Lie detectors. Any good?

Apparently lie detectors (polygraph tests) are routinely used in the US. Or are they? Perhaps I’ve been misled by american movies and tv-series.

They’re never used in Denmark. Are they any good?

I supposed no one expects them to be 100% accurate, but they must be a good deal better than 50% accurate before they’re any good. How accurate should they be before they are useful?

According to the mostly reliable Skeptic’s Dictionary the answer would be no:

In effect they can be quite good (higher than 70%, depending on who you ask), but they rely on things such as your heart rate and galvanic skin response. Every person is somewhat different so in standardizing these results a fairly valid analogy is the interpretation of the Rorschach inkblot test. Even though there is a a certain standardized way to interpret interpretations, that’s all your ever doing.

As the polygraph measures things that are not typically under the direct control of the subject, its validity as a test is open to debate. Does it measure what it thinks it measures? A statistically significantly high amount of the time the answer is yes, but also, often, the answer is no (it has failed to detect spies, for example).

There was a supreme court case, United States V. Scheffer, which ruled that jurisdictions were allowed to decide whether a polygraph is admissible as evidence. To me, this highlights that yes, it is quite accurate, but this thing could also fail just when we need it most.

What!? You are using a court decision as evidence that lie detectors work? That really makes no sense at all to me.

I’ve seen one used in a demonstration. The test subject was told to pick a number between 1 and 10, then the guy operating the lie detector asked him if 1 was his number, then if 2 was his number, etc. He was asked before the test to lie when it got to his number, i.e. always answer no. At the end of the test, the guy operating the machine asked if 6 was his number, and it was. The guy also said that our test subject happened to be a pretty bad liar, and that sometimes the results weren’t quite so clear. The guy giving the test also said that the test basically measures stress. Some people are able to control their stress reactions, and sometimes people have a false positive just because the question causes them emotional stress, not when they are lying.

Anyway, they are reasonably accurate, but as previous posters have said, they aren’t perfect. In most areas they are not allowed to be used as evidence in court, but they can be used as a “law enforcement tool”, which means that if the cops think you did something, they’ll give you a lie detector test. If you fail, they are going to think you are guilty and will try to find other evidence to prove it. It is fairly common that if someone is murdered, their spouse (who statistically is by far the most likely person to have comitted the crime) is often given a lie detector test.

A lot of intelligence agencies and the like require you to be able to pass a lie detector test as part of your job. This has led to some complaints, due to the fact that it’s not 100 percent accurate, and will cost some innocent folks their job.

A lot of businesses used to require lie detector tests, but all the ones I know have stopped using them. Some will give you “personality tests” instead now, where you have to answer questions on a computer.

Two extracts from the National Academies report on the polygraph speak volumes:

“Polygraph testing now rests on weak scientific underpinnings despite nearly a century of study, the committee said. And much of the available evidence for judging its validity lacks scientific rigor.”

“Despite its weak science base, polygraph testing is commonly believed to be a highly valid procedure for detecting lies. Popular culture and the mass media often portray lie detectors as magical mind-reading machines. The mystique surrounding the exams – instead of a solid scientific foundation – may account for much of their usefulness to authorities, the committee noted. Examiners’ field reports and indirect scientific evidence indicate that testing programs may deter potential security violators or elicit confessions from some offenders who, unaware of the tests’ weaknesses, believe that a lie detector would surely catch them.”

It’s akin to voodoo, the power of the polygraph comes from the unreasoning belief that people have in it, and the operators are the high priests of this mumbo-jumbo.

That’s because, IIRC, it’s illegal for a private business to administer a lie detector test as part of the job screening process. The federal (and maybe state nad local) governement(s) can all they want, but not your average business. I could be wrong, though. Anyone else have any real info on this, and not just what i think I read somewhere?

My position would be no, they are not very accurate and I would be very reluctant to submit to one. I say that, primarily,
because the results are subject to interpretation. I don’t know the stats, but quite often the results fall into the “inconclusive”
category, in which case the interpreter could shade the results to please whomever he is responsible to.
I also find it dubious that submitting to a polygraph is in the best interest of the subject. If you’re innocent and believe the test will
remove you from suspicion I think in most cases that you are deluding yourself. The authorities are most likely to believe results
that incriminate you and tend to dismiss results that support your innocence.
On a personal level, I submitted to a polygraph in 1981 as part of a pre employment examination. There were only myself and
the operator in the room. He conduct a brief interview, prior to the test, in which he went over some of the questions to be
asked. One of the question concerned marijuana use. I told him that I had been a very occasional user (true) and that my last
use was more than six months previous (false). I took the test and was very concerned that I wouldn’t get the job, but a day or
two later I was hired. BTW I’ve never considered myself much good at lying and I certainly don’t make a habit if it.

  1. You didn’t read the decision
  2. You took a single portion of my post out of context.

Whether or not polygraphs work is up to debate. It is true that polygraphs are most widely known for their use in criminal cases in my country. The most informed debates, where experts present the latest in scientific inquiry, occurs in the Supreme Court. Failing to present this information to the OP fails to answer his question as to whether or not they work, a question open to debate. Failing to read the decision and then attacking my post is, as the saying goes, an assumption.


  1. The United States Government is a de facto expert on polygraph usage. Cite:
  1. The United States Government has decided variously that polygraphs are both valid and invalid. Cite:

You continue to to present the argument that some people use polygraphs so they must be reliable. Not studies of their reliability.

I have yet to comment on test reliability. I have only commented on test validity. You may be unaware of it but these are technical terms. Please stop putting words in my mouth.

By the way, scientific papers are only a click away. As it is you are just pecking at parts of my posts. I’ll nod out until you produce something substantial.


Come on. I am not putting any words in your mouth. The only evidence I have seen in this thread are polygraph tests are used by lots of government agencies and there of scientific societies that don’t think they work.

“Typically” is a key word here; one can bring them under one’s control.

William Poundstone, in Big Secrets (or possibly the sequel, Bigger Secrets), claims that if the subject knows a bit about how the tests work, it’s trivially easy to defeat one. Whenever one is asked any question at all (including “What is your name?” or “What is today’s date?”), one should do something to produce a mild pain response: dig a fingernail into one’s palm, bite one’s tongue slightly, etc. If you’re really serious, put a tack in your shoe beforehand that you can push your toe onto when needed.

Apparently, a little bit of pain will produce a reaction much like the reaction to fear that polygraphs purport to measure. If every question, even the simplest and least threatening, produces this reaction, then there’s no evidence on which to base a judgment.

Poundstone also claims that the test’s effectiveness is largely based on psyching out the subject. The administrator will do a sample test like the one described above by engineer_comp_geek, guessing what card the subject has drawn from a deck. Obviously, there’s a card trick involved, and they know ahead of time what card it is. After “guessing” the card correctly, the test administrator will casually mention that the subject is “unusually easy to read,” thus making the subject feel even more on the spot.

I have no firsthand experience with any of this, but it all sounds plausible. And Poundstone is an extraordinary researcher; anyone who enjoys The Straight Dope would love his stuff as well.

How accurate polygraphs are is obviously a topic of hot debate and one that I’m not qualified to answer. However, I have seen how effective they can be, which is entirely different.

The test itself can almost never be used in court, or even a mention that it was administered (there are rare exceptions).

From a law enforcement perspective, the polygraph is just another tool used to get a confession. Once confronted with the evidence that we know the suspect is lying, he or she will often break down and confess. Even if they don’t confess completely, the test may begin to break down their defenses to where they will start admitting small things or start changing their story. Once that begins, it usually doesn’t take long for a skilled interrogator to get to the truth.

Somewhat anecdotal here, but I was talking to a CIA recruiter last year or the year before, and he mentioned that they require applicants for some program to take a polygraph test, but that the only thing the test was used for was to see if the person was willing to take the polygraph. Presumably, if the person refused to take it, there was something they wanted to hide that might make them an undesirable employee for the CIA.

Slight hijack, but thanks for this line. I was just wondering the other day what the “thumbtack in the shoe” was supposed to accomplish - I’d never heard what one was supposed to do, only that a thumbtack in your shoe could help you manipulate a polygraph.

The book is Big Secrets, incidentally. (There are two additional books by Poundstone, called Bigger Secrets and Biggest Secrets, BTW.)

Basically, when the police give someone a lie detector test, there are three types of questions: the relevant questions (“Did you kill that guy?”), irrelevant questions (“Is today Monday?” or “Is Albany the capital of New York?”), and control questions. Control questions typically ask about general honesty issues, like “Have you ever cheated on your taxes?” or “Did you ever steal something when you were a child?” These questions are used to establish the baseline for determining when the subject is lying. In other words, they assume that you are lying about the control questions. If you admit to stealing a candy bar when you were six, they’ll reword the question to exclude that, and then assume that you’re lying about the new question.

Knowing this, you can screw with the test results by elevating your stress level during the control questions. The tack method described above is one such method.

The type of polygraph test that used to be given by employers was a different type of test that doesn’t use control questions, only relevant and irrelevant questions. This type of test is based on the assumption that nobody lies about everything, so if you show elevated stress responses to some relevant questions, but not others, they assume that you’re lying about the ones where stress increases.

As others have mentioned, lie detector results are almost always considered inadmissable in court. It’s simply not that reliable a procedure.

Breaking news re: polygraphs: