Lie detectors are notorious liars. They implicate the innocent and exonerate the guilty. So why are they still used by intelligence agencies? Is that intelligent? Police use them in investigations, yet in most states the results cannot be used in court. Are the police wasting their time? Businesses use them for pre-employment screening. Doesn’t such stupidity make them less competitive?

In fact, lie detectors, formally known as polygraph machines, have been getting a bad rap. The National Academy of Sciences, for example, was strongly critical in their recent, widely publicized study. The web site of the prestigious Federation of American Scientists states bluntly, ” polygraphs are worse than useless-they are a significant threat to national security.

Let’s look a little deeper at the science. The standard method of polygraphy depends on four measurements: heart rate, breathing rate, blood pressure, and skin electrical conductivity. Of these, the last is the most informative, and you can buy a simple device to measure it for under $10. But the true lie detector is not the machine; it is the operator, and that person requires expensive training and extensive experience. In addition to the four measured bodily changes, the operator frequently watches body language and facial expressions, and pays attention to irregularities in verbal response. These are the same indicators we ordinary people use when trying to detect deceit, when sitting on a jury or just talking with a used car dealer.

Polygraph machines don’t actually detect lies. They measure hidden emotional responses. Given that information, the skilled polygrapher can redirect the interview in hopes of making evasiveness or deceit obvious. That makes the examination feel like an interrogation, and that is why subjects dread the process. If you would like to know what it feels like (I’ve gone through it three times), imagine interviewing for a job while sitting in full view of the interrogator-naked. Some people could do this with equanimity, but most of us would be extremely uncomfortable.

That discomfort is part of what makes polygraph enthusiasts enthusiastic. They themselves experienced the uneasiness of having the polygrapher return, repeatedly, to those very items that made them uncomfortable. They “know” that polygraphy “works.” In courtrooms, despite what is seen on the old Perry Mason reruns, the guilty almost never confess. But under a polygraph examination, they often do. They feel that the polygrapher is reading their minds, and they give up all hope of cover.

So claims that polygraphy is “worse than useless ” are too broad; the technology’s usefulness depends on the goal. Imagine you need to hire a new employee. You can do this on the basis of a submitted rsum, but I doubt you would trust that alone. You could conduct a telephone interview. Better yet would be a face-to-face discussion, and that’s what most employers do. They want to be able to look at demeanor, and perhaps vary the line of questions depending on what they see. Now imagine going one level deeper: asking penetrating questions (“Were you ever arrested?”) while listening to the heartbeat and sensing the sweat glands of the subject. Would you forgo that information, if it were easily available?

Only a few of the many reviews of the accuracy of the polygraph system meet rigorous scientific standards, according to the National Academy of Sciences’ report. Yet the credible studies show a surprisingly high accuracy rate. In my own informal survey, I was particularly impressed with an Israeli measurement (1) (they tricked their own policemen into cheating, then polygraphed them), a review of prior experiments conducted by Kircher et al. (2), and a review by Paul Ekman, a psychology professor at the University of California, San Francisco (3). I can summarize these results in an oversimplified but helpful way: the polygraph procedure has an accuracy between 80 and 95 percent. Let’s call it 85 percent.

If a subject is lying, that figure means that an examination by an experienced polygrapher will detect the deceit about 85 percent of the time. It will go undetected 15 percent of the time. That’s why the Russian spy Aldrich Ames could laugh at the test he was given; he was part of the 15 percent. Some people lie easily. Ekman implies that professional card players are particularly immune to exposure. They are probably people who are naturally good at minimizing or hiding emotional response.

It is also true that this accuracy figure implies that only 85 percent of truth-tellers will be exonerated; 15 percent will be falsely accused of lying. That is why the National Academy of Sciences’ report came down so hard on the process. The Academy was specifically tasked with examining whether polygraphy should be used for testing current employees of the U.S. Department of Energy. If 10,000 were tested, 1,500 would be incorrectly tagged as liars. Presumably this would include 15 percent of the best and the brightest. The effect on employee morale would be devastating, the Academy’s study committee concluded. The Federation of American Scientists’ quote about “worse than useless” referred specifically to the fact that despite the harm done to morale, true spies such as Aldrich Ames still slip through.

Police, on the other hand, have found a valid use for polygraphy in their investigations. If a suspect is willing to submit to a test, he may have more than his guilt to hide; he may have useful information. A nervous response to the mention of a specific location, for instance could help police find a murder weapon. 85 percent accuracy sounds pretty good, for this application.

When businesses use the process in pre-employment screening, they can be accused of being unfair, but not of being stupid. They may be willing to forgo 15 percent of potential good employees, provided that they can avoid 85 percent of potential troublemakers. It is a business decision, at least until lawmakers or courts decide that reading emotions during hiring is an illegal invasion of privacy.

Now we come to the true paradox. Lie detector results are inadmissible as evidence for criminal trials in most states. But I have been present at a trial in which the judge instructed the jury that it was their responsibility, not his, to determine the truth of the testimony. To do this, they were told to take into account “the demeanor” of the witness, his directness in answering questions, and anything else that they thought indicated truthfulness. Ironically, scientific tests show that the average person’s probability of catching a lie in this way is only “slightly better than chance,” according to Ekman. Moreover, the jurors who use this approach have the conviction that their accuracy is near 100 percent, despite their knowledge that most witnesses are extensively coached in methods of appearing sympathetic and truthful-in other words, in methods to defeat the system.

Polygraphy is not allowed in courts because 85 percent accuracy is not good enough. Instead courts use a system that is demonstrably worse-which could be a big part of the reason why so many convictions are now being overturned by DNA evidence. Where is the wisdom in that?

Notes 1) Avital Ginton, Netzer Daie, Eitan Elaad, and Gershon Ben-Shakhar,”A method for Evaluating the Use of the Polygraph in a Real-Life Situation”, Journal of Applied Psychology, vol 67 (1982), p. 132. 2) J. Kircher, S. Horowitz, D. Raskin, “Meta-analysis of Mock Crime Studies of the Control Question Polygraph Technique”, Law and Human Behavior vol 12 (1988), pp. 79-90. 3) Paul Ekman, Telling Lies: Clues to Deceit in the Marketplace, Politics, and Marriage, 4 (1985, revised 1991 and 2002), pp. 213-14.