Race/ethnicity and the Personality Assessment Inventory (PAI): The impact of culture on diagnostic testing in a college counseling center

This study examined the validity of using the PAI with African-Americans by comparing the PAI profiles of African-American and Euro-American college students for systematic differences as a function of race/ethnicity. These profiles were also analyzed to determine if they collectively reflected various Black/White diagnostic patterns identified in the research literature (e.g., Aldwin & Greenberger, 1987; Lawson, 1986; Strawkowski et al., 1995).; The PAI protocols of 656 college students (50% African-American and 50% Euro-American) who sought services at a large Midwestern college counseling center between 1995 and 2003 were analyzed. The two racial/ethnic groups were matched as closely as possible on gender and age.; A series of 2 (race) x 2 (gender) x 2 (geographical area of origin [i.e., urban vs. rural]) multivariate analyses of variance (MANOVA) were used to compare participants' scale scores. Participants scores on the PAI's validity scales, clinical scales, treatment scales, and interpersonal scales were used as the dependent variables. For each MANOVA yielding significant results, a univariate analysis of variance (ANOVA) was conducted to determine the specific variables contributing to the statistical significance. Although several significant differences were found between the mean scale scores of the two racial/ethnic groups, the magnitude of these differences could not support a conclusion of test bias. However, there was a relatively large difference (i.e., 8.82 t-scores) between the two groups on the Paranoia (PAR) scale. Possible explanations for this difference are discussed.; The PAI's factor structure for the two racial/ethnic groups was evaluated using the same factor analytical procedures reportedly used in the PAI Professional Manual (Morey, 1991). Specifically, exploratory factor analyses involving principal components extraction followed by varimax rotations was performed using the correlation matrices generated by the respective African-American and Euro-American samples. Factors with eigenvalues greater than one were retained for rotation. The results of each of these analyses were compared for differences, and compared to Morey's (1991) factor analytical results. Although there appeared to be significant differences between the factor structures of the two groups, there was also enough similarity to disallow a conclusion of test bias. Explanations of the observed differences are discussed, and suggestions for future research offered.