If you have issues viewing or accessing this file contact us at NCJRS.gov.

r I National Criminal Justice Reference Service l------~~------I: nCJrs

This microfiche was produced from documents received for inclusion in the NCJRS data base. Since NCJRS cannot exercise control over the physical condition of the documents submitted, the individual frame quality will vary. The resolution chart on this frame may be used to evaluate the document quality.

1.0

.0 1 1 11111 . 'I i -:0 111111.25 111111.4 111111.6

,

fI. i'

MICROCOPY RESOLUTION TEST CHART U.s. OGpartment of Jutrtlce I MaUontl !Mtltutll 01 Ju&tice NATIONAL BUREAU OF STANDARDS-1963-A I ThIs document has boon reproduced exactly as roostved from the ~ or OI'ganizatlOO OOglNltJl'l!J it. Points of view 01' opinions stated in thiS document oro thoSG of Iho authors and do not nocessC'lril> reprooGot tho QlIicial pos!tJon 01' poUcios of the National Institute 01 Justioo.

Permlsslon to reproduoo this ~ malerial has been gmnlodby It, Microfilming procedures used to create this fiche comply with Public Danain/United States CQngress the standards set forth in 41CFR 101-11.504. Office of Technology Assessment lathe NBlional Crlmlnol JustIeD Reference Servlee (NCJRS).

Points of view or opinions stated in this document are Fw1h01' rcpfOduetloo cutGIda of tho NC.lRS $)'tIlem requires pormls. those of the author(s) and do not represent the official !':lIon of tho COft'lfllh O'Mlor. position or policies of the U. S. Department of Justice.

a'. \ I J I National Institute of Justice ,f United States Department of Justice t Washington, D. C. 20531 , il\ '1 ( 6/1/84 4. ; 4 Q2E

·0 o

SCIENTIFIC~-. ... VALIDITY OF POLYGRAPH TESTING I A RESEARCH REVIEW AND EVALUATION

A TECHNICAL rv1EMORANDUM

NOVEMBER 1983

Technical Memoranda are issued by OTA on specific subjects analyzed in recent OTA reports or on projects presently in process at OTA. They are issued at the request of Members of Congress who are engaged in committee legislative actions which are ~. :c· ·~c- "1~(';,:"~: ~_ expected to be resolved before OTA completes its assessment. :. .'. ". ------~---~~ r->r-~~ ....""'- ....- ...... ~- ...... ,~...----~~- ~- -~ ------.-----~-~~ - -~ ------~ ------~. ------

Foreword

This technical memorandum presents the results of the Office of Technology Assess­ ment's (OTA) review and assessment of the scientific evidence on the validity of polygraph testing. Conducted at the request of Rep. Jack Brooks, Chairman, House Committee on Government Operations, and Rep. Frank Horton, the Ranking Minority Member, the OTA memorandum is intended to assist the committee in its deliberations on pro­ posed changes in polygraph use by the Federal Government. As requested, OTA has limited this technical memorandum to issues directly related to the scientific validity of the polygraph. OTA did not consider utility, privacy, con­ stitutional, and ethical issues, among others that have been raised in the debate over polygraph testing. We first discuss the various types of polygraph testing procedures and ways in which the polygraph is used, and then summarize the judicial, legislative, and scientific con­ troversy over polygraph testing validity. Next, we review and evaluate both prior reviews of the scientific research on polygraph validity and the individual research studies. Finally, we discuss the range of factors that may affect polygraph validity and the possibilities for future research, and present OTA's conclusions about the scientific evidence for cur­ rent and proposed Federal Government polygraph use. In preparing this memorandum, OTA has drawn on research information available from a wide variety of sources, including the major Federal Government polygraph users, the American Polygraph Association, various private polygraph practitioners, and polygraph researchers both in the United States, and abroad. Recommended Citation: Scientific Validity of Polygraph Testing: A Research Review and Evaluation-A Techni­ In addition to the members of the project advisory panel, this memorandum benefited cal Memorandum (Washington, D.C.: U.S. Congress, Office of Technology Assessment, from the consultation and review of a large number of persons in the Federal Govern­ OTA-TM-H-15, November 1983). ment, universities, and the polygraph community. It is, however, solely the respon­ sibility of OTA, not those who advised and assisted us in its preparation.

Library of Congress Catalog Card Number 83-600618

For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 Polygraph Validity Advisory Panel OTA Polygraph Validity Project Staff

Edward S. Katkin, Chairman Professor of Psychology, State University of New York at Buffalo John H. Gibbons, Acting Assistant Director, OTA Joseph P. Buckley Steve Pruitt Health and LiFe Sciences Division * President Director of Congressional Affairs John E. Reid & Associates Public Employees Department, AFL-CIO Clyde Behney, Health Program Manager Robert Edelberg Christopher H. Pyle Professor of Psychiatry and Psychology Associate Professor of Politics Fred B. Wood, Project Director UMDNJ-Rutgers Medical School Mt. Holyoke College (Communication and InFormation Technologies Program (CIT)) David C. Raskin Frank Horvath Professor of Psychology Leonard Saxe, Principal Investigator and Author (Boston University)** Associate Professor University of Utah Denise Dougherty, Co-author and Analyst (Health)f School of Criminal Justice Michigan State University Harold Sigall Theodore Cross, Co-author (Boston University)tt Professor of Psychology Jack Langenbrunner, Analyst (Health) David T. Lykken University of :Maryland Profesor of Psychiatry and Psychology George B. Trubow Katherine Locke, Research Assistant (Health) University of Minnesota Medical School Professor of Information Law and Policy Administrative Support Gail J. Povar The John Marshall Law School Ginny Cwalina, Assistant Professor of Medicine and Althea M. 1. Wagman Administrative Assistant (Health) Health Care Sciences Research Associate of Psychiatry Elizabeth Emanuel, Administrative Assistant (CIT) The George Washington University Medical Maryland Psychiatric Research Center Jennifer Nelson, Secretary (Health) Center University of Maryland School of Medicine Shirley Gayheart, Secretary (CIT) Other Contributors Michael Saks, Boston College Daniel Ozer, Boston University Yo ram Bar-Tal, Boston University Mary Beasley, Boston U{liversity Marie Calabrese, Boston University Molly Zane, Boston University

OT A Publishing Staff

John C. Holmes, Publishing Officer John Bergling Kathie S. Boss Debra M. Datcher Joe Henson Glenda Lawing Linda A. Leahy Cheryl J. Manning

'Since September 1983; H. David Banta served as Assistant Director until August 1983. "Assistant Professor of Psychology, Boston University. tOTA Analyst since September 1983; Boston University Graduate Research Assistant until August 1983. ttGraduate Research Assistant, Boston University, iv v """t .' • ------.------~ ---- ~----~------..

\ \ I

Acknowledgments Contents

OT A acknowledges the contribution of the following agencies and individuals that provided information, Chapter Page advice, and/or substantive reviews of draft materials: 1. SUMMARY Introduction ...... 3 4. REVIEW AND ANALYSIS OF POLYGRAPH Federal Government Agencies Fitz Godwin, Army Federal Polygraph Use ...... 3 FIELD STUDIES...... 47 Lewis R. Goldberg, University of Oregon Federal Polygraph Policy Changes ...... 3 Introduction...... 47 Department of Defense (Office of Deputy Under Gisli Gudjonson, London Institute of Psychiatry Secretary of Defense for Policy, Army, Air Force, Polygraph Validity ...... 4 Study Selection...... 47 David 1. Hammond, Psychologist Findings .... , ...... 4 Studies Excluded ...... 48 Navy, Marines, National Security Agency) James E. Hardy, Air Force Office of Special Department of Justice (Federal Bureau of Investigation, Personnel Security Screening ...... 4 Studies Included...... 49 Investigations Criminal Investigations ...... 5 Coding...... 51 Drug Enforcement Administration) Fred Hegge, Walter Reed Army Institute of Research Department of State False Negatives/Countermeasures ...... 5 Findings and Discussion ...... 51 Eric J. Holden, Baker, Holden & Associates Department of the Treasury (Secret Service, Bureau False Positives ...... 5 Variation Among Studies...... 52 Charles Robert Honts, University of Utah Voluntary v. Involuntary ...... 6 Studies Using Panel Criterion of Alcohol, Tobacco, and Firearms) William G. Iacono, University of British Columbia Postal Service Polygraph Theory ...... 6 and Examiners' Decisions ...... 52 Michel Pierre Janisse, University of Manitoba Office of Personnel Management Further Research ...... , ...... 6 Studies Using Confession as a Criterion Richard C. Johnson, Polygraph, Inc. Chapter-by-Chapter Overview ...... 6 and Blind Evaluations ...... 54 Central Intelligence Agency Scott Kingsley, National Security Agency Director of Central Intelligence (Security Committee) Conclusions ...... 7 Studies Using Judicial Outcomes Benjamin Kleinmuntz, University of lllinois at Chicago and Original Examiners' Results .... . 56 Circle 2. VARIETIES OF POLYGRAPH TESTING Other Considerations ...... 56 Individuals Brian E. Lynch, Canadian Police College AND USES ...... 11 Conclusions ...... 58 Lynn P. Marcy, American Polygraph Association Introduction...... 11 Maynard Anderson, Department of Defense Paul K. Minor, Federal Bureau of Investigation Polygraph Instrument...... 11 5. REVIEW AND ANALYSIS OF POLYGRAPH Norman Ansley, National Security Agency 1. D. Noland, Army Intelligence and Security Types of Testing Procedures...... 12 ANALOG STUDIES...... 61 Gordon H. Barland, Barland & Associates Command The Pretest Interview ...... 12 Introduction...... 61 Antonio S. Barrio, Central Texas College System Jesse Orlansky, Institute for Defense Analysis Louise Becker, Congressional Research Service Types of Questions...... 14 Characteristics of Analog Studies...... 61 John A. Podlesny, Federal Bureau of Investigation Polygraph Question Techniques...... 17 Study Selection ...... 62 William H. Bell, Department of Defense Roland Radloff, National Science Foundation Post-Test Interview ...... 23 Description of Studies...... 65 B. F. Bloomingburg, Naval Investigative Service Richard K. Riegelman, George Washington University Irv Boker, General Accounting Office Uses of Polygraph Testing ...... ~:;...:. .. 23 Control Question Technique...... 65 Medical Center Current Use ...... 23 University of Utah Studies...... 65 Robert A. Brisentine, Army Criminal Investigation r .. .. John Roberts, Civil Liberties Union of Massachusetts Conclusions ...... ",t'. . . .. ~;; Other St\ldies ...... 68 Command Robert Rosenthal, Harvard University Henry 1. Canty, Consultant Akihiro Suzuki, Japanese Institute of Police Science 3. CONTROVERSY OVER POLYGRkH li'¥J Richard P. Clayberg, SRI International C;; R~~~:~.I~~~~t~~~.~~~~s.:::::::::::: ;~ Julian J. Szucko, University of Illinois at Chicago TESTING VALIDITY ...... ~...... 29 Davidson...... 73 Eileen Correa, Veterans Administration Medical Circle Center Introduction i J ~'J 1 ~) Podlesny and Raskin...... 73 William A. Thomas, American Bar Foundation Judicial : : : : : : : : : : : : : : :;'. : : : : : :: i,:'}'1,,? Giesen Rollison...... 73 Ron Decker, Army Military Policy School Revi~~s' 29 ~nd Howard W. Timm, Southern Illinois University of Polygraph Findings as Evidenc~ .. A C. 30 Balloun and Ho~mes ...... 74 Jack Donnelly, Department of Defense Carbondale William R. Fedor, Department of Defense Laws Regula~ing Polygraphs it! Q U ff S E1'10 Bradley and Jamsse ...... 75 H. Herbold Wooten, Institute of Applied Polygraph Employment Settings ..... 31 rtSm ...... : ...... 75 Michelle Fine, University of Pennsylvania !...... Science Federal Debate Over Polygraph Validity.. 32 Preemployment Screerung ...... 75 Robert J. Gatchel, University of Texas at Dallas Lawrence S. Wrightsman, University of Kansas The 1960's ...... , ...... 32 Correa and Adams...... 75 The 1970's 34 Barland ...... 76 The 1980's : : : : : : : : : : : : : : : : : : : : : : : : :: 35 Findings ...... 78 Draft Revisions to DOD 5210.48 ...... 36 NSDD-84...... 36 6. FACTORS AFFECTING POLYGRAPH Administration Polygraph Policy ..... , 36 EXAMINATION VALIDITY ...... 83 Scientific Validity and Polygraph Introduction ...... 83 Research Reviews...... 37 Polygraph Examiner, Subject, and Setting. 83 Definitions of Scientific Validity...... 37 Examiner ...... 83 Research Design ...... 38 Subjects ...... 84 False Positives and Negatives ...... 38 Setting ...... 86 Reviews of Polygraph Validity...... 39 Countermeasures ...... 87 Conclusions ...... 43 Physical ...... , ...... 87

vi vii !. •.• """~~- \, I r I. \. /I Drugs...... 88 list of Tables . ~-;I Hypnosis/Biofeedback ...... 89 Mental...... 89 TabJe No. Page Research Implications of Factors 1. Polygraph Examinatiomi Conducted Affecting Validity ...... 91 by Federal Agencies, 1982 ...... 24 Theory...... 91 2. Reviews of Field Studies of Scoring ...... , .. . .. 91 Polygraph Validity ...... 40 Question Techniques...... 91 3. Characteristics of Field Studies ...... 48 Countermeasures ...... 91 4. Mean Detection Rates as a Percentage Comparison With Other Techniques. " 92 of Total in Field Studies ...... 52 Field Studies ...... 92 5. Outcomes of Field Studies of Validity ... . 53 Conclusions ...... 92 6. Outcomes of Control Question Analog Studies of Validity ...... 63 7. Outcomes of Guilty Knowledge Chapter 1 7. CONCLUSIONS...... 95 Analog Studies of Validity ...... 64 Introduction...... 95 8. Mean Detection Rates as a Percentage Overall Scientific Conclusions ...... 95 of Total in Analog Studies of Summary Specific Scientific Conclusions in Control Question Technique ...... 65 Policy Context ...... 97 9. Mean Detection Rates as a Percentage r.:::, Specific-Incident Criminal of Total in Analog Studies of Investigations...... 97 Guilty Knowledge Test ...... 65 Personnel Security Screening...... 99 B-1. Number of P.olygraph Exams False Negatives/Countermeasures...... 100 and Examiners ...... 108 Voluntary v. Involuntary ...... " 101 B-2. Purpose of Polygraph Exam ...... 109 Further Research ...... 101 B-3. Length and Results of Exams ...... 111 Concluding Comment ...... " 102 B-4. Long-term Confirmation of Exam Results ...... 111 Appendixes B-5. Selected Federal Agency A. Informed Consent Forms...... 105 Polygraph Research ...... 113 B. Results of the OTA Survey of Federal Government Polygraph Testing ...... 107 List of Figures C. Coding Form...... 114 Figure No. Page D. Acronyms and Glossary . , ...... , 120 1. Debate at the Federal Level 1963-83 ...... 33

I,;,

viii ---~~------.~. - ~----.---. --.....--- .. ----~ ,~.",,",""""'. \ I

r Chapter 1 Summary

INTRODUCTION

The primary purpose of this aTA technical The aTA technical memorandum is limited to '> memorandum is to review and evaluate current a critical review and evaluation of prior research. scientific evidence about the validity of polygraph The memorandum does not consider, in detail, testing. This memorandum responds to the Feb­ other polygraph issues such as utility, ethics, im­ ruary 3, 1983, letter of request from the Commit­ pact on employee morale and productivity, pri­ tee on Government Operations, U.S. House of vacy, and constitutional rights. The technical Representatives, and the need to provide infor­ memorandum, instead, focuses on the nature and mation that is relevant to congressional considera­ application of polygraph tests, scientific contro­ tion of the polygraph aspects of the President's versy over polygraph testing, data from field and National Security Decision Directive-84 (NSDD- simulation studies, and factors that affect test 84), proposed revisions to Department of Defense validity. (DOD) Directive 5210.48 governing the DOD polygraph program, and the recently revised ad­ ministration policy on polygraph use by Federal agencies.

FEDERAL POLYGRAPH USE aTA found that Federal Government use of ing--preemployment, preclearance, periodic, or polygraph tests has more than tripled over the last aperiodic-to establish initial and continuing 10 years, with about 23,000 examinations con­ eligibility for access to highly classified informa­ ducted in 1982 compared to about 7,000 exams tion. However, NSA accounted for almost half in 1973. Current use now exceeds the previous of all Federal polygraph examinations adminis­ known peak level of use (about 20,000 exams) in tered in 1982. Federal agencies at present make 1963. In all Federal agencies except the National only limited use of the polygraph for investiga­ Security Agency (NSA) and the Central Intelli­ tion of unauthorized disclosure of sensitive or gence Agency (CIA), more than 90 percent of classified information-261 examinations (exclud­ polygraph testing in 1982 was for criminal inves­ ing NSA and CIA) for this purpose over the 1980- tigations. Only NSA and CIA make significant 82 period. use of the polygraph for personnel security screen-

II Ii Ii Ii Ii FEDERAL POLYGRAPH POLICY CHANGES [I 11 The March 1983 draft proposed revisions to the would be on an aperiodic (i.e., irregular) basis. II 'iIi DOD polygraph regulations (5210.48) authorize These expanded uses of the polygraph would be "1\ ;\ the use of polygraph tests to determine initial and part of DOD personnel security screening. :J continuing eligibility of DOD civilian, military, 1\ and contractor personnel for access to highly clas­ Also, the proposed revisions to DOD 5210.48 'i d" sified information (Sensitive Compartmented In­ provide adverse consequences for refusal to take ;\ formation and/or special access). The use of poly­ a polygraph examination, when established as a graph tests in determining continuing eligibility requirement for selection or assignment or as a \

3 <~. I, I , I

4 5

condition of access. Refusal to take an examina­ would also permit Government-wide polygraph A 1980 survey conducted by the Director of deception better than chance, but with error rates tion may, after consideration of other relevant fac­ use in personnel security screening of employees Central Intelligence Security Committee con­ that could be considered significant. tors, result in nonselection for assignment or (ar.d applicants for positions) with access to highly cluded that the polygraph was the most produc­ In a typical criminal investigation, the poly­ employment, denial or revocation of clearance, clfl.ssified information. The new policy provides tive of all background investigation techniques. graph, if used at all, is used only after prior in­ or reassignment to a nonsensitive position. agency heads with the authority to give polygraph However, this was a utility study not a validity vestigation has been completed, and a prime sus­ examinations on a periodic or aperiodic basis to study, and had many qualifications. pect or suspects have been identified. To the ex­ NSDD-84, issued by the President on March randomly selected employees with highly sensitive 11, 1983, authorized agencies and departments to tent polygraph use in unauthorized disclosure in­ access, and to deny such access to employees who OTA recognizes that NSA and CIA believe that require employees to take a polygraph examina­ vestigations would be similar, then the available refuse to take a polygraph examination. the polygraph is a useful screening tool. However, tion in the course of investigations of unauthor­ OTA concluded that the available research evi­ research provides some evidence of polygraph ized disclosures of classified information. NSDD- Thus, the combined effect of NSDD-84, the dence does not establish the scientific validity of testing validity. However, for so-called "dragnet" 84 also provides that refusal to take a polygraph DOD proposals, and administration policy is to the polygraph for this purpose. screening where a large number of people would test may result in adverse consequences such as authorize substantially expanded use of the poly­ be given p;iJlygraph tests in the investigation of administrative sanctions and denial of security graph for purposes of personnel security screen­ In addition, there is a legitimate concern that unauthorized disclosures, relevant research evi­ clearance. ing and unauthorized disclosure investigations. the use of polygraph tests for personnel security dence does not establish polygraph testing validi­ screening may be especially susceptible to: 1) ty. There has been no direct scientific research on On October 19, 1983, the Department of Justice countermeasures by persons trained to use physi­ this application. (DOJ) announced that administration policy cal movement, drugs, or other techniques to avoid detection as deceptive; and 2) false positive er­ False Negatives/Countermeasures rors where innocent persons are incorrectly iden­ tified as deceptive. Theoretically, polygraph testing-whether for POLYGRAPH VALIDITY personnel security screening or specific-incident Criminal Investigations investigations-is open to a large number of coun­ In 1965 and again in 1976, the House Govern­ plications, the types of individuals tested, train­ termeasures, including physical movement or ment Operations Committee concluded that there ing of the examiner, purpose of the test, and types OTA found meaningful scientific evidence of pressuref drugs, hypnosis, biofeedback, and prior was not adequate evidence to establish the validity of questions asked, among other factors, can differ polygraph validity only in the area of investiga­ experience in passing an exam. The research on of the polygraph. OTA has assessed the research substantially. A polygraph test requires that the tions of specific criminal incidents. However, countermeasures has been limited and the re­ to determine the present state of scientific evi­ examiner infer deception or truthfulness based on OTA concluded that, even here, findings about sults-while conflicting-suggest that validity dence. a comparison of the person's physiological re­ polygraph validity must be qualified. This is may be affected. OTA concludfd that this is par­ sponses to various questions. For example, there OTA concluded that no overall measure or because prior research has used widely varying ticularly significant to the extent that the poly­ are differences between the testing procedures single, simple judgment of polygraph testing types of questions, examiners, and examinees, graph is used and relied on for national security used in criminal investigations and those used in validity can be established based on available among other differences. And there is, to date, purposes, since even a small false negative rate personnel security screening. Second, the research scientific evidence. Validity is the extent to which no consistently used and accepted methodology (guilty person tested as nondeceptive) could have on polygraph validity varies widely in terms of polygraph testing can accurately detect truthful­ for polygraph research. Also, the cases selected very serious consequences. not only results, but also in the quality of research ness and deception. in field studies and situations simulated in analog design and methodology. Thus, conclusions about studies may not be representative of most actual There are two major reasons why an overall scientific validity can be made only in the con­ polygraph testing conditions. Therefore the ability False Positives measure of validity is not possible. First, the poly­ text of specific applications and even then must to generalize from the results of prior research is OTA concluded that the mathematical chance graph test is, in reality, a very complex process be tempered by the limitations of available re­ limited. of incorrect identification of innocent persons as that is much more than the instrument. Although search evidence. deceptive (false positives) is highest when the the instrument is essentially the same for all ap- OTA found a wide divergence in the results of relevant research, due in part to variations in polygraph is used for screening purposes. The rea­ research quality and design. Six prior research son is that, in screening situations, there is usually reviews showed average validity ranging from a only a very small percentage of the group being screened that might be guilty. So, in a typical situ­ FINDINGS low of 64 percent to a high of 98 percent. OTA's own review of 24 relevant studies meeting mini­ ation, there may be, perhaps, one person per 1,000 engaged in unauthorized activity. There­ Personnel Security Screening directly relevant to personnel security screening mum acceptable scientific criteria found that, for use (one by DOD). None of these studies specif­ example, correct guilty detections ranged from fore, even if one assumes that the polygraph is OTA concluded that the available research t::vi­ ically assessed validity of polygraph testing for about 35 to 100 percent. Overall, the cumulative 99 percent accurate, the laws of probability indi­ dence does not establish the scientific validity of the purposes proposed by DOD or the administra­ research evidence suggests that when used in crim­ cate that one guilty person would be correctly the polygraph test for personnel security screen­ tion, and all had serious limitations in study inal investigations, the polygraph test detects identified as deceptive but 10 persons would be ing. OTA was able to identify only four studies design.

------~------~------~--~------~----~------~-- rr~~- ...... ,...... ---~.,..,...... --~-~----~-----~------,_-____~~ ---- _ -______------~------

6 7

incorrectly identified (false positives). This poten­ than the response to other questions. However, of discussion, no consensus has emerged about tions. These analog studies were conducted pri­ tial problem has not been researched in field or the examinee's intelligence level, state of psycho­ the accuracy of polygraph tests. The chapter de­ marily in psychology laboratories using college analog studies and clearly warrants attention. logical health, emotional stability, and belief in fines scientific criteria for establishing validity and students as subjects. Like the field studies, analog the "machine" are among the several other fac­ reviews previous efforts to evaluate the scientific studies have primarily investigated the control Voluntary v. Involuntary tors that may, at least theoretically, affect physi­ literature on polygraph testing. Disagreement question technique in specific-incident criminal in­ ological responses. about the validity of polygraph testing in the sci­ vestigations, although there are some studies of NSDD-84, the DOD proposals, and administra­ entific community reflects wide variations in the an alternative ("guilty knowledge") technique for tion policy authorize adverse consequences for re­ A stronger theoretical base is needed for the en­ criteria used for inclusion of studies in prior re­ criminal investigations and two studies of preem­ fusal to take a polygraph test. Apart from the eth­ tire range of polygraph applications. Basic poly­ search reviews, differences in research design and ployment screening, one using military intelli­ ical and legal implications, which OTA did not graph research should consider the latest research definitions of validity among specific research gence personnel as subjects. While using a more address, it is generally recognized that, for the from the fields of psychology, physiology, psy­ studies, and, perhaps most important, failure to standardized methodology than field studies, the polygraph test to be accurate, the voluntary co­ chiatry, neuroscience, and medicine; comparison clearly differentiate the scientific evidence in terms analog studies had other kinds of significant re­ operation of the individual is important. Thus, among question techniques; and measures of of the purposes for which polygraph examinations search design problems, and the range of error OTA concluded that imposing penalties for not physiological response. are conducted and the techniques employed. in polygraph results was greater than in field studies. The two studies of preemployment screen­ taking a test may create a de facto involuntary Chapter 4 presents OTA's own analysis of condition that increases the chances of invalid or Further Research ing were of poor methodological quality, and did polygraph field studies in order to make an inde­ not adequately reflect screening for national se­ inconclusive test results. However, no direct re­ pendent assessment of validity. Field studies in­ OTA identified a need for further research on curity purposes. search on this topic was identified. volve real-life uses of polygraph testing. With one polygraph countermeasures, polygraph theory, exception, all of the available field evidence meet­ and polygraph validity under field conditions (for Chapter 6 discusses a number of factors that Polygraph Theory ing minimal scientific criteria comes from cases both screening and criminal investigative situa­ may affect the accuracy of polygraph examina­ involving specific-incident criminal investigations The basic theory of polygraph testing is only tions). The currently planned Federal research on tions. Some of these account for the variation in using the control question technique. OTA found partially developed and researched. The most countermeasures appears to be inadequate. There study results discussed previously. Examiner, sub­ no field studies on the validity of polygraph test­ commonly accepted theory at present is that, is no known Federal research planned on poly­ ject, and setting characteristics are considered, ing for preemployment screening or periodic when the person being examined fears detection, graph theory. And the Army's current lO-year with special attention to the use of physical, drug, screening. Overall, the studies varied in impor­ that fear produces a measurable physiological re­ research program to develop a new, perhaps com­ and mental countermeasures that may be em­ tant ways with respect to, in particular, the action when the person responds deceptively. puterized, state-of-the-art polygraph instrument ployed by individuals to attempt to beat the poly­ criteria used to verify truth, and whether original Thus, in this theory, the polygraph instrument is should be reevaluated to determine if research graph. This chapter also presents some possible examiners' decisions or blind evaluation of charts measuring the fear of detection rather than decep­ priorities and direction need adjustment. Final­ priorities for further research on factors affecting were used as the basis of comparison with ground tion per se. And the examiner infers deception ly, the planned FBI-Secret Service polygraph va­ polygraph validity. truth. In addition, all studies had substantial prob­ when the physiological response to questions lidity study needs an extensive independent scien­ lems of research design, especially with case and about the crime or unauthorized activity is greater tific review. Chapter 7 highlights the major conclusions and examiner selection. As a result, the studies may policy implications of the scientific analysis. Ap­ represent a highly select sample of cases. These pendix A includes illustrative informed consent caveats limit the confidence that can be placed in forms use in Federal Government polygraph ex­ any conclusions about polygraph validity based aminations. Appendix B presents the results of CHAPTER·BY·CHAPTER OVERVIEW on field research. OTA's survey of Federal Government polygraph The preceding discussion summarizes OTA' s iological responses in order to detect deception Chapter 5 parallels chapter 4 and presents use and practice. Appendix C includes the coding major findings. This section provides a brief chap­ or establish truthfulness. This chapter discusses OTA's analysis of polygraph analog studies in form for OTA's analysis of field and analog ter-by-chapter overview of the technical memo­ the procedures and their common applications, which field methods of polygraph examinations studies. Appendix D provides a list of acronyms randum. and explains why different polygraph testing tech­ are used in simulated rather than real-life situa- and glossary of key terms. niques appear to be required depending on in­ Chapter 2 describes the varieties of polygraph tended uses. questioning techniques and a number of uses for polygraph examinations, with an emphasis on The validity of polygraph examinations to de­ Federal Government use. The chapter describes tect deception has long been a controversial issue CONCLUSIONS the polygraph instrument as relatively standard, Chapter 3 describes how the courts, State A major reason why scientific debate over poly­ application to another. The accuracy obtained in and, by itself, unable to detect truthfulness or de­ legislatures, and the executive and legislative graph validity yields conflicting conclusions is that one situation or research study may not generalize ception. What is often referred to as "the poly­ branches of the Federal Government have viewed the validity of such a complex procedure is very to different situations or to different types of per­ graph" is actually a set of relatively complex pro­ assessments of scientific validity as central to deci­ difficult to assess and may vary widely from one sons being tested. Scientifically acceptable re- cedures for asking questions and measuring phys- sions about polygraph use. Despite many decades 8 search on polygraph testing is hard to design and OTA concluded that, while there is some evidence conduct. for the validity of polygraph testing as an adjunct to criminal investigations, there is very little re­ Advocates of polygraph testing argue that thou­ search or scientific evidence to establish polygraph sands of polygraphs have been conducted which o test validity in screening situations, whether they substantiate its usefulness in criminal or screen­ be preemployment, preclearance, aPeriodic or ing situations. Claims of usefulness, however, are aperiodic, random, or "dragnet." Substantial re­ often dependent on information (e.g., confessions search beyond what is currently available or and admissions) obtained before or after the ac­ planned would have to be conducted in order to tual test, and on its perceived value as a deterrent. fully assess the scientific validity of the NSDD- o Chapter 2 The focus of the OTA technical memorandum 84, DOD, and administration polygraph pro­ is not whether the polygraph test has been useful, posals. but whether there is a scientific basis for its use. Varieties of Polygraph Testing i ·andlJses ;,>

o

j Q

I' ~ I l (J I' I* t I I I 0 I L

I I

f I i _,.l Chapter 2 Varieties of Polygraph Testing and Uses

INTRODUCTION Polygraph examinations have been likened to polygraph is being used. The test of a subject who psychological testing (cE. 89, 92, 101). As such, is suspected of a specific criminal activity typically polygraph testing is best described not in the involves application of a different polygraph tech­ singular but, instead, as a series of tests. These nique than the examination of a prospective Gov­ tests are designed to assess truthfulness and decep­ ernment employee. Some variation in technique tion in situations that range from screening job is also related to examiners' training, but such dif­ applicants to investigations of specific criminal ferences probably affect the way in which a tech­ incidents. Polygraph examiners, employed both nique is employed rather than which technique within and outside Government agencies, use a is used. A description of the instrument used in variety of polygraph testing techniques, each of polygraph testing and an analysis of the types of which has a somewhat different underlying logic test situations and polygraph techniques are pre­ and demonstrated validity" sented below. The choice of polygraph technique depends primarily on the circumstances under which the

POLYGRAPH INSTRUMENT Although there are numerous variations in test­ a moving chart. Usually, three different types of ing procedures, the polygraph instrument itself physiological responses are measured. The rate is fairly standard. The polygraph measures sev­ and depth of respiration is measured by pneumo­ eral, usually three, physiological indicators of graphs strapped around the chest and the abdo­ arousal. Changes in physiological arousal exhib­ men. A blood pressure cuff (sphygmomanometer) ited in response to a set of questions are taken to placed around the bicep is used to measure car­ indicate deception or truthfulness. The polygraph diovascular activity. In modern polygraph instru­ instrument, it should be noted, is not a "lie de­ ments, sphygmomanometer readings are electron­ tector" per sei i.e., it does not indicate directly ically enhanced so as to permit lower pressure in whether a subject is being deceptive or truthful. the cuff. The electrodermal response (EDR), a There is no known physiological response that is measure of perspiration, requires electrodes at­ unique to deception (108,122,123). Instead, a pol­ tached to the fingertips. This has also been re­ ygraph examiner obtains a subject's responses to ferred to as galvanic skin response (GSR) or skin a carefully structured set of questions, and based conductance response (SCR). Each of these physi­ on the pattern of arousal responses, infers the sub­ ological assessments has been shoVv'TI to be related ject's veracity. This assessment has been called the to physiological arousal (36). There is some lit­ "diagnosis" of truthfulness or deception (139). erature to suggest that one or more of the physi­ ological channels (EDR, in particular) is most sen­ In actual field testing, subjects' physiological sitive (e.g., 123). Actual field testing, however, responses are measured by a three- or four-chan­ almost always involves measurement of all three nel polygraph machine that records responses on types of responses.

11 12 13

right to obtain a copy of the test results, that the sizes gathering biographical data (77,139). Ques­ TYPES OF TESTING PROCEDURES subject will not be asked questions concerning tions deal with matters such as the subject's sus­ A polygraph examination normally takes any­ The pretest also allows the examiner to assess such areas as political activities, union affiliations, picions about who committed the crime and the where from 1 to 3 hours, although shorter or the effect of special conditions or circumstances racial or religious beliefs, or sexual activities unless subject's feelings about the test. Questions are in­ longer tests may result in a variety of circum­ which might affect physiological responsiveness. these areas are specifically related to the issue tended to provoke so-called "behavioral symp­ stances. The length of an examination depends on Thus, for example, subjects are typically queried under investigation (37). toms" (139) that are believed to be indicators of deception. These symptoms include evasiveness the purpose of the examination, as well as the sub­ about medical problems and use of drugs that Examples of consent forms used in criminal in·· in answering, or complaints that one's physical ject's attitude and a number of other factors. Ex­ could influence autonomic responding. Such as­ vestigations by Federal agencies are shown in ap­ disabilities will invalidate the recordings. When aminations may be very short because a subject sessments are usually made without collecting pendix A. The contents of Federal consent forms an examiner who uses the Reid method later "confesses" or may be lengthy when an examiner "hard" data, such as blood samples. vary somewhat by agency, although all require makes an assessment of truthfulness, this infor­ seeks to resolve an inconsistent or inconclusive that the subject "voluntarily" consent to the ex­ mation is used to supplement the data gathered pattern of responses. The examination can be di­ amination. Some agencies (e.g., Department of from the physiological measures. vided into at least three components: pretest in­ Consent Procedures the Treasury (186)) indicate that the subject has terview; question procedure; and post-test inter­ Depending on which polygraph method is em­ the right to stop the examination at any time. Al­ Whatever the format of the pretest interview, view. The actual questioning aspect of the exam­ ployed, as well as the subject's attitude and the though the National Security Agency (NSA) re­ if control questions are to be used in the test, the ination, which may be repeated three or four situation under investigation, pretest interviews ports that the full cooperation of the subject "is last part of the interview will be used to design times, lasts no longer than a few minutes for each may take from 20 to 90 minutes (20,27). One as­ essential or the results will be inconclusive," NSA such questions and review them with the subject. question set (limited, in some cases, because the pect of the pretest interview involves obtaining also reports (see app. B) that the polygraph exam­ In this phase, biographical and behavioral infor­ blood pressure cuff can be inflated for only 10 to the subject's consent to be examined. Consent pro­ ination is part of the Agency's security process­ mation collected earlier becomes essential. The in­ 12 minutes without causing the subject undue dis­ cedures vary depending on the nature of the in­ ing, and that failure to complete processing (which formation permits the examiner to tailor control comfort). Each aspect of a polygraph test is terview, most importantly between criminal or includes polygraph testing) may result in failure questions to the individual subject. The process described below in detail. Unless specifically preemployment polygraph tests. According to to be accepted for employment. As discussed more of designing control questions is complex and is noted, generally used polygraph procedures are Barland and Raskin (20), a typical polygraph ex­ fully below (see Current Federal Government discussed further in the section below which de­ described. Federal Government procedures are amination conducted as part of a criminal inves­ Use), NSA conducts polygraph examinations pri­ scribes the control question technique (CQT). often different and, where important such dif­ tigation requires that the examiner advise the ex­ marily in the context of preemployment and peri­ ferences are noted. aminee of his or her Miranda rights (or rights odic security screening; most other agencies con­ Testing Procedure under the Uniform Code of Military Justice). The duct polygraph examinations as part of specific­ Actual testing procedures have been described The Pretest Interview subject is also told that the polygraph examina­ incident criminal investigations. in detail by Barland and Raskin (20) and Reid The pretest interview has been considered an tion is voluntary. Subjects should also be in­ The remainder of the pretest interview also and Inbau (139). Polygraph measuring devices, formed whether or not the examination will be indispensable component of the polygraph exam­ varies. In the method taught to Federal exam­ including pneumographs, a sphygmomanometer, observed from outside the room or recorded. ination (121,139,194). The importance of the iners at the U.S. Army Military Police School and electrodes, are placed on the subject either These disclosures are usually included in a writ­ pretest is not only in its role to provide subjects (USAMPS), * the interview focuses on questions during the pretest interview or at its conclusion. ten form which the subject is asked to sign. Ac­ with information about the examination and to about the subject's background: employment, After the end of the pretest interview, the sphyg­ cording to Reid and Inbau (139), criminal suspects inform them of their legal rights, but also in its family, education, health, and any previous legal momanometer is inflated, and the recording of may already have been informed of their Miran­ ability to generate the psychological climate con­ problems (20). The examiner aims to learn enough responses begins. A short period, of about 10 to da rights and been asked to sign a consent form sidered necessary for a valid polygraph test. An to assess the subject's readiness for the examina­ 15 seconds, is used to observe initial respiratory before coming to the examination room. important purpose of the interview is to persuade tion and to prepare anxiety-provoking control cycles (baseline) and to allow any initial response a subject that the examination is professionally Applicants for employment need not be advised questions, if they are to be used. The polygraph to fade; then, the examiner asks the first question. conducted and that any deception attempted "will of their right to speak with an attorney but may, examiner then explains the polygraph technique Between each question, the examiner waits about be very obvious to the examiner" (20). Such in­ depending on local laws, need to be advised about to the subject and queries the subject in detail 15 to 20 seconds until the response to the last ques­ structions, it is thought, place truthful subjects at the voluntarism of the examination. In the case about the incident being investigated. tion is finished and physiological response is closer to baseline. The examiner notes on the chart when ease and increase anxiety in subjects who intend of such employment-related tests, along with a Another form of the pretest interview advo­ the exam begins, when question!:. ::ire asked, and to be deceptive. Persuading subjects about the ef­ provision concerning voluntary consent, subjects cated by Reid (founder of the Reid College of Lie when it ends. Extraneous behavior that affects the fectiveness of the examination should sharpen dif­ will be told how the results of the examination Detection) in criminal investigations makes use recordings may also be noted. When questions for ferences between deceptive and nondeceptive sub­ will be used. Thus, for example, they may be told of a structured series of questions and deempha- jects in their reactions to questions about a par­ that a copy of the test results will be provided to the first chart end, the examiner deflates the cuff. ticular incident. the sponsor of the exam, that the subject has a *The USAMPS provides polygraph examiner training for almost The examiner then inspects the chart and asks all Federal Government polygraph examiners, with the exception the subject about his or her reaction to the ques­ of CIA and NSA examiners. tions. The usual purpose for obtaining subjects' .~- 1 I 1

14 15

reactions is to allow refinements in the questions. recordings) to further convince subjects of the in­ In the investigation of a theft, for example, a rele­ of detection (or another hypothetically lying re­ The questions are reviewed again, and, when nec­ strument's efficacy. vant question might be "Did you steal that lated reaction (19)) for a subject to become physi­ essary, further clarified. The examiner may then money?" or even more specifically, "Did you take ologically aroused during the presentation of rele­ administer a stimulation test, designed to improve Types of Questions $750 from Jones' office?" Relevant questions can vant questions (48,108,136,194). Even with the test validity. The examiner will then continue to be broader, however. In preemployment screen­ addition of nonrelevant comparison items, it is test and obtain two or three more charts in the The central element of any polygraph examina­ ing and periodic or aperiodic screening, the area necessary to run several polygraph charts using same way. The examiner may use other stimula­ tion is the test of subjects' responses to a set of of interest may be the subjects' entire background. the same questions (though, perhaps in different tion tests between charts, and different question­ questions or items within questions. How these Thus, there may be a series of relevant questions, order) to be sure that reactions are consistent. If ing techniques (see below) to record different questions are structured represents the principal such as "Have you ever been fired from a job?" several charts are not run, a subject's responses charts. Different questioning techniques may then difference among polygraph techniques. There are or, "Have you stolen more than $50 in moneys could be attributed to surprise, physical move­ be used based on information revealed by the sub­ four different kinds of questions or items used in in anyone year from any of your employers?" ment, or some reasons for concern other than a ject. In most techniques, any new questions would polygraph testing, different combinations of ques­ (115). Intelligence agencies may ask broad ques­ lying-related cause (203). On the other hand, the be discussed with the subject before being asked. tions (generally referred to as question tech­ tions concerning unauthorized contact with for­ administration of several charts could theoretical­ The procedure in preemployment screening or in niques), and different applications for the various eign intelligence agents or involvement in com­ ly just repeat the initial situation leading to the other personnel screening tests may differ. techniques. Distinctions among questions and munist activities. Questions in an intelligence physiological response if the cause were not a ran­ techniques are important. Only one type of ques­ screening may also deal with areas which, poten­ dom one (e.g., presence at the scene, knowledge Stimulation Tests tion technique in one application (CQT in crimi­ tially, may make an applicant susceptible to of the incident, concern over being falsely iden­ nal investigations) has been extensively researched blackmail. It is important to note, however, that tified). Thus, the essence of polygraph testing is Polygraph examiners typically conduct what is (see chs. 4 and 5); and there are significant dif­ when several relevant questions relating to dif­ the comparison of responses to the relevant ques­ known as a "stimulation"or "stim" test, designed ferences between CQT and other techniques. The ferent issues are used, subjects are not expected tions with responses to nonrelevant questions, to further convince subjects of the accuracy of the range of questions, techniques, and applications which have been labeled control questions and polygraph examination. Although not actually a to exhibit physiological responses to all of them; is described more fully below. irrelevant questions. part of the pretest, stimulation tests can be given the relevant questions that do not evoke responses are used, after the fact, as a type of control either before the first actual set of test questions Questions question. or after the first chart has been recorded. Stimula­ Control Questions tion tests are intended to reassure truthful sub­ The kinds of questions that are used for poly­ To summarize, relevant questions are questions jects and provoke anxiety in deceptive subjects graph testing have been labeled relevant ques­ about the topic under investigation, but topics can Control questions, then, are used for purposes of comparison. Essentially, truthful subjects are (cf. 15). Their effect should be to increase differen­ tions, control questions, irrelevant questions, and be very specific (Did you take $750 from Jones' believed by polygraph examiners to be more con­ tial responsivity of deceptive and nondeceptive concealed information or guilty knowledge ques­ office?) or cover a long period of time and a varie­ cerned (and, thus, more physiologically aroused) subjects to different questions on the examination. tions. Basically, relevant questions are questions ty of acts (Have you ever stolen money from an about control than relevant questions. The re­ Some research suggests stimulation tests increase about the topic under investigation (a theft, drug employer? Have you ever had unauthorized con­ sponses to both control and relevant questions are the validity of polygraph examinations (35,149). use, contact with foreign agents). Suspects' re­ tact with a foreign agent?). It is not clear what compared. However, control questions, like rele­ sponses to relevant questions are of greatest effect, if any, the breadth of a relevant question The most common "stim" test is a "number" vant questions, vary in breadth and type. One interest to polygraph examiners. has on polygraph results, nor has there been any or "card" test. A subject is instructed to select, type of control question concerns what is hypoth­ research done on this issue. As is discussed fur­ from a deck, a card that has a number, word, or Control and irrelevant questions can be esized to be the same kind of issue that is under ther in chapters 4 and 5, the preponderance of suit on the back, or to write a number within a grouped together as questions used for purposes investigation at the time of examination. For ex­ certain range (50,57). Sometimes, the cards are of comparison to relevant questions. It is impor­ research evidence concerns the use of relevant questions to evoke reactions to specific acts. ample, a control question for "Did you take the secretly marked or otherwise arranged so that the tant to note, however, that the name one gives $750 from Jones' office?" might be "Other than examiner is sure to know the correct answer (139). to a question may depend on the specific context what you have told me [during the pretest inter­ Comparison Questions Many polygraph examiners claim this is unnec­ in which it is used. Thus, one cannot easily give view], have you ever stolen anything in your life?" essary, however, because the technique is accurate an example of a relevant question or a control In contrast to relevant questions, which con­ In an investigation of unauthorized disclosure of enough without use of such deception (cf. 123), question because in different situations and at dif­ cern issues of direct interest to the examiner, con­ classified information, a control question might and secret markings are not employed by Federal ferent times during an examination relevant ques­ trol and irrelevant questions are used for purposes be "Have you ever betrayed anyone who trusted agencies. The examiner then may repeat a range tions may be used as control questions. Likewise, of comparison. As noted above, there is no you?" Subjects innocent of the crime under in­ of suits, numbers or a set of words, asking the irrelevant questions may become relevant, de­ known physiological response unique to lying. vestigation are presumed to be more concerned subject if each is the concealed item. The suit, pending on a subject's response (201). Thus, a polygraph examination could not consist about having ever done anything of this sort (and, number, or word that is actually the concealed merely of relevant questions. If only relevant thus, being the "kind of person" who might have item is supposed to provoke the greatest physio­ Relevant Questions items were used, an examiner would not be able committed the crime under investigation). It is logical response. Often, the examiner will show Functionally, relevant questions are questions to establish the actual reason for the response. theorized that although guilty subjects will also the subject the polygram (i.e., the actual chart directly related to the focus of an investigation. There are a number of reasons, other than fear be concerned about control questions, they will ~------

17 16 be more concerned about and thus exhibit more Control questions, then, are questions for tails, but that innocent subjects will respond the control question, and is claimed to be a version physiological reactions to relevant questions. which the responses are designed to be compared same to all the items. Different types of concealed of the control question technique. The versions to responses to relevant questions. some screen­ information tests are described below (see Con­ discussed in this section are: There are a number of views, however, about In ing examinations, relevant questions may func­ cealed InFormation Tests). what distinguishes a control question from a rele­ 1. the traditional R/I; tion as control questions after the fact. That is, vant question. One distinction among control 2. the Federal version of the R/I; and if a relevant question produces a relatively mild questions is whether a question is inclusive or ex­ Summary 3. the R/I as used in typical preemployment physiological response, it may be compared to clusive. Inclusive control questions are questions screening tests. other relevant questions that produce greater re­ For any technique, deception is detected by which include the specific incident under investi­ sponse. Most often, control questions are designed comparison of suspects' physiological responses In a traditional RII examination, the two types gation. An example of an inclusive control ques­ to be arousing for innocent subjects (i.e., those on critical or "relevant" questions or items with of questions used are relevant and irrelevant ques­ tion in an investigation of an internal theft would their responses on noncritical (irrelevant or con­ tions. Deceptive subjects are assumed to have a be "Have you ever stolen money from an employ­ who are not being deceptive on the relevant ques­ tions), relative at least to relevant questions. This trol) items. Greater physiological responses to significantly greater reaction to the relevant ques­ er?" Exclusive control questions, on the other is usually the central point of control questions, relevant items than to noncritical (control, irrele­ tions than to the irrelevant questions. An under­ hand, cover a period of time not including the in­ and is central to the control question technique vant) items are assumed to be indicative of de­ lying assumption of this technique is that non­ cident under investigation. An example is, "Before (CQT) discussed below. ception. deceptive subjects should have an equal response age 18, did you ever take anything of value?" to all questions, because, being nondeceptive, they There is some controversy over how far back in Irrelevant Questions would not fear questions about the crime any time an exclusive control question must be set for Polygraph Question Techniques more than irrelevant questions. the subject to consider it psychologically separate Another type of question used, in part, for pur­ from the incident under investigation and, thus, poses of comparison to responses to relevant ques­ Three types of question techniques combining There are numerous well-recognized problems not a relevant question. Because inconclusive con­ tions is the so-called irrelevant question. Examples the four question types are described below: the with the traditional R/I technique, at least from trol questions may, from the suspect's perspec­ of irrelevant questions commonly used in inves­ relevant/irrelevant (R/I) technique, the control the perspective of psychologists who have eval­ tive, include the act under investigation, some tigations are; "Are you called [subject's name]?" question technique (CQT),and concealed informa­ uated polygraph test validity (cE. 108,126,136). polygraphers contend that they are really relevant or "Is today Tuesday?" Irrelevant questions are tion techniques. Each of these test types tends to First, the intent of the relevant and irrelevant ques­ questions; i.e., they cannot be used for purposes questions which are believed to have no, or very be used for particular purposes; e.g., the R/I tech­ tions is transparent, which means that the rele­ of comparison. The Federal Government, for ex­ little, emotional impact on a subject. Thus, such nique is used in the great majority of preemploy­ vant questions are likely to be more arousing for ample, typically uses exclusive control questions questions can be used as an indicator of a partic­ ment screening interviews, while CQT is normally the truthful as well as the deceptive subjects. Sec­ because it views inclusive controls as relevant ular subject's normal baseline level of arousal; no used in criminal investigations. There have been ond, questions in the RII technique are not usually questions. Examiners from the private polygraph universal standard of physiological arousal can adaptations of these techniques for other uses, reviewed with the subjects before the test. A larger firm of John E. Reid & Associates use both inclu­ be applied because individuals differ markedly. some of which are discussed below. Also, exam­ response to the relevant question may, thus, be sive and exclusive control questions. Irrelevant questions are hypothesized to serve pur­ iners may combine different techniques in an in­ due to surprise or misunderstanding, as well as poses other than providing a physiological base­ vestigation (see, e.g., 139). In general,. RII has the deception. Third, as with any question tedmique, Other kinds of nonrelevant questions other than line (139). Perhaps most important, irrelevant broadest potential use while the concl"!aled infor­ reactions may be flattened by drugs or by the gen­ those that cover the same kind of incident as the mation techniques are the least applicable. Within erally reduced responsivity of certain subjects one under investigation, or which cover it in a questions interspersed among relevant questions are hypothesized to provide a type of rest period each category, particularly CQT, there is consid­ (136). These effects are probably more difficult different way, are also considered to be control for the subject. erable variability and several versions of each to detect with RII than with other question tech­ questions. Thus, for example, "Have you ever fan­ technique are employed. niques. tasized about giving a document to a foreign Concealed Information Questions agent?" is a type of control question used in some Because of these problems, the confidence one intelligence investigations. In some screening ex­ Questions about concealed information are the Relevant/Irrelevant (R/I) Techniques can place in the R/I technique is limited (136). As a consequence, the RII technique is typically not aminations, in which contact with a foreign agent fourth type of question used in polygraph testing. The R/I technique was the first standard meth­ is of primary concern (i.e., constitutes the rele­ Unlike control and relevant questions, which ask used in the case of specific incident examinations od of polygraph questioning. The method was de­ by either public or private examiners. It is used vant question), "Have you ever done anything for subjects whether they have committed a crime, veloped by Marston (114), a psychologist and the which you are now ashamed?" could be a con­ concealed information items aim to detect infor­ almost exclusively with employees in nonspecific original proponent of polygraph examinations. investigations. The Federal Government occasion­ trol question. When a different issue than suscep­ mation about a crime that only a guilty subject An adaptation of this traditional technique is used tibility to blackmail is under investigation, "Have would have. Such information might include de­ ally uses the traditional RII and also a version in most of the preemployment screening con­ of the R/I which is claimed to function as a con­ you ever done anything for which you could be tails about the site of the crime or the means of ducted in the United States. blackmailed?" can be used as a control question. committing it, such as the type of murder weap­ trol question test. The Federal Government ver­ It is noteworthy that in a different context, such on used. It is hypothesized that guilty subjects will However, the R/I technique as used by the Fed­ sion of the technique is called the general ques­ as a broader screening examination, these would exhibit a different physiological response to the eral Government involves somewhat different tion test (GQT). Like the Reid CQT (discussed be considered relevant questions. correct (relevant) detail than to the incorrect de- types of questions than the traditional R/I, and below), it uses inclusive control questions, which it must be explained separately. As currently used pertain to the subject's entire life, such that a com­ by Federal examiners, the R/I relies on a type of plete answer would also include the specific inci- 18 19 dent being investigated. Thus, with a question Have you sold marihuana (sic) or other narcotics ille­ In the RII questioning technique, a diagnosis are relevant questions and apparent deception like, "Did you ever steal anything from a place gally in the past ( ) years? of truthfulness or deception indicated is made by (arousal) in response t.Q any of the items is ex­ Have you ever stolen more than ($ ) worth of mer­ where you worked?" the theft being investigated comparison of responses to each relevant ques­ plored. Examples of the kinds of questions used chandise in anyone year from any of your employ­ would in actuality be part of the answer. T ech­ tion with the responses to the irrelevant questions in aperiodic screening in a supermarket (204), nically these are seen as "relevant" questions, ers? Have you even stolen more than ($ ) in moneys in any and the remaining set of relevant questions (or include: because they are pertinent to the incident in ques­ in the Reid, and Army examples, control ques­ one year from any of your employers? Are you relatively satisfied with this job now? tion. Yet they are claimed to function as control Have you ever used a system to cheat one of your em­ tions). Presumably, an applicant will be decep­ questions,. because they are intended to provoke Do you, as far as you know at this time, intend to stay ployers? tive on no more than a few questions. These ques­ with this employer? a greater response in innocent subjects than ques­ Have you ever had your driver's license suspended or tions will provoke a greater physiological response Have you ever intentionally underpriced or under­ tions about the misdeed provoke. revoked? than the others, leading to further inquiries and weighed merchandise? Have you ever had any traffic citations in the past five An adaptation of the R/I technique is the prin­ an eventual diagnosis (56,204). Is there a particular person at your store that is respon­ (5) years? cipal method of questioning used in preemploy­ sible for damaging merchandise due to real careless­ Are you seeking a job with this company for any Other types of questions are used in some ness, not caring or intentionally? ment and periodic or aperiodic personnel screen­ reason other than legitimate employment? screening examinations, such as questions about ing. Unlike the questions used with other tech­ Have you deliberately lied to any of these questions? sexual practices or gambling. Such questions seek The relevant topic areas covered by NSA in a niques, R/I questions need not focus on one spe­ information about an applicant's character rather periodic screening are discussed later. Because of cific wrongdoing (20,108). The examiner can, The method used by John E. Reid & Associates its use of control questions, the Federal version employs four standard relevant questions: than his or her job performance and are consid­ thus, use the method to assess any number of ered by some to be unduly invasive (173). In re­ of RII is discussed in the next section. issues for which the subject's veracity is to be In the last five years did you steal any merchandise sponse to this practice, ethical standards have evaluated. from previous employers? been developed for use of the polygraph in pre­ In the last five years did you steal any money from Control Question Technique (CQT) In polygraph examinations used to screen em­ employment screening (d. 154), and some States previous employers? (e.g., Illinois) prohibit their use. Preemployment The CQT is the most common technique used ployees, the polygraph examiner usually presents In the last ten years did you take part in or commit polygraph examinations fall under the guidelines in investigations of a specific issue. The CQT was a series of relevant questions, with several irrele­ any serious crime? vant questions interspersed to provide a baseline. Did you falsify any information on your application? for emloyment interviewing of title VII of the developed to deal with some of the inherent prob­ Most relevant questions ask about past behavior Equal Employment Opportunity Commission, lems in the traditional R/I technique (139). Like

~ I and so examiners are obliged to conduct the ex­ the R/I that might disqualify the subject from a job (e.g., These standard questions may be modified de­ technique, it asks relevant questions about aminations in a way that would not discriminate employee theft, drug use, fighting on the job, in­ pending on admissions made during the pretest the crime like "Did you steal the $750 from Jones' on the basis of sex, race, etc. (d. 154). One cen­ curring a large debt). Some examinations may in­ (e.g., a revision may be, "In the last five years office?" As with R/l, the deceptive subject is I tral principle of ethical standards is that relevant clude questions about a potential emploYi:e's did you steal any merchandise other than minor assumed to produce a greater autonomic response questions be related to the job applied for. background or intentions regarding a job, for ex­ office supplies?"). In addition to the standard to the relevant than to other questions. But CQT Whether questions meet this criterion depends on ample, "Did you actually graduate from college?" questions a fifth relevant question (e.g., concern­ also adds control questions, which, as discussed the job; e.g., information about one's driving (201) or "Are you seeking a job with this com­ ing the illegal purchase or sale of merchandise; briefly above, are designed to provoke a greater record would be important in hiring a delivery response in subjects who are innocent and truthful pany for any reason other than legitimate employ­ use of narcotics) may be added depending on the person, but not in hiring a bank teller. Screening about the crime being investigated. ment?" (115). Listed below is an example of ques­ nature of the job. applicants for positions involving national securi­ tions from a preemployment screening protocol As discussed above, control questions are de­ used by a commercial firm (115; also see 56,204). ty apparently require questions about sexual be­ The Reid firm also uses what it regards as con­ havior, drug use, and mental health as well as signed to be arousing for nondeceptive subjects. Relevant _questions: trol questions in preemployment interviews. Con­ areas more directly related to national security The questions are designed to cause innocent sub­ jects to be doubtful Did you tell the complete truth on your job applica­ trol questions include, "Did you ever steal any­ (e.g., involvement in espionage). The range of and concerned about whether tion? thing in your life?" and "Did you lie to any of topic areas covered in national security pre­ they have actually told the truth or told a lie. Have you deliberately withheld information from your the questions you answered during the applica­ employment screening examinations by NSA is These questions usually probe for past misdeeds job application? tion process for this job?" It is not clear, however, of the same general nature as the crime being in­ Have you ever been fired from a job? discussed below under Current Federal Govern­ how the Reid preemployment control questions ment Use. vestigated but they are transgressions that poly­ Are you seeking a permanent position with this differ from the relevant questions. It seems rea­ graphers suspect most people have"committed" company? sonable to suppose that both truthful and non­ In so-called periodic or aperiodic checking for or considered committing in some form (139). An Since the age of ( ) have you committed an undetected truthful subjects (in terms of the relevant ques­ internal security purposes, employees are asked example of a control question might be, "Before crime? to submit to occasional polygraph examinations. Since the age of ( ) have you been convicted of a crime? tions) may be just as concerned with the subject the age of 25, did you ever steal anything from During the past year, have you used marihuana (sic) matter of the control questions as they are with These examinations can assess drug use: subjects' a place you worked?" Control questions are de­ more than ( ) per ( )? the relevant questions. It is also not clear why own or others' employee theft, and other matters signed to cover a long period of time, which may Have you used any other narcotic illegally in the past employers would be less concerned with the con­ including job satisfaction and commitment. In this make the subject even more doubtful about the ( ) years? trol than with the relevant questions. type of examination, almost all of the questions veracity of answers provided. 20 21

kind of person that could have committed the ical measure, the examiner derives a "plus" (truth­ Considerable attention in the pretest interview In reality, however, the examiner wants the su~­ crime referred to in the relevant questions. In is devoted to development of control questions ject to experience considerable dou.bt ab?ut hIS ful) score if the subject responds more to the con­ addition to relevant and control questions, irrele­ (139). The process of developing control ques­ or her truthfulness or even to be mtentIOnally trol question and a "minus" (deceptive) score if vant questions are included during the actual in­ tions, reviewing them with the subject, and then deceptive. According to Raskin (91), "Control the subject responds more to the relevant ques­ terview in order to provide a baseline of physio­ refining them is designed to develop the most ap­ questions are intentionally vague and extremely tion. A positive score above a certain criterion logical responsiveness. propriate questions, and to convince subjects to difficult to answer truthfully with an unqualified level is diagnosed as truthful, a minus score below view control questions as seriously as relevant 'No'." Several versions of CQT are regularly em­ a certain level is diagnosed as deceptive, and questions. In addition, the pretest review is de­ ployed and adaptations depend both on the train­ scores in between are considered inconclusive. , To produce the final version of ~ control q~es­ signed to get subjects either to be deceptive to con­ tion, the examiner begins by askmg the subject ing of the examiners and the testing situation. The A version of ZOC is used by Federal polygraph trol questions or at least to be concerned about Reid version can include relevant questions about a broad version of the question used in the pretest examiners. The Federal version differs from the the accuracy of their recollections (20,37,91,139). several aspects of the crime (139). For example, interview. Thus, for example, the question might Backster ZOC in that it may ask about several It is considered crucial to produce in the subject one chart could include questions about break­ be structured, "Did you ever steal anything in aspects of the crime in one chart. Relevant ques­ the right psychological set in relation to the con­ ing into an office, stealing a check, and then your life?" At this point, different polygraph ex­ tions are asked about primary involvement (e.g., tIol questions. This set is then thought to lead sub­ cashing it. Examiners who use Reid's CQT make aminers use slightly different methods to handle "Did you steal ?"), secondary involve­ jects to be more concerned about control ques­ a global comparison between the responses to the the discussion of past wrongdoing in response to ment (e.g., "Did you help steal ?"), tions than relevant questions, and so to respond relevant and the responses to the control ques­ the control questions asked during the pretest in­ and so called evidence connecting (e.g., "Do you more to them physiologically. This difference be­ tions. They also note the subject's behavior terview. In the USAMPS method (91), if the sub­ know where any of that money is now?"). In ad­ tween response to control and relevant questions throughout the interview (as discussed above, the ject confesses to a small transgression in the past dition to relevant, control, and irrelevant ques­ is then the basis for the diagnosis of deceptive or Reid technique includes a series of questions in (e.g., taking home pencils from work), the exam­ tions, the Government ZOC test contains a ver­ nondeceptive. Since the subject's psychological set the pretest interview designed to provoke certain iner will dismiss it as of no consequence. For other sion of the peak of tension test (see below), and is so crucial when control questions are used, dif­ "behavioral symptoms" in deceptive subjects). misdeeds, the examiner will rephrase the control "symptomatic" questions of two types. One type ferential responding to relevant or control ques­ The examiner uses the global comparison of poly­ questions to rule them out (e.g., "Other than w~at of symptomatic question (e.g., "Do you under­ tions (and ultimately, the validity of CQT), de­ graph responses supplemented by information we have discussed, did you ever steal anythmg stand that I'm not going to ask any trick or sur­ pends on the nature of the interaction between about the behavior of the subject to make a judg­ in your life?"). The examiner will actively in­ prise questions?") is designed to test whether the examiner and subject. This is true regardless of ment of the subject's veracity. An example of a tervene to prevent subjects from unburdening too examinee trusts the examiner's word that no sur­ the act in question, the particular CQT method Reid control question sequence, excluding the much of their anxiety over their past wrongs with prise questions will be asked. A large response is used, or the method of making assessments of pretest behavior provoking items, follows (139): the intention of keeping the~l concerned during symptomatic of distrust. A second type of symp­ truthfulness or deception. Even the validity of an the actual polygraph testing. Any troublesome 1. Do they call you "Red?" (where the pretest inter­ tomatic question (e.g., "Is there something else entirely computerized system of scoring and diag­ past transgressions the subject brings up are ex­ view had disclosed he is generally called "Red.") you are afraid I will ask you a question about, nosis would depend on the nature of the interac­ cluded (by such phrases as "Other than what we 2. Are you over 21 years of age? (or reference is made even though I have told you I would not?") is to tion between examiner and subject. In this sense, have discussed, . . . ?") so the subject is always to some other age unquestionably but reasonably, test whether there is some other issue the examinee CQT examinations, as the technology to conduct brought to the point at which he or she answers and not ridiculously, below that of the subject.) is concerned about (e.g., another crime) that may polygraph tests now stands, always require exa:n­ "No" to the control question. The control ques­ 3. Last Saturday night did you shoot John Jones? be absorbing his or her arousal. iners to make important judgments about and m­ 4. Are you in Chicago (or other city) now? tion is then ready to be used in actual testing. I Other versions of CQT or related techniques terventions in their interaction with subjects. i" S. Did you kill John Jones? The Reid method varies from the Federal meth­ I 6. Besides what you told about, did you ever steal are also used by Federal agency examiners. One, The polygraph examiner does not tell the sub­ od in some ways (139). If the subject does not ad­ anything else? the modified general question test (MGQT), re­ ject that there is a distinction between the two mit to a past wrongdoing, the examiner may 7. Did you ever go to school? sembles the Reid CQT with the following differ­ types of questions (control and relevant). Con­ 8. Did you steal John Jones' watch last Saturday ences: 1) only the polygraph charts are used to probe until the subject admits to one, even a cri:ne night? trol questions are described as intending to deter­ as small as stealing pocket change from a relatIve make determinations of truth and deception and 9. Do you know who shot John Jones? global evaluations using inferences about behavior mine if the subject is the "type of person" who during childhood. Such transgressions are then 10. Did you ever steal anything from a place where would commit a crime such as the one being in­ ruled out by adding the kind of exclusionary you worked? are dispensed with; 2) charts are numerically vestigated (136). The examiner stresses that the phrase discussed above (Le., "Other than what scored; 3) control questions exclusively concern subject must be able to answer the questions com­ we have discussed, ... '1"). However, as in the In contrast, Backster's (10) zone of comparison a time and place separate from the time and place pletely with a simple "yes" or "no" a.nswer~ t~at USAMPS method, it is assumed at this point that (ZOC) technique makes a diagnosis of deceptive of the crime under investigation, with the inten­ the polygraph will record any confUSIOn, mIsgIV­ the subject is either concealing other misdeeds or or truthful from a standardized numerical scor­ tion of clearly separating responses related to the ings, or doubts, and that the subject should discuss is worried that there are others he or she has ing of the charts. Each relevant question is paired crime and the control question; and 4) the con­ any troublesome questions with the examiner (~O). overlooked (139). This worry has been heightened with a control question. Scores are derived for tent of control questions is always related to the Thus, the situation is set up such that the subject because of the examiner's emphasis on learning each relevant question by comparing it only with crime under investigation, i.e., control questions is persuaded that the examiner wants the truth. the truth to "ascertain" that the subject is not the the previous control question. On each physiolog- about theft are used to investigate a theft, con------~--~---- f"o~--~--~~--~~------~------~---~ Hi( !'I "7"'t" ---

22 23

trol questions about assault are used to investigate tions will increase up to the critical question, it? Was it (a) in the men's room, (b) on the coat proponents do not adequately assess the conse­ assault, etc. Presumably, when unauthorized dis­ where they will reach a peak, hence the name, rack, (c) in the office, (d) on the windowsill, or (e) quences of false negatives. closures are at issue, control questions would con­ in the locker? and fall back down again. The card and number Concealed information tests have, according to cern some sort of unauthorized disclosures in the stimulation tests discussed above are actually ex­ past. There is a major difference, however, in the use several reviewers (e.g., 108,136), other important amples of POT. Barland and Raskin (20) provide suggested for GKT as compared to the use of the limitations. One problem is that they may not be To summarize, there are a number of control a hypothetical example of a POT in a criminal POT. POT is usually used as a supplement to a widely applicable. Knowledge about an incident question techniques, the most commonly used be­ case: CQT, or as an aid in investigation. GKT, how­ may not differentiate between a guilty and inno­ ing the Reid CQT, MGQT, and ZOe. Despite dif­ 1. Regarding the color of the stolen car, do you know ever, has been proposed as an alternative to con­ cent person where, for instance, a suspect is pres­ ferences among them, they share the same premise it was yellow? trol question techniques (92,107,108). Proponents ent at the scene of a crime but claims that some­ and underlying rationale. Use of each of the con­ 2. Do you know it was black? argue that GKT may reduce the number of false one else is responsible (108,136). Furthermore, trol question procedures relies on subjects' not 3. Do you know it was green? positives, because it focuses on specific details that concealed information tests require investigators knowing when they are being asked the relevant 4. Do you know it was blue? would be salient only to the perpetrator of a crime to gather information that is not always possible and control questions. If they know which ques­ 5. Do you know it was red? (108,110). Also, they claim, the validity of GKT to obtain, or must be disclosed to suspects in other tions are more important for scoring purposes 6. Do you know it was white? can be substantially improved by increasing the parts of the investigation (136). In some cases, they may be able to make anticipatory responses 7. Do you know it was brown? number of questions on the test. Critics claim that publicity about the details of a crime eliminates which could invalidate their charts (see ch. 6). Occasionally, criminal investigators use the it is especially susceptible to false negatives (136), the possibility of a concealed information test, POT technique to discover and develop additional that is, guilty persons not detected, and that GKT since the information is public knowledge (136). information about a case. The examiner asks the Concealed Information Tests suspect about a series of details, but does not Another polygraph questioning technique know which is actually relevant to the crime. The works on an entirely different premise than either detail that provokes an exceptional physiological POST-TEST INTERVIEW CQT or R/I. Instead of detecting deception about response is used as a clue in the investigation. For having committed a crime per se, concealed in­ example, an examiner might use POT to deter­ Interspersed among test questioning and meas­ the USAMPS Backster's ZOC and other methods formation tests aim to detect whether a suspect mine the exact location where stolen goods were urement of physiological responses are a number attempt to rely strictly on polygraph chart inter­ has information about a crime that only a guilty hidden. This kind of examination is called a of opportunities for examiners to discuss the test pretation (11,20). In examinations conducted by subject would have or, in some cases (e.g., the searching peak of tension test (20). The searching with the subject. At each occasion, the examiner the Federal Government, the final official deter­ actual amount of money embezzled) to detect the POT technique has been used, for example, in reviews the questions, and, depending on the re­ mination is made after supervisory review of poly­ information itself. Such information might include cases in which employees are suspected of hav­ sponses, questions subjects about their responses. graph charts. If the subject is judged to be decep­ details about the site of the crime or the means ing stolen money, but there is no evidence about At the end of the examination, the examiner will tive during the examination, the examiner will at­ of committing it (e.g., the type of murder weapon the extent of the theft (108). The examiner asks make an assessment of whether a subject is being tempt to elicit a confession. Usually, this is not used). Raskin (136) has aptly described these as the employee if he has stolen money ranging from deceptive or nondeceptive. In some methods, e.g., done directly but is couched in terms of providing "concealed information tests." Concealed infor­ a small amount to the entire amount taken. The Reid's (139), the assessment is a global one em­ the subject with an opportunity to clarify I explain mation tests take two forms: the peak of tension amount that provokes the largest response is ploying behavioral as well as polygraph data. But the responses and differences obtained. (POT) test and the guilty knowledge test (GKT). assumed to be the amount of the total that the employee stole. POT was developed by Keeler (d. 69) and has long been used in criminal investigations. The The GKT, described initially by Lykken (105, USES OF POLYGRAPH TESTING POT test uses a set of five to nine nearly identi­ 106) works in much the same way as POT. GKT, cal "yes or no" questions asking if the subject however, often includes a larger set of questions, As has been implied in much of the above dis­ graph testing, such as for national security pur­ knows about a particular detail related to a crime. and the questions may be of the multiple-choice cussion, polygraph examinations are used for a poses and for employment screening, is extreme­ The detail may be a type of object used, or the type rather than the "yes or no" type. Also, studies variety of purposes. The goal of all such applica­ ly limited. color of an item. One question actually includes investigating GKT have only used the electroder­ tions of the polygraph is the detection of decep­ the relevant detail, while the others include plausi­ mal response, while POT tests have employed tion or substantiation of truthfulness. The nature Current Use ble but false details of a parallel nature. The que,;­ standard three-channel polygraph recordings. An of the test situation, however, leads to important tions and the sequence in which they are asked example of two questions from a GKT used in a differences in the way a polygraph examination The majority of uses of polygraph testing ap­ are reviewed with the subject in the pretest inter­ laboratory study by Lykken (105) is listed below: is conducted. Unfortunately, the published re­ pear to be on behalf of private employers, the next view. The subject is usually instructed to answer 1. If you are the thief, you will know where the desk search literature deals almost exclusively with the greatest number are in the context of local criminal "no" to each question. The question with the true was located in the office in which the theft occurred. use of the polygraph by police and military ex­ justice investigations, and the remainder are done detail is usually presented in the middle of the Was it (a) on the left, (b) in front, or (c) on the right? aminers for criminal investigations. The research by the Federal Government. Of greatest concern sequence, so that the subject's physiological reac- 2. The thief hid what he had stolen. Where did he hide literature on a number of important uses of poly- for the present analysis are the numbers and types ------~ ------f ' 1IIIt4.' 7¥

\, 25 24

sumably, specific issue examinations would be investigations (see 136). In some cases (most typi­ of examinations currently conducted by agencies Table 1.-Polygraph Examinations Conducted by conducted using the control question technique. cally, rape and kidnapping cases, but also, for of the Federal Government. This section will de­ Federal Agencies, 19828 example, investigations of improper or illegal con­ Current DOD regulations also allow the use of duct by public officials (177)), witnesses and vic­ vote most attention to such uses, although local Agencyb Total government and private use are briefly discussed polygraph examinations to investigate situations tims whose veracity is in doubt are asked to take Department of Defense: in which credible derogatory information about in order to place Federal use in context. Army Criminal Investigation Command ...... 3,731 a polygraph examination. Suspects who claim in­ Army, Intelligence Command 279 an individual with clearance is provided to of­ nocence may be asked by their defense attorneys Current Federal Government Use Navy ...... 1,337 ficials. The frequency of this type of investiga­ or the prosecution to support their claim by tak­ Air Force ...... 3,019 tion, however, was not reported. Prior to the Marines ...... '" 263 ing a polygraph examination. In such cases, pros­ In order to assess the extent of polygraph use National Security Agency ...... 9,672 President's National Security Decision Directive ecutors and defense attorneys may make infor­ among Federal agencies, the Office of Technology Department of Justice: of March 11, 1983, use of the poiygraph in per­ mal agreements to drop the charges if the poly­ Assessment (OTA) conducted a survey of Gov­ Federal Bureau of Investigation ...... 2,463 sonnel investigations of competitive service ap­ graph examination indicates no deception. Or, the ernment use during May 1983. The request for Drug Enforcement Agency ...... 211 plicants and appointees to competitive service prosecution and the defense may formally stipu­ information was sent to the Departments of Department of the Treasury: Secret Service ...... 714 positions was limited to executive agencies with late that if deception is indicated, results of the Defense (DOD), State, Justice, Treasury, the U.S. Bureau of Alcohol Tobacco and Firearms .... . 256 highly sensitive intelligence or counterintelligence polygraph examination will be admissible at trial. Postal Service, and the Central Intelligence Agen­ U.S. Postal Service ...... 652 missions affecting the national security (e.g., a In some cases (New Mexico, Massachusetts, and cy (CIA), all of which were believed to employ Centrdl Intelligence Agency ...... n.a.c mission approachit,.· the sensitivity of that of CIA; the 9th Federal Circuit Court of Appeals (8,136, polygraph examinations. Information was re­ Totals ...... 22,597 see 188). ApprovCi~ . ..J use the polygraph could be 156,157)) polygraph evidence has been admitted quested about the number of examinations, pur­ aData were also reported for fiscal years 1980, 1981, and, In some cases year granted for only I-year periods. Refusal to con­ to date 1983. See app. B for complete raport. over objection. Polygraph evidence is also used poses, and results, as well as about research con­ "The U.S. Customs Service (Department of the Treasury), Department of Health sent to a polygraph could not be made a part of occasionally in postconviction proceedings such and Human Services, and Tennessee Valley Authority conduct a limited but ducted and/or planned (see app. B). At the time cunknown number of polygraph examinations. an applicant or appointee's personnel file. See as sentencing and motions for a new trial (136). of this technical memorandum, all agencies ex­ Classified or partly classified. chapter 3 for a description of proposed changes In polygraph examinations as part of criminal in­ SOURCE: Office of Technology Assessment. cepting CIA had provided written responses to in Federal use of polygraph testing. vestigations, some version of the control question the request for information about the number and technique is typically used. type of polygraph examinations that have been Non-Federal Government Use administered. tions-i.e., tests given to NSA (or affiliated) per­ The use of the polygraph examination by em­ sonnel who already have access to classified in­ Outside the Federal Government, polygraph ex­ ployers is reported to be widespread (144). Al­ CIA declined to respond because of the classi­ formation-would have relevant questions on the aminations are administered as part of criminal though it is illegal to ask employees to take an fied nature of the information. However, some I,. topics of involvement in espionage or sabotage L, investigations, as well as preemployment screen­ examination in 19 States and the District of Co­ data about CIA's use for background investiga­ i ing and periodic screening of employees for pur­ against the United States; knowledge of others in­ 'i lumbia, it is legal to do so in 31 States (8,156,157). tions were reported in a 1980 study (165). The ), volved in espionage or sabotage against the United poses of controlling internal crime and recom­ Polygraph examinations are used most commonly number of polygraph examinations are summa­ ! States; involvement in giving or selling clarsified i mending promotions. Less frequent uses include in commercial banking, investment banking, and rized in table 1. Table 1 indicates that Federal I materials to unauthorized persons; knowledge of \ examinations in such situations as paternity in­ retail operations. In such settings, both risk of agencies reported administering a total of 22,597 others giving or selling classified material to un­ l vestigations and workers' compensation cases. It theft and fraud are high and, in addition, employ­ polygraph examinations in fiscal year 1982. As authorized persons; and unauthorized contact I has been estimated that over a million polygraph ee turnover is high. The use of polygraph exam­ shown in appendix B, about half of these were p examinations are given a year (107), 300,000 of with representatives of a foreign government I, inations is also recommended to employers as a in the context of criminal investigations. Poly­ I (187). Examinations that are given to applicants i them for employment purposes alone (128). check before making promotion decisions (204). graph examinations are also reported to be used for employment and contractors who are apply­ I; for intelligence and counterintelligence investi­ Both private and police polygraphers use poly­ ing for access to Sensitive Compartmented Infor­ gations (some (NSA) at aperiodic intervals), and graph examinations in the process of criminal mation (SCI) consist of questions about the topics preemployment screening. The largest single num­ covered in counterintelligence-type aperiodic ber of polygraph examinations conducted in 1982 screenings (phrased as "Do you plan to com­ were conducted by NSA, a component of DOD, mit ... 7") as well as questions about a broader primarily for preemployment screening. These CONCLUSIONS range of issues: involvement in communist, fas­ numbers can be compared to previous surveys prospective employee. Because different poly­ cist .. or terrorist activity; commission of a serious What is often referred to as "the polygraph" is conducted in 1963, when Federal agencies, exclud­ graph procedures are required depending on in­ crime; involvement in adult homosexual activi­ actually a set of relatively complex procedures for ing NSA and CIA, conducted 19,796 polygraph tended use, it is necessary to consider validity by ty; involvement with illegal drugs or narcotics; asking questions and measuring physiological re­ examinations, and 1973, when 6,946 examinations polygraph technique and situation. In subsequent deliberate falsification of security processing sponses in order to detect deception or establish (including 3,081 by NSA) were conducted. chapters, such a variegated analysis is presented forms; treatment for a serious nervous or mental truth. Polygraph testing is employed for a varie­ and the scientific and policy contexts are more As shown in appendix B, NSA reports that it problem (187). According to NSA, the scope of ty of uses, ranging from ascertaining the guilt of uses primarily the R/I technique. NSA reports specific issue examinations is limited to questions a criminal suspect to assessing the honesty of a fully described. that counterintelligence-type screening examina- that are relevant to the issue to be resolved. Pre- Chapter 3 Controversy Over polygraph Testing Validity f""""":"':;-~" \ ~ , I

Chapter 3 Controversy Over Polygraph Testing Validity

INTRODUCTION The validity of polygraph examinations to de­ dential National Security Decision Directive (Mar. tect deception has long been a controversial issue II, 1983, NSDD-84). In part, the directive requires (d. 108,136,194,195). Since development of poly­ agencies and departments which handle classified graph techniques almost 80 years ago, their use information to revise existing regulations to per­ both within and outside the Federal Government mit use of polygraph examinations as part of inter­ has been the focus of numerous judicial opinions nal investigations of unauthorized disclosure of and, as well, legislative and executive branch classified information. Prior to the directive, in­ debate. Polygraph examinations have been advo­ vestigations of unauthorized disclosures had to cated as a way to ascertain guilt of criminal sus­ be referred to the Department of Justice (DOJ). pects, to exculpate innocent suspects, to protect Employees who refuse to submit to a polygraph national security, and to maintain employee hon­ examination could, if NSDD-84 is implemented, esty. Polygraph examinations have, at the same be subject to adverse consequences. In October time, been criticized for providing inaccurate and 1983, DOJ announced that administration policy misleading information, for failing to detect secu­ would also permit Government-wide polygraph rity risks (167), for interfering with the rights of use in personnel security screening of employees private citizens (128), and for lowering employees' (and applicants for positions) with access to highly morale. At the center of controversy over the use classified information. of polygraph examinations is the question of its Proposals to expand use of polygraph examina­ validity: does a polygraph examination actually tions to maintain national security have renewed identify truthful and nontruthful individuals? the debate about the appropriateness of various Recent interest in polygraph examinations and polygraph techniques and their ability to detect their validity stems from efforts to broaden Fed­ deception. In order to provide a context for the eral Government use. The Department of Defense present evaluation of scientific evidence on the (DOD), in late 1982, drafted revisions to existing validity of polygraph testing, previous assess­ regulations (5210.48). DOD proposed expansion ments of accuracy of polygraph testing are re­ of the use of polygraph tests for preemployment viewed in this chapter. Legal precedents regarding screening and periodic or aperiodic testing of polygraph testing and congressional hearings on employees who have access to highly classified its use, both within and outside of Government, information. Currently, only the National Securi­ are briefly considered. The chapter also describes ty Agency (NSA) and the Central Intelligence scientific criteria for establishing validity and Agency (CIA) are able to use polygraph tests in reviews other efforts to evaluate the scientific this way. Expanded use of polygraph testing in literature on testing. all Federal agencies was made explicit in a Pres i-

JUDICIAL REVIEWS When courts have been called on to resolve dis­ nique's validity and whether its use, however putes concerned with use of polygraph examina­ valid, interfaces with other values that the law .. tions, they have had to consider both the tech- seeks to protect. The varying decisions reached

29 ,.,,~. I, I 1

30 31

by State appellate courts and Federal circuits (see case, however, and recent discussion of the facts 8) may in large measure reflect varying beliefs mission of demonstrably more reliable techniques polygraph testing is felt particularly in industries of the case indicate that Frye was, indeed, guilty. is delayed until the requisite consensus has which have high risk of theft and fraud (e.g., com­ about the validity of polygraph examinations. In­ The crude polygraph examination conducted by deed, for many years, the leading case on the ad­ formed" (145; see, also, 60). mercial banks), high turnover (supermarkets, Marston, thus, appears to have yielded an inac­ other retail operations), or both. trdssibility of novel scientific evidence (Frye v. curate conclusion. When the courts examine polygraph testing, United States (58» was a case about the admissi­ they are faced with a series of dilemmas. To which According to Ansley (8), the use of private pol­ bility of polygraph evidence, and the opinion cen­ The Frye test is still used as precedent in most "particular field" of expertise can the courts turn: ygraph testing is limited by statute in 18 States tered on the question of validity. The issue of how Federal courts. Subsequent opinions (in areas physiology, psychology, polygraphy? If they look plus the District of Columbia. Most of these laws a court is to decide the question of any scientific other than the polygraph) have tried to better de­ to the data themselves, what are they to make of seek to protect employees from being requested, technique's validity has brought the Frye test into fine that line between "experimental" and "demon­ it? As the present report suggests, validity assess­ required, demanded, or subjected to polygraph question in recent years and makes salient the strative" stages of a scientific innovation. For ex­ ment involves a complex situation and technique­ examinations by their employers. Employers are problem of establishing judicial standards for ample, the court in United States v. Stifel (190) specific answer. Even if a final, single accuracy reported to be able to find ways around these assessing validity (60). held that "neither newness nor lack of absolute rate could be established, how should a court use laws. For example, employers may tell the em­ certainty in a test suffices to render it inadmissi­ it. How accurate must a diagnostic or predictive ployee that they suspect them of theft, but that ble in court." In a second case, United States v. technique be to be deemed valid for evidentiary if the employee can find a way to demonstrate Polygraph Findings as Evidence Brown (189), the court also seemed to be con­ purposes? Regularly admitted psychiatric evidence innocence, the employer will not discharge the cerned with validity: "The fate of a defendant in is widely recognized (including by the U.S. Su­ employee. In addition to polygraph validity, other The Frye case involved a 19-year-old defendant a criminal prosecution should not hang on his preme Court, see Addington v . Texas, (2» as hav­ polygraph-related concerns include issues of vol­ convicted of robbery and murder. Prior to his ability to successfully rebut scientific evidence trial, a well-known psychologist and one of the ing accuracy rates comparable to flipping coins untariness, invasions of privacy, being compelled which bears an 'aura of special reliability and (e.g., 55,208). In Barefoot v. Estelle (13) the Su­ to inform on othqr employees, inhibiting union originators of polygraph testing, Dr. William trustworthiness,' although, in reality the witness Marston, administered a IIsystolic blood pressure preme Court acknowledged that psychiatric pre­ activity, and the polygraph as a cover for racism ll is testifying on the basis of an unproved hypoth­ dictions of dangerousness and violent behavior and sexism. This list does not exhaust concerns test to detect deception (e.g., 114). Dr. Marston esis in an isolated experiment which has yet to determined, on the basis of this test, that Frye was do not exceed an accuracy level of 33 percent (see that have been expressed. gain general acceptance in its field." The Frye test 118). Yet, this evidence was held admissible in truthful when he denied involvement in the rob­ A survey of 143 private firms by Belt and Hol­ has been held to be too high a hurdle by some Barefoot and sufficiently valid to uphold a deci­ bery and murder. The trial judge, however, re­ den (25), regarding their use of polygraph testing, trial courts, which have replaced it with the test sion to execute a convicted person. fused to permit Dr. Marston to either testify about for admissibility of expert testimony generally: yielded a number of interesting findings. Twen­ the examination or conduct a reexamination using "testimony by a witness as to matters which are In summary, then, the courts have found them­ ty percent of respondents reported using poly­ the blood pressure test in court. beyond the ken of the layman will be admissible selves disagreeing on methods to establish validity graph examinations for preemployment screen­ Frye appealed his conviction on the grounds if relevant and the witness is qualified to give an for purposes of admissibility of evidence, where ing, periodic surveys, and investigations of spe­ that relevant exculpatory evidence had not been opinion as to the specialized area of knowledge" the critical focus of such judgment should rest. cific onsite crimes. It is interesting that of reasons given for using or not using polygnph tests, users admitted. The appeals court, how~ver, concurred (190). In addition, courts are inconsistent about what decision to make on the basis of judicial findings ranked moral or ethical considerations last and with the initial trial court judgment. The court A closely related question for the courts has reasoned that the systolic blood pressure decep­ of fact regarding the validity of a diagnostic or efficiency first; nonusers, however, ranked validi­ been who should determine whether some pro­ predictive device. ty and reliability second in importance, cost third, tion test was validated only by "experimental II cedure has gained general acceptance in its field. and the availability of qualified operators fourth evidence and was not based on a IIwell-recognized Some have held that the courts must look to the scientific principle or discovery." The decision in importance. The survey found a positive rela­ judgment of the scientific community (e.g., 191). Laws Regulating Polygraphs tionship between a State having a licensing re­ stated that, "while courts will go a long way in In other decisions, the court refused to "surrender in Employment Settings admitting expert testimony deduced from a well­ quirement for polygraphers and employers' use to scientists the responsibility for determining the of polygraph testing. According to Ansley (8),25 recognized scientific principle or discovery, the As described in chapter 2, screening employees reliability of (scientific) evidence," and that "a States have licensing requirements for polygraph­ things from which the deduction is made must be is the most frequent application of polygraph test­ determination of reliability cannot rest on a proc­ ers; licensing is optional in one State. sufficiently established to have gained general ac­ ess of 'counting (scientific) noses.' " ing. Many employers argue that use of polygraph ceptance in the particular field in which it belongs. "" testing for preemployment screening, periodic Although there is testimony that use of poly­ Just when a scientific principle crosses the line be­ Saks and Van Duizend (145) concluded that checking, and to resolve actual thefts is necessary. graph testing reduces employee crime (172), no tween experimental and demonstrable is difficult whichever set of tests is employed, the courts are Internal crime has been estimated to cost private formal cost-benefit analyses appear to have been to define." in a weak position to assess validity directly or industry up to $10 billion annually (see 172), and conducted. In addition, there is no research on to count scientific noses. The result has been: polygraph testing is regarded as a cost-effective the predictive validity of polygraph results Ironically, Frye's conviction was later reversnd 1) general deference by the courts to the judgments when another man confessed to the crime, thereby tool. Employers argue that screening applicants, (72,144). Although employee issues are critical to of scientific communities; and 2) "numerous in­ and periodic checking of employees, are the most proposed Government uses of polygraph testing, providing Frye with more convincing corroboni­ congruities ... where less reliable scientific and tion of his denials of guilt. This did not settle the efficient ways to control pilferage, embezzlement, few data are available un Government employees technological information is admitted but the ad- poaching, and other forms of theft. The need for (see chs. 4 and 5). ". \ I

32

One additional area of controversy has con­ destroys the trust relationship between employers cerned employee rights and employer-employee and employees, and creates employee dissatisfac­ relationships. The general matter of invasion of tion. However, the few employee surveys that privacy is particularly pertinent in preemployment have been conducted have not supported this ar­ screening and periodic checking. In preemploy­ gument. Apparently, five studies have examined ment screening, the range of questions that may whether the use of the polygraph causes private be asked has been subject to particularly heavy sector employees to be dissatisfied. In one study criticism. Questions have been reported to include (144),96 percent of applicants were willing to take items concerning union activity, sexual prefer­ a polygraph examination to get a job, 86 percent ence, and family problems (169); and, in addition, of the applicants thought the preemployment ex­ willingness to make a commitment to the job amination was fair, and 88 percent were willing (144), and whether the respondent has ever been to take it routinely as a condition of employment. tempted to steal (71). During periodic checking, A problem with the study was that applicants respondents are sometimes asked not only about were surveyed immediately after taking the poly­ their own possible improper behavior (e.g., un­ graph examination so they may havE: ihought their derringing in supermarkets), but also about their responses were part of the screening process. In level of job satisfaction, intention to remain with the one known survey of Federal employees, the the employer, and activities of their fellow em­ Air Force (183a) surveyed individuals who had ployees (204). There is some concern about volunteert!d to participate in a pilot project on the whether prejudices of the polygraph examiner use of the polygraph for counterintelligence/se­ based on racial, ethnic, and gender stereotypes curity examinations. About 99 percent of the re­ bias employees' responses (144). These assertions spondents felt that the examinaton was fair, and , do not appear to have been researched. And no were willing to take an examination for counter­ related claims under Title VII of the Civil Rights intelligence purposes. Acts have been upheld. One argument against the use of polygraph ex­ aminations in the employment situation is that it

FEDERAL DEBATE OVER POLYGRAPH VALIDITY

Concern about and debate over Federal Gov­ branch proposal to use lie detectors to find the ernment use of the polygraph have emerged at source of unauthorized disclosures of sensitive or several points during the past 20 years. As shown classifit::d information, sometimes known as in figure I, the histoJ.Y is essentially one of legis­ IJleaks" (192). The then chairman of the House lative concern triggered by some executive branch Committee on Government Operations asked the proposal or action regarding polygraph testing. Foreign Operations and Government Information The questions raised by Congress have included subcommittee to study the Federal Government's constitutional and ethical as well as validity issues. use of polygraphs. The study found that, exclud­ However, the scientific validity and reliability of ing the National Security Agency and Central In­ polygraph testing has been and is a central con­ telligence Agency (for which information was gressional concern. This chapter briefly aescribes classified), Federal agencies had conducted 19,796 the history of Federal Government involvement polygraph examinations in 1963. In 1964, the sub­ with the issue of polygraph validity. committee held hearings and received testimony from private polygraphers, researchers, and Fed­ The 1960's eral officials. In a 1965 report (167), the House Committee on Government Operations concluded Congressional interest first intensified in 1963 that there was no scientific evidence to support when controversy developed over an executive the theory of the polygraph, and that the research r

Figure 1.-Debate at the Federal Level 1963-83

1963-64 1965 1966 1967-74 1974 1975 1976 1977-78 1979 1980 1982 1983 Executive Proposal to use DOD Directive 5210.48 DOD issued Director of Cen- DOD Select President Issues branch polygraph to DOD research program revisions to tral Intelligence Panel Review of NSDD-84 activity investigate leak Civil Service regulations Directive study Investi- the DOD Person- DOD prepares 5210.48 gative Scope and nel Security further draft of Adjudicative Pro- Program revised regula- cedures Among Deputy Secre- tions Intelligence; Per- tary of Defense DOJ announces sonnel i3ecurity issues memos Survey administration instituting policy on poly- aperiodic poly- graph testing graph exams for those with SCI clearance DOD prepares proposed revi- sions to Direc- . tive 5210.48 Legislative Study and hear- Report Bills introduced Report Privacy. Report S.1845 intro- Hearings Hearings by Hearings by activity on ings by House issued by which would Polygraphs and issued by duced; would bar and report House Judiciary House Post Office Federal Government House restrict or ban Employment House Federal agencies by House Subcommittee and Civil Service agency Operations Government Federal Govern- issued by Government and private Permanent on Civil and Subcommittee on \. use Subcommittee on Operations ment use of Senate Judiciary Operations employers from Select Com- Constitutional Civil Service and Foreign Opera- Committee polygraph Subcommittee Committee requiring or mittee on Rights on DOD House Judiciary tions and Govern- on Constitutional requesting Intelligence proposed Subcommittee on ment Information Rights Hearings held on Subcommlt- revisions Civil and Con- Study and hear- S.1845 by Senate tee on stitutional Rights ings by House Judiciary Subcom- Oversight on NSDD-84 Government miltee on the Con- Congress passes Operations Sub- stitution amendment to committee on DOD appropria- Foreign Opera- tions bill pro- lions and hibiting changes Government in DOD polygraph Information regulations until Apr. 15. 1984 Hearings by Senate Commit- tee on Govern- ment Affairs on NSDD-B4 Legislative Some bills would Report referred S.1845 introduced Hearings HR2403 intro- activity on ban use of to private as Hearings held by duced; would private polygraphs by well as Federal House Ed- ban use of employer use private employers employer Privacy Protection ucation and polygraph by Study Commission Labor Sub- private employers issues report committee on Labor- Management Relations

\

-~~---~- ._------~----~~-~ ~-- -- ~~ ~~---~ ------~ ""'., » 'f~

34 35

evidence as to its accuracy was inadequate. The standards included directives concerning the spe­ the results therefrom, have done little to persuade to examine the utility of polygraph tests, but it committee recommended that further research be cific purposes for which the polygraph may be the committee that polygraphs ... have demon­ was not a validity study (165). strated either their validity or reliability in dif­ conducted and training for polygraph examiners used, and directives that a person to be examined As shown in figure 1, in addition to interest in ferentiating between truth and deception, other be upgraded, and that the President establish an must be informed as far in advance as possible Federal use of polygraphs, Congress has shown than possibly in a laboratory situation" (171). The interagency committee to study and work out so­ of the intent to use the polygraph and of the fact interest in the use of polygraph examinations by 1976 report concurred with the 1965 report that lutions to problems posed by Federal Government that refusal to consent to a polygraph examina­ private employers, in part because of constitu­ "There is no 'lie detector' " (171). Because of the use of polygraphs. tion will not be made a part of the person's per­ tional and privacy issues (see, e.g., 169,172,173; polygraph's "unproven technical validity" and the sonnel file. the Privacy Protection Study Commission Report Later in 1965, an interagency polygraph com­ suggestion that the "inherent chilling effect on in­ (128) mandated by Public Law 93-579; and several mittee of representatives from DOD, CIA, DO}, Also in response to the House Government dividuals subjected to such examination clearly laws introduced since 1967). Various congres­ Bureau of the Budget (now Office of Management Operations Committee's 1965 report, DOD pro­ outweighs any purported benefit to the investi­ and Budget), Office of Science and Technology posed, and in part undertook, an extensive poly­ sional committees have questioned the validity of i gative function of the agency," the Committee ! polygraph testing in a private employment con­ (now the Office of Science and Technology Poli­ graph research program. And in July 1965, DOD recommended a complete ban on the use of pol­ text, in particular as a condition for employment. cy), and other executive agencies was established. issued directive 5210.48 (177) to regulate the con­ ygraphs by all Federal Government agencies for Nevertheless, attempts to enact Federal legislation The interagency committee concluded that: 1) duct of polygraph examinations and improve se­ all purposes. However, 13 committee members regulating the use of polygraph examinations by there was insufficient scientific evidence concern­ lection, training, and supervision of its polygraph dissented, asserting both that the hearings had private employers and/ or the Federal Government ing the validity and reliability of polygraph test­ operators. Some of the results of the DOD re­ been held during an entirely different Congress, have not been successful. ing; and 2) the use of the polygraph constituted search program were later reported in a scientific and participated in by an entirely different group an invasion of privacy of the individual being in­ journal (29), but other reliability and validity of Members, and that, while testimony at the terrogated. The committee recommended that the studies proposed were never carried out (183). hearings represented a wide diversity of views, The 1980's "use of the polygraph in the executive branch Between 1967 and 1973 a number of bills were no witness had urged prohibition of the polygraph should be generally prohibited, and permitted In the recent past, the executive branch has introduced which would have either limited the for all purposes. The dissenters urged adoption only in special national security operations and again taken initiatives concerning the Federal use questions that could have been asked or banned of the recommendations originally proposed and in certain specified criminal cases" (166). The rec­ of polygraph testing. In April 1982, a DOD select altogether polygraph use by Federal agencies voted on by the members who had participated ommendations made at that time concerning per­ panel reviewed the DOD personnel security pro­ (170). None of these bills was enacted. in the hearings. These recommendations would sonnel screening were promulgated as Civil Serv­ gram (180) and expressed dissatisfaction because have, in part, prohibited the use of polygraphs of inconsistency in polygraph use across compo­ ice regulations on regulating the use of polygraphs in all cases except "1) those clearly involving the in personnel investigations of competitive service nent programs (as did the U.S. Congress (173)), The 1970's Nation's security, and 2) those in which agencies and the lack of reinvestigations. The panel ob­ applicants and appointees to competitive service can demonstrate in compelling terms their need positions (ch. 736, app. D, of the Federal Person­ Ten years after the 1964 hearings, this same served that military personnel, unlike civilians, for use of such devices for their law enforcement were appointed to NSA and allowed access to nel Manual). According to these regulations, House Government Operations subcommittee purposes, and that such uses would not violate which are still in effect, only executive agencies conducted another review of polygraph use by Sensitive Compartmented Information (SCI) with­ the fifth amendment or any other provision of the out undergoing a polygraph examination. In ad­ with highly sensitive intelligence or counterintel­ Federal agencies (169). In 1974 hearings, the sub­ Constitution." ligence missions directly affecting the national se­ committee found that the use of polygraphs in the dition, personnel could continue to get clearances curity such as "a mission approaching the sensi­ Federal Government had declined substantially The concern with scientific validity and its throughout their careers without ever being sub­ tivity of that of the Central Intelligence Agency" since 1963. In fiscal year 1973, a total of 6,946 implications for the Federal Government's use of ject to reexamination. The DOD panel recom­ are permitted to use the polygraph for employ­ examinations were conducted, including 3,081 by polygraph testing arose again in 1979 at hearings mended a broadened application of the polygraph ment screening and personnel investigations of ap­ NSA. This compared to 19,796 in 1963, excluding held on preemployment security clearance proce­ for security screening purposes, and selective use plicants for and appointees to competitive service NSA and CIA. Tne subcommittee also found that dures by the House Permanent Select Committee of counterintelligence scope polygraph examina­ positions. All other uses of a polygraph to screen there was not much additional research on poly­ on Intelligence, Subcommittee on Oversight (175). tions during periodic reinvestigations. The panel applicants for and appointees to competitive posi­ graph validity. The only federally funded studies The subcommittee found that there had been in­ noted that the recommended expanded use of the tions are forbidden. conducted had been those reported by the DOD sufficient research on the accuracy of the poly­ polygraph would require changes in DOD Direc­ Joint Services Group (183), and these studies were graph technique in screening job applicants and tive 5210.48. The regulations also set forth steps for deter­ considered by DOD to be inadequate for deter­ that "gaps in the statistics kept by the intelligence mining whether agencies met the criteria of hav­ On August 6, 1982, the Office of the Deputy mining the validity and reliability of Federal services do not make it possible to make the clear ing a highly sensitive mission, and stipulated that Secretary of Defense (39) issued a memorandum polygraph testing. judgment that the polygraph is unique and indis­ approval to use the polygraph would be granted requiring employees with SCI access to agree to pensable" (173). The Director of Central Intelli­ only for1-year periods. Agencies intending to use In a 1976 report based partly on the 1974 hear­ submit to polygraph examinations on an aperiodic gence (DCI) was urged to conduct a study to vali­ the polygraph for personnel screening were re­ ings, the House Government Operations Commit­ basis, and revised DOD Directive 5210.48 accord­ date the accuracy of the polygraph for preemploy­ quired to prepare regulations and directives meet­ tee concluded that lithe nature of research under­ ingly. Later in 1982 and again in early and mid- ment screening. DCI did conduct a study in 1980 ing certain minimum standards. The minimum taken, both federally and privately funded, and 1983, further revisions to DOD Directive 5210.48

----.-----~ "" -'-- ~ \ i 1

36 37

were drafted (181). In 1983, the President issued NSDD·84 SCIENTIFIC VALIDITY AND POLYGRAPH RESEARCH REVIEWS a National Security Decision Directive (NSDD-84) also authorizing broader use of the polygraph. On March 11, 1983, the President issued a Na­ Thus, recent polygraph policy actions have developed assessments of polygraph validity that Congress responded to these developments by tional Security Decision Directive intended, ac­ renewed interest in and debate over the scientific are accepted by the scientific community. conducting several sets of hearings, by requesting cording to DO] officials, to help safeguard against validity of the polygraph. Reviews of scientific In order to make explicit the criteria for validity OTA and General Accounting Office studies, and unlawful disclosure of properly classified infor­ literature form the principal means to cumulate used in this assessment, below are described sev­ by passing an amendment to the DOD appropria­ mation. One provision of NSDD-84 requires that research findings and are especially important in eral dimensions of validity and how they are as­ tions authorization bill (S.675) putting a morato­ persons with authorized access to classified infor­ order to assess the validity of polygraph testing. sessed. This description is based both on standards rium until April 15, 1984, on any revisions to mation sign a nondisclosure agreement, and that Single research studies, no matter how well con­ for psychological/psychometric tests (d. 3,5) and DOD Directive 5210.48 retroactive to August 5, persons with access to SCI must also agree to pre­ ducted, cannot answer global questions about va­ criteria to evaluate research designs (d. 41,147). 1982. On October 19, 1983, DO] announced a publication review. These provisions are outside lidity and must be considered in relation to other Although criteria for validity can be described ob­ new administration polygraph policy that would the scope of this memorandum, as is a full analysis evidence. Both because research evidence about jectively, it should be noted that it is essentially permit further expansion in polygraph use. The of NSDD-84. polygraph testing has rapidly increased, especially a qualitative judgment as to whether (or, to what DOD draft revisions, NSDD-84, and administra­ within the last 10 years, and because there have With respect to the polygraph, NSDD-84 in extent) a given criterion is met. In addition, tion polygraph policy are discussed in more detail been disagreements about the nature of evidence effect authorizes agencies and departments to assessments of the "preponderance" of evidence below. about polygraph testing, there have been a num­ require employees to take a polygraph examina­ necessary in order to assess the overall validity ber of such reviews. These reviews are important, tion in the course of internal investigations of of polygraph testing are similarly subjective. In because they are frequently cited in both legal and unauthorized disclosures of classified examina­ chapters 4 and 5, a systematic analysis of avail­ Draft Revisions to DOD 5210.48 legislative considerations and because they serve tions. NSDD-84 also provides that refusal to take able research is attempted, although it should be to shape future research. The draft revisions to the DOD polygraph reg­ a polygraph test may result in adverse conse­ recognized that there are a number of ways to ulations have gone through several iterations. For quences. NSDD-84 permits administrative sanc­ Underlying each of the reviews is the applica­ conduct such evaluations, each of which may the purposes of this validity study, a primary pro­ tions, including denial of security clearance, to tion of a set of criteria, only sometimes made ex­ yield a somewhat different outcome. posed revision (as of the March 1983 draft) is to be applied even when a person is not subject to plicit, regarding the validity of individual studies a criminal investigation (184). authorize the use of the polygraph for deter­ t, and their implications for overall assessments of Reliability polygraph testing accuracy. As introduction to mining initial and continuing eligibility of DOD Assessment of any test's validity is based on the Administration Polygraph Policy the scientific reviews, the nature of these criteria civilian, military, and contractor personnel for assumption that the test consistently measures the is described. The reviews, themselves, are then access to highly classified information (SCI and/or same properties. This consistency, known as relia­ On October 19,1983, DO] announced a com­ summarized and a preliminary analysis of discrep­ special access). The use of the polygraph in deter­ bility, is usually the degree to which a test yields prehensive administration policy on Federal agen­ ancies among reviews is presented. More d~tailed mining continuing eligibility would be on an repeatable results (i.e., the extent to which the cy polygraph use. The policy authorizes poly­ analysis of individual validity studies is provided aperiodic (i.e., irregular) basis (1B1). same individual retested is scored similarly). graph testing: in chapters 4 and 5. Also, the proposed revisions provide that re­ 1. as a condition of initial or continuing Reliability also refers to consistency across ex­ fusal to take a polygraph examination, when employment with or assignment to agencies aminers/scorers. A reliable polygraph test should established as a requirement for selection or with highly sensitive responsibilities direct­ Definitions of Scientific Validity yield equivalent outcomes when subjects are re­ assignment or as a condition of access, may, after ly affecting national security; tested and, as well, be scored similarly by indi­ consideration of all other relevant factors, result 2. as a condition of access to highly sensitive Validity viduals other than the initial examiner. For ex­ in adverse consequences for the individual. Ad­ categories of classified information; The validity of polygraph testing means, in ample, if a polygraph examiner reviewed a set of verse consequences are defined to include non­ 3. to investigate serious criminal cases; and charts and concluded that a subject was decep­ selection for assignment or employment, denial 4. to investigate serious administrative miscon­ nontechnical terms, accuracy of the test in detect­ ing deception and truthfulness. The problem of tive, any other polygraph examiner should be able or revocation of clearance, or reassignment to a duct cases including unauthorized disclosure to review the same charts and conclude that de­ nonsensitive position. of classified information (185a). assessing polygraph validity is especially difficult, not only because polygraph tests take a number ception was indicated. This illustrates interrater­ Technically, these expanded uses of the poly­ The policy in essence authorizes use of the poly­ of forms, but also because validity has different reliability. Such reliability might be affected by the amount and type of training of examiners. graph are considered to be part of personnel secu­ graph on a Government-wide basis for the ex­ dimensions and can be measured in 8. number of rity investigations. Use of the polygraph within panded polygraph uses proposed by DOD. Thus, ways. There are, as a result, a number of different The present study focuses primarily on validi­ DOD is already authorized under the existing 1975 for example, the policy provides agency heads forms of validity associated with polygraph ex­ ty because if a testing procedure is not measur­ v..ersion of 5210.48 for various criminal, counter­ with the authority to give polygraph examinations aminations depending on the type of polygraph ing what it purports to measure (validity), it mat­ illtelligence, and intelligence purposes. on a periodic or aperiodic basis to randomly test as well as on its use (e.g., employee screen­ ters little that it can measure the same thing again selected employees with access to highly sensitive A detailed review of the proposed changes is ing v. investigation of a criminal suspect). These and again. Examiners who consistently agree that information, and to deny such access to employ­ beyond the scope of this technical memorandum. difficulties underlie, in part, the failure to have they are seeing "deception" may in fact be measur- ees refusing to take a polygraph exam. r-r~, .- ...- ...... ,- ...... ~ ... ,.,....--~-~-- -~-- --- ~ ------~

I t I I

38 39

ing anxiety or some other form of arousal. Relia­ the usefulness of a test. A related set of validity The rate of false positives or negatives is In polygraph field studies, polygraph examin­ bility is, however, a necessary condition for va­ crtieria are also used to evaluate the validity of sometimes difficult to establish because, in re­ ers' decisions are compared against some post hoc lidity to be established. A test that is valid will, any single study design. These research design cri­ search studies, a number of criteria for deception/ determination of whether suspects are guilty or necessarily, be reliable. teria include, most importantly, internal and ex­ nondeception may be applied. Thus, for exam­ innocent; i.e., "ground truth." These post hoc ternal validity (d. 41,147). ple, in studies which employ numerical scoring determinations may, in different studies, consist Construct Validity Internal validity refers to the degree to which for polygraph charts, depending on the scoring of confessions by the presumably guilty party, Construct validity refers, in broad terms, to a study has controlled for extraneous variables system (e.g., cutoff points), different diagnoses decisions by a panel of attorneys or judges assem­ whether a test adequately measures the underly­ which may be related to the study outcome. Ex­ will be made. The rate of false positives and bled specifically for a particular study who base ing trait it is designed to assess. A polygraph test ternal validity refers to the established general­ negatives may also depend on the examiner's per­ their decisions on investigative files excluding is designed to detect deception. It is therefore im­ izability of a study to particular subject popula­ ception of the "base rate" of guilt/innocence. references to polygraph decisions, judicial out­ comes (dismissals, acquittals, convictions), as well portant to clearly define the construct of decep­ tions and settings. Internal validity in the case of In some cases, the examiner will deal mostly as other criteria. The fact that determinations of tion, and distinguish it from other concepts such a study of polygraph testing is usually enhanced with deceptive subjects (e.g., in certain criminal guilt or innocence are made post hoc makes draw­ as guilt. by the presence of control groups. Typically, such investigation contexts) and, thus, may be predis­ ing conclusions from field studies difficult (126). conditions of an experiment permit analysis of posed to make false positive diagnoses. In other To measure construct validity, it is necessary In real life situations, truth is seldom available variables such as different question formats. In settings (e.g., some personnel screenings), an ex­ to both describe the construct and show its rela­ (62). tion to a conceptual framework. Construct valida­ most field studies, internal validity is difficult to aminer may test only a small number of decep­ tion, thus, requires that a test be based on some establish since the investigation cannot control or, tive subjects and, then, may be predisposed to Attempts to use confessions, panel judgments, theory or conceptual model. Since different types in many cases, have definitive knowledge about false negative decisions. Regardless of rates, judicial outcomes, and other criteria as indicators of polygraph tests have different theoretical bases whether a subject is guilty or innocent. assessment of conditions that cont-ibute to either of truth have their own problems. Individuals (see ch. 2), there are multiple forms of construct External validity is simply the nature of the sub­ type of error is a focus of the reseach literature. may confess to crimes which they did not com­ validity for the polygraph. Construct validity is jects and settings tested. The broader the popula­ mit (108). In addition, individuals are sometimes established by various means. Most important­ tion examined and the type of setting investigated, Reviews of Polygraph Validity falsely convicted (34). Panel decisions may be gen­ ly, based on theoretical predictions of how items the wider that study'S results can be generalized. eralizable only to cases in which sufficient inves­ should interrelate or how other tests should inter­ In a parallel way, the more similar the research Since at least 1973, a number of polygraph re­ tigative information is available to make a deci­ correlate, actual evidence (e.g., scores from simi­ situation to the "real life" situation, the greater searchers and psychologists interested in physio­ sion without the addition of polygraph testing. lar tests) is examined. If no such predictions are a study's external validity. Evidence about exter­ logical detection of deception have reviewed avail­ One can never be certain that the panel decision possible, it is impossible to establish construct nal validity is developed both from investigations able scientific literature to assess the validity and is indeed correct, and the pallel and the polygraph validity. that test a broad range of subjects and situations reliability of polygraph testing. Most such reviews examiner may have been exposed to the same and from investigations that identify subject and focus on studies of criterion validity, although a prior information (62). Thus, while field studies Criterion Validity setting interactions with polygraph test outcomes. growing number of investigations deal with con­ provide the most direct evidence about polygraph The broader the population examined and the struct validity. The most important difference test validity, they have been criticized because Although from a theoretical point of view con­ among these criterion studies has to do with they do not adequately meet the standards of struct validity is most important, from a practical type of setting or the more similar it is to the situa­ tion for which one wants to use a test or a the­ whether they are conducted in actual field situa­ "ground truth" to establish criterion validity. point-of-view, criterion validity is the central tions or in "analog" situations. component of a validity analysis. This aspect of oretical construct, the greater a study's external validity. Comparison of Reviews validity refers, in the case of polygraph examina­ Field Studies tions, to the relationship between test outcomes A number of independent reviews (listed in and a criterion of ground trutt. In this respect, False Positives and Negatives For purposes of this technical memorandum, table 2) of the field evidence on polygraph testing criterion validity is what is meant by test accura­ field studies are those studies or "naturally" were assessed in order to determine reasons for With any test, the possibility exists of false pos­ occurring polygraph test situations; i.e., studies differences among reviews. The reviews differ in cy. In the absence of construct validity evidence, itives and negatives. False positives are decisions however, it is difficult to determine to what ex­ in which the researcher does not exercise experi­ a number of respects. In part, reviewers' conclu­ that individuals are being deceptive when they are mental control over the situation in which the sions differ because they include different kinds tent criterion validity data can be generalized. In providing truthful responses. Their charts are some situations, it is not clear which aspects of crime or other event occurred. Not exercising ex­ of studies and even different studies (despite, in scored as showing a "deceptive" reaction for some perimental control means that the researcher does several cases, having had the same studies avail­ a test are responsible for accuracy, and what fac­ other reason. False negatives are decisions that in­ tors cause a test to be inaccurate. not systematically assign people to conditions of, able to them). In addition, some reviews differen­ dividuals are not being deceptive when in fact they for example, guilt or innocence. We refer here to tiate between accuracy in detecting deceptive v. are being deceptive. There are a number of rea­ "field" studies but others (e.g., 7) use the ter­ nondeceptive subjects, emphasizing the problems Research Design sons why such false outcomes might be obtained minology "real" cases (v. "laboratory"). Abrams of false positives and false negatives; others ag­ and, in part, they depend on the criteria (e.g., The above validity criteria are those which are (1) differentiates between the laboratory and "ac­ gregated the overall accuracy rates across both amount of physiological change) used to indicate tual criminal cases." groups of subjects. Finally, there are differences typically assessed in considering evidence about deception or truthfulness. Table 2.-Reviews of Field Studies of Polygraph Validity c~ Raskin & Abrams Horvath Podlesny Lykken Ben-Shakhar Ansley Year (1973) (1976) (1979)a (1981) (1982) (1983) Reported "hit rate" (DIJ Bitterman and Marcuse (N=2Ex, 81S) .•...... •.....•.. 1947 RII NR NR EXg NR [100] Ben-Ishai (N = 1 E, 100C) ...... •.....•...... • 1962 C A NR NR NR NR [97.0]1 unpublished Ben-Ishai (N = 1 E, i0C) ...... •...... •...• 1962 C A NR NR NR NR [100] unpublished Bersh (N =4E, 216C) ..•...•.•.•...... •.•...... 1969 C G NR 85.6 I 89.3 A [74.6.92.4]1 [75-92]j [96.2]1 Horvath and Reid .....•...... •...... •....•..••...•..... 1971 C G [79:l-91.4]k EXI ·15.89k EXI 85.0 NR I 89·94k 90.5 Hunter and Ash ...•....•.••.....•..•...... •.....•...• 1973 C G EXI 88 EXI 87.1 NR I 86 85.7 Wicklander and Hunter ...... •..•...... •....•.•...... 1975 C G EXI 95 EXI 90.0 NR I 93 86.7 Siowick and Buckley •...... •..•.•...••..•...... •.... 1975 C G EXI 85 EX' 83.8 NR I 93 90.5 Raskin (3 studies) ...... •...... •.....•...... •.•.....••.. 1976" C G 93-100m NR NR NR I 69·95m Barland and Raskin (N = 102C) .•.•...... •..... _.•...••... 1976 C G NR 98 83.0 NR I 45 29.4 A [78.7·89.7]° [72] Horvath .....••..•.•...••.•...... •.• , ...... , . 1977 C G EXI NR 77 77.1 NR I 51 51.1 Davidson ...... •.•...... •.•.....•...... •..•..•...... • 1979 C NR NR EXI Edwards (N = 80E, 959C) ...... 1981 A NR NR [98.3]P unpublished Elaad (N = 184C) •...... •.•...... ••.••.•••....••..•. 1976 A NR NR NR [98.3]Q unpublished Peters (N = 172C) .•...••...•••.•...... ••.••...... •...... 1982 A [95.1]1 unpublished Wldackl (N =38C) ...•....•...... •...... •.••.....•...... •. 1982 z A [95.8]1 Author's conclusion (If any) ....•..••.....••.•...•..••.•... G 90 I 89 A 98tt NR 97.6 aNottechnlcallya review. Original review (126) relied on lab evidence. When crillcized by lykken, aUlhors responded In part IExcluded because they are seen as reliability studies. Horvath reasons that they used confessions as a criterion and because by ciling field sludles. confessions are not Independent of polygraphers' examinations "their usefulnass as a criterion measure for esllmallng validity bpresented In chronological order. Is IImlled." lykken Bxcludes them because both examiners and evaluators were Raid polygraphers.

Co a question type (C a control question; R/I _ relevanl/lrrelevant; Z D zone of comparison). mlower rate Is for nonnumerlcal evaluaUon; higher rate for numencal evaluallons. dO = correcl declsfon or rate of agreamenl (I.e., hll rale) for gulll}' (decepUve) (G) or Innocent (non deceptive) (I) subjects. nFocus of study was to test the "friendly polygrapher" hypothesis (see Raskin, Barland, and Podlesny, 1978). A = average hit rate (In cases where G and I are not reported separately). 078•7 percent = agreement between simple majorlly on panel and polygraph decision; 86.7 percent a agreement between

eN D number. Ex D number of examiners; E D number of evaluators; C D number of cases evaluated; S a number of sub­ 4/5 majority on panel and polygraph; 89.7 percent - agreement between Judicial outcome and polygraph (where Judiciary ~ects when actual subjects are tested In the study. not aware of polygraph resulls). Different studies use different criteria for valfdll}'. Thus, the hll rate can be ell her polygraphers' agreement with ground truth PDf 2,433 polygraph examinations In the original pool, only 959 (39.4 percent) could be verified as to truthfulness or not; (e.g., BIHerman and Marcuse), blind evaluators' agreement wllh polygraph examinations verified by confessions (e.g., Hor· of these 959, 943 were found to be correct. The 98.3 percent figure reported by Ansley represents 943/959 and does not vath and Reid, Hunter and Ash, WICklander and Hunter, Slowlck and Buckley, Horvath), or polygraph examinations valfdated take Into account the enllre pool of examlnallons. by a panel of judges using other criteria (e.g., Bersh, Barland, and Raskin) or actual Judicial outcome (WJdackl). Researchers qActuaily, of 1984 cases that were confirmed, 10 were Inconclusive and 6 were errors. The percent accuracy given In Ansley's have not agreed on what constitutes an appropriate criterion for validity In the field. review excludes the Inconclusive results and adds half the error rate to the accuracy rate. If one appropriately excludes both gExcluded by lykken as not useful evidence because of base rate problem. the examiners In the study assumed only one and Inconcluslves and the error rate, the accuracy rate Is at most 87 percent. hperson was guilty. Therefore, they would have had a 99·percent accuracy rate even If they had called everyone Innocent. rThe lower fdte fs for studies using confessions criteria; higher rate Is for studies using panel of experts criteria. Figures are NR - not reported by the reviewed Indicated. rounded. lit Is Important to note Ansley's method of calculating overali accuracy. When there are Inconcluslves (or disagreements be­ SUe control test only. tween the original examiners and the evaluator), Ansley divides the percent Inconclusive by 2 and adds one·half the percent­ tCalculalions based on lykken's conclusion that 6 out of 11 Innocent subjects were Judged deceptive. age to the accuracy percent reported by the original researcher. ttAbrams uses many other studies to come to this conclusion. Many are anecdotal. Other researchers (e.g., Horvath) say they jThe 74.6 percent figure Is for majority decision; 92.4 percent for unanimous decision. shOUld not be rolied on. kThe lower rate Is the accuracy rate for Inexperienced polygraphers; the higher rate Is the accuracy rate for experienced polygraphers.

\

- .. - ....':.. 41 in the way accuracy rates were calculated, in par­ 84 percent accurate for guilty suspects and 76 to ticular, how inconclusives are handled. Each of 81 percent accurate for innocent suspects. As a these differences has important implications for result, Ben-Shakhar, et al., concluded that exam­ the conclusions developed by the reviews. iners tend to value detection of guilty suspects highly, even at the risk of falsely classifying in­ Several reviews (1,81) conducted 5 to 10 years nocent suspects; their conclusion concurs with ago reported relatively positive conclusions based Lykken's. Ben-Shakhar, et al., in conductng their on an evaluation the scientific literature. (\f review, employ a utility theory approach based Abrams (1) in 1973 reviewed reports of the on Bayes' theorem. They predict dramatically dif­ polygraph's accuracy dating from 1917, including ferent utility rates based on different base rate anecdotal as well as experimental data. He cal­ assumptions. culated approximate estimates of overall accuracy from this data, noting, however, that "it is almost Although these recent reviews, by authors who meaningless to total and average these findings are not professional polygraphers, cast doubt on because of the great discrepancy in experii.'\ental the validity of at least the most common poly­ paradigms and the instruments employed." He re­ graph technique, a more recent review by Ansley ported that in studies with complete verification (7) comes to the most positive conclusions since of ground truth, diagnoses were 100 percent cor­ those of Abrams. Ansley's 1983 review is an im­ rect. In other field studies prior to 1963 Abrams portant review because it represents the views of calculated an accuracy rate of 98 percent. In NSA's chief polygraph exa.miner. (NSA conducts laboratory experiments prior to 1963, Abrams the largest number of polygraph examinations of estimated the average accuracy rate of 81 percent. any Federal agency.) As shown in table 2, Ansley Averaging the results of the reports between 1963 concludes that field research shows a 97.2-percent and 1973, Abrams estimate of laboratory and field validity rate and laboratory research a 93.2-per­ research accuracy was 83 and 98 percent, respec­ cent validity rate. Based on these validity calcula­ tively. Horvath's (6) review in 1976 used some­ tions as well as separate calculations for reliability what more stringent criteria in selecting data than and utility, Ansley concludes that the polygraph did Abrams. His review does not include an over­ is "clearly an excellent adjunct to the selection all average accuracy rate calculated across studies. process." The early positive views of the polygraph's Unfortunately, for the most part, polygraph re­ worth have recently been challeng'ed by Lykken views contained in table 2 do not explicitly state (108) and, to some extent, by Ben-Shakhar, et al. their study selection criteria (see 63). The result (28). Lykken in 1981 challenged the theoretical is that a number of different studies have been assumptions of the most prevalent question tech­ included in various reviews, each of which pre­ nique, the control question technique (CQT), and sents different problems for interpretations of asserted that an average 50-percent false positive validity. The kinds of studies include reports of rate supported his theoretical challenge. Lykken, single criminal investigations in which the actual however, continues to believe that particular pol­ solution to the crime is the criterion for validity; ygraph techniques are useful (Le., the detection studies in which "blind" polygraph interpreters of guilt by measuring physiological arousal) and compare their polygraph chart evaluations to offers the use of the guilty knowledge technique "ground truth" as established by confession; and as a way to increase overall validity. Adoption studies in which the judgment of legal profes­ of Lykken's suggestion would preclude the use of sionals, actual judicial outcome, or in one case, the polygraph for preemployment testing and pe­ the judgment of a single psychologist, is used to riodic checking. establish ground truth. Ben-Shakhar, et al.'s (28), analysis also limited Some reviews do specify criteria for exclusion. their assessment of the polygraph to CQT. Their Lykken, for example, does not include studies of 1982 assessment of existing polygraph field re­ single criminal investigations. Abrams, on the search indicated that polygraph testing was 83 to other hand, includes in his review a number of 43

42 favor of the examiners. For example, in the Bersh Apart from the different base rates on which study half of the difference between the typical­ such studies (e.g., 30,103). Lykken's reasoning Conclusions about the validity of the polygraph most of the reviewers calculated accuracy rates ly re~orted 92.4-percent rate and 1?0 perce~t. is was that in single criminal investigations, the ex­ may depend on whether the reviewer attends to (see above), one source of different accuracy rat~s 7.6 which Ansley divides in half, leaVIng a valIdIty aminer has a large chance of being accurate (de­ the average accuracy rate or to the accuracy for applies uniquely to Ansley (7). In any case In rate of 96.2 and an error rate of 3.8 percent. The pending on the number of suspects) merely by guilty and innocent subjects separately. The con­ which there is not 100-percent accuracy, the same method is used for the Peters, Elaad, and calling everyone innocent. The fact that other clusions of all decision statistics contributes to the Ansley review computes validity by dividing the Widacki studies, for which the preadjustment va­ reviewers do not include Bitterman and Marcuse, ability to make an accurate asses~ment of poly­ difference between the accuracy rate and 100 per­ lidity rates are 90.2, 96.6, an~ 91.6 I?ercent, and other such reports, implies that they accept graph testing validity, particularly in view of the cent (the so-called error rate) in half and adds half respectively. Each of these studIes, partIcularly Lykken's evaluation of the usefulness of such stud­ concern over both high false positive and high of the difference to the accuracy rate. Ansley uses Elaad (see ch. 4), have other problems of inter- ies as indicators of validity. It is possible that false negative decisions. If, for example, the in­ this procedure on the grounds that on the basis results of such reports could be useful in assess­ nocent correct rate is 80 percent but the remain­ of chance, errors were probably half in favor ?f pretation as well. ing polygraph screening of large numbers of in­ ing 20 percent consists of inaccurately calling the panel (or other criterion measure) and half In dividuals in specific incident cases, such as might innocent subjects guilty, a different policy con­ be the case in unauthorized disclosure investiga­ clusion may be drawn than if the remaining 20 tions. However, additional factors limit the ex­ percent consists of "inconclusives" or of false ternal validity of Bitterman and Marcuse and nega·jves. In some cases (e.g., preemployment CONCLUSIONS other such studies. In Bitterman and Marcuse, for screening), inaccurately designating nondeceptive Another explanation is that polygraph testing Central to legal, legislative, and scientific assess­ example, the investigators were psychology pro­ people as deceptive may have worse consequences has been viewed as a single technique. Thus, ment of polygraph tests are their validity. Yet, fessors apparently conducting their first polygraph for the employee than inaccurately deciding that despite testimony (e.g., which urge? differen­ despite many decades of judicial, legislative, and ~37) tests, and they did not use accepted polygraph deceptive individuals are nondeceptive. In some tial consideration of pOlygraphs used In, for ex­ scientific discussion, no consensus has emerged procedures or instruments. There are no recent cases (e.g., a heinous crime by a potential repeat ample, employment screening and cri~nal inves­ about the accuracy of polygraph tests. One ex­ systematic studies of specific incident investiga­ offender, infiltration by a foreign agent), a false tigations, the scientific evidence for part~cular pur­ planation is that scientific criteria for de?l tions involving a large number of suspects. negative may have serious consequences. validit~ poses has not been differentiated. As IS demon­ with a number of dimensions and that the CrIterIa strated by the analysis of scientific There is strong disagreement among reviewers vary widely among specific research studies. In litera~u~e (~e~e In only two reviews (Ben-Shakhar, Lykken) are and in chs. 5 and 6), in assessing valIdIty It IS about whether another group of studies should order to assess overall polygraph examjnation be included as indicators of validity. These studies summary percentages provided in terms of the necessary to separate clearly the purposes for percent accurately detected for both guilty and validity, it will be necessary to examine de~ai~s which polygraph examinations are conducted and were conducted with records selected from the of each of the relevant studies. Such analYSIS IS files of the John E. Reid & Associates polygraph innocent; in other reviews, these figures are pre­ the types of techniques employed. presented in chapters 4 and 5. firm. A group of cases was used which the authors sented as the average percent of accurate detec­ considered to be "verified" by confession of the tions. In some cases, the percent inaccurately guilty suspect (in most cases they were also veri­ "detected" as nondeceptive (when they were really fied by some form of corroboration (37)). The deceptive) or deceptive (when they were really polygraph charts in these cases are then reinter­ ncmdecepthre) as well as percent inconclusives preted by a group of polygraphers who are "blind" were also reported by reviewers. But for purposes to (Le., do not know) the suspect's guilt or inno­ of clarity these have been omitted from table 2. cence. The degree of agreement of the "blind" evaluators to verify guilt or innocence is the test Another reason reviews differ about the results of validity. Two reviewers (Horvath, Lykken) ex­ of the same studies is the fact that they make dif­ plicitly excluded the group of studies conducted ferent decisions about the base rate of subjects or based on Reid files. Horvath excluded them be­ cases that aTe included. If, for example, a panel cause they used confessions as a criterion (con­ cannot make a decision about 30 percent of the fessions not being independent of the polygraph cases (e.g., 22), some reviewers will omit the examinations), and Lykken because both examin­ number of nonagreements from the number in­ ers and "blind" evaluators were polygraphers from cluded in I[he accuracy rate and base accuracy the same firm. His claim was that the studies were, percentages on only the remaining cases. This ac­ thus, "merely demonstrations that Reid's examin­ counts foI' the difference between Horvath and ers score charts in a similar way" (108) and so Ben-Shakhar, et al., analyses of the Barland and were estimates of reliability rather than validity. Raskin results. In other studies (and reviews of However, reviews by Raskin and Podlesny (138) those studies, e.g., Ansley, Abrams) inconclusive and Ben-Shakhar, et al. (27), each use all four Reid polygraph results are excluded from the analysis. studies to assess validity. This has the effect of inflating the accuracy rates. *E.i ; 9. - ---~-- ~---

Chapier 4 Review and Analysis of Polygraph Field Studies

INTRODUCTION Chapter 4 As noted in the discussion of previous scientif­ incident criminal investigations using the control ic reviews of polygraph validity, considerable dis­ question technique (CQT). This is an important agreement exists among reviewers as to which limitation. Because a systematic review helps to field studies and what kinds of evidence constitute identify this kind of problem, researchers and Review and Analysis of acceptable tests of validity. This chapter presents policymakers have a better basis on which to the results of a systematic analysis of existing field determine what, if any, additional studies are studies of polygraph testing in order to make an necessary. Also, the analysis aids understanding Polygraph Field Studies independent assessment of validity. Field studies of which question techniques, test purposes, ques­ investigate actual polygraph examinations and tion designs, and scoring techniques have been constitute the most direct evidence for polygraph studied and which may require further research. test validity (27). Both quantitative and qualitative The analysis is designed to address many of the techniques are utilized in order to make an overall problems associated with qualitative or "literary" assessment of existing evidence (63,125,14.2). reviews of the research literature previously dis­ cussed. In particular, the analysis makes explicit The goal of this analysis is to synthesize avail­ the criteria used for both study selection and data able research. Almost all of the available Held analysis (63,125,142). evidence comes from cases involving specific-

STUDY SELECTION Studies were considered field studies of validi­ verification are unclear or unsystematic, have also ty if their sample consisted of actual instances of been excluded. polygraph examinations conducted by profession­ al polygraph examiners, used field-tested poly­ The population of field studies considered for graph techniques, and used some independent the present analysis was, in general, taken from criterion to assess actual guilt or innocenc:e. those studies referred to in existing reviews of the Although ground truth can probably never 'be scientific literature (see ch. 3). In addition, re­ known in an absolute sense, studies can be con­ searchers active in the field of polygraph research sidered studies of validity only if they included were contacted and asked to supply the names and some adequately described and systematically de­ publication information of any additional recent termined criterion of "truth" (e.g., panel decision, studies. A bibliography provided by the American judicial outcome, confession). Studies in which Polygraph Association (9) was also searched for judgments of one set of polygraphers are corre­ references to field studies of validity. The 10 stud­ lated with anothers' with no independent criterion ies finally included (and listed in table 3) in the of guilt or innocence are, in effect, reliability analysis are: Barland and Raskin (22), Bersh (29), studies. Such studies have been excluded from the Davidson (47), Horvath (82), Horvath and Reid primary analysis reported here. Reports of unsys­ (84), Hunter and Ash (85), Kleinmuntz and tematically collected cases from police agencies Szucko (92), Raskin (133), Slowick and Buckley and other organizations, in which the criteria for (ISS), and Wicklander and Hunter (205). The fol-

47 48 49

Table 3.-Characteristics of Field Studies McEvoy (116)), or surveYfi?of attitudes towards cover the source of decision errors; these findings validity of the polygraph (e.g., Cureton (44)). are discussed in chapter 6. The Raskin study in­ Type of validity affected cluded in the present analysis (133) was conducted The Bitterman and Marcuse (30) study was ex­ Basis of with only the 16 cases from Barland and Raskin's Examiner cluded because, as pointed out by Lykken (108) Study (22) sample able to be verified by confession. Criterion Decision Types of casesa and Horvath (81), among others, studies of single Bersh Panel of legal profes­ Original examiners' Criminal investigations/military crimes for which there is only one possible guilty sionals' assess­ decisions personnel ment of investiga­ person raises the probability of accurate decep­ Studies Included tive files tion, regardless of method used, to a level too high Barland and Raskin ...... Panel of legal profes­ Original examiners' Sex crimes, drug crimes, crimes of for the study to provide valid information. To The field studies included are listed in table 3 sionals' assess­ decisions violence, crimes of financial gain ment of investiga­ other crimes b ' give an extreme example, if there is one guilty sus­ in terms of the criterion used, the type of initial tive files pect among 100 examined, making an a priori de­ examiner decision, and the types of cases selected. Barland and Raskinc ...... Panel Blind evaluation Sex crimes, drug crimes, crimes of cision to call them all innocent yields a 99-percent These characteristics of studies relate to criterion, violence, crimes of financial gain other crimes ' accuracy rate. In addition, Bitterman and Mar­ construct, and external validity, respectively. Barland and Raskin ...... Judicial outcome Original decision cuse did not meet present criteria for field studies Sex crimes, drug crimes, crimes of The criterion dimension refers to the operation­ violence, crimes of financial gain, because the po1ygraphJrs were not professional alization of ground truth used in a study. In one other crimes examiners (they were psychology professors who Barland and Raskinc...... Judicial outcome Blind evaluation Sex crimes, drug crimes, crimes of type of validity study, polygraphers' original deci­ had read books and articles about the polygraph violence, crimes of financial gain, sions are compared against a criterion of ground other crimes technique), and they did not use field-tested meas­ Raskin...... Confession Blind evaluation truth established by a panel of experts (e.g., law­ Sex crimes, drug crimes, crimes of ures of physiological response. violenc~, crimes of financial gain, yers and judges). The panel makes their judgment other crimes Horvath and Reid ...... Confession Blind evaluation Theft, sexual misconduct, sabotage, Ben-Ishai's (26) paper reports on two studies, on the basis of information in an investigative file, bribery, criminal damage to property both of which were excluded. One consisted of from which polygraph results are excluded. In Hunter and Ash...... Confession Blind evaluation Theft, official misconduct, brutality, sex­ blind evaluations by Ben-Ishai of 10 polygraph another type of field study, a second set of ex­ ual assaults, homicide Siowick and Buckley ...... " Confession Blind evaluation Theft, industrial sabotage, drug abuse, charts. It is more accurately described as a study aminers evaluates charts taken from a file. In most rape of reliability. The other used a single psycholo­ cases, the evaluation is "blind;" i.e., the examin­ Wicklander and Hunter .... " Confession Blind evaluationd Homicide, sexual assault, theft, official gist's (Ben-Ishai's) judgments of guilt or innocence er/evaluator does not know the original examin­ misconduct Horvath...... Confession Blind evaluation Crimes against persons, crimes against based on investigative files as the criterion by er's decision, the disposition of the case, nor any property which to judge polygraph accuracy. It is difficult other information about the subject. In this situa­ Davidson ...... Confession Blind evaluation Crimes against property/military to justify use of the judgment of a single psychol­ tion, the original decisions have been verified by personnel Kleinmuntz and Szucko ...... Confession Blind evaluation Theft ogist as an adequate criterion of ground truth. confession of the guilty party. Verification by con­ ~AII studies use some version of control question technique. Likewise, the information used to ~stablish ground fession is used as the ground truth criterion. In truth for the Elaad, Peters, and Widacki reports the third, and the least common type of field ~~~:YI:~I~~e~21~a:he: ~~~~s~~~~:~~a:~~~ ~~~~u~~~~I~eihe text. Wicklander and Hunter also Included an evaluation In which evaluators were given additional case material. is not systematically collected and is inadequate­ study, original examiners' decisions (the construct ly described. These studies are more accurately validity component) are judged against guilt or described as a set of anecdotal reports. They use innocence established by judicial outcome, which lowing sections briefly describe the studies ex­ (206). One study, Kleinmutz and Szucko (92), not samples of cases collected from police files which is the ground truth criterion. cluded from the analysis and the kinds of studies included by various reviewers (because of its re­ are described as having been verified, sometimes included in the analysis. cent publication) has been included here. In ad­ Researchers disagree about whether blind eval­ by judicial outcome (Widacki), in others by con­ dition, a number of studies included by Abrams uations of polygraph charts or the decisions of fession (Elaad), and in the Edwards study, by (1), not shown in table 2, are also excluded from the original examiners constitute true tests of poly­ Studies Excluded "independent means." the present analysis. Many of the studies Abrams graph validity. Whether one uses examiner deci­ Not all studies refer.red to as field studies or ac­ cited are excluded by later reviewers (e.g., Hor­ A final set of studies excluded are two of the sions or physiological recordings depends on tual criminal investigations by other reviewers are vath (81)) because they are not actual validity three studies by Raskin (133). One analysis was whether one is testing examiner decisionmaking included in the present analysis. A comparison studies (and did not use external criteria of directed primarily at an -assessment of whether or physiological arousal in response to certain of studies shown in table 2 and the 10 studies in­ "guilt/innocence," e.g., MacNitt (113)), they did polygraph examinations are more favorable to de­ questions. Blind evaluations of charts are prob­ cluded in the present analysis indicates that eight not use appropriate polygraphic instrumentation fendants when conducted by polygraph examiners ably less useful as research evidence because, in studies included by one or another of the review­ (e.g., Summers; see Abrams (I)), or did not use chosen by defense attorneys than when they are the typical examination situation, the decision as ers are not included. The excluded studies are Bit­ testing procedures common today (e.g., Lyon conducted by examiners chosen by prosecutors to suspects' deception is made by the original ex­ terman and Marcuse (30), Ben-Ishai (26), two (111)). Other studies used by Abrams, but ex­ (the so-called "friendly polygrapher" hypothesis). aminer and not by a blind evaluator. Even when analyses reported in Raskin (133), Edwards (52), cluded from the present analysis, were unverified The purpose of the second analysis was to dis- examinations are subject to review (e.g., quality E1aad and Schahar (54), Peters (124), and Widacki self-reports published in popular magazines (e.g.,

,. . ' ~r.~..:r'~.

I "I

50 51

control procedures used by the Department of De­ of false positives (in the case of acquittals), or false fense (DOD)), final decisions are still based on negatives (in the case of dismissals). CODING review of all information. Although a blind anal­ ysis is the first task of the quality control office, The use of confessions, the most frequently used In order to conduct the present analysis, each rating of medium was given if numerical scores such quality control reviews do not fully control criterion of ground truth, is problematic in three field study was coded for a number of variables were assigned to subjective assessments of sus­ for the impact of a variety of factors, such as in­ ways: which had either been referred to as important pects' guilt or innocence (see, e.g., 22,92), low if terpersonal expectancy effects which would still 1. confessions, themselves, are not always factors in previous reviews of the literature, or ratings of deceptive or nondeceptive were based be reflected in the original polygraph charts. In­ valid; which were deemed relevant to the various com­ on global assessments of charts only, and very terpersonal expectancy effects (141) refer to the 2. if the confession occurs prior to or during ponents of validity described in chapter 3. If the low if decisions were based on charts plus other possibIlity that an examiner's preexamination de­ a polygraph examination, it cannot be con­ needed information was not available from the available information (in particular, observation and interaction with the subject). Objectivity cision concerning guilt or innocence affects con­ sidered an independent measure of guilt; and studies as published, the study author(s) were con­ ratings were made both for the original examiners' struction of examination questions or the psycho­ 3. those who confess may be a select sample tacted and asked to supply the information. Ap­ logical state of the suspect. Either of these could of subjects, as discussed further below. pendix C lists the coding categories including rele­ judgments and the blind evaluators or judges. vant validity components (panel decision or affect a suspect's physiological responses. There­ Finally, six categories of outcome data from In addition to the above problems, studies dif­ judicial outcomes; confession), as well as design fore, in studies for which results of both original each study were recorded: examinations and blind evaluations were in­ fer in the adequacy of their research design. The information (sample selection, attrition rate, ex­ cluded, as in Barland and Raskin (22), the pres­ most serious problems concern sampling. In most aminer / evaluators' knowledge of base rate of 1. guilty/deceptive subjects judged correctly; ent analysis uses results of the original examina­ reported studies, neither cases, examiners nor guilt). All codings were made by two reviewers 2. guilty/deceptive subjects judged incorrect­ tions instead of those for blind evaluations. It evaluators were selected randomly. In some stud­ and each instance of disagreement over coding ly (i.e., judged nondeceptive); should be noted, however, that in these cases it ies (e.g., 22,84), the cases of only one examiner was resolved before analysis. 3. guilty/deceptive suspects judged inconclu­ are sampled. Nonrandom selection leaves open sive; is difficult to determine to what extent the deci­ Data were coded directly from information pro­ the possibility that the studies are not investigating 4. innocentlnondeceptive subjects judged cor­ sions are based on the charts and to what extent vided within the study report or from informa­ "polygraph testing" in general, but instead only rectly; they are based on interaction with the suspect (see tion directly provided by the authors, with the a subgroup of practitioners or testing techniques. S. innocentlnondeceptive subjects judged in­ 27,92). exception of one variable. The exception was the When random sampling is used (as in Bersh (29)), correctly (i.e., deceptive); and coding category "objectivity of ratings," which Operationalizations of ground truth (the criteri­ high rejection rates of cases selected for analysis 6. innocent/nondeceptive subjects judged in­ required that the coder make a judgment from on component of validity) are also problematic. create other sample bias problems. conclusive. Studies using panel decisions have been referred high objectivity to low objectivity. Scoring was to as the only valid field research on the validity Some sample selectivity of unknown magnitude judged high if some actual standardized measure­ Categories 2 and 5 are the false negative and false of examiners' decisions (81), yet there is no way and importance occurs when confessions are used ment (e.g., using a ruler) was taken of the physi­ positive rates, respectively. to know whether panel decisions based on inves­ as a criterion. Studies using confessions may be ological recordings on the polygraph charts. A tigative files are, in fact, correct. Raskin (136) using only a select sample of examinations. The notes some of the problems with using judicial magnitude of this problem is illustrated by the fact outcomes and other criminal justice system resolu­ that in the sample of 92 cases obtained by Barland tions (dismissals, guilty pleas) as criteria for validi­ (22,133) only 16 were able to be verified by con­ FINDINGS AND DISCUSSION ty. Cases may be dismissed for lack of sufficient fession (132). evidence rather than actual innocence. If a jury Three questions are of particular importance for each category were converted to percentages, To summarize, because of problems in opera­ to an assessment of polygraph validity useful to and average percentages within each category acquits a defendant, it is not possible to determine tionalizing important components of validity, the extent to which the jury felt that the defend­ policymakers: were calculated. A measure of predictive associa­ none of the field studies of validity can be taken tion (lamb dab, see 64,73) was also calculated, ant was actually innocent or whether they felt that 1. Are polygraph examinations valid? by itself as an indication of polygraph testing va­ although the use of a single measure is very limited there was not enough evidence to meet the stand­ 2. Given the wide range of outcomes reported lidity. In addition, because of the different opera­ due to the wide variability in study design. ard of "guilty beyond a reasonable doubt." Many tionalizations of construct and criterion validity across studies, what accounts for their vari­ guilty pleas are actually confessions of guilty to and variations in research design, the studies are ability? The lambdab index shows the proportional re­ (lesser) crimes; as Raskin notes, it is difficult to 3. How generalizable are the results of studies not strictly comparable with each other. These duction in the probability of error in predicting interpret the meaning of such pleadings in regard to the current and proposed uses for national studies, however, constitute the most direct evi­ one category (in this case, deception) when a sec­ to guilt on the original charge. The result is that, security purposes? dence for validity currently available and are ana­ ond category (in this case, polygraph examina­ using criminal justice system outcomes, polygraph lyzed as a group in order to assess the current state examinations may appear to have a high number To answer the first question, data from the tion results) is known. If the information about of knowledge about polygraph testing. available field studies were analyzed to ascertain the second category does not reduce the proba­ whether polygraph examination accurately differ­ bility of error in predicting the first category at entiate deceptive suspects from nondeceptive sub­ all, the index is zero, and one can say that there jects. For this analysis, the outcome frequeT'des is no predictive association. On the other hand, \ I \

52

if the index is 1.00, no error is made in predict­ that almost all of them tested control question ing one category from another, and there is com­ techniques in criminal investigations, they differ plete predictive association. Essentially, lambda in operationalizations of ground truth and type provides an index that translates to the percent of examiner decision. The result is that there is improvement over the base rate and indicates the a great deal of variability in the results of studies. percent improvement in prediction when the poly­ Correct guilty detections range from 70.6 percent graph examinations are considered versus no fur­ in one condition of the Bersh study to 98.6 per­ ther information. There is almost no direct re­ cent in a condition of the Wicklander and Hunter search on the percent improvement of the poly­ study. Correct innocent detections are even more graph over other forms of investigation (cE. 207). variable, ranging from a low of 12.5 percent in The results of this analysis of predictive associa­ the Barland and Raskin judicial outcome study tion are shown in tables 4 and 5. The average to a high of 94.1 percent in one condition of the lambdab across studies is 0.65, which means that, Bersh study. Table 5 also indicates the range of on the average in these field studies, the polygraph incorrect judgments and inconclusives among diagnosis reduced 65 percent of the error of chance studies. False negatives range from 29.4 percent prediction. The lambdab for individual studies of the Bersh study to zero percent. False positives ranged from 0.13 to 0.90. range from 75 percent in Barland and Raskin (22) to zero percent in two studies.lnconclusives range To summarize, the analysis of the 10 field stud­ from zero to 25 percent. This section compares ies included in the analysis indicates that while studies that used comparable operationalizations polygraph examinations using CQT in criminal of construct and criterion validity in an attempt investigations detect deceptiveness and nondecep­ to discover reasons for the range of results. How­ tiveness better than chance, there is also what in ever, even using this method results in consider­ some cases might be considered a high error rate, able variability. The main point, however, is that particularly for nondeceptive subjects. The one no field studies exist to directly test the situations study which tested the validity of the relevant! for which DOD and the President propose to ex­ irrelevant question technique (the general ques­ pand polygraph use. tion test (GQT) portion of the Bersh study) also detected deceptiveness and nondeceptiveness bet­ ter than chance. Studies Using Panel Criterion and Examiners' Decisions Variation Among Studies Both Bersh (29) and Barland and Raskin (22) As implied in the introduction to this section, used a panel to establish the criterion for validi­ the use of a single statistic or summary number ty in their studies. The makeup of the panels and to describe the results of field tests of validity may the polygraph scoring systems were similar in each be misleading. As shown in table 3, although the study. In the Bersh study, which validated poly­ field studies of polygraph validity are similar in graph examinations conducted by military exam­ iners, the panel consisted of four Judge Advocate General (JAG) Attorneys; Barland and Raskin's Table 4.-Mean Detection Rates as a Percentage of panel consisted of two criminal defense attorneys, Total in Field Studies two criminal prosecuting attorneys, and a judge. "Ground truth" The examiners in the Bersh study used either GQT (a version of R/I) or the zone of comparison Examiners or Percent Percent evaluators' guilty innocent (ZOe) technique; for all but one subject in Bar­ diagnosis Mean S. D. Mean S.D. land's study, the Federal ZOC control question Deceptive ...... 49.3 (12.7) 8.2 (7.2) 57.5 technique was used and results evaluated using Nondeceptlve ...... 5.8 (5.1) 32.7 (16.7) 38.5 the Army scoring procedure. Assuming the ac­ Inconclusive ...... 2.0 (3.0) 2.1 (2.5) 4.0 curacy of the panel's decisions, the two studies' 57.1 43.0 100 % results are strikingly different. Barland and Raskin NOTE: lambda = 0.65. b S.D. = standard deViation. attained accuracy rates of 91.5 percent for guilty , "f"t' t

r ->-. ·~:r.

Table 5.-0utcomes of Field Studies of Validity

Guilty Innocent Number Incorrect Number Incorrect Total number of cases Correct (false negative) Inconclusive of cases Correct (false positive) Inconclusive of cases Lambda Bersh (29) (panel of 4) GOT unanimous ••..•.•.••.•••.••••••.• 32 96.9% 3.1% Oa 36 88.9% 11.1 % Oa 68 ZOC unanimous .•••••••••.•••••.•••••..•••.••.•••.•..•••. 38 89.5 10.5 0 51 94.1 5.9 0 89 Average unanimous ••••••••••.••••.•.•.•••..••••••.•..••.• 70 93.2 6.8 --0 87 91.5 8.5 0 157 0.84 Majority (ZOC and GOT) ...... 34 70.6 29.4 0 25 80.0 --20.0 --0 59 0.82 81.9 18.1 --0 85.6 14.3 0 216 Horvath and Reid (84) (1 examiner, 10 examiner/evaluators) .••.•• 20 85.0 15.0 Ob 20 90.5 9.5 Ob 40 0.76 Hunter and Ash (85)c-(1 examiner, 7 examiners/evaluators) ••.•• , 10 87.1 11.4 1.4 10 86.4 14.1 0 20 0.74 Slowlck and Buckley, (155)C(random selection; 7 examiner/evaluators) ..•••...••••••.•.•.•....•.•••••..••. 15 84.0 15.3 0.7 15 90.7 6.6 2.7 30 0.77 Wicklander and Hunter (205)-(2 examlners/6 evaluators) PG+ e •.•••••.••.•.••.•••••••.••••....••.•....••••••••• 10 98.6 1.3 0 10 86.6 8.3 5.0 20 PG ••••.••••••.••.•.•..•.•.•.••••..•••.•••...••.•.••. " 90.0 8.3 1.6 86.6 5.0 8.3 Average ••••..•.•.•••••••.••••••••••.•..•...•..•.•.•... 94.4 5.0 1.0 86.6 6.7 6.7 0.88 Horvath (82)f (10 examiner/evaluators) Verified cases •••••.•.•.•.•.•.•••..•.•...••..•.•.•.•••..• 28 77.1 22.9 og 28 51.1 48.9 og 56 0.28 Davidson (47)h_(random selection 7 examiners/evaluators) ..•••.• 10 90.0 10.0 0 11 9tO 0 9.0 21 0.90 Raskin, d,l ...... 12 4 16 (1 examiner, 25 evaluators) Numerical .••••.•••••••••••..• '" 91.7 0 8.3 75.0 a 25.0 0.75 Nonnumerical •••••••••••••••••••.•.•••••••.•••...•••.•.. 83.3 8.3 8.3 25.0 50.0 25.0 Barland and Raskin (22) (1 examiner, panel of 5) Panell .••••••.••••••••••.••••••..•••.•••••.••••..••..•• 47 91.5 0 8.5 17 29.4 52.9 17.6 64 0.29 Judicial outcome ••..•••••.•••••••••.••••...•••••••.••.•.• 33 90.9 0 9.1 8 12.5 75.0 12.5 41 0.13 Kleinmuntz and Szucko (92)-(6 examiners/evaluators)k ...... 50 75.0 25.0 01 50 63.0 37.0 01 100b 0.38 a Data for Inconcluslves not reported; Inconcluslves appear to total 27 (243 Initial N-216 decisions reported). b examiner/evaluators were not allowed to judge charts as Inconclusive as to overall deceptiveness. In another type of analysis on a question by question basis, ludgments of doubtful or Inconclusive were allowed. ~ Average of two blind chart analyses spaced at least 3 months apart. done by same examiners. Average frequencies divided by number of examiners. ~ PG+ Indicates evaluators had access to written Information In addition to polygraph charts (e.g., case details, SUbject behavior during examinations. etc.). Both PG only and PG+ examinations were done by the same examiners 2 months apart. Excludes Horvath's analysis of 28 unverified cases; because there Is no criterion. this Is a reliability study. ~ There were apparently 15 (1.3 percent) Inconclusive judgments out of 1.120 total Judgments (10 examiners x 112 cases) Which the author excluded from further analyses. Majority decision only. I Seven examiners used numerical scoring; 18 used nonnumerlcal scoring procedures. j Excludes 28 cases for which the panel was unable to come to a decision as to guilt or Innocence. ~ Decisions were based on one polygraph chart: standard practice generally employs at least three. Also. the evaluations were made by students with little polygraph experience. Inconcluslves were not allowed.

NOTE: GOT D general question test.

ZOC D zone of comparison.

\ 54 55

and 29.4 percent for innocent subjects; compa­ unavailable or, when they were available, unable rable figures in Bersh's study are 70.6-percent guil­ to recall the details of a crime. In many cases, few examiners with a median of 3 years experience, Barland (17) has suggested a number of reasons ty correct and 80-percent innocent correct. It is details were available. As a result, one-third of all of whom specialized in conducting polygraph why Horvath's results are lower than the Reid not clear why there should be this variation, al­ the 92 cases were judged inconclusive by the panel examinations for police agencies. The 25 evalua­ studies. One reason is that the blind reviewers did though differences in the nature of the cases, the merely on the basis of the investigative files. The tors in the Raskin (133) study were volunteers who not have access to "special charts" administered completeness of the case files, and sample selec­ figures reported in table 5 are for 64 of the original had trained in a variety of places. in 32 percent of the cases, primarily to subjects tion may account for some of the differences. 92 cases. the original examiner considered deceptive; these The results of the Reid studies do not vary sub­ charts were removed from the files before being In the Bersh study, cases were initially drawn It is not clear why there should be an inverse stantially. The greatest deviation from the mean reviewed by blind examiners. According to Bar­ at random from a pool of criminal investigations relationship between accurate detection of guilty occurred in one condition of the Wicklander and land, Horvath's original examiners had been 100 conducted by the three military services over a and innocent suspects in the two studies. It may Hunter study in which examiner/evaluators were percent correct in their judgments. A second rea­ period of 3 years (1963-66); then, any cases which be that both the panel and the examiner in the given additional information about the suspects son is that, as noted above, police examiners were had been judged "indeterminate" by the original Barland and Raskin study consistently tended to (verbal and nonverbal behavioral indicators, de­ used instead of private examiners; the difference polygraph examiner were eliminated. In addition, presume guilt in the absence of any a priori base mographic information) and the cases. This dif­ between the two kinds of examiners is not ex­ after polygraph charts were removed from the in­ rate (see 28,160). The cases in the Bersh study, ference, however, was net statistically significant. plained further . Yet a third reason, which Barland vestigative files, a preliminary panel of judges on the other hand, were initially selected to be Even so, it may be reasonable to consider it sep­ (17) believes may be the most important in terms eliminated from the sample all files containing in­ equally distributed among deceptive and nonde­ arately from the other Reid studies, because of of false positives, is that a number of victims and sufficient evidence to warrant a positive determi­ ceptive cases. It is not reported whether the panel the extra information available to evaluators. In witnesses were included in the sample (Le., were nation of guilt or innocence. Only those case.;; was aware of the base rate in the Bersh study. the Reid studies, guilty correct identification rates subjects). According to Barland (17), one theory which resulted in a unanimous decision by the ini­ ranged from 84 to 87.1 percent, with an average of detection of deception predicts that innocent tial JAG panel were retained in the validation sam­ of 86.5 percent (excluding the 98.6-percent Wick­ Studies Using Confession as a Criterion victims or witnesses may react emotionally dur­ ple. Altogether, one-quarter of the cases (80 cases lander result; 88.9 percent including it). The in­ and Blind Evaluations ing a polygraph examination because they expe­ out of 323) were eliminated because of insufficient nocent correct rates in the Reid studies range from rienced or witnessed the event regardless of evidence. This figure does not include the number The remainder of the field studies analyzed 86.4 to 90.7 percent with an average of 89 per­ whether they are telling the truth about specific initially eliminated on the basis of inconclusive tested the validity of polygraph testing by com­ cent. There is no difference when the Wicklander polygraph examinations. details of the incident. An analysis of the Hor­ paring the blind evaluations of polygraph exam­ and Hunter condition is included. vath data suggested by Barland, comparing results iners against a criterion of verification by confes­ In Barland and Raskin's (22) study, the initial An additional difference of note among the Reid for victims and witnesses with those for suspects, sion. Two exceptions are Barland and Raskin's ju­ pool of subjects consisted of 102 (nonmilitary) studies concerns the false negative rate, which is would be of interest (see Giesen and Rollison (61) dicial outcome analysis and one condition in the criminal suspects referred to Barland by police, highest in the studies which either used random for a comparison of innocent associations with Wicklander and Hunter study. The confession defense or prosecuting attorneys. These cases rep­ selection of cases (Slowick and Buckley) or elim­ guilty knowledge). resented the entire population of Barland's cases studies vary somewhat as to source of verified inated the most clear-cut charts from their original at that time. Then, 92 of these 102 cases were re­ files. The Horvath and Reid, Hunter and Ash, selection (Horvath and Reid). There is no appar­ Despite the generally anomalous results of Hor­ tained for further analysis on the basis of inde­ Slowick and Buckley, Wicklander and Hunter, ent explanation for the variatiun in false positive vath's (82) study, an interesting finding may help pendence (a case was considered independent and Kleinmuntz and Szucko studies all used files rates in the Reid studies, which ranged from 5 to to account for the results of the Kleinmuntz and from polygraph testing firms. Horvath's cases where two or more subjects had not been exam­ 14.1 percent. Szucko (92) study. Horvath found that suspects ined regarding the same crime). In one respect (the came from police files, Davidson's from military in crimes against property were less detectable The Davidson study results are basically similar fact that there was only one examiner), Barland cases, and Raskin's from the Barland cases re­ than suspects in crimes against persons. This may to those of the Reid studies, except for the absence and Raskin's sample was less variable than ported in Barland and Raskin (22; discussed be because crimes against persons are likely to of false positives. However, the study should be Bersh's. However, Barland and Raskin did not above). The first four studies used files from the have a greater amount of affect associated with interpreted with caution as one-third of the eliminate from consideration indeterminate exam­ firm of John E. Reid & Associates and involved them, and are, thus, more physiologically detect­ originally (randomly) selected sample was not able inations. Neither, and perhaps more important­ various criminal offenses. The firm used by Klein­ able. Barland and Raskin (22), on the other hand, to be used. ly, did Barland and Raskin eliminate cases in muntz and Szucko is not identified; all of their found no differences by type of crime. As noted cases involved theft. which investigative files without the polygraph The Horvath (82) and Kleinmuntz and Szucko previously (see table 3), Kleinmuntz and Szucko's were inadequate. As Barland (17) points out, In the first four studies, blind examiner evalu­ (92) studies have the lowest accuracy rates. As (92) study selectloed only cases from the files of a many of the investigative files that were given to ators also came from John E. Reid & Associates. with the Barland and Raskin (22) study, the low polygraph firm involving crimes of theft. How­ the panel were incomplete. The files had been The Reid studies did vary with respect to case accuracy rate may be related to the fact that Hor­ ever, although the crimes against property hy­ compiled by inexperienced student assistants who selection. Only one study (Slowick and Buckley) vath selected his sample from police files. Perhaps, pothesis is suggestive, it may not fully explain the often did not know where to obtain necessary in­ reports random selection of cases; in other studies, police records of verification are not reliable, or difference between Kleinmuntz and Szucko's and formation. The officials responsible for providing the cases of only one or two examiners were used. have greater variability than those of polygraph similar studies. The Davidson study, for exam­ the information were, more often than not, Horvath's (82) blind evaluators were field-trained firms. ple, only used theft cases, and it has a "0" false

-~ ~- ... --'------"----- .~. \ I I

56 57

positive rate (although it has a substantial incon­ Studies Using Judicial Outcomes clearances for current employees. Background in­ ground investigation techniques. For admissions, clusive rate). Analyses of other studies by crime and Original Examiners' Results vestigations are conducted through the use of per­ for example, the polygraph had an index of 6.59 type would be informative, although the number sonnel interviews, interviews with present and compared to 0.79 for "administrative screening," of cases would probably be too small for a mean­ Barland and Raskin's (22) analysis using judicial former neighbors, checks of educational and work 1.08 for "investigative interviews," and 0.28 for ingful analysis. outcomes as a criterion has the lowest accuracy "papers only." rate for innocent suspects-a 12.5-percent inno­ records, and checks with a consortium of other Szucko (159) has suggested that one possible cent correct and 75-percent false positive rate. The national agencies (the so-called National Agency Several aspects of the study should be noted. reason his results are so different from other problems with using judicial outcomes as a cri­ Check). Of the agencies surveyed, only the Cen­ polygraph firm studies' results, is that the in­ terion have already been referred to, in particular, tral Intelligence Agency (CIA) used the polygraph One is that the criteria for case selection and adverse information are not stated. Another issue, dividual who selected the charts in the Kleinmuntz the fact that the judicial outcome is not a highly to conduct background investigations. and Szucko study could not read polygraph noted by the DCI study authors, is that even accurate measure of guilt because of such char­ In the 4-month period covered by the study, charts. Therefore, case selection may have been though the polygraph is reported as the sole acteristics of the legal system as the nec~ssity for CIA conducted 507 background investigations. Of more variable than in some of the other studies. source in resolving adverse information, it was proof beyond a reasonable doubt, and the prev­ these, adverse information arose concerning 47 Alternative explanations are that: 1) Kleinmuntz only used after a thorough investigation using alence of plea bargaining. These problems are il­ percent of applicants or other individuals being and Szucko only evaluated one chart for each other sources had taken place. For this reason, lustrated here by the fact that only 41 of Barland investigated for security clearances. Thirty-five subject (at least three is standard); and 2) their it is difficult to assess its effectiveness separately and Raskin's original 92 cases were resolved by (83 percent) of the adverse cases were resolved evaluators were examiner-trainees at the end of from the effect of a thorough investigation. Fur­ the criminal justice system. Again, there is clear­ against the individual (Le., the applicant was not their internship period, not experienced examin­ thermore, as a result of being conducted at the ly greater agreement on guilty subjects. hired or clearance was not granted). In two-thirds ers* (see 91). end of a background investigation, in this case of the instances of adverse information resolved the polygraph examinations could be considered against the individual with the use of the poly­ a confrontation technique rather than an investi­ ·Some maintain that the evaluators in Kleinmuntz and Szucko's graph, subjects admitted to the adverse informa­ study were even less experienced than that. gative tool, according to DCI. Agencies surveyed tion. The kinds of issues admitted by subjects had by DCI were asked not to include confrontation primarily to do with drug and alcohol use (e.g., techniques in their responses. A third problem is marijuana use, alcohol abuse, abuse of other that there was no independent verification of the drugs; approximately 55 percent of the cases) and OTHER CONSIDERATIONS cases that were resolved. Perhaps most important, immoral conduct (e.g., sexual deviance; 24 per­ the effectiveness of polygraph examination cases Although the analysis above demonstrates that error rate problems is the observation that field cent of cases). Four cases involved irresponsibili­ involving most, if not all (i.e., irresponsibility) ty, a subcategory of which is violation of secu­ polygraph testing is better than chance at differen­ studies of polygraph testing have only been con­ of the kinds of adverse information uncovered tiating deceptive from nondeceptive subjects in ducted in criminal investigations. As is discussed rity regulations, and none involved the loyalty among applicants in the sample probably cannot criminal investigations, what might be considered more fully in chapter 6, criminal investigations category. It is not clear whether any of the four be generalized to investigations of unauthorized irresponsibility cases involved violations of securi­ as substantial false positive a..'1d false negative rates may generate different levels of affect. In addi­ disclosures. are obtained in several investigations. Although tion, different kinds of subject groups may be the ty regulations. Three of the eighty-four resolved against cases involved admissions of foreign con­ it is not possible to determine a "scientifically'" ac­ focus of expanded Government use of polygraph Edel and Jacoby (51), in a study reported in a nections, meaning in this case either that: ceptable rate of correct or incorrect judgments, testing. Only two field studies can be identified leading psychology journal, tested the reliability clearly if error rates are between 10 and 25 per­ that relate directly to polygraph testing in the na­ 1. the subject was not a U.S. citizen; of polygrapher judgments of physiological respon­ cent, a large number of incorrect decisions would tional security area: one by the Director of Cen­ 2. the subject's spouse was not a citizen; sivity in applicants for positions with "a large be made if the polygraph were widely employed. tral Intelligence (DCI,165; and a second by Edel 3. relatives were potential "hostages;" Government agency." Forty cases were random­ The base rate of guilt in actual situations may fur­ and Jacoby (51). Neither of these is a validity 4. alien relatives, "hostage" unlikely; or ly selected from the agency's applicants in 1966. ther complicate matters. It is not clear from the study but because they are the only field studies 5. life abroad cannot be verified. Ten practicing polygraph examiners acted as ac­ field studies conducted so far how many suspects with any relevance to national security, they will tual examiners in four cases each and raters in were involved in the cases selected for polygraph be described below in some detail. An analog The seriousness of the wrongdoings was not clear. eight additional cases. In each case, examiners testing, but if there were a large number of sus­ study of counterintelligence screening (16) is dis­ The crux of the DCI analysis was the construc­ (raters) judged three physiological responses to pects, more false positives could be expected (see cussed in chapter 5. tion of a productivity index for investigative tech­ each interview question as either "no specific reac­ ch.7). The DCI study consisted of a survey of 12 Gov­ niques from the CIA data and data from other tion" or "a specific physiological reaction." The Also problematic is the wide variability in ac­ ernment agencies (not including the National Se­ agencies. Based on the fact that a large number rate of agreement between examiners and raters curacy rates across studies. Although some dif­ curity Agency (NSA)). The study was conducted of cases were resolved against individuals by ad­ as to whether a physiological reaction took place ferences can be explained m~thodologically, other to evaluate the relative effectiveness of various mission, and the polygraph was the "unique averaged 96 percent. differences cannot. Of perhaps even greater im­ means of conducting background investigations source" (165) in all the CIA cases resolved against portance than the accuracy rate variability and for purposes of applicant screening and security the subject, DCI tentatively concluded that the Of course, as the authors note, demonstrating polygraph was the most productive of all back- consistency among examiners "is not equivalent "

:,!I I, ',' 58 f to demonstrating consistency in interpretations ical reactions indicate deception among applicants \) based on these physiological reactions" (51). For for positions in Government agencies has not been example, responses were not differentiated for rel­ tested. Because of the potential adverse conse­ "\ Ii,I evant v. irrelevant questions. Therefore, although quences for employment applicants (particularly 'I It Edel and Jacob's study indicates that the examiners in Government agencies where there is interagen­ II I, in the Government agency can reliably detect cy checking (see, e.g., 165)), such tests have sub­ Ii I, physiological reactions, whether these physiolog- stantial practical significance. II f ~ /1 II CONCLUSIONS ~f Chapter 5 Ii Although there is some evidence from available The following chapter reports on the effective­ field studies that polygraph testing is effective in ness of polygraph testing demonstrated by analog Revi,w and Analysis 0,' detecting deception by guilty criminal suspects, studies. As will be shown, the construct and cri­ there is also what in some cases might be regarded terion components of validity are stronger in ana­ as a substantial error rate. This is particularly so log studies, but because of problems with exter­ POlygr~ph Analog Studies for innocent subjects. There appears, as yet, to nal validity, they do not provide evidence about II be no scientific field evidence that polygraph ex­ actual polygraph testing that is as direct as that II aminations can be effectively used to investigate from field studies. Nevertheless, reviewing such IIII unauthorized disclosures or that they represent evidence is necessary to assess both the present Ii I' a valid test to prescreen or periodically screen and potential use of polygraph testing. ;t Government employees. Results of field studies ,I 'I are subject to additional problems of interpreta­ I tion because of inadequate measures of ground o truth. II! I I iI ! 1 j I I i I I I I

h t

! II 1 I, i ~-"'~.'...... ~---....,....,.. ... ~-.~-. --~ --_ .._------~~-

Chapter 5 -.. -1 Review and Analysis of Polygraph Analog Studies

INTRODUCTION Analog studies, for purposes of the present studies is presented. As with the field studies, the analysis, are investigations in which field methods studies concern the use of polygraph examinations of polygraph examinations are used in simulated for investigation of crimes. The two exceptions criminal or other situations. Such studies inves­ (16,43) use analogs to the type of relevant/irrele­ tigate either "mock" crimes set up by an experi­ vant (R/I) question technique typically used in the menter (with the knowledge and collaboration of personnel screening situation. subjects) or actual small crimes "induced" by the The present chapter is organized. as follows: experimenter. Such analog studies are not actual first, the characteristics of analog studies and the criminal investigations and subjects are usually varieties of ways in which they differ from field aware that they are participants in polygraph re­ studies are discussed. Then, the criteria used for search. Analog studies differ from other labora­ includLng studies in the analysis are described. The tory studies of polygraph testing in that they sim­ coding procedure, which is essentially the same ulate actual field examinations. However, in ana­ as that used to code the field studies, is described log studies, typical components of field examina­ briefly. Analog studies of the control question tions are replicated to the extent it is possible to technique (CQT), guilty knowledge technique do so. Such studies test the validity of various (GKT), and personnel screening examination are polygraph techniques under controlled conditions. then reviewed. The findings of a statistical anal­ In chapter 4, the results of a systematic review ysis of the analog studies complete the chapter. of field studies of validity were presented. In the present chapter, a similar analysis of analog

CHARACTERISTICS OF ANALOG STUDIES

The "crimes" utilized in analog studies in order The consequences of failing a polygraph exam­ to establish ground truth have taken different ination (e.g., a possible prison sentence) cannot forms. For the most part, they are "mock crimes;" be replicated in the laboratory. In analog studies, i.e., crimes in which subjects know they are "role punishment takes such forms as losing the chance playing" at being criminals for purposes of an ex­ for a monetary reward. Some researchers have periment. Mock crime studies may be further dif­ experimented with other punishments such as elec­ ferentiated by whether or not the experimenter tric shock (105) or the threat of shocks (35). The controls the guilt or innocence of research par­ analog studies that use real crimes provide another ticipants. In some studies, subjects know that the alternative, in that subjects can be threatened with crime is part of the experimental situation but they real punishment (e.g., academic sanctions for are more or less free to go through with the crime cheating on an examination). In still other cases, or not. Two analog studies have utilized actual subjects are led to believe that "stable" individuals small crimes. In these studies, apparently real sit­ can avoid detection. u.ations were embedded in an experimental situa­ tion in which subjects were given an opportuni­ Analog studies represent, thus, a "tradeoff" to ty to commit a crime or not. the investigator interested in polygraph testing

61 f ,. fiIli(*5' "t"t' --~ ~ ------

62

validity. On the one hand, because the researcher training is as polygraph examiners and who are sets up the crime, ground truth is known; and be­ usually experienced. This would suggest that field cause "ground truth" is established, analog studies examinations may be more accurate. are superior to field studies in terms of criterion The characteristics of subjects who participate validity. Furthermore, they provide the investi­ in analog studies also vary from subjects in field gator with more control of the polygraph situa­ studies. Several use college students, others recruit tion and conditions of testing. The experimenter community members through the newspaper, one can select particular subject groups, can standard­ uses police candidates, and another prison in­ ize testing procedures for all subjects, and can sys­ mates. In many studies, subjects are probably bet­ tematically vary guilt or innocence. With this con­ ter educated and more highly socialized than the trol, the experimenter can also directly compare average field examinee. In the case of student sub­ the effects of variations in polygraph techniques, jects, they are probably younger on the average physiological measures, information given to sub­ and from a higher social class as well. Raskin (132) jects, and scoring methods. notes that analog studies using students yield a On the other hand, although analog studies lower accuracy rate than other studies. As will have greater criterion validity and offer greater be discussed below, this may be due to subject experimental control, their use as indicators 0f differences between field and analog studies be­ polygraph testing validity is potenti~lly problem­ cause a realistic fear of failure does not playa cen­ atic. The reasons have to do primarily with ex­ tral role for subjects. The consequences of failure ternal validity (20,136; see, also, 1,7,108); i.e., for analog studies are usually minimal in contrast the crime situation differs, the testing situations to typical field investigations. in the field and the laboratory differ, the train­ ing of the examiners differs, the subject popula­ tion differs, and, apparently most important, the Study Selection consequences for "suspects" differ dramatically be­ For present purposes, studies were only in­ tween the field and the laboratory. In addition, cluded as analog for the primary analyses if they in analog studies, the questions and question tech­ employed actual field polygraph techniques to de­ niques most often are not tailored to individual tect deception or concealed information, and if subjects. In actual criminal field investigations, the studies pertained to some use of polygraph case information about the crime and the subject testing in the real world. The studies selected are usually provides a basis for tailoring questions. listed in tables 6 and 7. Studies of components Numerous specific differences can be noted. of the polygraph examinations, such as studies Perhaps most importantly, the laboratory crime which used only card tests (97,101), number tests and the consequences of detection are much less (120), or tests concerning concealed personal in­ serious. In addition, in an analog study, demand formation (e.g., parents' first name; see, e.g., 106) characteristics (which suggest to the subject de­ were not included. sirable responses) may create a somewhat different In addition, studies were excluded because their polygraph situation than found in typical field sit­ primary focus was on a theoretical factor thought uations (20). In terms of factors that may increase to affect validity, such as variability in physio­ validity of analog studies, there is some evidence logical recordings (45), nonstandard means of in­ that laboratory researchers are, in general, able terpreting such recordings (163), or the role of '1y­ to use more sophisticated and stable equipment ing" (96). Such studies will be referred to as lab­ than portable machines often used in the field oratory investigations and are distinguished from (136). On the other hand, examinations in analog analog studies. studies are often conducted by researchers who are primarily psychophysiologists (e.g., 49) or Analog studies of the guilty knowledge test psychologists (43) with only limited training in (GKT) have been included, although analyzed .. field techniques. Field examinations, in contrast, separately, because this form of the polygraph ex­ are conducted by individuals whose primary amination represents an alternative proposed for r

Table 6.-0utcomes of Control Question Analog Studies of Validity

Guilty Innocent Number Incorrect Number Incorrect Total number of subjects Correct (False negative) Inconclusive of subjects Correct (False negative) of subjects Inconclusive Lambdab Barland and Raskin (21) ••• 36 63.9% 8.3% 27.8% 36 41.7% 16.7% 41.7% 72 0.47 Podlesny and Raskin (127) 20 69.0 16.0 15.0 20 91.0 4.0 5.0 40 0.75 Raskin and Hare (137) ••.• 24 87.5 0 12.5 24 75.0 4.0 20.8 48 0.83 Rovner, et al.a (143) •••..• 36 77.8 8.3 13.9 36 SO.5 13.9 5.5 72 0.72 Kircher (89a) ...... 50 60.0 4.0 36.0 50 76.0 2.0 22.0 100 0.72 Dawson (49) •.••••••••.•• 12 91.7 0 8.3 12 58.3 25.0 16.7 24 0.33 Wldackl and Horvathb,1 k (207) ••••••.•••.•••••• 20 90.0 5.0 5.0 20 - Bradley and Janlsse (35) •• 96 96 192 0.33 EDR •••••••.•••••••••. 60.4 13.5 26.0 58.3 9.4 32.3 Heart rate ••.••••••••.• 35.4 20.8 43.8 33.3 19.8 46.9 Szucko and KlelnmuntzC (160) ••••••••••.••••.• 15 71.3 28.7 d 15 49.3 50.7 d 30 0.22 Glnton, et al. (62) •••••••• 2 100.0 0 0 13 84.6 15.4 0 15 O.ooj Honts and Hodes (75)e: No countermeasures ••• 9 67.0 0.0 33.0 12 33.0 17.0 50.0 0.25 f Countermeasures •••••• --19 58.0 5.5 36.6 -g 28 12 40 Honts and Hodes (76): No countermeasures ••. 19 84.2 0.0 15.8 19 31.6 15.8 52.6 0.53 Countermeasures •••••• 19 36.8 26.3 36.8 -g 38 19 57 Heckel, et al. (74)1: k Normals •••••••••.•.•• - h 5 100.0 0.0 0.0 Nondeluslonal psychiatric •••••••••• _h 5 70.0 10.0 20.0 Delusional psychiatric .••..••••• h 5 45.0 35.0 20.0 15 Hammond (74a) •••••••••• 32 71.9 3.0 25.0 30 40.0 20.0 40.0 62 a Summed across conditions. b Examiner's task was to detect the one guilty person In each of 20 groups of four suspects. ~ Based on ratings of 5' on a 1 to 8 scale of certainty of nondeceptlon/deceptlon. Examiners were not allowed to categorize an examination as Inconclusive. e Original subject assignments, 12 to each of 4 (Including 2 countermeasure) conditions. A total of five countermeasure subjects were eliminated from the analysis of results for guilty subjects for failure te follow f countermeasure Instructions. Three no countermeasure subjects were eliminated for spontaneously using countermeasures. Nine guilty subjects used pain countermea:lures (tongue biting) and 10 used a muscle (toe pressing) countermeasure. ~ Innocent subjects used no countermeasures. There was no guilty condition. I Not Included In analysis reported In table 8. l Lambda Is a poor statistic when the base rate Is skewed. Lamba was not calculated when only guilty or Innocent subjects were used.

\ ------~------r ... M( .W ''''t'''"*"

r

Table 7.-0utcomes of Guilty Knowledge Analog Studies of Validity

Guilty Innocent Number Incorrect Number Incorrect Total number of subjects Correct (false negative) Inconclusive of subjects Correct (false positive) Inconclusive of subjects Lambdab Lykkena (105) •..••.•••..• 50 88.0% 12.0% 48 100.0% 0% 98 - c Davidson (46) ...... 12 91.7 8.3 36 100.0 0 48 0.92 Podlesny and Raskin (127) ..•..•.•••••..•.. 10 80.0 20.0 0 10 80.0 0 20.0 20 0.80 Balloun and Holmes (12) .. 18 34 Test 1 ...•.•...... •. 61.1 38.9 87.5 12.5. 0.44 Test 2 .••..•....•..•.. 16.7 83.3 93.7 6.3 Giesen and Rollison (61) •• 20 95.0 5.0 20 100.0 0 40 0.95 Bradley and Janlsse (35) •• 96 96 192 EDR .••..•...•...... 59.4 40.6 88.5 11.5 0.38 Heart rate ..•.•...... 44.8 55.2 82.3 17.7 Tlmm (163) ..••.••.•....• 237 237 Liberal cutoff •.•••.••.• 80.8 19.2 - b - d Conservative cutoff .... 70.4 29.6 ~requencles for detection of two mock crimes were combined. -There were no Innocent subjects. ~Lambda cannot be calculated because crimes were not reported separately. Lambda cannot be calculated with only one condition.

\ 65 use in the field (92,107,108), even though it has Table B.-Mean Detection Rates as a Percentage of not been put into general practice. Total In Analog Studies of Control Question Technique Ground truth Description of Studies Percent Percent guilty innocent The following sections discuss each of the ana­ Examiners' diagnosis Mean Mean log studies organized into three categories accord­ Deceptive ...... 33.0 6.8 ing to questioning technique. The discussion of Nondeceptive ...... 5.4 27.9 CQT analog studies is first. Studies of CQT rep­ inconclusive ...... 13.4 13.5 resent available studies, much like the case for 51.8 48.2 NOTE: lambda = 0.43. field investigations (see ch. 4). Six studies of the b concealed information or GKT and two of R/I follow. In only one study (16), involving the R/I Table 9.--Mean Detection Rates as a Percentage of technique, were subjects Government employees. Total in Analog Studies of Guilty Knowledge Test The results of individual studies are summarized in tables 6 (CQT) and 7 (GKT). The description Ground truth of the studies is followed by a systematic statisti­ Percent Percent cal analysis of the results of the CQT and GKT guilty innocent studies. The RII studies were not analyzed as a Examiners' diagnosis Mean Mean group because of the paucity of studies. Guilty ...... , ..... , ...... 27.9 2.2 Not guilty ...... '" .. , ...... 17.3 52.6 Essentially, as shown in tables 6 to 9 the anal­ Inconclusive ...... o o ysis of the analog studies yields conclusions sim­ 45.2 54.8 NOTE: lambda = 0.70. ilar to those of the field study analysis-Le., al­ b though there is a greater-than-chance probability of detecting deceptive and nondeceptive subjects, have an effect on validity. For example, Barland there is what might be regarded as a significant and Raskin (22) examined the effect on validity error rate, and a great deal of variation across of different types of feedback about the poly­ studies. However, as has been found in some re­ graph, and Dawson (49) investigated the effects views (1,7), analog studies of CQT had lower ac­ of countermeasures. These factors are examined curacy rates than field studies of CQT. more systematically in chapter 6; the emphasis of In the studies detailed below, some experiments the present chapter is on the validity of different also tested the effect of factors hypothesized to forms of polygraph examinations.

CONTROL QUESTION TECHNIQUE

Fourteen analog studies of the control question graph. A description of these studies follows dis­ technique were located. The largest group of these cussion of the University of Utah studies. studies emanate from the research program of Professor David C. Raskin at the University of University of Utah Studies Utah. The remaining studies were conducted at a number of settings in the United States and Despite longstanding controversy about poly­ elsewhere. Raskin and colleagues have conducted graph validity, the first research project conduct­ a systematic analog research program, and these ing an analog study that simulated field polygraph studies are described as a group. Other researchers techniques was not conducted until the 1970's have published individual studies testing specific (136). It was then that an ongoing research pro­ hypotheses relevant to the validity of the poly- gram headed by Professor Raskin at the Univer------~

66 67

sity of Utah began to study the validity of the tions. They also tested the accuracy of behavioral sions made by the independent blind evaluator half for low levels. One purpose of their study polygraph through analog experiments. In addi­ observations of the examinee (80,139) in detect­ to assess the validity of the polygraph test. This was to investigate whether deceptive psychopaths tion, these studies also examined the relationship ing deception, since this type of information is was, however, equivalent to using the polygraph could more easily escape detection than normal to validity of different polygraph techniques (e.g.,' used in many field examinations and could pos­ examiner's decision, because the independent rater subjects (see ch. 6). Overall, assessments of decep­ the stimulation test), different physiological meas­ sibly affect the validity of the technique (107,108). and the examiner agreed on 100 percent of their tion from the field evaluations from all charts were ures, different methods of assessing the results They compared as well the capability of different decisions. 88 percent correct, 4 percent wrong, and 8 per­ different types of information provided to sub~ cent inconclusive. There were only two errors, physiological measures in differentiating between The results for both types of control questions jects, and different subject and situation factors both false positives. No significant differences guilty and innocent subjects. A GKT was also combined (with an inconclusive zone used) were that could potentially affect polygraph validity. were found between psychopaths and nonpsycho­ conducted with 20 subjects (see below). 80 percent correct, 10 percent incorrect, and 10 The experiments conducted by Raskin and col­ paths, suggesting that a CQT polygraph examina­ In Podlesny and Raskin's study, subjects were percent inconclusive. There were three false neg­ leagues use similar procedures to set up the mock tion is equally valid for both. Also, a quantitative community members who responded to news­ atives (8 percent) and one false positive (2 per­ crime and to conduct polygraph testing. In each analysis showed that all the physiologj"al meas­ paper advertisements. The experimenters drew cent). The accuracy of CQT differed depending of their studies, subjects are randomly assigned ures were significantly different between guilty from the Reid method in t.heir design of the pretest on whether exclusive or inclusive control ques­ to an "innocent" condition or to a "guilty" con­ and innocent subjects. Psychopathy did not ob­ interview (see ch. 2). One experimenter asked the tions were used. When the exclusive control ques­ dition. The mock crime is the theft of a small scure these differences and in some cases enhanced subjects three questions from Reid and Horvath's tions were used, 85 percent of the subjects were amount of money or a valuable object from a desk them. structured pretest interview designed to provoke identified correctly, 5 percent incorrectly, and 10 in a nearby room. To increase their motivation the subjects into displaying "behavioral symp­ percent inconclusively. Of the assessments of the subjects are offered a financial bonus for convinc~ toms" of deception (80,139). 20 subjects in this condition, one (5 percent) was Rovner, Raskin, and Kircher ing the examiner they are innocent. In the testing a false negative and there were no false positives. the examiner employs the Federal zone of com­ During the polygraph examination they in­ Rovner, Raskin, and Kircher (143) studied the When nonexclusive control questions were used, effect of information and practice on the accuracy parison (ZOC) control question technique, includ­ cluded two special types of control questions 75 percent were correct, 15 percent incorrect, and ing a pretest interview. A numerical field scoring among the set of questions asked of the subjects. of polygraph examinations. Seventy-two subjects 10 percent inconclusive. Of these 20, two (10 per­ recruited from the community took part in this method developed by the Utah group (21) is used One was a "guilt complex question," which asked cent) were false negatives and one (5 percent) was to make the diagnosis of truthfulness or deception. the subject if he committed a fictitious crim~ of mock crime experiment. One third of the subjects a false positive. The evaluative scores for each (12 innocent and 12 guilty) were given in-depth the same nature as the real crime. In this study, physiological component were analyzed to deter­ Barland and Raskin information about the polygraph and about coun­ the guilt complex question was, "Did you take mine if the scores differed between guilty and in­ that watch from room 7027" (127). There was, termeasures used to appear innocent (information In the initial analog study using CQT (21), 72 nocent subjects. Only the EDR and plethysmo­ of course, no watch stolen from room 702. The condition). Another third received this informa­ student "guilty" and "innocent" volunteers were graph scores yielded significant differences. experimenters also varied the wording on some tion and underwent two practice polygraph ex­ randomly assigned to one of three "feedback" con­ of the control questions, so that half the subjects Behavioral observations, by themselves, aminations about which they received "feedback" ditions. The positive feedback subjects were in­ received "nonexclusive" and half "exclusive" con­ yielded a significant number of correct decisions, (information and practice condition). The other structed that the polygraph was effective, the neg­ trol questions. but this differed greatly between innocent and third had no such intervention (standard). A blind ative feedback students were told that the machine guilty subjects. Of the guilty subjects, 86 percent field evaluation performed some time later pro­ was not working properly, and the other students In the pretest interview, the examiners followed of the decisions made were correct (25 deceptive, duced the scores for decisions of guilt or inno­ received no feedback. Subjects then underwent the usual field procedure of reviewing the con­ 4 nondeceptive, and 1 inconclusive); however, cence, and for analysis of the physiological re­ a complete polygraph examination including a trol questions with the subjects, and the questions only 48 percent of the innocent subjects were cor­ sponses. Accuracy for the standard group and the pretest interview. The Federal version of the ZOC were adjusted until they elicited a "no" response. rectly identified (12 deceptive, 11 nondeceptive, information group was identical: 88 percent cor­ technique was employed, with standard control The control question polygraph test then took 7 inconclusive). An objective quantitative analysis rect, 4 percent incorrect, and 8 percent inconclu­ questions used for all subjects. On average, the place, with three or more charts obtained from for each physiological measure was employed to sive. But accuracy for the information and prac­ CQT identified 53 percent of all subjects correct­ each subject, although only the first three were determine if each was effective in discriminating tice condition was lower: 62.5 percent correct, 25 ly. Twelve percent were identified incorrectly and used in the objective scoring. Immediately after between guilty and innocent. Most of the meas­ percent incorrect, and 12.5 percent inconclusive. 35 percent of the examinations were inconclusive. testing, the first three charts obtained were scored ures yielded significant discriminations, with the There was one error in the standard group and Of the errors, three (4 percent of the entire sam­ blind on electrodermal response (EDR), respira­ exception of a few of the cardiovascular measures. one in the information group-both false posi­ ple) were false negatives and six (8 percent) were tion, and cardio measures. Later, an independ­ tives. The six errors in the information and prac­ false positives. ent rater scored the tests, using the numerical scor­ Raskin and Hare tice conditions were three false positives and three ing system devised by Barland and Raskin (21). false negatives. Podlesny and Raskin The experimenters also used objective measure- A special population, prisoners, especially rele- ments of all physiological response measures with vant to the field use of the polygraph, was the Podlesny and Raskin (127) conducted a more the aid of computers and persons who had no focus of a study by Raskin and Hare (137). In their Kircher extensive experiment to examine the accuracy of knowledge of the field evaluations or treatments sample of 48 inmates of a Canadian prison, half Some of the latest work of th7 Utah labora~ory CQT using three different types of control ques- administered. The experimenters used the of'd- were selected for high level~ of psychopathy, an~ ~plore_s !he u~.~ ()f computers In t~e anaJysls of ,------~------~------~~.~~~~~~------~ 68 69 polygraph recordings. Kircher (91a) compared the questioning: while subjects listened to questions ures along with the knowledge of the control ques­ Because of its experimental design that had the accuracy of a computer decisionmaking process and while they responded. tion or other technique. examiner make decisions on four suspects as a to the accuracy of assessments of a field examiner. Dawson's sample consisted of 24 student actors, group, the study produces data about the accu­ The computerized analysis cannot be included in half of whom were randomly assigned to the "guil­ racy of the polygraph that is difficult to interpret. the statistical analysis of this technical memoran­ ty" group and half to the "innocent" group. They Widacki and Horvath But it does shed light on the efficacy of the poly­ dum, because it is not presently a field scoring graph relative to other investigatory techniques were instructed to use the Stanislavsky method Widacki and Horvath (207) designed an experi­ method, but the decisions of an independent that might be the alternative. Certainly, it is to appear innocent on the polygraph examination. ment to examine the polygraph's efficacy in com­ evaluator who was used can be. This mock crime crucial in policymaking to judge the validity of After the mock crime, four charts were obtained parison to other techniques in the mock investiga­ study followed the basic procedures of Podlesny the polygraph relative to other techniques that from ZOC control question test about the crime. tion of a mock crime. They recruited 80 Polish and Raskin (127) with 100 subjects from the com­ would be used in its stead. More research is On two of the charts, the subjects were instructed student volunteers and had all of them provide munity. The accuracy of the original examiner needed in which the polygraph is compared to not to respond until they received a signal 8 sec­ writing specimens, photographs of themselves, was not reported though the results of an inde­ other investigatory techniques, and the quality of onds after a question. This served to separate re­ and fingerprints. Subjects were then assigned to pendent evaluator were. The independent evalu­ information across techniques is held constant. sponding associated with the questions from re­ 20 groups of four subjects each. \tVithin each ator, who numerically scored the charts blindly, Such a comparative analog study would be espe­ sponding associated with answering. Numerical group, one subject was randomly assigned to be correctly diagnosed 87 percent of the subjects; cially valuable if it included different techniques scoring based on Barland and Raskin's (21) system the perpetrator, and the other three were innocent misdiagnosed 6 percent; and made a judgment of used in investigations of Federal personnel, such was done separately on three different types of suspects. Each group was thus an "investigative inconclusive on 7 percent. The six errors were as those reported in the Director of Central In­ physiological responses: case." Because of this feature of the design, the evenly divided between three false negatives and telligence (DCI) survey mentioned in chapter 4 decisions of guilty and innocent were not inde­ three false positives. In comparison, different 1. responses when the answers were immediate; (" administrative screening," "investigative inter- pendent. Therefore, Widacki and Horvath's find­ computer decision models, on the average, cor­ 2. responses during the questions when the VIews,. "t)e c .. rectly identified 84.9 percent of subjects, misiden­ answers were delayed; and ings could not be included in the statistical anal­ ysis of the control question analogs and must be tified 7.85 percent, and placed 7.2 percent in an 3. responses during the answers when the Bradley and Janisse inconclusive category. answers were delayed. considered separately. A similar situation holds for Kubis' (93) mock crime experiment (see be­ Bradley and Janisse (35) studied the effects of Dawson found that the subjects' immediate low). two other variables hypothesized to influence the Other Studies physiological responses to the questions, whether validity of the polygraph: the degree of threat in­ they were answering immediately or not, led to The mock crime proceeded as follows: the guil­ volved in the punishment for being judged guil­ A range of other studies has been conducted decisiuns which were 88 percent correct, 8 per­ ty subject picked up a parcel from one of two per­ ty, and successful demonstration to the subjects in recent years to evaluate aspects of polygraph cent incorrect, and 4 percent inconclusive (fre­ sons acting as a "doorkeeper" of a building in the of the technique's accuracy. A mock crime was test validity. Such studies usually manipulate one quencies across two conditions were summed). area. The perpetrator gave some experiment­ carried out using procedures similar to those used or two variables that are hypothesized to be im­ The delayed answer response yielded a rate of 29 related papers to the doorkeeper and then signed by Barland and Raskin (21). Subjects were also portant determinants of polygraph validity. For percent correct, 8 percent incorrect, and 62 per­ for the parcel. Thus, an eyewitness account (by given a series of stimulation tests. Results of these the most part, these experiments use procedures cent inconclusive, The incorrect decisions made the doorkeeper), fingerprints, and handwriting tests were manipulated such that they made the that are similar to Raskin's mock crime paradigm. were entirely false positives. A quantitative anal­ specimens were all available. Blind polygraph ex­ polygraph test appear perfectly effective, partially Some of the discussion of the procedures in each ysis revealed that the EDR and cardiovascular aminations then were conducted using the Reid effective, and ineffective. In addition, half the sub­ study is omitted, because they closely follow this measures differentiated significantly between in­ control question method (including the examiners' jects were told they would receive a painful elec­ paradigm. nocent and guilty, but respiration did not. The behavioral observations of the subject). Analysis tric shock if found guilty, though no shock was major outcomes of this study suggested that the of the three other sources of evidence was car­ ever given. Dawson polygraph was not susceptible to cognitive coun­ ried out. termeasures of the sort used by the actors and that The degree of manipulated effectiveness had no Dawson (49), for example, focused on the ef­ direct effect on scores, but did tend to increase scorable responses generally occur immediately Widacki and Horvath found that the polygraph fect of "cognitive countermeasures" on validity. the accuracy of detection. Threat of punishment after questions. produced the most correct decisions (n = 18), the His study was unique in that the subjects were did not affect accuracy of detection, although it fewest (along with handwriting) incorrect deci­ actors trained in the Stanislavsky method of acct­ This experiment does not, however, test cog­ did have an overall effect on heart rate. EDR and sions (n=l), and the fewest inconclusive decisions ing, which teaches actors to use their own expe­ nitive countermeasures in a situation in which the heart rate change were significantly accurate in (n =1). Widacki and Horvath note, however, that rience to create emotional states appropriate for subjects know the essentials of CQT and apply differentiating guilty and innocent, although a direct comparison of these four investigative a role. Studying the attempts of "method" actors cognitive countermeasures differentially to rele­ another measure, pupil size change, was not. to foil the polygraph may help determine whether vant and control questions. The average criminal methods may be invalid because the experimen­ tal procedures could not ensure a comparable level guilty subjects can be trained to use cognitive subject is likely to attempt cognitive measures Honts and Hodes countermeasures to appear innocent (see ch. 6). naively, but a sophisticated subject-perhaps the of quality of evidence for each method (e.g., fin­ Dawson was also interested in analyzing separate­ type more likely to appear in a national security gerprints were not detectable in the majority of Two recent analog studies of the Backster ZOC ly responses during two distinct phases of the investigation-may learn cognitive countermeas- cases). method of testing (76,77) were conducted primari- I.·~· \ I

70 71

ly for the purpose of testing whether polygraph percent false positive rate. Examiners were not jects actually underwent the polygraph testing, jects, and 3) student examiners would not be as examiners could detect the use of physical coun­ able to detect the use of countermeasures. only two of whom had actually cheated. . accurate as an expert examiner. He was also in­ termeasures by subjects. In the first study, sub­ terested in the overall value of polygraph ex­ jects were college students who received extra In their second experiment on countermeasures, A CQT was administered, and each subject was aminations for forensic psychology. The subjects credit toward their final grades for their participa­ Honts and Hodes used approximately the same evaluated by three polygraph experts who had in Hammond's study were volunteers solicited tion. "Guilty" suspects participated in a mock procedures and subject pool, with the exception conducted or witnessed the particular examina­ through sign-up sheets in a col1~ge fraternity (nor­ crime (theft of an examination); innocent suspects that subjects were asked to employ both coun­ tion being evaluated. One examiner (an observer) mals), alcoholism treatment centers (alcoholics), were only told of the theft. All subjects were moti­ termeasures simultaneously, were given 30 min­ relied on behavioral observation, another (a rater) and ex-offender programs (psychopaths) as well vated to produce truthful outcomes on the poly­ utes of training, including a practice session, and used only the charts, and a third (the actual ex­ as through newspaper advertisements and other graph test by an offer of twice the number of cred­ were asked to practice at home. A cardio cuff was aminer) used both sources of information. The means. Psychological tests (e.g., subscales of the its if the examiner reported them as truthful. added to the polygraph instrument, and a card evaluations were made globally. Five other poly­ MMPI) as well as polygraph examinations were test was conducted prior to the administration of graph examiners evaluated the charts later using given to the subjects. The polygraph examinations In addition to participation in the mock crime, the first test. both the Utah group's scoring system (21) and 24 of the guilty subjects participated in IS-minute were conducted by students near the end of their global evaluations. The original three perfor:ned training sessions in which they were told about training at the Backster School of Lie Detection. Overalt results of the second study replicated a second analysis in this way, too. ConcluslOns the theory of CQT and shown how to use either Examiners used a version of Backster's control the first. The categorizations of the original ex­ about this study are limited because of a large no­ tongue biting (12 subjects) or toe pressing (12 sub­ question technique, and Backster's numerical scor­ aminer were 51 percent correct, 14 percent incor­ show rate among the guilty subjects. Both guilty ing system. Charts were scored using several levels jects) as countermeasures during presentation of rect, and 35 percent inconclusive. Twenty-six per­ subjects who took the test were correctly detected. of inconclusive zone by both the student exam­ the control questions. They were also instructed cent of the countermeasure subjects compared to However 15 percent of the noncheaters were in­ to try to relax as much as possible during presen­ iners and an expert examiner who scored the none of the guilty/no countermeasure subjects correctly identified as deceptive. tation of the relevant questions. charts blindly. Two polygraph charts, rather than were incorrectly classified as truthful. Examiners the standared three, were conducted for each sub­ were not able to detect successful countermeasure The actual polygraph examinations took place Heckel, et al. ject. a week after the theft and training sessions. All users. guilty subjects were instructed to have the "stolen" Another analog study (74) used a staged crime Table 6 shows the results of Hammond's study to investigate the differential accuracy of CQT using the standard +8 inconclusive zone. As examinations with them, presumably to enhance Ginton, et al. subject involvement. Four charts were obtained with psychotic, neurotic, and normal subjects. Fif­ shown, approximately 72 percent of the guilty subjects and 40 percent of the innocent subjects from each subject ll~ing a standard Backster ex­ Only two CQT analog studies have used reat teen subjects (five from each of the above three amination administered by an experienced poly­ albeit smalt crimes. Another study using a real groups) were given the opportunity to steal money were scored correctly. Neither alcoholics, nor­ graph examiner. The examiner was aware of the crime tested the GKT technique and is discussed from the wallet of an experimenter who was stag­ mals, nor psychopaths showed differences in de­ details of the experiment, including a knowledge in the section on that technique. An experiment ing a session of psychological testing. The exper­ tectability. In addition, there were no differences of the base rates of guilt and the countermeasures by Ginton, et al. (62L aimed to create even more imenter later alleged that $20 had been stolen, and between the numerical scores of the student ex­ that would be attempted, but was blind to the verisimilitude than usual in the analog study. An arranged for polygraph examinations of the 15 aminers and the blind expert examiner. However, group assignment of individual subjects. At the auxiliary purpose was to test the relative effec­ subjects by a field examiner. No money had ac­ using the +8 cutoff, expert evaluators had more end of each examination, the examiner made a tiveness of behavioral observations, global evalu­ tually been stolen, so the subjects were actually inconclusives (and fewer innocent "hits") than the yes/ no decision regarding the subject's use of ations, including behavioral observations, and nu­ innocent. Four polygraph experts later rated the student examiners. While Hammond concluded countermeasures. After all subjects had been merical scoring based on the charts alone. charts. Averaging the results for these independ­ that his study supported the validity of polygraph tested, the original examiner made a decision as ent evaluators, 11 of the subjects were correctly testing, he believed that certain factors in his study Subjects in Ginton, et al.'s, investigation were to deception by blindly evaluating the charts using labeled innocent, 1 was called guilty, and 3 were could account for the failure to show differences 21 Israeli policemen. They were given paper and by subject category. In particular, all subject the Backster numerical scoring technique, and placed in an inconclusive category. The one error pencil tests that were presented as required apti­ groups actually turned out to be relatively heavy made another decision about the use of counter­ and one inconclusive were with psychotic sub­ tude tests. Subjects were asked to score their own drinkers. Hammond also contended that overall measures based on inspection of the charts. Charts jects, and the other two inconclusives were ~ith tests, which provided an opportunity to cheat, accuracy rates would have been higher with more were also examined and scored by a second ex­ neurotic subjects. Because only innocent subjects i.e., to revise their initial answers. The test answer experienced polygraph examiners. He observed aminer who was blind to all aspects of the were included, a lambda was not calculated for sheets, however, were chemically treated so that that the examiners in his study were unskilled at experiment. this study. cheating could be detected. Seven of the twenty­ detecting countermeasures and at calibrating the As shown in table 6, while there was a low rate one subjects actually changed their initial answers. polygraph instrument. of false negatives (5.5 percent), examiners were Later, subjects were told they were suspected of Hammond not able to make a decision on one-third of coun­ cheating, were offered an opportunity to take a Hammond (64a) conducted a mock crime study Szucko and Kleinmuntz termeasure and no countermeasure guilty subjects, polygraph examination, and were told their ca­ to test the hypotheses that: 1) alcoholics would A somewhat different approach to assessing the and half of the innocent subjects. There was a 7 reers might depend on the outcome. Fifteen sub- be less detectable than normal subjects, 2) psy­ chopaths would be as detectable as normal sub- validity of the polygraph was taken by Szucko ",~:-~" '\ \ I

73 72

ty knowledge test was accurate to a significant first to test GKT with physiological measures and Kleinmuntz (160). They directly compared guilty subjects, but also called half of the inno­ degree in identifying subjects who committed other than EDR. To make assessments of guilt, the ability of polygraph examiners to assess decep­ cent subjects deceptive (false positive). Szucko and both, either, or neither of the crimes. On the basis they used the traditional polygraph respiration tion against the ability of computers to do the Kleinmuntz state that 80 percent of the protocols of this experiment, Lykken argued that GKT, with and cardio measures, and another vascular meas­ same using a digitalized form of the same data. could be classified correctly using a purely statis­ some refinements, could be applicable in criminal ure that was a composite of finger blood volume They had a sample of 30 psychology undergrad­ tical analysis, but they do not state the detection investigations. and finger blood amplitude. This latter measure uate volunteers and randomly assigned them to rate, false positive rate, and false negative rate was recorded by the photoplethysmograph men­ of their statistical analysis. the guilty or innocent conditions. The mock crime Davidson tioned'above. In addition, Podlesny and Raskin involved the "theft" of a $5 bill. Polygraph tests Kircher and Raskin (91) contend on the other performed a quantitative analysis of differences were administered by four examiner-trainees from hand thalt evaluators using numerical evaluations Other researchers have used Lykken's GKT par­ between guilty and innocent subjects on several a polygraph firm near Szucko and Kleinmuntz's can be "at least as accurate as those produced by adigm to explore further its validity as a poly­ other physiological measures. university. The recordings of the physiological graph examination technique. Davidson (46) ex­ any kno'wn statistical decision model and that the GKT was conducted after the same mock theft measures were transformed into digital form for accuracies of both clinical and statistical methods amined the GKT's validity under conditions that computer analysis. varied motivation level and that he claimed were, Podlesny and Raskin (127) used to study CQT. exceed 90 percent." Kircher and Raskin reanalyzed Twenty subjects (10 guilty and 10 innocent) were Six experienced polygraph examiners independ­ charts from the Rovner, et al. (143), study de­ in general, more "ego-involving" for subjects. In Dawson's study, 48 college students were recruited examined with GKT, which included five ques­ ently evaluated the charts. No inconclusive cate­ scribed above and used a lens model, similar to tions with six alternatives each. The rele\ant alter­ gory was allowed in the study. Digital polygraph that employed by Szucko and Kleinmuntz. The and assigned randomly to 12 groups of 4. Three of the four were instructed to attempt to commit natives were placed among the other alternatives data was evaluated by computer. A lens model studies, however, differed in a number of ways, in a "pseudo-random" order (127). The GKT equation drawn from studies of human judgment which could account for the variation in their specific mock murders, and the fourth served as a control. The mock crimes were arranged such charts were scored by the same method used by was used. The results of this analysis indicated results. Probably the most important difference Lykken (105) and Davidson (46). Podlesny and that five of the six polygraph raters were able to is that Kircher and Raskin's interpreters were that one subject would "commit" the crime, one would try to fail, one was motivated but never Raskin also scored the charts in another way, with detect deception significantly better than chance, trained in numerical scoring procedures (136), the addition of an inconclusive zone of scores five but four of them also had fairly high rates of false whereas interpreters in the Szucko and Klein­ had the opportunity, and one (the control) had no knowledge of the crime. Half of the subjects or six. This scoring system for assessing guilt was positives. Szucko and Kleinmuntz estimate that muntz study used global evaluation prooedures used with the photoplethysmograph, respiration, the judges detected on the average 71 percent of (139). who "committed" the murders received a large amount of money ($25 to $50) and half received and cardio measure as well as EDR. Their findings a small sum ($10 to $1). The different amounts were that GKT with EDR was correct for 90 per­ were presumed to create a different level of cent of the subjects and incorrect for 10 percent, all false negatives. Using an inconclusive zone did CONCEALED INFORMATION TESTS motivation in the subjects. The subjects then were examined with the use of GKT. Six multiple­ not add significantly to the accuracy of the tech­ nique, however: 80 percent of assessments were Although the largest number of analog studies Each test in Lykken's study (105) included six choice questions with five alternatives were presented to the subjects, and the EDR was re­ correct, 10 percent incorrect (all false negatives), investigate CQT, several analog studies have ex­ questions about details related to the "murder" and 10 percent inconclusive. amined the validity of the guilty knowledge test, situation and "theft" situation (e.g., asking the corded. The scoring method followed Lykken's one type of concealed information test. A search subject to identify an object present in the "mur­ (105) exactly (see above). Using a weighted aver­ age, 98 percent of the classifications were correct of the literature revealed no analog studies of the der" room). To make subjects anxious about the Giesen and Rollison peak of tension test as a distinct technique. accuracy of their responses during the examina­ against a chance level of 25 percent. The only er­ tion, they were told they would be given an elec­ ror was one false negative. Giesen and Rollison (61) studied the effects on Lykken tric shock if the examiner felt their responses in­ GKT of the subjects' trait anxiety levels and of clicated guilt; in fact, subjects received an electric Podlesny and Raskin the possibility that crime-related details could be In one early investigation of GKT, Lykken (105) shock after every question. The relevant alterna­ relevant to innocent subjects because of asso­ Podlesny and Raskin (127) included GKT in attempted to demonstrate that the detection of in­ tive in each question was randomly varied among ciations unrelated to the crime. Trait anxiety is their study of a variety of polygraph techniques criminating knowledge about a crime can be done an average of five possibilities. If the question anxiety that is characteristic of one's personality and physiological measures. Their experiment was more accurately than the detection of a lie about about the relevant detail produced the EDR with and would be relatively stable over time. Both unique in that it employed GKT in the same con­ the crime. In Lykken's study, 49 male college stu­ the greatest amplitude, it received a score of "2." trait anxiety and "innocent associations" could text as CQT (see above). Thus, they were able dents were randomly assigned to four categories If it was the second largest in amplitude, it re­ conceivably confound the detection of guilt with to compare the accuracy rates of the two tech­ of guilt in conducting two mock crimes. Subjects ceived a "I." A perfect guilty score on each test GKT. niques, although they claimed that a different either committed a staged "theft," a staged was "12," and a perfect innocent score was "0." statistical comparison was impossible because the Giesen and Rollison selected 40 female under­ "murder," both, or neither. An experimenter then A score of seven or greater was categorized as two techniques use very different methods to graduates who responded positively to a question­ conducted two GKT polygraph examinations with guilty for the purpose of analysis, and a score of assess guilt. Podlesny and Raskin also were the naire item on "palmar sweating." EDR is related each subject, one for each crime. six or less was categorized as innocent. The guil- <.~~. I \ \

74 75

to sweating. Thus, this sample may have tended Balloun and Holmes Bradley and Janisse all 270 college student subjects committed a mock to produce higher EDRs than the norm. This crime. There were no "innocent" subjects. Before group was divided into two groups of 20: those Balloun and Holmes (12) used GKT to detect In their study of the influence of threat and the mock crime, subjects were either: 1) given a who scored high on a questionnaire measure of guilt in a "real" crime arranged by the experi­ demonstrations of accuracy on the polygraph ex­ placebo and told it would help them "beat" the anxiety (Lykken's activity preference question­ menters. They were also interested in the effect amination (see above), Bradley and Janisse (35) test; 2) given a placebo and told it would make naire) and those who scored low. Ten subjects in of psychopathy and of repeated examinations on also tested the 192 subjects with the GKT after it more difficult to deceive the examiner; or 3) not each group were then assigned to the guilty know­ the accuracy of GKT. They selected 18 male col­ the CQT had been conducted. The questions con­ given a placebo. The stimulation or number test ledge condition, and to the "innocent associations" lege students with high scores on the psychopathic cerned four relevant details. They were scored was arranged to produce three different feedback condition. The guilty subjects were told to pre­ deviate (Pd) scale of the Minnesota Multiphasic using the Lykken (105) method. With EDR data, conditions. One-third of the subjects' numbers tend to be secret agents who had committed a Personality Inventory (MMPI) and 16 with low the GKT classified an average of 74 percent of sub­ were detected, one-third were not, and one-third murder. They read a narrative about the crime, scores. The Pd scale was originally designed to jects correctly, and 26 percent incorrectly with 11 did not receive the results of the stimulation test. and role-played the act of burning an incriminat­ make the diagnosis of psychopathic personality false positives and 39 false negatives. With the After the GKT was conducted on each subject, ing picture. Innocent subjects also played secret and was used as a scale to measure relative measure of heart rate change, the GKT catego­ charts were scored according to the Lykken (105) agents, but read a narrative containing several de­ "amounts" of psychopathy. The experimenters rized 63.5 percent of subjects correctly and 36.5 method. Adequate charts were obtained for 237 tails (e.g., how much money was involved), which acknowledge, however, that the Pd scale may be incorrectly, with 17 false positives and 53 false subjects. Of these subjects, 70.4 to 80.8 percent in the guilty condition were related to the crime. an inadequate measure of this diagnosis. These negatives. Neither the degree of threat nor the of them produced scores indicative of guilt, de­ They had, therefore, as much exposure to this in­ subjects took a fake intelligence test with two demonstrated effectiveness of the polygraph test pending on how conservative a cutoff point for formation as the guilty subjects, but in an inno­ other students (actually confederates of the ex­ had a significant effect on the discrimination be­ the score was used. Neither the placebo condition cent context. Using GKT with EDR, experimenters aminer). The confederates urged subjects to cheat tween deceptive and truthful subjects. nor the feedback condition produced a significant asked subjects eight crime-related questions, each and supplied test answers to those who were will­ effect on detection ability. Because of the absence with five alternatives. Those details common to ing. Eighteen of the thirty four students cheated. of "innocent" subjects in this study (i.e., a base both conditions were used as the crime-relevant Later, the subjects undelwent a polygraph exam­ Timm rate of guilty of 100 percent), the study tells us items in GKT questions. Scoring followed Lyk­ ination using GKT. They were reminded that nothing about the accuracy of GKT with the in­ ken's (105) method. cheating on exams could lead to academic dismis­ Timm (163) examined the effect of the admin­ nocent subjects. And even the results with guilty sal, and that the experimenters knew that some istration of a placebo on the validity of GKT. Also subjects are difficult to interpret when there is no Giesen and Rollison found that GKT was highly had cheated on the "intelligence test." Informa­ included in the experiment was an investigation comparison to results with innocent subjects. accurate, correctly classifying all of the innocent tion from the intelligence tests that only the of the effect on GKT accuracy of differential feed­ Also, without innocent subjects, a lambda is im­ subjects and detecting all but one of the guilty sub­ chea.ters would know served as the incriminating back from the stimulation test. In the experiment possible to calculate. jects (an average of 97.5 percent correct). In ad­ details on GKT. Another GKT with the same con­ dition, they found that the EDR measure was sig­ tent, but a different order of questions was then nificantly different between guilty and innocent administered to see if the subjects would adapt subjects. Trait anxiety level had no effect on EDR to GKT and, thus, reduce its accuracy. PREEMPLOYMENT SCREENING by itself, but the more anxious subjects in the guil­ ty condition had significantly greater EDR than Balloun and Holmes scored GKT using Lyk­ Despite its widespread use in the field, there are background. Half the group was instructed to lie the less anxious, especially in response to the rele­ ken's (105) method with three physiological meas­ few analog studies of the preemployment screen­ to nine relevant questions and half to tell the truth. vant items. These findings would suggest that anx­ ures (EDR, heart rate, and finger pulse volume), ing polygraph examination. The two that are The polygraph test was conducted, and three iety alone does not confound GKT results, but but only EDR produced significant results. On the known to employ post-1960 polygraph screening charts of 32 questions each were recorded. Most anxiety in guilty subjects might indeed augment first test, guilty subjects scored significantly higher techniques are reviewed. Correa and Adams (43) of the relevant questions concerned information the accuracy of the technique. The study also sug­ and were detected with significant accuracy. How­ conducted an analog investigation of this type of from the questionnaire, but also included were gests that GKT may be accurate even when in­ ever, on the second test, though the guilty sub­ examination with 40 undergraduate subjects. Bar­ three questions about events staged by the re­ nocent subjects have greater associations with jects had significantly greater scores, they were land (16) conducted an analog study with Federal searcher in the initial interview (e.g., giving the crime-relevant items than with neutral items. This not great enough for significantly accurate detec­ Government personnel. subject a glass of water). These latter questions finding, however, must be tempered by the fact tion of guilt at the criterion level (5.5 out of 10) served as a check on the honesty of subjects in that the entire sample was selected for their tend­ used. There was no difference between the high completing the questionnaire, and were consid­ ency for palmar sweating under stress and, thus, and low Pd subjects on either administration of Correa and Adams ered relevant questions in the evaluation of decep­ may be unrepresentative of polygraph subjects in GKT. tion or nondeception. The examiner subjectively general. Like the usual preemployment screening test, made assessments of veracity based on the poly­ the examination in Correa and Adams' study in­ graph recordings. When questions about the cluded a number of relevant questions. Subjects staged events and the application were diagnosed were interviewed prior to the polygraph examina­ by the examiner, all 40 of the subjects were cor­ tion and completed a questionnaire about their rectly identified as being deceptive or truthful. r ,." fili(!4' 4 ..,."t'"

~r::'~" I., \. I 1

77 76

control question response in a zone. In Barland's • 23 percent incorrect; Correa and Adams conducted a question-by­ the control questions against which responses to • 4 percent inconclusive. (16) zone method, each physiological ~eas~re for question analysis of the charts of deceptive sub­ relevant questions are compared. DLCQ differs • 80 percent correct identification of deceptive subjects; each relevant/irrelevant control question paIr was • 13 percent incorrect; jects. A mean of 75 percent of the relevant items from the control question discussed previously rated on a point scale using interpretive criteria from the screening application were correctly clas­ (see ch. 2). With the DLCQ technique, the con­ • 7 percent inconclusive. taught at USAMPS. In the relevant-irrelev~nt sified, and a mean of 25 percent were incorrectly trol questions are not designed to provoke the sub­ method of interpretation, each relevant questIon Presumably, the correct identification rates classified. When change scores were calculated for ject to lie or be concerned about the telling the was evaluated without making specific reference would be lower if only those cases in which the each physiological response, all physiological truth. The '1ies" do not constitute deception since to the control question nearest it; emphasis "was truly deceptive relevant response was counted as measures (EDR, respiration, cardiovascular) sig­ the examiner instructs the subject to tell lies that placed on the size and consistency of reactions at a "hit." To test this hypothesis, the authors co~­ nificantly discriminated truthful from deceptive they both know are false. However, the directed the relevant questions" and scored globally rather ducted a question-by-question analysis. In t~IS subjects. Correa and Adams suggest that these lies are believed to generate concern in innocent than numerically. The "greatest control" met~od method, identification of truthful responses m- findings provide evidence for the validity of . subjects because the subjects are told that to ap­ consisted of evaluating all five relevant questIons creased but identification of deceptive responses prescreening polygraph examinations. There are, pear nondeceptive on the rest of the examination, against the single control question on that chart declined quite a bit. Using the zone technique, 77 however, a number of problems with the Correa they must appear deceptive on the directed ques­ which had the largest overall reaction. In addi­ percent of the truthful questions and only 57 per­ and Adams' study that may compromise its validi­ tions. tion to the comparisons of the three chart inter­ cent of the deceptive questions were correctly ty. Several features of the experiment are prob­ The question of whether CQT can be used out­ pretation methods, charts were. analyz~d global­ identified. With the greatest control scoring meth­ ably highly unrepresentative of or unrelated to side of specific issue investigations (e.g., in pre­ ly and on a question-by-questlon baSIS. od, 85 percent of truthful response~ and less than field preemployment polygraph examinations: the employment or periodic screening) is controver­ half (43 percent) of deceptive ques~lOns were :or­ length of the interview (96 questions); the number In the global method of analysis, subjects were sial. It is difficult to construct standard control rectly identified. The R/I scormg techmque of deceptive responses subjects made (9); and the categorized as either deception indicated, no. de­ questions when much of a person's past is irrele­ showed the best results. With this method, 88 per­ inclusion of questions about the experiment itself. ception indicated, or inconclusive on the baSIS of vant to the purpose of the examination, since past cent of the truthful subjects and 67 percent of Furthermore, the experimenters fail to discuss the appearing deceptive to any of t~e relevant qu~s­ misdeeds (i.e., other than the specific issue being deceptive questions were correctly identified criteria by which the assessments of veracity were tions. That is, if a subject was m fact deceptIve investigated) typically comprise the subject area (although global results were better with the. ~/I made, so it is difficult to ascertain whether these to any relevant question, and reacted of control questions. ~e ~ecep­ technique). This interpretation should be modified assessments correspond to field assessments. tively to any of the questions, It was cons~~ered by the fact that each examiner us~d all three scor­ In this 1981 study, Barland solicited volunteers a hit even though the examiner may have mISIden­ ing techniques and the RII technIque was the .last Barland from the military intelligence community. Sub­ tified which relevant question the subject was de­ one used. Thus, the interpreter had the benefI~ of jects were told the purpose of the study and that ceptive to. Using this method of asses~ing decep­ his previous judgments. The results of a bhnd The Barland (16) study is important for several testing would be limited to the subject's date of tiveness, the three methods of chart mterpreta­ analysis using other interpreters ~ere ;tot retl.dy reasons. One, subjects were actual military per­ birth, place of birth, education, employment, and tion achieved the following results: to be reported by Barland at the tIme hIS 1981 re­ sonnel who in Barland's opinion might be the residences (these were the relevant items), and that port was submitted. types screened for counterintelligence purposes. Zone: b' some subjects would be instructed to furnish the • 62 percent correct identification of truthful su Jects; All subjects were assigned to intelligence duties. examiner with false information. Approximate­ • 19 percent incorrect; The results of the Barland study raise serious It is, thus, unique in being the only validity study ly half the subjects were instructed to lie to one • 19 percent inconclusive; questions about the usefulness of directed lie con­ of preemployment screening in an intelligence of the relevant items; these subjects were offered • 70 percent correct identification of deceptive subjects; trol questions in screening procedures as ~ell as, context. However, because it did not ask any a $20 reward if they could appear truthful on the • 17 percent incorrect; in general, the validity of polygraph testmg for questions related to security interests, it cannot polygraph examination. Unlike the data in the • 13 percent inconclusive. preemployment and counterintelligence 'p~~ses, be considered a full analog to field situations. Greatest control: . Correa and Adams' study, the experimenter was • 77 percent correct identification of truthful subjects; especially if used alone. Of course,. the hmIt~tlOns Second, it tested the validity of a type of CQT, able to check the information given by the sub­ • 15 percent incorrect; of analog studies should be taken mto conSIdera­ jects against data obtained from background in­ the directed lie control question (DLCQ) tech­ II • 8 percent inconclusive. tion Because of these limitations, Barland con­ nique, in a screening situation. DLCQ is part of vestigations. The three polygraph examiners in • SO percent correct identification of deceptive subjects; side;s his results a "worst case" scenario. Final­ a counterintelligence screening test developed by the study had 3, 6, and 9 years of polygraph ex­ • 23 percent incorrect; ly, interpretations must depend 0;t the false pos­ Army Intelligence examiners in 1971. During the perience and had been trained at the U.S. Army • 27 percent inconclusive. itive and false negative rates whIch are deemed Military Police School (USAMPS) polygraph Relevant-irrelevant: . pretest phase of this technique, subjects typical­ • 73 percent correct identification of truthful subjects; acceptable for particular purposes. ly answer "yes" to certain questions. When they course. answer yes, the examiner instructs them that when Examiners used three methods of chart inter­ they are asked such questions during the actual pretation: zone of comparison, greatest control polygraph examination, they should respond with method, and relevant-irrelevant method. As ex­ a "no" rather than a "yes." Thus, they are directed plained in chapter 2, in the zone method, relevant to lie, and their lies to these questions constitute questions are evaluated against the larger of either ------

\i' -V

I,

'-

\

4 78 " 79

FINDINGS oJ: the preemployment screening interview, one relatively high degree of accuracy, but also to be olf them with Government personnel. The analog subject to false positive errors. There was a large Separate statistical analyses were performed for aminers' judgments markedly inferior to a pure­ studies systematically explored many of the tech­ amount of variability among the control question the guilty knowledge and control question analog ly statistical analysis of the charts. However, it nical variables associated with the polygraph (d. analogs, especially the more they diverged in tech­ studies. The following data for the analog studies is unclear how comparable their method of meas­ the Utah group's studies of CQT), and also studied nique from the field method. The guilty knowl­ discussed above were reviewed: uring validity is to the usual method of using an the effect of several situational variables on the edge test had a slightly lower rate of detection accuracy rate, and it is also not clear how appli­ validity of the polygraph. The control question of guilt, more false negatives, but fewer false • percentage of guilty subjects judged decep­ cable the lens model they use is to the question test was found to detect guilty subjects with a positives. tive; of the validity of the polygraph. • percentage of guilty subjects judged nonde­ ceptive (false negatives); Another exception is Ginton, et al.'s, study (62), • percentage of guilty subjects judged incon­ in which field numerical scoring was found to be clusive; inferior to the global evaluation method in detect­ • percentage of innocent subjects judged decep­ ing deception. However, the examiners in that tive (false positives); study were Israeli polygraph professionals who • percentage of innocent subjects judged truth­ may characteristically use a global method of ful; and assessment, and who may have been unfamiliar • percentage of innocent subjects judged incon­ with the Utah numerical scoring system, clusive. Accuracy of detection differed sizably between Also, as with the field studies, an index of control question analog studies using students as predictive association (lambdab) was calculated. subjects (Barland and Raskin, Bradley and Janisse, The results (see tables 8 and 9) indicate that the Szucko and Kleinmuntz; Widacki and Horvath control question test provides a 43-percent im­ is excluded as discussed above) and other control provement in prediction over the base rate for question analog studies (Podlesny and Raskin, these analog studies, and the guilty knowledge test Raskin and Hare, Rovner, et aI., Dawson, Gin­ a 70-percent improvement in prediction over the ton, et a1.). Experiments using students had lower base rates. Because the studies differed so much, percentages of correct decisions for both guilty lambdas were calculated separately for each and innocent, and more false negative:s and false study. As shown in tables 6 and 7, individual positives. Given the small number of studies in lambdas ranged from zero to 83 percent for the each category when the studies are divided in this CQT studies and 38 to 95 percent for the GKT way, it is unclear whether this difference is at­ studies (see ch. 4). These figures should be inter­ tributable to the nature of the subjects (student preted with caution as in real life the base rate v, nonstudent) or other characteristics of these of guilt will vary considerably from approximate­ experiments. ly 50/50 distributions in laboratory experiments. As shown in tables 8 and 9, GKT analog stud­ Thus, it is difficult to draw unqualified conclu­ ies detected a slightly lower average percentage sions from the analog studies given the wide varie­ of the guilty subjects than the CQT analog studies. ty of designs used. They also had a relatively higher proportion of The false negative rate for the analog studies false negatives but a lower rate of false positives. of CQT technique ranged from 0 to 29 percent. It should be noted, however, that GKT was not Inconclusives ranged from 0 to 44 percent for guil­ assessed under conditions that deviated as much ty subjects and from 0 to 53 percent for innocent from the ideal as the control question test devi­ subjects. There is a wide range of false positives ated. Nor were there as many studies testing GKT (4 to 51 percent). Global evalu51tions by the ex­ as CQT. This suggests that the confidence one can aminers, field scoring techniques, and purely sta­ have in the GKT findings is, in general, less than tistical analyses of the data all seem to produce the confidence one can have in the CQT findings. high detection rates in most studies. One excep­ In summary, there exists a number of studies tion is Kleinmuntz and Szucko's (92) study, which of CQT; a smaller number of the concealed infor­ found the validity coefficients of polygraph ex- mation test, all using GKT; and only two studies Chapter 6 Factors Affecting Polygraph Examination Validity

INTRODUCTION Chapter 6 The analyses of both field and analog studies jects have been known to use, or might be trained reported in chapters 4 and 5 indicate that there to use, a number of countermeasures to "beat" is considerable variability in accuracy rates of the polygraph. For many of these factors the re­ Factors Affecting polygraph examinations. To interpret these varia­ search evidence is contradictory. For others, there tions, numerous factors, such as the restricted has been little or no empirical testing. This chapter range of techniques and applications tested in describes evidence from field and analog studies, these studies, need to be considered. In addition, as well as from laboratory investigations, on fac­ Polygraph Examination Validity researchers have attempted to explain the varia­ tors that may affect the accuracy of polygraph bility in accuracy scores by proposing a number tests. The chapter also discusses possible priorities of factors that theoretically may affect polygraph for additional research on factors affecting poly­ test validity. These include characteristics of ex­ graph validity. aminers, settings, and subjects. In addition, sub-

POLYGRAPH EXAMINER, SUBJECT, AND SETTING The previously described analyses of field and ferences in accuracy rates (137,138). There are analog studies (see chs. 4 and 5) emphasize the some data to indicate that experienced examiners characteristics of polygraph tests and their rela­ have better accuracy rates. In recognition of this tion to accurate or inaccurate outcomes. In the outcome, training has been accorded a high priori­ present section, the focus shifts away from the ty both within and outside Government agencies tests themselves, to additional factors that may which conduct polygraph examinations and by affect validity. These factors are sometimes re­ polygraph examiner groups (d. 3). An extensive ferred to as dimensions of external validity and array of training facilities now exists, offering a aid in the assessment of the generalizability of somewhat diverse set of orientations to polygraph research findings. Considerations of these factors testing. will enable evaluation of the conditions under which various levels of validity may be expected Experience from polygraph examinations. Differential validi­ ty in polygraph tests may be obtained with dif­ A number of studies have tested how examiner ferent examiners, subject populations, and with experience relates to validity of polygraph ex­ examinations conducted in different settings. aminations. Horvath and Reid (84), for example, had charts utilized in their validity study reex­ Examiner amined by a group of 10 polygraph examiners. Seven of the examiners were experienced and three It has long been recognized (d. 108,122,135, of them were examiner-interns (each with less than 154) that the examiner's skill has an important ef­ 6 months' experience). According to Horvath and fect on the validity of polygraph tests. Examiner Reid, experienced examiners made an average of experience is an essential element reported by in­ 91.4 percent correct judgments, while the average vestigators and has often been used to explain dif- for inexperienced examiners was 77.5 percent. , \ I

84 85

Training Subjects detectable were significantly less socialized than Other Psychopathology those who were detectable. Susceptibility to detec­ Experience in conducting polygraph examina­ Guilty psychopaths may escape detection be­ Much effort in recent years has been devoted tion seemed to be mediated by socialization; tions suggests that there are a number of clinical cause they are not concerned enough about a mis­ to development of systematic training. Less atten­ results indicated that low socialization subjects components to detection of deception. To some deed to create interpretable physiological re­ tion appears to have been paid to the character­ showed reduced EDRs. Highly socialized subjects extent, training programs capture these clinical sponses. Individuals with other forms of psycho­ istics of subjects of polygraph testing. Frequent­ were more responsive electrodermally, and as a elements by extensive training in "proper" ex­ pathology may escape detection or be classified ly, research reports of polygraph examination do result, several of them were misclassified as aminer attitude and relationship with subjects. In­ as false positives for other reasons (e.g., emotional not report even the most easily available data on deceptive. instability, delusional thinking). The one study creasingly, however, training programs emphasize subject characteristics (e.g., proportion of males Raskin (136) has criticized Waid, et al.'s (198), that has investigated this possibility (74) found, standardized techniques for constructing questions and females). There have, however, been a num­ research as not having practical significance for in fact, that innocent neurotics and particularly and scoring examinations. In this respect, the U.S. ber of studies of specific population groups (e.g., evaluations of polygraph validity. According to Army Military Police School (USAMPS) is per­ psychopaths) hypothesized to be less detectable. psychotics were likely to be identified as decep­ Raskin, simply demonstrating that there is a dif­ tive. There were no guilty subjects in this "real haps the best example. The school serves as the In addition to subjects' psychopathy, other diag­ ference in responsivity on the first set of questions crime" analog study. central training site for almost all Government nostic categories and subject variables such as does not mean that subjects would not be correct­ agencies which maintain polygraph examiner gender, intelligence, motivation, and responsivi­ staffs. USAMPS teaches several versions of the ty to arousal may also affect validity. ly detected in an actual polygraph examination Gender control question tech..'1ique (CQT) (including what (which may involve three to four charts). Some One of the most obvious subject differences is they call the modified general question technique Subject factors are often described in the liter­ of Raskin's own studies (e.g., 21,137) suggest that gender. Males and females may have different pat­ (MGQT) and the original Backster's zone of com­ ature as personality or individual difference fac­ psychopathic individuals are not less detectable terns of autonomic arousal, and such differences parison (ZOC) method) and several specific pro­ tors (136,194). They refer to traits associated with than nonpsychopathic individuals. In Raskin and may affect polygraph testing validity (136,194). tocols for selecting question sets and scoring individuals that may make them differentially de­ Hare's study, convicted felons, half of whom were Unfortunately, few data exist to examine this polygraph charts. Trainees receive both didactic tectable in a polygraph examination. Understand­ diagnosed as psychopathic, performed a mock hypothesis; most research only studies male sub­ classroom training and supervised experience con­ ing these effects should enable determination of crime. These subjects were then administered a jects. The one study by Cutrow, et. al. (45), that ducting polygraph examinations. The current cur­ the conditions under which polygraph testing will polygraph examination and offered a substantial specifically tested for sex differences did not find riculum for USAMPS uses Reid and Inbau's (139) yield particular levels of validity. The mechanism monetary bonus if they could produce a truthful any. In another study (61), all female subjects text on polygraph testing, supplemented by ma­ by which subject variables affect polygraph ex­ outcome. In contrast to Waid, et al.'s. findings, were tested in a mock-crime situation using the terials prepared especially for its trainees (179). amination validity has to do with differential Raskin and Hare found that individuals diagnosed guilty knowledge test (GKT). GKT was found to USAMPS is one of a number of training programs autonomic arousal. Validity is affected when an as psychopathic and/ or low in socialization were be highly accurate, but because males were not certified by the American Polygraph Association interaction results between arousal and polygraph more reactive and easily detectable than those not also tested, it is impossible to determine if males (cE. 3). testing. psychopathic and high in socialization. Earlier would have been less detectable. The two Honts research by Raskin (21) supports this finding. On the basis of presently available data, it is and Hodes (76,77) analog studies described in Barland and Raskin's (22) field study, on the other not possible to determine whether types of train­ Psychopathy and Level of Socialization chapter 5 included both females and males; the hand, found that subjects who scored high on the researchers do not report any gender differences ing have an effect on outcomes. A study by Ra~­ One aspect of potential subject effects that has psychopathic deviate (Pd) scale of the Minnesota kin (133) indicates that examiners trained in in detectability. received considerable attention is the effect of level Multiphasic Personality Inventory (MMPI) (a schools that emphasize numerical scoring were of socialization and psychopathy on detectabili­ measure of psychopathy) had smaller cardio (but Intelligence significantly more accurate than examiners who ty. In a series of studies by Waid and his col­ not respiration or skin conductance) scores than attended other schools (97.1 v. 86.9 percent). It leagues (193,198,199) significant relationships low Pd subjects. Intelligence is an additional variable which po­ is difficult to determine, however, if training in were found in the laboratory between socializa­ tentially might affect detectability. The ability of numerical scoring is more efficient or if better ex­ In a previously described study, Balloun and tion and autonomic responsiveness. An initial intelligent subjects to anticipate questions may af­ Holmes (12) conducted an analog study of col­ aminers/schools select such techniques. The fact finding (193) was that college students who scored fect polygraph accuracy. One possibility is that that examiners who were trained in numerical lege students using a "cheating" situation. Their low on socialization (on a standard psychological intelligent subjects are less detectable because, if results indicated that subjects who scored high on techniques, but who did not use them, did more inventory), gave smaller electrodermal responses trained, they are able to anticipate questions and the Pd scale of the MMPI were just as easy to poorly than examin.ers trained in numerical tech­ (EDRs) to stimuli than did high scoring subjects. employ countermeasures. Another possibility is detect as were those individuals who scored low niques who used them (88.5 v. 98.9 percent) sug­ In a more directly relevant investigation (198), a that because intelligent subjects better understand on the scale. It is important to note, however, that gests that numerical evaluation rather than exam­ group of college students was asked to deceive or the implications of a polygraph examination, they iner selection (or some other aspect of the train­ the polygraph test was a concealed information not to deceive a professional polygraph examiner. will respond to relevant questions with heightened ing) provides an advantage. type of test, not a CQT or relevant/irrelevant Results indicated that subjects who were not arousal when they are attempting to deceive (20). (R/I) test. A doctoral dissertation by Hammond (64a) also found no differences between normal There has been relatively little research on in­ and psychopaths. telligence and polygraph testing. In one of the few ______~ ______-L----- .. ~- \ 1

86 87

experiments which look at intelligence and detec­ Setting Bogus Pipeline conditions using paper-and-pencil tests would not tability, Kugelmass (95) found no correlation be­ reveal such atHtudes. Later research has shown One theory underlying lie detection using the An interesting and potentially important aspect tween intelligence and responsivity on a peak of that this findings holds for attitudes toward hand­ polygraph is that the threat of punishment leads of how the polygraph achieves valid results is tension (POT) card test. On the other hand, re­ icapped individuals and for "confessing" to hav­ an individual to manifest a physiological reaction based on what social psychologists such as Jones search by Gustafson and Orne (65) found that mo­ ing prior knowledge about a psychological experi­ (48). This suggests, then, that settings in which and Sigall call the "bogus pipeline" (87). The tivation to deceive increased the probability of ment. detection. Barland and Raskin (20) feel this is an individual is more certain of being detected and bogus pipeline is a procedure used to elicit truthful evidence of the potential role of intelligence. in which the consequences are greatest, will per­ attitudes in situations where social desirability ef­ Although the bogus pipeline research suggests Barland and Raskin's study (22) which compared mit higher levels of detection. Furthermore, in fects (Le., subjects' desire to express socially ac­ that the conditions of testing (in particular, the detection rates among subjects of different educa­ order to be certain of being detected, a subject ceptable opinions) may mask actual attitudes. The perceived complexity and accuracy of equipment) tion levels, found no difference. However, a sepa­ must believe in the efficacy of the polygraph pro­ procedure involves attaching subjects (via skin may have important effects on polygraph sub­ rate analysis of the sources of false positive er­ cedures in order for it to function. According to electrodes) to an ostensible physiological record­ jects, it is not clear how or to what extent these rors by Raskin (133) found that the majority of some (e.g., 194), the polygraph is often used ing device called the "electromyograph" (EMG) effects influence the validity of the test itself. In false positives occurred among subjects who had somewhat like a "stage prop," and its presence and providing subjects with a "steering wheel" a substantial number of criminal investigations college degrees. Level of education, of course, is is meant to "enhance the subject's concern." device to record their attitudes. In a typical study subjects voluntarily confess after having the poly­ not a perfect indicator of intelligence, and there Stimulation tests, used in almost all field (87), subjects were told that the EMG measured graph procedure explained or being shown the re­ is a need to better understand these relationships. polygraph examinations, serve the same function, implicit muscle potentials and that it was an im­ sults of the examination. In personnel screening, albeit more directly. There is considerable discus­ proved polygraph or "lie detector. II The recording subjects often admit to errors in their job appli­ sion (e.g., 202) in the literature about how fre­ device is actually "electrical junk" (87), and the cation or past indiscretions (24,165). Most avail­ quently within a polygraph examination such purpose of the procedure is simply to convince able field and analog research does not permit de­ Ethnic and Group Differences stimulation tests should be utilized in order to in­ subjects that their actual attitudes are detectable. termination of the extent to which the polygraph Another category of subject differences that crease the validity of the examination. is used in this way. may affect polygraph validity has to do with Results from a number of investigations which have used the bogus pipeline procedure (e.g., ethnic and group differences in physiological re­ Instrument Specific Settings sponse. Research conducted cross culturally (e.g., 131,150) support Jones and Sigall's premise. Sev­ Polygraph examinations take place in a 97,104,158), indicates that there are ethnic dif­ Some research, reported by Orne and his col­ eral studies indicate that when subjects believe number ferences in response to stress. Such differences leagues, addresses the question of the situational that their attitudes are detectable by a physiolog­ of settings, ranging from facilities specifically de­ may, in tum, affect detection of deception. As features necessary for a polygraph examination. ical recording device, they more readily express signed for this purpose to motel rooms. Specifical­ noted earlier, these effects may interact with the In one component of a study reported by Orne, their actual attitudes. Although it is difficult to ly designed facilities usually include one-way mir­ ethnic identification of the examiner. However, et al. (123), subjects were led to believe that the know what "actual" attitudes are, higher truth­ rors for observation and audio recording capa­ effects of ethnic differences have not been direct­ polygraph recording equipment was not opera­ fulness is assumed with the bogus pipeline because bilities, and are located so as to prevent interrup­ ly tested with respect to polygraph examinations. tive. There was some indication that the pretest the procedure yields more socially undesirable re­ tions during the examination. It is reasonable to condition in which subjects were led to believe sponses than when it is not used. For example, assume that the setting may interact both with that the polygraph instrument was inoperative in Sigall and Page's (150) initial experiment, they subject and examiner characteristics to affect the produced a lower detectability; however, results found that subjects in the bogus pipeline condi­ validity of polygraph tests. No research, however, Autonomic Lability were not statistically significant. In an earlier tion would admit to negative attitudes about directly tests the impact of different settings on A final individual difference is what Waid and study (161), detectability was not affected by sub­ "Negroes." Similar subjects in nonbogus pipeline polygraph testing validity. Orne (194) have referred to as autonomic labili­ jects' belief in whether the machine was recording. ty. Regardless of other differences among subjects, Both of these studies involved use of concealed there may be consistent individual differences con­ information tests. nected with their level of autonomic arousal. A more recent study by Orne's group (198) COUNTERMEASURES Although there is considerable variance for an tested a similar hypothesis using a different pro­ Countermeasures are deliberate techniques used tion. The research on polygraph countermeasures individual in autonomic responses to most phys­ cedure. In this study, subjects saw the polygraph by deceptive subjects to avoid detection during is summarized below by type of countermeasure. iological measures of autonomic nervous system machine turned off, although the experimenters a polygraph examination (23,108,139,194,195). (ANS) arousal, electrodermal lability may be dif­ actually ran the leads to a second polygraph de­ Countermeasures can range from simple physical Physical ferent. Given the importance of the EDR for poly­ vice and were able to record responses during a techniques, to so-called mental countermeasures, graph examinations, it may be essential to under­ pretest review of questions. The results indicated to the use of drugs and biofeedback techniqu~s. Physical measures taken by a subject during a stand more about this factor. Unfortunately, most that subjects who were aware of being recorded There is a potentially large list of such counter­ polygraph examination are, perhaps, the most fre­ of this research (e.g., 200) has been conducted had significantly higher responses to relevant measures, and there are a number of plausible, quently discussed countermeasures used by sub­ with concealed information tests and not with questions and not significantly different responses but not yet validated, techniques to avoid decep- jects (20,108). Any physical activity which could CQT or R/I tests. to control questions. \ i \

89 88

research is now being planned under the auspices the test. As noted earlier, most polygraph exam­ affect physiological response is a potential prob­ cological agents have the potential to produce in­ of the National Security Agency and the Army inations rely on the subject's motivation to avoid lem for interpretation of a polygraph test record. correct or uninterpretable polygraph records. Intelligence and Security Command. There are detection rather than on any response directly There is no question that physical measures, from also no data as to the use of common drugs by connected with "telling a lie." Simple cognitive tensing muscles to biting the tongue, to squeez­ Research on drugs' f.::lctors is only beginning to countermeasures include patterns of thinking that be conducted. Recent research by Waid (197) in­ actual polygraph examination subjects. Although ing toes, to shifting one's position can affect examiners normally ask subjects to report use of suppress responses to control or irrelevant ques­ physiological response. dicates that the tranquilizer, meprobamate (Mil­ tions. More complex cognitive countermeasures town®), permits subjects who are being deceptive medications or other drugs, blood samples or There are frequent references to the use of such other detection means are typically not employed. are based on knowledge of the results of the ex­ to increase their ability to avoid detection in a amination and lack of belief in one's detecta­ measures (see e.g., 40,108). But little systematic polygraph examination. One feature of tranquil­ It is thus difficult to assess the magnitude of drug use by subjects in previous research on the validity bility. research has been conducted to establish the im­ izers such as meprobamate is that they suppress pact of the use of such measures on polygraph autonomic activity which may not be accompa­ of polygraph testing. decisions. Kubis (93) found that when subjects nied by any observable psychomotor differences. In addition to drugs, there have also been re­ Controlling Thoughts press their toes towards the floor they were able In Waid, et al.'s, study a GKTwas used in a poly­ ports of the use of various chemicals to confuse Any individual who under-stands the basic to reduce the probability of detection from 75 to graph test. Subjects were all male and divided into physiological recording (see 20). Placing antiper­ structure of a particular polygraph examination 10 percent. A replication of this experiment, how­ three groups: 1) a tranquilizer group; 2) a placebo spirant powder, clear nail polish, or other agents should be able to differentiate irrelevant and con­ ever, by More (119) found that there was no de­ group; and 3) a control group. Only 3 of 11 guil­ on the balls of one's finger's may make EDRs less trol from relevant questions (when an R/I or CQT crease in detectability caused by toe movements. ty subjects who had taken meprobamate were reliable. Such measures, however, should be de­ examination is conducted). Given that when a In two more recent studies discussed in chapter scored as deceptive. tectable by a trained examiner. CQT is employed, the questions are reviewed 5, by Honts and Hodes (76,77), the efficacy of It should be noted that because Waid, et al.'s, prior to their presentation, a subject intent on de­ two physical countermeasures was tested in ana­ Hypnosis/Biofeedback ception could try to alter cognitively responses log situations. Both studies found that counter­ investigation involved GKT, the ability to generalize from the results is limited. According to various questions (although, since the order of measures allowed subjects to "beat" the poly­ There is a substantial literature in psychology presentation is varied, this is made difficult dur­ graph. Furthermore, there were a large percent­ to Raskin (136), a different problem would be en­ about the use of hypnosis and biofeedback to alter countered by attempts to utilize tranquilizers to ing an actual test). The procedure would be to age of inconclusives. In addition, both studies and condition physiological responses. There is try to dissociate oneself from the relevant qULS­ found that experienced examiners were not able defeat an examination employing CQT. The use some evidence (see 146) that hypnosis, for exam­ of such drugs in a CQT polygraph examination Hons and heighten response to control questions. to detect use of the countermeasures. A recent ple, induces declines in skin conductance levels. Various means of such mental dissociation have study by Honts, Raskin, and Kircher (78) also would be more likely to yield inconclusive find­ A number of investigations have attempted to ings, rather than errors, because the drugs would been described (d. 20,108). They range from con­ found that the use of physical countenneasures show that hypnotically suggested amnesia is an centrating on an irrelevant object or by convinc­ decreased detectability; the false negative rate for likely res~t in no difference between the responses effective countermeasure. Such research seems to to control and relevant questions. This interpreta­ ing oneself that the question means something dif­ countermeasure subjects was 78 percent. How­ indicate that hypnosis is not effective (see 20). ferent than intended. ever, examiners using a separate EMG analysis tion is supported by the recent analog study of were able to detect the use of countermeasures 80 Gatchel, et al. (59), which found that the use of Recent research by Corcoran, Lewis, and Gar­ There is little research evidence available to as­ percent of the time. propranolol, a beta-blocking drug, resulted in a ver (42) has examined the effects of biofeedback certain the potential of these techniques. Only one 32.2-percent inconclusive rate, although the over­ training on suppressing EDR. They found that study on the use of such cognitive countermeas­ Thus, the evidence, while limited, is that decep­ all error rate was low. An additional finding was both hypnosis and biofeedback groups were able ures appears to have been conducted, and it does tive subjects who use physical countermeasures that examiners could not tell which subjects had to reduce detectability after training as compared not provide much information. In this study, and who can distinguish nonrelevant from rele­ used the drug. Conclusions drawn from this study to a control group. In another study, Rovner, Dawson (49) used as subjects actors trained in the vant questions (in a CQT or RII test) can increase must be limited by the fact that subjects were stu­ Raskin, and Kircher (143) reported that subjects "method" school of acting. Subjects were in­ their chances of avoiding detection. dents. Other studies using college students (e.g., who received extensive information about the na­ structed to try to defeat the polygraph in a CQT 76,77) have also resulted in large numbers of in­ ture of lie detection and practiced using counter­ examination. The results indicated that without Drugs conclusives. measures were detected significantly less than sub­ exception the actors were detected and were not jects without such training. It seems clear that if able to control their affective responses. It should In contrast to physical measures, which poten­ A recent study by Iacono, et aI. (86), found that hypnosis or biofeedback operate as countermeas­ ingestion of neither 10 milligrams of diazepam be noted, however, that the actors in Dawson's tially may be detected by an observant polygraph ures, especially with commonly used tests such study were merely tolJ to "appear innocent" and examiner by running multiple charts or by careful (Valium®) nor 20 milligrams of methylophenidate as CQT, that extensive training would have to (Ritalin®) affected the accuracy of detection. were not told what would make them appear non­ comparison of particular physiological measures, accompany their use. deceptive on a polygraph test. the use of various pharmacological agents is prob­ Results in both active drug conditions were more accurate than when subjects ingested a placebo ably more difficult to detect. Not only may drugs Mental Knowledge of Results be difficult to detect by observation, but they may (a capsule containing lactose). also not be detected by multiple polygraph tests. Research on other psychoactive drugs has not Another category of countermeasures involves Another set of countermeasures is based on sub­ Some theorists have suggested that such pharma- been reported in the literature, although such those that get the subject to think differently about jects having knowledge of the results of their poly- ,~. 'I, I I

90 91

graph examination. In criminal situations, partic­ ever, 25 percent of the guilty subjects in the in­ ularly when an inconclusive outcome is obtained formation plus practice group were able to avoid RESEARCH IMPLICATIONS OF FACTORS AFFECTING VALIDITY or when the subject disagrees with the outcome detection. Raskin (136) maintains that this 25-per­ of an initial polygraph test, subjects are retested. cent error rate should be considered the "upper If further research on polygraph testing is car­ An additional priority is to develop and test In noncriminal situations, individuals are often limit" because, in actual field situations, motiva­ ried out, a number of research priorities can be question techniques that may be employed in tested at the beginning of their employment and tion would be much higher. Although Raskin is, identified on the basis of the present analysis. screening situations. Some examiners for exam­ at a number of subsequent points. In addition, perhaps, correct, it is also possible that in actual These priorities include research on the theory of ple claim to use a version of CQT for screening subjects who know that they will receive a poly­ situations (where motivation is high), subjects polygraph testing, polygraph techniques, coun­ examinations (see ch. 2). This application of CQT graph examination may seek training in methods might engage in more practice. termeasures, comparison with other techniques, has not been subjected to scientific tests of validi­ to avoid detection. For all of these reasons, it is and field-based studies. ty. In addition, efforts should be devoted to test­ important to understand how feedback about ing the use of CQT with diff,erent subject groups polygraph examinations affects validity and Theory and in national security investigations. Belief in :lMachine" whether prior experiences represent a potential Polygraph testing is premised on the belief that countermeasure. A final countermeasure is based on research lying produces reliable physiological reactions. about the bogus pipeline (87) and the role of the Countermeasures In an early laboratory study (67), subjects were Testing the efficacy of this assumption is an im­ setting in inducing valid outcomes. If the validi­ given a stimulation test and feedback concerning portant research need. Basic research could ex­ If polygraph testing is to be more widely em­ ty of polygraph testing is dependent on the belief its outcome. Feedback was manipulated so that amine the physiological reactions to different ployed in national security investigations, there by subjects in the efficacy of the procedure, then some subjects thought they had successfully types of lies and under different conditions of is an urgent need for research on countermeasures. a possible countermeasure would involve train­ avoided detection and others thought that they arousal. Particular priorities would be research on drugs, ing subjects to believe that the polygraph does not had not. Subjects were motivated on a subsequent biofeedback training, and subject gullibility, and work. This might be done, for example, by pro­ trial to avoid detection (they were told, "only motivation. Such research needs to be carried out viding subjects with false feedback on a polygraph Scoring mature and stable individuals are able to fool the both in field situations and in the laboratory. examination. Unfortunately there is little research lie detector"). The results indicated that subjects Research is currently being conducted by the in this area, and the two studies that have been who believed that they had avoided detection U.S. Anny on development of computer scoring conducted come to different conclusions about the There are a number of drugs that are suspected were much less detectable on the second trial (13 systems and more reliable measures of physiolog­ effect of belief in the techniques' effectiveness. In of lowering ANS arousal and that theoretically out of 16 were not detected, while only 1 of 16 ical arousal. There is some evidence (e.g., 92) that one study, Bradley and Janisse (35) tested the may be able to invalidate the results of a poly­ in a ,control group were able to avoid detection). the validity of polygraph examination decisions hypothesis by rigging a stimulation test at various graph examination or compel an "inconclusive" It shO\.tld be noted that a stimulation test is a form is improved if the clinical judgment of examiners levels of effective detection. Depending on the finding. A first priority is to extend Waici, et al.'s of a con<:ealed information test and the result may is removed (see also, 27) and related evidence that condition, subjects were "detected" on one, two, (197), research on meprobamate (which reduced be due to lowered overall arousal. "Beating" a numerical scoring is more accurate (91,133) than or three trials to create the impression that the detectability) to other psychoactive drugs. CQT represents a somewhat different problem. nonnumerical scoring. Research should proceed detection measures were ineffective, sometimes ef­ on developing analogs to digital scoring systems. A recent study, by Rovner, et al. (143), tested fective, or perfectly effective. For the EDR meas­ Such research, however, would not address the Biofeedback training, as well as other forms of a similar hypothesis in a CQT examination. Sev­ ure, the more effective the apparatus appeared to impact of examiner-examinee interaction, and this training have not been investigated, yet their ef­ eral groups of subjects were placed in a mock be, the more the innocent subjects scored as non­ area also needs further study. fects on polygraph examinations may be substan­ crime situation. One group was given informa­ deceptive and the more the guilty subjects scored tial. Subjects' beliefs about the accuracy of the tion about the nature of a CQT examination and as deceptive. In an earlier study, however, Timm Question Techniques polygraph may also be critical. As suggested by information on what physiological reactions they (162) found that feedback about the techniques' the research on the "bogus pipeline," individuals should try to simulate. Another group was given effectiveness had no effect on whether subjects Another research priority is to develop addi­ who believe their underlying thoughts are detect­ information plus two practice tests involving ac­ deceptiveness or nondeceptiveness could be de­ tional protocols for question development. able are more likely to provide truthful responses. tual physiological recordings after which they tected. The theoretical support provided by re­ Perhaps the most important research need in this The reverse phenomenon seems feasible and it were told whether or not they had beat the poly­ search on. the bogus pipeline indicates that sub­ regard is to develop and field-test the concealed would seem possible to train individuals to believe graph. A third group served as a control and was jects' belief in the technique may be important, information test. Basic research and theory (see, that the polygraph is ineffective. Such training given a typical polygraph examination. The re­ and that additional research is needed to assess e.g., 27,108,136) suggests that such examinations might be accomplished by providing individuals sults indicated that the information only and con­ the effects of belief in the machine on actual poly­ ~\ have the highest likelihood of detecting deception, with false feedback on the polygraph as well as trol group were not able to avoid detection; how- graph tests. \ although no field research has yet been conducted by specific instructions during simulated poly­ to examine their use. Such research should both graph examinations. Similarly, subjects who can establish means of constructing GKTs and their be easily trained to beat the polygraph may be validity in actual use. more desirable as intelligence agents.

____----<111 __ ~~ ~-- -~--~------...... ------~~ ~ -- - ,------

'J d

92

Comparison With Other Techniques cause they can only only be conducted where an independent criterion of guilt or innocence is Only one study in the available literature (207) available. As such, these studies may represent systematically compares the polygraph with other a select sample of cases (e.g., where guilt is over­ investigatory tools. There is a need to examine whelming) and a select set of examiners. Analog whether the polygraph provides independent or studies have a different set of problems and have corroborative evidence and whether the judg­ not adequately motivated subjects or may not ments made by polygraph examiners are merely have the appropriate number of cases. What is a function of their clinical judgment as investiga­ needed is research which deals with the problems Chapter 7 tors, or whether it is a function of the polygraph of the available field and analog studies. examination itself. One "theoretical" solution to the problem of A complication with this research is that the conducting systematic field studies is to conduct Conclusions polygraph functions, in many situations, as a "ABSCAM"-like investigations using bogus un­ threat. Individuals' fear of taking the ex~mination, authorized disclosures (instead of bribes) that in many instances, may lead them to confess or would enable investigators to set up situations provide incriminating evidence. The threat poten­ where they have knowledge of who is guilty or tial, however, is in part a function of theirs and innocent. The polygraph could be used to select others' knowledge of research results. If, for ex­ guilty from innocent with a known base rate and ample, it became widely known that the poly­ ground truth. Such methods, of course, raise a graph was "beatable," it is likely that this threat number of ethical, Lgal, and pragmatic questions, would be reduced and, hence, the validity (and and it is not clear whether they could provide de­ utility) of the polygraph would be reduced. finitive answers. They could not be used frequent­ ly nor with a wide range of techniques/situations. Field Siudies Conducting polygraph research presents serious conceptual and methodological problems; in the As described in chapters 3, 4, and 5, there are absence of such r~search, however, it will not be numerous problems with the available field and possible to develop fully an assessment of the va­ analog evidence. Field studies are problematic be- lidity of polygraph examinations.

CONCLUSIONS

The description in this chapter of factors affect­ central problem, not adequately addressed by ing validity and potential countermeasures sug­ either the literature on improvements in validity gests that there is a great deal more to understand or countermeasures, is the extent t.o which these about polygraph tests before one can be assured factors affect false negative and positive error rates of their validity. Despite our lack of full under­ or affect numbers of inconclusives. For policy pur­ standing, however, several factors that affect poses, clearly such distinctions and a sense of the validity are known. In part, the history of poly­ magnitude of false decisions is needed. Substan­ graph development over the past 15 to 20 years tial research, beyond what is currently available, has been to systematize and improve polygraph would have to be conducted in order to answer testing procedures based on these factors. One such questions. Chapter 7 Conclusions

INTRODUCTION The primary purpose of this technical memo­ examinations by th~ Federal Government for in­ randum is to evaluate the scientific evidence on vestigations of specific incidents (i.e., unau­ the validity of polygraph tests. The memorandum thorized disclosures), and, most significantly, for responds to concerns of the Committee on Gov­ personnel security screening. In addition, NSDD- ernment Operations, U.S. House of Representa­ 84, administration policy, and the DOD proposals tives, about significant changes in Federal Gov­ authorize adverse consequences for refusal to take ernment policy concerning polygraph testing. As a polygraph examination. discussed in chapters 1 and 3, National Security By letter of February 3, 1983, the Committee Decision Directive 84 (NSDD-84), issued by the on Government Operations asked OTA to assess President on March 11, 1983, authorized executive the scient.ific evidence on the validity of polygraph agencies and departments to require employees testing, based primarily on a critical review and to take a polygraph examination in the course of evaluation of existing research. In order to con­ investigations of unauthorized disclosures of clas­ duct this assessment, OTA studied the actual pol­ sified information. On October 19, 1983, the De­ ygraph examination process, reviewed the results partment of Justice announced that administra­ of prior research reviews, analyzed a wide range tion policy would also permit Government-vvide of relevant field and analog studies, and surveyed polygraph use ih preemployment, preclearance, Federal agencies as to their polygraph use and any periodic, and aperiodic personnel security screen­ past, present, or planned polygraph research. This ing of employees with access to hIghly classified chapter highlights the overall scientific conclusions information. Draft proposed reyisions to Depart­ of the OTA evaluation and then discusses in some ment of Defense (DOD) polygraph regulations detail specific scientific conclusions and the im­ (DOD 5210.48) would also authorize the ex­ plications for recent and proposed changes in Fed­ panded use of polygraph testing as part of per­ 1 eral policy on polygraph testing. l sonnel security screening of employees with highly ! sensitive access. I :1 I The combined effect of these changes is to au­ j q thorize substantially expanded use of polygraph Ii ~ ! 11 :1 :1 ;1 ." OVERALL SCIENrlFIC CONCLUSIONS 'I :1 OTA concluded that, as shown in chapter 2, response (as measured by the instrument) to the 'I polygraph testing is, in reality, a very complex different questions, and infer deception or truth­ '1I, :1 process that varies widely in application. Al­ fulness based on these comparisons. !I though the polygraph instrument itself is essen­ 'I tially the same for all applications, the purpose One general type of polygraph question tech­ of the examination, type of individual tested, ex­ nique (called the control question technique) is amin.er training, setting of the examination, and commonly used for investigations of specific crim­ type of questions asked, among other factors, can inal incidents and has received most of the re­ differ substantially. The instrument cannot itself search attention. Another technique (known as detect deception. Therefore, polygraph tests re­ relevant/irrelevant) typically used for preemploy­ quire the examiner to develop questions to be ment screening and periodic screening purposes asked in each case, compare the physiolohical has been only minimally researched. Based on a

95 96 97

detailed review of these and other question tech­ cluded that this accuracy range could be partial­ niques in chapter 2, OTA concluded that there ly explained by variations in research design but SPECIFIC SCIENTIFIC CONCLUSIONS IN POLICY CONTEXT are significant differences, and that the results of perhaps to a greater extent, as is discussed in research on one technique cannot be generalized chapter 6, by differences in examiners, examinees, Specific-Incident Criminal Investigations • Ten individual field studies: to other techniques. Also, differences between question techniques, and conditions of testing. -correct guilty detections ranged from 70.6 techniques are so significant that the results of A principal use of the polygraph test is as part to 98.6 percent and averaged 86.3 percent; research on one technique in one application can­ OTA concluded, therefore, that no overall of an investigation (usually conducted by law en­ -correct innocent detections ranged from not necessarily be extrapolated to other applica­ measure or single statistic of polygraph validity forcement or private security officers) of a specific 12.5 to 94.1 percent and averaged 76 tions. Chapter 2 also reviewed the Federal Gov­ can be established based on available scientific situation in which a criminal act has been alleg­ percent; ernment's use of polygraph testing and found that, evidence. The amount and quality of the evidence ed to have, or in fact has, taken place. This type -false positive rate (innocent persons found with the exception of the National Security Agen­ depends on the design and conduct of specific of case is characterized by a prior investigation deceptive) ranged from 0 to 75 percent and cy (NSA) and Central Intelligence Agency (CIA), studies and the particular application researched. that both narrows the suspect list down to a very averaged 19.1 percent; and most current u')e, even in DOD, is for investiga­ Some applications (e.g., the use of the polygraph small number, and that develops significant in­ -false negative rate (guilty persons found tion of specific crimes using the control question in criminal investigations) have been fairly heavily formation about the crime itself. When the nondeceptive) ranged from 0 to 29 A per­ technique. researched, while others (e.g., polygraph use in polygraph is used in this context, the. ~p~lic~tion cent and averaged 10.2 percent. preemployment screening) have had very little is known as a specific-issue or specifIc-mcident • Fourteen individual analog studies: In chapter 3, OTA reviewed the legal, govern­ criminal investigation. mental, and scientific controversies over poly­ research attention. -correct guilty detections ranged from 3504 graph testing. OTA found that previous debates to 100 percent and averaged 63.7 percent; Further, regardless of whether polygraph testing Results of OTA Review at the Federal level have focused heavily on -correct innocent detections ranged from 32 is used in specific-incident investigations or per­ whether polygraph testing is scientifically valid. The application of the polygraph to specific­ to 91 percent and averaged 57.9 percent; sonnel screening, OTA concluded that polygraph The conclusion of previous congressional inquiries incident criminal investigations is the only one to -false positives ranged from 2 to 50.7 per­ accuracy may also be affected by a number of fac­ has been that there is little or no scientific basis be extensively researched. OTA identified 6 prior cent and averaged 14.1 percent; and tors: examiner training, orientation, and experi­ for the use of polygraph testing. Prior scientific reviews of such research (summarized in ch. 3), -false negatives ranged from 0 to 28.7 per­ ence; examinee characteristics such as emotional reviews, on the other hand, have contradicted as well as 10 field and 14 analog studies that met cent and averaged lOA percent. stability and intelligence; and, in particular, the each other, some concluding that polygraph test­ minimum scientific standards and were conducted use of countermeasures and the willingness of the using the control question technique (the most ing is almost 100 percent accurate, others that it The wide variability of results from both prior examinee to be tested. In addition, the basic common technique used in criminal investiga­ is little better than chance. OTA determined that research reviews and OTA's own review of indi­ theory (or theories) of how the polygraph test tions; see chs. 2, 3, and 4). Still, even though part of the problem in reaching conclusions about vidual studies makes it impossible to determine actually works has been only minimally devel­ meeting minimal scientific standards, many. of polygraph testing validity is that several scientific oped and researched. a specific overall quantitative measure of poly­ these research studies had various methodologIcal criteria must be taken into account when assess­ graph validity. The preponderance of research problems that reduce the extent to wh~ch results ing validity. Also, previous scientific reviews have In sum, OTA concluded that there is at pres­ evidence does indicate that, when the control can be generalized. The cases and exammers were not been conducted systematically. In addition, ent only limited scientific evidence for establishing question technique is used in specific-incident often sampled selectively rather than rand~mly. previous reviews, whether legal, governmental, the validity of polygraph testing. Even where the criminal investigations, the polygraph detects For field studies, the criteria for actual gUilt or or scientific, have not differentiated polygraph use evidence seems to indicate that polygraph testing deception at a rate better than chance, but with innocence varied and in some studies were inade­ by type of question technique or application. detects deceptive subjects better than chance error rates that could be considered significant. quate. In addition, only some versions of the con­ OTA conducted its own systematic review of (when using the control question technique in spe­ trol question technique have been r~searched~ and prior research studies on the validity of polygraph cific-incident criminal investigations), significant error rates are possible, and examiner and exam­ the effect of different types of exammers, subjects, The figures presented above are strictly ranges testing (see ch. 4 for discussion of field studies of settings, and countermeasures has not been sys­ or averages for groups of research studies. actual polygraph examinations and ch. 5 for inee differences and the use of countermeasures may further affect validity. tematically explored. Another selection of studies would yield different discussion of analog or simulation studies). OTA results, although OTA's selection represents the Nonetheless, this research is the best available four:d that there are almost no studies relevant More specific scientific conclusions and the im­ set of studies that met minimum scientific criteria. source of evidence on which to evaluate the scien­ to proposed Federal Government expansion of plications for recent and proposed changes in Fed­ Also, some researchers exclude inconclusive re­ tific validity of the polygraph for specific-incident polygraph testing for preemployment, periodic, eral policy on polygraph testing are presented sults in calculating accuracy rates. OTA elected criminal investigations. The results (for research or aperiodic screening. This finding has major pol­ below. The discussion is organized in terms of to include the inconclusives on the grounds that on the control question technique in specific­ icy implications discussed later. OTA also found conclusions and implications, first, for specific­ an inconclusive is an error in the sense that a guilty incident criminal investigations) are summarized that, even among the rather extensive studies of incident investigations and personnel security or innocent person has not been correctly iden­ below: the control question technique in criminal inves­ screening use of the polygraph; second, for poly­ tified. Exclusion of inconclusives would raise the tigations, there is a wide range of accuracy (and graph countermeasures and for the voluntary na­ • Six prior reviews of field studies: overall accuracy rates calculated. In practice, in­ thus, inconclusive and error) rates. OTA con- ture of testing; and finally, for further research. -average accuracy ranged from 64 to 98 conclusive results may be followed by are-test

percent or other investjgations 9. ______~ ______- - --~ -~------

, .~. \ I'I'

99 98

the percentage of the group of persons being to the way the polygraph is used in criminal inves­ nel for access to highly classified information (Sen­ Relevance to NSDD-84 and Administration Policy screened that has engaged in the criminal (or tigations, there is at least some although far from sitive Compartmented Information and/or special otherwise proscribed) activity. If the percentage While the results of the OTA review indicate conclusive scientific basis for polygraph validity. access). The use of polygraph tests to determine that the control question technique has some va­ of guilty is small, say 5 percent (1 guilty person continuing eligibility would be on an aperiodic out of every 20 persons screened, or 50 out of (Le., irregular) basis (181). These are all known lidity in criminal investigations, there is only a Large-Scale Screening limit~d scientific basis for generalizing the results 1,000), then even assuming a very high (95 per as personnel security applications of the poly­ of the OTA review to the context of NSDD-84 The polygraph test is used by some private cent) polygraph validity rate, the predictive value graph. In addition, administration policy an­ and the October 19, 1983, administration policy firms and on rare occasions by some Federal agen­ of the screening use of the polygraph would only nounced on October 19, 1983, would permit Gov­ on polygraph use. NSDD-84 and administration cies to screen a large number of people in con­ be 50 percent. That is, for each 1,000 individuals ernment-wide use of polygraph tests in person­ policy authorize the use of the polygraph in ad­ nection with the investigation of a crime. Unlike screened, about 47 out of the 50 guilty persons nel security screening of employees (and appli­ ministrative as well as criminal investigations of the typical specific-incident criminal investigation, would be correctly identified as deceptive, but 47 cants for positions) with access to highly classified unauthorized disclosures of classified information. in a large-scale screening investigation, typically out of the 950 innocent persons would be incor­ information. The new policy provides agency the suspect list has not been narrowed down to rectly identified as deceptive (false positives). Thus heads with the authority to give polygraph exam­ First, there is no validity research directly on one or a few persons and only limited informa­ of the 94 persons identified as deceptive, one-half inations on a periodic or aperiodic basis to the use of the polygraph in unauthorized disclo­ tion about the crime is available. would be innocent persons. For every person cor­ employees with highly sensitive access. sure investigations. The subject matter and per­ rectly identified as deceptive, another person NSDD-84 appears to permit such use of the haps subjects of these investigations will vary would be incorrectly identified. Results of OTA Review from the typical criminal investigation as might polygraph in unauthorized disclosure investiga­ the conditions and techniques of testing and use tions, although the actual extent of NSDD-84 is As another ex"'J. ". ': a lower polygraph valid- Personnel security screening involves a different of countermeasures. unclear. Administration policy appears to be am­ ity rate is assum( <;.,.)0 percent), then the pre- type of polygraph test than specific-incident inves­ bivalent. While on the one hand providing guide­ dictive value WOlli _ expected to drop to about tigations, and very little screening research has Second, the investigative conditions authorized lines for "carefully limited use of the polygraph," 33 percent. That is, for every person correctly been conducted. Three studies were cited by the by NSDD-84 and administration policy may be the policy implies that DOD polygraph regula­ identified as deceptive, two persons would be in­ intelligence agencies (NSA and CIA) as providing quite different from conditions under which prior tions are acceptable. DOD regulations have been correctly identified (false positives). support for personnel security use of polygraph research was conducted. NSDD-84 does not speci­ used, albeit infrequently, to authorize polygraph tests. These are, of course, hypothetical examples, fy what type of investigative procedures will be screening of large numbers of individuals (rang­ and have not been systematically investigated in A 1975 field study (6) of polygraph screening followed, how subjects will be selected or iden­ ing from about 2 dozen up to 80) in investigation either field or analog research, although some of government job applicants (from an unidenti­ tified, who will conduct the examinations, or what of specific incidents. question techniques will be used. Administration reviewers (e.g., Ben-Shakhar (28)) have careful­ fied Federal agency) showed high consistency in policy provides some specific guidelines such as There is no scientific basis for generalizing the ly worked through a number of possibilities. Also, readings of physiological arousal by different ex­ requiring that polygraph testing be used only results of the OTA review to establish polygraph operating procedures of Federal agencies (e.g, aminers. But this study concluded nothing about when "other information or means of investiga­ validity in this large-scale screening application. quality control review, consideration of other in­ validity. First, no scientifically acceptable research has been vestigatory information) might catch, correct, or tion have produced a substantial objective basis In a 1981 analog study (43) of preemployment conducted on large-scale specific-incident screen­ minimize erroneous polygraph decisions. for seeking to examine the employee" and there screening use, 75 percent of the responses of ing use of the polygraph. Second, the screening is "no other reasonable means of resolving the Nonetheless, the FBI, which outside of DOD deceptive individuals were detected accurately. conditions here are likely to vary even more from matter" (185a). However, in general, the extent and CIA, is the principal Federal agency that con­ Twenty-five percent were detected incorrectly. the conditions of the research studies reviewed by to which employees will be requested or required ducts polygraph examinations, believes that large­ Any conclusions based on this study must be lim­ OTA. For one thing, much less information is like­ to take polygraph examinations in unauthorized scale screening is not an appropriate use of poly­ ited by the fact that the subjects were students, ly to be known about circumstances surrounding disclosure investigations is largely left to the graph testing. FBI regulations prohibit the "use the questions and context had nothing to do with an unauthorized disclosure and possible suspects. discretion of agency heads. of the polygraph for dragnet-type screening of national security, and the test format was atypical This could translate into differences in the ques­ large numbers of suspects or as a substitute for of personnel screening examinations. Third, even the Federal Bureau of Investigation tions used, the behavior of the polygraph exam­ logical investigation by conventional means" [FBI (FBI) has concluded that, "to date, no methodo­ iner, the motivation and response of the subject, Polygraph Regulation 13-22.2 (2), 1980]. A 1980 survey conducted by the Director of the logically adequat~ study of control question tech­ and the effectiveness of countermeasures. Central Intelligence Security Committee con­ niques has been reported .... Inferences regard­ Third, the large-scale screening use of poly­ cluded that the polygraph was the most produc­ ing the validity of control question examinations Personnel Security Screening tive of all background investigation techniques. ... rest upon the results of laboratory studies graph testing theoretically can be expected to result in significantly higher error rates than when However, this was a utility study not a validity conducted under highly dissimilar conditions." Draft revisions to the DOD polygraph regula­ the list of suspects is narrowed down to a very study, and had many limitations and qualifica­ The FBI is planning its own validity research. tions would authorize the use of polygraph tests small number, as in a typical criminal investiga­ tions. For example, the criteria for case selection to determine initial and continuinb eligibility of On the other hand, to the extent polygraph use were not stated and there was no independent tion. The screening use of polygraph tests is most DOD civilian, military, and contractor person- in unauthorized disclosure investigations is similar dependent on the so-called base rate of guilt, i.e., verification of the cases that were resolved. Also,

u.....-'I'__ ---o.------~~-~~--.------~ ------

I .,,,,,,,,,,,,,,,",- i .~ I

101 100

sanctions and denial of security clearance. And the polygraph was used only after a thorough in­ ce.nt polygraph validity is assumed, there would experience in passing an exam. The research on administration policy authorizes denial of clear­ vestigation based on other sources had taken place stIll be 10 false positives for every correct detec­ polygraph countermea.sures has been limited and ance, transfer or reassignment, and, under some (see ch. 4 for further discussion). tion of a guilty person. the results-while conflicting-suggest that validi­ circumstances, termination of employment for re­ ty may be affected. Further, some research (e.g., fusal to take a polygraph test. OT~ inquiries to all DOD components using Again, these are hypothetical examples that 75) suggests that polygraph examiners may not the pOlygraph identified only one DOD research have not been systematically studied in field or be able to easily detect certain physical counter­ Under these conditions, polygraph examina­ study on personnel screening use of the polygraph analo~ research. NSA claims that they in fact have measures. The research results for drug and psy­ tions would not be voluntary in the strict sense, (!6). The results of this study raise more ques­ ~xpenenced a very low false positive rate and that, chological countermeasures are mixed. The possi­ since a refusal could result in penalties. Apart tI?ns than they answer, and certainly do not pro­ In any event, polygraph test results are only one ble effects of countermeasures are particularly from the ethical and perhaps legal implications, vIde s~pp~rt f~r high polygraph validity in a ~actor in making decisions and are subject to qual­ significant to the extent that the polygraph is used which OTA did not address, conducting poly­ screenll;g sItua:lOn. The limitations of the study Ity control checks and other reviews. It appears and relied on for national security purposes, since graph tests on this basis could affect test validi­ reduce Its apphcability, but it is the only DOD that NSA (and possibly CIA) use the polygraph even a small false negative rate could have serious ty. It is generally recognized that, for the poly­ polygraph screening research known to OTA not to determi~e deceI;'tion or truthfulness per se, consequences. In addition, those individuals who graph test to be accurate, the voluntary coopera­ OTA inquiries to other executive agencies and but as a techmque of Interrogation to encourage the Federal Government would most want to de­ tion of the individual is important. For example, departments using the polygraph identified no admissions. NSA has stated that the agency "does tect (e.g., for national security violations) may NSA has stated that, in conducting screening ex­ research on personnel security screening use of the n~t use the :tru~h v. deceptive' concept of poly­ well be the most motivated and perhaps the best aminations, "[t]he full cooperation of the individ­ polygraph. gLaph examInatIons commonly used in criminal trained to avoid detection. ual taking the test is essential or the results will cases. Rather, the polygraph examination results OTA recognizes that the administration as well be inconclusive." The polygraph only detects that are most important to NSA security adjudi­ physiological arousal, and under involuntary con­ as NS~, CIA, and DOD believe that the poly­ Voluntary v. Involuntary ~ators are the data provided by the individual dur­ ditions, the arousal response of the examinee may graph IS a useful screening tool. However, OTA Ing the pretest or posttest phase of the examina­ As currently used in the Federal Government, concluded that the available research evidence be very difficult or impossible to interpret. How­ tion" (187). with few exceptions, polygraph examinations are ever, no direct research on this topic was iden­ does not establish the scientific validity of the voluntary. That is, a person cannot be forced to The validity of the polygraph as used by NSA tified. Overall, OTA concluded that imposing polygraph for this purpose. take a polygraph test against his or her will. A has not been researched. And, in general this kind penalties for not taking a test may create a de facto refusal to take a polygraph test does not, or at In comments to OTA, CIA agreed that the cu­ of application is potentially different i~ so many involuntary condition that increases the chances mulative unclassified research evidence reviewed least is not supposed to, result in adverse conse­ of invalid or inconclusive test results. w?y~ fro~ the polygraph use in specific-incident quences. The only exceptions are NSA (and by by OTA is not directly relevant to national securi­ crImInal. InvestIgations (e.g., with respect to type ty applications. However, CIA does claim to have extension, CIA) and, under certain conditions, the of questIons asked and question techniques em­ FBI. NSA notes that "the polygraph examination Further Research classified research to support their use of poly­ ployed) that results of the OTA research review graph tests. OTA did not review this research. is part of the Agency's security processing. Failure OTA concluded that, to the extent that poly­ previously discussed cannot be generalized to the to complete processing may result in failure to be No. other Federal agency! including NSA, has NSA situation. graph testing is going to continue to be used by claImed to have relevant research results that were accepted for employment" (187). FBI regulations the Federal Government, further research is not .available for OTA review on an unclassified require that "polygraph examinations will be ad­ needed. Possible research priorities include the False Negatives/Countermeasures ministered only to individuals who agree or basIs. following. The primary purpose of polygraph testing volunteer to take an examination" [FBI Regula­ tion 13-22.2(3}]. The only exception is for certain False Positives under N~I?D-8~, the DOD revised regulations, Polygraph Theory and ad~mstratIon policy is to detect persons who FBI employees and apphcants under specified cir­ One area of special concern in personnel securi­ ~?ve or Intend to participate in proscribed activ­ .~".mstances where "a refusal to be examined by The basic theory of polygraph testing is only ty screening is the incorrect identification of in­ ItIes (e.~., unauthorized contact with a foreign polygraph may lead to an adverse inference be- partially developed and researched. The most ~ocent persons as deceptive. All other factors be­ agent, ?isclosure of classified information). A con­ ing drawn." commonly accepted theory at present is that, g e~ual, the low base rates of guilt in screening u: cez:n Wlt?false negatives (guilty persons incorrect­ when the person being examined fears detection, SItuatIons would lead to high false positive rates The DOD proposal would provide that refusal ly IdentIfIe~ as nondeceptive) is that, apart from that fear produces a measurable physiological even assuming .very high polygraph validity. Fo; to take a polygraph examination, when estab­ any errors Inherent in the polygraph test itself reaction when the person responds deceptively. ex~mp~e, a typIcal polygraph screening situation lished as a requirement for selection or assignment the guilty person may be able to escape Thus, in this theory, the polygraph instrument is mIght Involve a base rate of one guilty person detectio~ or as a condition of access, may result in adverse through the use of countermeasures. measuring the fear of detection rather than decep­ (e.g., one person engaging in unauthorized dis­ consequences for the individual. These include tion per se. And the examiner infers deception closure) out of 1,000 employees. Assuming that Theor~ti\cally, polygraph testing-whether for nonselection for assignment or employment, de­ when the physiological response to questions th: polygraph is 95 percent valid, then, the one ?erso~ne~ security screening or specific-incident nial or revocation of clearance, or reassignment about the crime or unauthorized activity is greater guilty person. would be identified as deceptive but InveshgatIons-is open to a large number of coun­ to a nonsensitive position. NSDD-84 also provides than the response to other questions. However, so would 50 Innocent persons. The predictive va­ termeasures, including physical movement or that refusal to take a polygraph test may result this theory has been challenged by some psycholo- lidity would be about 2 percent. Even if 99 per- pressure, drugs, hypnosis, biofeedback, and prior in adverse consequences such as administrative 102 gists and others who believe that various factors research design is as scientifically sound as possi­ -e.g., the examinee's intelligence level, psycho­ ble. Also, the U.S. Army's current 10-year re­ logical health, emotional stability, and belief in search program to develop a new state-of-the-art the "machine" -may, at least theoretically, affect polygraph instrument should be reevaluated to the physiological response. determine if research priorities and direction need adjustment. As it stands now, validity issues will OTA concluded that a stronger theoretical base not be addressed by the Army research until the is needed for the entire range of polygraph appli­ late 1980's. cations, incbding current and proposed Federal Government applications. Basic polygraph re­ Personnel Security Screening VaJidity search should consider the latest research from the Appendixes fields of psychology, physiology, psychiatry, Given the almost total lack of res~arch on this neuroscience, and medicine; comparison among application, further research is clearly necessary question techniques; and measures of physiologi­ if there is to be any possibility of establishing a cal response. scientific basis for the personnel security screen­ ing use of polygraph testing. (I Criminal Investigation Validity Research on Polygraph Countermeasures There are still many unanswered questions about the validity of use of the polygraph in spe­ Since NSA and CIA are: already heavily de­ cific-incident criminal investigations. A planned pendent on the polygraph, their use alone justifies FBI-Secret Service validity study is intended to an intensified research effort on countermeasures. meet this need. However, OTA did not review NSA and the U.S. Army Intelligence and Securi­ the research plan, which would benefit from an ty Command are planning such research, but the independent review by the scientific community level of effort appears low (e.g., $65,000 pilot ) I and others before the research approach is final­ study in NSA) considering the consequences of ;/ ized. Such a review would help ensure that the false negatives. G

CONCLUDING COMMENT A major reason why scientific debate over poly­ The focus of the OTA technical memorandum graph validity yields conflicting conclusions is that is not whether the polygraph test has been useful, i1 ) ( the validity of such a complex procedure is very but whether there is a scientific basis for its use. H I; n! '.J difficult to assess and may vary widely from one OTA concluded that, while there is some evidence ~ application to another. The accuracy obtained in for the validity of polygraph testing as an adjunct Ij 1 I one situation or research study may not generalize to typical criminal investigations of specific in- i j ) to different situations or to different types of per­ cidents, ahd more limited evidence when such in- J sons being tested. Scientifically acceptable re­ vestigations extend to incidents of unauthorized I search on polygraph testing is hard to design and disclosure. However, there is very little research I, conduct. or scientific evidence to establish polygraph test validity in large-scale screening as part of un- authorized disclosure investigations, or in person- I Advocates of polygraph testing argue that thou­ nel security screening situations, whether they be sands of polygraphs have been conducted which preemployment, preclearance, periodic or aperi- (J " ./.1 " substantiate its usefulness in criminal or screen­ odic, random, or "dragnet." Substantial research ing situations. Claims of usefulness, however, are beyond what is currently available or planned 1\ [J often dependent on information (e.g., confessions would have to be conducted in order to fully II,1 .. , and admissions) obtained before or after the ac­ assess the scientific validity of the NSDD-84, i1 ~. ~\ tual test, and on its perceived value as a deterrent. DOD, and administration polygraph proposals. 1\ I. Ii .. ' . 0 .. I b· , Mt .i 'lilt'" f ~~ ----- ~------r -----~-..------...------I 1---- I. r·

Appendix A Informed Consent Forms

POLYGRAPH EXAMINATION WAIVER

.Place: ______

Date & Time: ______

1,--____------have been requested by Special Agent ______of the Naval I nvestigative Service

to submit to a polygraph examination relative to my (knowledge of) (participation in)

With respect to that request, I have been advised:

(a) that I have the right to consult with a lawyer prior to making any decision concerning the examination; ___ (b) that the polygraph examination will be conducted only with my prior written consent; _____ (c) that no adverse action will be taken against me solely on the grounds that I refuse to consent to this examination; _____ (d) that the area in which the examination is to be conducted (does) (does not) contain a two-way mirror or similar device; ____ (e) that the area in which the examination is to be conducted (does) (does not) contain a camera; _____ (f) that the are.a in which the examination is to be conducted (does) (does not) contain an electronic audio recording device and the polygraph examination (will) (will not) be monitored. ______

With an understanding of the above conditions, I have decided that I do not desire to consult with a lawyer at this time. I freely consent to be examined by polygraph and I agree to cooperate fully with the examiner during that examination. I make these decisions freely and voluntarily and they are made with no threats having been made or :)romises extended to me.

(Signature) Time: ______

Witnessed:

NISFORM 01o-F/04·80 GPO BB~-411 105 !1\ ~------.------~ ----- ~ .~--

106 Appendix B

POLYGRAPH EXAMINATION For u •• of .hlG form. .. 8 AR 195..6; the proponent openey Results of the OTA Survey of STATEMENT OF CONSENT I. U.S. Army Criminal Invullpatlon Command. PLACE OATE TIME Federal Government Polygraph Testing

ST A TEMENT OF CONSENT OF (N"m~, G, .. d~, "nd SSN) PI.ACE OF BIRTH Introduction However, long-term trends in the number of poly­ OATE OF BIRTH graph examinations show a substantiai increase since In May 1983, OTA surveyed selected Federal Gov­ 1973. In fact, the total number of examinations in 1982 ernment agencies including the Departments of State, ( Strike out inapplicable portions indicated in parentheses. ) was more than triple the 1973 total (22,597 examina­ Defense (DOD), Treasury, and Justice, Central Intel­ tions in 1982 compared to 6,946 in 1973) and actually In the presence of the witness(es) whose signature(s) appear(s) below, Article 31 of the Unifonn ligence Agency (CIA), Office of Personnel Manage­ surpassed the previous known high (19,796 in 1963, Code of Military Justice and The Fifth Amendment to the Constitu tion of the United States have ment, and U.S, Postal Service (USPS), with respect excluding NSA). As illustrated below, the FBI, Air been explained to me by to their use of polygraph testing. The survey requested Force, and NSA experienced the largest absolute in­ detailed information about agencies' current and past who infonned me that he is a polygraph examiner of the United States Army. He has informed me creases in polygraph examinations over the 1973-82 use of polygraph testing and research conducted or period. that this statement is being completed in connection with planned by the agency. The request for information was sent to all Federal agencies believed to conduct He explained to me the nature of the polygraph examination and told me that I cannot be required to polygraph examinations. A follow-up survey was sent, Number of examinations conducted take such examination without my consent. He explained to me that I have the right to consult with in July 1983, with respect to use of polygraph testing Agency Fiscal year 1963 Fiscal year 1973 Fiscal year 1982 in unauthorized disclosure investigations. Army CIC ...... '... 4,4oo} 2,028 3,731 counsel prior to this examination and to have counsel present to observe the examination. He explained Army ISC ...... ' 8,094 279 Results of the survey are described below. All agen­ Navy...... 1,200 665 1,337 that this counsel may be civilian counsel retained at my expense, or counsel appointed for me at no ex­ Air Force ...... , ...... 1,912 482 3,019 cies responded to OTA's inquiry; however, the CIA Marines ...... 812 62 :<63 pense to me (,or if a member of the United States Armed Forces, that I may select military counsel of considers all such operational and research informa­ NSA , ..••.•.•..•... ,. Not available 3,081 9,672 Other DOD...... 140 6 0 my choice if such counsel is reasonably available). I (do) (do not) want to consult with counsel prior tion to be classified. In addition, the results do not in­ ~ DOD subtotals . , •..• , . 16,558 6,325 18,301 to this examination. I (do) (do not) want to have counsel present to observe the examination. I was II clude information from the Customs Service (a Depart­ FBI. ... , ...... 2,314 79 2,463 ~ ment of the Treasury component), Department of DEA ...... , ...... 211 further advised that the examination room (does) (does not) contain a two-way mirror or observation /1 SS ...... 65 SO 714 Health and Human Services, and Tennessee Valley BATF ' ...... 256 device and that the examination (will) (will not) be monitored or recorded. He explained to me that I do Authority, which conduct a limited but unknown USPS, .. , ...... 338 485 652 Other ., ...... 521 7 0 not have to make any statement whatsoever but that any statement I do make may be used as evidence , number of polygraph examinations. OTA supple­ Totals ..•...... •. ,.,. . 19,796 6,946 22,597 ~! against me in a trial by court-martial (if subject to the Uniform Code of Military Justice) or in any " mented the survey results with site visits to polygraph SOURCE: Data from the Office of Technology Assessment, 1982; 1973 and 1963 data from U.S. Congress, House of Representatives, Committee on Government " facilities at the U.S. Army, National Security Agency other military or judicial proceedings. Understanding my unqualified right to refuse, I, I' Operations, reports, The Use of Polygraphs and Similar Devices by Federal understand that I will be questioned prior to, during and after the ! ,/1 (NSA), and Federal Bureau of Investigation (FB!), and Agendes, 1976 and Use of Polygraphs as Ue Detectors by the Federal Govern­ ! II discussions with officials from several Federal agency ment,1965. instrument portion(s) of the polygraph examination a~d . ~ Ii do hereby, this date, voluntarily and without duress, coercion, unlawful mducement, or promise of ~ , polygraph programs. j~ 1; reward, consent to a polygraph examination. H I Ii Number of Polygraph Examinations I II Number of Polygraph Examiners WITNESSES SIGNATURE OF PERSON TO BE TESTEO I " 11 SIGNATURE (I For 1982, the agencies reported conducting a total For 1982, agencies reported employing a total of 209 I of 22,597 individual polygraph examinations. Of this polygraph examiners, Of these examiners, the majority total, 18,301 examinations were conducted by DOD (130) were employed by DOD component agencies. TYPEO NAME ORGANIZATION ANO/OR ADDRESS ORGANIZATION ANO/OR AODRESS component agencies, including the Army, Navy, Air Individual agency totals are shown in table B-1. The Force, Marines, and NSA. Individual agency totals are U.S. Army has the largest number of examiners, fol­ shown in table B-1. NSA conducts the largest number SIGNATURE lowed closely by the FBI, and then by the U.S. Air of examinations, 43 percent of the total. Next, in terms Force and NSA. The reason that the number of exam­ SIGNATURE 01'" EXAMINER of number of tests, is the Army Criminal Investiga­ iners is not directly related to the number of examina­ TYPEO NAME ORGANIZATION ANO/OR AOORESS tion Command, followed by the Air Force Office of tions is that examinations are conducted by agencies Special Investigations, Naval Investigative Service, for different purposes and under different conditions. and FBI. The NSA and the Air Force have steadily in­ J..------,-----1 TYPED NAME ANO ORGANIZATION For example, NSA examinations are conducted for EXHIBIT HUMBER creased the number of examinations conducted each screening purposes in a central location; in contrast, year during the 1980-82 period, while the number of Army examinations are conducted primarily as part polygraph examinations appears to be relatively stable of criminal investigations, and examiners frequently DA FORM 2801 I'tEPI.ACES EDITION 01'" , JAN H, WHICH IS OBSOI.ETE. over this period in other agencies. travel to sites within a geographic region. 1 AUG 17 ,"u.s. Gove'n"",-nt ",lnUn, OfrlCit 1'.0-:110."'1'012 · \

108 109

Table B·1.-Number of Polygraph Exams and Examiners Table B·2.-Purpose of Polygraph Exam

Number of polygraph exams Number of examiners Criminal Counter Agency Fiscal year Fiscal year investigation intelligence Intelligence Other 1980 1981 Department of Defense: 1982 1980 1981 1982 Department of Defense; . Army Criminal Investigation Army Criminal Investigation Command .•...... 3,977 3,832 3,731 39 42 Command ...... 1980 3,968 9 polygraph examiner applicants Army Intelligence and Security Command ...... 230 44 Naval Investigative Service ...... , , ...... ••...... 260 279 9 9 12 1981 3,820 12 polygraph examiner applicants 1,317a 1,185a a Air Force Office of Special Investigations ...... 1,33,a 11 12a 14a 1982 3,713 19 polygraph examiner applicants 1,474a 1,485a 3,019a a a a Marines ...... '" 8 26 29 28 Personnel Limited Polygraph 376 245a 263a a a a National Security Agency ...... a a 8 6 6 security access applicants 5,676 7,418 9,672a 13a a a Subtotals ...... " ...... •...... -- 30 26 Army Intelligence and 13,050 14,425 18,301 106 -- Security Command ...... 1980 NA 44 NA o 58 5 Department of State ...... •...... 128 130 Does not conduct polygraph exams 1981 NA 33 NA 9 34 1 Department of Justice: 1982 NA 78 NA 58 62 2 Navy ...... 1980a 30 Federal Bureau of Investigation ...... 2,121 1,209 78 2,162 2,463 NA NA 1981 a Drug Enforcement Administration ...... 230 40 1,049 50 86 Department of Treasury:b ~OO 211 NA NA 14 1982a 1,210 45 82 a Secret Service ...... Air Force ...... 1980 1,296 NA NA NA 466 a Bureau of Alcohol, Tobacco and Firearms ...... 714 NA NA 16 1981 1,298 NA NA 176 254 a 256 4 4 4 1982 1,750 NA NA U.S. Postal Service ...... Marines NA Polygraph examiner applicants 714 725 652 NA ...... Central Intelligence Agency ...... NA 5 National Security Agency ...... NA NA NA Applicant screening Does conduct polygraph exams but specific opera. tional information is classified Department of Justice: Offico of Personnel Management ...... Federal Bureau of Investigation 1980-82 6,038 474 234 personnel security Does not conduct or use polygraph exams TotallS ...... Drug Enforcement Administration ... 1980-82 449 192 internal investigations Department of the Treasury: 16,291 18,232 8Calendar year. 22,597 209 Secret Service ...... 1982 562 65 59 other agency bExcludes Customs Service. 16 bond NA = Not available 12 inspection Bureau of Alcohol, Tobacco and Firearms ...... •...... 1980-82 686 U.S. Postal Service ...... 1980-82 2,091

Other Federal Agency Polygraph Users DEA .. '" ...... Immigration and Naturalization Service •• 2 (1981-1982) aCalendar Year. NA = Not available. U.S. Marshall's Office ..•.. " ...... •• 3 (1981-1982) Department of State ...... •...• " 2 (1981-1982) The Federal agencies listed in table B-1 are the pri­ Internal Revenue Service • .••.•....•..•. 1 (1981-1982) mary users of the results of polygraph tests conducted Secret Se.vice ...... •. Internal Revenue Service . .. '" ..•..• } Specific data not graph examinations are for applicant screening! i.e., by their personnel. However, these agencies reported U.S. Attorney's Office ...... •. " . . available. but Use of Polygraph in Unauthorized Department of Treasury...... • total is less than 8 for use in personnel security evaluations of applIcants Department of Agriculture. . . • ...... percent of all Disclosure Cases that during 1980-82, polygraph examinations were also for employment. In 1982, OTA estimates that NSA conducted by their staff for other Federal agencies, Federal Reserve Bank. . . . • ...... • . Secret Servke exams. conducted about 6,700 applicant screening polygraph Polygraph exams are use? by s~ve~al Federal agen­ both those with polygraph capability and those with­ BATF .•..•...•..••.• Other Agencies (very limited) ..•...... Data not available USPS ...... Internal Revenue Service...... • .. 4 exams. No other Federal agency, except CIA, conducts cies in connection with the mveshgatI0r: .of t~e unau­ out. A listing of the number of examinations conducted U.S. Marshall's Office...... 1 routine applicant screening polygraph ~xams. ~IA, as thorized disclosure of sensitive or claSSIfIed mforma- for agencies that do not employ their own polygraph U.S. Congress...... 1 noted above, did not provide informat~o~ on b1e p.ur­ tion; however, such use at present is limited. . staffs follows: The polygraph use by these other agencies represents pose of their exams. However, publIc mformatIon Federal agencies responding reported the followmg a very small percentage of total Federal agency use. polygraph use in unauthorized disclosure cases over NUllIberof available from a report of the Permanent Select C?m­ Exams conducted by Exams conducted for exams 1980-82 the 1980-82 period: Army. CIC ...... Department of State ...... mittee on Intelligence, U.S. House of Representatives 26 Number of polygraph examinations (1980-82) Internal Revenue Service . .•.•••...•.••• Purpose of Polygraph Examinations Agency 1 (173), indicates that the CIA ~tilizes polygraph tests Very rew Defense Investigative Service . ...•....•. Army. CIC ...... 1 as part of its applicant screemng. Army.ISC ...... 1 Department of Defense (other) .....••... 14 Army. ISC ...... Defense Intelligence Agency .•...... •.• As shown in table B-2, with the exception of NSA, Navy ...... •.. 78 7 Navy ...... Coast Guard ...... over two-thirds of Federal agency use of the polygraph The following agencies also conduct a small num~er Air Force ...... 112 1 Marines ...... o General Services Administration ...•••.. 1 is for criminal investigative purposes. In the major Fed­ of polygraph exams for counterintellige~ce and/ or m­ NSA •...... Data not available Department of State ...... 2 Air Force...... Defense Investigative Service ...•...... telligence purposes (see table B-2 for estimates!: Army State Department ...... o 16 eral polygraph user agencies, such as the Army, Navy, FBL ...... 26 (since 1978) Defense Intelligence Agency •...... •.. 21 Security and Intelligence Command, Na~, AIr Force, 33 Coast Guard ...... Air Force, and FBI, over 90 percent of polygraph use DEA ...... 1 11 Department of State ...... NSA, FBI, and Secret Service. Ot~er n:Iscellaneous Secret Service ...... Marines ...... None 1 is for criminal investigations, for example in the veri­ BATF ...... o NSA ...... •...... DOD components ...... Data not available fication of information provided by suspects, victims, purposes for polygraph exams are lIsted m table B-2. USPS ...... o FBI ...... Bureau of Prisons...... 39 (1982) and witnesses. The one exception, for which data are Other Agencies. • .. . . • ...... •. 10 per year available, is NSA. About two-thirds of NSA poly- . --"."'-. , "'~4 \ \

110 111

For agencies providing detailed statistics, the results of the exams were as follows: quently used. The Army Intelligence Command, FBI, Table B·3.-Length and Results of Exams DEA, Secret Service, and BATF also use the rele­ Not No Deceptive Deceptive Average length of exam Results of exams deceptive Incondusi"e opinion vant/irrelevant technique to a limited extent. All agen­ Anny. ISC •...•• 0 confinned 1 0 0 Navy ...... 26 0 cies reported that examiners use a standardized numer­ 51 1 (j Fiscal Fiscal Fiscal Fiscal year 1980 percent Fiscal year 1981 percent Fiscal year 1982 percent Air Force ...... 26 18 85 1 0 ical scoring system for interpreting results of exams year year year FBI .....•...... 16 21 10 0 0 1980 1981 1982 NO NO o NO NO NO NO DEA ...•...... 2 14 conducted with a control question technique. For o o 31 0 0 Secret Service .... 0 Data not available Department of Defense: 11 0 0 exams conducted with the relevant/irrelevClllt tech­ 0 Army Criminal Confirmation of deceptive exam results was primari­ nique, the examiner looks for significant, consistent Investigation ly though a pre- or post-test confession or admission. reactions. See chapter 2 for further discussion of ques­ Command ...... 2:53 2:54 3:03 68.7 21.4 0.6 9.2 66.3 24.1 0.9 8.7 64.0 28.3 0.6 7.1 tion techniques. Arm:; Intelligence and Very few of the not deceptive test results were con­ Security Command 4 hours average 34 59 4 3 33 62 2 3 34 59 3 4 firmed. Except for the FBI, information was not avail­ Air Force .....••...... 50 50 50 50 50 50 able on what action, if any (€.g., administrative sanc­ Marines ....•...... At least 1.5 hours 51 NA NA NA 51 NA NA NA 52 NA NA NA Methods of Quality Control National Security tion, removal of security clearance, criminal prosecu­ Agency •...... 1.5 to 2 hours average NA NA NA NA NA NA NA NA NA NA NA NA tion), was taken based on the deceptive exam results. All Federal agencies reported that essentially the Department of Justice: The FBI reports that in 12 closed cases, deceptive ex­ same polygraph instruments and physiological meas­ Federal Bureau of Investigation ....•.... NA 3,527 deceptive out of 6,646 total exams for FY 80-82 amination results contributed (at least in part) "I) 3 con­ ures are employed in conducting polygraph examina­ Drug Enforcement victions, 1 dismissal, 1 diSciplinary action, 3 resigna­ tions. All Federal agencies use primarily Stoelting and Administration ...... 2 to 3 hours average 171 deceptive out of 641 total exams for FY 80·82 tions, 3 censures, and 1 voluntary retirement. 464 NO Lafayette polygraph instruments (purchased from pri­ 4 I vate manufacturers). The physiological measures in­ Department of the Treasury: Polygraph Examiner Training clude respiration (breathing), perspiration (galvanic Secret Service ...... NA 46.8 46.9 4.5 6.8 3 year average and Techniques skin response), and cardiovascular (blood pressure and Bureau of Alcohol, pulse rate). Tobacco and Firearms -...... 3:49 3:47 3:37 51.1 37.5 7.4 4.0 40.2 47.2 5.5 7.1 28.5 57.4 6.3 7.8 Federal agencies reported a high degree of consisten­ Agencies also indicated that all polygrams (charts) U.S. Postal Service ...... • 1.86 hours average 11 83 4 1.0 10 83 4 2 17.3 73.1 6.8 1.2

cy in the training of and techniques used by Federal are reviewed independently by a supervisor and/ or a NOTES: o a Deceptive polygraph examiners. All agencies, except NSA, re­ polygraph coordinator at a headquarters location. This NO - Nondeceptlve I a Inconclusive ported that examiners are required to be graduates of quality control review includes checking the original NO a No Opinion the 12-week U.S. Army Polygraph Training Course examiner's chart interpretation as well as reviewing NA a Not available at Ft. McClellan, Ala. (a component of the U.S. Army question construction and other aspects of the exam. Military Police School). NSA requires examiners to be Agencies vary in the specifics of their quality control Table B·4. Long·term Confirmation of Exam Results (percent confirmed) graduates of either the U.S. Army School or the Mary­ process, but any disagreement between the chart in­ Fiscal year 1980 Fiscal year 1981 Fiscal year 1982 (incomplete) land Institute of Criminal Justice. All examiners are terpretations of the original examiner and quality con­ D ND D ND D ND trol examiner usually requires a reexamination. required to have at least 2 years investigative experi­ Department of Defense: ence. USPS requires 3 years investigative experience, Army Criminal Investigation 20 9 68.8 20.9.. 72.8 the Secret Service requires 4 years investigative experi­ Length of Polygraph Examinations 49.3 ence, and the Navy, FBI, and BATF require 5 years. Command ...... D ~~.4 tion via confession court conviction, legal determmatlon via legal or location of other suspec t In addition, all examiners are required to have an Agencies reported that the length of polygraph ex­ N~~~n~~~ation det~rmination Army Intelligence and 70 82 undergraduate degree from an accredited college. The aminations ranges from about 1.5 to 4 hours, as indi­ 83 Drug Enforcement Agency (DEA), Secret Service, and cated in table B-3. Security Command ...... Confirmation primarily via examinee 4~dmission BATF require examiners to participate in an advanced 42 46 N~vy ...... 50 10 3 year average or refresher course every year; DOD components and Results of Examinations and Air Force ...... Confirmation by other evidence the FBI require such participation every 2 years; and 24 25 USPS requires such participation every 3 years. All Subsequent Confirmation ~:;il~~:I'S~~~~itY 'Ag~~~y': :Data not available examiners are required to complete an internship or Department of Justice: probationary period after graduation from polygraph The results of polygraph examinations vary widely school. among Federal agencies. The number of deceptive ex­ FederalInvestigation Bureau of...... , 1 966 0 f 3527, deceptives confirmed by confession With respect to examiner technique, examiners at amination results ranges from about 10 percent of total exams (for USPS in 1981) to abcut 69 percent (Army Drug Enforcement 0 dece tives confess during post-tes~ int~rrogation all agencies reporting except NSA make primary use Administration ...... '~~~:o~f ND c6nflrmed by subsequent Investigations of one or more control question techniques. The mod­ Criminal Investigation Command, 1980), with most ified general question and zone of comparison are the agencies in the 40 to 60 percent deceptive range. See Department of the Treasury: r 90% of opinions are confirmed . most frequently used control question techniques. Ex­ table B-3 for agency specific statistics. Secret Service ...... ?;~o of D confirmed by admissions or confeSSions aminers at most agencies also use the peak of tension Confirmation of results also varies widely, as shown Bureau of Alcohol, Tobacco II bl technique (a concealed information technique). At in table B-4. Independent confirmation rates for decep­ and Firearms ...... Data not ava a e NSA, the relevant/irrelevant technique is the most fre- tive exam results range from about 25 percent for the U S Postal. Service...... ~%~Df d 36% of D 39% of D Marines to 70 to 80 percent for the Army Criminal In- . . con esse confessed confessed ,.~-

\I

112 113

vestigation Command, Army Intelligence Command, cific investigation. In fact, agency regulations generally and Secret Service. Confirmation of deceptive exam require that polygraph results "be used selectively as Table B·S.-Selected Federal Agency Polygraph Research results is primarily by examinee admissions or confes­ an investigative aid" and not lito the exclusion of other Past polygraph research Present/future research sions. Confirmation of nondeceptive exam results is evidence or knowledge obtained during the course of Department of Defense: generally more difficult, with nondeceptive confirma­ a complete investigation" (FBI regulation 13-22.2(2), Army Criminal Investigation tion rates of less than 50 percent indicated by all agen­ 1981). Command ...... 1965·67 Validation study of polygraph ex- 1981-90 Instrument research and cies reporting except DEA. aminer judgments (reliability development project (to develop study known as "Bersh" study a state-of-the art polygraph Federal Agency Polygraph Research instrument) Use of Polygraph Examination Results 1966-67 Calibration study of polygraph Based on information provided by Federal agencies, instrument In general, with the exception of NSA, polygraph the major past, present, and future Federal polygraph test results are used as an investigatory tool in specific research is summarized in table B-5. Research on the 1973 Comparison of voice analysis criminal, counterintelligence, intelligence, or person­ and polygraph (U.S. Army Land polygraph instrument itself includes a 1966-67 calibra­ Warfare Laboratory) nel security cases. Polygraph examinations are volun­ tion study (U.S. Army), a 1966-67 technical evalua­ Army Inteliigence Command .. 1979-81 Validation and reliability study of 1981-90 Instrument research and tary L" the 5Cnse that agencies in general are proscribed tion study (Navy under contract to National Bureau counterintelligence screening development project from forcing individuals to take an examination, or of Standards), 1969-70 and 1975-77 cardioac:tivity test Planning research on polygraph from penalizing or taking adverse action against indi­ Navy ...... 1966-67 Technical evaluation study of None countermeasure and monitor studies (Air Force), a current cardioactivity polygraph instrument (under con- anticountermeasures viduals who refuse to take an examination. However, monitor study (FBI), and the current 10-year instru tract to National Bureau of at NSA, where a polygraph examination is part of the mentation research sponsored by the Army Criminal Standards) preemployment security screening process for all job Investigation Command and Army Security and Intel­ Air Force ...... 1962 Polygraph validity study None applicants, refusal to take polygraph examination may ligence Command and intended to develop a new poly­ 1965 Analysis of polygraphic data result in failure to be accepted for employment. Also, 1969-70 Development and validation graph instrument utilizing state-of-the-art technology. studies of cardioactivity monitor the FBI noted that in cases where an FBI employee is Research on polygraph validity and reliability, broadly 1975-78 Reliability and validity studies of asked to take a polygraph examination but refuses, the defined, includes a 1962 validity study (Air Force), a cardioactivlty monitor refusal may lead to an adverse inference being drawn. 1965-67 reliability study (Army Criminal Investiga­ Marines...... None None Overall, agencies were not able to provide specific National Security Agency ..... 1983 Review of scientific literature on 1983-84 $65,000 pilot study of effect of tion), 1979-81 counterintelligence screening test va­ polygraph validity, reliability and drugs/hypnosis/nonverbal tech­ information on how the results of polygraph exams lidity st'.-dy (Army Intelligence), and the planned 1984- utility niques on polygraph validity were actually used, since the agency office conducting 85 validity and reliability study cosponsored by the Department of Justice: the examination is usually different from the office FBI and Secret Service. Also, in 1976-78, the Depart­ Federal Bureau of conducting the investigation and taking action. Sta­ ment of Justice sponsored validity and reliability Investigation ...... None 1984-85 Polygraph validity and reliability tistics on use of examination results apparently are not studies by university researchers David Raskin and research (criminal investigatory maintained, at least not on a centralized basis. Also, context) David Lykken. Finally, both Army Intelligence and 1984 Instrumentation research (on the results of a polygraph examination are usually only NSA are planning research on polygraph counter­ monitoring blood pressure) one of several sources of information relevant to a spe- measures. Law Enforcement Assistance Administration ...... 1976-78 Raskin and Lykken studies of polygraph validity and reliability Department of the Treasury: Secret Service ...... Participated in Raskin study Cooperation with planned FBI study on polygraph validity and reliability Bureau of Alcohol, Tobacco and Firearms ...... None None U.S. Postal Service ...... None None ...:.D..:..e..:..p-art.....:m-=--e-nt-o-f:-S-t-a-te-.- ..-.-. -. -.. -.-. -. ---7:'None None Office of Personnel Management ...... None None Central Intelligence Agency ..... Classified research Data not available .. *"7111f ~------

,-i

Appendix C Coding Form 115 Polygraph Coding Form Page 2

DRAFT CODING FORM 7/18/83 POLYGRAPH CHARACTERISTICS Coder BASERATE, Base rate of guilt AUT HOR:..... ______--- GROUNDME, Method of establishing ground truth (1) Majority judgment, (2) YEAR Unanimous judgment, (3) ---- Confession, (4) Court STUDY ID decision, (5) Mock crime or OUTCOME ~N=O-.------contrived story, (6) real TOTAL NUMBER OF OUTCOMES IN THIS ANALYSIS Qrime ·set up· by ------experimenter, (7) not TYPSTUDl, Type of study: analog or field verified, (8) not specified TYPESTU2, Type of study: (1) analog, (2) field ACCUR, Experimenter's judgment of accuracy of (1) detection, (2) blind basis for ground truth (see Barland, :valuation of charts, (3) 1982) (1) low, (2) high Judgment of accuracy based on QUESDES, Method for design~ng control questions other criteria, (4) "utility or pretest interview (1) Standard for all Ss; study," (5) jUdgment of (2) Customized accuracy based in pg and other criteria --- TECHNIQU, Type of question technique (l)ZOC, (2) MGQT, (3) POT, (4) ZOC & MGQT, (5) ZOC & SUBJECTS NSUBJS, Number of subjects or cases POT, (6) MGQT & POT, (7) GQT, TYPSUBJS, Type of subj pop (8) ZOC & GQT, (9) MGQT & (1) college students, (2) GQT, (10) POT & GQT, (11) GK, general pop, (3) non-crim. (12) RI, (13) RCQT i military personnel, (4) i STIM, Stirn test included? (1) Yes, (2) No I non-military criminals or j MACHINE, Machine type (1) Lafayette 4 channel Model suspects, (5) military 76058, (2) Narco Bio-system criminals or suspects, (6) I polygraph, (3) 3 channel police informants, (7) prison Stoelting, (4) 4 channel inmates, (8) police I Stoelting, (5) Sanborn 150 applicants, (9) private Recorder, (6) Keeler employment applicants, (10) I polygraph, (7) Stoelting with gov't employees or CAM, (8) Grass Model 7, (9) applicants, (11) v~ctlms, t physiograph, (10) 5 Channel (12) witnesses " *CASESRC, Source of cases for jUdgment I Reid, (ll)~v~a~r=i~e~d ______(1) polygraph school files (12) ______(2) police files, (3) , (13) ______PCTMALE, military files PASTE, Type of contact paste (1) Sanborn, (2) Beckman, (3) PURPOSE NaCl w/cornstarch (1) pre-employment, (2) crim investigation

114 f ! I ) 116 Ii 117 \ I Polygraph Coding Form I: Page 3 Polygraph Coding Form f: Page 4 PHYSMEAS, Phys. measure used for results (1) SCR/GSR, f: INCZONE, Inconclusive zone (+ or - ~) (2) Respiration, (3) Blood It I PGPCTMAL, % of Polygraphers Male pressure, (4) Heart rate, (5) t! PGEXP, Avg. yrs Poly training and experience Cnrdiovascular unpsecified, I,I' EXAMEXPF, Examiners' expo range from (in r;s. ) (6) finger pulse volume, (7) I EXAMEXPT, Examiners' expo ranged to some combination ii Ji" PHYSME2, Were other phys. measures taken and (1) Ye s, ( 2) No not used in analysis? II PGTYPE, Type of initial examiner (1) private, (2) police, (3) military, (4) other govt, PHYSME3; If ans. to PHYSME2 is yes, why? (1) results inconclusive, [1 Ii (5) trainees, (6) not a prof. (2) not given !; CHARTS, Number of charts on which examiners' I' examiner judgment based ii PGTRN, Place polygraph examiner trained (1) Reid, (2) Army, (3) H PROCED, Did procedure differ from standard in I) i! any way (e.g., Podlesny & Raskin did *JUDGES, Judge characteristics (1) Polygraphers trained at , same school, (2) Polygraphers not review control questions with Ss) (1) Yes, (2) No PROCVARY, Way procedure varied II trained at different school, (Use variable list I!" code#) 11 (3) law enforcement agents, n II (4) legal professionals (1) (2) [I (lawyers, judges), (5) Same EXAMEQ, Did examiners do own init. ratings Yes, No ~ (i.e., chart interpretations) 'I as initial examiners f1 I' If answer to EXAMEQ is "No,~ answer following I, ("utility" studies), (6) i with respect to those who did do init. ratings 1, Statistical analysis, (7) fi (Note these are not ultima~judges in field studies) d Other methods of **PGBLIND, Were raters blind to subj P ideniification (fingerprints, II condition? II handwriting, eyewitness), (8) (1) Yes, (2) No Other, (a) Polygraphers **KNm-1RATR, Did raters know rate of guilt? (1) Yes, (2) No II **RATEXPF, Raters expo ranged from (in yrs.) [other than (1) & (2)] **RATEXPT, Raters expo ranged to (in yrs.) II NJUDGES, Number of judges or evaluators (not II initial examiners) KNOWRATJ, Did judges know base rate of guilt? ( 1) Ye s , (2) No OBJRAT, Was orig. rating objective? (1) high (specific JUDGAGRE, Method of judge agreement (if panel) (1) UI!animous, (2) Majority measurement of phys. *JUDGEXPF, Judges expo ranged from (in yrs.) variables), (2) medium (score iII *JUDGEXPT, Judges expo ranged to (in yrs.) assigned to subjective II *JUDGEXP2, Judges expo ranged to (1) less than 1 yr., assessment, (3) low (rating [1 lJ (2) greater than 1 yr. of guilt or innocence based A *JUDGEXP3, Judges expo ranged from (1) less than 1 yr., an visual assessment, (4) I lJ (2) greater than 1 yr. very low (rating of guilt or r *AVJUDEXP, Av. judge exp (yrs.) innocence based on case files, clinical assessment Ii1 etc. ) NEXAM, Number ~f initial examiners II

11 tl I

!I j I ~'! ______-----L 1,II ____ ~ _____'____'______~~~,__" ____ ,__ ~,_~ \~~~..

\

118 119

Polygraph Coding Form Page 5 polygraph Coding Form Page 6 DESIGN SAMPLING, Random selection of Ss or cases? (1) Yes, (2) No OUTCOME EXCLU, If not randomly selected, % of population DETECTION STUDIES not included in sample GC BASISSEC, Basis of selection (use variable code listing) GNC GIN ATTRIT, % attrition from sample IC BASISATT, Basis of attrition (use variable code listing INC lIN KNOWRATE, Did init. examiners know rate of guilt? (1) Yes, (2) No UNIT, Unit of analysis for outcome l) Persons MOTIV, Were subjects offered inducement to (2) Questions, beat machine? (analogue only) (1) Yes, (2') No PGBLIND2, Did examiners know Ss were in an exp? (1) Yes, (2) No JUDGMENT STUDIES INDEPEND, Was initial polygraph rating blind JGC (independent of examinntion?) (1) Yes, (2) No JGNC *OBJRAT2, Were "judges" objective? r~tings (1) high (specific JGIN measurement of physo JIC variables), (2) medium (score JINC assigned to subjective JIIN assessment, (3) low (rating of guilt or innocence based OTHER CROSS-VALIDATION STUDIES on visual assessment, (4) GC2 very low GNC2 GIN2 FACTORIA, Factorial effect tested (use variable IC2 code listing) INC2 FACTORIB, Was factorial etfect IA significant? (1) Yes, ( 2 ) No IIN2 FACTOR2A, Second factorial effect tested? FACTOR2B, Was factorial effect 2A significant (1) CONTINUOUS SCORES (Means and signif. te~ts) FACTOR3A, Yes, ( 2) No Third factorial effect tested? GUILTY, Mean for guilty (deceptive) sU~Je~:s FACTOR3B. Was factorial effect 3B signifiicant? (1) INNO, Mean for innocent (truthful) sUbJec FACTOR4A, Yes, ( 2) NO Fourth fac~orial effect tested? SIGTEST, Significance test used (l) F, (2) t FACTOR4B, Was factorial effect 4A (3) , (4) significant? ( 1) Yes, ( 2) No Ii ( l) Ye s, (2) No i,' SIGDIFF, Was difference significant? j f 121 Appendix D

Acronyms and Glossary interaction: An occurrence which affects validity of about the examination and their rights. In some pre­ polygraph testing because individual character traits test interviews, examiners also make observations OT situational factors might result in unexpected about subjects' behavior to assist in determinations Acronyms physiological responses. of decf.!ptiveness or nondeceptiveness. baseline: The readings on a polygraph chart that form internal validity: The degree to which a study has con­ psychopathy: A psychiatric diagnostic category sig­ a point of comparison for the physiological re­ trolled for extraneous variables which may be nifying a character style prone to criminal activity ANS -autonomic nervous system sponses to the polygraph questions. related to the study outcome. and amoral, manipulative behavior. CIA -Central Intelligence Agency cla~sified information: Information that pertains to na­ irrelevant questions: Neutral questions designed to as­ random sampling: A procedure used to obtain repre­ CQT -control question technique tIonal security and by definition, cannot be disclosed sess the subject's baseline physiological response to sentative samples from a population. In complete DCI -Director of Central Intelligence to others without clearance. questioning and to provide a rest between relevant random sampling, each subject in the population DLCQ -directed lie control question [technique] clinical components: Components of a polygraph test questions. must have an equal chance of being selected and the DOD -Department of Defense procedure, including "proper" examiner attitude and numerical scoring: The assignment of numbers to poly­ selection or nonselection of one subject cannot in­ DOJ -Department of Justice relationship with subjects, that attempt to ensure graph chart responses. fluence the selection or nonselelction of another. EDR -electrodermal response accuracy. physiological arousal: Responses related to increases relevant/irrelevant technique: An examination tech­ GKT -guilty knowledge test construct validity: The extent to which a test or pro­ in anxiety. Those measured in polygraph examina­ nique that utilizes two types of questions: relevant GSR -galvanic skin response cedure measures what it is designed to measure. tions include electrodermal response, blood pres­ questions and neutral questions intended to assess GQT -general question test control. question. technique: A polygraph question sure, and respiration rate. the subject's baseline response. JAG -Judge Advocate General techmq~e that Incorporates control questions which polygraph chart: A continuous graph on which a sub­ relevant questions: Polygraph test questions about the LEAA - Law Enforcement Assistance ?re deSIgned to be arousing for nondeceptive sub­ ject's physiological responses are registered. topic or topics under investigation. Administration Jects and less arousing for deceptive subjects than predictive association: An index which measures the reliability: The degree to which a test yields repeatable MGQT -modified general question test the relevant questions. proportional reduction in the probability of error results. Reliability also refers to consistency across MMPI -Minnesota Multiphasic Personality counte~intelligence: Efforts of an organization to stop in predicting one category (in this case, deception) examiners / scorers. Inventory ?utside groups from gaining information about when the second category (in this case, polygraph Sensitive Compartmented Information: Classified in­ NSA -National Security Agency Itself. examination results) is known. formation above the top secret level. NSDD -National Security Decision Directive counterintelligence screening examinations: Examina­ predictive validity: The accuracy with which criterion socialization: The process in and by which individuals OPM -Office of Personnel Management tions given to personnel who already have access scores obtained in the future can be estimated from learn the ways, ideas, beliefs, values, patterns, and Pd -psychopathic deviate [scale] to classified information. test data obtained in the present. norms of a particular culture and adapts them as POT -peak of tension [test] electrodermal response: A physiological measure that preemployment screening: The use of polygraph test­ a part of their own personalities. RCQT -Reid control question technique has been ~hown to be related to physiological ing to question employee applicants. validity: A measure of the extent to which an observed RII -relevant/irrelevant [technique] arousal. It IS measured as the electrical resistance of pretest interview: The first portion of the polygraph situation reflects the "true" situation. SCI -Sensitive Compartmented Information the skin through the use of electrodes attached to testing procedure in which subjects are informed SCR -skin conductance response the fingertips. USAMPS-U.S. Army Military Police School external validity: The established generalizability of USPS -U.S. Postal Service a s~udy to particular subject populations and ZOC -zone of comparison [technique] settIngs. false negative: An erroneous decision that an individ­ ual is not deceptive when she or he is actually Glossary deceptive. analog studies: Analog studies are laboratory studies false .positive: ~n erroneous decision that a person is beIng deceptIve when he or she is actually being of polygraph testing that simulate actual field exam­ truthful. inations. Typical components of field examinations Held te~ting: Actual techniques used by polygraph are replicated. The goal of such studies is to test the examIners. validity of various polygraph techniques under con­ trolled conditions. generalizability: The extent to which results of previous investigations can be used in evaluation of aperiodic checking: Polygraph tests conducted at ir­ present investigations. regular times with randomly or otherwise selected ground truth: The establishment of actual guilt or in­ personnel to ask questions for internal security pur­ poses. ~ocence. In a field study it is based on a criterion autonomic lability: Term used to describe individual ~nd~~endent of the polygraph test (e.g., confession, differences in autonomic arousal. JudICIal outcome, panel decision). Outcome of an examination in which base rate: The number of guilty (or innocent) subjects in~onchlsives: as a percentage of the total. It cannot be determined from the subject's responses whf~ther he or she is deceptive or nondeceptive.

Llo"______~1~20L_ ____~ ______~ ______~ ______~

~ -- ~------~ ------~- ",0:; '1't"

References

1. Abrams, S., "Polygraph Validity and Reliabili­ port Battalion, Military Intelligence Group, ty: A Review," Journal of Forensic Sciences U.S. Army, Fort George Meade, Md., 1981. 17:313-327, 1973. 17. Barland, G. H., "On the Accuracy of the Pol­ 2. Addington v. Texas, 441 US 418, 1979. ygraph: An Evaluative Review of Lykken's 3. American Polygraph Association, Manual for Tremor in the Blood," Polygraph 11:258-272, Polygraph School Inspections (Lithicum 1982. Heights, Md.: American Polygraph Associa­ 19. Barland G. H., "Detection of Deception tion, 1983). Theory," unpublished, November 1982. 4. American Polygraph Association, "Polygraph 20. Barland, G. H., and Raskin, D. c., "Detection School Accreditation Standards," 1983. of Deception," Electrodermal Activity in Psy­ References 5. American Psychological Association, Stand­ chological Research, W. F. Prokasy and D. C. ards for Educational and Psychological Tests Raskin (eds.) (New York: Academic Press, (Washington, D.C.: American Psychological 1973), pp. 418-471. Association, 1974). 21. Barland, G. H., and Raskin, D. c., "An Eval­ 6. American Psychological Association, Joint uation of Field Techniques in Detection of De­ Technical Standards for Educational and Psy­ ception," Psychophysiology 12:321-330, 1975. chological Testing, unpublished draft, Wash­ 22. Barland, G. H., and Raskin, D. c., "Validity ington, D.C., 1983. and Reliability of Polygraph Examinations of 7. Ansley, N., "A Review of the Scientific Litera­ Criminal Suspects," report No. 76-1, contract ture on the Validity, Reliability and Utility of No. 75- Polygraph Techniques" (Ft. Meade, Md.: Na­ Nl-99-0001 (Washington, D.C.: National In­ tional Security Agency, 1983). stitute of Justice, Department of Justice, 1976). 8. Ansley, N., Quick Reference Guide to Poly­ 24. Barton, M., "A Study of the Admissions Made graph Admissibility, Licensing Laws, and Lim­ During Pre-Employment Polygraph Examina­ iting Laws (Severna Park, Md.: American Pol­ tions (1964-1975) and Their Significance to the ygraph Association, 1983). Business Community," unpublished masters 9. Ansley, N., Horvath, F., and Barland, G. H., thesis, Sam Houston State University, 1976. Truth and Science: A Bibliography, 2d ed. 25. Belt, J. A., and Holden, P. B., "Polygraph (Lithicum Heights, Md.: American Polygraph Usage Among Major U.S. Corporations," Per­ Association, 1983). A comprehensive index to sonnel Journal 80-86, 1978. international literature on the detection of de­ 26. Ben-Ishai, A., "Some Remarks on Polygraph ception and the polygraph (lie detector) tech­ Research," paper presented at the Ninth Annual nique. Meeting of the American Academy of Poly­ 10. Backster, c., "Methods of Strengthening Our graph Examiners, Chicago, Ill., August 1962. Polygraph Technique," Police 6:61-68, 1962. 27. Ben-Shakhar, G., Bar-Hillel, M., and Lieblich, 1. Backster, c., Standardized Polygraph Note­ 1., Trial by Polygraph: Scientific and Juridical pack and Technique Guide (San Diego, Calif.: Issues in Lie Detection, Hebrew University, C. Backster, 1979). 1983. 12. Balloun, K. D., and Holmes, D. S., "Effects of 28. Ben-Shakhar, G., Lieblich, 1., and Bar-Hillel, Repeated Examinations on the Ability to Detect M., "An Evaluation of Polygraphers' Judge­ Guilt With a Polygraphic Examination: A Lab­ ments: A Review From a Decision Theoretic oratory Experiment With a Real Crime," JoU" Perspective," Journal of Applied Psychology nal of Applied Psychology 64:316-322, 1979. 67:701-713, 1982. 13. Barefoot v. Estelle, SILW 5190, June 28, 1983. 29. Bersh, P. J. , "A Validation Study of Polygraph 14. Barland, G. H., "A Survey of the Effect of the Examiner Judgements," Journal of Applied Psy­ Polygraph in Screening Utah Job Applicants: chology 53:399-403, 1969. Preliminary Results," in: U.S. Congress, Senate 30. Bitterman, M. E., and Marcuse, F. L., "Cardio­ Judiciary Committee, hearings, 1978. vascular Responses of Innocent Persons to 15. Barland, G. H., "An Introduction to the Num­ Criminal Interrogation," American Journal of ber Test," Polygraph 7:203-208, 1978. Psychology 60:407-412, 1947. 16. Barland, G. H., "A Validation and Reliability 32. Blaylock, K. D., "Use of Polygraphs and Pre­ Study of the Counterintelligence Screening publication Review," testimony before joint Test," unpublished manuscript, Security Sup- hearing of House Post Office and Civil Service •

126 127

Subcommittee on Civil Service and House Ju­ 48. Davis, R. c., "Physiological Responses as a diciary Subcommittee on Civil and Constitu­ Means of Evaluating Information," The Manip­ fects of Trait Anxiety and Stimulus Context on of the Society for Psychophysiological Re­ tional Rights, Apr. 21, 1983. ulatiDn of H!. .'!TlCm BehaviDr, A. D. Biderman Skin Conductance," JDurnal 'Of Research in Per­ search, Minneapolis, Minn., 1982. 33. Blum, R. H., and Osterloh, W., "The Poly­ and H. Zimmer (eds.) (New York: Wiley, sonality 14:1-11, 1980. 76. Honts, C. R., and Hodes, R. L., 'The Effects graph Examination as a Means for Detecting 1961), pp. 142-169. 62. Gintox1, A., Netzer, D., and Elaand, E., "A of Multiple Physical Countermeasures on the Truth and Falsehood in Stories Presented by 49. Dawson, M. E., "Physiological Detection of Method for Evaluating the Use of the Polygraph Detection of Deception," PsychDphysiDlDgy Police Informants," The Journal of Criminal Deception: Measurement of Responses to Ques­ in a Real-Life Situation," JDurnal 'Of Applied 19:564-565 (abstract), 1982. Law, Criminology and Police Science 59:133- tions and Answers During Countermeasure PsycholDgy 67:131-137, 1982. 77. Honts, C. R., and Hodes, R. L., "The Effect of 137, 1968. Maneuvers," PSychDphysiDlDgy 17:8-17, 1980. 63. Glass, G. V., McGaw, B., and Smith, M. L., Simple Physical Countermeasures on the Detec­ 34. Borchard, E., Convicting the Innocent (Garden SO. Decker, R. E., 'The Army Stimulation Test: A Meta-Analysis in SDcial Research (Beverly tion of Deception," PsychophysiDlDgy 19:564 City, N.Y.: Garden City Publishing Co., 1932). Control Procedure," PDlygraph 7:176-178, Hills, Calif.: Sage Publications, 1982). (abstract), 1982. 3.5. Bradley, M. T., and Janisse, M. P., "Accuracy 1978. 64. Goodman, L. A., and Kruskal, W. H., "Meas­ 78. Honts, C. R., Raskin, D. c., and Kircher, J. Demonstrations, Threat, and the Detection of 51. Edel, E. c., and Jacoby, J., "Examiner Reliabili­ ures of Association for Cross-Classification," c., "Detection of Deception: Effectiveness of Deception: Cardiovascular, Electrodermal, and ty in Polygraph Chart Analysis: Identification JDurnal 'Of the American Statistical AssDciatiDn Physical Countermeasures Under High Motiva­ Pupillary Measures," Psychophysiology 18: of Physiological Responses," JDurnal 'Of Applied 49:732-764, 1954. tion Conditions," paper presented at meetings 307-314, 1981. PsychDlDgy 60:632-634, 1975. 65. Gustafson, L. A., and Orne, M. T., "Effects of of the Society for Psychophysiological Re­ 36. Brown, C. c., MethDds in PsychDphysiolDgy 52. Edwards, R. H., A Survey: Reliability 'Of PDl­ Heightened Motivation on the Detection of De­ search, September 1983. (Baltimore: The Williams & Wilkins Co., 1967). ygraph ExaminatiDn CDnducted by Virginia ception," JDurnal 'Of Applied PsycholDgy 43: 79. Horvath, F. S., "The Police Candidate Poly­ 37. Buckley, J., letter to Fred B. Wood, Office of PDlygraph Examiners, report for the Common­ 408-411, 1963. graph Examination: Considerations for the Po­ Technology Assessment, 1983. wealth of Virginia (Richmond, Va.: Bureau of 66. Gustafson, L. A., and Orne, M. T., 'The Ef­ lice Administrator," PDlice 33-39, 1972. 39. Carlucci, F. c., "Degradation of Operational Forensic Science, Department of General Serv­ fects of Task and Method of Stimulus Presen­ 80. Horvath, F. S., "Verbal and Nonverbal Clues Readiness/Mission Accomplishment Due to ices, 1981). tation on the Detection of Deception," JDurnal to Truth and Deception During Polygraph Ex­ Personnel Security Investigation Shortfalls," 53. Ekman, P., Levenson, R. W., and Friesen, W. 'Of Applied PSYChDlDgy 49:412-417, 1964. aminations," JDurnal 'Of PDlice Science and Ad­ memorandum, U.S. Department of Defense, y., ~'Autonomic Nervous System Activity," 67. Gustafson, L. A., and Orne, M. T., "Effects of ministratiDn 1:138-152, 1973. Aug. 6, 1982. Science 221:1208-1210, 1983. Perceived Role and Role Success on the Detec­ 81. Horvath, F. S., "Detection of Deception: A 40. Cimmerman, A., "They'll Let Me Go Tomor­ 54. Elaad, E., and Schahar, E., "Polygraph Field tion of Deception," Journal 'Of Applied PsychDl­ Review of Field and Laboratory Research," PDl­ ,. Dgy 49:412-417, 1965. ygraph 5:107-145, 1976. row," Criminal Defense 8, 7-10, 1981. Validity," Scientific InterrogatiDn in Criminal j 41. Cook, T. D., and Campbell, D., Quasi-Exper­ InvestigatiDn, 1. Nachson (ed.), selected papers I 67a. Hammond, D. L., "The Responding of Nor­ 82. Horvath, F. S., "The Effect of Selected Varia­ imentatiDn: Design and Analysis Issues fDr Field presented at the First National Conference on mals, Alcoholics and Psychopaths in a Labor­ bles on Interpretation of Polygraph Records," Settings (Chicago: Rand McNally, 1979). Scientific Interrogation in Criminal Investiga­ atory Lie-Detection Experiment," unpublished JDurnal of Applied PsychoIDg'/ 62:127-136, 42. Corcoran, J. F. T., Lewis, M. D., and Garver, tion at Bar-Han University and Ramat-Gan, Is­ doctoral dissertation, California School of Pro­ 1977. R. B., "Biofeedback Conditioned Galvanic Skin rael, November 1976. fessional Psychology, 1980. 83. Horvath, F. S., "Detecting Deception in Eye­ Response and Hypnotic Suppression of Arous­ 55. Ennis, B. J., and Litwack, T. R., "Psychiatry 68. Hamon, E., "Evidence Obtained by Polygraph: witness Cases: Problems and Prospects in the al: A Pilot Study of Their Relation to Decep­ and the Presumption of Expertise: Flipping An Israeli Perspective," The Criminal Law Re­ Use of the Polygraph," Advances in the Psy­ tion," Journal DfFDrensic Sciences 23:155-162, Coins in the Courtroom," CalifDrnia Law Re­ view 340-349, 1982. chDlogy 'Of Eyewitness TestimDny, J. Wells and 1978. view 62:693 ·752, 1974. 69. Harrelson, L. H., The Keeler Technique, 2d ed. B. Loftus (eds.) (New York: Cambridge Univer­ 43. Correa, E. 1., and Adams, H. E., "The Validi­ 56. Ferguson, R. J., The PDlygraph in Private In­ (Chicago: Keeler Polygraph Institute, 1964). sity Press), in press. ty of the Pre-employment Polygraph Examina­ dustry (Springfield, Ill.: Charles C. Thomas, 71. Hayden, T., "ACLU Project Privacy," testi­ 84. Horvath, F. S., and Reid, J. E., "The Reliabili­ tion and the Effects of Motivation," PDlygraph 19Q6). mony of Trudy Hayden, in U.S. Congress, Sen­ t:, of Polygraph Examiner Diagnosis of Truth 10:143-156, 1981. 57. Fingerhut, K. R., "Use of the Stimulation Test ate Judiciary Committee, hearings, 1978. and Deception," The Journal 'Of Criminal Law, 44. Cureton, R., "A Consensus as to the Validity in Preemployment Testing," PDlygraph 7:185- 72. Hayden, T., "Employers Who Use Lie Detec­ CriminolDgy and PDlice Science 62:276-281, of Polygraph Procedures," Tennessee Law Re­ 188, 1978. tor Tests," Business & SDciety Review 41:16-21, 1971. view 22:18-32, 1953. 58. Frye v. United States, 293 F. 1013 (DC Cir. 1982. 85. Hunter, F. L., and Ash, P., "The Accuracy and 45. Cutrow, R. J., Parks, A., Lucas, N., and 1923). 73. Hays, W. L., Statistics, 3d ed. (New York: Consistency of Polygraph Examiners' Diagno­ Thomas, K., "The Objective Use of Multiple 59. Gatchel, R. J., Smith, J. E., Kaplan, N. M., et Holt, Rinehart, & Winston, 1981). ses," Journal 'Of PDlice Science and Administra­ Physiological Indices in the Detection of Decep­ al., "The Effect of Propanolol on Polygraphic 74. Heckel, R. V., Brokaw, J. R., Salzberg, H. c., tiDn 1:370-375, 1973. tion," PsychophysiDlDgy 9:578-588, 1972. Detection of Deception," unpublished manu­ and Wiggins, S. L., "Polygraphic Variations in 86. Iacono, W. G., Boisvenu, G. A., and Flt:!ming, 46. Davidson, P.O., "Validity of the Guilty-­ script, 1983. Reactivity Between Delusional, Nondelusional, J. A., 'The Effects of Diazepam and Methyl­ Knowledge Technique: The Effects of Motiva­ 60. Giannelli, P. c., 'The Admissibility of Novel and Control Groups in a Crime Situation," phenidate on the Electrodermal Detection of tion," JDurnal 'Of Applied PSychDlDgy 52:62-65, Scientific Evidence: Frye v. United States, a JDurnal 'Of Criminal Law, CriminolDgy and PD­ Guilty Knowledge," unpublished manuscript, 1968. Half Century Later," CDlumbia Law RevieW 80: lice Science 53:380-383, 1962. University of British Columbia, Vancouver, 47. Davidson, W. A., "Validity and Reliability of 1198-1250, 1980. 75. Honts, C. R., and Hodes, R. L., "The Effect of Canada. the Cardio Activity Monitor," PDlygraph 8: 61. Giesen, M., and Rollison, M. A., "Guilty Simple Physical Countermeasures on the Detec­ 87. Jones, E. E., and Sigall, H., "The Bogus Pipe­ 104-111, 1979. Knowledge Versus Innocent Associations: Ef- tion of Deception," paper presented at meetings line: A New Paradigm for Measuring Affect ... -

.~.

1 " 128 129

and Attitude," Psychological Bulletin 76: tors," Journal of Criminal Law and Criminol­ Interview and Its Role in the Detection of De­ 135. Raskin, D. c., "Science, Competence, and Pol­ 349-364, 1971. ogy 12:390-399, November 1921. ception," Polygraph 9:74-85, 1980. ygraph Techniques," Criminal Defense 8:11-18, 89. Keeler, L., ''Debunking the 'Lie Detector,'" The 104. Lazarus, R. S., Psychological Stress and the 122. Orne, M. T., "Implications of Laboratory Re­ 1981. American Journal of Police Science XXV:153- Coping Process (New York: McGraw Hill search for the Detection of Deception," Legal 136. Raskin, D. "The Scientific Basis of Poly­ 159, 1934. c., 1966). ' Admissibility of the Polygraph, N. Ansley (ed.) graph Techniques and Their Uses in the Judicial 89a. Kircher, J. c., "Computerized Decision-Making 105. Lykken, D. T., "The GSR in the Detection of (Springfield, Ill.: Charles C. Thomas, 1975), Process," proceedings of the Stockholm Sym­ and Patterns of Activation in the Detection of Guilt," Journal of Applied Psychology 43:385- pp.94-119. posium on Witness Psychology, 1982, Deception," unpublished doctoral dissertation, 388, 1959. 123. Orne, M. T., Thackray, R. I., and Paskewitz, pp. 319-371. University of Utah, 1983. 106. Lykken, D. T., "The Validity of the Guilty D. A., "On the Detection of Deception: A 137. Raskin, D. c., and Hare, R. D., "Psychopathy 90. Kircher.> J. c., and Raskin, D. c., "Computer­ Knowledge Technique: The Effects of Faking," Method for the Study of the Physiological Ef­ and Detection of Deception in a Prison Popula­ ized Decisionmaking in the Detection of Decep­ Journal of Applied Psychology 44:258-262, fects of Psychological Stimuli," Handbook of tion," Psychophysiology 15:126-136, 1978. tion," Psychophysiology 18:204-205, 1981. 1960. PsydlOphysiology, N. Greenfield and R. Stern­ 138. Raskin, D. c., and Podlesney, J. A., "Truth 91. Kircher, J. c., and Raskin, D. c., "Clinical Ver­ 107. Lykken, D. T., "Psychology and the Lie Detec­ bach (eds.) (New York: Holt, Rinehart, & Win­ and Deception: A Reply to Lykken," Psycho­ sus Statistical Lie Detection Revisited: Through tor Industry," American Psychologist 29:725- ston 1972), pp. 743-785. logical Bulletin 86:54-58, 1979. a Lens Sharply," unpublished, Department of 739, 1974. 124. Peters, R B., "A Survey of Polygraph Evidence 139. Reid, J. E., and Inbau, F. E., Truth and Decep­ Psychology, University of Utah, Salt Lake City, 108. Lykken, D. T., A Tremor in the Blood: Uses in Criminal Trials," American Bar Association tion-The Polygraph Technique, 3d ed. (Bal­ Utah, 1983. and Abuses of the Lie Detector (New York: Journal 68:162-165, 1982. timore: The Willliams & Wilkins Co., 1977). 92. Kleinmuntz, B., and Szucko, B., "On the Fal­ McGraw Hill, 1981). 125. Pillemer, D. B., and Light, R. J., "Benefiting 140. Robbins, N. E., and Penley, W. J., "Review of libility of Lie Detection," Law and Society 109. Lykken, D. T., "The Psychopath and the Lie From Variation in Study Outcomes," New Dir­ Polygraph Charts on Nondeceptive Subjects," Review 17:84-104, 1982. Lctector," Psychophysiology 15:137-142, 1978. ections for Methodology of Social Behavior Sci­ Polygraph 4:199-207, 1975. 93. Kubis, J. F., Studies in Lie Detection: Computer 110. Lykken, D. T., "The Detection of Deception," ence: Quantitative Assessment of Research Do­ 141. Rosenthal, R, Experimenter Effects in Behav­ Feasibility Considerations contract No. AF-30- Psychological Bulletin 86:47-53, 1979. mains, R. Rosenthal (ed.) (San Francisco: ioral Research (New York: Irvington, 1976). (602)-2270, project No. 5534 submitted to U.S. 111. Lyon, V. W., "Deception Tests With Juvenile Jossey-Bass, 1980). 142. Rosenthal, R., and Rubin, D. B., "Comparing Air Force, available from the Armed Services Delinquents," Journal of Genetic Psychology 126. Podlesny, J. A., and Raskin, D. c., "Physio­ Effect Sizes of Independent Studies," Psycho­ Technical Information Service (AD-284902), 48:494-497, 1936. logical Measures and the Detection of Decep­ logical Bulletin 2:500-504, 1982. Arlington, Va. 1962. 112. Macy J. W., Jr., Chairman, U.S. Civil Service tion," Psychological Bulletin 84:782-799, 1977. 143. Rovner, L. 1., Raskin, D. c., and Kircher, J. 94. Kubis, J. F., Analysis of Polygraphic Data, Commission, letter to Hon. John E. Moss, 127. Podlesny, J. A., and Raskin, D. c., "Effective­ c., "Effects of Information and Practice on De­ Technical Report No. RADC-TDR-64-101, Jan­ Chairman, Foreign Operations and Govern­ ness of Techniques and Physiological Measures tection of Deception," paper presented at the uary 1965. ment Information Subcommittee, House Com­ in the Detection of Deception," Psychophysi­ meetings of the Society for Psychophysiological 95. Kugelmass, S., "Reactions to Stress," contract mittee on Government Operations, Nov. 22, ology 15:344-358, 1978. Research, Wisconsin, 1979. No. AF-61-(052)839, U.S. Air Force Office of 1965. 128. Privacy Protection Study Commission, Person­ 144. Sackett, P. R., and Decker, P. J., "Detection Scientific Research, 1967. 113. MacNitt, R. D., "In Defense of the Electroder­ al Privacy in an Information Society (catalog of Deception in the Employment Context: A 96. Kugiemass, S., Lieblich, I., and Bergman, Z. mal Response and Cardiac Amplitude as Meas­ No. Y3.P93/5:1/9177) (Washington, D.C.: Review and Critical Analysis," Personnel Psy­ 'The Role of 'Lying' in Psychophysiological De­ ures of Deception," Journal of Law and Crimi­ U.S. Government Printing Office, July 1977). chology 32:487-507, 1979. tection," Psychophysiology 3:312-315, 1967. nology 33:266-275, 1942. 130. Pyle, C. H., ''The Defense Department's Poly­ 145. Saks, M. J., and Van Duizend, R., The Use of 97. Kugelmass, S., and Lieblich, I., ''The Relation 114. Marston, W. M., "Systolic Blood Pressure graph Regulations of 1982," statement before Scientific and Technological Evidence in Litiga­ Between Ethnic Origin and GSR Reactivity in Changes in Deception," Journal of Experimen­ the Subcommittee on Civil and Constitutional tion: Report and Indexed Bibliography (Wil­ Psychophysiological Detection," Journal of Ap­ tal Psychology 2:143-163, 1917. Rights, Committee on the Judiciary, U.S. liamsburg, Va.: National Center for State plied Psychology 52:158-162, 1968. 115. Matte, J. A., "The Art and Science of the Pol­ House of Representatives, Jan. 25, 1982. Courts, 1983). 98. Kugelmass, S., Lieblich, I., Ben-Ishai, A., ygraph Technique" (Springfield, Ill.: Charles C. 131. Quigley-Fernandez, B., and Tedeschi, J. T., 146. Sarbin, T. R., and Slagle, R. W., "Hypnosis Opatowski, A., and Kaplan M., "Experimen­ Thomas, 1980). ''The Bogus Pipeline as Lie Dectector: Two Va­ and Psychophysiological Outcomes," Hypno­ tal Evaluation of Galvanic Skin Response and 116. McEvoy, J. P., "The Lie Detector Goes Into lidity Studies," Journal of Personality and sis: Developments in Research and New Per­ Blood Pressure Change Indices During Criminal Business," The Reader's Digest, January 1941, Social.Psychology 36:247-256, 1978. spectives, E. Fromm and R. E. Shor (eds.) (New Interrogation," The Journal of Criminal Law pp. 69-72. 132. Raskin, D. c., personal communication, Aug. York: Aldine; 1979). Criminology and Police Science 59:632-635: 118. Monahan J., The Clinical Prediction of Violent 23, 1983. 147. Saxe, L., and Fine, M., Social Experiments 1968. Behavior (Washington, D.C.: U.S. Govern­ 133. Raskin, D. c., Reliability of Chart Interpreta­ Methods for Design and Evaluation (Beverly 100. Lahri, S. K, and Ganguly, A. K., "Experimen­ ment Printing Office, 1981. tion and Sources of Errors in Polygraph Exam­ Hills, Calif.: Sage Publications, 1981). tal Study of the Accuracy of Polygraph Tech­ 119. More, H. W., "Polygraph Research and the inations (report No. 76-3, contract No. 75-NI- 148. Schroeder, J., "Zone Comparison Test Con­ nique in Diagnosis of Deception With Volun­ University," Law and Order 14:73-78, 1966. 99-0001, National Institute of Law Enforcement struction (ZCT) Summary Sheet," unpublished teer and Criminal Subjects," Polygraph 7:89-94, 120. Moroney, W. F., and Zenhausern, R. J., "De­ and Criminal Justice, Law Enforcement Assist­ manuscript, 1983. 1978. tection of Deception as a Function of Galvanic ance Administration, U.S. Department of Jus­ 149. Senese, L., "Accuracy of the Polygraph Tech­ 101. Larson, J. A., Lying and Its Detection (Chicago: Skin Response Recording Methodology, II The tice, University of Utah, 1976. nique With and Without Card Test Stimula­ University of Chicago Press, 1932). Journal of Psychology 80:255-261, 1972. 134. Raskin, D. C. "Scientific Assessment of the Ac­ tion," Journal of Police Science and Administra­ 103. Larson, J. A., "Deception Tests and Lie Detec- 121. Mullenix, P. A., and Reid, J. A., "The Pretest curacy of Detection of Deception," Psycho­ tion 4:274-276, 1976. physiology 15:143-147, 1978. 150. Sigall, H., and Page R., "Current Stereotypes: , ------~ ~ i. 44 4'" •

130 131

A Little Fading, A Little Faking," Journal ofPer­ Community Agencies: Personnel Security Sur­ sonality and Social Psychology 18:247-255, vey," 1980. 175. U.S. Congress, House Permanent Select Com­ 187. U.S. National Security Agency, initial and sup­ 1971. 166. U.S. Civil Service Commission, letter to the mittee on Intelligence, Preemployment Securi­ plemental responses to OTA survey, 1983. 151. Sigall, H., and Page, R., "Reducing Attenua­ President, July 29, 1966. ty Procedures of the Intelligence Agencies,. hear­ 188. U.S. Office of Personnel Management, Federal tion in the Expression of Interpersonal Affect 167. U.S. Congress, House Committee on Govern­ ings before the Subcommittee on Oversight, Personnel }.1anual System, ch. 736, app. D, May 16, 17, 24, June 21, 1979 (Washington, via the Bogus Pipeline," Sociometry 35: ment Operations, Subcommittee on Govern­ "Use of the Polygraph in Personnel Investiga­ D.C.: U.S. Government Printing Office, 1980). tions of Competitive Service Applicants and 629-642, 1972. ment Information and Foreign Operations, Use 176. U.S. Department of Commerce, The Cost of Appointees to Competitive Service Positions" 152. Simon, M. A., "Shall We Ask the Lie Detec­ of Polygraphs as "Lie Detectors" by the Federal (Washington, D. U. S. Office of Personnel tor?" Science, Technology, and Human Values Government, 89th Cong., 1st sess., House Re­ Crimes Against Business, 1976. c.: 177. U.S. Department of Defense, Management, 1973). 8:3-13, 1983. port No. 198 (Washington, D.C.: U.S. Govern­ Department of 7 153. Skolnick, J. H., "Scientific Theory and Scien­ ment Printing Office, 1965). Defense Directive Number 5210.48, Conduct of 189. United States v. Brown, 557 F.2d 541, 19 7. tific Evidence: An Analysis of Lie Detection," 168. U.S. Congress, House Committee on Govern­ Polygraph Examinations and the Selection, 190. United States v. Stifel, 433 F.2d 431 (6th Cir. 1970). Yale Law Journal 70:694-728, 1961. ment Operations, Subcommittee on Govern­ Training and Supervisiun of DOD Polygraph July 13, 1965. v. 583 F.2d (2nd Cir. 154. Slowik, S. M., "Ethics in Pre-employment Pol­ ment Information and Foreign Operations, Use Examiners, 191. United States Williams, 178. U.S. Department of Defense, Department of 1978). ygraph Examinations," Polygraph 8:144-149, of Polygraphs as Lie Detectors by the Federal 5210.48, 192. "Uproar Over News Leaks and Lie Detectors," 1979. Government (Part 2), 89th Cong., 2d sess., Defen.';e Directive Number Polygraph 54:8, Aug. 15, 155. Slowick, S. M., and Buckley, J. P., "Relative House Report No. 280V (Washington, D.C.: Examinations and Examiners, Oct. 6, 1975. U.S. News and World Report Accuracy of Polygraph Examiner Diagnosis of U.S. Government Printing Office, 1966). 179. U.S. Department of Defense, Department of the 1963. Army, Army 193. Waid, W. M., "Skin Conductance Response to Respiration, Blood Pressure, and GSR Record­ 169. U.S. Congress, Senate Committee on the Judi­ Law Enforcement Investigations, Field Manual 19-20 (Washington, D.C.: De­ ings," Journal ofPolice Science and Administra­ ciary, Privacy, Polygraphs, and Employment, Both Signaled and Unsignaled Noxious Stimu­ tion 3:305-309, 1975. a study prepared by the staff of the Subcom­ partment of the Army, 1977). lation Predicts Level of Socialization," Journal 156. Smith, R., "Compilation of State and Federal mittee on Constitutional Rights (Washington, 180. U.S. Department of Defense, Select Panel Re­ of Personality and Social Psychology 34, 923- Privacy Laws," Privacy Journal 1:79, 1981. D.C.: U.S. Government Printing Office. 1974). view of the Department of Defense Personnel 929, 1976. 157. Smith, R., "Supplement to Compilation of 170. U.S. Congress, House Committee on Govern­ Security Program, Apr. 16, 1982. 194. Waid, W. M., and Orne, M. T., "Cognitive, State and Federal Privacy Laws 1982, 1983," ment Operations, The Use of Polygraph and 181. U.S. Department of Defense, Department of Social, and Personality Processes in the Physi­ Privacy Journal. Similar Devices by Federal Agencies, hearings Defense Polygraph Program, DOD Directive, ological Detection of Deception," Advances in 158. Sternback, R. A., and Turskey, B., "Ethnic Dif­ before Subcommittee on Foreign Operations June 1983. Experimental Social Psychology, Berkowitz, L. ferences Among Housewives on Psychophysi­ and Government Information, 93d Cong., 2d 182. U.S. Department of Defense, Department of the (ed.) 14:61-106, 1981. cal and Skin Potential Responses tu Electric sess. (Washington, D.C.: U.S. Government Army, The Validity of Polygraph Testing 195. Waid, W. M., and Orne, M. T., "The Physio­ Shock," Psychophysiology 1:241-246, 1965. Printing Office, 1974). (Washington, D.C.: U.S. Government Printing logical Detection of Deception," American 159. Szucko, J. J., personal communication, 1983. 171. U.S. Congress, House Committee on Govern­ Office, 1983). Scientist 70:402-409, 1982. 160. Szucko, J. J., and Kleinmuntz, B., "Statistical ment Operations, The Use of Polygraphs and 183. U.S. Department of Defense, Present Status of 196. Waid, W. M., Orne, E. c., Cook, M. R., and Orne, M. T., "Effects of Attention, as Indexed Versus Clinical Lie Detection," American Psy­ Similar Devices by Federal Agencies, Thirteenth DOD Research on the Polygraph, report of the chologist 36:488-496, 1981. Report by the Committee on Government DOD Joint Services Group on a Coordinated by Subsequent Memory, on Electrodermal De­ 161. Thackray, R. 1., and Orne, M. T., "Effects of Operations Together with Separate and Dis­ R&D Program of Lie Detection Research, Aug. tection of Information," Journal of Applied the Type of Stimulus Employed and the Level senting Views, 94th Cong., 2d sess. (Washing­ 28, 1968. Psychology 63:728-733, 1978. 197. Waid, W. M., Orne, E. Cook, M. R., and of Subject Awareness on the Detection of De­ ton, D.C.: U.S. Government Printing Office, 183a. U.S. Department of Defense, Department of the c., ception," Journal of Applied Psychology 1976). Air Force, response to OTA survey, 1983. Orne, M. T., "Meprobamate Reduces Accuracy 52:234-239, 1968, 172. U.S. Congress, Senate Committee on the Judi­ 184. U.S. Department of Justice, background briefer of Physiological Detection of Deception," Sci­ 162. Timm, H. W., 'The Effect of Placebos and ciary, Polygraph Control and Civil Liberties on Presidential Directive, 1983. ence 212:71-73, 1981. Feedback on the Detection or Deception," un­ Protection Act, hearings before the Subcommit­ 185. U.S. Department of Justice, Presidential Direc­ 198. Waid, W. M., Orne, M. T., and Wilson, S. K., published doctoral dissertation, College of So­ tee on the Constitution (Washington, D.C.: tive on Safeguarding National Security Infor­ "Effects of Level of Socialization on Electroder­ cial Science, Michigan State University, 1979. U.S. Government Printing Office, 1978). mation, 1983. mal Detection of Deception," 16:15-22, 1979. 163. Timm, H. W., "Analyzing Deception From 173. U.S. Congress, House Permanent Select Com­ 185a. U.S. Department of Justice, "Presidential Direc­ Psychophysiology 199. Waid, W. M., Orne, M. T., and Wison, S. K., Respiration Patterns," Journal of Political mittee on Intelligence, Subcommittee on Over­ tive on Safeguarding National Security Infor­ Science and Administration 10:47-51, 1982. sight, Security Clearance Procedures in the In­ mation and Polygraph Examinations of Federal "Socialization, Awareness and the Electroder­ 164. Timm, H. W., "Effect of Altered Outcome Ex­ telligence Agencies (Washington, D. c.: U. S. Employees," statement of Richard K. Willard mal Response to Deception and Self-Disclo­ pectancies Stemming From Placebo and Feed­ Government Printing Office, 1979). before the House Government Operations Sub­ sure," Journal of Abnormal Psychology 88: back Treatment on the Validity of the Guilty 174. U.S. Congress, House Committee on Education committee on Legislation and National Securi­ 663-666, 1979. ty, Oct. 19, 1983. 200. Waid, W. M., Wilson, S. and Orne, M. T., Knowledge Technique," Journal of Applied and Labor, Pressures in Today's Workplace, K., "Cross-Model Physiological Effects of Electro­ Psychology 4:391-400, 1982. vol. III, hearings before the Subcommittee on 185b. U.S. Department of Justice, response to OTA survey, 1983. dermal Lability in the Det-ection of Deception," 165. U.S. Central Intelligence Agency, Director of Labor-Management Relations, Dec. 15, 1979 Central Intelligence, "Investigative Scope and (Washington D.C.: U.S. Government Printing 186. U.S. Department of the Treasury, response to Journal of Personality and Social Psychology Adjudicative Procedures Among Intelligence Office, 1980). OTA survey, 1983. 40:1118-1125, 1981. .,

132

201. Weir, R. J., "In Defense of the Relevant-Irrele­ 206. Widacki, J.; "Analiza Przestanek Diagnozo­ vant Polygraph Test," Polygraph 3:119-166, wartia w Bac,mich Poligraficznych (The Anal­ 1974. ysis of Diagnostic Premises in Polygraph Ex­ 202. Weir, R. J., "Stimulation Procedures-A Con­ aminations)" (Katowice, Poland: Uniwersytetu servative View," Polygraph 7:209-214, 1978. Slaskiego, 1982). 203. Weir, R. J., and Atwood, W. F., "Applicant 207. Widacki, J., and Horvath, F., "An Experimen­ Screening Polygraph Examinations," Polygraph tal Investigation of the Relative Validity and 10:129-142, 1981. Utility of the Polygraph Technique and Three 204. Whiteside, G., Profits vs. Polygraph (Phoenix, Other Common Methods of Criminal Identifi­ Ariz.: Associated Grocers, 1979). cation," Journal of Forensic Sciences 205. Wicklander, D. E., and Hunter, F. L., "The In­ 23:596-601, 1978. fluence of Auxiliary Sources of Information in 208. Ziskin, J., Coping With Psychiatric ancJ Psy­ Polygraph Diagnoses," Journal of Police chological Testimony (Venice, Calif.: Law and Science and Administration 3:405-409, ]975. Psychology Press, 1982).

o ,r'"' ------

"'.'