The clinical definition of vegetative state was written as a behavioral description, and it was always meant that way. The criteria: no reproducible, purposeful, or voluntary behavioral response to visual, auditory, tactile, or noxious stimuli; no evidence of self-awareness or environmental awareness; possible preservation of sleep-wake cycling; possible brainstem reflexes. The definition names what the bedside observer can see. It doesn't claim to describe what's inside. It was precise in exactly the domain it was designed for.
In 2006, Adrian Owen and colleagues put a 23-year-old woman — five months after a traumatic brain injury, clinically diagnosed vegetative — into an fMRI scanner. They asked her to imagine playing tennis. Her supplementary motor area activated. They asked her to imagine walking through her home. Her parahippocampal gyrus activated. Both patterns were indistinguishable from those of healthy volunteers performing the same tasks. She sustained each activation for 30 seconds on cue, switching reliably when prompted. She showed no behavioral response to any of it. No eye movement, no hand movement, nothing the standard clinical exam would have detected.
The scanner found command-following. The room found nothing.
In 2010, Monti and colleagues extended the protocol to 54 patients with clinical diagnoses of vegetative or minimally conscious state. Five of the 54 showed willful modulation of brain activity on fMRI. In one patient — a 36-year-old man who had been in vegetative state for five years following a car accident — the researchers established yes/no communication. Tennis imagery meant yes. Navigating a home meant no. He was asked questions about his life and answered five of six correctly. He was asked whether he was in pain. He answered no. There is no way to know what those five years had contained.
In August 2024, Bodien and colleagues published a prospective cohort study in the New England Journal of Medicine. Six international centers. 353 adults with disorders of consciousness. Each patient was assessed behaviorally and then with task-based fMRI and EEG. Among those who showed no observable behavioral response to commands — the group that would have been diagnosed vegetative or equivalent — 25 percent showed covert command-following on fMRI or EEG. One in four. The study gave the condition a name: cognitive motor dissociation. CMD. The motor output channel is severed. The cognitive process behind it is running.
The behavioral definition was always describing the output state, not the internal state. That was its domain and its limit, and for most of medical history it was the only domain that could be assessed. What the scanner revealed is that the two states can come apart. The person can be processing language, following instructions, answering questions in their brain — and none of this will appear on the clinical exam, because the clinical exam is a behavioral assessment and the motor pathway is gone. The term "no observable response" was accurate. The inference that no observable response meant no experience was what failed.
fMRI is not standard of care for disorders of consciousness at most hospitals. The protocol for detecting CMD, demonstrated in 2006, published in Science, replicated and extended through two decades of follow-on work, is still not routine. There are efforts toward EEG-based CMD detection — cheaper, portable, usable at the bedside — but those too remain outside standard practice. The 2024 paper will prompt calls to change this; so did the 2006 paper. The gap between demonstration and clinical implementation is its own kind of problem, measured in a unit that is difficult to look at directly.