News Conference

Opening Statement
by

Stephen E. Fienberg

Professor of Statistics and Computer Science, Carnegie Mellon University, Pittsburgh
and
Chair, Committee to Review the Scientific Evidence on the Polygraph

Good morning. On behalf of the National Academies and my colleagues on the committee, I welcome those of you in the room as well as those listening to the live audio webcast. We are pleased to be here to release our report, The Polygraph and Lie Detection. I will spend a few minutes outlining the major findings and conclusions of the report, and then we will answer your questions, including those sent by e-mail from listeners on the Web.

Our committee was charged with conducting a scientific review of the research on polygraph examinations, particularly when they are used for personnel security screening. Tens of thousands of individuals undergo polygraphs by federal agencies every year. We also looked at other screening techniques, such as those involving brain-related electrical activity, voice tremor, and thermal imaging. And we considered the security needs of the labs that are governed by the U.S. Department of Energy, which sponsored this study. It is important to emphasize that our report does not specifically assess the validity or the usefulness of polygraph testing for other purposes, such as criminal investigations.

Polygraph exams include a pre-test interview between the examiner and test taker, followed by interrogation that is coupled with the use of a device that records an individual's physiological responses. Government agencies use these exams in personnel security screening in an attempt to identify spies and other major security risks among current or prospective employees. Agencies also use them to deter employees from breaking security rules.

Our main policy conclusion is that the polygraph's accuracy in distinguishing actual or potential security violators from innocent test takers is insufficient to justify the federal government's reliance on this method for employee screening. We believe that testing yields a choice between two unsatisfactory results: Either too many loyal employees may be falsely judged as deceptive, or too many major security threats could go undetected. 

We also concluded that overconfidence in polygraph screening presents a danger to national security objectives. The belief in its accuracy goes beyond what the evidence supports, and this overconfidence can create a false sense of security among policy-makers, employees in sensitive positions, and the general public. This may lead, in turn, to the neglect of other methods of ensuring safety, such as periodic security reviews. 

In our examination of several techniques that have been proposed as alternatives or enhancements to the polygraph, we found that some show promise. However, there is no viable alternative in hand that can reliably identify actual or potential threats to security. The shortcomings of the polygraph and all other available techniques for employee screening underscore the importance of finding new ways to pursue security goals. 

At this point, I will discuss how the committee went about its work, as well as other key findings from the report.

We considered both the validity and utility of the polygraph. Validity has to do with whether the tests can spot deception. Utility has to do with whether the exams lead to admissions of security violations or deter security threats. 

We examined studies of how closely the act of deception is linked to the psychological and physiological phenomena that determine test results. We also looked at carefully conducted research studies of how well results correspond to actual truthfulness or deception by specific examinees.

The science base is severely limited in three key respects. First, there is almost no evidence assessing accuracy in realistic security-screening situations. Second, we found no studies on testing alleged terrorists and spies. And third, there is very limited evidence on whether efforts to beat the tests, known as countermeasures, can deceive experienced examiners.

Basic scientific knowledge on physiological responses leads to the expectation that testing would have some diagnostic value, at least among naive examinees. However, the research base shows only limited correspondence between deception and the physiological responses monitored by the polygraph. In particular, responses typically viewed as an indication of deception can have other causes, so polygraph testing is intrinsically susceptible to producing errors. The technique's accuracy also varies across situations and probably depends on the examiners' expectations as well as on the reason people are being tested -- whether, for example, they are criminal suspects or job applicants. Moreover, countermeasures may be effective. 

The difficult policy choice involved in security screening is illustrated in a table in the report's executive summary. The table shows the results of a hypothetical polygraph screening program. The example assumes that a federal agency uses the test to screen 10,000 employees, that 10 of them are spies, and that none of the spies uses countermeasures. If the test were set sensitively enough to detect 80 percent of spies, it would misclassify and pass two spies. About 1,600 people, virtually all of them innocent, would have failed. And the agency would have to weed out the eight spies from this large group. If the agency set the test to reduce the number of innocent people who fail to about 40, the test would catch only two spies, and eight would go completely undetected. In the committee's view, neither of these possibilities is acceptable. National security is too important to be left to such a blunt instrument.

Even if the validity of the technique is weak, could it still have utility? Indirect evidence supports the idea that the polygraph may have utility for deterring violations and eliciting confessions, but this utility depends on examinees believing that the polygraph is accurate. In the long run, utility depends on validity.

The Energy Department and some other federal agencies now operate polygraph programs that yield only a small percentage of positive test results. These programs may have value for deterring security violations, but screening will not spot more than a small proportion of major security violators. The tests may deter, but they will not do a good job of detecting.

The committee considered how research might improve the government's ability to identify and thwart major security risks. Given the high stakes, an expanded research effort could be worthwhile. We believe that such an effort should be broad-based – and not focused narrowly on the polygraph. It should be centered on methods for detecting and deterring major security threats, including efforts to improve screening techniques, and should include both basic and applied research. To ensure the highest quality science, a considerable part of the program should be run by an impartial organization that is not responsible for any aspect of detecting deception.

Many of the issues that we studied are not new. For as long as human beings have deceived one another, people have tried to develop ways to detect lies and uncover the truth. It's up to the research community to point out when these techniques rest on a weak scientific foundation. We hope our report will inform public policy in this area.

My colleagues and I will now take your questions. Reporters in the room, please step to a microphone. Whether you are here or participating over the Web, please be sure to identify yourself and your news organization. Thank you.