In the summer of 2016, I sat across from a clinician who asked me a series of questions. Had I been sleeping poorly? Had I lost interest in activities I once enjoyed? Did I feel worthless or guilty? Had I experienced changes in appetite, concentration, or energy?

I answered yes to enough of the questions to receive a diagnosis: major depressive disorder.

What no one asked—what the instrument in front of them was not designed to ask—was whether the depression I was describing was the whole story, or one face of something larger and more dangerous that the checklist could not see. It could not see the bipolar disorder beneath the depressive episode. It could not see the syphilis infection rewriting my neurology from the inside. It saw what it was built to see: symptoms, arranged in clusters, matched to a label.

The checklist those clinicians were using came from a book. That book—the Diagnostic and Statistical Manual of Mental Disorders, or DSM is the subject of today’s post. Last week, I described what it felt like to wear a diagnosis that didn’t fit. This week, I want to look at the manual that cut the coat.

The Man Who Built Psychiatry’s Bible

Allan Frances was once called “perhaps the most powerful psychiatrist in America” by The New York Times. He chaired the task force that built the DSM-IV (1994)—the book that determined which forms of suffering counted as mental illness, which treatments insurers would cover, and which diagnoses would follow millions of patients through the rest of their lives.

Years later, the psychologist Gary Greenberg asked Frances a simple question: could he define what a mental disorder actually was?

Frances’s answer has followed him ever since. “There’s no definition of a mental disorder,” he said. “And it’s bullshit. I mean, you can’t define it.”

He later clarified that he meant the bright line is impossible to draw—that good clinicians must exercise judgment where science cannot yet provide certainty. But the admission stood. The bible of the field had been built on a foundation its own architects could not fully describe.

How We Got Here

The DSM-III, which appeared in 1980, was designed to save psychiatry from itself. Its architect, Robert Spitzer, had watched the field spend decades under the influence of psychoanalysis—a framework that could explain almost anything and predict almost nothing. “Psychiatry was regarded as bogus,” Spitzer said. He believed the field would only survive if it was “accepted as a medical discipline.”

His solution was radical. The DSM-III excised the language of Freud entirely—no more neuroses, no more unconscious motivations, no more speculation about underlying causes. In its place, Spitzer installed something that looked like science: a taxonomy of discrete disorders, each defined by a checklist of observable symptoms. If a patient met, for example, five of seven criteria, she received a diagnosis.

The result, as Greenberg described it, was a transformation. The spiral-bound, 134-page DSM-II became a 500-page hardbound tome that “vastly expanded the manual’s scope, turning it into an entirely new psychopathology of everyday life.”

The taxonomy kept growing—from 106 disorders in 1952 to more than 300 in the current edition. Allan Horwitz and Jerome Wakefield, in The Loss of Sadness, demonstrated how this expansion had turned normal grief and sadness into clinical depression by stripping context from the criteria. Frances himself—who would go on to become the most prominent critic of the DSM-5, waging what Greenberg called a “scorched-earth campaign” against his successors—described the accumulation as the “wholesale imperial medicalization of normality.”

But the deeper problem was not the expansion. It was the logic underneath it.

The Gap Between Agreement and Truth

What Spitzer achieved was reliability—the extent to which clinicians using the same checklist would arrive at the same diagnosis. Before the DSM-III, two psychiatrists examining the same patient might produce two entirely different diagnoses. The checklist fixed that. It gave the field a common language.

What reliability is not, however, is validity—the extent to which the diagnosis describes an actual disease. A checklist can produce consistent agreement about a label without that label corresponding to anything real in the body. As Greenberg put it, a doctor who diagnosed strep throat entirely on the basis of symptoms—sore throat, fever, swollen glands—without ever running a culture would be practicing bad medicine. A doctor who diagnosed depression entirely on the basis of symptoms was practicing standard psychiatry.

Spitzer’s attempt to make psychiatry look more like medicine had, paradoxically, “only highlighted the differences.”

This is the gap I lived inside for years. The labels I received were reliable. Any clinician using the DSM’s criteria would likely have arrived at the same diagnosis. They were not valid. They did not correspond to what was actually happening in my brain. The agreement was real. The disease it pointed to was not.

Biology Never Read That Book

By the early 2000s, the gap between psychiatry’s categories and the biology they claimed to represent had become impossible to ignore from within the field itself.

Thomas Insel, after thirteen years leading the National Institute of Mental Health (NIMH)—the largest funder of mental health research in the world—announced that the agency would reorient its research away from DSM categories entirely. The manual’s diagnoses, he said, were still based on nothing more than consensus about symptom clusters. He told The New York Times: “As long as the research community takes the DSM to be a bible, we’ll never make progress. Biology never read that book.”

That sentence appeared in 2013, more than three decades after the DSM-III had promised to ground psychiatry in science. The promise remained substantially unfulfilled. The field had achieved agreement. It had not achieved understanding.

What the Lantern Can and Cannot Do

None of this means the DSM should be abandoned. Frances himself—after all the blistering critiques—argued that its categories are “useful constructs,” provisional frameworks that capture how people commonly suffer, even if they cannot explain why. For many patients, a diagnosis brings the first coherent name for a lifetime of experiences that had previously resisted language. That matters. Even a rough syntax is better than silence.

But there is a version of a psychiatrist who knows what the DSM is—who treats it as a first approximation, a working hypothesis, an invitation to keep asking—and a version who does not. The version who does not is the one who handed me a label fitted for someone else’s suffering and closed the inquiry. The version who knows would have understood that naming is not the same as knowing—that between the description and the disease, there is always a gap, and that patients live inside that gap, sometimes for years, sometimes nearly dying there.

The DSM is a lantern, not a destination. It illuminates a small circle around the clinician’s feet. What it cannot reach, it tends to declare does not exist. I know something about living in that darkness. I know what it costs.

The question—for psychiatry, for families, for anyone who has watched a diagnosis land on someone they love and wondered whether it fit—is not whether we need a map. We do. The question is whether we can learn to hold the map more lightly, to remember that the territory is always larger than the map can show, and to keep walking when the lantern’s light runs out.