Lie Detector Looks Inside Your Brain

MRI May Check Truthfulness, But Is It More Accurate Than Polygraphs?

(AP) 
But Dr. Mark George, the genial neurologist and psychiatrist who let me lie in his scanner and be grilled by his computer, said he doesn't see a privacy problem with the technology.

That's because it's impossible to test people without their consent, he said. Subjects have to cooperate so fully — holding the head still, and reading and responding to the questions, for example — that they have to agree to the scan.

"It really doesn't read your mind if you don't want your mind to be read," he said. "If I were wrongly accused and this were available, I'd want my defense lawyer to help me get this."

George and colleagues recently reported that the technology spotted lies in 28 out of 31 volunteers. I joined an extension of that study.

That's why I found myself lying in George's fMRI scanner, focused on questions popping up on a computer screen.

Some were easy: Am I awake, is it 2004, do I like movies? Others were a little more challenging: Have I ever cheated on taxes, or gossiped, or deceived a loved one? As instructed, I answered them all truthfully, pushing the "Yes" button with my thumb or the "No" button with my index finger.

Then, there it was: "Did you remove a watch from the drawer?"

Just a half-hour or so before, in an adjacent room, I'd been told to remove either a watch or a ring from a drawer and slip it into a locker. This was the mock crime the study used. So I took the watch. As I lay in the scanner I remembered seizing its gold metal band and nestling it into the locker.

So, the computer was asking, did I take the watch?

No, I replied with a jab of my finger. I didn't steal nuthin.'

I lied again and again when asked if I'd taken the watch, but replied truthfully when asked the same questions about the ring.

It would be a different computer's job to figure out which I was lying about, the watch or the ring. It would compare the way my brain acted when I responded to those questions versus what my brain did when I responded to routine questions truthfully. Whichever looked more different from the "truthful" brain activity would be considered the signature of deceit.

The computer asked me 160 questions over the course of 16 minutes — actually, it was 80 questions two times apiece. The verdict would take a few days to produce, since it required a lot of data analysis.

In a real-world interrogation, George said, the subject of the questioning would go through an exercise like this ring-or-watch task as well as being quizzed about the topic at hand. That way, if the computer failed in the experimental task, it would be obvious that it couldn't judge the person's truthfulness.

But ethical and legal experts said they were wary of quickly applying fMRI tests for spotting lies.

"What's really scary is if we start implementing this before we know how accurate it really is," Greely said. "People could be sent to jail, people could be sent to the death penalty, people could lose their jobs."

Judy Illes, director of Stanford's program in neuroethics, also has concerns: Could people, including victims of crimes, be coerced into taking an fMRI test? Could it distinguish accurate memories from muddled ones? Could it detect a person who's being misleading without actually lying?