We investigated methodological and sample-based characteristics that might contribute to discrepancies in the structure of the 22-scale Personality Assessment Inventory (PAI; Morey, 1991, 2007). In Study 1, we used parallel analysis, Velicer's minimum average partial procedure, and random variables to determine the appropriate number of principal components to retain in a clinical sample (N = 227). We retained 3 oblique dimensions that broadly emphasized (a) general distress, (b) elevated mood and dominance, and (c) substance abuse and psychopathy. In Study 2, we applied the same uniform criteria and procedures to 5 previously published samples and conducted orthogonal vector matrix comparisons to determine how congruent 3- and 4-dimensional structures were across samples. Results suggested the PAI has 3 dimensions that are highly congruent across samples. Using Morey's normative sample, we provide the formulas needed to compute T scores for each component so they can be used in clinical work with patients. We discuss clinical implications and directions for future PAI research.
