Towards a genuinely medical model for psychiatric nosology

Two patients walk into the same clinic on the same day, describe the same symptoms, and walk out with different diagnoses. That isn't a fluke — it's a structural feature of how psychiatric diagnosis works. For decades, the field's answer has been to improve criteria, create sharper checklists, and establish cleaner categories. Randolph Nesse and Dan Stein believe this answer is aimed at the wrong problem. Their argument is that psychiatry has been measuring itself against a model of medicine that most of medicine does not actually use. To understand why, you need to look at the history. By the late nineteen seventies, psychiatry had a credibility problem. Diagnoses varied wildly between clinicians, and the field was seen as more art than science. The Diagnostic and Statistical Manual of Mental Disorders, or DSM-III, published in nineteen eighty, was the fix: out went psychoanalytic theory, and in came checklists. For major depression, for instance, at least five of nine possible symptoms had to be present for a minimum of two weeks. The DSM-IV, released in nineteen ninety-four, added that those symptoms had to cause clinically significant distress or impairment. The shift worked in its own way. It made large-scale epidemiological studies possible, allowed neurobiologists to search for pathology tied to reliably defined conditions, and provided regulatory agencies and insurers with a common language. However, the very studies that operationalized diagnosis kept turning up the same four issues. Comorbidity is rampant — most people who meet criteria for one disorder also qualify for additional diagnoses. Heterogeneity is enormous — two patients with no symptoms in common can both legitimately receive a diagnosis of major depression. Boundaries between disorder and normality are fuzzy — diagnostic groups often aren't separated from each other or from healthy populations by any clear gap. And biomarkers are essentially absent. After three decades of neuroimaging and laboratory research, not one of the main DSM mental disorders can be validated by a biological test. Allen Frances, who chaired the DSM-IV Task Force, put it bluntly: "the disappointing fact is that not even one biological test is ready for inclusion in the criteria sets for DSM-V." So the field attempted to fix it. A twenty-nine member DSM-5 Task Force, coordinating six study groups and thirteen work groups, prepared revisions for two thousand thirteen — merging substance dependence and substance abuse into "substance use disorder," separating agoraphobia from panic disorder, and exploring quantitative severity dimensions. None of it satisfied the critics, and few expected it to. The deeper hope has been biomarker-driven diagnosis: to find the biological signature that maps cleanly onto the clinical category. Nesse and Stein acknowledge this might eventually work for some disorders, but three decades of negative results are hard to ignore. Brain-circuit models earned serious attention too, with the idea that psychiatric conditions might be better defined by the neural pathways involved. The problem is that evolved brain systems do not behave like engineered circuits. They have indistinct boundaries, massively distributed functions, and redundant connections, which is precisely why neuroimaging has such low sensitivity and specificity for DSM categories. The Research Domain Criteria, or RDoC, framework made a more constructive move: it organized research around five functional domains — negative valence, positive valence, cognitive systems, social processes, and arousal and regulatory systems — intersecting seven units of analysis from genes to behavior. That's a genuine improvement in framing. However, Nesse and Stein note that the RDoC still "remains committed to the hope that most psychiatric diagnoses will eventually be based on biomarkers." The core assumption hasn't changed. Here's where Nesse and Stein make their turn. The problem, they argue, isn't that psychiatry has failed to be medical enough. It's that psychiatry has been emulating the wrong slice of medicine. Consider a cough. A cough isn't defined by its cause — it's defined by its function, which is to clear foreign material from the respiratory passages. That functional understanding is what tells a clinician when a cough is protective and when it's pathological. Patients who cannot cough are likely to die from pneumonia. Therefore, medicine treats cough first as a signal worth investigating and considers suppression only after determining what's causing it. Now think about congestive heart failure. It has multiple causes, blurry boundaries, and nonspecific biomarkers — exactly the features that make psychiatric diagnoses seem scientifically suspect. But no one claims congestive heart failure isn't a real medical category. What makes it useful is that it's grounded in a clear understanding of what the cardiovascular system is supposed to do. The rest of medicine is full of syndromes defined by failures of functional systems or failures of feedback control, not by single discoverable causes. Psychiatry's failures look much less like failure when you hold them up against that broader picture. The part of that picture most missing from psychiatry, Nesse and Stein argue, is an understanding of what emotions are actually for. Emotions like fear, anxiety, and low mood aren't glitches. They are evolved, adaptive defenses — they adjust physiology, cognition, behavior, and motivation in ways that helped our ancestors survive and reproduce. The very existence of systems that regulate these responses on and off confirms they serve real functions. Once you see them that way, a particular logic becomes clear. Nesse and Stein call it the smoke detector principle. False alarms are expected and even built in because the cost of expressing a defense — the metabolic cost of anxiety and the behavioral cost of low mood — is usually small compared to the potentially catastrophic cost of failing to respond to a real threat. Just as a smoke detector that only fires when there's an actual fire would be dangerous, an anxiety system that only activates under genuine danger would be inadequate. The system is calibrated toward sensitivity, not precision. This changes the diagnostic question entirely. If emotions are defenses calibrated by context, then whether a given emotional response is pathological depends on whether the situation warrants it. A cough prompts a search for what's causing it. An emotion should prompt a search too. However, psychiatry has largely done the opposite: emotions of sufficient duration and intensity get categorized as disorders regardless of the situation that prompted them. Nesse and Stein note that sixty-one percent of DSM diagnoses already include criteria about context — but that context is included without the functional framework that would explain what to do with it. To make this actionable, the paper proposes a simple situational coding approach. Rate each case on two axes, each scored from none through severe: one for trait vulnerability and one for the intensity of current situations likely to arouse the emotion. Nesse and Stein walk through two vignettes. A college student whose depressive symptoms are tied to losing his girlfriend and an unwanted living situation gets coded as moderate on current situation and none on trait vulnerability — treatment planning looks very different than if you ignored that context entirely. A young man with similar symptoms but a history of childhood abuse and lifelong isolation gets coded as severe on trait vulnerability and mild on current situation — a different clinical picture and different therapeutic priorities. The point isn't that the coding scheme is the final answer. It's that this kind of reasoning — grounding the diagnosis in what the emotion is responding to — is exactly what the rest of medicine does with pain, fever, and cough, and psychiatry has been missing it. This also reframes the field's most frustrating features. Comorbidity isn't a sign that the categories are wrong — it's what you'd expect from nonspecific defenses activated by many different problems. Heterogeneity isn't a failure of diagnostic rigor — it's what you'd expect from syndromes with multiple upstream causes. Fuzzy boundaries aren't an embarrassment — they're the norm for any system built on adaptive responses that exist on a continuum with normality. Nesse and Stein's conclusion isn't a surrender. They are not saying the DSM is fine as it is. They are saying the dissatisfaction with it has been fueled by an unrealistic expectation — that mental disorders should resemble strep throat, with a specific cause, a clean diagnostic test, and a targeted treatment. That expectation describes a minority of medical conditions. The rest of medicine operates with heterogeneous syndromes, overlapping categories, and multiple causes, and does so productively. The goal for psychiatry isn't to find the magic biomarker that makes depression look like a single-cause disease. It's to understand depression the way cardiologists understand heart failure — as a syndrome defined by what has gone wrong with a system we understand, evaluated in the context that shaped it. The untidy landscape is the landscape. The task is to learn to read it better. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.