In Praise of Scientific Open Relationships: A Commentary on Caspi et al. (2026)

Caspi et al. (2026) marshal impressive evidence to demonstrate that mental disorders so commonly co-occur—across couples, generations, and lifespans—that the usual practice of conducting research on a single mental disorder should no longer be the norm in clinical psychological science. They recommend that every aspect of psychopathology research, from study design and measurement to publication and clinical translation, be revised accordingly, such that “single-disorder loyalty” is no longer the starting point of the research enterprise. I agree. Research focused on a single diagnosis, or its symptoms, is poorly suited to understanding psychological phenomena that are rarely contained within nosological boundaries.

But, probably like many readers, I am also genuinely interested in some clinical symptoms in some populations, and not that interested in others. (Callous and aggressive children? Endlessly fascinating to me. Adults who can’t stop worrying? Yawn.) The causes and consequences of mental disorders are rarely specific to individual disorders, but the intrinsic interest necessary to sustain a research program often is quite specific. Is the alternative to an entirely monogamous intellectual relationship with one’s favorite clinical disorder a dilettantish flirtation with every disorder? Or is there a research equivalent of a happy open relationship, in which widening one’s attentions beyond the bounds of a single disorder also deepens and enriches one’s relationship with the primary intellectual love object? Happily, my experience in psychiatric genetics suggests scientific open relationships can thrive—but they require a commitment to collaboration that is still too rare in clinical psychology.

Psychiatric genetics, as a field, has had some time to grapple with Caspi et al.’s central point, because it provided foundational evidence, in the form of genetic correlations, that the causes of mental disorders cannot be presumed to be specific to any individual disorder (Bulik-Sullivan et al., 2015; Consortium et al., 2018). Genetic correlations, unlike ordinary phenotypic correlations, can be estimated for any two phenotypes that have been the focus of a genome-wide association study (GWAS), even if they have never been measured in the same individuals or manifest at opposite ends of the lifespan. Moderate to high genetic correlations among mental disorders have now proven to be ubiquitous: Much of the genetic architecture of any one disorder overlaps, sometimes nearly entirely, with other disorders (Lee et al., 2019; Mallard et al., 2022). Genes are almost never “for” just one thing.

Crucially, research studies that leverage the genetic sharing among mental disorders, rather than ignoring it, can make discoveries that would not have been possible if genes had as much single-disorder loyalty as scientists do. Our work on externalizing spectrum disorders is an instructive example. GWASs of externalizing disorders, such as alcohol use disorder, opioid use disorder, conduct disorder, and attention deficit hyperactivity disorder, have steadily increased in sample size (Deak & Johnson, 2021; Demontis et al., 2023), but the available data for any one disorder is still too small to identify more than a tiny fraction of the specific genes that confer risk for that disorder. In contrast, a multivariate genetic model, which jointly analyzes multiple externalizing-related disorders, behaviors, and personality traits, has substantially more statistical power than any single-disorder study (Dick et al., 2024; Karlsson Linnér et al., 2021). As a result, we can discover many more genes important for all externalizing disorders than were previously discoverable for any one disorder individually—in some cases, up to a hundred times more. Put differently, I would never have learned so much about the genes affecting child conduct problems, the form of psychopathology that I am most interested in, if I had not learned to sometimes ignore conduct problems and play with other data instead (Tanksley et al., 2024).

The astute reader will note that the line of research I just described aims to identify genetic associations multiple externalizing disorders, not with a general factor of all psychopathology (‘p’). This decision is grounded in previous research: Although internalizing, externalizing, and thought disorders are indeed genetically correlated with each other, existing evidence suggests that a general factor of psychopathology does not cohere at the level of individual genetic variants (Clapp Sullivan et al., 2024). Thus, I also appreciate the authors’ caveat that “our claim that psychopathology research should stop studying one mental disorder at time is not tantamount to claiming that mental disorders should only be studied at the broadest level.” How broad or narrow one’s scientific aperture needs to be is an empirical question, and the answer to it might differ depending on whether one is examining, e.g., genetic causes, neurobiological correlates, early attachment experiences, or economic contexts.

Psychiatric genetics been able to produce transdiagnostic research because it has developed strong norms around collaboration, communication, and credit-sharing. Data, results, methods, and code are all shared as widely and openly as possible, and authorship lists can be hundreds of people long, with multiple shared senior authors. These practices, unfortunately, are not the norm in much of clinical psychology. Caspi et al. recommend that training programs “teach students about the intergenerational and developmental context of all mental disorders,” but if we really are to abandon single-disorder loyalty, then more sweeping changes to how we train students are necessary. Too much of the field is dominated by faculty who have defined their entire professional identity around a single diagnosis, who have trained their students primarily to collect new data but not to share it (and certainly not to analyze other people’s data), and who seem as if they would rather crawl across broken glass than share credit with other senior investigators (partly because they would be penalized for tenure and promotion if they did). If the causes and consequences of mental disorders were specific to individual disorders, then perhaps labs that function as cottage industries for small-N, single-PI, case-control studies could make substantial scientific progress. As the target paper so powerfully demonstrates, however, that is not the reality of the disorders we claim to care about. No longer being in a scientifically monogamous relationship with a single disorder doesn’t mean you need to divorce it entirely, but it does mean you will need to learn to share.