Abstract
Although psychopathy is a correlate of recidivism, including violent offending, the inclusion of antisocial behavior items in the Psychopathy Checklist (PCL) assessment instruments has been debated. Further, the latent factor structure underpinning the PCL measures has not been comprehensively validated in Australia. This study examined the construct validity of the PCL:SV in a sample of 192 Australian male violent offenders using structural equation modeling. The incremental validity of including the antisocial behavior items for predicting recidivism was also assessed. Approximately one-third of the sample had high scores on the PCL:SV. Construct analyses indicated strongest support for two distinct three-factor models of PCL psychopathy, with the established two- and four-factor models demonstrating poorer fit. The Antisocial factor improved predictive accuracy incrementally relative to the other three factors (Interpersonal, Affective, and Lifestyle). Area under the curve analyses revealed predictive accuracy for recidivism only for factors containing the antisocial behavior items. Findings contribute to the construct and predictive validity of the PCL measures and are discussed relative to contemporary forensic risk assessment practice.
Notes
1 Testlets are used in measurement models to reflect item dependencies (Steinberg & Thissen, 1996). Cooke and Michie (Citation2001) proposed that the descriptions for items in the PCL measures loading on a testlet prior to the respective factor have substantial content overlap (e.g., descriptions for Item 1 (Superficial) and Item 2 (Grandiose)), and therefore, items are more accurately represented in this manner in structural models.
2 PCL psychopathy is increasingly considered as a dimensional construct rather than a categorical label (Walters, Duncan, & Mitchell-Perez, 2007), and the use of cut-scores is discouraged for making clinical decisions with legal implications about a person (Balsis, Busch, Wilfong, Newman, & Edens, 2017; Hare et al., Citation2018).
3 Walters et al. (Citation2011) found no significant difference between using factor scores or factor score composites – taking into account the ordinal nature of PCL items – for the PCL measures in predicting recidivism. Factor scores were used for analyses in the current article.