From chewing rice to scanning brains, the perfect lie detector remains elusive
An early form of lie detection existed in India 2,000 years ago. Back then, a potential liar was told to place a grain of rice in his mouth, and chew. If he could spit out the rice, he was telling the truth. If he could not, that meant fear of being caught had parched his throat, and his deceit was confirmed.
Since that time, scientists have searched for a truth tool more reliable than Uncle Ben's—one that can separate fibs from facts with the push of a button. Such a device could slash trial length, aid job screeners and protect borders. The person to fashion this magical instrument—as precise as DNA, and far more applicable—would shift the entire landscape of forensic discovery. It could create a gap in the dictionary between "periwinkle" and "perk," where "perjury" once stood, and a crater in the TV Guide, where "CSI" and all its spin-offs once reigned supreme.
But each advance in the field of lie detection has met with a hitch. Polygraph machines have drawn considerable scientific scrutiny and remain inadmissible in courtrooms. Functional imaging has pinpointed which areas of the brain become active when people lie, but the results are based on group averages and become less accurate when a single person is tested. Even people with incredibly accurate facial analysis skills, so-called lie detection "wizards," were called into question last month in the journal Law and Human Behavior.
What follows is an overview of the long and continued struggle to find the perfect lie detector.
In the early 20th century, Harvard psychologist William Mouton Marston created his "systolic blood pressure test," more commonly known as the polygraph machine. Marston's hodgepodge of gizmos included a rubber tube and a sphygmomanometer—that childhood favorite the pediatrician wraps around a bicep and inflates with each squeeze of an egg-shaped ball. Polygraph 101 is clear enough: a person has typical levels of heart-rate, respiration and blood pressure when answering a basic question like "Is it true you live at 520 Elm Street?" If these levels remain the same during questions such as "Did you kill Jane Doe?" then the person is telling the truth. If not, he or she is lying.
Despite its reputation as the default lie detector, the polygraph has never received much credibility. In 1922, a federal judge ruled that Marston's device could not be used in a murder case; it did not hold "general acceptance" among the scientific community, wrote Justice Josiah Alexander Van Orsdel of the United States Court of Appeals. This decision, known as the "Frye standard," has essentially kept the polygraph out of courtrooms ever since.
In 2002, the National Academy of Sciences orchestrated a massive review of the polygraph. The Academy concluded that the tool was not consistent enough to be used as a screening device when hiring national security employees. The physiological responses measured by the machine can be the result of many factors other than lying, including mere nervousness.
"There are many people who will speak in favor of the polygraph," says William Iacono, who is a professor of psychology and law at the University of Minnesota. "The argument is, if the government uses it 100,000 times a year, how can it be so wrong? The reason they believe it is because of the nature of the feedback they get. Occasionally, people fail the test and they're asked to confess, and they do. But if a guilty person passes, he doesn't turn around on his way out and say: ‘Hey, I really did it.' They never learn of their errors, so they don't think there are any errors."
In the end, Marston's reputation made out better than that of his machine; he went on to earn fame as the creator of Wonder Woman.
The Guilty Knowledge Test
In the late 1950s, modern deception research took a new turn, when psychologist David Lykken of the University of Minnesota adapted polygraph interrogation with his guilty knowledge test.
A typical polygraph question asks a suspect whether he or she committed a crime. The guilty knowledge test focuses its questions on knowledge that only a perpetrator would have. Say, for example, you stole a purse from a woman wearing a bright green dress. A polygraph examiner might ask: "Did you steal the dress?" A good liar could control his response and pass the exam. Lykken would ask two questions: "Did you see a green dress?" and "Did you see a blue dress?" Regardless of your answer, the mere mention of the incriminating detail would cause a noticeable blip in your physiological reactions.
In 1959, Lykken published the first study showing the effects of this method. He had some 50 subjects enact one or two mock crimes, while others enacted none. Then he asked everyone to take a guilty knowledge test. Based on physiological responses, Lykken correctly categorized about 90 percent of the subjects, he reported in the Journal of Applied Psychology.
One of the subjects, it so happens, was a Hungarian refugee who had twice fooled the KGB about his anti-Soviet involvement. After a 30-minute interrogation, Lykken had identified which of the two mock crimes this subject had committed.
One day in 1983, the phone rang in J. Peter Rosenfeld's psychology lab at Northwestern University. It was a CIA agent. He wanted to know if Rosenfeld would run the agency's new lie detection program.
Rosenfeld froze. How did the CIA know he had planned to start researching deception? After all, he had only told a trusted colleague, and his mother. But it soon became clear that the agent had been calling several researchers in the hopes of luring one to direct the new program. Rosenfeld declined but recommended a promising graduate student, and for the next several months, broad-shouldered men in suits popped out from behind trees on Evanston's north campus.
Finally, the agency decided to hire the student. She flew to Washington, D.C. and took a polygraph test as standard job-screening procedure. But as her husband and children prepared for a new life, she failed the test on a question about her sexuality and lost the job, Rosenfeld says. "It was a simple case of the polygraph making a mistake, but the CIA has to be more safe than sorry," he says. "At that point, I said we might as well try to have one [a lie detector] that's based on science."
Rosenfeld settled on a method that combined Lykken's guilty knowledge test with brainwave research performed by Columbia University researcher Samuel Sutton. In the 1960s, Sutton had discovered that human brains show a burst of activity 300 milliseconds after a person sees a distinct image. Rosenfeld's premise was simple: If a woman wearing a green dress is robbed, then the perpetrator's mind will store an image of the dress, and his brain will respond a certain way when later confronted with this image.
The basic science behind the idea is not much more difficult. Brain cells emit electronic signals in a rhythmic, up-and-down pattern. These signals can be recorded from a person's scalp, and the resulting sequence of peaks and dips is called a brainwave. One of these waves, the P300, swoops enormously when it recognizes an image. The "P" aspect stands for positive, and the "300" refers to the number of milliseconds the wave occurs after recognition.
In 1987, Rosenfeld tried his P300 test on ten subjects. Each subject "stole" one item from a box of nine desirables. By actually touching the item, subjects formed a bond with the object that would result in a P300 response, Rosenfeld predicted. The subjects then watched names of the items flash across a monitor. When non-stolen items appeared, the brainwaves showed up normal. But when the stolen item flashed on the screen, the subject's brainwave formed a distinct P300 response.
The main advantage of this method over the traditional polygraph is striking: deception is implied without the suspect saying a single word. In fact, the P300 cannot even be considered a lie detector. "You're looking at recognition, not lying," Rosenfeld says. "However, I think the inference is justified if you take the proper measures."
In the 1990s, a scientist named Lawrence Farwell combined the guilty knowledge test and the P300 technique to create a commercial lie detector called Brain Fingerprinting. In 2000, Brain Fingerprinting almost gained admission to the courtroom during an appeal of a murder case in Iowa. (A district court judge rejected the appeal but ruled that the technique could have been admissible. A State Supreme Court judge eventually upheld the appeal, but did not take Brain Fingerprinting results into account.)
But a drawback of lie detectors based on the P300 method is that investigators must work very hard to find unusual items that only the criminal would have seen. Take the case of the bright green dress. If that dress is truly unique to the crime, the suspect will produce a powerful P300 response. But if the criminal's wife happens to wear a lot of green dresses, the P300 wave could be blunted down to regular size.
Functional imaging, often called fMRI, allows scientists to monitor brain activity in real time. Subjects are wheeled on a padded platform into a noisy magnetic resonance imaging machine that scans their brains every two seconds in search of increased neural activity. A small mirror allows them to see and react to prompts shown on a screen outside the machine. Meanwhile, from another room, investigators collect brain activity for statistical analysis.
The first fMRI study of lie detection to receive widespread attention was published in 2002 by Daniel Langleben of the University of Pennsylvania. Langleben handed his subjects a playing card—the five of clubs—before sliding them into the MRI machine. He encouraged them to deny having the card, offering a $20 reward for those who successfully deceived the machine, which was more than enough incentive for his undergraduate subjects.
During the test, subjects saw various playing cards on a screen and pushed a button indicating whether or not they had the card being shown. Most of the time, when subjects denied having the card on the screen, they were telling the truth. Only when the five of clubs appeared was the response a lie.
Langleben compared truthful brain activity with deceptive activity and found that a person's mind is generally more active when lying. This result suggests that truthfulness might be our default cognitive status, and that deception requires additional mental effort.
But a lie detector based on functional imaging would suffer from a few potentially fatal flaws. Critics of the method often point out that functional imaging results are averaged from a group, not based on individual subjects. Such a limitation causes obvious problems in the world of criminal law.
In the fall of 2005, Langleben found encouraging evidence that functional imaging can detect deception on an individual basis. Using a modified version of his previous test, Langleben reported being able to correctly classify individual lies or truths 78 percent of the time. His results are the first evidence that functional imaging can detect deception for an individual person regarding an individual question. Still, 78 percent accuracy, while promising, is far from fool-proof.
While driving on a dark night in northern California, Maureen O'Sullivan listened to J.J. Newberry, a former agent in the Bureau of Alcohol, Tobacco and Firearms, discuss how he had been betrayed by a friend. Newberry seemed very upset by the incident, and very involved in the telling of it, O'Sullivan recalls. Then, suddenly, Newberry asked O'Sullivan to pull over. In the middle of his engrossing story he had spotted a man slumped over behind the wheel of a parked car across the street.
Such preternatural awareness has helped make Newberry a lie detection "wizard," says O'Sullivan, who coined the term with her colleague Paul Ekman at the University of San Francisco. The distinction is a select one: in 30 years of testing, the researchers have found fewer than 50 wizards. These people score in the upper ranks on a battery of deception tests developed by Ekman and O'Sullivan.
"These people are super hunters," O'Sullivan says. "What they see is unbelievable."
Ekman and O'Sullivan began testing for people who could identify deception with great accuracy in the late 1980s. They eventually settled on series of three tests. The first involves spotting people lying about their feelings. For this test, potential wizards watch a videotape of ten women, half of whom are lying about their current emotions, half of whom are telling the truth.
The second test shows ten men describing an opinion they have, and the third shows ten men discussing whether they had stolen money. Again, in both cases, half the people are lying and half are telling the truth.
For a person to become a wizard, he or she must first correctly identify nine people in the emotional test, then go on to identify at least eight people in one of the two other tests. As of 2003, having studied more than 10,000 people, the researchers had found just 29 wizards. That number has grown to about 50, O'Sullivan said recently.
Many wizards spent time in the Secret Service, says O'Sullivan. The practice of scanning large crowds for odd behaviors has honed their acuity. Whereas regular people make a quick decision when watching the test videotapes, wizards hold their final analysis until the end, tracking intonation changes, word choice and eye gaze. Therapists also score high on the tests.
Social psychologist Charles F. Bond Jr. of Texas Christian University is unconvinced. Bond believes the wizards are mere statistical outliers—the eventual result of testing thousands of people on the same task.
"They presented the fact that a small number of people did well out of a huge number of people who took the test, as evidence that those people had a special skill, "says Bond, whose argument was published online recently in Law and Human Behavior. "If a lot of people play the lottery, someone wins."
Before government and legal agencies begin consulting these wizards, Bond would like to see outside sources conduct additional tests on them—a measure of validity that O'Sullivan says is now in the works.
But even with additional tests, perfection will have to wait until the next generation lie detector. To date, says O'Sullivan, nobody has scored perfectly on all three tests.