Assessing the External Correlates of Alternative Factor Models of the Psychopathic Personality Inventory–Short Form Across Three Samples

Abstract

This study investigated various theoretically relevant correlates of a short form of the Psychopathic Personality Inventory (PPI; Lilienfeld & Andrews, 1996) using archival data from large undergraduate, foster care, and juvenile offender samples. External correlates of the 2 primary scales (PPI–I and PPI–II) and the Coldheartedness subscale were for the most part consistent with prior findings. Analyses for an alternate factor model in which the Fearlessness subscale loaded onto PPI–II (rather than PPI–I) resulted in relatively few substantial changes to the pattern of correlations with criterion measures, but a third factor that included the Coldheartedness and Carefree Nonplanfulness subscales functioned differently than Coldheartedness alone in these data.

Acknowledgment

Research involving the foster youth sample was supported by National Institute on Mental Health grant, R01MH 61404.

Notes

Unfortunately, the alternate factors identified in the Neumann et al. (2008) study and PPI–R manual were not supplemented by examining any associations with theoretically relevant external correlates, which is one of the seminal means of addressing construct validity (see, e.g., Cronbach & Meehl, 1955; Messick, 1980).

Several researchers have cautioned against rigid adherence to factor-analytic results, which can reflect spurious findings due to numerous factors, such as restricted variance stemming from individual sample characteristics (Floyd & Widaman, 1995) and correlated residuals that are difficult to anticipate and account for in CFA approaches (Hopwood & Donnellan, 2010). Other authors have raised concerns about “replacing our previous idolatry of p values with an idolatry of fit indices” (Westen & Rosenthal, 2005, p. 410; see also Tomarken & Waller, 2003), the utility of which increasingly are being questioned in relation to the assessment of personality inventories as newer statistical approaches are being developed and refined (see, e.g., Marsh et al., 2010). Hopwood and Donnellan (2010) recently urged researchers to continue to utilize various methods to evaluate the construct validity of personality measures, including methods such as criterion-related validity, and to “think more critically … about how common factor methods are used to evaluate the structure of personality inventories” (p. 25).

Alpha levels for PPI–I and PPI–II, respectively, were .83 and .78 in the undergraduate sample, .62 and .78 in the foster youth sample, and .65 and .78 in the juvenile justice sample.

Variables with base rates < 25% (GAD, PTSD, ASPD, substance abuse, employment or school attendance, arrest between ages 17 and 19, socioeconomic status, any mental illness diagnosis, potential psychotic symptoms, and ADHD) were examined further using receiver operating characteristic (ROC) curves, given that low BRs can significantly attenuate the magnitude of biserial correlations. These results generally provided stronger support for the magnitude of the PPI–I/PPI–II effect size differences. For instance, in predicting GAD diagnoses, PPI–I had an Area Under the Curve (AUC) value of .38 (SE = .06), whereas PPI–II had an AUC value of .68 (SE = .04). For ASPD, PPI–I had an AUC of .54 (SE = .04) whereas PPI–II had an AUC of .71 (SE = .03). For arrest between ages 17 and 19, PPI–I had an AUC of .51 (SE = .04) and PPI–II had an AUC of .73 (SE = .03).

Using ROC analyses, PPI–I had an AUC of .39 (SE = 03) and PPI–II had an AUC of .69 (SE = .03) for this variable.

Additionally, researchers might consider using alternative self-report measures (e.g., Witt, Donnellan, & Blonigen, 2009) that have recently been developed to operationalize Fearless Dominance (PPI–I) and Impulsive Antisociality (PPI–II).

Somewhat oddly, the correlations in the college student sample were relatively similar, despite the fact that there was no evidence for this factor in the exploratory factor analysis.