The Latent structure of the Dark Triad : unifying Machiavellianism and psychopathy

The Dark Triad (DT) has emerged as a popular extension to the extant literature on personality psychology. The DT is comprised of three similar and yet distinct constructs: Machiavellianism, narcissism, and psychopathy. Recent research has criticized the DT for suffering from a number of measurement problems, including a failure of existing Machiavellianism measures to adequately capture the construct. Prior research has chiefly been conducted using domain level information, thus neglecting item level information. The present thesis investigates the possibility of empirically distinguishing between DT constructs with particular focus on Machiavellianism and psychopathy, using item level information. In Article I, the Dirty Dozen inventory is analyzed using latent variable models in order to replicate its structural properties. Results indicate that narcissism is more independent from Machiavellianism and psychopathy than the latter two are with each other. These results provide initial evidence that Machiavellianism and psychopathy are empirically indistinguishable. In Article II, the Short Dark Triad (SD3) is analyzed in a series of factor analytic models testing whether Machiavellianism and psychopathy can be jointly modeled. The models fit similarly across two and three factor solutions, thus favoring a two factor model based on the principle of parsimony. In Article III, the SD3 is modeled using Item Response Theory in order to analyze how much information the SD3 provides across the latent trait continuum. Results indicate that SD3 items from the domain of Machiavellianism may be less severe in content than psychopathy items, thus yielding differential item endorsement rates, albeit along a unitary dimension. In Article IV, the overarching literature on Machiavellianism is extended by analyzing Machiavellianism items from 7 different measures using an hierarchical analysis. This analysis is subsequently compared with expert rated Five Factor model prototype scores o