A Review and Comparison of the Reliabilities of the MMPI-2, MCMI-III, and PAI Presented in Their Respective Test Manuals

Review of Wise, E., Streiner, D. & Walfish, S. (2010).  “A Review and Comparison of the Reliabilities of the MMPI-2, MCMI-III, and PAI Presented in Their Respective Test Manuals.”  Measurement and Evaluation in Counseling and Development, 42(4), 246 – 254.

The Minnesota Multiphasic Personality Inventory (MMPI) test originally was developed in 1939 and amended from time-to-time until now.  It has 567 true-false questions.  These are parsed into various “scales” including clinical scales; sub-scales; content scales; supplementary scales; and critical item scales.  There also are scales purporting to measure the test’s validity.  The basic claim underlying the MMPI is incredible – that a series of true-false questions can accurately assess personality traits.  The questions originally were normed against a group of middle-class persons in Minnesota, who self-selected as relatives of in-patients at a psychiatric facility (this population since has been somewhat expanded).  Many of the questions are stereotypes (“do you like to read mechanic’s magazines”), compound, assume premises and are phrased as double-negatives or colloquialisms.  In some cases a sub-scale is based on endorsement of as few as a half-dozen items.  Nonetheless the MMPI is widely used and has become the de-facto standard of personality testing.

This review article statistically assesses various aspects of the MMPI, particularly with regards to validity and reliability.  The authors surveyed the last decade’s worth of literature.  Conventional interpretation of the MMPI is that an elevated t-score (in excess of around 65 – 70) on a particular scale indicates potential psychopathology for the construct the scale is supposed to measure, subject to caveats such as elevation on related scales, a clinical interview and the like.  The authors of this study believe a better measure is around 90.  None of the MMPI scales meet this criterion.  Even lowering the value to 70, few of the scales are reliable.  In particular the clinical scales are much less reliable than the content and supplementary scales.  The authors conclude the MMPI is only weakly reliable.  In my opinion no series of true-false questions can measure the complex dimensions of human personality, so I concur with this conclusion.