The controversies embroiling the latest version of the Diagnostic and Statistical Manual of Mental Disorders (DSM; American Psychiatric Association 2013) culminated in the head of the National Institute of Mental Health (NIMH) proclaiming that the DSM would cease to be the essential framework for grant applications submitted to the Institute (Insel Reference Insel2013). NIMH would now be “re-orienting its research away from DSM categories” to fund proposals targeting trans-diagnostic mechanisms as embodied in the Research Domain Criteria (RDoC) initiative (Insel et al. Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn, Sanislow and Wang2010). Remarkably, even the chief architect of the new manual had seemingly lost faith in the categorical approach. He and his co-authors announced, “We are now coming to the end of the neo-Kraepelinian era” (Regier et al. Reference Regier, Narrow, Clarke, Kraemer, Kuramoto, Kuhl and Kupfer2013, p. 68).

As the diagnostic manual was undergoing its periodic overhaul, a team of clinical psychometricians developed a hierarchical dimensional alternative to the DSM's categorical system (Kotov et al. Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller, Moffitt, Morey, Mullins-Sweatt, Ormel, Patrick, Regier, Rescorla, Ruggero, Samuel, Sellbom, Simms, Skodol, Slade, South, Tackett, Waldman, Waszczuk, Widiger, Wright and Zimmerman2017). Capping this structure was the p factor (Caspi et al. Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel, Meier, Ramrakha, Shalev, Poulton and Moffitt2014), the general factor of psychopathology akin to g, the general factor of intelligence. And just as intelligence researchers have sought to identify the biological referent of g (e.g., Deary Reference Deary2012), so have some psychopathologists suggested that the p factor “may have a physical reality” (Lahey et al. Reference Lahey, Van Hulle, Singh, Waldman and Rathouz2011, p. 187), perhaps genetic. However, it is questionable whether the p factor amounts to an empirical discovery about a major cause of psychopathology. As van Bork and her colleagues have emphasized (Van Bork et al. Reference Van Bork, Epskamp, Rhemtulla, Borsboom and van der Maas2017), any dataset consisting of highly intercorrelated measures is mathematically bound to yield a general factor, even if the factor has no existential referent independent of the data themselves.

The categorical and dimensional perspectives have long been the only nosological games in town (McNally Reference McNally2011, pp. 184–211). Yet, the landscape of our field dramatically changed when Borsboom, Cramer, and their associates introduced the network approach to psychopathology (Borsboom Reference Borsboom2017; Borsboom & Cramer Reference Borsboom and Cramer2013; Cramer et al. Reference Cramer, Waldorp, van der Maas and Borsboom2010). On the heels of the first few empirical network studies (e.g., Cramer et al. Reference Cramer, Borsboom, Aggen and Kendler2012; McNally et al. Reference McNally, Robinaugh, Wu, Wang, Deserno and Borsboom2015; Robinaugh et al. Reference Robinaugh, LeBlanc, Vuletich and McNally2014), a tsunami of studies has appeared with many more in the pipeline (for reviews, see Fried et al. Reference Fried, van Borkulo, Cramer, Boschloo, Schoevers and Borsboom2017; McNally Reference McNally2016).

Pushback has been inevitable, especially from those favoring traditional approaches. As Bringmann and Eronen (Reference Bringmann and Eronen2018) observe, many psychometric critics have argued that network and latent variable models are mathematically equivalent. Although that may often be the case, they are certainly not ontologically equivalent (cf. geocentric versus heliocentric models of the solar system; Galilei Reference Galilei, Shea and Davie1615/2012). Just as it makes a great deal of difference whether the earth revolves around the sun or vice versa, so does it matter whether symptoms arise from a latent variable having an existential referent (e.g., undetected brain tumor) or not. Treating a headache with aspirin is inadvisable in the former case, but not in the latter (Borsboom & Cramer Reference Borsboom and Cramer2013).

Analyzing two epidemiologic datasets, proponents of the latent dimensional approach launched a spirited critique of network analysis in psychopathology (Forbes et al. Reference Forbes, Wright, Markon and Krueger2017a; Reference Forbes, Wright, Markon and Krueger2017b). Their central claim was that “networks” do not replicate – or at least not as much as one would have thought. Unfortunately, their work was hobbled with errors, and when Borsboom et al. (Reference Borsboom, Fried, Epskamp, Waldorp, van Borkulo, van der Maas and Cramer2017) correctly re-did their analyses, the substance of Forbes et al.’s critique evaporated.

In their excellent target article, Borsboom et al. convincingly criticize contemporary biological reductionist accounts of mental disorders. The purpose of my commentary is to elaborate on several themes adumbrated by these authors.

Given the unsuitability of common-cause biological reductionist theories of mental disorder, what roles might biology play in the psychopathology enterprise? One possibility would be to use biological measures as grouping variables in network analyses. For example, some studies have shown that people with at least one copy of the short allele of the serotonin transporter promoter polymorphism exhibit more symptoms of depression in stressful circumstances than do those with two copies of the long allele (e.g., Caspi et al. Reference Caspi, Sugden, Moffitt, Taylor, Craig, Harrington, McClay, Mill, Martin, Braithwaite and Poulton2003). Van Borkulo et al.’s (Reference Van Borkulo, Boschloo, Borsboom, Pennix, Waldorp and Schoevers2015) network comparison test applied to depression symptoms of subjects grouped by genotype (or other biological variables) could illuminate how such risk variables predict network topologies in psychopathology.

Another possibility would be to unpack the biology of signs and symptoms rather than syndromes. The classic psychiatric distinction between form and content may help us here (e.g., Taylor Reference Taylor1981, pp. 2–7). That a person hears voices in the absence of anyone speaking denotes the form of psychopathology, whereas what the voices are saying denotes its content. The intentional content – “aboutness” – of a symptom renders it elusive to biological reductionist explanations, as Borsboom et al. emphasize. On the other hand, the formal features of symptoms such as auditory hallucinations, experiences of influence, thought broadcasting, and so forth have better prospects for biological elucidation. And some symptoms appear to lack intentional content altogether. Consider posttraumatic stress disorder: Although flashbacks and nightmares possess intentional content, as they are about the trauma, emotional numbing and exaggerated startle do not.

Finally, the discipline of biology itself has been undergoing its own network makeover (Barabási & Oltvai Reference Barabási and Oltvai2004). Network genomics (Forst Reference Forst2002), network neuroscience (Bassett & Sporns Reference Bassett and Sporns2017), and network medicine (Barabási et al. Reference Barabási, Gulbahce and Loscalzo2011) are flourishing, non-reductionist enterprises. The goal is to build bridges between networks emerging at these levels of analysis rather than to reduce “higher” levels to “lower” ones. And now the “network takeover” (Barabási Reference Barabási2012, p. 14) has reached psychopathology, inspired by Borsboom and his colleagues.