The short version
- Thirteen of 41 patients who showed no response at the bedside learned, within days of a brain injury, to change the pitch of a beeping tone by imagining a hand movement.
- Those 13 recovered consciousness and independence far more often than the 28 who could not control the tone, though the groups are small and the study cannot say the tone helped.
- The system uses the ordinary EEG wires already on most ICU patients, which is why its makers think it could one day spread beyond research hospitals.
A patient lies in a neurocritical care bed a few days after a severe brain injury. Their eyes may open, but they do not follow a finger, squeeze a hand or answer a question. On the chart they are “clinically unresponsive”. Headphones sit over their ears, and a beeping tone plays. A recorded voice asks them to try to open and close their hand, over and over, and to make the beep change.
For about one patient in three in a new Columbia University study, the beep changed.
The paper, published October 22 in the journal Brain, describes a brain-computer interface built from equipment that is already in almost every intensive care unit. It is a small study with a careful claim. It is also one of the clearest demonstrations yet that some people who look absent are listening.
The problem it tries to solve
Doctors judge consciousness mostly by behaviour: does the patient follow a command, track with their eyes, respond to pain? Over the last decade a body of research has shown that this misses people.
The Columbia group, led by the neurologist Jan Claassen, has been one of the main contributors. In a 2022 study of 193 patients with acute brain injury, it played spoken motor commands while recording EEG, and used machine-learning software to look for brain responses to them. Twenty-seven patients, 14%, showed responses their bodies could not. Doctors call this cognitive motor dissociation, or CMD. Within a year, 41% of those patients made a full recovery, against 10% of the rest.
In 2024, a seven-centre study across the US, the UK, Belgium and France, with Claassen among the authors, tested 241 unresponsive patients with EEG or functional MRI. As many as one in four showed the same hidden responses.
Finding a hidden response is one thing. Those tests are one-way: the patient is asked to do something, and the researchers look afterwards for a sign that they tried. The patient never learns whether anyone noticed. The new study asks a different question. If the patient can hear their own effort, can they steer it?
How the tone works
The setup is simple, which is the point. Standard EEG wires, the kind already used to watch for seizures in the ICU, record brain activity. A bedside computer reads the signal in real time. When it detects the pattern of an attempted hand movement, the pitch of the beep in the headphones shifts. A recorded voice encourages the patient when the pitch moves and congratulates them when it reaches a target.
“EEG is used in clinical practice every day in the ICU,” Claassen said in Columbia’s announcement. In his account, the only sophisticated part is the software.
The system is self-paced. Rather than waiting for a cue, the patient can try at any moment, and the sound responds when they do. The abstract calls it a “one-way” interface: the patient can move the tone, but the tone does not yet carry a message such as yes or no.
What happened
The team tested 41 clinically unresponsive patients within days of their injuries, along with 22 critically ill patients who were conscious and 8 healthy volunteers for comparison. Each had a brief training period.
- All 8 healthy volunteers learned to control the tone.
- 9 of the 22 conscious patients did, 41%.
- 13 of the 41 unresponsive patients did, 32%.
The unresponsive patients were slower. The median time for their brain signal to move the tone was about 50 seconds, against about 17 seconds for conscious patients and 16 for the volunteers. Separately, 8 of the 41 tested positive for CMD on the team’s established EEG test. According to Columbia, every one of them could also control the tone.
The striking numbers come from what happened next. Of the 13 patients who controlled the tone, 4 regained consciousness before leaving hospital. Of the 28 who did not, 1 did. At 12 months, 6 of the 13 had reached functional independence on the standard recovery scale the team used. Only 2 of the 28 had.
The study also surveyed patients and families afterwards. All 37 people who answered supported using such systems in future. Their suggestions were practical and moving: use a loved one’s voice instead of a recording, add culturally familiar details, and give patients a way to signal pain or hunger.
The honest read
This is a meaningful result, and it is easy to read too much into it.
First, the groups are small. The odds ratios in the paper are large, about 21 for recovering consciousness and about 10 for independence at a year, but the confidence interval on the first runs from 2.3 to 930.7. That width is what 13 against 28 people gives you. The direction is clear. The size is not.
Second, the study is observational and run at one hospital. Patients who could move the tone were very likely less badly injured, or recovering already. Controlling the beep is a sign of a brain that was going to do better. Nothing here shows that practising with the tone helped anyone recover, and the authors do not claim that. Their claim is that control “strongly predicted” outcomes.
Third, two-thirds of unresponsive patients did not control the tone, and that does not mean they were unaware. An aware patient could plausibly fail a task like this through fatigue, sedation, hearing problems or a bad moment, and the abstract does not sort out why the others failed. A negative result is not a verdict.
Fourth, we read the abstract and Columbia’s account, not the full paper. Details such as how the team defined “control” and what range of pitch the tone covered were not in what we could read.
What the study does show is that an auditory feedback loop, a beep that answers effort, can reach some patients that bedside examination misses, using wires they are already wearing. Claassen’s own summary is modest. “Our brain-computer interface is only a first step,” he said.
Why sound? A patient whose eyes do not focus cannot use a screen, and headphones need nothing from the patient but their ears. A tone that shifts when you try is about as plain a signal as a brain can be given.
What to watch
The next step the team describes is two-way communication: turning tone control into answers, so a patient could say whether they are in pain. That will need the larger, multi-centre studies this field has started to run, and a test of whether patients can hold control over repeated sessions. For now, Columbia says, tests for CMD are available only in research settings, not in routine care.
“Patients with hidden consciousness or cognitive motor dissociation are unable to show any expression,” Claassen said. A beep that moves is not yet a voice. For 13 people in this study, it was a way to show they were there.
