The CAPP Provides Support for a Bi-Factor Model of Psychopathy

Martin Sellbom – Australian National University
David J. Cooke – Glasgow Caledonian University
Stephen D. Hart – Simon Fraser University

Psychopathy is an important clinical and forensic psychopathology construct; however, its optimal conceptualization continues to be a source of significant controversy. The Comprehensive Assessment of Psychopathic Personality (CAPP; Cooke, Hart, Logan, & Michie, 2012) considers 33 personality traits that integrate historical and contemporary conceptualizations of the disorder. The current study examined the internal structure of self-ratings the CAPP traits in a large international sample of community-dwelling participants (N D 719; 52% women). Results indicated that a bi-factor (one general factor, three residual bi-factors) model representing global psychopathy, as well as residual factors of boldness/ emotional stability, emotional detachment, and disinhibition, best fit the data. Associations with nine additional self-rated items revealed a generally expected pattern of convergent and discriminant validity. Finally, a Spearman rank-order correlation between CAPP item loadings on the global psychopathy factor and prototypicality ratings by experts (Kreis, Cooke, Michie, Hoff, & Logan, 2012) was .76, reflecting substantial content validity as well as agreement about relative importance of psychopathy traits using widely different conceptual and empirical procedures.

psychopathy, personality, Comprehensive Assessment of Psychopathic Personality, confirmatory factor analysis

Psychopathy is a personality disorder that is characterized by callousness, a lack of remorse, impulsivity, and manipulativeness. It has serious implications for interpersonal functioning and the risk of crime and violence. However, despite extensive attempts, researchers and clinicians continue to struggle to conceptualize and operationalize psychopathy for adequate assessment or diagnosis. The Comprehensive Assessment of Psychopathic Personality (CAPP) was developed to encompass 33 personality traits of psychopathy into layman terms. The tool is divided into six domains: Attachment, Behavioral, Cognitive, Dominance, Emotional, and Self. The current study also included “foil” items which are typically not associated with psychopathy to examine the divergent validity of the CAPP items. The “foil” items were: Dependent, Perfectionistic, Conscientious, Considerate, Strange, Restrained, Shy, Cautious, and Self-Conscious. The current study consisted of 719 participants who made self-ratings using the CAPP through an online survey service.
While the CAPP has shown moderate to strong levels of construct validity in the past, the present study set out to examine the internal structures of the self-ratings of symptoms in the CAPP using a large, international sample. Their second objective was to “determine the degree to which there is observed agreement between the prototypically ratings (using the forensic expert sample) and latent factor loadings (an indication of which items are best explained by the latent psychopathy domain(s)),” (p. 175).
The current study found that, “across evaluations of various factor models, there seemed to be consistent support for the existence of a robust higher-order factor reflecting general psychopathy. There was also, however, evidence of specific sub-factors representing lower-order facets. The best overall factor model—one which fit as well as or better than any other model, was reasonably parsimonious, was readily interpreted, and demonstrated some stability across gender—was a bi-factor model comprising one general factor and three residual and correlated sub-factors. Of further importance, a rank order of the latent factor loadings observed for the general psychopathy factor was substantially correlated with that of the prototypicality ratings…indicating substantial convergence of relative importance of CAPP items to the theoretical construct of psychopathy using two highly divergent methodologies” (p. 178).
“The finding of a 1-factor model of psychopathy is relatively novel in the psychopathy literature, but the current findings indicate that a hierarchical model of a global psychopathy factor with residual sub-facets reflecting boldness/emotional stability, emotional detachment, and disinhibition best fit self-ratings of CAPP items” (p. 178).

The authors first tested six models of psychopathy proposed in the literature. However, they all demonstrated only poor to acceptable fit. In order to find a better model fit, the authors used exploratory factor analyses and determined that a bi-factor model best fit the data. This model consists of a general psychopathy factor and three residual bi-factors. Under the general psychopathy factor, “the items in which the latent psychopathy factor predicted the most variance were from the Attachment and Dominance domains, whereas those that loaded the least meaningfully on this factor were those from the Cognitive and Emotional domains…Inspection of the factor loadings for the three bi-factors suggested that they reflect aspects of boldness/ emotional stability, emotional detachment, and disinhibition” similar to previous established literature of the triarchic model of psychopathy. (p.176)
The study examined whether the resulting bi-factor model (one general factor, three bi-factors) were equivalent across men and women. There were non-meaningful differences in the CAPP items to factor loadings. The researchers concluded that a “partial invariance model represented support for metric measurement equivalence across genders,” (p. 177).

“The current findings indicate that a hierarchical model of a global psychopathy factor with residual sub-facets reflecting boldness/emotional stability, emotional detachment, and disinhibition best fit self-ratings of CAPP items”(p. 178). The bi-factor resembled some aspects of traditional triarchic psychopathy “model constructs of boldness (i.e., sense of invulnerability, lacks anxiety, emotionally stable, pleasure-driven), meanness (lacking in emotional depth, attachment, and empathy), and disinhibition (lack of planfulness and concentration, unreliability), respectively. Moreover, initial validity data were supportive as well, especially for the emotional stability/boldness and disinhibition factors. More specifically, the former was negatively associated with ratings of shyness, self-conscious, and being cautious (i.e., a socially confident, self-assured, and risk-taking style), whereas disinhibition was primarily associated with lacking in perfectionism and conscientiousness,” (p. 178).
For clinicians connecting this study to general psychopathy theory, “affective-interpersonal” features, meanness, or callous-unemotional traits were at the core of the disorder. “Across both latent modeling findings and expert ratings, which substantially converge, we can begin to move closer to what we might consider to be at the core of the psychopathic personality. As further research along these lines continue to accumulate, potential revisions to the CAPP concept map, especially with regard to items that both perform poorly from a psychometric perspective and have low prototypicality (especially across cultures), might be in order,” (p. 179).

“Furthermore, the latent psychopathy structure underlying the CAPP items was generally (albeit not completely) invariant with respect to gender. By and large, factor loadings were invariant across genders, but there were some meaningful differences when thresholds were constrained to be equal. It is noteworthy that the majority of the CAPP items that were variant at this level were those that are typically deemed less prototypical of psychopathy, and most belong to either the cognitive or emotional domain. Given their lack of centrality to the psychopathy construct (per ratings) and differential performance in contributing to average psychopathy scores across genders, further research should examine the degree to which they are necessary for the conceptualization of the psychopathic personality” (p. 178).

“The emotional detachment factor was only meaningfully associated with restraint, potentially indicating disaffiliativeness in social relationships. More research is needed, with more extensive extratest data, to further elaborate on the meaningfulness of the general vs. residual factors within this hierarchical model,” (p. 178).

As always, please join the discussion below if you have thoughts or comments to add!