Racial Bias in Personality Assessment: Using the MMPI-2 to Predict Psychiatric Diagnoses of African American and Caucasian Chemical Dependency Inpatients Matthew J. Monnot, Stuart W. Quirk, Michael Hoerger, and Linda Brewer Central Michigan University An assessment of predictive bias was conducted on numerous scales of the Minnesota Multiphasic Personality Inventory–2 (MMPI-2; J. N. Butcher, W. G. Dahlstrom, J. R. Graham, A. Tellegen, & B. Kaemmer, 1989), including the Restructured Clinical (RC) scales, in the prediction of clinical diagnostic status for African American and Caucasian male veterans seeking substance abuse treatment. Patients completed a battery of self-report instruments and were administered structured diagnostic interviews. African American patients obtained higher scores across most MMPI-2 scales compared with Caucasians with clinically meaningful elevations (T scores  5 points) on 3 scales. The RC scales demonstrated strong correlations with diagnoses, however, like other MMPI-2 scales examined in this study, they displayed a general trend of predictive bias. Step-down hierarchical regression procedures (G. J. Lautenschlager & J. L. Mendoza, 1986) indicated the presence of predictive bias for a majority of the scales examined; however, most of these effects were small to modest (accounting for 3%–5% of variance). The pattern of slope and intercept biases across types of MMPI-2 scales differs from prior research and indicates the importance of evaluating bias in various populations and settings. Keywords: MMPI-2, RC scales, racial bias Test bias in psychopathology assessment is an important con- cern due to the magnitude of decisions made as a result of such assessments. The most widely used self-report inventory of psy- chopathology, the Minnesota Multiphasic Personality Inventory–2 (MMPI-2; Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989), has been criticized as potentially biased in predicting psy- chiatric status of racial minorities (Adebimpe, 1981; Aponte & Johnson, 2000). However, empirical efforts have yielded equivocal results with regard to both mean differences and biased predictions of clinically relevant extratest variables (Arbisi, Ben-Porath, & McNulty, 2002; McNulty, Graham, Ben-Porath, & Stein, 1997; Timbrook & Graham, 1994). Additional research of racial bias using relevant clinical criteria (such as symptom reports or diag- nostic status) in various clinical settings and with differing patient populations is required in order to clarify previously discrepant findings. Such research is also necessary to help determine the importance of patient group and clinical setting with regard to predictive bias in the MMPI-2. Early efforts to explore racial bias in the original MMPI and subsequent MMPI-2 examined group differences in mean scale elevations (e.g., Gynther, 1972). Evaluating mean differences be- tween ethnic minorities and Caucasian groups has yielded ambig- uous results. For example, researchers (Frueh, Smith, & Libet, 1996) have reported that the profiles of African American veterans indicated greater maladjustment (i.e, higher mean scores on Scales 6 and 8) compared with Caucasian patients; however, this effect was not replicated (Frueh, Gold, de Arellano, & Brady, 1997). Such inconsistency is reflected in a meta-analysis of scale differ- ences between African American and Caucasian groups (Hall, Bansal, & Lopez, 1999). A fundamental limitation of this line of research is that the presence of mean differences between groups is insufficient for establishing test bias. Mean differences may reflect genuine differences between groups or settings rather than biases in clinical conclusions or behavioral predictions (Archer, Griffin, & Aiduk, 1995). Tests of predictive equivalence require the inclusion of relevant external criteria against which patterns of prediction can be scru- tinized (Arbisi et al., 2002; Morrison, Edwards, & Weissman, 1994; Munley, Busby, & Jaynes, 1997). Small but reliable mean differences between Caucasian and African Americans have been found without a corresponding difference in the prediction of psychopathology in normal samples (McNulty et al., 1997) and partner ratings of interpersonal behaviors and personality charac- teristics (Timbrook & Graham, 1994). In one of the most comprehensive evaluations of racial bias in the MMPI-2 to date, Arbisi and colleagues (2002) collected clin- ically relevant criterion information from inpatient charts in a general inpatient psychiatric hospital setting and a large VA med- ical center to evaluate relevant MMPI-2 scales for racial bias. They examined the comparative associations between MMPI-2 scales and criterion variables by group status (tests of slope bias) as well as systematic over- or underprediction for each group (intercept bias). In this general sample of inpatients, the MMPI-2 yielded a number of small but reliable intercept bias effects indicating un- Matthew J. Monnot, Stuart W. Quirk, Michael Hoerger, and Linda Brewer, Department of Psychology, Central Michigan University. Linda Brewer is now at Central Arkansas Veterans Healthcare System. Correspondence concerning this article should be addressed to Matthew J. Monnot, who is now at Organization Development, Genentech, South San Francisco, CA 94080. E-mail: monnot.matthew@gene.com Psychological Assessment © 2009 American Psychological Association 2009, Vol. 21, No. 2, 137–151 1040-3590/09/$12.00 DOI: 10.1037/a0015316 137