On top or underneath: Where does the general factor of psychopathology fit within a dimensional model of psychopathology?

Background: Dimensional models of psychopathology are increasingly common, and there is evidence for the existence of a general dimension of psychopathology (‘p’). The existing literature presented two ways to model p: as a bifactor or as a higher-order dimension. Bifactor models typically fit sample data better than higher-order models,and are often selected as better fitting alternatives but there are reasons to be cautious of such an approach to model selection. In this study, the bifactor and higher-order models of p were compared in relation to associations with established risk variables for mental illness. Methods: A trauma-exposed community sample from the United Kingdom (N= 1,051) completed self-report measures of 49 symptoms of psychopathology. Results: A higher-order model with four first-order dimensions (Fear, Distress, Externalizing, and Thought Disorder) and a higher-order p dimension provided satisfactory model fit, and a bifactor representation provided superior model fit. Bifactor p and higher-order pwere highly correlated (r = .97) indicating that both parametrizations produce near equivalent general dimensions of psychopathology. Latent variable models including predictor variables showed that the risk variables explained more variance in higher-order p than bifactor p. The higher-order model produced more interpretable associations for the first-order/specific dimensions compared to the bifactor model. Conclusions: The higher-order representation of p, as described in the Hierarchical Taxonomy of Psychopathology, appears to be a more appropriate way to conceptualise the general dimension of psychopathology than the bifactor approach. The research and clinical implications of these discrepant ways of modelling p are discussed.