Psychometric evaluation of the restructured clinical scales of the MMPI-2

Conceptual overlap and heterogeneity have long been noted as weaknesses of the Minnesota Multiphasic Personality Inventory's clinical scales. Restructured clinical (RC) scales recently were developed to address these concerns (A. Tellegen et al., 2003). The authors evaluated the psychometric properties of the RC scales in psychology clinic clients (N=285) and military veterans (N=567). The RC scales were as internally consistent as the clinical scales and correlated strongly with their original counterparts (except for RC3/Hysteria). They also were less intercorrelated, produced conceptually clearer relations with measures of personality and psychopathology, and yielded somewhat greater incremental utility than the clinical scales. Thus, the RC scales demonstrated several psychometric strengths while utilizing 60% fewer items, but the 2 sets of scales cannot be used interchangeably. Interpretive considerations are discussed.     

An evaluation of the usefulness of the MMPI-2 F(p) scale

The Minnesota Multiphasic Personality Inventory-2 (MMPI-2) F(p) scale was developed by Arbisi and Ben-Porath (1995) by identification of 27 items endorsed by fewer than 20% of individuals in both normal and psychiatric samples. The F(p) scale was designed for applications in settings characterized by high base rates of serious psychopathology, such as psychiatric inpatient units, and is proposed as a useful scale in discriminating overreported protocols from those produced by patients with serious psychopathology. In this study we investigated the characteristics of this scale in a sample of 617 psychiatric inpatients who responded to the MMPI-2 under standard conditions, and 203 overreported protocols derived in research studies conducted with normal adult participants instructed to simulate various forms of serious psychopathology. Results of this study are consistent with prior reports of a relatively low frequency of item endorsement for F(p) scale items in psychiatric samples, and intercorrelations between the F(p) scale and the MMPI-2 basic clinical scales in clinical samples that are generally lower than those produced between either F or Fb and the basic clinical scales. However, this intercorrelational pattern between F(p) and the MMPI-2 basic scales was not as consistent for the overreported sample. Additionally, the F(p) scale appears to be effective in discriminating overreported from accurate MMPI-2 protocols, with some evidence that the optimal cutting scores for this and other MMPI-2 infrequency scales may differ as a function of gender. Finally, these findings do not show clear evidence of improved group prediction derived from the use of the F(p) scale in contrast to results obtainable through the use of the MMPI-2 F scale.     

The factor structure of the MMPI-2 Restructured Clinical (RC) scales

We investigated the MMPI-2 Restructured Clinical (RC) scales (Tellegen et al., 2003) to determine if they had a more differentiated factor structure than the MMPI-2 Clinical scales. When factored alone, the RC scales had a 5-dimensional structure; the Clinical scales had 3 dimensions. When factored in combination with the Content scales, both sets of scales produced 5 dimensions. However, the RC and Content factors generally provided more efficient and logical markers of psychopathology than the Clinical and Content factors. We discuss interpretive considerations.     

Psychometric evaluation of the MMPI-2 Restructured Clinical scales in two Dutch samples

In this article, we evaluate internal validity, internal consistency, and test-retest reliability of the MMPI-2 Restructured Clinical (RC) scales in the Dutch MMPI-2 normative sample (N = 1,244) and a Dutch outpatient psychiatric sample (N = 1,066). We pay special attention to a critique regarding construct drift of RC3 and the redundancy of the RC scales with existing MMPI-2 scales. The results indicate that the RC scales in both samples show comparable or better internal consistencies than the Clinical scales. Also, in both samples, the RC scales demonstrate lower scale-level intercorrelations than the Clinical scales. As to the structural characteristics, principal component analysis of the RC scales provided a clearer pattern than an analysis of the Clinical scales. Furthermore, mean raw scores on the RC scales for men in the Dutch normative sample corresponded highly with those in the U.S. normative sample except for RC2 and RC4. Less correspondence was found for women. Overall, we conclude that the RC scales show satisfactory reliability and promising internal validity in our Dutch samples. We suggest that U.S. validation studies on the RC scales may be generalized to the Dutch-language version of the MMPI-2 RC scales.     

An examination of three sets of MMPI-2 personality disorder scales

Three sets of personality disorder scales (PD scales) can be scored for the MMPI-2 (Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989). Two sets (Levitt & Gotts, 1995; Morey, Waugh, & Blashfield, 1985) are derived from the MMPI (Hathaway & McKinley, 1983), and a third set (Somwaru & Ben-Porath, 1995) is based on the MMPI-2. There is no validity research for the Levitt and Gotts scale, and limited validity research is available for the Somwaru and Ben-Porath scales. There is a large body of research suggesting that the Morey et al. scales have good to excellent convergent validity when compared to a variety of other measures of personality disorders. Since the Morey et al. scales have established validity, there is a question if additional sets of PD scales are needed. The primary purpose of this research was to determine if the PD scales developed by Levitt and Gotts and those developed by Somwaru and Ben-Porath contribute incrementally to the scales developed by Morey et al. in predicting corresponding scales on the MCMI-II (Millon, 1987). In a sample of 494 individuals evaluated at an Army medical center, a hierarchical regression analysis demonstrated that the Somwaru and Ben-Porath Borderline, Antisocial, and Schizoid PD scales and the Levitt and Gotts Narcissistic and Histrionic scales contributed significantly and meaningfully to the Morey et al. scales in predicting the corresponding MCMI-II (Millon, 1987) scale. However, only the Somwaru and Ben-Porath scales demonstrated acceptable internal consistency and convergent validity.     

Convergent validity of the MCMI-III personality disorder scales and the MMPI-2 scales

The MCMI-III personality disorder scales (Millon, 1994) were empirically validated in a sample of prisoners, psychiatric inpatients, and outpatients (N = 477). The scale intercorrelations were congruent with those obtained by Millon, Davis, and Millon (1997). We conclude that our Flemish/Dutch version shows no significant differences with the original version of the MCMI-III as far as intercorrelations are concerned. Convergent validity of the MCMI-III personality disorder scales was evaluated by the correlational data between the MCMI-III personality disorder scales and the MMPI-2 clinical (Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989) and personality disorder (Somwaru & Ben-Porath, 1995) scales. Improved convergence was obtained compared with previous versions of the MCMI-I. Only the compulsive MCMI-III personality disorder scale remains problematic. The scale even showed negative correlations with some of the related clinical scales and with the corresponding personality disorder scales of the MMPI-2.     

An investigation of the psychometric properties of the MMPI-2 Restructured Clinical (RC) scales with mental health inpatients

In this study, we compared the internal psychometric properties and external correlates of the MMPI-2 (Butcher et al., 2001; Tellegen et al., 2003) Restructured Clinical (RC) scales and corresponding MMPI-2 non-K-corrected Clinical scales in a sample of 280 men and 264 women admitted to inpatient psychiatric units. We evaluated the internal psychometric properties of the RC and Clinical scales by intercorrelation analyses and alpha coefficients. We investigated external correlates of scores from the RC and Clinical scales using selected chart variables, selected Symptom Checklist 90-Revised (Derogatis, 1983, 1994) items, and clinician ratings on the Brief Psychiatric Rating Scale (Overall & Gorham, 1988). Results showed psychometric support for the RC scales in terms of improved intercorrelation patterns and comparable or greater alpha coefficients. Further, external correlational analyses revealed varying degrees of evidence of improvement in convergent and discriminant validity.     

A taxometric analysis of the MMPI-2 Depression scales

The MMPI (Hathaway & McKinley, 1943) and MMPI-2 (Butcher et al., 2001) have long been used as measures of psychopathology. Both clinicians and researchers have noted the widespread existence of negative affectivity on the MMPI and MMPI-2 that may elevate scale scores and eclipse the tests' ability to differentiate depression from other clinical disorders. Using taxometric analyses, in this study we sought to test directly whether the MMPI-2 depression scales could differentiate patients with depressive symptoms from patients with other disorders. A large psychiatric sample (N = 2,000) was utilized and analyses were run separately for men and women. Taxometric analyses did not find a MMPI-2 Depression scale cut point that categorizes patients with depressive symptoms from other patients. Rather, these findings support previous studies finding an underlying dimensionality of depression. We discuss implications for MMPI-2 scale use and depression nosology in light of these findings.     

Comparability of MMPI-2 scales and profiles over time

This study investigated Minnesota Multiphasic Personality Inventory-2 (MMPI-2; Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989) scale and profile comparablilty for MMPI-2 profiles completed on 2 separate occasions by mental health patients receiving treatment at a Veterans Affairs Medical Center (n = 114). Patients were predominantly men (96.5%), with an average age of 44.08 and an average of 12.39 years of education at the time of initial testing. MMPI-2 tests were completed on 2 separate occasions as a routine part of treatment with a mean interval between test administrations of 688 days. Findings were analyzed for the complete sample and for 3 subsamples with different test-retest intervals. MMPI-2 scale test-retest correlation coefficients for the entire sample ranged from .48 to .69 for the Basic scales, .49 to .80 for the Supplementary scales, and .56 to .78 for the Content scales with scale high-point agreement = 38.60%, high 2-point agreement = 16.67%, and high 3-point agreement = 19.30%. High-point agreement for subsets of participants with well-defined high points, 2-points and 3-points was 41.07%, 27.50%, and 25.93% respectively. Pearson r correlation coefficients for T scores across the Basic scales for pairs of profiles averaged .78, suggesting similarity of profile shape across testing occasions. MMPI-2 profiles were also examined in relation to Skinner and Jackson's 3 modal MMPI profile types.     

Validation of selected MMPI-2 basic and content scales

To investigate whether some of the MMPI-2 basic and/or content scales measure the factor structure represented by the Big-Five, 288 undergraduates (178 women and 110 men), between the ages of 18 and 34 years, were administered MMPI-2 and MARS (Michill Adjective Rating Scale) which measures four factors that are conceptually similar to the first four of the Big-Five. It was hypothesized that (a) three MMPI-2 scales (basicDepression or D, contentAnxiety or ANX, and contentDepression or DEP) would be correlated with MARSUnhappiness; (b) two MMPI-2 scales (basicSocial Introversion or Si and contentSocial Discomfort or SOD) would be correlated with MARSExtraversion; (c) two MMPI-2 scales (contentAnger or ANG and contentType A Behavior or TPA) would be correlated with MARSAssertiveness; and (d) two MMPI-2 scales (contentWork Interference or WRK and supplementaryDominance or DOM) would be correlated with MARSProductive Persistence. Results corroborated the hypothesized relationships between the components of the aforementioned four dyads.     

Incremental validity of the MMPI-2 Restructured Clinical (RC) scales in a private practice sample

Building on results reported in Sellbom, Graham, and Schenk (2005), in this study, we examined the incremental validity of the newly introduced MMPI-2 (Butcher et al., 2001) Restructured Clinical (RC) scales (Tellegen et al., 2003) over both the Clinical and Content scales. Participants were 647 clients in private practice who were administered the MMPI-2 and the Multiaxial Diagnostic Inventory (Doverspike, 1990) early in therapy. The results indicate that the RC scales had acceptable internal consistency, reduced intercorrelations (compared to the Clinical scales), and promising convergent and discriminant validity. Hierarchical regression analyses revealed that the RC scales added incrementally to both the Clinical and Content scales in predicting self reported clinical symptoms. These findings suggest that the RC scales are relatively homogenous measures of core clinical constructs that can add unique information to the understanding of private practice clients above and beyond the Clinical and Content scales.     

Convergent validity of alternative MMPI-2 personality disorder scales

The Morey, Waugh, and Blashfield (1985) MMPI (Hathaway et al., 1989) personality disorder scales provided a significant contribution to personality disorder research and assessment. However, the subsequent revisions to the MMPI and the multiple revisions to the diagnostic criteria sets that have since occurred may have justified comparable revisions to these scales. Somwaru and Ben-Porath (1995) selected a substantially different set of items from the MMPI-2 (Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989) to assess Diagnostic and Statistical Manual of Mental Disorders (4th ed.; American Psychiatric Association, 1994) personality disorder diagnostic criteria. In our study, we compared the convergent validity of these alternative MMPI-2 personality disorder scales with respect to 3 self-report measures of personality disorder symptomatology in a sample of 82 psychiatric outpatients. The results suggested that Somwaru and Ben-Porath's scales are as valid as the original Morey et al. scales and might be even more valid for the assessment of borderline, antisocial, and schizoid personality disorder symptomatology.     

A cross-validation of MMPI-2 substance abuse scales.

We evaluated the efficacy of the Addiction Potential Scale (APS) and the Addiction Acknowledgement Scale (AAS), two new scales designed to assess substance abuse problems with the Minnesota Multiphasic Personality Inventory-2 (MMPI-2), in samples from settings different from those used in their development. Results replicate earlier findings that both scales discriminate between psychiatric and substance abuse samples and do so more effectively than other substance abuse scales designed for use with the MMPI and carried over to the MMPI-2. Results also suggest that APS may be more resistant than AAS to response distortion.     

The effect of MMPI-2 on scoring of special scales derived from MMPI-1.

Twenty-nine special scales derived from the Minnesota Multiphasic Personality Inventory (MMPI-1; Hathaway & McKinley, 1983) were scored from MMPI-I and MMPI++-2 (Hathaway et al., 1989) items for a normal and a psychiatric sample. Resulting pairs of mean scores were compared. Absolute differences were found to be small but statistically significant. We concluded that, despite the statistical significance, the small absolute differences suggest that most of the MMPI-I special scales probably can be scored and interpreted from MMPI-2 items. Further research along this line is necessary.     

An evaluation of the incremental validity of the MMPI-2 Superlative (S) scale in an inpatient psychiatric sample

The MMPI-2 Superlative (S) scale was developed by Butcher and Han (1995) to assess individuals' tendencies to present themselves in an unrealistically positive light. The current study examined the performance of the L, K, and S scales in accurately distinguishing the MMPI-2 profiles of 379 psychiatric inpatients who produced one or more elevations on the basic scale profiles, from 82 psychiatric inpatients who produced no clinical range elevation on any of the eight basic clinical scales (excluding scales 5 and 0). The findings from the present study indicate that the S scale appears to be effective in the identification of defensiveness among psychiatric inpatients, with mean effect sizes in the moderate to large range. The optimal T-score cutoff for identifying defensive responders was 70 for both men and women. Further, results of hierarchical regression analyses showed that the S scale added incrementally to the prediction levels achieved by the optimal combination of the L and K scales.     

Profile validity standards for MMPI and MMPI-2 F scales

Although numerous indices of validity have been developed for the MMPI and MMPI-2, interest in the F scale and its variants continues, especially among practicing clinicians. The use of the binomial for assessing standards for random answering and possibly for judgments of malingering offers another approach for the interpretation of F-scale scores. The theoretical binomial distribution and Monte Carlo data are in accord. Cut-off scores of 24 for the MMPI and 23 for the MMPI-2 suggest random responses, and scores of 40 and 37, respectively, suggest clinical interpretation rather than randomness of responding.     

A comparison of the correlational structures and elevation patterns of the MMPI-2 restructured clinical (RC) and clinical scales

The relationship between the MMPI-2 restructured clinical scales and the original clinical scales was evaluated using an outpatient clinical sample (N = 150). Similar patterns of correlations to those reported by Tellegen et al. in 2003, such as high correlations between the restructured scales and their original scale counterparts and lower intercorrelations among the restructured scales than among the original scales, were found in the outpatient sample. The former provided evidence of convergence between the two sets of scales, whereas the latter increases the potential for discriminant validity in the restructured scale profiles. Mean T-scores of the restructured scales were significantly lower than their original scale counterparts for every clinical scale except Scale 1 (hypochondriasis). Individual profiles exhibited fewer scale elevations using the restructured clinical scales (M = 2.15, Mdn = 2.0) than the original clinical scales (M = 3.29, Mdn = 3.0). The majority of client profiles (56%) had fewer scale elevations when plotted using the restructured scales versus the original clinical scales.     

The utility of the NEO-PI-R validity scales to detect response distortion: a comparison with the MMPI-2

In this psychometric study, we compared the recently developed Validity Scales from the Revised NEO Personality Inventory (NEO PI-R; Costa & McCrae, 1992b) with the MMPI-2 (Butcher, Dahstrom, Graham, Tellegen, & Kaemmer, 1989) Validity Scales. We collected data from clients (n = 74) who completed comprehensive psychological evaluations at a university-based outpatient mental health clinic. Correlations between the Validity Scales of the NEO-PI-R and MMPI-2 were significant and in the expected directions. The relationships provide support for convergent and discriminant validity of the NEO-PI-R Validity Scales. The percent agreement of invalid responding on the two measures was high, although the diagnostic agreement was modest (kappa = .22-.33). Finally, clients who responded in an invalid manner on the NEO-PI-R Validity Scales produced significantly different clinical profiles on the NEO-PI-R and MMPI-2 than clients with valid protocols. These results provide additional support for the clinical utility of the NEO-PI-R Validity Scales as indicators of response bias.     

Factor analysis of MMPI-2 basic scales among college students

The MMPI-2 was administered to 288 college students (178 women and 110 men) between the ages of 18 and 34 years. Scores on the basic 13 (three validity and ten clinical) scales were correlated and subjected to principal component analyses separately for men and women. In each case, four factors were retained and subjected to varimax rotation. Coefficients of congruence were calculated for all possible pairs of factors across two samples of college students and the normative adults, separately for men and women, and across genders within each sample. Results indicated that the factorial structure of the MMPI-2 basic scales seemed to possess considerable uniformity across samples and genders, but not without some notable exceptions.