A long-time law enforcement favorite, the lie detector, now finds itself sweating the hot lights of scientific inquiry.
Crime dramas have long depicted the polygraph's tangle of wires and wiggling chart lines uncovering lies during a hard-boiled criminal interrogation. As suspects are questioned, the device checks for sweaty skin or racing hearts to root out deception, but the machine's accuracy has long been in dispute.
Nonetheless, the polygraph has a higher-than-ever profile. It's an ongoing bone of contention on Capitol Hill and a factor in recent spy investigations of FBI turncoat Robert Hanssen and physicist Wen Ho Lee. In the Lee case, the FBI's contention the physicist had lied on a polygraph test in 1998 led to 59 charges, all but one dropped in a plea bargain two years later. That sparked a request for a National Academy of Sciences (NAS) report, due as soon as the first week of October, on the validity of the polygraph.
Now, some of the same politicians who called for polygraphs of federal employees are involved in an FBI investigation aimed at finding who's responsible for a classified intelligence leak about two intercepted messages that hinted at the Sept. 11 terrorist attacks. Many House-Senate intelligence committee staffers and legislators, perhaps most prominently Sen. Richard Shelby, R.-Ala., have declined to take polygraph tests.
"Allowing the executive branch to submit the legislative branch to lie-detector tests raises constitutional issues of separation of powers," Shelby says, in a statement.
Polygraph critics such as Alan Zelicoff of Sandia National Laboratory in New Mexico find the sensitivity of Shelby and his committee peers ironic, noting two years ago that the committee helped instigate the polygraph screening of weapons scientists designed to root out spies.
A physician biodefense researcher at the weapons lab, Zelicoff has led opposition to the polygraph there, saying that for screening purposes, the device's measures - pulse, blood pressure, breathing and sweating - reveal deception about as well as a coin flip. He likens the polygraph to a defective medical test, one whose high false-positive rate, depicting honest people as liars, makes it unreliable as a diagnostic tool. Last year, Attorney General John Ashcroft estimated the false-positive rate of polygraphs at 15%, about a one-in-six chance, at a news conference.
Polygraphs are perhaps the most controversial tool in law enforcement. Some states and federal court judges now accept lie-detector results, but many states ban them outright. A 1998 Supreme Court decision allowed such bans, but read in part, "There is simply no consensus that polygraph evidence is reliable: The scientific community and the state and federal courts are extremely polarized on the matter."
The disagreements have become so entrenched that the NAS deliberately sought members for its report committee who had never staked out a position on the issue. "My primary qualification is I've never worked on the topic," noted committee chair Stephen Fienberg, a statistics expert at Carnegie Mellon University in Pittsburgh.
The key question before his committee is whether the lie-detector test is a scientific test of deception.
In theory, the polygraph relies on the different responses of the guilty and innocent to "relevant" and "control" questions during an interview. People innocent of the case under investigation, whether stealing or spying, would feel more stress about the control questions (e.g. "Before 1992, did you ever tell a lie?"), which discuss common crimes or lapses. Guilty people should show more stress during the relevant questions (e.g. "Did you shoot your husband?"), whether out of fear, effort to hide their deception or guilty remorse.
"Lying drives everything," says Dan Sosnowski of the American Polygraph Association. People inherently act differently when they are being deceptive, he and other advocates suggest. In the hands of a well-trained examiner, he says, the polygraph reveals deceptive answers in more than 90% of cases.
Studies offer a mixed picture. In 1996 the Journal of General Psychology looked at 41 criminal cases and found that control-question tests were 93-96% accurate. Tests where some study participant pretended to steal $5-20 have produced similar findings.
"We know it works, the basic idea behind the procedure is sound," says criminologist Frank Horvath of Michigan State University in East Lansing. He argues polygraphy is most effective when police can structure questions around knowledge known only to the person who committed a crime.
However, in some studies, researchers may toss "inconclusive" tests results, which if added back in, may lower the accuracy rate for detecting deception to the 70% range. Critics like retired FBI scientist Drew Richardson suggest the high accuracy rates reported in some studies result more from the high likelihood of guilt among those tested - criminals or study volunteers who act as faux criminals.
A 1997 survey of 421 psychologists estimated the test's average validity at about 61%, a little better than chance. And University of Utah psychologists published a 1994 report that suggested biting your tongue, pressing your toes to the floor and counting backwards by 7's during control questions would screw up the accuracy of polygraphs.
"A big problem is that it's not really a test of anything," says psychophysiologist William Iacono of the University of Minnesota. He agrees the polygraph can measure physical reactions, but beyond that, nobody knows how the nervous system acts when it is lying. People who don't believe in the polygraph may be more likely to fail tests, he says. Their disbelief and non-responsiveness may look like deception.
Horvath says polygraphy should be compared against other techniques for solving crimes, for accuracy. "Even if we split the difference in the debate about polygraphy and assume it's only 80% accurate, how good are the alternatives? The evidence is clearly on the side of polygraphy."
Iacono suggests that spy catchers and police officers have historically embraced the device, regardless of its lack of a truly scientific rationale, because people do spontaneously confess on occasion during exams. Obviously, guilty people who pass an exam don't turn around and confess, he suggests, with the result that polygraphers never learn about any of their mistakes.
Until then, "I would just tell Shelby and his colleagues to answer the 'relevant' questions and pass on the rest," says Zelicoff. That's what he advises scientists taking the polygraph tests at the national labs.
"I believe that the National Academy of Sciences report will be a significant milestone in the field and I hope that it will encourage those within and without the field to engage in more and better quality research," says Horvath. "Plainly there is more going on here than meets the eye."