The truth about the polygraph

It's junk science, but proponents say it can be a useful tool in interrogations, and even a deterrent.

By Susan McCarthy
- - - - - - - - - -

March 02, 2000 | China knows something about an American miniaturized nuclear warhead called the W-88, developed at Los Alamos, N.M. We're not sure exactly how much it knows, and it doesn't seem to have built any, but naturally we're curious about how China found out. People at the Department of Energy and the FBI have been brooding over this since 1988. In 1998 the DOE's then-director of counterintelligence told a congressional committee that he thought China had gotten this information from the United States (rather than via Russia), and that he had a suspect in view.

The man he had in mind was Wen Ho Lee, a Los Alamos nuclear weapons scientist who had traveled to China in the mid-'80s. Also arousing suspicion, it seems, is the fact that he is of Chinese descent. It was then discovered that Lee had downloaded large amounts of classified material, for which he was fired. He is now in jail awaiting trial, charged with mishandling classified material. He's never been charged with espionage.

Alarmed by the idea of Chinese spies gaining access to our weapons secrets, and wishing to be seen taking action, Congress ordered that security be increased at the nuclear laboratories at Los Alamos; Sandia, N.M.; and Livermore, Calif.

Measures to be taken by the Department of Energy under that mandate included regulating e-mail and data transfers, toughening background checks on visiting foreigners and a new policy requiring that employees report any sexual encounter with foreigners longer than a one-night stand. (Not all foreigners, just foreigners from certain countries. Other foreigners, go for it; we don't wanna know.) But the measure that upset the lab employees was the one saying they had to take polygraph -- lie detector -- tests.

As many as 13,000 people might be tested, it was announced. The employees, many of them actual rocket scientists, hit the roof. They wrote letters of protest, thronged hearings to denounce polygraph testing, issued reviews of the scientific literature slamming polygraphs and created dissident Web sites. They called it a witch hunt, described the polygraph as the tool of a police state and asked what was next, the rack?

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

The polygraph is an American phenomenon, with limited use in a few countries, such as Canada, Israel and Japan, as a result of American influence. In the 1980s, in the wake of one of those spy scandals that the British are so good at, U.S. intelligence agencies urged the U.K. to use polygraphs for "security vetting." The House of Commons' employment committee, "concerned at the Government's apparent faith in the polygraph test procedures and its implications," held an inquiry, at which the British Psychological Society, among others, roundly denounced polygraphs, and the scheme was dropped.

David Lykken, emeritus professor of psychology at the University of Minnesota, is a leading critic of polygraphs and author of an influential book, "A Tremor in the Blood" (1981). He can hardly believe he still has to tell people that the polygraph isn't science.

"There's something about us Americans that makes us believe in the myth of the lie detector. It's as much of a myth as the Tooth Fairy," he exclaims in frustration.

The first proponent of lie detection machines, in the early 1900s, was William Marston, a publicity-hungry psychologist who also created Wonder Woman. The polygraph was taken up and promoted by the Berkeley, Calif., police department in the 1930s. Dozens of polygraph schools sprang up around the country. The industry thrived, with three branches: pre-employment testing, criminal investigation and counterintelligence.

The first polygraphs measured blood pressure and respiration. Then galvanic skin resistance -- sweatiness -- was added. A modern polygraph, which costs around $8,000, measures changes in all three. (You can get one for less if you supply your own printer and PC, and for just $9.95 you can order a kit from RadioShack to make a bare-bones device that measures changes in galvanic skin resistance.)

These parameters are considered to be measures of tension or anxiety. The polygraph can't read your mind, but it can detect bodily changes that are often the result of tension. The idea is that, with mind and body being connected, if you are lying you are apt to be nervous and this may show up on at least one of the parameters the polygraph records. It doesn't necessarily show up on all three; apparently it's not unusual for your pulse to say "No, no" when your sweat is saying "Yes, yes." "The correlations among the different measures can politely be described as modest," writes one British researcher.

Polygraph examiners take a training course that lasts from six to 10 weeks, but operating the machine itself is a cinch if you read the manual, according to a salesman I talked to at Axciton Systems. The need to take a lengthy training course should be your first clue that something other than a pure scientific test is going on, critics say. The enormously variable human element makes it, at best, an art, not science.

In 1988, after congressional hearings at which critics made some very unkind remarks about the polygraph, the Employment Polygraph Protection Act (EPPA) was passed, severely limiting polygraph testing. You cannot be fired for refusing to take a polygraph test. Pre-employment screening is forbidden in all except a few job categories -- security guards, armored-car drivers, etc. "There was a huge business of doing pre-employment screening that went away," says Skip Webb, a polygraph administrator who is the American Polygraph Association's public relations chairman. Judicial decisions also prevent polygraph tests from being admitted in court in almost all cases.

It might sound as if polygraph testing were dead, but there are some significant loopholes. The EPPA's provisions apply to private industry, not the government. Police departments regularly give polygraphs to suspects. And because so many people aren't aware of the problems with polygraphs, innocent people may volunteer to be tested in an effort to clear themselves of suspicion.

Old hands call taking a polygraph test "going on the box." But the box is the least of it. Webb says that if you beat the polygraph test, "you have beaten the examiner, not the instruments."

Before you take a polygraph, the examiner interviews you extensively. The examiner says that first he'll explain how the machine works. Then he'll tell you the questions he's going to ask so there will be no surprises, and the two of you will go over the questions to make sure you understand them before the actual test. The entire pretest phase generally runs about an hour, according to Webb.

This seems reassuring, since you may have been concerned that the examiner will suddenly ask who broke that slide projector in seventh grade, or whether you have sex fantasies about carp, or whether it was you who said that when they were passing out brains, your boss was standing in line to get seconds on anal-retentive.

This is also when you hear about "control" questions. These are questions about wrongdoing so sweeping that almost no one can honestly answer "No." "Have you ever told a lie to get out of trouble?" "Have you ever broken a traffic law?" They are intended to evoke an emotional reaction that the polygraph can detect when you answer "No," which can then be compared with your reaction when the tester asks you the questions he is really interested in.

At this time you might find yourself confiding to the examiner that, OK, maybe you have run a stop sign or two. (One American Polygraph Association official has testified that in pre-employment screening, three out of four applicants admitted to stealing something from a previous employer.) The examiner will appear very concerned, and ask for all the details. Your dossier has been opened, and you haven't even gone on the box yet.

The examiner also runs an "acquaintance test," usually a demonstration with a card or a number. He hooks you up to the polygraph and has you pick a card from a deck. The examiner then asks a series of questions to determine the card: "Is it a red card?" "Is it a black card?" Is it a face card?" Your job is to answer "No" every time. Then the examiner names the card, ostensibly because the charts spike when you're lying.

This is intended to convince you of the polygraph's infallibility, which makes the testing more effective, because the better you think it works, the more apt you are to be nervous about being caught in a lie. Of course, if the examiner already knew that the card was the queen of diamonds because you picked it from a stacked deck, your faith might be misplaced. In laboratory studies, without a stacked deck examiners can tell from the polygraph charts which card you picked between 30 percent and 73 percent of the time.

The examiner determines the exact form of the control questions, the non-control questions and all the follow-up questions. He may ask some questions repeatedly. As a result, every polygraph testing session is different.

Examiners don't get much feedback other than the confessions they extract at the time of testing. As Webb says, "If I were to test someone who was in fact guilty of a crime, and I mistakenly test them as not guilty, they're not likely to tell me about it. By the same token, if I test someone as guilty and they're in fact innocent, like many people who are guilty, they'd say they're innocent."

On the other hand, Webb says, in law enforcement polygraphs, the examiner may get confirmation in the form of later confessions, new evidence or convictions. Since innocent people can be convicted, most scientists don't like this last form of confirmation. In criminal cases the examiner also has information about the crime and the suspect; he needs this information to formulate the questions, but it means that he has a lot to rely on besides the machine's readings in declaring that the person is "deceptive" or "not deceptive" or in extracting a confession -- anything but a double-blind experiment.

Webb also points to studies of polygraph testing as proof of their validity. But there are a number of problems with the studies: They tend to look at either criminal investigations or staged crimes. In staged crimes volunteer subjects may be told to go into a room where half of them "steal" an object and then are tested to find out which are the "thieves." It seems unlikely that people would feel the same kind of anxiety about staged misdeeds that they feel about real misdeeds.

The biggest problem with the studies, however, is the results. In studies, polygraph diagnoses are often wrong, with rates hovering around 80 percent correct -- and sometimes much lower.

False positives -- people whom the examiner says are "deceptive" but who are in fact telling the truth -- are more common than false negatives -- people whom the examiner says test "not deceptive" but who are in fact lying. Of one study situation Lykken says, "An accused person who is innocent who takes a polygraph test has almost a 50-50 chance of failing it." Says Lykken, "Those odds are worse than Russian roulette!"

Moreover, two examiners looking at the same polygraph charts will not always agree on what they mean. In fact, one study found that 10 to 20 percent of the time the same examiner may read the same charts differently if he reads them after a six-month interval.

Thus polygraph testing gets poor marks for both reliability and validity. While the polygraph has always been wrapped in the trappings of science, scientists reject it. A 1997 poll of psychologists and psychophysiologists showed that most of them view it as junk science that should not be admissible in court.

Lykken compares it to astrology: You may think there's some validity to astrology, but would you use it to find spies? You would? Well, Mrs. Reagan, you're entitled to your opinion.

Manuel Garcia, a physicist at the Lawrence Livermore Nuclear Laboratory (LLNL) compares it to phrenology. Edward Teller, director emeritus of the LLNL, put it more tactfully in a letter to DOE Secretary Bill Richardson, calling it "a clumsy and imperfect tool" that produces "rather dubious evidence."

Nuclear lab workers who fail polygraphs, but for whom there is no evidence of guilt, can't be fired but may be transferred to less sensitive work -- a transfer that would destroy the careers of most scientists.

The courts tend to agree with the scientists; polygraph results are not admissible in court in most situations. Sometimes they are allowed as evidence, if both sides agree beforehand, in cases in which people volunteer to take a test, hoping they will be vindicated. But since even innocent people can test as "deceptive," most lawyers strongly advise their clients to shun the box.

Why does the Department of Energy want to do polygraph testing if it's junk science? Is it so stupid it doesn't know that?

It is not stupid, though some congresspeople may be.

When the scientists at the nuclear labs went public with their protest against being given polygraphs, retired Air Force Gen. Eugene Habiger, in charge of the DOE's security, told the San Francisco Chronicle that the test is a powerful deterrent.

Polygraphs don't have to work to be a deterrent. People just have to believe that they work and can reveal whether they have committed crimes. The DOE doesn't have to believe they work, either.

More important, polygraphs are an immensely effective interrogation tool; they need not detect lies. Lykken tells an anecdote of two cops interrogating a suspect at a time when copy machines were not familiar objects. Lacking a lie detector, the cops put a piece of paper in the copier that said "He's lying!" They made the suspect place his hand on the strange machine while they asked him questions. When they didn't like his answers, they'd hit a button on the machine. It would groan, whir, stink and shoot out a piece of paper that read "He's lying!" Realizing that denial was useless, he confessed.

"If I was in the police business I would use [the] polygraph," says Lykken. "It's a powerful inducer of confessions, and you don't have to hit 'em with any clubs. I can't blame the police for using it; I only blame them for believing it."

A 1983 report from the Office of Technology Assessment says, "It appears that the NSA [National Security Agency] (and possibly CIA) use the polygraph not to determine deception or truthfulness per se, but as a technique of interrogation to encourage admissions."

The FBI, interrogating Wen Ho Lee in March 1999, told him that he had failed his polygraph tests. He had taken two. The first was given by the DOE in December 1998, and three examiners agreed he'd passed "with flying colors." The DOE apologized.

A few weeks later the FBI said Lee had not passed after all. In February 1999, the FBI retested him. Unusually, they did not say at the time that he had failed, as they later charged. In March 1999, the FBI interrogated him, telling him that he'd flunked both tests. In a standard interrogation technique, they asked him to explain this. "If I don't have something that I can tell Washington as to why you're failing those polygraphs, I can't do a thing," said one agent, according to transcripts.

Lee said he didn't understand. If he didn't explain, said the agent, "I can't do anything for you, Wen Ho." The agent listed the things he couldn't do for Lee unless he explained why he hadn't passed the test: He couldn't get him his job back, keep reporters from calling his family, keep his wife from being given a polygraph or keep "them" from going to Lee's house and taking him away in handcuffs. Lee repeated that he was telling the truth.

The agents then pointed out that they also wouldn't be able to keep Lee from dying in the electric chair like Julius and Ethel Rosenberg if he didn't tell them why he had failed the polygraph test. "They electrocuted them, Wen Ho."

If Lee is really a spy, he might know how to fool the polygraph test. Lykken is confident that a spy would have received the simple training required to beat the polygraph.

Spies Aldrich Ames and Harold J. Nicholson passed routine CIA polygraph exams. In fact, it's not clear that any spies have ever been caught by polygraph testing. "It's inconceivable that [a spy] won't know how to beat the test," Lykken says. "So the spy won't be caught, and a few innocent people will have their careers ruined, and the secretary of energy will say that we did everything we could. I think it's a scandal."

Was Lee telling the truth? DOE polygraph examiners thought he was truthful in December 1998. Then the FBI said the same test results showed that he failed. When they interrogated Lee, agents told him he had failed both that test and the later FBI test. Did he really fail? Since polygraph examiners often disagree with one another (and even with their earlier diagnoses), the question is not very meaningful. Neither is the question of whether the agents genuinely thought that he'd failed, since even if they believed he'd passed, telling people they've failed is such a powerful interrogation technique.

Could polygraphs ever work? All they do is record certain poorly correlated physical reactions to mental stimuli. Since people don't think and react identically, since even when we're being offensively honest we don't always agree about what is the truth and what is a lie, the attempt seems doomed by definition.

What if polygraphs worked perfectly, if they could somehow sort out the tension caused by telling lies from the tension caused by fear of being called a liar, and could do it every single time? What if they never made an innocent person look guilty, never let a guilty person get off and could always tell who was likely to steal from an employer or betray the employer's trust? What if two examiners never reached different conclusions from the same charts and the same examiner never read the same chart two different ways? Would that change everything?

Those who administer polygraphs would be even more impelled to hook people up. But many people would be just as eager not to be hooked up, even if they weren't guilty of crimes and didn't plan to be guilty of crimes in the future. Do we want government minions or corporate flunkies reading our minds and feelings? The American Civil Liberties Union has called polygraph tests "strip searches of the mind."

We would have people flying about demanding to be hooked up and answer just one question: "I did not break your Lego tower, I swear. I demand a lie detector test!" "I did not think you were stupid for acting that way at the party, honestly. How can I convince you? I'll gladly go on the box!"

We would also have people carrying polygraphs around with them at all times: "You really believe people pick up their welfare checks in Cadillacs? Put these electrodes on and say that!"

Lee, on trial not for spying but for mishandling classified data, is in solitary confinement in a New Mexico jail. He is the only person to be prosecuted for such offenses. Some angry scientists have pointed out that former CIA Director John Deutch committed a similar offense in keeping classified material on unprotected home computers. He lost some of his security clearances, but not others, enabling him to do lucrative consulting.

Many of Lee's colleagues view him as a hapless scapegoat who made serious errors. "What he is accused of is a very serious security violation," says electronic engineer William O'Connell, a past president of the Society of Professional Scientists and Engineers (SPSE). "There may be some real spy problems," he adds, saying that he favors improved computer security, "but there's no evidence to connect Wen Ho Lee to those issues."

The charges that Lee was singled out because of his ethnic background, charges supported by statements made by several DOE officials, are also creating great unhappiness in the labs.

At this writing, the DOE has redefined the category of those who must be given polygraphs to scientists in a few highly sensitive categories, reducing the number from 13,000 to 800. It has also said it will ask Congress to soften its mandate that the DOE do polygraph testing. These steps have muted the storm, but opposition remains. The SPSE continues to take the position that no one should be subjected to the voodoo vagaries of the polygraphs, and to fret that the DOE still has the power to order sweeping testing at any time.

In a recent criminal investigation of the murder of an LLNL worker, Livermore police, who apparently haven't been following the news, were surprised by the unwillingness of many of the worker's colleagues to undergo polygraph testing.

As for the future of the polygraph, although it has almost been banned from private business, it continues to find a home in various arms of law enforcement. Its popularity there may rest less on any misguided belief in its infallibility than on its real efficacy as a tool of interrogation.

Yet the tide may be slowly turning. The 1988 Employment Polygraph Protection Act dealt the industry of pre-employment screening a serious blow. The gradually increasing body of research has convinced scientists that the polygraph is a form of interrogation, not science. But as long as general public awareness of the subject is uncritical, spy scandals will be followed by politicians calling for testing so they can be seen taking action and by law enforcement calling for testing in hope of extracting confessions.

Return to Rants Page