Can We Trust Counterintelligence Polygraph Tests?
Vance MacLaren, University of New Brunswick, Canada
Reproduced with permission from Polygraph, the journal of the American Polygraph Association.
Citation: MacLaren, V.V. (2000). Can we trust counterintelligence polygraph tests? Polygraph, 29, 2, 151-155.
Polygraph interviews are an important part of security clearance procedures in many branches of the American government. This paper briefly reviews open-source literature on polygraph security screening procedures currently in use by the US Department of Defense. Current polygraph security screening procedures make a valuable contribution to the maintenance of national security.
Recent developments at America's national laboratories have drawn public attention to the issue of polygraph security screening. In the wake of allegations of penetration of the nuclear lab's security by the People's Republic of China, the Department of Energy has initiated polygraph counterintelligence screening of some employees with access to sensitive information. Some critics argue that such tests are inaccurate, and that they infringe upon the dignity of government employees (see www.stoppolygraph.com for criticisms of security screening polygraph tests). In the present paper, it is argued that current polygraph screening practices make a valuable contribution to the maintenance of national security.
Polygraph interviews have been an integral part of security clearance procedures for both civilian and military personnel since the early days of the Cold War. They are also widely used in pre-employment screening of police recruits (Meesig & Horvath, 1995). Fear of an increase in espionage activity aimed against the United States led to a healthy skepticism of security vetting procedures in the 1980s (e.g. United States Congress Office of Technology Assessment, 1983). Polygraph screening techniques in use at that time were subjected to systematic appraisal in which several studies (e.g. Barland, Honts & Barger, 1989; Honts, 1992) found the Counterintelligence Scope Polygraph (CSP) technique to be inadequate as a means of identifying persons involved in espionage activity. As the magnitude of the Aldrich Ames spy cse was revealed in the early 1990s, it became clear that something had to be done to improve the methods used to detect subversives operating within the government and the military.
Around 1992, the Department of Defense Polygraph Institute (DoDPI) began development of an improved test to be used in security screening (DoDPI, 1994). The result was the Test for Espionage and Sabotage (TES). The TES format differs in many ways from the CSP, and is generally more standardized and less intrusive than older methods. Also, it makes use of Directed-lie comparison questions (Abrams, 1991; Honts & Raskin, 1988; Horowitz, Kircher, Honts & Raskin, 1997). According to the DoDPI annual report to Congress for fiscal year 1994 (DoDPI, 1995), by that time, "All DoD agencies [had] switched from the CSP to TES for security screening" (p. 28).
To date, only two large-scale studies designed to evaluate the accuracy of TES have been published (DoDPI Research Division Staff, 1997; 1998; see also Reed, 1994). In both studies, participants in programmed guilty groups were required to take part in very realistic simulations of espionage activity. In the two studies, 50 of the 60 guilty participants were correctly identified (83.3%). Of the 108 participants in programmed innocent groups, 98 were correctly identified (90.7%). Although further replication of these findings should be sought, it appears that the TES procedure has an acceptable level of discriminative accuracy under controlled laboratory conditions.
At present, no systematic field study of TES has been published, but some information is available in the form of DoDPI's annual reports to Congress, which present statistics on the activities of the Department of Defense screening program. In the last five fiscal years, the Department of Defense has administered 43,648 counterintelligence screening tests. This does not include testing performed under authority of other federal agencies, such as the National Security Agency and the National Reconnaissance Office. In those tests, 902 people made admissions relevant to security issues. In addition, 148 of the tests produced either "significant physiological responses" or "inconclusive" outcomes. Of the 963 cases in which either relevant admissions were made, or in which inconclusive or significant physiological responses were found, only 24 resulted in adverse action following subsequent investigation. An additional 97 were still pending investigation or adjudication at the time the reports were presented to Congress, so it is not known how those cases were resolved. The actual number of cases in which investigation subsequent to the polygraph screening revealed evidence sufficient to result in adverse action can therefore be estimated as falling between 24 and 121.
The tiny percentage (less than 0.3%) of cases in which access to information was denied or withheld may seem like a meagre harvest, in light of the expensive resources devoted to the polygraph screening program. However, if one considers the scale of damage that can be caused by individuals like Edwin Pitts, David Boone, or Harold Nicholson, even these small numbers are anything but insignificant (see intellit.muskingum.edu for information on these and other spy cases). It should also be remembered that many of the individuals identified by the polygraph screening program were also subjected to background investigations and other means of screening, which failed to identify their illegal activity. In addition, others working for the government may have been deterred from engaging in espionage activity in the first place, out of fear of being detected by periodic or aperiodic screening.
The False Negative Problem.
In a CSP study by Barland, Honts & Barger (1989), approximately 20% of the 207 participants made admissions about real-world violations of security protocols in the course of their mock-espionage CSP examinations. Based on this estimate, Honts (1994) erroneously concluded that the CSP program had an extraordinarily large rate of false negative error. While a large proportion of employees may violate security procedures at some time in their careers, the vast majority of such violations are of trivial importance. They may indicate sloppy security practices, but certainly not involvement with organized espionage. It is laughable to suppose that one in five government employees is a spy. A more reasonable estimate of the prevalence of espionage that one might draw from the Barland et al study is the number of security violations considered "significant". After all, the purpose of the screening program is not to detect petty infractions of institutional policy; it is to protect national security from legitimate threats. Of the 207 persons examined in that study, only 2 made such admissions (1%). This rate is similar to that found in the Department of Defense screening program, in which admissions were obtained in 902 of 43,648 cases (2%).
If we follow the example of Honts' (1991) conditional probability analysis and assume that 1% of persons screened by the polygraph are actually guilty of taking part in activities that might undermine national security, we might then assume that about 436 of the 43,648 people tested under the Department of Defense polygraph screening program had committed security violations of some consequence. Because only between 24 and 121 were actually identified, according to the DoDPI reports, it seems reasonable to concur with Honts' assertion that a large number of serious security violations may remain undetected, although the prevalence of false negatives is probably much less severe than he had supposed.
In any case, most of those who received adverse action subsequent to their polygraph interview would not have been detected if the screening program were not in place. No one technology can be expected to detect all subversives, but if each of the security measures used by the federal government (e.g. financial records analysis, background interviews, etc.) manages to identify a few, then they may collectively provide an effective obstacle against those who would do harm to national security.
Two Strikes and You're Out... Maybe.
In the two simulation studies of TES, 90.7% of innocent participants were correctly identified. If we tentatively assume that these results generalize to field conditions, then we must conclude that 9.3% of innocent government employees would be falsely accused of espionage by a TES exam. However, that does not appear to have happened, according to the statistics reported in the DoDPI annual reports to Congress. Why, then, are so few people classified as deceptive or inconclusive by TES? Without a proper field study of the TES procedure, it is not correct to rule out the possibility that the detection rates found in the simulation studies might not generalize to the field situation. Another possibility, as suggested by Barland, Honts & Barger (1989), is that some important precautions are used to protect employees from false accusation. No test is immune to error but, if the test is applied carefully, the negative ramifications of errors may be minimized.
One simple way to minimize possible harm to employees would be to require repeated examinations whenever significant responses or inconclusive results are found. With repeated administrations of a test, the likelihood of being falsely classified by each repetition tends to diminish. Suppose that 10,000 innocent people are given TES examinations and that 930 (9.3%) fail the test. To preclude the possibility of falsely smearing those individuals' careers, all 930 are retested. This time 844 (90.7%) of them pass the test and 86 (9.3%) fail it. The prior likelihood that an individual would fail either of two tests is p = .093 + .093 = .186, but the likelihood that an individual would fail both of the tests is only p = .093 * .093 = .0086. By simply enacting the policy that all individuals who fail one test are subsequently re-tested to verify the results, the likelihood of false positive error is greatly reduced. Under this scheme, over 99% of innocent employees would pass the security screening. Whatever the true specificity of TES, and whatever the effect of prior testing on the likelihood of false positive error in subsequent interviews, a requirement of corroborative results from repeat testing should tend to ameliorate the chances of falsely accusing innocent employees. One important, and as yet unanswered question, is the level of test-retest reliability of results found in the TES screening procedure.
By the same logic, we would expect 83.3% of guilty operatives to be identified by the first test. These employees would be subjected to added scrutiny, perhaps including the initiation of other investigative procedures, any of which migth reveal evidence of their illegal activity. In any case, the chance that a spy would fail both tests in p = .833 * .833 = .694, or approximately 69%.
Now suppose that we were to randomly test 10,000 employees, and for simplicity sake, 1% of them are spies. On the first test, 921 of the 9,000 innocent employees and 83 of the 100 guilty ones fail. On a second test, 86 innocent people fail the test, and 69 guilty ones fail. We now have 155 cases of people who failed the test twice. Of these, 45% are foreign agents, and 55% are not. Additional investigative techniques would need to be deployed to sort the innocent from the guilty. However, the investigators now have only 155 cases to sift through; not 10,000. In this scenario, use of the polygraph produces tangible benefits to the efficiency and cost-effectivencess of counterintelligence efforts.
By applying the counterintelligence polygraph tests in an orderly and careful way, large numbers of individuals can be effectively screened. This must lead to reductions in the financial expecditures related to security investigations. However, care must be exercised because false negative errors can occur, and if a person passes a polygraph screening, they should not be exempted from other screening procedures.
There is an important distinction that must be made between criminal investigative applications of polygraph tests and the types of situation involved in security screening. Security screening tests are not specific-issue tests. They are a way to probe large numbers of employees in search of a tiny proportion that are involved in undesirable security breaches. It is a needle - in - a - haystack problem. In some respects, the accuracy of the technique may be of importance secondary to its utility. Whereas the identification of individual saboteurs and spies may be important, knowing the nature of their activities may be equally useful. Admissions obtained in the course of polygraph testing may be a valuable source of counterintelligence information. Consider the following excerpt from the Office of Technology Assessment report (1983):
"It appears that NSA (and possibly CIA) use the polygraph not to determine deception or truthfulness per se, but as a technique of interrogation to encourage admissions. NSA has stated that the agency does not use the 'truth v. innocence' concept of polygraph examinations commonly used in criminal cases. Rather, the polygraph examination results that are most important to NSA security adjudicators are the data provided by the individual during the pretest or posttest phase of the examination." (p. 100).
In some cases, such "data" may be very important, indeed. Whether or not the squiggly lines produced by a polygraph can detect deception is an academic dispute that bears little importance to the applied issue of whether the technique is a useful, fair and cost-effective contributor to protecting the nation's most treasured secrets.
The polygraph technology used in the Department of Defense screening program appears to have considerable validity and utility. Previous assessments of the situation (e.g. Honts, 1994; 1991) were written at a time when older techniques (i.e. CSP) were in use. The newer TES procedure appears to be a great imporvement over those methods. The results of actual examinations summarized in the DoDPI annual reports to Congress are not consistent with the idea that large numbers of innocent public servants, civilian contractors and military personnel are being unfairly "flutered" out of their jobs (e.g. Lykken, 1998). The polygraph screening program has a respectable record of identifying persons with hostile or selfish intentions that are contrary to the maintenance of national security.
Abrams, S. (1991). The directed lie control question. Polygraph, 20, 26-31.
Barland, G.H., Honts, C.R. & Barger, S.D. (1989). Studies of the accuracy of security screening polygraph examinations. Department of Defense polygraph institute, Fort McClellan, AL.
Department of Defense Polygraph Institute (1994). Polygraph program, United States Department of Defense. Polygraph, 23, 61-84.
Department of Defense Polygraph Institute (1995). Department of Defense polygraph program report to Congress for fiscal year 1994. Polygraph, 24, 13-33.
Department of Defense Polygraph Institute (1996). Fiscal year 1995 report to Congress on the Department of Defense polygraph program. Ft Jackson, South Carolina: Author.
Department of Defense Polygraph Institute (1997). Department of Defense polygraph program annual report to Congress for fiscal year 1996. Polygraph, 26, 240-254.
Department of Defense Polygraph Institute research Division Staff (1997). A comparison of detection accuracy rates obtained using the counterintelligence scope polygraph and the test for espionage and sabotage question formats. Polygraph, 26, 79-106.
Department of Defense Polygraph Institute (1998). Fiscal year 1997 report to Congress on the Department of Defense polygraph program. Polygraph, 27, 171-180.
Department of Defense Polygraph Institute research Division Staff (1998). Psychophysiological detection of deception accuracy rates obtained using the test for espionage and sabotage. Polygraph, 27, 68-73.
Department of Defense Polygraph Institute (1999). Fiscal year 1998 report to Congress on the Department of Defense polygraph program. Ft Jackson, South Carolina: Author.
Honts, C.R. (1991). The emperor's new clothes: Application of polygraph tests in th American workplace. Forensic Reports, 4, 91-116.
Honts, C.R. (1992). Counterintelligence scope polygraph (CSP) test found to be poor discriminator. Forensic Reports, 5, 215-218.
Honts, C.R. (1994) Psychophysiological detection of deception. Current Directions in Psychological Science, 3, 77-82.
Honts, C.R.& Raskin, D.C. (1988). A field study of the validity of the directed lie control question. Journal of Police Science and Administration, 16, 56-61.
Horowitz, S.W., Kircher, J.C., Honts, C.R. & Raskin, D.C. (1997). The role of comparison questions in physiological detection of deception. Psychophysiology, 34, 108-115.
Lykken, D.T. (1998). A Tremor in the Blood: Uses and Abuses of the4 Lie Detector. New York: Plenum.
Meesig, R. & Horvath, F. (1995). A national survey of practices, policies and evaluative comments on the use of pre-employment polygraph screening in police agencies in the United States. Polygraph, 24, 57-136.
Reed, S. (1994). A new psychophysiological detection of deception examination for security screening. Psychophysiology (Abstract), 31, S80.
United States Congress Office of Technology Assessment (1983). Scientific Validity of Polygraph Testing: A Research Review and Evaluation - A Technical Memorandum. OTA-TM-H-15. Washington, D.C.: U.S. Government Printing Office.