The Basics: The PAI is a personality test designed to provide information relevant for a clinical diagnosis and to measure both the severity and breadth of any psychological defects discovered. The test is composed of 344 questions whose responses are rated on a 4 point scale ranging from false, not at all true, to very true. The test takes about an hour with someone with a 4th grade reading level or above to complete. The exam is designed to be given in writing with the test taker sitting in a desk or a table by himself.
Strengths and Weaknesses: The results of the PAI have been shown as useful in predicting future violence and recidivism rates. However, this test is relatively new and its validity isn’t as established as some other personality tests, like the MMPI-2.
Critical Issues: The test requires a 4th grade reading level and may not be appropriate for people who fall below that level. In addition, because the PAI is a newer exam, its results are easier to challenge in court. (text from wordpress.com)
The Personality Assessment Inventory (PAI) is a self-report 344-item personality test that assesses a respondent’s personality and psychopathology. Each item is a statement about the respondent that the respondent rates with a 4-point scale (1-“Not true at all, False,” 2-“Slightly true,” 3-“Mainly true,” and 4-“Very true“). The PAI is often used in forensics and corrections settings. The test construction strategy for the PAI was primarily deductive and rational. It shows good convergent validity with other personality tests, such as the Minnesota Multiphasic Personality Inventory and the Revised NEO Personality Inventory.
The PAI has a number of strengths for applied psychological assessment. First, the 4-point scale contributes to greater scale reliability and validity, as it provides respondents with the opportunity to give nuanced ratings of themselves (as opposed to a true-false scale). Second, it is relatively economical, assessing most of the constructs that are widely considered important in clinical personality assessment with only 344 items. Third, nearly all of the PAI items are readable at the 4th grade level. Brevity and straightforward item wording makes it accessible to those with lower levels of intelligence and reading ability. Fourth, the responses from each individual’s profile can be compared to multiple large samples. The first sample consists of 1,000 people with similar demographic characteristics (e.g., age, gender, ethnicity) to the U.S. Census data. Comparison with this group is useful to detect and estimate the severity of clinical problems relative to the average person. The second sample consists of 1,265 psychiatric patients. Comparison with this group helps assess the severity of psychopathology among other patients. There is also a sample of 1200 offenders from multiple correctional settings and a further sample of 1051 college students. Finally, there are also a number of psychometric (statistical) strengths of the PAI, importantly including content validity and discriminant validity.
Review of the Personality Assessment Inventory [2007 Professional Manual]
Thorpe, G. L., & Dawson, R. F. S. (2010).
In R. A. Spies, J. F. Carlson, & K. F. Geisinger (Eds.), The eighteenth mental measurements yearbook. Retrieved from http://marketplace.unl.edu/buros/
DESCRIPTION
The Personality Assessment Inventory (PAI) is a 344-item, multiscale self-report inventory for use in the clinical assessment of adults. Respondents select an answer to each item from a 4-point scale: F = (false, not at all true), ST = (slightly true), MT = (mainly true), and VT = (very true). The typical administration time is 40 to 50 minutes. The test produces scores on 22 discrete scales: 4 validity scales, 11 clinical scales, 5 treatment scales, and 2 interpersonal scales. The validity scales assess for inaccuracies in and distortions of responding attributable to Inconsistency, Infrequency, Negative Impression, and Positive Impression. Six supplemental validity indicators include a Rogers Discriminant Function, drawn from weighted combinations of 20 scale scores, designed to distinguish genuine from simulated profiles.
The clinical scales assess Somatic Complaints, Anxiety, Anxiety-Related Disorders, Depression, Mania, Paranoia, Schizophrenia, Borderline Features, Antisocial Features, Alcohol Problems, and Drug Problems. Most of the clinical scales also yield subscale scores reflecting recognized components of the syndromes. For example, the Antisocial Features scale encompasses the subscales Antisocial Behaviors, Egocentricity, and Stimulus-Seeking. The Treatment Consideration Scales address features important to treatment planning that overlap across diagnostic categories: aggressive attitudes and behaviors, thoughts and ideas related to death and suicide, current life stressors, perceived lack of support, and attitudes toward treatment-including “unwillingness to participate actively in treatment, a refusal to acknowledge problems, and a reluctance to accept responsibility for problems in one’s life” (professional manual, p. 46). The Interpersonal scales assess the levels of dominance and warmth that respondents display in their relationships with others.
Test users may also examine critical items and interpret various supplemental indexes, profile codetypes, and modal cluster profiles. The first 160 items of the PAI have received a level of validation as a short form of the instrument.
Raw scores on the PAI scales and subscales are plotted on a multi-sided profile form that indicates the corresponding T scores. These values were derived from the standardization sample of 1,000 U.S. Census-matched adults from the general community. The test author suggests that T scores of 70 and above may indicate clinically significant problems.
The PAI inventory itself still bears only the original copyright dates 1990 and 1991 and thus has not been changed, but the professional manual (2007) is a second edition that provides “updated information about the technical aspects of the test norms, reliability, and validity” (professional manual, p. 1).
DEVELOPMENT
The PAI scales were selected to reflect five constructs that the developer believed were “most pertinent to a broad-banded assessment of mental disorder … (a) validity of an individual’s responses, (b) clinical syndromes, (c) interpersonal style, (d) treatment complications, and (e) characteristics of the individual’s environment” (professional manual, p. 106). Potential test items were generated by a team of researchers, faculty members, practitioners, graduate students, and others drawing from the research literature, classic texts, diagnostic manuals, and clinical experience. The resulting item pool of over 2,200 was eventually distilled into the 344 items of the PAI.
Item selection and scale development of the PAI took place in two stages. In the first stage, each of the potential items was evaluated for the conceptual meaning of its content. This evaluation consisted of: (a) ratings of the quality of the item and appropriateness of the subscale assignment by members of the research team, (b) a sorting study of item content by experts in the specific fields, and (c) a review of item content to eliminate those that could be offensive to male or female respondents or to those with particular racial, religious, or ethnic group identifications. The second stage was a two-tiered empirical evaluation of the 776 items that survived the first stage. In the first tier, the alpha version of the test was administered to a sample of college students for the purposes of “evaluating item distributions, item social desirability, possible gender effects, and studies of the manipulations of response set to investigate the effects of malingering or faking on item responses” (professional manual, p. 123). This process resulted in a beta version of the test with 597 items, next administered to a heterogeneous sample of individuals from community and clinical settings to examine the items’ internal consistency, specificity, and internal validity in addition to differential responding by groups varying in age, gender, race, or ethnicity. The 344 items that constitute the PAI itself were those that demonstrated the optimal psychometric properties and appeared to best reflect the constructs that the test was intended to measure.
TECHNICAL
The PAI was standardized on three samples of respondents: the general community sample noted earlier (N = 1,000), a sample of patients from 69 clinical sites (N = 1,265), and college students from seven universities (N = 1,051).
Reliability
The data for internal consistency showed that the median alpha coefficients for PAI scales and subscales were .81 (general community sample), .86 (clinical sample), and .82 (college student sample). Test-retest reliability was assessed with additional community and college samples who were retested after 24 or 28 days. Taken together, those two groups produced test-retest correlations ranging from .79 to .92 for the clinical scales.
Validity
At 133 pages, the chapter entitled “Validity Studies” is by far the longest in the professional manual. The test author emphasizes convergent, construct, and discriminant validity, and has amassed data not only from his own validation studies but also from hundreds of others that have appeared in print since the PAI was first issued in 1991. Validity data are presented for all five groups of scales and indexes. The chapter documents the correlations of PAI components with other well-known inventories and structured clinical interview protocols, and presents studies of various criterion groups and their differential responses to the PAI elements. For example, among clinical participants, PAI Depression correlates .66 with the MMPI D scale (Scale 2), and PAI Schizophrenia correlates .55 with the MMPI Sc scale (Scale 8).
Modern test theory methodology was used in concert with classical test theory procedures in some applications. In a tour de force of documentation, the test author provides in this chapter 26 figures and 50 tables that, among other things, delineate the relationship between the indexes and scales of the PAI and dozens of external instruments as responded to by thousands of individuals in a variety of settings.
Standardization
Tables for converting raw scores to T scores are found in six of the manual’s appendixes. The U.S. Census-matched community sample of 1,000 respondents is the standardization sample proper. Similar tables for clinical, African American, senior (age 60 or over), and college student samples are provided for comparison purposes only. Test users in many settings and with a diversity of respondent populations are likely to find the raw and T scores of appropriate comparison groups tabulated within these appendixes and in the chapter on validation.
COMMENTARY
The test author paid elaborate attention to detail in developing and validating the PAI, which has also been extensively researched by others. The most cursory of informal surveys revealed half a dozen articles on the PAI in a handful of recent issues of Psychological Assessment, the topics being short versus full forms of the PAI, clinician-assessment versus self-assessment using the PAI, predicting sex offender adjustment, assessing antisocial personality disorder, and back irrelevant responding (completing later test items in a less valid manner than earlier items) as a PAI validity indicator (two articles). As the test author notes (professional manual, p. 1), the PAI is ranked among the top four personality tests in terms of its general popularity and widespread professional acceptance. Recent citations attest to its utility in assessing critical clinical variables in a broad spectrum of populations in clinical and forensic settings (e.g., Caperton, Edens, & Johnson, 2004; Guy, Poythress, Douglas, Skeem, & Edens, 2008).
The level of detail involved in scoring supplemental indexes can be excessive. To take the most extreme example, calculating the Rogers Discriminant Function involves multiplying each of 20 scale and subscale scores by a different weighting, the value of which is expressed to eight places of decimals-in effect, working at the level of a 100 millionth of a raw score unit. In the example provided in the professional manual (Figure 2.4, p. 17), this results in a sum of -1.35172, which is then interpolated on the profile form between the values of -1.25 and -1.50 to give a T score of 47. When we performed the same calculation after rounding the weighting values to 2 places of decimals we obtained a sum of -1.34 and the same T score.
A final quibble-the test’s label as the Personality Assessment Inventory may mislead some professionals who do not follow the early tradition of the Minnesota Multiphasic Personality Inventory (MMPI0; Hathaway & McKinley, 1943) in drawing parallels between (if not even equating) personality and psychopathology. Many would expect a personality inventory to consist of scales measuring extraversion, neuroticism, and at least the other components of the familiar “big five” traits rather than a list of constructs with names similar to those of mental disorders. Yet, with his typical thoroughness, the test author provides a table of correlations between the PAI scales and the NEO Personality Inventory (NEO-PI) domains for a community sample (professional manual, Table H.9, p. 358), and the relationships seem consistent with expectation: For example, PAI Anxiety correlates .75 with Neuroticism, PAI Paranoia correlates -.54 with Agreeableness, and PAI Borderline Features correlates -.31 with Conscientiousness.
SUMMARY
The PAI is a self-report inventory assessing personality and psychopathology that is convenient to administer, psychometrically sound, extensively researched, and suitable for use in clinical, forensic, and other applied settings. It can be recommended as an alternative to the MMPI-2 (Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989; 11:244) for its relatively short administration time and its inclusion of scales directly measuring a respondent’s amenability to treatment.
REVIEWERS’ REFERENCES