[image]
[title]With its weathered blue sofa, textbook-littered desks and corkboard plastered with postcards from spring-break spots like Florida and Hilton Head, Room 202 in Northwestern University’s Cresap Hall could easily belong to a frat brother. But no keg parties or Guitar Hero marathons have taken place within its yellowing walls. It’s a laboratory where psychologist and professor Peter Rosenfeld is working to read minds—by building the ultimate lie detector.
Rosenfeld isn’t tinkering with knob-covered polygraph machines like the ones seen on cop shows. Instead, the burly 72-year-old is strapping electrodes to people’s heads and grilling each subject with questions designed to activate brain waves called P300s, which are triggered when a person recognizes a stimulus, like a picture or a word. For example, if you show a jewel thief a picture of a diamond he stole, he’ll most likely emit a P300 signal. (P stands for positive, and 300 is the minimum number of milliseconds it takes for the person to recognize the image.) Rosenfeld’s claim: Besides helping pinpoint whodunit, P300s can be used to predict and prevent future crimes by plucking essential information from the brains of would-be perpetrators.
With his thinning gray hair and a New York accent that’s survived four decades in Chicago, Rosenfeld has been studying deception detection at Northwestern since the early ’80s. But after September 11, he started to think about his research in a new light. “I realized the government would be interested in an antiterror scenario,” he says. Last June, he published the findings of an unprecedented mock terrorism study in the scientific journal Psychophysiology.
Researchers split 29 Northwestern students into two groups, instructing one to plan a vacation; the other, a terrorist attack. Then, subjects were outfitted with swim-cap-like hats covered in electrodes that would, essentially, read their minds. The lab’s aging PCs flashed hundreds of dates and pictures of cities; the “terrorists’ ” P300s shot up when relevant images or words—like Houston, the site of the attack, or July, the month—appeared. When researchers knew some of the attack’s details beforehand (comparable to the CIA having limited intelligence on a potential threat), they nabbed 12 out of 12 terrorists. When they didn’t, they identified ten of 12.
“We don’t claim we’re revealing lies; we’re revealing recognition,” Rosenfeld says. “But if you can show that someone recognizes something they verbally deny, you can infer deception with a high probability of accuracy.”
The method isn’t perfect. If a subject has an innocent mental link to something suspicious, his or her P300s will still spike at levels indistinguishable from the guilty suspects. (In the case above, that means anyone who happened to be planning a vacay to Houston in July may have been flagged as a terrorist threat.) Of course, the more details available about a crime, the less probable such coincidences would be. But what happens when you don’t have that crucial intel? When interrogating a terrorism suspect, you can name all major U.S. cities pretty quickly to see if one is a trigger. But when grilling a potential bank robber, it would be near impossible to name every bank in the country to determine his or her target.
According to Rosenfeld and his disciples, even with these flaws, P300 exams are superior to the polygraph tests favored by U.S. government and law-enforcement agencies. “You can be taught how to beat a poly in ten minutes,” says Mike Winograd, a 25-year-old Ph.D. candidate who’s been studying with Rosenfeld for five years. The machines measure physiological arousal—factors like heart rate or blood pressure, which allegedly rise when you lie. But you can throw off the system, Winograd says, by causing yourself physical pain—say, biting your tongue—to get your heart racing during the less-heated part of the interrogation. Then when you lie on the big question (“Did you murder your wife?”), any unintentional reaction will seem muted in comparison, and bam: The poly perceives you as innocent.
On a typical day in Rosenfeld’s lab, psych undergrads doubling as test subjects shuffle in and out while Ph.D. candidates conduct trials. “They’re not all as sexy as the terrorism scenario,” says Winograd, whose most recent study centered on a mock petty theft. The data sniffed out the guilty subjects with 92 percent accuracy.
Rosenfeld says his research is “close to accurate enough” to use in court. The only hurdle left is a study on real-world subjects. He’s working on getting a proposal through the Institutional Review Board, which monitors research ethics. His goal is to get federal, law-enforcement and security agencies to recognize the potential of P300s. But the holy grail of lie detection, he adds, would be a system in which responses are directly correlated with deception—à la Pinocchio’s nose. “I think that kind of thing may be possible one day,” Rosenfeld says. “But you’re not going to get it with a polygraph.”