Figure 2. Schematic illustrations for thinking about the positive statistical correlations among psychiatric disorders.

Panel A shows a Correlated-Factors Model, which is the original structural model used in psychopathology research. In this model, the latent variables represent variance shared (or comorbidity) among the disorders within each of three spectra: Internalizing, Externalizing, Psychotic Experiences. The high correlations between these latent traits suggested the possibility that they could be accounted for by a general factor of psychopathology, labelled “p”. Panels B1 and B2 show two ways to conceptualize p, respectively. The Higher-Order Factor Model (B1) shows that there is a second-order factor arising from the Internalizing, Externalizing, and Psychotic Experiences first-order latent variables; p represents the variance shared among the three spectra. The Bi-Factor Model (B2) shows that there is one common liability to these three forms of psychopathology (p) as well as a set of independent factors which influence a smaller subset of symptoms and disorders. The use of the term “bi-factor model” is an unwieldy historical and statistical legacy; it harkens back to the early days of intelligence research which first proposed a general factor that is common to all items on a test (g) and more specific factors that are common to a smaller subset of related items and are thought to represent independent cognitive modules. Panel C shows an alternative conceptualization of positive correlations among disorders. Here there is no common cause, but instead there is a causal network in which disorders influence each other (straight arrows) and themselves over time (looped arrows) (131). Panel D shows that rather than a cause of disorders, p is constructed from the disorders, reflecting a common manifestation that is shared by the different disorders.