The ‘Lie Detector’ Test Revisited: A Great Example of Junk Science
Recently I came across one of television’s true crime programs that presented a provocative example from actual case files: A woman had been brutally murdered in her apartment. Her former boyfriend with whom she had recently had a vigorous altercation became the leading murder suspect. During the investigation, this man was “offered” a polygraph (lie detector) test in the attempt to establish his likely guilt or innocence. He agreed to be tested and was found to have “failed,” thus presumably indicating his likely guilt. Despite this result, the evidence presented at the trial was deemed insufficient for a guilty verdict, and he was acquitted. Convinced that the polygraph test was accurate, his local community made him a pariah; he was shunned and even threatened with bodily harm. Several months later, however, another man, the actual murderer, was apprehended and convicted. Thus at least for this first suspect, despite his ordeal, the story had a satisfactory ending, and the lie detector test itself proved inaccurate and misleading. This outcome leads to an obvious question: How often does “lie detection by machine” prove false?
Much of the American public seems to be convinced that the “lie detector” is valid, as indicated by its ubiquitous use in “whodunit” literature and on television crime, psychology, talk, and news shows. After all, faced with such an avalanche of widespread approbation, who could doubt the validity of such a test? Supporting this illusion is the fact that federal, state, and local police departments and law enforcement agencies across the United States are generally avid proponents of this method.
But let’s take a closer look at this subject. Questions about the accuracy of this test should be amenable to modern scientific methods. Interestingly, this challenge is strikingly similar to those we face regularly as medical researchers and practitioners when we evaluate various tests in the attempt to establish the presence or absence of many diseases. Through this lens, therefore, I can provide some insight on a method that is often uncritically analyzed.
The Procedure and Its History
The “lie detector” test has been used for nearly a century, and it employs a “polygraph,” which, during questioning, continuously records an examinee’s blood pressure, respiration, pulse rate, and skin resistance (an indirect measure of perspiration).
The usual format compares responses to “relevant” questions with those of “control” questions. The control questions are designed to control for the effect of the generally threatening nature of relevant questions. Control questions concern misdeeds that are similar to those being investigated, but refer to the subject’s past and are usually broad in scope; for example, “Have you ever betrayed anyone who trusted you?”
A person who is telling the truth is assumed to fear control questions more than relevant questions. This is because control questions are designed to arouse a subject’s concern about their past truthfulness, while relevant questions ask about a crime of which they are suspected. A pattern of greater physiological response to relevant questions than to control questions leads to a diagnosis of “deception.” Greater response to control questions leads to a judgment of no deception. If no difference is found between relevant and control questions, the test result is considered inconclusive.
In the effort to improve the test’s accuracy, alternative means of questioning have been suggested, such as the “guilty knowledge test” (Ruscio 2005). Rather than attempting to determine the truthfulness of an examinee’s responses to relevant items, this technique aims to expose “hot” responses to questions only a guilty individual could display. This is done with a series of multiple-choice questions, one of which contains the incriminating information. This type of interrogation is limited only to instances of specific wrongdoing, but it has not been sufficiently investigated and is quite likely to suffer from the same shortcomings as the conventional procedure.
The test records the activity of the sympathetic branch of the autonomic (involuntary) nervous system that influences heart rate, respiratory rate, blood pressure, and perspiration. Although this part of the nervous system is active at all times, it increases during excitement, rage, anxiety, fear, or fright, any of which could be caused by lying. But deception is a cognitive function that defies direct measurement. Indeed, throughout the entire history of medical science, there have been no scientific studies that have shown that the emotional response linked to lying could be measured. Moreover, reactions associated with lying and any other assumed emotional stresses can be quite variable. Some people may stay calm with a gun at their head. By contrast, others may respond excessively with heart thumping and sweaty palms at simply shaking someone’s hand. The polygraph examination itself often causes fear and anxiety, and if such responses are excessive in response to a given question, then one may be deemed to have failed that question by a polygraph examiner.
Because of this obvious biologic improbability in this era of evidence-based medical science, the premise of lie detection by polygraph has fallen under a cloud of skepticism, and it is justly considered pseudoscience by most of the scientific community (Iacono 2001).
The American Polygraph Association (APA), a professional organization for polygraph examiners, predictably has complete faith in the accuracy of the test. They have licensing procedures in twenty-eight states and their own trade journal, Polygraph, in which they report scientifically questionable studies and provide anecdotes of the accuracy of their trade. The majority of these members complete a six-week to six-month post–high school training course in the art of polygraphy. They are not required to have formal training in medicine, psychology, physiology, or behavior—the very disciplines on which the testing is based. The majority of their members cater to the legal system upon which their economic livelihood depends, thus creating the background for a clear conflict of interest.
As expected, polygraph examiners will usually confidently state that the exam is highly accurate, around 95 percent. This implies that if 100 guilty suspects are given a polygraph exam, ninety-five of them will be detected through the test, meaning that only five of these 100 will be a false negative and not be detected by this method. On the other hand, they will state that if you are telling the truth, you have almost a 100 percent chance of being cleared by the test (Reid and Inbau 1977).
However, to be acceptable to modern scientific standards, the two characteristics that must be established are the sensitivity and specificity of this test: Sensitivity is that percentage of positive test results when lying is known to exist. Specificity is the percentage of negative results in the absence of prevarication. Under both scenarios, to establish the test’s accuracy, the presence of both lying and honesty must be determined independently of the test procedure itself. True accuracy must also be derived from real-life conditions because, for obvious reasons, it cannot be derived from volunteers in laboratory settings that lack the emotional pressures of real suspicion. To accomplish this, a group without known prevarication is tested to assess the results. But absolute proof of honesty is evasive in such individuals, because even if lying is absent, the anxiety associated with the test may cause false positive test results, reducing test specificity, which is a major confounder.
But what about test results in those who are actually guilty? We have little knowledge of the frequency with which liars are judged to be truthful. Moreover, guilty subjects—and many others—can purposely control their reactions using what are called “countermeasures,” sufficient to confuse the results enough to produce a “false negative.” One outstanding example of a false negative was that of Aldrich Ames, who in 1995 had successfully passed five polygraphs during his long career in intelligence, and, despite this, he was subsequently arrested and convicted of spying. After Sandia National Labs Senior Scientist Alan P. Zelicoff published a strong commentary in Skeptical Inquirer calling polygraphs “a dangerous ruse” (Zelicoff 2001), Ames himself wrote a remarkable letter to the editor from federal prison confirming Zelicoff’s points, adding that polygraphs are indeed “junk science,” “a superstition,” and “a refuge from responsibility.” “Like handing fate to the stars, entrails on the rock, bureaucrats can abandon their duties and responsibilities to junk scientists and interrogators masquerading as technicians,” Ames wrote (Ames 2001). In 2003, another example was provided by Gary Ridgway, who eventually was found to be the Green River Killer, having murdered forty-nine women in the Seattle area. Ironically, Ridgway had passed a lie detector test in 1987, while another man—who was proved to be innocent—failed. Although such isolated examples are admittedly anecdotal, they raise suspicions that the test itself may be flawed and should be carefully scrutinized through critical scientific means that employ large numbers of tests, as described below.
The polygraph test was not subjected to modern scientific investigation until three decades ago (Saxe et. al 1983; Lykken 1981). Since then there have been several studies employing improved methodology. Despite the impossibility of achieving a completely satisfactory research design, these tests clearly refute the high accuracy previously claimed. These studies have appeared in reputable peer-reviewed journals (Horvath 1977). They generally report a sensitivity ranging around 76 percent. This means that out of 100 liars only seventy-six of them will be detected by the polygraph. But adding doubt about this apparent test sensitivity is the fact that establishing true “guilt” often cannot be dissociated from prior polygraph testing, which may have contributed to confessions and/or guilty verdicts in court, skewing the data toward a falsely high sensitivity value of the test. The testing examiner also may have prior suspicions about the honesty of the examinee, which can in turn cause bias in the test’s interpretation and further skew the data. The skill of the examiners also is associated with considerable variability of results, creating another major source of inaccuracy. Some smaller police departments with limited budgets may designate a police officer to be their department examiner instead of using an outside professional examiner. The designated testing officer may have little or no formal training, other than the limited training provided by the company selling the instrument to the police department. This can lead to situational bias based upon an officer’s predisposition to believe most suspects are generally guilty. As a result, the officer may categorize an inconclusive result as untruthful. If the testing officer is aware of the circumstances of the individual being tested, and the weight and extent of the evidence indicating the individual being tested is the likely perpetrator of the crime, unbiased testing is virtually impossible. Officers are also generally trained to ask questions using phrasing that may cause someone to answer in a manner that may appear to be an admission of guilt. This type of questioning format would be inappropriate in a polygraph examination. For these various reasons given above, establishing true sensitivity of this test is therefore unlikely ever to be achieved, but the estimations above are likely overstated.
Even more damning—but unsurprising—is that studies report an average specificity of 52 percent, meaning that out of 100 people who are not lying, only fifty-two will be identified as telling the truth while forty-eight of these honest individuals will be branded as liars. These odds are similar to a coin toss, which would have a specificity of 50 percent. Other studies (Brett et al. 1986; Kleinmuntz and Szucko 1984; Lykken 1981) have shown even lower specificity values, indicating that “positive” test results are virtually useless. In 2003, the National Academy of Sciences, after a comprehensive review issued a report, The Polygraph and Lie Detection (National Academy of Sciences 2003), stating that the majority of polygraph research was “unreliable, unscientific and biased” and concluding that fifty-seven of approximately eighty research studies—upon which the American Polygraph Association relies—were significantly flawed. It concluded that, although the test performed better than chance in catching lies—far from perfect—perhaps most important, the test produced too many false positives.
If anything, there is perhaps one minor advantage to subjecting suspected felons to such testing (Lykken 1981; Lykken 1991): 25 to 50 percent of examinees will, under the intense psychological pressure of the exam, confess to the misdeed at hand, having been persuaded that they have been proven dishonest by “scientific” means. It is usual for the polygraph examiners to interrogate the subjects whom they believe have “failed” the test. Examiners may state that there is no way now to deny the objective guilt demonstrated by this “impartial” scientific device, and that the only available option is to confess, which they often do. But, while perhaps effective, this in no way exonerates the test itself. Perhaps various forms of torture such as water boarding might be just as effective. And one might ask further how often this form of interrogation might lead innocent persons to falsely admit, out of fear or other threats, to some form of wrongdoing.
Justification for Continued Use?
For all these reasons, the continued use of polygraph lie detection has the potential to cause much harm to those many innocents who are falsely judged dishonest by its results. A single failure could conceivably ruin one’s life. Since 1923, polygraph evidence has not been admissible in federal court cases because the test was deemed to lack scientific validity. Sadly, however, it is still used widely by the court systems of many states. Moreover, suspects are frequently “offered” this test prior to criminal proceedings, but if, for any reason, they decline to be tested, this refusal alone may cause them to be presumed guilty. Conversely, if they consent to be tested, they are risking the commonly occurring false positive outcome, which in the view of many prosecutors and juries, supports a guilty verdict. Thus from the standpoint of the accused, he/she is caught in a catch 22 situation.
Even more regrettable is the attempted application of this test for pre-employment or security clearances. In this context, testing large groups with a low base rate of dishonesty will disclose a large number of false positive responses, as encompassed in the mathematical principles of Bayes theorem (Tavel 2012). Understanding these concepts, the American Medical Association’s Council on Scientific Affairs (1986) recommended that the polygraph not be used in pre-employment screening and security clearance, with which I fully concur.
Adding clear restrictions to this testing, The Federal Employee Polygraph Protection Act, passed in 1988, virtually outlawed using polygraphs in connection with employment. That law covers all private employers in interstate commerce, which includes almost every private company that uses a computer, the U.S. mail, or a telephone system to send messages to someone in another state.
Under the Act, it is illegal for all private companies to:
- Require, request, suggest, or cause any employee or job applicant to submit to a lie detector test;
- Use, accept, refer to, or inquire about the results of any lie detector test conducted on an employee or job applicant;
- Dismiss, discipline, discriminate against, or even threaten to take action against any employee or job applicant who refuses to take a lie detector test.
Federal applicants and employees are also generally protected from lie detector tests by civil service rules. Despite all these apparent safeguards, they often must submit to a polygraph examination in the quest of a coveted security clearance for federal employment or to retain such a job. The Employee Polygraph Protection Act allows polygraph tests to be used in connection with jobs in security and handling drugs, or in investigating a specific theft or other suspected crimes.
As a result, these examinations continue to be used by federal agencies such as the FBI, CIA, and National Security Agency, where they are commonplace in decisions concerning employment. Applicants might expect them as a form of initial screening even before they start working at the agency. They may also be required to take follow-up polygraph tests from time to time. Prior to a likely job offer, the polygraph test is often the last remaining hurdle. Under these circumstances, the need to pass can be a very nervous event for anyone—especially those who have not been subjected to a polygraph before, and of course, as explained, this can easily trigger a false positive response, resulting in an unjustified rejection. Thus use of such a test for this purpose is very difficult to defend.
Given the fallacies of such testing, one would assume that enterprising individuals would provide instruction to subjects on how to pass such a test (through “countermeasures”), thus leading to the conclusion—rightly or wrongly—that one is being truthful. Moreover, the prior administration of certain drugs to block the sympathetic nervous system could also be expected to impair test accuracy. Actually, instruction on passing these tests is easily available on some Internet sites. Therefore it is difficult to understand why anyone providing personal instruction to this end would actually be prosecuted as if this were a felony. This is exactly what occurred recently (Taylor 2013) when an Indiana man was accused of “threatening national security” by teaching government job applicants and others how to pass lie-detector tests. He was sentenced to eight months in prison after federal agents had targeted him in an undercover sting. At the time of this writing, at least one additional case is pending on similar grounds. Although legal intricacies extend beyond this discussion, this issue appears to threaten First Amendment rights. Moreover, that such instruction is even possible exposes the flaws of a procedure largely regarded as pseudoscience by the scientific community. To exemplify this point further, let us present an intriguing—although farcical—analogy: Suppose we discovered that, through coaching, one could teach persons how to transfer firearms past metal detectors at air terminals without triggering alarms? Such a revelation would logically cause one of two reactions by the supervising authorities: 1) Eliminate such a flawed test in favor of one that is accurate; or 2) Attempt to silence those coaching this “deception” with threatened legal penalties, including incarceration. The obvious answer to this question—number one—requires little thought. But with regard to polygraph testing, this answer has thus far not been chosen by our authorities, for they have exercised the second option. The mere fact that authorities have chosen to suppress information of this type might, in itself, be considered as a tacit admission that such testing is fatally flawed.
In summarizing its many pitfalls, Iacono (2001) concluded in an article titled “Forensic Lie Detection: Procedures without Scientific Basis,” the following, which I paraphrase for clarity: Although this form of testing may be useful as an investigative aid and tool to induce confessions, it does not pass muster as a scientifically credible test. Its theory is based on naive, implausible assumptions about its accuracy. It is biased against innocent individuals and can be beaten simply by artificially augmenting responses to control questions. Although it is not possible to adequately assess the error rate of this test, both of these conclusions are supported by published research findings in the best social science journals.
Given such overwhelming evidence of inaccuracy as I have presented here, how can we, as a society, react to such a perversion of science? The logical solution is to completely abandon this method of testing. The most urgent necessity is the complete elimination of this testing in connection with employment. All remaining state and federal laws that allow use of the polygraph, in or outside of court settings, should be abandoned. Moreover, as long as it remains in use, there is simply no justification for prosecuting those who provide information about how to “pass” such a test.
Unfortunately, various state and national polygraph certifying and licensing organizations—whose livelihood depends upon their own continued existence—are well entrenched in our society. Their provision of services to most law enforcement agencies creates a symbiotic relationship that is difficult to overcome. Nevertheless, to eradicate this blight the scientific community, as well as others who understand these concepts, must educate the public and relentlessly urge the responsible authorities to discontinue the present unsatisfactory status quo.
- American Medical Association Council on Scientific Affairs. 1986. Polygraph. Journal of the American Medical Association 256: 1172–75.
- Ames, Aldrich. 2001. The polygraphs controversy (letter to the editor). Skeptical Inquirer 25(6): 72–73.
- Brett, A.S., M. Phillips, and J.F. Beary. 1986. Predictive power of the polygraph: Can the “lie detector” really detect liars? The Lancet. 1: 544–547.
- Horvath, F. 1977. The effect of selected variables on interpretation of polygraph records. Journal of Applied Psychology 62: 127–136.
- Iacono, W.G. 2001. Forensic “lie detection”: Procedures without scientific basis. Journal of Forensic Psychology Practice 1: 75–86.Kleinmuntz, B., and J. Szucko. 1984. A field study of the fallibility of polygraphic lie detection. Nature 308: 449–450.Lykken, D.T. 1981. A Tremor in the Blood: Uses and Abuses of the Lie Detector. McGraw-Hill, New York.
- ———. 1991. Why (some) Americans believe in the lie detector while others believe in the guilty knowledge test. Integrative Physiological and Behavioral Science 26: 214–222.
- National Academy of Sciences. 2003. The Polygraph and Lie Detection, Committee to Review the Scientific Evidence on the Polygraph. The National Academies Press, Washington, D.C.
- Reid, J.E., and F.E. Inbau. 1977. Truth and Deception: The Polygraph (“Lie Detector”) Technique. Williams & Wilkins, Baltimore.
- Ruscio, J. 2005. Exploring controversies in the art and science of polygraph testing. Skeptical Inquirer 29(1): 34–39.
- Saxe, L., D. Dougherty, and T. Crosse. 1983. Scientific validity of polygraph testing: A research review and evaluation. Conference: OTA-TM. U.S. Congress Office of Technology Assessment.
- Tavel, M.E. 2012. Snake Oil Is Alive and Well: The Clash between Myths and Reality. Reflections of a Physician. Chandler, AZ: Brighton Press, 51.
- Taylor, M. 2013. Indiana man gets 8 months for lie-detector fraud. Seattle Times (September 6).
- Zelicoff, Alan P. 2001. Polygraphs and the national labs: Dangerous ruse undermines national security. Skeptical Inquirer, (July/August). Online at http://www.csicop.org/si/show/polygraphs_and_the_national_labs_dangerous_ruse