March 2026 โ In laboratories separated by thousands of miles of ocean and continent, two human beings exchanged a message using nothing but their thoughts. No speech. No keyboard. No gesture. Just neural electricity, decoded by a machine, fired across the internet, and written directly into another person’s brain. It sounds like the premise of a science fiction thriller. It is, instead, the subject of four peer-reviewed studies published in some of the world’s most respected scientific journals.
The human brain has always been the most isolated organ in the body. Sealed inside bone, wrapped in membrane, bathed in fluid, it communicates with the outside world only through the imperfect intermediaries of muscles, nerves, and sensory organs. Every word you have ever spoken began as an electrochemical cascade in your motor cortex, traveled down the brainstem, animated your larynx and lips, and only then became sound waves that another nervous system could detect and decode. The journey from one mind to another has always been, at its core, a translation problem thought into signal, signal into perception, perception back into thought.
Scientists have now found a way to cut out the middle steps entirely.
Using a combination of two well-established brain technologies one that reads electrical signals from the scalp and one that writes them back in through magnetic pulses researchers have built what they call a brain-to-brain interface, or BBI. The architecture is deceptively simple. A sender wears a cap of electrodes that records the faint oscillations of their neural activity. A computer decodes those oscillations in real time, extracts a meaningful intention or decision, and transmits it over a standard internet connection to a second laboratory. There, a magnetic coil positioned against a receiver’s head fires a precisely timed pulse into the cortex, triggering either an involuntary muscle twitch or a flash of light that exists only inside the receiver’s mind. The sender thought something. The receiver experienced something. In between, no sensory organ was involved.
The first time this happened between two human beings was in August 2013, at the University of Washington in Seattle. Rajesh Rao, a computational neuroscientist, sat in his laboratory wearing an electroencephalography cap EEG for short and watched a simple computer game on his screen. The game required him to imagine firing a cannon at specific moments. Across campus, his colleague Andrea Stocco sat wearing a swim cap fitted with a transcranial magnetic stimulation coil positioned over his left motor cortex, the region of the brain that controls the right hand. Stocco was wearing noise-cancelling headphones and could not see the game. At the moment Rao imagined firing, the EEG system detected the characteristic suppression of brain oscillations that accompanies motor imagery, decoded it as a “fire” command, and sent that command across the campus internet to the TMS device at Stocco’s head. Stocco’s right hand moved involuntarily, as though controlled from afar. The result was published the following year in PLOS ONE by Rao and colleagues, and it marked the first peer-reviewed demonstration of direct brain-to-brain communication between humans.
What makes the signal readable in the first place is a phenomenon that neuroscientists have studied for decades. When you imagine making a movement without actually moving the motor cortex of your brain produces a distinctive pattern of electrical activity. Specifically, it suppresses its usual rhythmic oscillations in the 8 to 13 hertz range, a frequency band called the mu rhythm. This suppression, known as event-related desynchronization, is reliably detectable through EEG electrodes on the scalp and has long been used to allow paralyzed individuals to control computer cursors and robotic limbs through imagination alone. Rao’s innovation was not in the EEG decoding itself, which was built on decades of prior BCI research, but in where he sent the decoded signal. Instead of routing it to a cursor or a machine, he routed it into a colleague’s brain.
The technology for doing so TMS is also well-established, having been used since the 1980s for both clinical and research purposes. A TMS coil generates a brief, focused magnetic field that passes painlessly through the skull and induces a tiny electrical current in the cortex underneath. Depending on which part of the brain is targeted, different effects occur. Over the motor cortex, a single pulse produces an involuntary twitch in the corresponding muscle group. Over the visual cortex at the back of the head, a pulse generates a phosphene a fleeting, subjective impression of light that appears in the visual field without any light actually being present. These phosphenes are entirely real to the person experiencing them: a genuine conscious perception, produced entirely by magnetic stimulation of neural tissue, with no involvement of the eyes or the optic nerves.
It was the phosphene that made the next step possible.
If a TMS pulse to the visual cortex reliably produces a perceived flash of light, and if the presence or absence of that flash can be made to encode a binary digit a one or a zero then a receiver equipped with a TMS coil and a coding scheme could, in principle, receive and decode a stream of binary information written directly into their conscious experience. That is precisely what Carles Grau and colleagues at the University of Barcelona and Starlab demonstrated in a 2014 study published in PLOS ONE that remains one of the most remarkable experiments in the history of neuroscience.
Grau’s team placed a sender in Thiruvananthapuram, in the southern Indian state of Kerala, and a receiver in Strasbourg, France a distance of more than 5,000 miles. The sender wore an EEG cap and had been trained to generate distinct brain signals for binary one, achieved by imagining a right-hand movement, and binary zero, achieved by imagining a foot movement. The target message was the word “hola,” encoded as its 8-bit ASCII binary representation. The sender worked through each bit in sequence, performing the appropriate motor imagery for each, while the EEG system decoded the signal and transmitted it across the internet to France. At the receiving end, a robotized TMS coil precisely positioned over the occipital cortex using a neuronavigation system that mapped the coil’s location against an MRI scan of the receiver’s brain fired pulses that induced phosphenes for binary ones and remained silent for binary zeros. The receiver, wearing eye masks and earplugs to eliminate any peripheral sensory cues from the equipment, perceived the sequence of flashes or absences and transcribed the binary string. When decoded, it read “hola.”
Then they sent “ciao.”
The error rate across the full set of experiments was 15 percent, split roughly between a 5 percent error on the sender’s EEG decoding side and an 11 percent error on the receiver’s phosphene perception side. That performance is far from perfect, but it is statistically robust well above what chance alone would predict and it was replicated in a separate experimental corridor between Cรกceres and Barcelona in Spain. The authors were careful in their language when describing what had occurred. They noted that because both the encoding and decoding of the message required conscious, deliberate participation from both participants the sender consciously choosing each motor imagery, the receiver consciously attending to and reporting each perceived flash the term “mind-to-mind” communication was arguably more accurate than “brain-to-brain.” Both ends of the channel were inhabited by conscious, intentional agents.
It was not telepathy. But it was the closest thing to it that science has ever produced.
The University of Washington group, meanwhile, was asking what else a brain-to-brain interface could do. In a 2015 study published in PLOS ONE, Andrea Stocco, Chantel Prat, and their colleagues designed an experiment based on the structure of the “Twenty Questions” game. One participant the answerer knew the identity of a secret object drawn from a pre-defined list. The other participant the questioner had to identify the object by asking yes-or-no questions. The answerer encoded each response as a brain signal, yes corresponding to right-hand motor imagery and no to foot imagery, which was decoded by EEG and transmitted to the questioner’s motor cortex via TMS. A TMS-evoked hand twitch meant yes. Silence meant no. Round by round, the questioner accumulated neural information and used it to formulate their next question, narrowing down the possibilities until the object was identified.
What made this experiment conceptually significant was its iterative structure. The BBI was not delivering a pre-formed message in a single transmission. It was sustaining a genuine dialogue a back-and-forth communicative exchange in which each signal received shaped the next signal requested. The loop between minds was not just open; it was dynamic. The questioner’s brain was performing inference on neural information, using it to make decisions, and those decisions were generating the conditions for the next neural signal. The two brains were, in a meaningful sense, thinking together.
The field’s most ambitious achievement arrived in 2019, when Linxing Jiang, Andrea Stocco, and Rajesh Rao published a paper in Scientific Reports describing a system they called BrainNet the world’s first multi-person brain-to-brain interface network. Instead of one sender and one receiver, BrainNet connected three people simultaneously. Two of them served as Senders and one as the Receiver, collaborating on a Tetris-like game in which a falling block either needed to be rotated before landing or left as it was. Each Sender could see the game and made an independent neural recommendation rotate, encoded as right-hand imagery, or do not rotate, encoded as left-hand imagery. Both recommendations were EEG-decoded in real time and transmitted via internet to the Receiver’s laboratory. The Receiver, who could not see the game, received TMS phosphenic signals encoding both Senders’ votes, integrated them, made a decision, and implemented it in the game.
Across five groups of three participants, BrainNet achieved a mean task accuracy of 81.25 percent. But the most striking result was what happened when the researchers introduced a saboteur. In a subset of trials, one Sender was programmed to provide systematically incorrect recommendations. The Receiver’s brain, receiving conflicting signals from the two Senders, began over successive trials to downweight the unreliable source without being told to do so, without any explicit feedback about which Sender was trustworthy. The neural network was performing implicit social cognition. One brain was learning, in real time, to evaluate the credibility of another brain’s transmissions and adjust its behavior accordingly. This was not just signal relay. It was something that begins to look like collective intelligence.
The technology, in its present form, has serious limitations that no honest account should minimize. The bandwidth is extraordinarily low transmitting an eight-letter word across 5,000 miles took the better part of an hour when the encoding and decoding steps are fully accounted for. Normal human speech delivers roughly 150 words per minute. The error rates, while above chance, would be unacceptable in any practical communication system. EEG signals are sensitive to movement, fatigue, electrode placement, and individual variation in the strength and character of the mu rhythm. TMS-evoked phosphenes are similarly variable some individuals perceive them reliably, others inconsistently, and the perceptual quality and threshold of the experience differs substantially between people. Scaling the technology beyond highly controlled laboratory conditions, and beyond small samples of healthy young adults, remains an open and serious challenge.
But these are engineering limitations, not fundamental physical barriers. The researchers working in this field are candid that the present systems demonstrate principle rather than practicality. What the experiments have established definitively, reproducibly, across independent laboratories is that the human brain can serve as both the origin and the destination of a communication channel that bypasses every conventional sensory pathway. Thought can be measured. It can be transmitted. It can be received. And the mind at the receiving end can respond consciously to what it has been given.
The implications spiral outward in several directions simultaneously. For patients with severe motor and sensory disabilities individuals with locked-in syndrome, advanced ALS, or brainstem strokes who have lost the ability to communicate through any peripheral channel a sufficiently refined BBI could restore a form of communicative contact with the world. For neuroscience, the ability to directly link two brains creates experimental opportunities that did not previously exist: researchers can now study how neural signals are interpreted when they arrive through channels that bypass the sensory cortices entirely, what the brain does with information that enters through TMS rather than through the eyes or ears, and how social cognition functions when the medium is not language but induced phosphenic experience.
The ethical questions are equally significant. Every BBI study conducted so far has involved full informed consent from all participants, completely non-invasive procedures, and the active conscious engagement of both sender and receiver. Under those conditions, the technology raises no ethical concerns that are not already present in conventional BCI research. But the trajectory of the field points toward systems of greater bandwidth, greater precision, and potentially greater range of signal types. A sufficiently advanced CBI could, in principle, deliver information to a receiver’s brain without triggering conscious awareness of the source or even, in extreme hypotheticals, without the receiver’s active consent. The boundary between communication and influence, between sharing a thought and implanting one, becomes a legitimate object of ethical scrutiny at some point along that development path.
These are not present-day regulatory emergencies. The gap between “hola” encoded in binary phosphenes over 5,000 miles and any form of covert neural influence is vast, and the current technology sits far closer to the former than anything more troubling. But the time to develop ethical frameworks for neurotechnology is before the technology forces the question, not after. Researchers including Rafael Yuste at Columbia have been among the advocates for a formal framework of “neurorights” protections for mental privacy, cognitive liberty, and psychological continuity that would apply to technologies operating at the level of the brain itself.
What the science has shown, stripped of speculation and hype, is this: two conscious human minds, separated by ocean and continent, have exchanged a message that originated as a thought and arrived as a thought, with no sensory organ involved at any point in its journey. Three human brains have collaborated on a spatial task through a direct neural network, with one brain learning implicitly to trust and distrust the inputs of the other two. The channel is narrow, the errors are real, and the technology is primitive by the standards of what researchers believe is ultimately possible. But the channel is open. The first message has been sent, received, and published in a peer-reviewed journal.
The age of brain-to-brain communication has begun. It is moving slowly, carefully, and with rigorous scientific scrutiny. That is exactly how it should.
References
Rao, R. P. N., Stocco, A., Bryan, M., Sarma, D., Youngquist, T. M., Wu, J., & Prat, C. S. (2014). A direct brain-to-brain interface in humans. PLOS ONE, 9(11), e111332.
Grau, C., Ginhoux, R., Riera, A., Nguyen, T. L., Chauvat, H., Berg, M., Amengual, J. L., Pascual-Leone, A., & Ruffini, G. (2014). Conscious brain-to-brain communication in humans using non-invasive technologies. PLOS ONE, 9(8), e105225.
Stocco, A., Prat, C. S., Losey, D. M., Cronin, J. A., Wu, J., Abernethy, J. A., & Rao, R. P. N. (2015). Playing 20 questions with the mind: Collaborative problem solving by humans using a brain-to-brain interface. PLOS ONE, 10(9), e0137303.
Jiang, L., Stocco, A., Losey, D. M., Abernethy, J. A., Prat, C. S., & Rao, R. P. N. (2019). BrainNet: A multi-person brain-to-brain interface for direct collaboration between brains. Scientific Reports, 9, 6115.
Wolpaw, J. R., Birbaumer, N., McFarland, D. J., Pfurtscheller, G., & Vaughan, T. M. (2002). Brain-computer interfaces for communication and control. Clinical Neurophysiology, 113(6), 767โ791.
Schalk, G., McFarland, D. J., Hinterberger, T., Birbaumer, N., & Wolpaw, J. R. (2004). BCI2000: A general-purpose brain-computer interface system. IEEE Transactions on Biomedical Engineering, 51(6), 1034โ1043.
Barker, A. T., Jalinous, R., & Freeston, I. L. (1985). Non-invasive magnetic stimulation of human motor cortex. The Lancet, 325(8437), 1106โ1107.
Yuste, R., Goering, S., Bi, G., Carmena, J. M., Carter, A., Fins, J. J., Friesen, P., Gallant, J., Huggins, J. E., & Illes, J. (2017). Four ethical priorities for neurotechnologies and AI. Nature, 551(7679), 159โ163.
Ienca, M., & Andorno, R. (2017). Towards new human rights in the neurotechnology age. Life Sciences, Society and Policy, 13(1), 5.
Pascual-Leone, A., Walsh, V., & Rothwell, J. (2000). Transcranial magnetic stimulation in cognitive neuroscience: Virtual lesion, chronometry, and functional connectivity. Current Opinion in Neurobiology, 10(2), 232โ237.
