The world’s first brain-to-brain connection has given rats the power to communicate by thought alone.
“Many people thought it could never happen,” says Miguel Nicolelis at Duke University in Durham, North Carolina. Although monkeys have been able to control robots with their mind using brain-to-machine interfaces, work by Nicolelis’s team has, for the first time, demonstrated a direct interface between two brains – with the rats able to share both motor and sensory information.
The feat was achieved by first training rats to press one of two levers when an LED above that lever was lit. A correct action opened a hatch containing a drink of water. The rats were then split into two groups, designated as “encoders” and “decoders”.
An array of microelectrodes – each about one-hundredth the width of a human hair – was then implanted in the encoder rats’ primary motor cortex, an area of the brain that processes movement. The team used the implant to record the neuronal activity that occurs just before the rat made a decision in the lever task. They found that pressing the left lever produced a different pattern of activity from pressing the right lever, regardless of which was the correct action.
Next, the team recreated these patterns in decoder rats, using an implant in the same brain area that stimulates neurons rather than recording from them. The decoders received a few training sessions to prime them to pick the correct lever in response to the different patterns of stimulation.
The researchers then wired up the implants of an encoder and a decoder rat. The pair were given the same lever-press task again, but this time only the encoder rats saw the LEDs come on. Brain signals from the encoder rat were recorded just before they pressed the lever and transmitted to the decoder rat. The team found that the decoders, despite having no visual cue, pressed the correct lever between 60 and 72 per cent of the time.
The rats’ ability to cooperate was reinforced by rewarding both rats if the communication resulted in a correct outcome. Such reinforcement led to the transmission of clearer signals, improving the rats’ success rate compared with cases where decoders were given a pre-recorded signal. This was a big surprise, says Nicolelis. “The encoder’s brain activity became more precise. This could have happened because the animal enhanced its attention during the performance of the next trial after a decoder error.”
If the decoders had not been primed to relate specific activity with the left or right lever prior to the being linked with an encoder, the only consequence would be that it would have taken a bit more time for them to learn the task while interacting with the encoder, says Nicolelis. “We simply primed the decoder so that it would get the gist of the task it had to perform.” In unpublished monkey experiments doing a similar task, the team did not need to prime the animals at all.
In a second experiment, rats were trained to explore a hole with their whiskers and indicate if it was narrow or wide by turning to the left or right. Pairs of rats were then connected as before, but this time the implants were placed in their primary somatosensory cortex, an area that processes touch. Decoder rats were able to indicate over 60 per cent of the time the width of a gap that only the encoder rats were exploring.
Finally, encoder rats were held still while their whiskers were stroked with metal bars. The researchers observed patterns of activity in the somatosensory cortex of the decoder rats that matched that of the encoder rats, even though the whiskers of the decoder rats had not been touched.
Pairs of rats were even able to cooperate across continents using cyberspace. Brain signals from an encoder rat at the Edmond and Lily Safra International Institute of Neuroscience of Natal in Brazil were sent to a decoder in Nicolelis’s lab in North Carolina via the internet. Though there was a slight transmission delay, the decoder rat still performed with an accuracy similar to those of rats in closer proximity with encoders.
Christopher James at the University of Warwick, UK, who works on brain-to-machine interfaces for prostheses, says the work is a “wake-up call” for people who haven’t caught up with recent advances in brain research.
We have the technology to create implants for long-term use, he says. What is missing, though, is a full understanding of the brain processes involved. In this case, Nicolelis’s team is “blasting a relatively large area of the brain with a signal they’re not sure is 100 per cent correct,” he says.
That’s because the exact information being communicated between the rats’ brains is not clear. The brain activity of the encoders cannot be transferred precisely to the decoders because that would require matching the patterns neuron for neuron, which is not currently possible. Instead, the two patterns are closely related in terms of their frequency and spatial representation.
“We are still using a sledgehammer to crack a walnut,” says James. “They’re not hearing the voice of God.” But the rats are certainly sending and receiving more than a binary signal that simply points to one or other lever, he says. “I think it will be possible one day to transfer an abstract thought.”
The decoders have to interpret relatively complex brain patterns, says Marshall Shuler at Johns Hopkins University in Baltimore, Maryland. The animals learn the relevance of these new patterns and their brains adapt to the signals. “But the decoders are probably not having the same quality of experience as the encoders,” he says.
Patrick Degenaar at Newcastle University in the UK says that the military might one day be able to deploy genetically modified insects or small mammals that are controlled by the brain signals of a remote human operator. These would be drones that could feed themselves, he says, and could be used for surveillance or even assassination missions. “You’d probably need a flying bug to get near the head [of someone to be targeted],” he says.
Nicolelis is most excited about the future of multiple networked brains. He is currently trialling the implants in monkeys, getting them to work together telepathically to complete a task. For example, each monkey might only have access to part of the information needed to make the right decision in a game. Several monkeys would then need to communicate with each other in order to successfully complete the task.
“In the distant future we may be able to communicate via a brain-net,” says Nicolelis. “I would be very glad if the brain-net my great grandchildren used was due to their great grandfather’s work.”
Journal reference: Nature Scientific Reports, DOI: 10.1038/srep01319