Wired Science

Can Detailed Brain Images Really Show Us When We’re Lying?

Functional magnetic resonance imaging—fMRI—was all but unknown a decade ago. Today it’s the go-to technology for neuroscience, allowing phenomenally detailed images of the workings of the human brain. But that success has also allowed people using it to make broad claims about what they can and can’t see. Not least among them: the idea that fMRI can be tuned into a sophisticated, unbeatable lie detector. If it worked, it’d be almost magical—it’s no coincidence that one of the inventors of the original lie detector also created the superhero Wonder Woman and her truth-compelling magic lasso.

Magnetic resonance isn’t magic, though. The room-sized scanners use super-powerful magnets to snap the protons in every atom of a person’s body to a new orientation. Radio waves ping the protons’ positions and report back to a computer. Water has a lot of protons, so tissues with fluid—organs—show up on an MRI.

fMRI goes a step further. As you’ll hear about in a few weeks on Wired Science, one of the main tasks of blood is to carry oxygen throughout the body. The molecules of O2 bind to a carrier protein called hemoglobin. Oxygenated hemoglobin (also called oxyhemoglobin) responds differently to a magnetic field than hemoglobin alone (deoxyhemoglobin). Hit a brain with the scanner and you can see which parts are using more O2. It’s called fMRI—perhaps unsurprisingly—blood oxygen-level dependent functional magnetic resonance imaging.

Here’s the cool part: A brain is really just billions of neurons, cells specially designed for thinking. When cells work harder, they metabolize oxygen, but for a while people weren’t sure whether blood oxygen levels were a good proxy for an increase in neural action potential, the electrical changes that indicate activity. Then, in an article in the journal Nature back in 2001, Nikos Logothetis —apt name, huh?—at the Max Planck Institute for Biological Cybernetics in Germany showed that neurons using more oxygen according to fMRI were indeed active. (He did it by scanning macaque monkeys with fMRI at the same time as he watched the responses of electrodes stuck into specific neurons in their brains. Neuroscience can be pretty rough on macaques.)

Those kinds of findings turned fMRI into a darling of neuropsychology. With fMRI, researchers had a tool that promised to show which region of the human brain was responsible for, well, just about anything you could have someone do while stuffed into the aforementioned coffin-shaped tube. And wow, did they use it. In 1993, fewer than 20 journal articles cited the use of fMRI. By 2003 that number had grown to 1,800.

What did they find? Some pretty cool stuff. Among the most famous results is MIT researcher Nancy Kanwisher’s work. She has long argued that mammalian brains have a part of the cortex—the thin, wrinkled outer covering—specially designed to see faces. In 2003 Naomi Eisenberger at UCLA reported in Science that the emotional pain of social rejection was perceived in the same part of the brain that’s active during actual, physical pain. (It’s also the same part of the brain with which I remember high school.) And for the last couple of years, the University of Montreal’s Mario Beauregard has been using fMRI to localize religious experiences.

And, of course, there was research that showed fMRI could pick up deception. Daniel Langleben, a researcher now at the University of Pennsylvania and a specialist in addiction medicine, averaged together the brain scans of his lying subjects and saw a pattern. An area of the interior cingulated cortex lit up—if you could poke your finger through your right eye socket, you’d just about touch the spot. Theoretically, fMRI picks up the difference between when the brain is processing information from memory versus when it’s generating new facts from imagination.

Obviously, Langleben’s work generated some interest—the technology transfer office at Penn was able to license it to a company called No Lie MRI. But fMRI has its limits. The smallest area the scanner can image is about 1 millimeter by 1 millimeter, and 3 millimeters thick. More than that, if you think of the scanner as a sort of camera, then the shutter speed—the amount of time captured by a single image—is about 100 milliseconds. Martha Farah, a neuroscientist at Penn whom we talked to for the show, focused on this limitation. “Patterns of activation that are on a scale of less than a millimeter, you are not going to be able to capture with brain imaging,” she said. “Things that come and go on a time scale of milliseconds, we’re not going to be able to understand them.”

To believe that fMRI can detect lies, though, you have to accept more than the idea that deception takes place in the brain on a spatiotemporal scale suited to fMRI. You also have to believe that people understand what a lie is. It seems simple, at first—a lie is something that’s not true. But what about a fictional short story? It’s not true, but it’s not a lie, exactly. What if I read a fictional short story, in the first person? “We don’t have a taxonomy of lying,” Farah said. “There are a lot of psychological entities that you can name that don’t correspond to a part of the brain. … Face recognition, yes, that looks like it really is a component of the way mind and brain function. Patriotism? Doubtful. And the question is, which one is lying more like?”

Langleben had similar concerns (though, to be fair, he’s pretty sure he got it right). “The idea that came from that first paper is that there is a common denominator,” he said. “There may be different neural patterns of deception under different conditions, different circumstances, different forms. But the question is, is there a common denominator? Because if there isn’t, we have a problem.”

Here’s what makes me most suspicious that fMRI lie detection is as easy as it sounds: corvids. Those are birds—crows, ravens, jays, and jackdaws—and they are wicked smart. They make and use tools, learn behaviors from each other, hoard food against future shortages, and, most importantly, deceive one another about the location of those caches. They don’t want their raven buddies to swipe them. But birds don’t have interior cingulate cortices—or any cortex at all. Their tiny little brains are smooth as a trackball. That means they have to have an entirely different neurological basis for a complex behavior that looks very, very human.

Meanwhile, though, No Lie MRI is open for business.  About $3,000 will get you scanned. If you can manage to lie still in the coffin-sized bore of the tube, and if you can press the right buttons—something apparently beyond my capacity as a test subject—the machine may actually be able to look into your heart.