A mother sat across from me recently, her twelve-year-old beside her, and asked a question I have heard in one form or another hundreds of times over two decades of practicing psychiatry: “Can we just get her tested? I want to know, scientifically, what she has.”

She wasn’t asking for a diagnostic interview. She was picturing something closer to what happens elsewhere in medicine when a symptom is uncertain: a scan, a panel, a definitive number on a printout. Usually, no such test exists, and the reason runs deeper than simply lacking the right instrument. Even where a biological marker of some kind can be measured, it still has to be checked against a diagnostic category, and most psychiatric categories are not single conditions. They are syndromes: clusters of symptoms grouped together for clinical convenience, often encompassing patients with different underlying mechanisms. A test is only as meaningful as the label it is being asked to confirm.

Start with the images most people have actually seen: the brightly colored brain scans that circulate in documentaries and magazine spreads, seeming to show exactly where a disorder lives. Functional MRI measures blood oxygenation, an indirect proxy for brain activity during a task; PET tracks radiolabeled tracers. Both are legitimate research tools. Neither is a routine diagnostic instrument. They cannot reliably distinguish a pathological brain pattern from a compensatory or medication-related one in an individual patient, and they do not steer everyday treatment decisions. Their value lives almost entirely in the research lab, not the exam room.

Structural imaging, ordinary MRI and CT, has a real clinical role, but not the one patients expect. We order these scans to rule something else out: a tumor, a stroke, a degenerative process masquerading as a mood or behavior change. What they do not do, in routine practice, is diagnose depression, bipolar disorder, ADHD, autism, or schizophrenia. Group-level research has found average structural differences across populations with certain diagnoses, enlarged ventricles in some patients with schizophrenia being the textbook example, but a finding that only holds up averaged across hundreds of people tells us almost nothing about the one patient in front of us. Part of why is the diagnosis itself: “Schizophrenia” is not one biologically uniform condition but a syndrome that likely encompasses several, so an average drawn from that group blurs together patients who may have little in common beyond meeting the same symptom checklist.

EEG suffers a similar fate in the public imagination. Patients sometimes describe it, with a note of hope, as “the brain wave test,” as though wiring the scalp might finally settle a diagnostic question. In practice, it detects seizure activity and other electrical abnormalities, not anxiety, bipolar disorder, or ADHD, even when those conditions are on the differential for the referral. There are rare, genuine exceptions worth naming: Temporal lobe epilepsy, in particular, can present with irritability, derealization, or behavioral change that closely mimics a primary psychiatric disorder, and EEG has real value in catching it. But that, too, is a story of ruling out a neurological condition wearing a psychiatric mask, not of EEG diagnosing anxiety, bipolar disorder, or ADHD in their own right.

Then there is the phrase that has probably misled the public more than any other in modern psychiatry: chemical imbalance. Serotonin, dopamine, and norepinephrine are unquestionably important to mood, motivation, and attention. But no blood test measures serotonin and thereby diagnoses depression, just as none measures dopamine and diagnoses ADHD or schizophrenia. Blood work is genuinely useful (it can uncover thyroid disease, anemia, or hormonal problems that mimic psychiatric symptoms), but that is narrower than what most patients have in mind when they ask for it. Pharmacogenetic panels, increasingly marketed direct-to-consumer, mostly describe how a body metabolizes drugs, not whether a given medication will actually work.

Even neuropsychological testing, which can look the most rigorous of all (hours of structured tasks, a report dense with percentile scores), does not function the way referring physicians and patients often expect. Performance on these tasks is shaped by anxiety, sleep, pain, medication, motivation, and educational background. A low attention score does not automatically mean ADHD any more than slowed processing automatically means a mood disorder. But the deeper limitation sits beneath any individual score: Those percentiles are only meaningful relative to a norm, and the norms clinicians lean on are built around diagnostic categories (ADHD, autism spectrum disorder) that are themselves heterogeneous. If the reference group used to define “typical” or “impaired” performance is itself a mix of patients with different underlying conditions lumped under one syndromic label, the comparison inherits that imprecision. A percentile score can look exact to two decimal places while resting on a norm that was never especially coherent to begin with. The testing can still meaningfully inform care, particularly for conditions with a narrower, better-defined phenotype such as dementia, brain injury, or specific learning differences. It rarely settles a broader psychiatric diagnosis on its own, however precise the numbers look on the page.

The consequences are not merely semantic. Patients delay treatment waiting for a test that isn’t coming, or pay out of pocket for panels that promise more certainty than they deliver. Others lose confidence in an otherwise sound diagnosis once they learn no scan confirmed it. For physicians outside psychiatry, the stakes are just as real: A patient referred for “testing” is often really asking us to validate a diagnosis that was, and still is, reached the way psychiatric diagnoses have always been reached, through history, longitudinal observation, and collateral information, not a printout.

None of this makes psychiatry unscientific, or its tools worthless. It means those tools differ from what the rest of medicine has trained patients to expect, and that the field has done a poor job explaining why. It also means the missing test is not simply a gap that better technology will eventually fill. A validated biomarker, should one emerge for some slice of psychiatric illness, would still need a diagnostic category precise enough to make the match meaningful, and psychiatric nosology, built around symptom clusters rather than shared mechanism, is not yet built for that kind of precision. A diagnosis reached through careful history and observation over time is not a lesser kind of medicine for lacking a lab value to point to. But letting a patient believe a test exists, or is coming, when the deeper obstacle is the diagnosis itself, serves no one, least of all the patient still waiting for an answer.

Nebyou Belay is a practicing psychiatrist with twenty years of experience treating children, adolescents, and adults. He is affiliated with the Brown University Department of Psychiatry and practices at Rhode Island Hospital and Newport Hospital.

He writes about the practice of psychiatry itself: how diagnoses are made, how they’re communicated to patients and families, and how psychiatric language is understood (and often misunderstood) in the wider culture. His work touches on the science and the limits of psychiatric assessment, the boundaries of what medication and diagnosis can realistically offer, and the growing overlap between clinical language and everyday identity.

Across all of it, his aim is the same: to help patients, families, and the public engage with mental health more honestly, with less oversimplification and less noise. He is currently writing a book, The Myths and Misunderstandings of Psychiatry, and is a contributor to KevinMD. He shares updates on LinkedIn.