Polygraphs Don't Give True Story
May 14, 2004
By Noah Shachtman
Nearly 75 years since the introduction of the polygraph, there's still nothing close to a foolproof lie detector. Traditional methods for catching a fibber have been battered by scientific study. And, despite endless waves of hype, the high-tech alternatives -- brain scans, thermal images and voice analysis -- have withered under scrutiny, or remain largely unproven.
"Everybody would love to have a lie detector that works. But wanting it isn't going to make it happen," said Stephen Kosslyn, a Harvard University professor of psychology.
"You can flip a coin, and get the same results," said Mike Ritz, a former Army interrogator who now trains people to withstand questioning.
In a 2002 report, the National Academy of Sciences concluded that traditional polygraph screening was so flawed that it "presents a danger to national security." The group found that too many innocent people who took polygraphs were labeled guilty, and too many guilty people slid by undetected.
Federal and local governments have carried on with polygraphs anyway. U.S. military investigators, armed with the devices, have been deployed to Iraq, to question candidates for detention. The Energy and Defense departments give out thousands of the tests every year to filter out potential security threats. And the Supreme Court has ruled that it's up to the states to decide whether evidence from lie detectors is admissible in court.
Polygraphers contend that -- especially when they start out with a piece of damning evidence -- they can catch liars at rates of 90 percent or better. The problem is that polygraphs check only for physical responses that indicate deceit: heavy breathing, high pulse rate, sweat. But panting or sweating don't necessarily mean that a person is guilty of anything. All these responses indicate is that someone is anxious, said University of Arizona psychology professor John JB Allen. And innocent people get jumpy, too -- especially when there's a bull-necked interrogator in the room.
These same limitations also apply to many newfangled approaches to catching a liar. The area around the eyes supposedly heats up when a person gets agitated. But an experienced con man is unlikely to get flushed when he's fibbing.
Voice stress analysis uses computer programs to look for hints of prevarication in someone's speech. Washington University researcher Mitchell Sommers found the method worked "consistently less than chance." A Chicago firm, V, is pushing a competing approach called layered voice analysis. But company spokesman Jayson Schkloven admits there have been "no significant studies" of the technology.
There are other ways to look for guilt, however. Hidden beneath the conscious mind, many researchers believe, are indicators of recognition that could potentially be traced. And those signs can show whether or not someone has intimate knowledge of a crime or terrorist plot. An interrogator could show a suspect a murder weapon, or a suicide bomber's belt, and the suspect's mind would involuntarily and almost instantly flash signals of familiarity.
The most promising pointers, many researchers believe, are so-called P300 waves -- electrical impulses that peak about 300 to 500 milliseconds after the brain has seen something recognizable.
But P300s have been fickle under examination. In a recent study of 75 University of Arizona students who had just acted out a mock crime, "guilty" undergrads were only spotted half of the time. Those results dropped to just 25 percent when the simulated suspects used countermeasures during their interrogations, such as thinking of their professor giving them a hard slap when a particular item was discussed.
The numbers stand in stark contrast to the claims of Larry Farwell, the chairman of Brain Fingerprinting Laboratories. Farwell has become involved in a number of high-profile court cases -- including, most recently, working with the defense in the death row appeal of Oklahoma inmate Jimmy Ray Slaughter. And for that, Farwell has been showered with media attention, from 60 Minutes, Time Magazine and Wired News, too.
Farwell asserts that he's been just about 100 percent accurate in finding out whether a person knows something or not. The reason he's so successful, Farwell says, is that he looks at not just the P300 electrical spike, but also at a trough that comes nearly a second later. The questions he asks his subject are different, as well. And so are his algorithms for crunching brain waves.
But Allen, of the University of Arizona, is one of several researchers who aren't buying Farwell's results.
"I've replicated his procedure in the lab, and only detected 50 percent of the criminals," Allen said. "There's some promise to it. But it's not ready for prime time."
Farwell says that he has run his tests on hundreds of subjects. But he's only published one peer-reviewed study. It only had six people in it. Farwell contends he's been too busy in the real world to write up his other experiments.
"I invented a lifesaving technology. I withheld it for 15 years. And it has yet to make an error," Farwell said. "What am I supposed to do when Jimmy Ray Slaughter comes to me?... 'I want to do a few more years of research. Too bad you're going to be executed in the meantime?'"
Regardless, Farwell sees the limitations of his approach. At best, P300s will only tell an interrogator that his suspect knows or doesn't know something. If a murderer doesn't pick up on fashion, for example, it might not register that his victim was in a green dress.
Which brings interrogators back to their old workhorse: the polygraph.
Ritz, the former Army interrogator, says a well-trained person doesn't need a machine to spot lies. Like a good poker player, a good interrogator can spot a fabricator's "tells" -- the small changes in body language that become dead giveaways.
But Steven Aftergood, with the Federation of American Scientists, sees a place for the machines "as props in the investigative process."
For those who believe in the omniscience of American hardware, a polygraph test can be absolutely terrifying. Hey pal, the machine says you're hiding something. Might as well confess.
"It's not science. It's not technology," Aftergood noted. "But it's sometimes effective theater."