What Kind of Thing a Diagnosis Is

There is no blood test for depression. No scan identifies schizophrenia, no genetic marker diagnoses ADHD, and no laboratory finding confirms or excludes any category in the Diagnostic and Statistical Manual.

What a DSM diagnosis is, instead, is a list: a set of symptoms, a minimum number of them, a duration, and a requirement that they cause distress or impairment. Meet five of nine criteria for two weeks and you have major depressive disorder. Meet four and you do not.

Those lists were produced by committees of experts, debated, voted on, and revised between editions. Asperger's syndrome existed as a diagnosis from 1994 and was removed in 2013. Homosexuality was a mental disorder until 1973, when it was removed following a vote of the American Psychiatric Association's membership.

This is not a scandal, and framing it as one gets the history wrong. Medicine has worked this way for most of its existence, and syndromal description, noticing that these symptoms cluster, is how you begin, not a failure to do something better. Tuberculosis was a syndrome before it was a bacillus.

What it does mean is that a diagnosis is a useful summary and a shared vocabulary, rather than the identification of an underlying thing that has been found.

Which produces a specific error, and it is worth naming because clinicians make it as readily as patients. "Why is she so low?" "Because she has depression."

That explains nothing. Depression is the name for the cluster of which low mood is a member. The label cannot be the cause of the symptoms that define it, any more than "having a headache" explains a headache. It feels explanatory because a named thing sounds like an entity, and the feeling of having an explanation is exactly what stops further questions. About what is happening in this person's life, what maintains the pattern, what changed recently, what they have stopped doing.

The empirical pressure on the categories comes mainly from comorbidity. Roughly half of people meeting criteria for one DSM disorder meet criteria for another, and the overlap is systematic rather than random: anxiety and depression co-occur enormously more than chance would predict. If these were distinct diseases, that would be surprising. If they are overlapping descriptions of a smaller number of underlying dimensions, it is exactly what you would expect.

This is the main driver behind the alternatives now being developed: the Hierarchical Taxonomy of Psychopathology, which models symptoms on continuous dimensions rather than in boxes, and the NIMH's Research Domain Criteria, which organises research around functional systems like threat response and reward learning instead of around diagnoses. Neither has replaced the DSM, and both exist because the categories are straining.

The categories also proliferate, and the reason is structural rather than sinister. Subdividing a heterogeneous category produces more homogeneous groups, which genuinely helps research. Specialists advocate for their area. Patients benefit from having a recognised name, because names get you services, accommodations and insurance. There is no comparable force pushing categories back together. The count has risen from around a hundred in DSM-I to several hundred, and nothing in that process requires anybody to be acting badly.

Finally, the study that shaped public understanding of psychiatric diagnosis more than any other, and which has largely fallen apart.

David Rosenhan's "On Being Sane in Insane Places", published in Science in 1973, reported that eight healthy people gained admission to psychiatric hospitals by claiming to hear a voice saying "empty", "hollow" or "thud". Once admitted they behaved normally, and none was detected as a fake; all were discharged with diagnoses in remission, after an average of nineteen days.

It was devastating and enormously influential, feeding directly into deinstitutionalisation and into a generation's scepticism about psychiatry.

Susannah Cahalan spent years trying to find the eight pseudopatients for a book. She could identify two. Rosenhan's own hospital record, which she obtained, described symptoms considerably more severe than the paper reported, including that he had said the voices were saying things could be put in his head. She found one participant whose experience had been positive and whose data appear to have been excluded. And she could find no trace of the other five.

None of this establishes that the study was fabricated, and it does mean that the most famous demonstration in this area cannot bear the weight that has been put on it for fifty years. The interesting question it raised (what does a diagnosis actually track) remains open, and now needs somebody to answer it properly.

One thing that should not be concluded from any of this: that because a category is constructed, the suffering is not real. Whatever the diagnostic system is doing, the person cannot get out of bed. A constructed category can organise treatment, enable research, secure services and give someone a language for what is happening to them. The critique is narrower than the use it usually gets put to.