Perils of Partialing Redux: The Case of the Dark Triad

The Dark Triad (DT) is a collection of overlapping aversive personality profiles constituting psychopathy, narcissism, and Machiavellianism. Debates remain regarding the optimal way to examine the unique outcomes associated with each construct, with several scholars advocating partialing these profiles in multiple regression analyses (i.e., removing their shared variance). The present paper details the pitfalls inherent in this approach by comparing the convergence and divergence of relations derived from raw and residualized DT composite scores. In Sample 1 (N = 393), DT scores were examined to determine the extent to which their raw and residualized components manifested similar relationships with the Five-Factor Model and the DSM-5 Section III personality disorder traits. In Sample 2 (N = 542), the same approach was taken in relation to an array of associated behaviors (e.g., antisocial behavior, promiscuity). Findings from Samples 1 and 2 demonstrate that the use of residualized (vs. raw) coefficients presents important interpretative challenges for both narcissism and Machiavellianism. This study illustrates the substantial interpretive difficulties that can arise when using findings from residualized analyses (e.g., multiple regression) to build nomological networks around Dark Triad constructs. We argue that bivariate relations be given preferential treatment, given their more direct ties to the assessments, and that if multivariate approaches are to be used, they must be accompanied by strong theory about the components of DT constructs. General Scientific Summary Partialing dark triad constructs from one another, so as to examine the unique variance associated with relevant criteria, is a common used but potentially problematic approach to understanding these important personality constructs. The removal of overlapping variance among the three DT constructs results in widely disparate constructs and nomological networks for Machiavellianism and narcissism, such that their basic trait profiles and associations with key outcomes change dramatically.