The translation followed a multistage procedure recommended by the International Test Commission and required by the instrument’s publisher. Two Brazilian psychologists independently translated all items and reconciled differences through discussion, producing a third translation that was compared with a version prepared by a professional linguistic-services company. Reviewing these alternatives led to minor changes in 14 items where small wording differences could alter interpretation. Examples included the English word “smashing,” for which “quebrar” and “esmagar” carry different implications of intensity in Brazilian Portuguese, and an item referring to legal issues, which was adjusted for differences in how minor infractions may be perceived in Brazil. The revised form was examined by a psychiatrist and another psychologist, back-translated into English, and reviewed by the University of Minnesota language team. Bilingual participants then evaluated item wording, instructions, and possible refinements before the final Portuguese version was used in the larger community study.

The bilingual comparison included 18 men and 18 women aged 21 to 56 who were native Portuguese speakers, had lived in the United States for at least six months, and demonstrated C2-level English proficiency. Each participant completed the English and Brazilian Portuguese forms two weeks apart, with administration order counterbalanced within sex. Correlations for the higher-order scales ranged from 0.81 for Emotional/Internalizing Dysfunction, EID, to 0.92 for Behavioral Dysfunction, BXD. Restructured Clinical scale correlations ranged from 0.69 for RC1 to 0.91 for RC4 and RC7. The revised PSY-5 scales ranged from 0.81 for Aggressiveness, AGGR-r, to 1.00 for INTR-r. These coefficients represent both similarity between language versions and stability over the two-week interval, meaning they do not isolate translation differences from ordinary variation over time. The investigators therefore described the results as cross-language score correspondence rather than a formal test of measurement equivalence.

The community sample initially included 910 Brazilian adults aged 18 to 75 with at least six years of formal education. After 146 protocols, or 16.0% of the initial group, were excluded under MMPI-2-RF invalidity criteria, the final sample comprised 764 adults: 382 men and 382 women, with a mean age of 33.16 years. Most participants came from Brazil’s Southeast region, and the group was predominantly single, White, relatively highly educated and middle-income. The study was intentionally not designed to represent people with severe psychiatric disorders. Recruitment excluded people diagnosed with schizophrenia, psychotic depression in an acute crisis, people with psychiatric hospitalizations, and those with suicide attempts in the previous six months. Accordingly, the results cannot be extended to those populations, including people with psychotic-spectrum conditions whose symptom variability was deliberately limited.

Internal consistency, assessed with Cronbach’s alpha, was uneven across the inventory. The highest alpha among substantive scales was 0.90 for RCd, followed by 0.87 for EID, while several short problem-specific scales were less internally consistent. HLP had an alpha of 0.38, DSF, Distress/Shortness of Breath, had 0.44 and SFD, Somatic Dysfunction, had 0.76. The revised PSY-5 scales clustered between 0.64 for AGGR-r and DISC-r, Disorganization, and 0.74 for NEGE-r, Negative Emotionality. Alpha depends on both the number of items and their average covariance, so coefficients from brief scales cannot be interpreted in the same way as those from longer scales. Still, low internal consistency can limit the precision with which the Brazilian form distinguishes individual differences. The response-inconsistency scales VRIN-r and TRIN-r require special caution because they are built from item pairs associated with inconsistent or imbalanced responding rather than interchangeable indicators of one substantive construct.

Standard errors of measurement were 4 to 9 T-score points. These figures describe expected score imprecision under the classical test theory assumptions and the internal consistency estimates obtained in this sample, not thresholds for clinical significance. Most scales had mean standard errors of 8 points or less. VRIN-r and TRIN-r had the highest values, at 9 points, whereas F-r had values of 4 to 5. The higher-order and Restructured Clinical scales generally had the smallest standard errors, while HLP, BRF and DSF were among the problem-specific scales with the highest values. The results suggest that the Brazilian form may be better suited to broad domains and some clinical scales than to several narrower problem scales, although the study did not test whether the observed precision is sufficient for any particular application.

An exploratory model examined demographic associations with NUC scores in the community sample. The model was statistically significant but accounted for only 7.4% of the variance in NUC scores, with an adjusted R² of 0.060. Predicted scores declined across early and middle adulthood, reached an estimated minimum at about age 51 and then increased, indicating a small nonlinear age association. The linear age coefficient was −0.209, with a 95% confidence interval from −0.305 to −0.114, while the quadratic coefficient was 0.006, with a 95% confidence interval from 0.002 to 0.010. NUC scores also differed by sex, and education showed a small association in the robust analysis that was not significant in the corresponding conventional analysis of variance. Geographic region was not associated with NUC scores. Because this analysis used only score-level demographic associations, it did not test measurement invariance or differential item functioning and cannot explain NUC’s comparatively low cross-language correlation.

Intercorrelations among the Brazilian scale scores ranged from weak to strong, with several comparatively large associations appearing among scales assigned to the same broad domains. EID was strongly associated with RCd, RC7 and other scales related to emotional dysfunction. RC6 and RC8 showed very high associations with PSYC-r, while RC4 and RC9 were strongly associated with the broader behavioral dysfunction scales; RC9 also correlated strongly with AGG and ACT, and RC4 was strongly related to JCP, Justified Problems. These patterns resemble the broad organization of the original MMPI-2-RF, but they were derived entirely from correlations among scores obtained from the same self-report inventory. Without independent measures of personality, psychopathology, functioning, diagnosis or clinical outcomes, the study cannot establish convergent, discriminant or criterion-related validity. It also cannot determine whether correlated scores represent distinct syndromes, genuine distress, response style or shared item content.

The authors emphasize that the findings are not a release of clinical interpretation for the Brazilian form. The study did not test a latent structural model, measurement invariance, differential item functioning, diagnostic classification or criterion-related prediction, and the cross-language design did not separate language effects from two-week stability. The community sample also lacked adequate representation of severe psychiatric and psychotic disorders, limiting conclusions about scale performance in those groups. Previous work on other MMPI adaptations has shown that a well-translated instrument can still differ across countries in score distributions, item functioning and scale associations. Even the closer European Portuguese adaptation of the different Personality Assessment Inventory produced normative differences and did not reproduce its originally proposed full-scale structure without modification. For the Brazilian MMPI-2-RF, the current evidence supports a carefully adapted instrument with promising score correspondence and preliminary internal coherence, but further research with representative community and clinical samples, external criteria, formal structural models and invariance testing is needed before interpretations developed for the North American version can be evaluated for use in Brazil.

Subject of Research: Initial Brazilian Portuguese adaptation and psychometric evaluation of the MMPI-2-RF personality and psychopathology inventory

Article Title: Filling the gap of the MMPI-2-RF in Brazil: translation, reliability, and preliminary internal structural coherence of scale scores

Article References: Lerman, T. G., Andreoli, S. B., Sellbom, M., Ben-Porath, Y. S., Gonçalves, A. P., & de Francisco Carvalho, L. (2026). Filling the gap of the MMPI-2-RF in Brazil: translation, reliability, and preliminary internal structural coherence of scale scores. Current Psychology, 45(20), Article 1635. https://doi.org/10.1007/s12144-026-10168-8

Image Credits: AI Generated

DOI: 10.1007/s12144-026-10168-8

Keywords: MMPI-2-RF, Brazilian Portuguese, psychological assessment, personality inventory, psychopathology, test adaptation, measurement reliability, cross-language validation, internal structure, Brazil, Filling, translation