Hook
Brain scans of people diagnosed with ADHD show three distinct patterns. One group has extreme deficits in a specific brain network. Another shows milder, more scattered differences. The third falls in between. Same diagnosis, different brains.
This isn’t a breakthrough. It’s how diagnostic categories evolve. You start with a broad label — “attention deficit hyperactivity disorder” — and then measurement gets better. You can see more. And what you see is variation that the single label hides.
The question: what do you do with that variation? Do you split the category into subtypes? And if you do, what does that buy you — and what does it cost?
What Subtyping Is
Subtyping is the move from “you have X” to “you have flavor A of X, not flavor B.” Medicine does this constantly.
Diabetes was one disease until 1936, when researchers noticed some patients responded to insulin and others didn’t. Now: Type 1 (autoimmune, insulin-dependent) and Type 2 (insulin-resistant, often linked to metabolic factors). Same symptoms at the surface — high blood sugar, fatigue, increased thirst — different mechanisms underneath.
Breast cancer was one disease until molecular testing revealed different tumor types. HER2-positive tumors respond to targeted drugs like trastuzumab. Triple-negative tumors don’t. Knowing the subtype changes treatment from the first decision.
The point of subtyping: finer categories improve prediction. If I know your subtype, I might predict which treatments work, which complications you’ll face, how the condition evolves. I can study mechanisms separately instead of averaging across groups that work differently.
The trade-off: more complexity. Harder to communicate. Risk of over-splitting — creating distinctions that don’t matter clinically, fragmenting knowledge into categories too narrow to be useful.
What The Scans Show
The ADHD study used brain scans to measure connectivity — how different regions of the brain communicate with each other. Think of it as mapping the wiring diagram. Some regions talk a lot; some stay quiet; some form tight clusters.
Three patterns emerged among people diagnosed with ADHD. One group showed severe deficits in the frontoparietal network — the system that coordinates attention, working memory, and cognitive control. Another group showed milder, more distributed differences across multiple networks. The third group sat in between.
These aren’t arbitrary clusters drawn from noise. They mapped onto differences outside the scanner. The severe-deficit group had worse working memory scores, more impulse control problems, more functional impairment in daily life. The milder group looked more like people without ADHD on some measures but still met diagnostic criteria.
Brain connectivity matters because it’s a window into how a brain organizes itself. ADHD has always been diagnosed by behavior — inattention, hyperactivity, impulsivity. But behavior is the output. Connectivity is part of the machinery producing that output.
Why This Matters And Doesnt
What does subtyping buy you?
Better prediction. If your subtype links to specific connectivity deficits, researchers might predict which medications work, which skills training helps, which comorbidities appear. A person with severe frontoparietal deficits might respond differently to stimulant medication than someone with distributed, milder differences.
Better research. If you study three subtypes separately, you might find mechanisms invisible when you average them together. Maybe one subtype involves dopamine signaling and another involves norepinephrine. Lumping them hides that.
What subtyping does NOT buy you yet: a scan-based diagnostic test. These findings aren’t validated across labs, aren’t accessible outside research settings, aren’t necessary for clinical diagnosis. A psychiatrist still diagnoses ADHD by talking to you, observing behavior, reviewing history.
It doesn’t buy you a treatment algorithm. We don’t yet know if subtype predicts medication response reliably enough to guide prescribing. The scans might correlate with outcomes, but correlation at the group level doesn’t guarantee prediction at the individual level.
And it doesn’t buy you certainty. These three subtypes are provisional. Five years from now, better scans or different analysis methods might split the categories differently — or reveal that splitting this way doesn’t improve outcomes.
The Lumping Splitting Tension
Every diagnostic system faces this tension. Lump or split?
Lumping is simpler. One label, easier to communicate, easier to study in large groups, easier to build guidelines around. But it hides variation. Two people with “ADHD” might have almost nothing in common beyond meeting the same checklist.
Splitting is more precise. Subtypes capture real differences in mechanism, trajectory, treatment response. But precision costs you. Harder to operationalize — does every clinic need a brain scanner? Harder to communicate — explaining three ADHDs is harder than explaining one. Risk of fragmenting knowledge into silos too narrow to be useful.
This plays out everywhere. Depression: one category or a dozen subtypes (melancholic, atypical, anxious, psychotic)? Autism: spectrum or distinct profiles? Personality disorders: ten categories or dimensional traits?
The question isn’t “which is right.” It’s “what does each choice cost and buy?” A category is a tool. You judge it by whether it helps you predict, treat, and understand better than the alternative.
Right now, ADHD is a lumpy category. Subtyping is a move toward splitting. Whether these three subtypes stick depends on whether they outperform the single label in clinical use — not just in research papers.
Close
Diagnostic categories aren’t discovered intact. They’re built, tested, revised. ADHD subtypes are a draft, not a destination. The scans don’t reveal hidden truth — they offer a new way to carve the territory. Whether these three patterns become standard depends on whether they help doctors and patients make better decisions than the old single label. That’s the test. And it’s how every diagnostic category evolves — slowly, messily, always provisional.