Just by looking at the pattern of firing in your brain, neuroscientists can tell whether you are thinking about moving your hand to the left or to the right. They can tell if you have seen something you didn’t even know you saw, and, now it seems, they can tell which mathematical operation you secretly have in mind.
“We wanted to see how far we could go with reading peoples’ thoughts from their brain activity,” says John-Dylan Haynes, lead author of the a study just published in Current Biology1.
Patterns of brain activity have most commonly been studied to see how they relate to thoughts about motion, such as the brain signals that make our fingers type and our legs walk. A study published last year even showed that an electronic implant could translate the brain activity generated by a man thinking about moving a cursor on a computer screen into actual movement of the cursor2. And prosthetic arms have been made that translate thought into movement (see ‘Re-wiring brings back touch for amputated limb’).
But can a simple map of neural activity also expose more abstract thoughts — such as those to do with planning and decision-making?
To find out, Haynes and colleagues put three men and five women into a functional magnetic resonance imaging (fMRI) scanner, and asked them to decide whether they would add or subtract a set of numbers, and to concentrate on that decision.
After a delay, two numbers appeared on the screen, and the volunteers performed the mental maths they had decided on. They then selected the correct answer from a list, revealing to the researchers which option they had chosen.
Think of a number…
Pick an equation. Think about it hard. A brainscan should tell which you have chosen.
To see what was happening while volunteers were privately making their decision to add or subtract, Haynes’ team looked at changes in oxygen in the blood across the brain. These data, picked up by the scanner, indicate changes in metabolism and so, in neural activity.
A computer program was then fed all these results, so that it could learn to distinguish the patterns of activity associated with a thought about addition from that for subtraction. When they tested the computer’s abilities to predict the correct answer on a fresh set of data, the computer correctly determined what the volunteer was planning to do 75% of the time — quite a lot better than the 50% expected from random chance.
“We’ve shown that it’s not just simple motor commands that you can read out,” says Haynes. “You can also read out something really high level, if you know where in the brain to look.”
Although Haynes is enthusiastic about the possibilities their research might have for research on communication between brains and computers, he might yet have to convince his peers of the usefulness of the discovery.
Brad Dickerson, of Harvard Medical School and Massachusetts General Hospital in Boston, wonders what exactly a prosthetic or other interface between the brain and a computer might do with this information. “It’s difficult to see what exactly is being measured in terms of thought and how that would be useful,” he says. As for the notion of making a really thorough mind-reading machine, distinguishing a thought from just two choices is very different from when you have many options, as is more often the case in real life. “If there were ten options, would our predictions be any better than chance?”
Todd Braver, a neuroscientist at Washington University in St Louis, Missouri, is more excited about the study’s implications for understanding when we become conscious of the decisions we make. Haynes agrees: “The next step to take is to see whether we can read thoughts about intention before a person has even become conscious of them themselves.”