (men = 76.0%, women = 85.0%; black or “brown” = 23.1% of men, 14.3% of women). No additional demographic information was gathered.

Measures

Demographic information

Participants self-reported their age, height, weight, race/ethnicity and highest educational level.

Multidimensional Body-Self Relations Questionnaire-Appearance Scales [5]

The MBSRQ-AS has 34 items divided into 4 subscales that assess satisfaction and the affective, cognitive, and behavioral components of body image. The questionnaire is self-administered, and all of the items are evaluated using a 5-point Likert scale. Most items measure agreement (1 = definitely disagree to 5 = definitely agree), satisfaction (1 = very dissatisfied to 5 = very satisfied) or frequency (1 = never to 5 = very often). The Self-Classified Weight Scale has five specific response options (1 = very underweight to 5 = very overweight). The results of each subscale are derived from the average of its items. On the subscale Appearance Evaluation (7 items), high scorers feel mostly positive and satisfied with their appearance. On the subscale Appearance Orientation (12 items), high scorers place more importance on how they look, pay attention to their appearance, and engage in extensive grooming behaviors. On the Body Areas Satisfaction Scale (9 items), high scorers are generally content with most areas of their body. On the subscale Overweight Preoccupation (4 items), higher scores indicate higher preoccupation. On the subscale Self-Classified Weight (2 items), individuals are classified according to their own perception in categories from “very underweight” to “very overweight”.

Appearance Schemas Inventory-Revised (ASI-R) [39, 40]

The instrument has 20 items and evaluates dysfunctional investment that an individual may have with his or her physical appearance. Its items are divided into two factors: (1) self-evaluative salience (SES), which accurately reflects the intensity to which beliefs about appearance influence the social and personal life of an individual and (2) motivational salience (MS), which is related to the intensity of concern about people’s appearance and how individuals adopt behaviors to control it. Items are scored over a Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree). The final score corresponds to the average of the items. Higher scores indicate higher levels of investment. In Brazil, the instrument has 13 items and was adapted by our research group (data available with authors). The scale showed good internal consistency in the present study (men: α = .71 for SES and .73 for MS, women: α = .73 for SES and .74 for MS).

Sociocultural Attitudes Toward Appearance Questionnaire-3 (SATAQ-3) [41]

The instrument has 30 items and measures 4 aspects of appearance-related media influences. This study used the nine-item General Internalization subscale, which evaluates the degree to which individuals have internalized media messages showing unrealistic body ideals as their personal standard. Items are scored over a Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The final score corresponds to the average of the items. Higher scores indicate greater internalization of socially established standards. In Brazil, the instrument was adapted by Amaral et al. [42], and the subscale has seven items. The scale showed good internal consistency in the present study (men: α = .93, women: α = .89).

Eating Attitudes Test (EAT-26) [43]

The instrument has 26 items distributed across 3 factors and assesses symptoms and concerns characteristic of eating-disordered populations. Items are scored over a Likert scale ranging from 0 (a few times, almost never and never) to 3 (always). The final score corresponds to the sum of the items, requiring inversion of one of the questions. Higher scores indicate greater evidence of inappropriate eating behavior. In Brazil, the instrument was translated by Nunes et al. [44]. The scale showed good internal consistency in the present study (men: α = .76, women: α = .77).

Rosenberg Self-Esteem Scale (RSE) [45]

The scale has 10 items and evaluates self-worth. Items are scored over a Likert scale, ranging from 1 (strongly disagree) to 4 (strongly agree). Five items are reverse coded. The final score corresponds to the average of the items. Higher scores indicate higher self-esteem. In Brazil, the instrument was adapted by Hutz and Zanon [46]. The scale showed good internal consistency in the present study (men: α = .88, women: α = .87).

Procedures

Translation and semantic validity

The original author granted permission for translation and validation of a Brazilian version of the MBSRQ-AS. Five independent translations were conducted by three experts in body image, one specialist in psychological instrumentation, and one professional translator. A single version of the instrument was created through discussions among the researchers and translators. A pilot study was carried out with the preliminary version of the Brazilian MBSRQ-AS administered to 31 students and university staff (16 women and 15 men) with different levels of education (11 had completed only elementary school, 10 had completed high school, and 10 had graduate education). No additional demographic information was gathered. Participants were personally recruited from various university campus locations by the researchers. Interested participants were provided with information about the study and those who agreed to participate gave written informed consent. All testing was conducted in a laboratory setting. Pilot participants read and answered each of the items and then discussed whether the terms were clear, appropriate, and well written. Some changes were made to produce a final version, which was back translated. Back translations were performed by two independent translators, whose first language was English, and who were blind to the original questionnaire. The final version was sent to the author of the original instrument.

Content validity

Content validity was evaluated by presenting the Brazilian MBSRQ-AS to a committee that was composed of six experts in body image and eating disorders. After accepting and signing an informed consent form, experts were provided a letter that contained a description of the questionnaire, conceptual definitions, the dimensions involved in it, and a characterization of each of the five subscales. The judges were requested to read the instrument, identify to which subscale each item pertained (i.e., Appearance Evaluation, Appearance Orientation, Body Areas Satisfaction Scale, Overweight Preoccupation, and Self-Classified Weight) and to provide their opinions and suggestions on the questionnaire. An accuracy rate of at least 80% for each question was used as the adequacy criterion [47].

Verbal comprehension

To assess verbal comprehension, participants were personally recruited from universities, technical schools and libraries by the researchers. The sample was similar to the one used in the pilot study with the preliminary version of the Brazilian MBSRQ-AS. Interested participants were provided with information about the study and those who agreed to participate gave written informed consent. All the testing was conducted in the educational settings, where participants self-reported their age, height, weight, race/ethnicity and highest educational level and completed the paper-and-pencil Brazilian MBSRQ-AS. Afterward, they were instructed to indicate how much they comprehended each of the items using a 5-point Likert scale ranging from 0 (did not understand anything) to 5 (understood perfectly). To analyze the verbal comprehension score for each question, we calculated the mean, standard deviation, and frequency of the participants’ responses to get an overview of verbal understanding of the instrument. Responses of 0, 1, 2, and 3 indicated insufficient understanding.

Factor structure and psychometric characteristics

To evaluate the MBSRQ’s factor structure, internal consistency, test–retest reliability, association between factors and differences between sex and weight categories, SurveyMonkey was hired. The company recruited individuals from their member site called SurveyMonkey Contribute and selected a random group from their member database that aligned with the inclusion/exclusion criteria approved by the IRB. Participants were pre-screened based on their age range (older than 18 years old), information that has been provided by respondents upon joining SurveyMonkey’s member base. For all other screening criteria, a question was included at the beginning of the survey. It was used as a standard template email notification to notify potential respondents that they had a new survey to take, and those who said yes to any of the conditions used as exclusion criteria were routed to a “Thank you” page, that thanked them for their time, and they were then directed to SurveyMonkey home site. Participants who were eligible to complete the study were provided with a link to the survey where they first read the informed consent online, provided their consent (or not), and then moved on to the questionnaires. All participants completed the questionnaire and were asked to self-report their age, height, weight race/ethnicity and highest educational level. To evaluate test–retest reliability, SurveyMonkey randomly invited 210 participants (90 men and 120 women) from the first sample, who answered the MBSRQ-AS again, within a 3-week interval.

Convergent validity

To evaluate convergent validity, participants were opportunistically recruited from various university campus locations by two trained research assistants. Interested participants were provided with information about the study and those who agreed to participate gave written informed consent. All testing was conducted in a laboratory setting, where participants completed paper-and-pencil, anonymous questionnaires.

Data analysis

Data were screened for missing values and to assess distributional properties. A descriptive analysis (mean, standard deviation, frequencies) was performed. To evaluate whether the original model provided a good fit to the data, a structural equations model or confirmatory factor analysis (CFA) was used. From the original model (estimated by the maximum-likelihood method), other models were adjusted. Information about discrepancies in the adjusted model, available from the covariance between the errors, was used to make post hoc changes and consequently improve the model fit. To determine how well the original model fit the data, we used six fit indices: χ2, goodness-of-fit index (GFI), normed fit index (NFI), comparative fit index (CFI), Tucker–Lewis index (TLI) and root mean square error of approximation (RMSEA). To assess model fit, the cutoff points purposed by Hu and Bentler [48] were considered: GFI, NFI, CFI and TLI ≥ .95 and RMSEA ≤ .06.

To verify that the items of the Brazilian MBSRQ-AS had adequate common variance for factor analysis, the Kaiser-Meyer-Olkin measure of sampling adequacy and the Bartlett’s test of sphericity were performed [49]. Exploratory factor analyses (EFA) were conducted using principal axis factor analysis (PAF) followed by Promax rotation (κ = 4), because some factors were theoretically correlated [5, 11]. The number of factors to be extracted was determined by factor eigenvalues above 1.0, the Velicer’s minimum average partial (MAP) test and Parallel Analysis using Principal Axis Factoring (PA-PAF) with raw data permutation. Factors were retained whenever the eigenvalues from the original data for a given factor exceeded the eigenvalues corresponding to the 95th percentile of the distribution of random data eigenvalues [50]. In addition, an extraction criterion of .40 was used.

Internal consistency was evaluated using Cronbach’s α coefficient. Intraclass correlation tests (ICC) were carried out to assess test–retest reliability using a two-way mixed effects model and type consistency. The association between the MBSRQ-AS factors was tested using Pearson’s correlation test. Separate multivariate analyses of variance (MANOVA), followed by pairwise comparisons using Bonferroni correction were used to evaluate differences between sexes and weight status. The Box’s test of equality of covariance matrices was conducted, and because the assumption of homogeneity of variance was violated (p < .05), the results were interpreted using Pillai’s trace test. Levene’s test of equality of variances was conducted for all tests, and results were not computed if p < .01.

Results

Even with some changes in the layout of the test, response scale, and some of the items made during the translation process, the author of the original instrument stated that the back-translated version was nicely equivalent to the English language version in meaning. The final version of the instrument is available from the authors.

Content validity

The level of agreement among the experts was 92.16% for the scale in its entirety. On the Appearance Evaluation subscale, the concordance was 85.71%. On the Appearance Orientation subscale, the concordance was 90.28%. On the Body Areas Satisfaction Scale, the concordance was 94.44%. On the two remaining subscales, the concordance was 100%. When evaluated separately, 29 items had over 80% concordance among the experts (25 items with 100% and 4 items with 83.33%), and only five were below that level (items 15, 16, 17, 18, and 34).

Verbal comprehension

The items were easily understood by the target population. Among women, the overall mean was 4.86 (SD = .54), varying from 4.56 (SD = 1.07) to 4.97 (SD = .16). A lower mean was given to item 31, which received scores of 0–3 from 12.3% of the sample. Among men, the overall mean was 4.87 (SD = .06), varying from 4.70 (SD = .83) to 4.94 (SD = .23). A lower mean was also given to item 31, which received scores of 0–3 from 7.9% of the sample.

MBSRQ-AS factor structure

First, to test whether the factor model proposed by Cash [5] suited our data, a CFA was conducted. However, the results were unsatisfactory [33] (χ2 = 1705.30 (DF = 467), GFI = .866, NFI = .844, CFI = .880, TLI = .856, and RMSEA = .066), which led us to switch to an exploratory approach.

The correlation matrix showed an intercorrelation of r = .80 for the two items of the subscale Self-Classified Weight. The squared multiple correlation (SMC) was close to 1.0 for both items. Therefore, an estimation problem due to multicollinearity could emerge. Consequently, we excluded these items from the factor analysis and conducted an EFA with only the first four factors.Footnote 1 The significance of Bartlett’s test of sphericity [χ2(496) = 9735.21, p < .001] and the size of the Kaiser-Meyer-Olkin measure of sampling adequacy (KMO = .90) revealed that the items of the Brazilian MBSRQ-AS had adequate common variance for factor analysis. The eigenvalue criterion and the MAP test indicated it was possible to extract six factors. However, parallel analysis suggested four factors should be retained. Based on the factor structure of the original MBSRQ-AS [5] and its other versions [10, 12], we decided to conduct a second EFA and set the number of factors extracted at four. The results demonstrated that the fourth factor was defined solely by reverse-worded items (14, 16 and 20), item 19 cross-loaded on factors 3 and 4, and item 11 did not load on any factor. We, therefore, decided to exclude all six reverse-scored items and to perform another EFA.

The size of the Kaiser-Meyer-Olkin measure of sampling adequacy (KMO = .91) suggested the remaining items of the MBSRQ-AS had adequate common variance for factor analysis, and the significance of Bartlett’s test of sphericity, χ2(325) = 7888.01, p < .001, suggested the correlation matrix was factorable. The eigenvalue criterion indicated it was possible to extract four factors, which was confirmed by the MAP test. Based on these criteria, the four-factor solution explained 51.58% of the variance after extraction. The first factor corresponded to the subscale Appearance Orientation (8 items) and explained 25.35% of the variance. The second factor corresponded to the Body Areas Satisfaction Scale (9 items) and explained 18.35% of the variance. The third factor corresponded to the subscale Appearance Evaluation (5 items) and explained 4.14% of the variance. The fourth factor corresponded to the subscale Overweight Preoccupation (4 items) and explained 3.73% of the variance. Rotated factor loadings (from the factor pattern matrix) and communalities are depicted in Table 1.

Table 1 Explained variance, factor loadings and communalities (h2) for each item of the Brazilian MBSRQ-AS
Full size table

Internal consistency

Internal consistency was satisfactory for all subscales. Cronbach’s α value was .87 for Appearance Orientation, .90 for Body Areas Satisfaction Scale, .83 for Appearance Evaluation, .73 for Overweight Preoccupation, and .89 for Self-Classified Weight.

Test–retest

Results demonstrated good test–retest stability on the five subscales: ricc = .86; p < .001 for Appearance Orientation; ricc = .76; p < .001 for Body Areas Satisfaction Scale; ricc = .81; p < .001 for Appearance Evaluation; ricc = .85; p < .001 for Overweight Preoccupation; and ricc = .92; p < .001 for Self-Classified Weight.

Relationships among MBSRQ-AS factors

Pearson’s correlations among the MBSRQ-AS factors ranged from weak to moderate. All correlations were significant except for Appearance Orientation, which was not associated with the Body Areas Satisfaction Scale or Self-Classified Weight, or for the Body Areas Satisfaction Scale, which was not correlated with Overweight Preoccupation (Table 2).

Table 2 Correlations among MBSRQ-AS factors
Full size table

Differences between sex and weight status

Separate multivariate analyses of variance (MANOVA), followed by pairwise comparisons using Bonferroni correction, were carried out to compare sex and BMI categories on the MBSRQ-AS subscales. To approximate group size, individuals who were classified as underweight (12 men and 18 women) were excluded, and participants who were overweight or obese were combined and compared to normal weight individuals. The MANOVA revealed a significant main effect of sex [Pillai’s Trace = .07, F (5,600) = 9.01, p < .001]. Follow-up univariate analyses demonstrated that women reported significantly higher levels of investment in appearance, lower levels of satisfaction with specific parts of the body, higher concerns with being overweight, and classified themselves as heavier than men (Table 3).

Table 3 Means (standard deviations) of the scores on the Brazilian MBSRQ-AS subscales by sex
Full size table

The MANOVA also revealed a significant main effect of weight status among men (Pillai’s Trace = .44, F (5,280) = 43.95, p < .001) and women (Pillai’s Trace = .43, F (5,284) = 42.57, p < .001). Follow-up univariate analyses demonstrated that overweight and obese men reported lower satisfaction with specific parts of their bodies, higher concerns with being overweight, and classified themselves as heavier than their normal weight counterparts. Similarly, overweight and obese women reported lower satisfaction with specific parts of their bodies, lower overall satisfaction, higher concerns with being overweight, and classified themselves as heavier than normal weight women (Table 4).

Table 4 Means (standard deviations) of the scores on the Brazilian MBSRQ-AS subscales by sex and weight status
Full size table

Convergent validity

As we hypothesized, Appearance Orientation was positively related to the Self-Evaluative and Motivational Saliences scales from the Appearance Schemas Inventory-Revised, the General Internalization subscale from the SATAQ-3 and the Eating Attitude Test. As can be seen in Table 5, all correlations were significant in the expected direction for both sexes.

Table 5 Correlations between the MBSRQ-AS scales and measures of body image and psychosocial functioning for men and women
Full size table

It was hypothesized that the Appearance Evaluation subscale and the Body Areas Satisfaction Scale would be related in a negative direction to both subscales from the Appearance Schemas Inventory-Revised, the General Internalization subscale from the SATAQ-3 and the Eating Attitude Test, and in a positive direction to the Rosenberg Self-Esteem Scale. As can be seen in Table 5, all the correlations were significant for women, except for Motivational Salience. For men, body satisfaction was not associated with Motivational Salience and eating behavior.

Finally, it was expected that the Overweight Preoccupation subscale would be positively related to the Self-Evaluative and Motivational Saliences scales from the Appearance Schemas Inventory-Revised, the General Internalization subscale from the SATAQ-3 and the Eating Attitude Test. Again, all correlations were significant in the expected direction for both sexes (Table 5).

Discussion

The aim of the present study was to cross-culturally adapt the MBSRQ-AS into Brazilian Portuguese and to investigate its factor structure and reliability in a nonclinical sample. After the back translations were processed, the author of the original questionnaire found the two versions were quite similar in meaning. The instrument was found to be easily understood by the target population, with verbal comprehension indices higher than 4.5 for all of the items. The majority of the items were classified as 4 (understood almost everything) or 5 (understood perfectly and had no doubt) by more than 90% of the sample.

Compared with the original MBSRQ-AS, the Brazilian version revealed some particularities. First, the items of the Self-Classified Weight subscale were removed from the exploratory factor analysis due to multicollinearity issues and because one factor must contain at least three items [51]. This choice was based on the fact that the absence of multicollinearity is an assumption for factor analysis [51, 52]. Importantly, Vossbeck et al. [13] faced the same situation with the German MBSRQ-AS and opted for the exclusion of these items from their factor analysis. Nevertheless, considering its unique contribution in the assessment of one’s perceived weight, we decided to retain this subscale.

Second, all negatively worded questions were excluded. Many researchers, with the intention of avoiding response bias, particularly acquiescence, adopt reverse-worded items. However, negative items adversely affect internal consistency and might produce a separate factor that is not substantively meaningful, which reflects method effects associated with the wording of the items [53, 54]. Moreover, negatively worded items are particularly troublesome in translations from English into other languages [6, 55, 56]. Measurement error caused by negatively worded items in Brazilian samples has also been observed with other scales (e.g., [42, 57]).

The abnormal psychometric characteristics of negatively worded items have been attributed either to carelessness or to difficulty in interpreting the questions [56]. In fact, a factor composed by negative items appears when only 10% of the respondents are careless [58]. Moreover, these items lead to increased difficulty, and thus more bias, without any clear advantage [59]. In our study, participants from the pilot study declared that they understood all items appropriately, but Swain et al. [60] argue that an error can be made due to the high level of difficulty of reverse-worded items, even when respondents think they understood the item well.

Nonetheless, the remaining items of the Brazilian MBSRQ-AS displayed the same factors as those reported by Cash [5], and all items had significant loadings ≥ .40 on at least one factor. The factor solution explained 51.58% of the variance after extraction, and this result is similar to results found by other authors worldwide, whose variance ranged from 41.44 to 85.41% [6, 7, 9, 10, 16, 17].

The subscales Appearance Orientation and Appearance Evaluation emerged as distinct factors, corroborating the results from the original instrument [4] and other versions [11, 13, 16, 17]. These two subscales are the only ones belonging to the original Body-Self Relations Questionnaire [4]. The Appearance Orientation subscale taps into the extent of investment in one’s appearance and is not considered pathognomonic, thus it may only reflect taking care of or taking pride in one’s looks [61], while Appearance Evaluation assesses overall satisfaction with one’s appearance. In this sense, it is plausible to suppose that the significant correlation found between them reflects that body satisfaction tends to increase as individuals place more importance on how they look, pay attention to their appearance, and engage in extensive grooming behaviors.

Another important result was the distinction between Appearance Evaluation and the Body Areas Satisfaction Scale. Even though they were correlated with each other, because both assess body satisfaction, the Body Areas Satisfaction Scale was not associated with some of the factors correlated with Appearance Evaluation. This fact demonstrates that they provide different information and must be treated as distinct factors. Moreover, they have different response formats which give another reason to treat them as separate factors.

Self-Classified Weight was associated with Appearance Evaluation and Body Areas Satisfaction, but not with Appearance Orientation. This result was already expected given that the investment made in appearance is not related to one’s weight [40].

In our study, internal consistency was found to be satisfactory for all subscales [3], and test–retest reliability demonstrated good stability on the five factors. Our findings are similar to those found by Cash [5] and other authors worldwide [11, 13, 16].

The significant differences in scores between men and women provided further support for the sensitivity of the scale. Similar results were found with the MBSRQ [5, 11, 17, 25]. Therefore, these differences were already expected as women generally report more dissatisfaction, are more invested in their appearance and are also more concerned with their weight [26, 62].

The association between the MBSRQ-AS and weight status demonstrated that overweight and obese participants reported higher dissatisfaction and concerns with being overweight and classified themselves as heavier than normal weight participants. Previous studies showed a link between being overweight, body dissatisfaction and weight concerns [36, 37]. Thus, these differences were also expected, supporting, once more, the sensitivity of the Brazilian MBSRQ-AS. Similar findings were described worldwide [16, 17, 25].

Convergent validity was demonstrated by the significant correlations between the MBSRQ-AS subscales and several measures of body image and psychosocial functioning. Appearance Orientation was positively related to the Self-Evaluative and the Motivational Saliences scales from the Appearance Schemas Inventory-Revised. This result was expected because the three subscales measure psychological investment in appearance. The significant correlations found between Appearance Orientation, the SATAQ-3 and the Eating Attitude Test were also expected, because men and women who give more importance to appearance often report greater internalization of appearance-related media ideals and more disordered eating [40]. Satisfaction with appearance was negatively related to the Appearance Schemas Inventory-Revised, the SATAQ-3 and the Eating Attitude Test, and positively associated with the Rosenberg Self-Esteem Scale. These results were anticipated because a large body of research has already demonstrated the inverse relationship between body satisfaction and psychological investment in appearance, internalization of media ideals and disturbed eating behavior and a positive relationship between body satisfaction and self-esteem [11, 17, 63,64,65]. Finally, the Overweight Preoccupation subscale was positively related to investment in appearance, internalization of appearance-related media ideals and more disordered eating. These results make sense because this subscale measures concerns about being or becoming fat, vigilance over weight fluctuations, dieting, and eating restraint [5]. All these data confirm that the exclusion of the reversed items did not influence the test’s ability to behave as expected in relation to other well-established measures.

On one hand, our study has several strengths. This study is the first study to adapt the MBSRQ-AS into Brazilian Portuguese and to provide evidence of the psychometric quality of the adapted test. We used a large sample composed of men and women from 18 to 65 years, recruited from different settings (e.g., community, universities), personally and via internet, instead of mainly female college students. Importantly, some studies have demonstrated that the mean scores for self-reported surveys using paper-and-pencil and internet data collection methods are generally equivalent [66, 67], validating the results of the study. We also evaluated the readability and comprehension of the items, and conducted a retest, proving the temporal stability of the Brazilian MBSRQ-AS. We presented evidence of construct validity, reliability and convergent validity with individuals from the community and college students. Our goal was to provide data supporting the use of the measure in samples that represent a variety of adults. On the other hand, the study has its limitations. White individuals mainly composed our sample. Another study attempting to replicate our findings with more diverse racial/ethnic groups may provide further support of the current results. The use of self-reported instruments and measures of weight and height are also limitations of the study, as well as the differences in educational level and body mass index across the samples. Finally, given the relevance of body image in eating disorders, future studies should analyze its psychometric properties in a clinical sample.

Conclusions

The Brazilian MBSRQ-AS presented good indices of verbal comprehension and the same factor structure as the original instrument. Reliability indices were found to be satisfactory, and the instrument was able to capture sex and weight status differences. The present study also provides evidence for the convergent validity of the Brazilian MBSRQ-AS. Overall, these results demonstrate that the questionnaire is a good option for researchers in Brazil to measure the construct of multidimensional body image. Furthermore, the study provides substantial parameters for comparison with other adaptations of the instrument around the world.