Psychiatric diagnosis is indeed imperfect, subjective and not based on pathophysiology or causation (Nature 496, 416–418; 2013). But quests for biological markers and dimensional approaches, such as the Research Domain Criteria project mentioned, are unlikely to be clinically appropriate solutions because they run the risk of oversimplifying complex psychiatric illnesses and introducing predicative logic.

For instance, people with post-traumatic stress disorder (PTSD) or psychosis may show similar abnormalities of frontal-lobe function on a magnetic resonance imaging scan, just as those with anxiety or psychosis may show similar overactivation of the amygdala in the brain. But the underlying reasons for these superficial similarities are different, and so are the treatments. For example, lithium is not an effective treatment for psychosis, anxiety or PTSD, but it is near-curative for some people who meet current diagnostic criteria for bipolar disorder.

The genetics of some conditions may overlap (for example, schizophrenia, autism and bipolar disorder), as do the genetics of, say, multiple sclerosis and Crohn's disease. However, these are distinct clinical conditions that require different interventions.

Caution and healthy scepticism are essential before embracing fashionable trends to revise psychiatric diagnosis.