THE ACCURACY AND UTILITY OF POLYGRAPH TESTING

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP07-01280R000400400001-7
• Release Decision: RIFPUB
• Original Classification: K
• Document Page Count: 191
• Document Creation Date: December 23, 2016
• Document Release Date: April 16, 2013
• Sequence Number: 1
• Content Type: REPORT

Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
THE ACCURACY AND UTILITY
OF POLYGRAPH TESTING
DEPARTMENT OF DEFENSE
WASHINGTON, D.C.
1983
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
CONTENTS
Foreword
Overview
Chapter 1 - Research on Polygraph Criterion Validity and Reliability
Field Studies of Criterion Validity
Relationship of Blind Examiners and Ground Truth
Relationship of Examiner Decision, Ground Truth Unknown
Laboratory Studies
Control-Question Test, Criterion Validity
Control-Question, Reliability
Relevant-Irrelevant Technique, Criterion Validity
Relevant-Irrelevant, Reliability
Guilty Information Tests, Criterion Validity
Guilty Information Tests, Reliability
Chapter 2 - Analysis of Research
Analysis
The Bersh Field Validation Study
Experience of Professional Investigators
and Quality Control Personnel
Laboratory Experiments with Mock Crimes
Summary
_Chapter 3 - Utilization of the Polygraph
The Polygraph in the Federal Government
Federal Agencies
Polygraph Examiners in the Department of Defense
Number of Examiners (Table)
Annual Polygraph Utilization (Table)
Criminal Investigation
History
Polygraph as an Aid to Investigation
Tables on Use in Criminal Investigations
Criminal Investigation Case Examples
Criminal Conduct Revealed During Screening Examinations
Intelligence, Counterintelligence and Security
History
Specific Investigations of Espionage
Examples of Polygraph Cases - Espionage
Suitability - Statistics
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Polygraph and the Traditional Methods
1953 Study of Polygraph/Background,
1983 Study of Polygraph/Background
1981 Polygraph Results from Cleared Military Personnel
1974 - 1979 Polygraph Results from Cleared Military Personnel
1983 Study of Interviews/Polygraph
Use of the Polygraph in Army Personnel Security Cases
Surveys of Examinees
1982 Air Force Survey
1.951 Survey of Atomic Energy Commission Employees
1956 Survey of NSA Employees
Surveys in Law Enforcement and Commerce
Quality Control
U.S. Army Polygraph Training Facility
Basic Polygraph Training
Advanced Training
Primary Polygraph Testing Techniques
Chapter 4 - Summaries of Research
Field Abstracts
Gordan H. Barland and David C. Raskin (1976)
Akiva Ben-Ishai (1962)
Philip J. Bersh (1969)
M. E. Bitterman and F. L. Marcuse (1947)
Avital Ginton at al. (1982)
Eugene C. Ede]. and Jacob Jacoby (1975)
Eitan Elaad and Esther Schahar (1976)
Robert H. Edwards (1981)
Frank Horvath (1977)
Frank Horvath and John E. Reid (1971)
Fred L. Hunter and Philip Ash (1973)
Verne W. Lyon (1936)
Jesse Orlansky (1962)
Robert B. Peters (1982)
David C. Raskin (1976)
Stanley M. Slowik and Joseph P. Buckley (1975)
Douglas E. Wicklander and Fred L. Hunter (1975)
Jan Widacki (1982)
Laboratory Abstracts: Control Question Techniques
Gordan H. Barland (1981)
Gordon H. Barland and David C. Raskin (1975)
M. T. Bradley and Michel P. Janisse (1981)
Robert J. Gatchel, John E. Smith, and Norman M. Kaplan (1983)
David L. Hammond (1980)
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Charles R. Honts (1982)
D. C. Raskin and R. D. Hare (1978)
L. I. Rovner, D. C. Raskin and J. C. Kircher (1978)
William M. Waid, Emily C. Orne and Martin T. Orne (1981)
William M. Waid, Martin T. Orne and Stuart K. Wilson (1979)
Jan Widacki and Frank Horvath (1978)
Laboratory Abstracts: Relevant-Irrelevant Techniques
Richard H. Blum and William Osterloh (1968)
Eileen J. Correa and Henry E. Adams (1981)
Laboratory Abstracts: Guilty Knowledge and Peak of Tension Techniques
Kristan H. Balloun and David S. Holmes (1979)
P. O. Davidson (1968)
Miroslav Dufek (1969)
Jan Kronbergerova and Miroslav Dufek (1969)
David T. Lykken (1959)
David T. Lykken (1960)
Kazuo Ohnishi, et al. (1976)
Robert M. Stern, et al. (1981)
William M. Waid, et at. (1978)
William M. Waid, et al. (1981)
Kazunoba Yamaoka and Akihiro Suzuki (1980)
References Cited
Most of the references cited in the text will be found in the abstracts
section. Those references which are not in the abstracts are listed after the
abstracts.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Foreword
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved ForRelease2013/04/16 : CIA-RDP07-01280R000400400001-7
FOREWORD
This analysis of the scientific literature on the accuracy of the polygraph,
with supporting information on use and utility, was prepared at the direction of
Deputy Under Secretary of Defense Richard G. Stilwell, General, USA (Ret.).
General Stilwell, with the concurrence of the Director of the National Security
Agency, appointed Norman Ansley to lead a working group of senior examiners from
the military services to prepare a report, setting forth the basis upon which the
polygraph is utilized in the Department of Defense. In addition, the report was to
include information on the use and results of polygraph testing in the Department.
The working group, listed below, provided information from their resources
on the use of the polygraph and case examples. They also served as an editorial
board, reviewing the text. This report was prepared by Norman Ansley, Chief,
Polygraph Division, Office of Security, National Security Agency, and Marcia
Garwood, Ph.D., Research and Special Branch, Polygraph Division, Office of
Security, National Security Agency. Gordon H. Barland, Ph.D., served as a
scientific editor of the report. William Fedor, Deputy Director (Personnel
Security) Counterintelligence and Investigative Programs, and William H. Bell,
Security Specialist, Counterintelligence and Investigative Programs Directorate,
ODUSD (P), represented DoD in the development of this report.
THE WORKING GROUP
Norman Ansley
Chief, Polygraph Division
Office of Security
National Security Agency
Robert A. Brisentine, Jr.
Chief Polygraph Examiner
U.S. Army Law Enforcement
U.S. Army Criminal Investigation Command
Louise S. Fuse
Deputy Polygraph Program Manager
U.S. Army Intelligence and Security Command
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
B. Frank Bloomingburg
Special Assistant
Scientific Investigations
Headquarters, Naval Investigative Service
James E. Hardy
Chief, U.S. Air Force Polygraph Program
Headquarters, USAF Office of Special Investigation
This book was produced in less than 30 days. It is based on the material
currently available. The tables on DoD use are based on the semiannual reports of
the military services and NSA. An enlarged and final edition will be available
later this year. It will contain more summaries, particularly foreign material.
There will be additional topics addressed, and a more detailed analysis of the data.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
OVERVIEW
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
OVERVIEW
The federal government has over sixty-five years of experience with polygraph
techniques in criminal and counterintelligence testing. Beginning in 1951, the
Department of Defense established a polygraph school for the training of military
examiners. Today, that course is considered second to none. It conducts basic and
advanced courses to train almost all of the polygraph examiners for the federal
service.
The use of the polygraph for the investigation of crimes is a well established
practice. Known errors in the field use are exceedingly rare. Where examinees are
found to be deceptive during testing, the confession rate is consistently high,
despite the fact that all of these criminal suspects had already been interrogated
by an experienced investigator. Moreover, there is nothing about the polygraph
technique that likely to cause a false confession because of the requisite low key
questioning. The use of the polygraph to confirm admissions and confessions with
additional testing also reduces the possibility of false confessions. This study
includes a number of representative cases of criminal investigations in which the
polygraph has been a significant factor.
The polygraph is extremely useful in intelligence and counterintelligence
operations. This is evidence of the deterrent effect of screening examinations.
Examples of espionage and attempted espionage cases detected by polygraph
examinations are included in this study. Without the polygraph as an investigative
tool, a number of espionage cases never would have been solved. Helmich, Kampiles,
and Barnett probably would not have been successfully prosecuted without the
skillful application of polygraph techniques. In addition, there is definite
evidence that some extremely sensitive U.S. intelligence operations would have been
penetrated by hostile intelligence services if the polygraph had not been employed
in screening for clearance and access. Examiners conducting screening cases have
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
obtained confessions from applicants of :heir intention to commit espionage. In
other cases they developed such significant admissions that penetration attempts by
hostile intelligence were detected and neutralized. Screening has also kept our
intelligence agencies from hiring some e::tremely undesirable people. Examiners, in
FY 82, obtained admissions from applican:s of undetected crimes involving murder,
attempted murder, arson, rape, and numerous other felonies.
The polygraph field is one of those :are situations where the practice has
outpaced the research. While scientists are debating how the various testing
formats work, and are trying to devise a satisfactory theoretical framework, others
are conducting research on the development of improved techniques and advanced
instrumentation. Much of the literature which addresses the issue of polygraph
validity was conducted for other purposes. Some of the research is very limited in
scope, some is dated, and some is flawed in design. Despite these limitations they
produce results significantly above chance. Those studies which approximate field
conditions by using field instruments, trained and experienced examiners,
standardized techniques, and a complete field pretest interview, produce higher
rates of accuracy than those conducted with one physiological channel, make-shift
or novel techniques, and untrained personnel.
The polygraph is not limited to our culture. It is significant that research
results are similar throughout the world, regardless of varied cultures and
languages.
It is important to realize that polygraph examinations are not conducted in
isolation. Their use is always in the context of a larger program. They play a
role in investigations but they are never a substitute for invegtigation.
Polygraph examiners do not make recommenrlations or decisions in regard to further
investigation or the adjudication of the results. In screening, the polygraph
examination is not used alone for clearance and access. It is an adjunct to the
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
national agency check and background investigation. Decisions and adverse actions
are not made solely on the results of an anlysis of polygraph charts. All other
factors in a situation are considered before decisions are made.
Used with prudence, and a full knowledge of its limitations, the polygraph will
continue to play a role in our criminal justice system and counterintelligence
operations.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
Chapter 1
Research on Polygraph Criterion Validity and Reliability
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
RESEARCH ON POLYGI PH CRITERION VALIDITY AND RELIABILITY
FIELD STUDIES
Field studies have investiga ad both the validity and reliability of the
polygraph using various criteria These studies typically use field model
polygraphs and record respiratiol, electrodermal activity, and cardiovascular
activity. Professional polygrap examiners are typically used, but they differ
widely in experience. Some exam -kers are still in internship training programs;
others have as much as 20 years ( f field experience. Field studies have used both
relevant-irrelevant and control--lestion techniques. A description of the
standardized formats appears in le Chapter on Utility.
The original examiner's deci ion of truth or deception is often a clinical
judgment based on information frcu physiological recordings, subject behavior, and
case facts, although some data oI the numerical scoring of polygraph charts also
will be discussed. Most of the 'ield studies suffer from significant
methodological problems.
Two types of studies will be iiscussed. First, studies investigating the
relationship between the examine 's decision and some criterion of truth or
deception will be presented as s idies of polygraph criterion validity. All of
these studies use the clinical d agnosis of the original examiner to predict the
criteria. However, two of the s idies (Harland and Raskin, 1978; Ginton, Daie,
Elaad and Ben-Shakhar, 1982) als use both numerical scoring by the original
examiners, and decisions of exam lers not present during the test based on
numerical evaluation and chart s utiny to predict truth or deception.
Second, studies investigatin the relationship between the decisions of a blind
examiner and actual truth or dec )tion using verified polygraph cases will be
discussed. Some of these studie have been used as evidence for polygraph
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
criterion validity and there are problems with this interpretation. The final
section focuses on the reliability of the polygraph examiner's decision.
FIELD STUDIES OF CRITERION VALIDITY
Field studies of criterion validity have used various criteria of ground
truth. Determination of ground truth is the major problem for such studies. The
ideal criterion is independent of the polygraph examiner decision of truth or
deception and accurately reflects subject guilt or innocence. If examiner decision
and the criteria are not' independent, the accuracy of the polygraph in correctly
classifying innocent and guilty subjects will be reported as spuriously high.
However, most of the independent determinations of ground truth probably produce
spuriously low correct classification rates due to their nature. There is only one
study where the criterion is both independent and accurate (Ginton, et al. 1982).
This study will be discussed in detail at the end of this section.
Several studies have used the criterion of case disposition to assess the
accuracy of the examiners' decision (Lyon, 1936; Peters, 1982; Edwards, 1981).
Case disposition as a criteria may be neither independent nor accurate. Lyon
(1936) determined the dispositions of randomly-selected cases involving 100
juvenile criminal suspects (81 boys and 19 girls over age 15). Twenty polygraph
tests were classified truthful and 80 were classified as deceptive. These cases
were followed to determine if examiners' results were corroborated by positive
evidence or confession. Only 40 cases could be verified. Seven cases classified
as truthful by the examiner were verified as truthful by additional evidence.
Thirty-three of the deceptive cases could be verified by confessions; some of the
confessions occurred immediately after the test. Thus, the examiner had correctly
classified all cases. Lyon, however, gives insufficient detail about the
verification procedure. Testing involved a unique procedure but was in the
relevant-irrelevant category. Respiration and cardiovascular activity were .
recorded.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
Peters (1982) reported on the disposition of 220 cases involving stipulated
polygraph examinations conducted at a state crime laboratory from 1976 to 1979.
Stipulation means that the prosecution, defense, and judge agree to admit results
of the polygraph examination into evidence with the examiner testifying and subject
to cross?examination, although the examiner actually testified about the subject's
truth or deception in only 11 cases. Control?question procedures and numerical
chart analysis (with an inconclusive range) were used. Of the initial 220
examinations, 25 were eliminated because they were inconclusive, 23 cases were
eliminated because the test could not be completed. The disposition of 100 cases
indicated that the subject was innocent or telling the truth (95 criminal suspects
had charges dropped, 1 criminal suspect was acquitted, and 4 victims or witnesses
were involved in trials where the accused was convicted). The polygraph examiner
had classified 88% of these cases as truthful. The disposition of 63 cases
indicated the subject was guilty or lying (10 criminal suspects were convicted at
trial, 51 pleaded guilty, and 2 alleged victims were involved in trials where the
accused was acquitted). The polygraph examiner had classified 98% of these cases
as deceptive. Six cases were pending and 3 cases had charges dropped with
conditions (these cases are not included in the correct classification figures).
Edwards (1981) surveyed licensed Virginia polygraph examiners to determine the
correspondence between polygraph examiner decisions and dispositions of case in
specific issue examinations. Only results from law enforcement examiners will be
discussed; the response rate for private firms was so low that it is meaningless.
Of the 2,433 polygraph examinations, 959 (594 guilty and 365 innocent) were
verified by means such as acquittal, conviction, confession, etc. The examiners
had correctly classified 99% of the guilty cases and 97.5% of the innocent cases.
Lack of independence between examiner decision and the criterion can occur in
several ways and it is impossible to determine the extent the nonindependence
influences the percentage of correct classifications. The examiner can influence
?
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
the trial process either by increasing the probability of conviction or that
charges are dropped. This is a particular problem in the Peters study involving
stipulated polygraph tests. Peters does mention that the examiner testified about
subject guilt or innocence in only 11 cases, but it is unclear the extent to which
the results of the polygraph test were influential in the decision to go to trial
or drop charges. Confession also is not independent of the polygraph examination.
Examiners usually interrogate only those showing physiological responses to
relevant questions. Thus, those showing no deceptive responses will be under less
pressure to confess. It also is possible to induce a false confession, although
that is extremely rare. Some researchers think the use of confession invalidates
the study, other think confession is acceptable as a criterion (see Horvath, 1971;
Lykken, 1981; Raskin and Podlesny, 1979).
Field studies using case disposition also are problematical because they report
data on a restricted sample. -Typically, a large number of cases cnnnot be verified
and it is not clear if verified cases differ from unverified cases in the ease with
which examiner can make a decision about the case. Horvath (1977) provides
preliminary data that verified and non-verified cases may be quite Aimilar.
Two field studies have used panel judgments as the criterion by which to
validate polygraph examiner judgment (Bersh, 1969; Barland and Raskin, 1976).
Barland and Raskin (1976) is one of the studies Lykken (1979; 1981) cites in
maintaining that polygraph tests have a high rate of false positive errors.
Bersh (1969) investigated the reltionship between examiner judgment in criminal
investigations by the military services and unanimous guilt-innocence decisions by
a panel of four Judge Advocate General Attorneys who had complete access to case
files except for the polygraph examinations. Examiners also had access to a
subject's case file, but the author maintained the file was not complete and that
polygraph tests are only given when there is doubt about a person's guilt or
innocence. Cases were selected at random from 1963 to 1966 with some
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
restrictions. Cases involving confession by the subject and inconclusive judgments
by the examiner or the panel were eliminated. Cases were selected so that there
were similar numbers of deception and no deception examiner judgments and a mix of
zone comparison tests, Modified General Question Tests and relevant-irrelevant
tests. The first two formats are control question tests. However, Bersh grouped
the Modified General Question Test and the relevant-irrelevant test in reporting
the data. This was referred to as the General Question Test. Decisions by the
panel were unanimous in 157 cases. With the General Question Test group, the
examiner classified as deceptive 97% of the cases the panel judged as guilty, and
classified as truthful 90% of the cases the panel judged as innocent. With the
zone-comparison test, the examiner agreed with 90% of the cases the panel judged as
guilty and 94% of the cases the panel judged as innocent.
Barland and Raskin (1976) investigated the correspondence between polygraph
examiner decisions and panel decision (I judge, 2 prosecuting attorneys, 2 defense
attorneys) for 92 sequential independent cases involving criminal suspects. Panel
decisions were based on evidence collected by untrained investigators and the panel
had no knowledge of the outcome of the polygrAilph examinations. Polygraph
examinations used control-question procedures. Panel decision was based on a
numerical score by each panel member. The original examiner used field numerical
scoring techniques and made a clinical decision from both this and other (e.g.
subject's behavior) information.
The panel unanimously made decisions in the same direction in only 22 cases
(24%) so these data will not be discussed separately but are included among those
66 cases (72%) where the criterion was a majority of the panel making a decision in
the same direction. Among the 66 cases, 100% of the guilty subjects and 35.7% of
the innocent subjects were correctly classified (omitting inconclusives) by the
original examiner's clinical diagnosis. With the same criterion, original
examiner's numerical score correctly classified 100% of the guilty subjects and
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved ForRelease2013/04/16 : CIA-RDP07-01280R000400400001-7
27.3% of the innocent subjects (omitting inconclusives). Numerical scoring by an
independent examiner correctly classified 97.5% of the guilty subjects and 45.5% of
the innocent subjects.
When judicial outcome was the criterion, 100% of the guilty subjects were
correctly classified by the original examiner's clinical diagnosis, by the original
examiner's numerical analysis, and by the independent evaluator (omitting
inconclusives). The original examiner's clinical diagnosis and numerical analysis
correctly classified 14% and 17% of the innocent subjects, respectively (omitting
inconclusives) with the judicial outcome criterion. The independent examiner
correctly classified 43% of the innocent subjects.
Thus, the Bersh, and Barland andRaskin studies both find high agreement
between panel decision of guilty and a polygraph examiner decision of guilt, but
differ dramatically in agreement between panel and examiner for innocent cases.
Several possible reasons have been suggested for this discrepancy (see Lykken,
1979; 1981; Barland, 1982; Raskin, 1978). In the Bersh study, examiner decisions
were not totally independent of panel decisions since the examiner had access to
limited case data - eritugh to conduct the polygraph test. This may have increased
the correspondence between examiner decision and panel decision beyond that which
is representative. However, Berland and Raskin note that the panel in their study
often received case files with only small amounts of information. The panel did
not necessarily have the police reports, but did have the subject's explanation of
the events (Berland, 1982). This lack of information is reflected by the few cases
of unanimous panel decision. Barland and Raskin also note that 34 of the cases
were tested with a polygraph in poor mechanical condition suggesting that the
polygraph examiner had poor charts in certain instances. Thus, the figures
reported by Barland and Raskin may underrate the typical agreement between examiner
results and a panel decision.
Panel judgments pose a further problem as a criterion. Although the panels
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
.4/ Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
X.
were to judge cases only on merit and ignore legal technicalities, our judicial
system classifies a person as guilty only with overwhelming evidence. Thus,
panels composed of persons with legal training may be biased towards calling people
innocent if there is any doubt about their guilt.
There is only one study where actual guilt/innocence of the subject is
independent of the polygraph examination and in which there is certain knowledge of
ground truth (Ginton, et al., 1982). This study has been classified as a field
study because it used a real-life situation and subjects thought there would be
consequences from failure to pass the polygraph test.
Twenty-one Israeli policemen were administered a series of pencil-and-paper
tests presented as part of a police course. Beneath the answer sheet for one of
the tests was a hidden chemical page that recorded what was written on the answer
sheet. The chemical page was removed, the answer sheet returned and subjects
scored their own tests. Seven subjects actually changed their answers during the
scoring. Several days later subjects were informed that cheating was suspected and
were offered an opportunity to take a polygraph test. They were told their career
liplay depend on test results. Fifteen subjects, 13 innocent and 2 guilty, were
actually tested with control-question procedures administered at police
laboratories by professional polygraphers. Electrodermal activity, cardiovascular
activity, and respiration were recorded. There were three participants in each
case - an examiner, an observer in an adjacent room, and a blind chart evaluator.
Thus, each participant had different information by which to make a decision of
truth or deception. The examiner made,his decision based on the polygraph charts
and subject behavior, the observer made his decision based only on subject
behavior, and the chart analyst used only the polygraph charts. In addition to the
their overall impressions, the examiner and the blind chart evaluator also scored
the tests using field scoring techniques. Each relevant-control pair was scored
from +3 to -3 and summed across physiological response systems, questions, and
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
.4, Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
charts. Total scores from +5 to -5 were called inconclusive. Additionally, 8
other examiners analyzed the charts using both general chart scrutiny, and the
numerical scoring method.
Only the data from the examiner, the blind chart evaluator, and the 8
additional examiners will be presented. Using only his clinical judgment, the
original examiner correctly classified 100% of the guilty subjects and 85% of the
innocent subjects; there were no inconclusive decisions. Numerical field scoring
increased the number of inconclusives; one guilty subject and six innocent subjects
were classified inconclusive. Omitting inconclusives, the original examiner
correctly classified 100% of the-guilty subjects and 86% of the innocent subjects
using numerical field scoring techniques. The blind evaluator correctly classified
50% of the guilty subjects (none of the guilty were classified inconclusive) and
70% af the innocent subjects using only general chart scrutiny (omitting
inconclusives). Numerical field scoring increased the number of inconclusive
decisions and improved the accuracy (omitting inconclusives) of the blind
evaluator: 100% of the guilty subjects and 83% of the innocent subjects were
correctly classified.
The 8 additional examiners correctly classified 94% of the guilty subjects with
chart scrutiny and 100% (omitting inconclusives) of the guilty subjects with
numerical analysis. Omitting inconclusives, their decisions from general chart
scrutiny correctly classified 82% and their decisions from numerical analysis
correctly classified 83% of the innocent subjects.
There are a number of additional field studies but they have severe
methodological defects. Ben-Ishai (1962) reports 94% correspondence between
examiner judgment and case outcome as determined by a student who had collected
additional data on the case, but presents no detail on methods used in the
polygraph examination or in case verification. Bitterman and Marcuse (1947) report
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
100% agreement between polygraph decisions and actual truth with 81 subjects in an
actual crime but this study contained no subjects guilty of the crime. Elaad and
Schahar (1978) report that polygraph examiners correctly classified as deceptive
96.6% of the subjects who were actually guilty; however, this study contained no
innocent subjects. It is difficult to draw conclusions from studies using only
guilty or only innocent subjects because the examiner would have high correct
classification rates merely by a bias to call all subjects innocent or all subjects
guilty. There also are older studies reporting high correct classification rates
(e.g. Mac Nitt, 1942), but these studies give insufficient detail about procedures
and are impossible to evaluate.
Field studies of criterion validity are fairly consistent in their results for
guilty subjects. When original examiner subjective decision is used to'predict the
criteria, 90% to 100% of the guilty subjects have been correctly classified.
_Results are similar for studies using a criterion that is independent of examiner
decision and studies using a criterion correlated with examiner decision.
Numerical field scoring by the original examiner, chart scrutiny by a blind
examiner, and numerical analysis by a blind examiner does not appear to alter the
percentage of guilty subjects correctly classified when inconclusives are omitted.
The results for innocent subjects are much more variable. When original
examiner's clinical judgment is used to predict the criterion, omitting
inconclusives, from 14% to 100% of the innocent subjects are correctly classified.
The Berland and Raskin (1976) was the only one to show poor results; problems with
this study have been discussed. Omitting the Berland and Raskin data, 85% to 100%
of the innocent subjects were correctly classified. It is important to note that
the study reporting a 85% correct classification figure used a criterion that was
both independent of original examiner decision and accurate (Ginton
et al., 1982).
Percentage of correct classification for innocent subjects does not appear to
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
change when the original examiner uses numerical scoring and when a blind evaluator
makes either clinical or numerical decisions. Additional data on the relationship
between different types of examiner decision is discussed in a subsequent section.
Relationship Between the Decisions of Blind Examiners and Ground Truth Using
Verified Polygraph Cases
The studies discussed below investigate the ability of examiners working in the
blind to correctly classify subjects as truthful or deceptive using verified
cases. Since ground truth is known, these studies are cited by some researchers as
providing evidence on polygraph validity. For example, Lykken (1978; 1981) cites
Horvath (1977) as a validity study and uses it as evidence that polygraph tests
misclassify large numbers of people.
There are several problems with these studies. First, they do not provide
information about the validity of the original examiner's decisions unless the
original examiner makes his decision using only information from the polygraph
charts (this is more likely when numerical scoring is used). In most cases the
original examiner uses both the polygraph charts and additional information in his
decision (e.g. subject behavior and case information). Since the two decisions
were based on different types of information, these studies will only provide
information on the validity of the blind evaluator's decision based
polygraph charts (Barland, 1982).
It is debatable whether these studies are even validity studies of the blind
evaluator's decision because the description of sample selection in most of the
studies suggests only verified cases confirming the original examiner's decision
were used. If cases where the original examiner's decision is incorrect are
eliminated from the sample, the sample has a very restricted range. A true
validity study also would investigate how the blind examiner scores the cases not
correctly classified by the original examiner. However, incorrect decisions were
solely on the
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
not necessarily eliminated. All of
with the original examiner's decisio
reporting of sample selection in all
them problematical.
It could be said that use of a
confirming the original examiner's df
blind examiner's decision and that v.
reported in these
original examiner
examiner making a
studies. However,
makes influences t
clinical judgment
polygraph charts may make a decision
examiner using only numerical scorinf
influences the examination procedure
evidence for the validity of the bli;
A set of studies were conducted ;
specializing in polygraph examinatio-
1973; Wicklander and Hunter, 1975; S
similar basic procedures. Polygraph
deceptive cases verified by confessi,
contain similar numbers of verified
Polygraph examiners of varying exper
Control?Question Technique and the e
guilt or innocence based on the poly,
knowledge of case facts. Electroder
respiration were recorded. These ch
subject guilt or innocence, and a de
polygraph charts was made. Thus, th
original examiner bases a decision o
e verified cases Horvath (1977) found agreed
Thus, there is enough confusion in the
f the studies using verified cases to make
stricted sample of only those verified cases
ision would overestimate the validity of the
idity could not be any higher than what is
t is possible that the type of decision the
polygraph test. For example, the original
om subject behavior, case facts, and
ooner, possibly with fewer charts, than an
If the type of decision an examiner makes
these studies should not be considered as
examiner's decision.
John E. Reid and Associates, a private firm
(Horvath and Reid, 1971; Hunter and Ash,
wick and Buckley, 1975). These studies use
harts from equal numbers of truthful and?
selected from the files. The cases
uthful and verified deceptive questions.
nce administered the standard Reid
miner made a clinical decision of subject
aph charts, subject's overt behavior, and
1 activity, cardiovascular activitiy and
ts were then reanalyzed by examiners blind to
sion based only on the information in the
Reid studies indicate the extent to which the
the polygraph charts as opposed to!other
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
information. It is important to consider that these studies were conducted at John
E. Reid & Associates because the Reid technique emphasizes the use of nonverbal
indicators (Mullenix and Reid, 1980; Reid, 1980). None of the Reid charts were
scored numerically. Most other schools and techniques give little or no weight to
nonverbal behavior.
Horvath and Reid (1971) reported the reanalysis of 40 cases by 10 polygraph
examiners. Seven examiners had more than one year experience and 3 were interns,
having 4 to 6 months of experience. The 40 cases contained no charts that were
obvious to any examiner and none that were uninterpretable. Examiners could call
individual questions inconclusive but had to make a guilt/innocence judgment for
each case. The 10 examiners correctly classified 85% of the guilty cases and 90.5%
of the innocent caes. Omitting inconclusives, 89.6% of the truthful questions and
86.7% of the deceptive questions were correctly classified. The experienced
examiners were more accurate', averaging 91.4% correct case classifications and the
inexperienced examiners averaging 79.2%. Experienced examiners also had higher
correct classification rates for individual questions than inexperienced
examiners.
Hunter and Ash (1973) reported on data on the blind rescoring of polygraph
records. Seven examiners (6 with more than 1 year of experience and 1 intern)
reanalyzed data from 20 cases on two occasions separated by at least 3 months.
Examiners did not know they were rescoring the same charts. Similar correct
classification rates were found on each occasion. Occasion-to-occasion agreement
for the same examiner on the individual cases ranged from 75% to 90% (average 85%)
and on individual questions ranged from 67% to 91% (average 81%). Combining
occasions and omitting inconclusives, examiners correctly classified 86% of the
truthful cases, 88% of the deceptive cases, 85% of the truthful questions, and 88%
of the deceptive questions. The authors note that the poorest results were
obtained from the examiner with the least experience.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
The effect of receiving all of the case information on the scoring of polygraph
charts was reported by Wicklander and Hunter (1975). Six experienced examiners
(mean experience = 4 years) rescored 20 cases using only the polygraph charts.
Omitting inconclusives, 95% of truthful cases, 92% of deceptive cases, 91% of
truthful questions, and 91% of the deceptive questions were correctly classified.
For the second review, examiners were given the polygraph charts, a description of
the case, verbal and nonverbal behavioral symptoms noted by the original examiner,
and relevant test questions. This additional information primarily reduced the
unmet of inconclusive decisions. Omitting inconclusives, 91% of the truthful
cases, 98% of the deceptive cases, 92% of the truthful questions, and 97% of the
deceptive questions were correctly classified.
Slowick and Buckley, (1975) investigated the effect of using all physiological
measures versus each measure independently on percentage of correct decisions.
Seven'experienced examiners (mean experience = 3.8 years) first rescored the charts
from 30 cases using all physiological data and then analyzed each physiological
measure independently (at 3-month intervals to avoid recall). Using all
physiological measures and omitting inconclusive results, the seven examiners
correctly classified 93% of the truthful and 85% of the untruthful cases, and 93%
of the truthful and 80% of the untruthful questions. Analysis of individual
physiological measures revealed slightly lower correct classification rates;
differences between percentage of correct classifications with each measure were
small.
The remaining studies used cases from a variety of sources. Kleinmuntz and
Szucko (1982a; 1982b) report the lowest figures of all the studies. Polygraph
examinations from 50 verified innocent and 50 verified guilty subjects suspected in
crimes of theft were obtained. Six polygraph trainees, in the last few weeks of
their internship training program, reanalyzed each chart on an 8-point scale with 1
representing definitely truthful and 8 definitely untruthful. Trainees knew the
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
proportion of guilty subjects but did not know who was guilty or innocent. When a
single question sequence from one polygraph chart was scored, correlations between
trainees' ratings and actual guilt or innocence ranged from .26 to .52. Three of
the examiners also scored complete charts. Correlations between their ratings and
actual guilt/innocence ranged from .45 to .55.
Horvath (1977) reported data on the effect of several variables on analysis of
polygraph recordings. Ten field-trained polygraph examiners from law enforcement
agencies (5 having more than 3 years experience and 5 having less than 3 years
experience, including some interns) judged a sample of polygraph records from 112
criminal suspects from independent cases. The records were sampled so that there
were similar numbers of verified.and unverified cases, truthful and deceptive
subjects, and crimes against people and crimes against property. Data on the
unverified cases will be presented in a later section. Polygraph examinations used
standard control-question procedures; electrodermal activity, respiration, and
cardiovascular activity were recorded. All examinations included at least 2 charts
of the control-question test and 1 stimulation test; 48 examinations (43%) included
additional charts. The original examiner decision was based on all of the
information in the polygraph examination; decision of the 10 other examiners was
solely based on a scrutiny of polygraph charts.
For verified cases, 51% of the truthful cases were classified as truthful by
the polygraph examiner, and 77% of the deceptive cases were classified as deceptive
by the examiner. Results were similar for verified cases of person crime and
verified cases of property crime.
Experience of the polygraph examiner did not influence accuracy. There were
effects for number of charts. Seventy-one percent of the records containing only
the basic two relevant charts were correctly classifed, whereas 52.5% of the
records containing more than the basic battery were correctly classifed (verified -
and unverified cases combined).
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-sr
Raskin (1976) reports on the rescoring of records from 12 confirmed guilty
subjects and 4 confirmed innocent subjects from the Barland and Raskin (1976)
study, by 25 field polygraph examiners with different types of training and
experience. Eighteen examiners had at least one year of experience, and 13 had
formal training in numerical scoring, of whom 7 explicitly used numerical scoring
in their reanalysis of the data. Excluding inconclusives, 90.5% of the 400
decisions were correct. It is not possible from the published data to determine
the percentage of innocent and guilty subjects correctly classified because Raskin
does not report the percentage of inconclusive results for guilty and innocent
subjects separately. If it is assumed that inconclusives are evenly divided, 73%
of the innocent subjects and 95% of the guilty subjects would have been.correctly
classified. Examiner experience did not affect accuracy but other variables did.
Examiners who attended schools emphasizing numerical scoring had higher accuracy
than examiners attending other schools (97.1% versus 86.9%). Examiners explicitly
using numerical soring had higher accuracy (98.9%) than all the remaining examiners
(87.9%) and the subsample of examiners trained in, but not explicitly, using
numerical scoring (88.5%).
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
This study also provides some data on the reliability of the inconclusive
category. This information is not available in the other studies because they only
used cases where the original examiner made a guilt/innocent decision. When the
original examiner's decision was based on more than numerical chart analysis, the
blind examiner agreed with 85% of cases classified as truthful, 78.9% of the cases
classified as deceptive, and 53% classified as inconclusive. When the original
examiner's decision was based only on numerical chart analysis, the blind examiner
agreed in 91% of the cases classified as truthful, 87% of the cases classified as
deceptive, and 71% of the cases classified as inconclusive.
Edel and Jacoby (1975) reported on agreement between the original examiner and
blind examiners, and between blind examiners, when both were making the same type
of decision. The forty randomly selected cases were from pre-employment screening
examinations for a federal agency and contained 25-30 relevant, irrelevant,
control, and overall truth questions. Ten experienced examiners (interview
experience from 3 months to 10 years) analyzed the charts. Each examiner conducted
the original examination in 4 cases and was the blind rater in 8 cases. Thus, each
case had one examiner and one blind rater. Examiners and raters were to decide
whether a significant physiological response or no significant physiological
response occurred to each question in each physiological system recorded
(electrodermal activity, cardiovascular activity, and respiration). The analysis
by the original examiner was not done at the time of the examination. It was done
later when the examiner also was analyzing the charts from the other examiners'
cases.
The examiner and the raters agreed on 96% of the decisions, averaged across
physiological measures. The blind raters agreed in 94% of the cases when
agreements were averaged across physiological measures. The percentages of correct
decisions for each physiological measure were similar.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
The studies using blind evaluators to analyze polygraph records of verified
cases report a range of correct classifications for guilty and innocent subjects.
They are difficult to interpret because the method of sample selection is not
clear. If, as many studies indicate, the only cases used were those in which case
outcome confirmed the original examiner's decision, then the studies only provide
information on how a blind evaluator agrees with the original examiner. When
original examiner decision is a clinical diagnosis from case facts, subject
behavior, and polygraph charts, it is not surprising that agreement would be low.
Furthermore, the type of decision an examiner makes may influence how a test is
conducted, making chart usage from one type of examination to make another type of
decision problemmatical.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Relationship Between Different Types of Examiner Dec aions When Ground Truth
is not Known
Studies investigating the relationship between c fferent types of examiner
decision when ground truth is not known and the exan lers make decisions using the
same type of information are reliability studies.
Test reliability refers to whether test results re repeatable; for example,
when the test is administered on different occasionE 3r when it is administered by
different people are the same results obtained. Te E reliability provides some
clues as to potential test validity as a test will t able to predict a criterion
only to the extent it is reliable. It should be rec ;nized, however, that even if
a test is very reliable, it may not be valid.
There are different types of reliability. Critl al issues for polygraph
examinations are the extent to which different exami ars make the same decision of
truth or deception from the same polygraph examinati a, and the extent to which an
examiner will draw the same conclusions when the exE ination is administered to the
same subject on different occasions. There is no it, 3rmation on
occasion-to-occasion repeatability of polygraph exam aations. There is some
information on agreement of different examiners whet their decision is made from
the same type of information.
Barland and Raskin (1976) report data on the ag/ ament between a blind examiner
and the original examiner when the original examiner iecision was not necessarily
correct. Details of this study are already presente in the Validity section. The
original examiner had numerically scored the polygrE 1 charts although the
decision, in some instances, was based on more than americal chart analysis.
Omitting inconclusives, the blind examiner classifiE as truthful 100% of the cases
classified as truthful by the original examiner, anc :lassifed as deceptive 98% of
the cases classified as deceptive by the original e3 nicer. When the original
examiner decision was based only on numerical chart lalysis, the two examiners
agreed in 100% of the cases, excluding inconclusiveE
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
The Kleinmuntz and Szucko (1982) study already discussed reported information
on the relationship between the ratings of 6 examiner trainees. Trainees analyzed
one chart on an 8-point scale with 1 representing definitely truthful and 8
representing definitely untruthful. Correlations between the scores assigned to
the records by the different trainees ranged from .24 to .56. Three of the
trainees also scored complete charts; correlations between the three trainees'
ratings ranged from .45 to .61.
The Horvath (1977) study provided information on the agreement between the
original examiner and ten blind evaluators for unverified cases in addition to the
verified case data already discussed. Since the blind evaluators used different
information than the original examiner, this study cannot be used as evidence for
the reliability of original examiner decision. For the unverified cases, the blind
evaluator agreed with 52% of the cases the original examiner classified as truthful
and 72% of the cases the original examiner classified as deceptive. However, there
was considerable difference in the agreement for person and property crimes with
more agreement for innocent subjects in cases of property crime and more agreement
for guilty subjects in cases of person crime. Horvath also reported significant
consistency among the blind evaluators both for verified (r=.89) and unverified
(r=.85) cases.
All but one of the field studies investigating the repeatability of examiner
decisions based on the same type of information report high reliability.
Unfortunately, all of the data is either on the repeatability of numerical scoring
or the repeatability of blind examiner chart scrutiny. There appears to be no data
on the relationship between clinical diagnosis of the original examiner and the
clinical diagnosis of another examiner when both, have equivalent information about
the case. The Kleinmuntz and Szucko (1982) study was the only one to repeat low
reliability. As previously noted, the examinersi in this study were trainees and
used only one chart to make their decision.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
There are various other reports of the reanalysis of cases that were verified
and confirmed the original examiner's decision. None of these studies report how
the original examiner made the decision (this was not necessarily the purpose of
the study) and it cannot be determined whether errors reflect differences in
information available to make a decision or whether they are random. Davidson
(1982) provided evidence on the reanalysis of 10 confirmed deceptive and 11
confirmed truthful property crime cases. Seven examiners (experience 6 months to
14 years) numerically scored records (U. S. Army Polygraph School procedures)
containing different types of physiological recordings. The examiners first
evaluated only Cardio Activity Monitor recordings (a device designed for the U. S.
Air Force), then only the electrodermal, respiration, and cardiovascular activity
records, and, finally, all recordings to make judgments of truth deception or
inconclusive. For guilty cases, 90% of the decisions with the Cardio Activity
Monitor, 87% of the decisions with.the polygraph and 90% of the decisions with both
were correct. For innocent cases, 85.7% of the decisions with the Cardio Activity
Monitor, 84% of the decisions with the polygraph, and 88% of the decisions with
both were correct. Inconclusive decisions were considered incorrect decisions and
it is not possible from the presented data to determine the percentage of guilty
and innocent cases correctly classified, omitting inconclusives.
Ben Ishai (1962) reports 10 criminal cases (number of guilty and innocent not
known) were accurately classified by comparing average electrodermal response to
relevant and control questions. Widacki (1982) reported that 91.6% of guilty
(n = 16) and innocent (n = 22) confirmed cases using control-question tests were
correctly classified when charts were reanalyzed using the Backster numerical
scoring method. Separate percentages are not given for guilty and innocent
subjects. Guertin and Wilhelm (1954) reported that factor analytically derived
measures of electrodermal activity from relevant-irrelevant tests on 19 guilty and
15 not guilty criminal suspects correctly classified 95% of the guilty subjects and
94% of the innocent subjects. ?
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4 Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Laboratory Studies
In laboratory research the criterion (ground truth) is both independent and
accurate because it is controlled by the experimenter. Many laboratory studies
have used a mock crime procedure. Subjects are designated as guilty or innocent by
the experimenter and the guilty subjects participate in the crime (typically
stealing money). The consequence for failure to produce a truthful outcome on the
test (i.e. guilty subjects deceiving the examiner and innocent subjects
demonstrating their innocence) is generally loss of a monetary bonus. Only a few
laboratory studies have used a professional polygraph examiner to test subjects and
score records. Many of the examinations are conducted by graduate students under
the supervision of psychologists-or psychophysiologists with experience in
psychophysiological recording techniques. Control-question, peak-of-tension,
guilty knowledge, and relevant-irrelevant techniques have been used. Most of the
studies measured respiration, cardiovascular activity, and electrodermal activity,
although many of the studies using the peak of tension or guilty knowledge tet
formats measured only electrodermal activity.
Many of the studies have used control-question techniques in the context of
determining how other variables such as drugs (Gatchel, Smith and Kaplan 1983),
subject characteristics (Raskin and Hare, 1978; Hammond, 1980), information and
practice (Rovner, Raskin and Kircher, 1978), and countermeasures (Honts, 1978)
effect detection and deception. Although this information is important, it is the
purpose of this report to review polygraph accuracy on samples that have not
received,special procedures. Thus, when studies have manipulated other variables,
only information from the normal guilty and normal innocent subjects receiving
standard polygraph tests will be analyzed. The results of the manipulated variable
are listed in the summary of each study.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Control-Question Test - Criterion Validity
A set of studies by Raskin and his colleagues (Berland and Raskin, 1975;
Rovner, Raskin, and Kircher, 1978; Raskin and Hare, 1978; Podlesny and Raskin,
1978) have used similar procedures to investigate detection of deception in a
laboratory situation. Subjects are designated guilty or innocent; guilty subjects
steal money or an item from a particular location and receive a monetary award for
deceiving the examiner. Motivation for innocent subjects to appear innocent varies
with the study. The examiners used in the Raskin studies were relatively
inexperienced, but Barland, Raskin and Podlesny had attended basic polygraph
training courses. Several of the studies recorded many physiological responses,
but classified guilty and innocent subjects only from a combination of
electrodermal, respiratory, and cardiovascular recordings. All of the studies use
a ten-item control question test. Charts are evaluated with numerical field
scoring techniques in which each physiological response measure on each
control-relevant question pair is scored from +3 to -3-based on the magnitude of
the difference, and summed across physiological response measures and questions.
Scores below a certain level indicate deception, above a certain level indicate
truthfulness. An inconclusive range also is used.
Berland and Raskin ( 1975) reported on 36 innocent and 36 guilty subjects
receiving three presentations of the test. Charts were scored both by the original
examiner (experience not discussed) and 5 additional examiners who were
professionally trained and experienced in numerical scoring techniques. The
authors note that the electrodermal and cardiovascular recordings were not of high
quality. Innocent subjects were told the details of the crime and that, if they
did not appear innocent on the test, they may not get course credit for
participating. An inconclusive range from +4 to -4 was used. The original
examiner correctly classified 88.5% of the guilty subjects and 71% of the innocent
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
(
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
subjects when 25 inconclusives were omitted from the analyses. Correct
classifications for the five additional examiners ranged from 79% to 86% (mean =
82%, results are not presented separately for innocent and guilty subjects).
Rovner, Raskin, and Kircher (1978) reported on the effects of information and
practice on detection of deception. Only the group receiving the standard
polygraph test will be discussed in this report. There were 12 innocent and 12
guilty subjects in the standard group and they received three presentations of the
test. The inconclusive range in this study was from +5 to -5. Money was used to
motivate innocent subjects to appear truthful on the polygraph test. Innocent
subjects were not told the details of the crime; they were merely told that a theft
had occurrred. Classification results are from the scoring of another evaluator
who did not observe the test. Omitting the 2 inconclusives, 100% of the guilty
subjects and from 90% to 92%* of the innocent subjects were correctly classified.
Similar correct classification rates were reported by Raskin and Hare (1978) in
a study on the effect of psychopathy on detection of deception using a professional
polygraph examiner. Only the results from the 12 inncocent non-psychopaths and 12
guilty non-psychopaths will be discussed. Money was used to motivate innocent
subjects to produce a truthful outcome on the test. The authors state that the
innocent subjects were informed about the crime. The examiner continually stressed
the importance of being truthful to all subjects thereby directing the attention of
innocent subjects to the control questions and guilty subjects to the relevant
questions. Subjects received from 3 to 7 presentations of the test and were
initially classified from the first three charts using inconclusive range of
between +5 and -5.
Since the study does not report the actual status of the inconclusive
subjects, it is not possible to cite a precise figure for the percentage of
innocents subjects correctly classified.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
If this classification was inconclusive, additional charts were scored. Between
each chart, questions were discussed with the subjects and reworded if necessary.
Omitting the 4 inconclusives, 100% of the guilty subjects and 91% of the innocent
subjects were correctly classified. Comparsion of classification accuracy for
subjects receiving more than three charts with those only receiving three charts
demonstrated that additional charts reduced the number of inconclusive decisions.
Podlesny and Raskin (1978) reported results both on the control-question test
and the guilty knowlege test. Only the control-question results will be reported
here; guilty knowlege tests result will be discussed in a subsequent section.
Twenty subjects were guilty and twenty subjects were innocent. Half of the
subjects received a control-question test with nonexclusive control questions.
Half received tests with exclusive control-questions. A nonexclusive
control-question has no limit as to time, and thus includes the period in which the
offense occurred. The exclusive control-question excludes the offense under
investigition by referring only to an earlier period in the subject's life. The
innocent subjects received a brief description of the crime but they did not know
its details. Money was used to motivate innocent subjects. Subjects received 3 or
more presentations of the test. Between charts subjects were questioned and the
wording of the control-questions was changed if necessary. Results for different
methods of quantifying results are reported; those using the conventional
inconclusive range of from +5 to -5 will be presented. Charts also were scored by
a blind rater who had not observed the polygraph test. Omitting the 4
inconclusives, 82% of the guilty subjects and 95% of the innocent subjects were
correctly classified. Slightly higher correct classification rates were found for
exclusive control-questions than non-exclusive control-questions.
There are a number of other studies using control-question procedures in a mock
crime situation. Dawson (1977) presents a brief report on detection accuracy with
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4 Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
ten guilty and ten innocent subjects. Different methods of scoring the responses
were used; the results with the standard procedure are of interest here. When
average vasomotor and electrodermal responses to relevant and control questions
were compared, 89% of the guilty subjects and 71% of the innocent subjects were
correctly classified omitting the 4 inconclusives. An additional study (Dawson,
1977, 1980) investigated the ability of actors using the Stanislavsky method of
acting to deceive the polygraph examiner, but will not be discussed since the
purpose of the test was to evaluate a countermeasure method and a variation in
polygraph technique.
Hammond (1980) reported on normal college students, alcoholics, and
psychopaths, but only the data from the college students will be presented. There
were 11 innocent and 10 guilty subjects. Instructions for the mock crime were
presented by tape and were designed _to motivate the subjects. Innocent subjects
listened to the same tape and, thus, knew the details of the crime. All subjects
received money for a truthful outcome of the test. Inexperienced examiners from
the Backster School of Lie Detection in the fifth and sixth weeks of their 7-week
training course conducted the tests using zone-comparison procedures and the
zone-comparison method of scoring. Respiration, electrodermal activity, and
cardiovascular activity were recorded. Testing involved a minimum of three charts,
but in most cases only two charts were used for analysis. The inconclusive range
in this study was between +8 and -8 . The charts also were analyzed by an expert
examiner. Omitting the 7 inconclusives, the student examiners correctly classified,
86% of the normal guilty subjects and 57% of the normal innocent subjects. The
expert examiner had 15 inconclusives and correctly classified none of the normal
innocents and 100% of_the normal guilty subjects.
Bradley and Janisse (1981) reported on 96 guilty and 96 innocent subjects.
Half of the innocent subjects and half of the guilty subjects were told that they
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4 Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
would receive an electric shock if they were judged guilty. Both control-question
and guilty knowledge tests were administered in counterbalanced order. Only the
data from the control-question will be discussed in this context. The test
consisted of three presentations of nine questions; subject were allowed to discuss
questions with the examiner in order to eliminate ambiguities. The numerical
scoring in this study used a more limited range than other studies; each
control-relevant pair was scored 1, 0, or -1 and the inconclusive range was from +2
to -2. Electrodermal activity, heart rate, and pupil size change were recorded and
subjects were classified as guilty or innocent for each physiological measure
independently. Only the electrodermal results will be presented, since the most
correct classifications occurred.with this measure, and heart rate alone has seldom
been a factor in field testing. Pupillometry is not in common use anywhere in the
field. Omitting the 56 inconclusives, 82% of the guilty subjects and 86% of the
innocent subjects were correctly classified. The authors note that their
procedures for the control-question test were not optimal because they did not
combine the physiological measures to classify the subjects, and their
semi-objective scoring procedure had a limited range.
The effect of physical countermeasures on the detection of deception was
investigated by Hants (1982). Only the data from the 12 innocent-and 12 guilty
subjects using no countermeasures will be discussed. Innocent subjects knew the
general nature of the crime, but were naive to its details. All subjects were told
they would receive extra points applied to their final course grades if they
produced a truthful outcome on the test. Respiration, electrodermal activity, and
cardiovascular activity were recorded by a licensed examiner with 5 years
experience. Subjects received four presentations of eight questions. Immediately
after the test, the examiner was required to make an innocent/guilty/inconclusive
decision based on both subjects' behavior and limited (less than 5 minutes) chart
analysis. The examiner also scored the charts by semi-objective methods using an
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
inconclusive range of from between +12 and -12. Another experienc
scored the charts using the semi-objective method. Results are ba
most productive charts. The original examiner's clinical decision
classified 86% of the guilty subjects and 60% of the innocent subjf
the 7 inconclusives. Semi-objective scoring by the original exami-
correctly classified 100% of the guilty and 67% of the innocent sul
the 11 inconclusives. Omitting inconclusive decisions, the origin,
the blind evaluator agreed in 1007. of the cases when decisions wer(
semi-objective scoring.
The effect of propranolol on detection of decption was studied
Smith, and Kaplan (1983). Only the results from the 7 guilty and
subjects not receiving drugs will be discussed. All subjects woul:
for appearing truthful on the test. Innocent subjects were told tl
had taken place but were not told its details. Blood pressure, elc
evaluator
d on the three
orrectly
ts, omitting
r evaluator
ects, omitting
examiner and
made with
y Gatchel,
innocent
receive a bonus
t a mock crime
.trodermal
activity, and respiration were recorded by two professional examin( s administering
a nine-item control-question test. Charts were numerically scored
inconclusive range of +5 to -5. Subjects appear to have received (
presentation of the test. Omitting the 6 inconclusives, 75% of th(
and 100% of the innocent subjects were correctly classified.
Kleinmuntz and Szucko (1982) reported on 15 innocent and 15 gu
Subjects were administered a control-question polygraph test by fol
were trainees in the first half of training. Tests were scored by
trainees with three months of experience and 3 with more experienc(
but only one chart of each set of three or more was used for the a:
were classified on an 8-point scale with 1 being definitely truthf
definitely untruthful. When a rating of 5 or more was used to ind
1th an
Ay one
guilty subjects
ty subjects.
examiners who
ix examiners (3
up to 8 years)
lyses. Tests
and 8 being
ate deception
and less than 5 indicated truthfulness, 71% of the guilty subjects nd 51% of the
innocent subjects were correctly classified. No inconclusive rang was used.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4 Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Correlations between examiner ratings and actual guilt/innocence ranged from .02 to
.43. Experience of the examiners scoring the tests did not influence the results.
Waid, Orne, and Wilson (1979) reported results on 15 innocent and 15 guilty
subjects. This study did not use a mock crime. Rather, guilty subjects learned
six code words during timed interpolated tasks. Innocent subjects performed the
same tasks but learned no words. Innocent subjects were told it was often
difficult to prove ones innocence on a polygraph test and, the authors note, care
was taken to involve the innocent subjects in the task. Guilty subjects were told
that highly mature and intelligent people could deceive the examiner. A
professional polygraph examiner administered several types of polygraph tests. The
first test consisted of three relevant and two control questions presented twice.
Between presentations, subjects were questioned about their answers and received a
stimulus test. Subjects were classified as deceptive if the response to any of the
three critical question was larger than the response to the control questions;
thus, this study used no inconclusive range. Only skin conductance was used to
classify the subjects. Eighty percent of the innocent and 80% of the guilty
subjects were correctly classified on each test presentation.
Waid, Orne, and Orne (1981) used a similar situation to investigate detection
of deception with 34 innocent and 40 guilty subjects. The polygraph test in this
study consisted of 3 presentaions of 9 relevant questions, 8 control questions and
5 irrelevant qustions and was administered by a professional polygraph examiner.
Different physiological measures were recorded and classification of guilty or
innocent was done for individual physiological measures. The best -results were
obtained for skin conductance. Subjects were classified as deceptive if the
average electrodermal response to the relevant questions was larger than the
average electrodermal response to the control questions and were classified as
innocent if the opposite relationship occurred; thus, no inconclusive range was
used in this study. Seventy-three percent of the guilty subjects and 76.5% of the
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
innocent subjects were correctly classified. This study also administered guilty
knowledge tests after the control-question procedure; these results will be
discussed later.
Widacki and Horvath (1982) compared the polygraph with fingerprints,
hand-writing specimens, and eyewitnesses identifying persons from full-faced
photographs in solving 20 mock-crimes. Subjects were divided into 20 groups of
four people. In each group, three subjects were innocent and one subject was
guilty. Each guilty subject was involved in an independent but similar case. The
guilty subject took an envelope to a specific building, gave it to a specific
person, and received a small parcel containing a cosmetic. Upon receiving the
parcel, he signed the receipt with a fictitious name. Polygraph examiners made
subjective decisions of guilty or innocence from a polygraph test using Reid
control-question procedures and recording respiration, cardiovascular activity, and
electrodertal activity. The examiner tested each group of four before making A
decision about who was guilty, Using the polygraph, 18 of the 20 cases were solved
(there was one error and one inconclusive decision). Thus, omitting inconclusives,
95% of the cases were solved. Handwriting and fingerprinting also were accurate
although there were many inconclusive decisions with fingerprinting. The
eyewitness identification results were very poor.
Berland (1981) is the only study to use control-question test in a mock
screening situation. Polygraph tests were administered by three experienced
examiners and respiration, electrodermal activity and cardiovascular activity were
recorded.
Subjects initially supplied biographical information which was verified by a
background investigation. Thirty subjects then supplied false biographical
information and 26 subjects supplied accurate biographical information to the
polygraph examiner. The polygraph test consisted of a minimum of three charts of
thirteen questions. If no decision could be made, additional charts were run with
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
a maximum of six charts. The examiner was not allowed to interrogate the subject.
Three methods were used to evaluate the charts - the zone method, the greatest
control method, and a relevant-irrelevant method. The data for the
relevant-irrelevant method will be presented in a subsequent section. The zone
method is similar to the method used is the previously cited control question
laboratory studies. Each control-relevant pair was scored from +3 and -3 and
summed; scores between +3 and -3 were inconclusive. The greatest control method
used the same procedure except that the relevant questions were evaluated against
the one control question with the largest reaction.
When subjects were classified deceptive or non-deceptive (irrespective of
whether the specific question was identified), omitting the 9 inconclusives, 81% of
the guilty and 76% of the innocent were correctly classified with the zone method.
With the greatest control method, omitting the 10 inconclusives, 68% of the guilty
and 83% of the innocent were correctly classifed. There were 250 truthful
questions and 30 deceptive questions (each deceptive subject lied to 1 question and
answered the remaining questions truthfully). Omitting inconclusives, the zone
method correctly identified 91% of the nondeceptive questions and 63% of the
deceptive questions, whereas the greatest control method correctly identified 54%
of the deceptive questions and 97% of the nondeceptive questions.
The laboratory studies of control-question techniques report a range of correct
classification figures for guilty and innocent subjects. Most studies report that
from 85% to 100% of the guilty subjects are correctly classified. Results for
innocent subjects are much more variable. Few of the studies used professional
examiners, but it must be recognized that most of the other studies used graduate
students under the supervision of psychophysiologists who were well-versed in
psychophysiological recording techniques. The lowest correct classification rates
for innocent subjects are reported by two studies using student examiners
(Kleinmuntz and Szucko, 1982; Hammond, 1980). The Kleinmuntz and Szucko study also
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
classified subjects from review of only one polygraph chart out of each set of
three or more. Other studies (Waid et al., 1979; 1981; Bradley and Janisse, 1982)
classified subjects as guilty or innocent by using each physiological measure
independently. It is well-known that all people do not respond in the same
physiological system (Engel, 1960; Garwood, Engel, and Capriotti, 1982), and higher
correct classification rates would likely be found if subjects were classified from
several psychophysiological measures as is done in field studies. For example, in
a Japanese study (Ohnishi, Matsuno, Arasuna, and Suzuki, 1976), the detection rate
for breathing was 46% and for electrodermal was 72%. When combined, the detection
rate rose to 92%. The possible importance of another variable should be
considered. In most of the studies, innocent subjects knew that a crime had been
committed but were naive to its details. In the Hammond, (1980) study, however,
innocent subjects heard the same tape that guilty subjects heard and thus, knew all
the details of the crime. The subjects in the Harland and Raskin (1975) study also
knew the details of the crime and this study reported the lowest correct
classification figures of the studies by Raskin and colleagues. However, it
appears that subjects in the Raikin and Hare (1978) study also knew the details of
the crime (although details are minimal) and this study reported high correct
classification figures. This study also used more sophisticated procedures than
the earlier Raskin studies. The effect of knowledge of crime details by innocent
subjects on classification accuracy certainly deserves futher consideration.
Several interpretations of these laboratory studies of control question .
technique studies are possible. In the worst case, the range of accuracy in
supporting truthfulness is 0 to 100%, and the range of detecting deception is 717.
to 100%. If those cases conducted by student examiners are removed from
consideration then, the range for truthfulness improves to 71% to 100%, and the
range for detecting deception is 73% to 100%. If the weighted average for all of
the studies is used, the truthful rate is 77.1% and the deceptive rate is 81.3%.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
If the cases done by students are removed; the truthful rate is 81.2% and the
deceptive rate is 81.9%. However, weighted averages presume a similarity in the
research methods, which was not always the case and these figures should not be
cited as reflecting the percentage of guilty and innocent people the polygraph
correctly classifies. One conclusion is clear, the closer the research
approximates field conditions by using field techniques, field instruments, trained
field examiners, and the standard field methods for chart analysis, the higher the
accuracy.
Control-Question Test - Reliability
A few of the studies investigated the relationship between the decisions of the
original examiner and those of an independent evaluator. All of these studies used
numerical field scoring techniques and examiner decisions were not based on subject
behavior. The procedures of each of the studies investigating reliability were
discussed. in the previous section.
All of the studies demonstrate that numerical field scoring techniques are very
reliable. Barland and Raskin (1975) reported that the correlations for the
numerical scores of pairings by examiners (one original and five additional) ranged
from .78 to .95 (mean = .86). Of the 1080 pairwise comparisons, examiners agreed
on 98% of the cases when inconclusive decisions were eliminated. Podlesny and
Raskin (1978) reported a correlation of .97 between the numerical scores of the
original examiner and one additional examiner, who had not observed the test. When
both examiners had made decisions, there was agreement in 100% of the cases. Honts
(1982) reported that the numerical scores of the original examiner and a blind
evaluator had a correlation of .88 and that there was 100% agreement when both
examiners had made decisions. Hammond reports that the correlation for student
examiners' numerical scores and expert examiner's numerical scores were correlated
.64 for the innocent condition and .61 for the guilty condition. It should be
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
JO.
noted that these correlations are computed for the guilty and innocent conditions
separately. If the correlation between the two examiners' numerical scores was
computed across all subjects, the correlation would be very high.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Relevant-Irrelevant Technique - Criterion Validity
Only a few laboratory studies have used relevant-irrelevant techniques in the
detection of deception. The chapter on Utility contains a description of one of
these'techniques. The studies vary considerably in their procedures.
Blum and Osterloh (1968) presented data on the ability of the
relevant-irrelevant technique to detect false information in stories involving
crimes in a laboratory situation. Subjects were police informants who had
previously given information on criminal cases to local or federal agencies. It
should be recognized that when informants are talking to the police their lives may
be in danger; thus, subject arousal in this study may have been high. The
situation was structured so that some subjects told completely true stories (n =
9), some told partially false stories (n = 7), and some told completely false
stories (n = 4). True stories were those that had been told on a previous occasion
and had withstood the scrutiny of investigation. Partially false stories were true
stories that had been told in the past, but with some items changed. Totally false
stories were invented, but contained credible information.
Three professional polygraph examiners conducted the tests. The authors note
that each story provided from five to eight relevant questions, but do not give
details of the polygraph test other than it took two hours. Physiological measures
used are not mentioned; however, since the examiners used a three-channel Stoelting
polygraph, presumably electrodermal activity, respiration, and cardiovascular
activity were recorded. Examiner decision was a clinical judgment of truth or
deception based on a scrutiny of the charts.
The examiners were able to correctly categorize informants telling totally
true, totally false, and partially false stories in 100% of the cases. There was
some error in classifying the 42 deceptive questions and 64 truthful questions.
Ninety-three percent of the deceptive questions and 98% of the truthful questions
were correctly classified.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4t. Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Similiar results were obtained by Correa and Adams (1981) who used the
relevant-irrelevant technique in a mock pre-employment screening session. Subjects
initially completed a Preemployment Data Sheet and nine items were selected. On
the subsequent polygraph test, the 20 designated deceptive subjects lied regarding
these items and the 20 designated truthful subjects answered questions about them
truthfully. Motivation also was systematically varied with certain subjects
competing for a $25 bonus based on their performance. The examiner's training and
experience was not mentioned. The polygraph test consisted of three series, each
series containing 32 questions about the Preemployment Data Sheet with 3 of the
potential lie questions in each series. Subjects were questioned between charts
about any questions where the reeponse to the relevant question was larger than the
response to the irrelevant questions. These questions were rephrased and inserted
at the beginning of the next series. Electrodermal activity, respiration and
cardiovascular activity were recorded, and examiner decision was a subjective
judgment of truth or deception.
All of the subjects were correctly classified as lying or telling the truth.
Accuracy was not as high for individual lies. Eighty-eight percent of the lies in
the lying motivated and 79% of the lies in the lying unmotivated were correctly
identified.
The Berland (1981) study discussed earlier, used a relevant-irrelevant
evaluation procedure in addition to two numerical scoring methods to detect false
biographical information. It should be noted that the polygraph. test was a
control-question test format, not relevant-Irrelevant. There were 30 deceptive and
26 truthful subjects with each guilty subject responding deceptively to one of five
questions. Contrary to Blum and Osterloh (1968), and Correa and Adams (1981), the
examiner was not allowed to question subjects between charts. Omitting
inconclusives, 86% of the deceptive and 76% of the truthful subjects were correctly
identified, as compared to 68% and 83% respectively, for the zone comparison
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
numerical evaluation. Individual question analysis revealed that, omitting
inconclusives, 69% of the deceptive and 92% of the truthful questions were
correctly indentified. Of the three methods of chart analysis,. the
relevant-irrelevant method was the only one able to correctly detect both truthful
and deceptive answers to individual questions. The fact that the
relevant-irrelevant method of chart analysis was the last one to be applied may
have contributed to this finding.
Heckel, Brokaw, Salzberg and Wiggins (1962) used the relevant-irrelevant
technique with subjects who believed they were accused of a real theft. The study
used normals, and delusional and non-delusional psychotics, but only the results
from normals will be discussed. Subjects were administered a series of
psychological tests during which the experimenter briefly left the room leaving his
wallet on the desk. When the experimenter returned he accused the subject of
taking $20 from the wallet, although nothing was actually missing. A professional
polygraph examiner administered two series of an 11-item General Question Test (a
type of relevant-irrelevant) with an interspaced stimulation test. Respiration,
heart rate, and electrodermal activity were recorded. Charts were scored by four
experts in lie detection" who did not know the nature of the study. Each examiner
made a clinical judgment of deception, no deception, or inconclusive. All of the
five normals were correctly classified as non-deceptive by the four judges.
Thus, two of the studies of the relevant-irrelevant technique report that they
were able to correctly classify 100% of the lying and 100% of the truth telling
subjects. Correct classifications for specific truthful and deceptive questions
were slightly lower. Heckel et al. (1962), using only innocent subjects, also
classified all subjects correctly. One factor that may be important is subject
interrogation between the polygraph charts. Barland (1981) did not allow
interrogation. Questioning subjects and restructuring questions, as is typical in
field usage, would be expected to increase polygraph accuracy.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Relevant-Irrelevant Test - Reliability
Heckel et. al (1962) appears to be the only laboratory study with information
on the reliability of the relevant-irrelevant technique, but they used only
innocent subjects. There was 100% agreement between the four examiners. Edel and
Jacoby (1975) report a field study on reliability of the relevant-irrelevant test
in which they found an agreement rate of 96% between the original examiner and
independent evaluators and 94% between the independent evaluators. A detailed
discussion of this research appears in an earlier section.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7 .
_Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Guilty Information Tests - Criterion Validity
Both the guilty knowledge test and the peak of tension tests use information
about a crime or situation that only the guilty person would know. The techniques
are described in the Chapter on Utility. The critical difference between the tests
is item placement. In the guilty knowledge test the guilty information is randomly
placed among the alternatives. In the peak of tension test the item is placed in a
specified sequence known by the suspect.
All of the studies investigating the guilty knowledge test have been laboratory
studies. Professional polygraph examiners have rarely been used and electrodermal
activity frequently is the only physiological response recorded. Several studies
have used the specific guilty knowledge test procedure of Lykken (1959). Several
multiple choice questions are administered. Each question contains one critical
alternative (information known by guilty but not by innocent people) and several
similar, but not relevant, alternatives. The -position of the relevant alternative.
to each question is randomly determined. Each question is scored, 2, 1, or 0
depending on the size of the response to the critical information in relation to
the size of the response to the other items. The first alternative is not scored
to reduce the effect of orienting. A question is scored 2 when the largest
response occurs to the critical alternative and a 1 when the second largest
response occurs to the critical alternative. All other possibilities are scored
0. Scores are summed across questions with those above a certain level indicating
deception and those below a certain level indicating truthfulness. There is no
inconclusive range.
Lykken (1959) reported on subjects guilty of two mock crimes, a murder and a
theft (n = 13), subjects guilty of only one crime, either the murder (n = 12) or
the theft (n = 12), and subjects innocent of both crimes (n = 12). Subjects were
told they would receive an electric shock if any of their responses indicated
deception but really received random shocks. The guilty knowledge test consisted
of two lists of six multiple choice questions, one for the murder and one for the
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
theft. The number of alternatives to each question averaged 4.67 in the murder,
list and 5 in the theft list. A score over 6 indicated deception and a score of 6
or less indicated innocence. Subjects made no response to the alternatives. These
procedures correctly classified 100% of the innocent, 100% of the subjects
committing only the theft, 92% of the subjects guilty of murder, and 69% of the
subjects involved in both crimes. Thus, the only errors involved calling the
guilty innocent. Lykken (1960) obtained 100% detection when he used the guilty
knowledge test to match subjects with their biographical information in an
experiment where subjects were taught how to defeat the test. Results will not be
discussed since the study used a special sample.
Davidson (1968) reported high correct classification of innocent and guilty
subjects in a mock crime situation with a guilty knowledge test consisting of 6
multiple-choice questions each with five alternatives. Apparently the questions
were presented only once. Forty-eight students were divided into groups of four
with each group containing three thieves and one innocent subject. Each group
participated in a theft of money from a different victim with the thieves in each
group personally planning the details of the crime. The situUtion for each group
, I
was structured so that one thief tried but failed, one thief had no opportunity to
attempt the crime, and one committed the crime. Only the thieves that completed
the crime were considered guilty so there were 12 guilty and 36 innocent subjects.
Motivation also was varied with the group of guilty subjects receiving variable
amounts of money for deceiving the examiner. Examiner decision was based on
electrodermal activity scored by the method devised by Lykken (1959) with a score
of 6 or less indicating innocence and greater than 6 indicating guilt. Subjects
made no response to the questions. Ninety-two percent of the guilty subjects and
100% of the innocent subjects were correctly classified.
Podlesny and Raskin (1978) used the guilty knowledge test with 10 innocent and
10 guilty subjects in a mock crime situation. The test consisted of one
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
presentation of 5 questions, each with 6 alternatives. The test was scored using
the Lykken (1959) procedure and electrodermal activity was the physiological
measure. Scores of 6 or greater indicated deception and less than 6 indicated
truthfulness. This study also investigated the effect of an inconclusive zone on
the guilty knowledge test with scores of 5 and 6 considered inconclusive. Subjects
were required to answer no to each alternative. One hundred percent of the
innocent and 80% of the guilty were correctly classified when no inconclusives zone
was used. When the inconclusive zone was used there was similar accuracy and 10%
were inconclusive.
A real cheating situation was used by Balloun and Holmes (1979) to investigate
the guilty knowledge test in the-detection of deception. Subjects were completing
a bogus intelligence information test along with two confederates as part of a
laboratory situation. Students were encouraged to do well on the test and told if
their performance was poor they would have to meet with a panel of psychologists.
One of the confederates urged the subjects to cheat by supplying a copy of the
answer sheet. Nearly half of the subjects actually cheated; thus there were 16
non-cheaters and 18 cheaters. The guqty knowledge test consisted of two series of
questions each series containing five relevant questions each question with five
alternatives. The relevant questions contained only the information the guilty
subjects could have known from the answer sheet provided by the confederate.
Subjects were required to verbally respond after each alternative by repeating the
alternative. The Lykken (1959) scoring system was used with scores greater than
5.5 indicating cheating and less than 5.5 indicating no cheating. When
classification was based on electrodermal activity, 61% of the cheaters and 87.5%
of the non-cheaters were correctly classified on series 1. On series 2, 17% of the
cheaters and 94% of the non-cheaters were correctly identified. Thus, there was a
significant reduction in the detection of cheaters from series 1 to series 2.
Heart rate and finger pulse volume were not effective in detecting cheating.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
NA, Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
The Bradley and Janisse (1981) study used a guilty knowledge test in addition
to the control-question procedures. The control-question and guilty knowledge
tests were counterbalanced. There were 96 innocent and 96 guilty subjects of a
mock crime. The guilty knowledge test consisted of one presentation of 4
questions, each question containing 5 alternatives. It is not clear whether
subjects were required to make a response. The scoring method of Lykken (1959) was
used. A score of 4 or more indicated deception and less than 4 indicated
innocence. Skin resistance was the best discriminator, with 89% of the innocent
subjects and 59% of the guilty subjects correctly classified. Pupil size and heart
rate change were not as accurate in determining guilt and innocence. The authors
noted that their version of the guilty knowledge test was a weak version since tt
did not contain many items of information and was only presented to the subjects
once.
Stern, Breen, Watanabe, and Perry (1981) had 26 innocent and 26 guilty subjects
play the parts of American Embassy hostages in Iran. Guilty subjects received
written instructions about their involvement in a murder plot. Innocent subjects
received instructions that had some of the same relevant words as the instructions
for guilty subjects, but involved no murder plot. Feedback also was manipulated
but these results will not be discussed. The guilty knowledge test consisted of
one presentation of 10 questions, each with 5 alternatives. Four of the ten
questions were control questions. Subjects were to make no response during test
presentations and responses were scored by the method of Lykken (1959). Ninety-two
percent of the guilty and 85% of the innocent subjects were correctly classified.
Several studies by Waid and associates (Waid, Orne, Cook, and Orne, 1978; 1981;
Waid and Orne, 1982; Waid, Orne and Wilson, 1979) used smiliar procedures to
investigate the effect of several variables on the guilty knowledge test. Only the
results for classification accuracy will be reported. Guilty subjects learned six
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
code words during a series of tasks. Depending on the study, innocent subjects
either completed a questionnaire or performed the same tasks as the guilty
subjects, but learned no code words. Guilty subjects were told that highly
intelligent mature individuals could deceive the examiner. Care was taken to
involve innocent subjects in the task. The guilty knowledge test consisted of a
variable number of presentations of a list of 24 words. There were six categories
of words with each category containing four words, one word being the critical code
word. A word was considered detected if the code word elicited a larger response
than the three alternatives. Subjects were required to make a response in all of
the studies. Waid et al. (1978) reported data from three experiments all using
electrodermal activity to detect.deception. Experiment 1 used 29 guilty subjects
and 11 innocent subjects with two presentations of the 24 word list. Electrodermal
activity was measured. Seventy-nine percent of the guilty and 73% of the innocent
were correctly classified. Experiment 3 used the same procedures except that 15
innocent and 15 guilty subjects were tested and each of the 6 categories of words
contained 6 words. Seventy-three percent of the guilty and 80% of the innocent
were correctly classified. The 18 guilty and 10 innocent subjects in Experiment 2
were tested in groups of 3 to 7 and subjects were told they would receive a shock
in the end of the test proportional in strength to the number of detected lies.
The list of 24 words was presented five times. Sixty-one percent of the guilty and
90% of the innocent subjects were detected. Waid at al (1981) correctly classified
100% of the innocent subjects and 82% of the guilty subjects when electrodermal
activity was the physiological measure. Cardiovascular activity and respiration
were not as good at classifying subjects. This study used two presentations of the
24-item list. There were 11 innocent and 11 guilty subjects.
Waid, Orne, and Wilson (1979) correctly classified 93% of the 15 innocent
subjects and 53% of the 15 guilty subjects using electrodermal activity. The list
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
of 24 words was repeated four times. The guilty knowledge test was administered by
a professional examiner after several types of polygraph tests were administered.
Waid and Orne (1980) reported data from two experiments, both recording
electrodermal activity. Experiment I used the same procedure as Experiment 2 in
Waid et al. (1978). There were 18 guilty and 10 innocent subjects. Sixty-seven
percent of the guilty and 90 percent of the innocent subjects were correctly
classified. The 15 guilty and 15 innocent subjects in Experiment 2 were
individually tested by a professional examiner and the guilty knowledge tests were
preceded by other types of polygraph tests. The 24-item list was presented four
times. Sixty percent of the guilty subjects and 100% of the innocent subjects were
correctly classified.
Giesen and Rollison (1980) reported on college students who were selected
because they reported showing high palmar sweating responses to stress. There were
20 guilty and 20 innocent subjects. ,Guilty subjects read and followed detailed
instructions involving a murder. Innocent subjects read and followed instructions
that contained some of the same information but involved no murder. The common
items were used as relevant items on the subsequent guilty knowledge test. The
test included 8 questions (6 relevants) each with five alternatives. Electrodermal
activity was recorded and the test was scored using the procedure of Lykken
(1959). Ninety-five percent of the guilty and 100% of the innocent subjects were
correctly classified.
Apparently, there are few studies using the peak of tension test with both
innocent and guilty people. The Waid, Orne, and Wilson (1979) study, already
discussed used a peak of tension test in addition to two control-question tests and
a guilty knowledge test. The peak of tension was always presented after the
control question tests. Forty-seven percent of the 15 guilty subjects and 80% of
the 15 innocent subjects were correctly classified. On both control-question
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
tests, 80% of the guilty and 80% of the innocent were correctly classified, whereas
with the guilty knowledge knowledge test 53% of the guilty and 93% of the innocent
were correctly classified. However, the tests were administered in the same order
to all subjects and this would be expected to influence the results.
There are a number of other studies using the guilty knowledge test and peak of
tension tests to detect information (Gustafson and Orne, 1963, 1965; Miyake, 1978;
Ben-Shakar, 1977; Ben-Shakhar and Lieblich, 1982; Kugelmass and Lieblich, 1968;
Lieblich, Kugelmass, and Ben-Shakhar, 1970; Ohkawa, 1963; Timm, 1982; Stern et al.
1981; Kronbergerova and Dufek, 1969; Yamaoka and Suzuki, 1980; Dufek, 1969;
Ohnishi, Matsuno, Arasuna, and Suzuki, 1976). These studies only use guilty
subjects or subjects possessing information to be detected. Since there is no
field analogue of this procedure, they will, not be discussed. However, they are
important in providing information on variables that influence detection of
deception.
The guilty knowledge tests are consistent in demonstrating higher accuracy with
innocent than guilty subjects. From 73% to 100% of the innocent subjects have been
correctly identified, whereas from 53% to 95% of the guilty have been detected. It
is not clear why there is such a range with guilty subjects. The number of
stimulus repetitions (Balloun and Holmes, 1979), whether subjects are required to
make a verbal response (Gustafson and Orne, 1965), and electrodermal lability (Waid
and Orne, 1980; Giesen and Rollison, 1980) may be important. Whether the use of
physiological measures in combination improves the accuracy should be investigated.
Guilty Information Tests - Reliability
Since guilty knowledge tests use an objective scoring system it would be
expected to be highly reliable. The scoring of the peak of tension test is more
subjective because there is judgment on the correspondence between maximum
physiological activity and a critical item of information.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
Chapter 2 Analysis of Research
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Chapter 2
Analysis of Research
The previous chapter reviewed specific laboratory and field studies of the
validity and reliability of lie detection. This chapter seeks to discuss some of
the problems and issues of lie detection and to draw conclusions from the research
literature.
Several observations can be made about the research. First, the research
literature is burgeoning. There has been more scientific research conducted on lie
detection in the last six years than in the previous 60 years. Second, the
polygraph technique is a robust technique. Although there are numerous variables
which affect its accuracy, it works better than chance in a wide range of testing
situations including criminal investigations, intelligence operations, security
screening, mock crimes in a laboratory environment, and even games in which a
person denies what number he selected. Indeed, it is difficult to find experiments
in which lie detection did not work better than chance. The technique obviously
works well with many cultures because many of the studies were conducted in Poland,
Israel, and Japan. Third, there is no such thing as a perfect study. Each study
is flawed; sometimes in its design, often in its execution, and sometimes in the
selection, analysis, and reporting of the data. Occasionally the shortcomings are
so serious as to disqualify the study from serious attention. Fortunately, the
various studies usually contribute to the mosaic of our understanding of the
polygraph technique and the factors which affect its accuracy. Our understanding
of these variables has expanded rapidly within the last five to ten years with the
accelerating pace of research.
Nonetheless, it is difficult to estimate the precise accuracy of the polygraph
technique in everyday applications because of the numer of variables involved.
There are a variety of polygraph applications, each with its peculiar problems
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
and issues: criminal investigation, intelligence operations, pre-employment
screening, and aperiodic security screening of current employees. Within each of
these applications are many subcategories. Polygraph accuracy is probably
different for each of these settings; however, some general comments about error
rates can be made.
There is some evidence to suggest that control question tests may be more
accurate in detecting the deception of the guilty person than in verifying the
truthfulness of the innocent person; that is, given equal numbers of guilty and
innocent persons, more inconclusive results may occur with the innocent persons,
and the number of false positives may outnumber the number of false negatives.
Studies of this type have primarily found either more false positives than false
negatives or essentially no difference in the false positive and false negative
error rate. Since the variables determining whether or not a study showed
differential error rates have not been determined, it is unknown whether a
preponderance of false positive errors is characteristic of field polygraph usage.
Additional evidence that the guilty may be easier to detect was reported by
Barland & Raskin (1975). They found that the absolute values of scores from guilty
subjects in a mock crime were more extreme than the scores from the innocent
subjects, indicating that the size of the reactions of the guilty subjects on the
relevant questions were larger than those of the innocent subjects on the control
questions. They also found that on numerically scored control question tests of
criminal suspects, the scores of the suspects called deceptive by the examiner were
more extreme than the scores of the suspects called truthful by the examiner
(Barland & Raskin, 1976). Not all of those decisions could be verified, however.
It is of interest to note that in two of the studies reporting a preponderance
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
of false positives (Horvath, 1977; Kleinmuntz and Szucko, 1982), the errors
occurred during the blind analysis of the polygraph charts. When verified cases
were selected for study, no known errors were found regarding the polygraph tests
as conducted by the original examiners.* This suggests the possibility that the
clinical judgment of the examiner who is able to observe the subject's demeanor
throughout the test may serve as a safeguard against false positive errors that may
occur when the polygraph charts are interpreted in isolation from nonpolygraphic
sources of data.
Peak of tension tests and especially the guilty knowledge test incorporate an
extremely effective safeguard against false positive errors--the innocent person
cannot determine which question is the critical question, and therefore cannot
consistently react to it regardless of how nervous or fearful he is. Thus, with
the peak of tension and guilty knowledge tests, virtually all errors are false
negative errors. However, these tests Can rarely be used in many applications of
the polygraph. More research is needed to clarify the relationship between false
positive and false negative errors with the control question test.
The base rates of truthfulness and deception within the specific population
being examined can have significant effects on the proportion of false positive and
false negative errors. The confidence to be placed in a given test outcome will
vary according to the proportion of false positive and false negative results
compared to the base rates of truthfulness and deception. As the base rate for
deception decreases there will be an increase in the number of false negative
errors and a decrease in the number of false positive errors. It should be
recognized, even if there is doubt about a particular polygraph test outcome,
* Personal Communication, Frank Horvath, 1983; Personal Communication,. Julien
Szucko, 1983.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
the test can nonetheless be useful. If, before the test, there is 1 chance in
1,000 that a person is deceptive, but afterwards there is a 50/50 chance the person
is deceptive, the test has been helpful. This is particularly important to
consider since the polygraph is supplemented by other investigative tools.
Another variable affecting the error rate is the type of issue to be resolved.
The polygraph is believed to be most accurate when the subject denies having
committed a specific physical act, such as a theft. Although there is, as yet, no
research on this matter, several theories predict that the accuracy is reduced when
the issue is vague or ambiguous, such as when the subject admits having shot the
victim, but claims he had only intended to frighten, not kill, him. The polygraph
issue is then one of intent, not the act. The extent to which the accuracy is
reduced depends upon a number of factors unique to each case, such as the nature of
the issue, the psychological makeup of the person being examined, the length of
time elapsed since the event occurred, and the skill and experience of the examiner
in formulating and defining the precise test questions.
The polygraph is believed to be most accurate when only one issue is to be
resolved. When more than one issue must be included in the test, as often happens
in many intelligence applications and in preemployment screening, the accuracy
would be expected to be reduced. In the absence of research on this issue, it is
not known whether the accuracy is reduced. The Correa & Adams (1981), Blum and
Osterloh (1968), and the Berland (1981) studies suggest that there may be little if
any decline in the accuracy of gross categorization of "entirely truthful" versus
"deceptive to one or more questions", and that the decline in accuracy occurs in
decisions
It is
decisions.
experience
regarding the precise question(s) to which the person was deceptive.
clear that the examiner's level of experience affects the accuracy of
It may be that once a certain threshold has been reached additional
results in only slight increases in accuracy, due to the inherent
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
robustness of the technique. However, the critical level of experience has not
been determined. Within the government, the closely supervised internship and the
quality control review process assists the intern in conducting his examinations
properly.
The approach to decision making (clinical versus numerical analysis of
polygraph charts) undoubtedly affects the accuracy of the decisions, but this
factor has not been adequately researched. It may be that the clinical approach
may be important in safegaurding against false positive errors while the numerical
evaluation approach minimizes false negative errors. If this proves to be true,
then a combination of both approaches should minimize both types of error. It
would be a useful experiment to employ the clinical approach followed by a quality
control review which incorporates numerical analysis of the charts to see if the
combination is more effective than either approach alone. There is some evidence
suggesting that when several polygraph examiners concur in the interpretation of
the polygraph charts, the accuracy of the group decisions is higher than any
individual opinion (Berland and Raskin, 1975). More research is needed to
determine the most effective approach to decision making.
The number of polygraph charts obtained from a given subject may be related to
the accuracy of the decisions. Generally, the more physiological information
available to the examiner the more accurate his decisions are (Barland & Raskin,
1976). Beijk (1980) reported increasing accuracy rates on a card test with
increasing numbers of polygraph charts. Studies reporting lower accuracy rates
occur when the reviewers saw only one chart or only used a portion of the
information available from the charts (Kleinmuntz & Szucko, 1982; Bradley &
Janisse, 1981). The Kleinmuntz & Szucko study reported data from an experiment
where the reviewers saw only one polygraph chart of a set, and had one of the
lowest accuracy rates reported in any study. Multiple charts are taken in all
federal applications of the polygraph.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
One of the complicating factors in trying to use the available research to
estimate the accuracy of the polygraph in real life stivations is the fact that
many laboratory studies are conducted by researchers who have not been trained in
the field uses of the polygraph. Among the many differences that lack of field
training generates is the relative lack of interaction between the researcher and
the subject. There are many things that can cause responses to appear to indicate
deception on the polygraph charts. These include extraneous noises, visual
distractions, emotional complexes associated with some word in a question, and even
random thoughts. The accuracy of the polygraph technique depends largely upon the
ability of the examiner to control the testing environment so that extraneous
sources of reactions are eliminated. "Lie detection" can properly be thought of as
an inference based upon a process of elimination. One important way in which the
field examiner determines whether a subject may be reacting to a question for some
reason other than deception is to talk to the subject between charts to find out
what was going through his mind on various questions. The questions can then be
restructured based on the information the subject provides. This procedure would
almost certainly reduce the number of innocent subjects classified as deceptive in
laboratory tests. For example, Correa and Adams (1981) questioned the subject
during the procedures and rephrased the questions based on the subject's replies.
That study reported 100% accuracy in classifying the innocent and guilty subjects.
In addition to polygraph accuracy, polygraph utility also is important.
Utility refers to the ability of a test or procedure to produce the results
desired. If 100 tests are administered, it is desired that 100 correct decisions
be obtained. Anthing which reduces the number of desired results reduces the
utility of the test.
As applied to the polygraph, utility has several aspects. In one sense it
refers to the ability of the polygraph examiner to obtain useful information from a
subject as a result of the test regardless of the actual accuracy of the polygraph
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
in terms of chart interpretation. During the interview conducted by the examiner
immediately prior to attaching the subject to the polygraph, many subjects confess
or reveal information they had not mentioned to previous investigators. The
pyschological "demand characteristics" of the polygraph situation are very powerful
and should not be underestimated. Thus, even if the polygraph were not at all
accurate, it would have a certain utility as long as people believed that it worked.
A second aspect of utility relates to the inconclusive rate. Even the most
accurate test has diminishing utility as the inconclusive rate increases.
Fingerprints, for example, have limited utility in investigations despite their
extremely high accuracy because only occasionally can identifiable prints be
recovered. In reference to the polygraph, it is necessary to distinguish between
its utility and its accuracy. Utility in this respect refers to the total number
of correct decisions out of the total number of cases; that is, the total number of
cases minus both the inconclusive results and the errors. Accuracy, on the other
hand, refers to the number of correct decisions out of the total number of
decisions, after the inconclusive results have been set aside.
It is not possible to estimate the accuracy of the polygraph in DoD, at
present, for the reasons discussed in this chapter. A great deal of government
polygraph usage involves screening, but there is little evidence on the accuracy of
the polygraph in screening situations. Two laboratory studies, Blum and Osterloh
(1968) and Correa and Adams (1981) used procedures that are similar to the
screening situation. Both correctly classified 100% of the innocent and 100% of
the guilty subjects. Although there are not enough data to make definite
statements about polygraph accuracy in screening, it can be said that there is no
evidence to indicate that large numbers of people are misclassified.
There is more evidence on potential accuracy in DoD criminal applications. The
Bersh (1969) field validation study was designed to estimate the accuracy of the
polygraph technique in DoD criminal investigations.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
This study found that in criminal investigations there was an agreement rate of
90% for guilty subjects and 94% for innocent subjects between the decisions of
military examiners and the criterion of unanimous decisions of JAG attorneys who
had reviewed the investigative dossiers minus the polygraph outcome. However,
since the examiners knew some of the case information which was later forwarded to
the panel of attorneys, these figures should probably be considered to be the upper
limit for the accuracy of the polygraph technique in military criminal
investigations in the early 1960's. Since that time the accuracy may have
increased somewhat due to the higher selection standards, more thorough training
(both the initial polygraph training and periodic seminars on advanced topics), and
the establishment of quality control offices which review every polygraph case.
For all of these reasons, it is likely that the quality and accuracy of federal
polygraph examiners is higher than the polygraph profession generally.
Several laboratory studies have utilized numerically scored control question
tests in mock crime situations with the physiological measures typical of the
field. Thus, these studies use procedures which are similar to DoD procedures
(Harland and Raskin, 1975; Rovner et al., 1978; Raskin and Hare, 1978; Podlesny and
Raskin, 1978; Hants, 1982; Gatchel et al., 1983; Hammond, 1980). Bradley and
Janisse (1981) used numerical scoring, but their scoring system differs from that
used at DoD and they classified subjects as truthful or deceptive from only one
physiological channel. Those experimental studies which used other test formats,
such as the guilty knowledge test, or in which subjects were instructed to engage
in countermeasures, were excluded from the data in order to make the results as
applicable as possible to use the polygraph within DoD. Only those studies using
numerical scoring are of interest because numerical scoring currently is used in
DoD criminal investigations. The decisions of the original examiner were used for
all studies except Rovner at al. (1978) because this study only reported results
for a blind evaluation.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
1. Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
These studies correctly classify from 757. to 100% of the guilty subjects and
from 57% to 100% of the innocent subjects. The mean correct classification rate
weighed for number of subjects in the study is 90% for guilty subjects and 80% for
innocent subjects.
There are numerous differences between mock crimes conducted in laboratories
and real crimes. Some of these differences, such as a lower level of emotional
involvement, probably made it harder to detect deception in the laboratory, whereas
other differences, such as a greater homogeneity of the research subjects in terms
of age, IQ, criminal history, etc., probably tend to increase the accuracy of
laboratory results by permitting the polygraph procedure to be optimized for the
subject population. Some of the.more recent research has employed subjects
recruited off the street or has used convicted criminals (e.g. Podlesny & Raskin,
1978; Raskin & Hare, 1978).
Since there is a limited amount of research data, it is important to consider
the experience of professional investigators and quality control personnel in
assessing field polygraph accuracy. The polygraph has been used as a lie detector
in hundreds of thousands of criminal investigations from the moment of its
? inception in 1921. Police detectives and other professional investigators have
worked closely with polygraph examiners, and have continued their investigations of
countless cases after learning of the polygraph results. By their nature and
training, investigators tend to be skeptical pragmatists. Yet, they rely upon the
polygraph.
One of the functions of the various federal quality control offices is to
follow up the cases, maintain records, and investigate causes of polygraph errors.
Although it is impossible for all polygraph errors to be detected, the number of
known errors in federally-administered polygraph examinations of criminal suspects
is less than 1% of all polygraph examinations. One reason why the actual number of
errors is unknown is that military prosecutors give a great deal of weight to
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
polygraph results when deciding whether to proceed with court martial proceedings.
It is extremely rare for a person who has been found truthful by a federal
polygraph examiner to be prosecuted. Consequently, false negative errors are not
likely to be found. On the other hand, instances in which the subject is found
deceptive are almost always investigated further and are usually prosecuted. Thus,
the probability of false positive errors being detected by additional investigation
is much greater. It is possible that the actual number of combined errors may be
four or five times the number of known errors. This type of analysis would
therefore suggest that the polygraph in_criminal investigations is almost certainly
more than 90% accurate, and that it probably is about 95% accurate.
Thus, the Bersh study, experience of investigators and quality control
personnel, and mock crime laboratory studies give different estimates of the
accuracy of control question tests in criminal investigations, ranging from about
80% to 95%. Until additional research is conducted we do not feel comfortable in
specifying a precise figure for polygraph accuracy in DoD criminal investigations.
Additional research also is needed to draw precise conclusions about polygraph
accuracy for other DoD applications where different techniques are used, such as
personnel screening and intelligence investigations. It is noted, however, that
there are no data suggesting that the various polygraph techniques and applications
at DoD have high false positive or high false negative error rates.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Chapter 3 Utilization of the Polygraph
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
THE POLYGRAPH IN THE FEDERAL GOVERNMENT
FEDERAL AGENCIES
The polygraph is currently used by the Department of Defense, United States
Secret Service, Federal Bureau of Investigation, Postal Inspection Service,
Alcohol, Tobacco and Firearms Administration, Drug Enforcement Administration,
Central Intelligence Agency, United States Marshals, United States Customs Service,
and the Department of Labor (Anti-Racketeering). These agencies often provide
polygraph service to those federal law enforcement agencies and intelligence
agencies and activities which do not have polygraph examiners. The Department of
Defense elements which currently have polygraph programs are the United States Army
Criminal Investigation Command, the United States Army Intelligence and Security
Command, the Naval Investigative Service, the Air Force Office of Special
Investigations, the U.S. Marine Corps Criminal Investigation Division, and the
National Security Agency.
These fifteen agencies are represented on the Federal Interagency Polygraph
Committee which meets quarterly, or more often if necessary. Although informal,
the committee has achieved the adoption of minimum standards for the selection and
training of federal examiners. The Committee has also established and operated an
advanced polygraph seminar. Each year, the FBI hosts a one-week seminar attended
by seventy or more federal examiners at its training facility at Quantico,
Virginia. The program of instruction is planned and coordinated by a different
federal agency each year, although in some years two agencies combine to put on the
instruction.
Polygraph Examiners in the Department of Defense
All polygraph examiners in the Department of Defense must meet the minimum
requirements set forth in DOD regulation 5210.48 plus those established in each
service or agency. DoD examiners are college graduates, experienced investigators,
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
U. S. citizens, at least 25 years old, and specifically selected for their maturity
and judgment. All have been subjected to a thorough screening examination with a
polygraph, and have been the subject of an extensive background investigation. The
basic polygraph training of military examiners is the twelve-week course conducted
by the Army at the Military Police school, Fort McClellan, Alabama. All examiners
then serve a closely supervised internship of six to twelve months.
Number of Examiners
As of September 1983, the Department of Defense had 153 certified examiners.
That number does not include interns or those in basic training.
DoD CERTIFIED EXAMINERS
,Army Army Marine
CID INSCOM Navy Air Force kia..1- NSA Total
1980
39
9
11
20
8
13
100
1981
42
9
12
20
6
30
119
1982
44
12
14
20
6
26
122
'
1.983
50
15
15
40
'6
27
153
Annual Polygraph Utilization
ANNUAL POLYGRAPH UTILIZATION (Number of Examinations)
Marine
Army Army Navy Air Force Corps
CID INSCOM NIS OSI CID NSA Total
1980
4005
214
1317
1415
258
5676
12885
1981
3690
281
1185
1418
245
7418
14237
1982
3686
390
1337
2026
263
9672
17374
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
CRIMINAL INVESTIGATION
History
The Federal Bureau of Investigation and its predecessor, the Bureau of
Investigation, made use of the polygraph technique occasionally from 1917 into the
1930's. In the mid-1930's Leonarde Keeler, who was then at the Northwestern
University Crime Laboratory, trained an FBI agent in polygraph technique. A few
years later he trained an agent from the United States Secret Service. The Army
began to have agents trained as examiners in World War II, and when Leonarde Keeler
established the first polygraph school in 1948 the federal government began to
expand the use of the polygraph for criminal investigations. Today fifteen federal
agencies or departments have polygraph programs.
Polygraph As An Aid To Investigation
In criminal cases, the polygraph is not used in the Federal Government
suirstitute for investigation. Field investigation is usually conducted as
as a
far as
circumstances permit before the polygraph plays a role. This requirement for prior
investigation plus a prior and separate interview of all prospective polygraph
subjects is mandatory in the Department of Defense, and may be waived by DoD
officials only in the most unusual circumstances. Despite these restrictions, the
polygraph is widely used in criminal investigations. Defense counsel in criminal
cases frequently ask for exculpatory examinations.
The Army, which is the greatest user of the polygraph for law enforcement
purposes in the Federal Government, has utilized the polygraph in 17% to 19% of its
investigation of felony crimes during the last year and a half. Further, the Army'
has used the polygraph in 95% of its criminal investigations pertaining to crimes
for which the maximum penalty is 15 years or greater. The effectiveness of this
effort is illustrated by comparing the percentage of felony cases solved by the
Army with the national rate. The Army solved 64.7% of their felonies while the
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
national average for the same period was only 19.5%. The use of the polygraph was
one of the significant reasons for this extraordinary achievement.
The following tables depict the extent of use of polygraph examinations in
criminal investigations in DoD, the pretest and post test confession rates, and the
administration of exculpatory polygraph examinations which are requested by the
suspect or his attorney:
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
CRIMINAL INVESTIGATION CASES
Army CID
Navy NIS
Air Force OSI
US Marine Corps
Totals
1980
3990
1209
1336
255
6790
1981
3677
1049
1304
240
6270
1982
3665
1210
1745
261
6881
PRETEST ADMISSIONS OR CONFESSIONS
Army
Army
Navy
Air Force
Marine Corps
CID
INSCOM
NIS
OSI
CID
1980
12.5%
21.5%
11.5%
10.0%
1.9%
1981
10.3%
16.0%
8.9%
13.3%
3.7%
1982
10.7%
13.3%
11.0%
12.0%
3.4%
POST TEST ADMISSIONS OR CONFESSIONS OF DECEPTIVE SUBJECTS'
Army
Army
Navy
Air Force
Marine Corps
CID
INSCOM
NIS
OSI
CID
1980
46.4%
19.4%
34.4%
35.6%
50.0%
1981
44.7%
27.7%
36.5%
33.3%
52.4%
1982
48.7%
51.3%
30.1%
31.2%
36.7%
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
EXCULPATORY EXAMINATIONS
(Requested by the suspect or his attorney)
Percentage of total cases in parenthesis
Army Navy
CID NIS
Air Force
OSI
Marine Corps
CID
1980
328
(8.2%)
239
(18.1%)
506
(35.6%)
39
(15.1%)
1981
349
(9.5%)
164
(13.8%)
466
(32.9%)
24
( 9.8%)
1982
324
(8.8%)
209
(15.6%)
542
(26.8%)
31
(11.8%)
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
Criminal Investigation Case Examples
In 1980, an Army captain was killed when his parachute failed to open.
Investigation revealed that the static line of the officer's parachute had been
intentionally cut. As this crime occurred in an airborne division, the possibility
existed that another parachute had been cut and that additional military members
would be fatally injured during parachute jumps. As 160 parachute riggers had an
opportunity to commit this crime they were all considered suspects. A decision was
made to use the polygraph swiftly to determine the individual responsible for
sabotaging the parachute. Five days after the death of the officer, use of the
polygraph resulted in the identification of the person who sabotaged the parachute,
who then confessed.
In 1982, a serviceman had been scheduled for trial for the rape and sodomy of a
female military member. Based on a polygraph administered tothe alleged
assailant, who requested this test in exculpation, it was determined that he had
not committed any sexual crimes with or upon the complainant. A subsequent
polygraph examination of the female complainant resulted in a confession that she
had made a false complaint because she was angry at some members of her former
military unit. In this instance, the polygraph prevented an innocent man from
going to trial for the serious crimes of forcible rape and forcible sodomy.
In June 1982, a female military person reported in a sworn statement that a
male military person had assaulted her with intent to commit rape. A criminal
investigation of the alleged incident failed to confirm or deny her complaint.
Subsequently, the alleged incident became a security clearance matter and was
reopened. The alleged perpetrator admitted "playing around" with the complainant,
but denied any aspect of assault or attempted rape and he volunteered to undergo a
polygraph examination. As a result of the examination, the male military person
confessed to all the charges made by the complainant.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A ci Man employee operating a government owned service (gas) station aboard a
military Installation was suspected of falsely reporting a $1,800.00 robbery of the
service :ation. HE was administered a polygraph examination which indicated
deceptio The employee then admitted stealing the $1,800.00 himself and
identifi I a local military member as an accomplice.
An i lividual was tried and sentenced to life in prison for the murder of a
fellow s ldier. This individual asked the Army to give him a polygraph examination
for excu )ation. The results revealed that he had not committed the murder. The
investig :ion was reopened and the actual perpetrator of this crime was apprehended
and the inocent solder was released from prison. This matter would not have been
resolved ;ithout the use of the polygraph.
The )dy of a nine-year-old girl was found in a trash dumpster on a military
installa ion. An autopsy revealed the child had been sexually molested and the
cause of ieath was asphyxia. Investigation revealed that an individual reported he
had been fobbed in the vicinity of the dumpster on the night in question. Further
investig :ion revealed the complainant of the robbery had been in a bowling alley,
which we the last place the girl was seen alive. The subject was offered a
polygrap examination and he agreed. The examination results indicated deception
on the i mes of kidnapping and killing the girl. During the subsequent
interrog :ion by the polygraph examiner, the subject made a full confession.
A U. 3. military member was suspected of setting a fire aboard a ship that
resulted Ln excess of one million dollars damage; he denied the charges. The
military member was administered a polygraph examination which indicated deception,
where up i he confessed to intentionally setting the fire.
An i lividual charged with Possession and Sale of Controlled Substance
maintain i his innocence during the investigation and requested an exculpatory
polygrap examination. The examination was administered by an Army examiner and
the indi Ldual was found to be truthful when he denied criminal participation in
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
the possession and sale of the controlled substance. The polygraph evidence was
not considered during the pre-court martial hearings nor was it allowed to be '
introduced during the court martial proceedings. The individual was convicted and
sentenced to serve time in prison. Following the conviction, the examiner
continued the investigation and identified a civilian subject who confessed to the
examiner that he committed the crime for which the military member was sent to
prison. The examiner then coordinated efforts to obtain the release of the
innocent military member from prison.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Criminal Conduct Revealed during Screening Examinations
A detailed analysis was made of the admissions made during the testing of
20,511 applicants for clearance and access at NSA from 1 October 1975 through 28
February 1979. There were 695 persons, 3.4% of the total, who admitted to the
commission of a felony. In almost all of these cases the perpetrator had gone
undetected. Among the few who had been caught, there were five or six who had been
convicted, but who had falsified their forms to conceal the arrest, conviction, and
time served. The-admissions included murder, armed robbery, forcible rape,
burglary, arson, embezzlement, hit and run driving with personal injury, thefts of
expensive items or large amounts, smuggling and wholesale selling of illegal
drugs.
In addition, 2,489 persons (12.1%) admitted to misdemeanors. These admissions
included petty theft, simple assault, tax evasion, falsification of travel
vouchers, malicious damage, indecent.exposure, prostitution, selling illegal drugs
and controlled substances for profit, deliberately writing bad checks, and other
forms of fraud. Some of these cases may have been felonies. When there' was doubt,
they were statistically classified as misdemeanors.
admissions were disqualifying. The purchase, use or
was not listed in this category.
There were 8,383 (40.1%) who admitted to the use
one occasion. There were 2,228 (10.9%) who admitted
barbiturates, hallucinogens and similar substances.
admitted to the use of heroin, cocaine, or opium.
The statistical rates for the admissions of crimes and drugs have changed very
little since that survey.
Examples of Crimes Admitted during Screening Examinations
The following admissions were made to NSA examiners during screening
examinations. All are from FY 1982 cases:
Not all of the misdemeanor
possession of illegal drugs
of an illegal drug on at least
using amphetamines,
There were 704 (3.4%) who
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
? .? Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
An applicant for an engineering position at NSA, employed as an engineer by
another government agency, admitted that his engineering degree was phony (he
bought it through mail order from London for $100). He also admitted that he shot
and wounded his second wife (his prison term was not on his SPH) and his present
wife is missing under unusual circumstances that he would not explain.
An applicant admitted to setting fire to the trailer that his ex-wife and child
lived in. He had been questioned by the Michigan State Police as a suspect
(attempted murder/arson) but denied it, and refused their offer of a polygraph test.
An applicant said he killed a young girl while in combat in Vietnam. He
recognized her as a young girl when he shot her the first time, then for no reason
he could give, he shot her several more times at close range. He called it
murder. He also admitted stabbing a stranger in the face with a knife in an
argument over some beer.
.An applicant admitted to the forcible rape of his ten year old niece, which was
never reported to the police.
An applicant for a position in which be would carry a firearm, admitted he had
been charged with attempted murder but not tried for lack of evidence. He admitted
firing his shotgun at six people, and hitting all of them.
An applicant admitted to firing a rifle into his estranged wife's home in an
attempt to murder her. He fled Connecticut which had a current warrant outstanding
for his arrest.
?
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
'
POLYGRAPH and the TRADITIONAL METHODS
The polygraph examination is part of the processing for clearance and access in
NSA and CIA. Both are necessary for complete coverage. Although many agencies
rely entirely on the background investigatiOn, neither NSA nor CIA have considered
authorizing SCI access without a polygraph in more than 30 years. During the
Korean emergency in 1951, when resources wee inadequate to conduct background
investigations in a timely manner, NSA granted temporary access based on polygraph
results and a National Agency Check, with full access on receipt of the background
investigation.
1953 - Study of Polygraph/Background
In 1953, the National Security Agency conducted a study of the effectiveness of
their emergency procedure of using the polYgraph and National Agency Check.* That
study considered 3,926 individuals who were processed for clearance and access by
NSA (and its predecessor, AFSA) between 1 !November 1951 through December 1952.
Of the 3,926 individuals concerned, 3,878 volunteered for a polygraph
interview, and 48 refused to do so. These 3,926 individuals are further broken
down into
1,024 who
employed.
two categories: 2,902 who werel, interviewed prior to employment, and
were recruited in the field and! were interviewed after they were
Of those who refused to take a polygraph test, 41 were applicants and 7
were employees. There were
reviewed during this study.
2,772 complete or partial background investigations
The disparity between this figure and the total
persons in the study is because, generally, when applicants revealed information
during the polygraph examination which disqualified them from further
consideration, no background investigation was initiated. Another contributing
factor is that in the case of 64 employees and 209 applicants the reports of the
*NSA does not now authorize clearances On a polygraph and an NAC because PL 88-290
requires a full field investigation prior to clearance.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
complete background investigations had not yet been received by the agency at the
time of the study. Of the 1,017 employees who were interviewed a total of 455 of
the tests indicated no deception and the employees gave no information requiring
evaluation, while 569 indicated deception, gave information requiring evaluation,
or they were physically or emotionally unfit for testing.
Of the 569 employees who required further security evaluation after the
polygraph interview, 83 were terminated or permitted to resign in lieu of
termination. In 65 of these cases the information which formed the basis for this
personnel action was furnished solely as a result of the polygraph interview.
Eight persons were terminated solely on the basis of information. furnished from the
background investigation, and 10.persons were terminated on a composite of the
Information from the two investigative methods. Of the 8 people who were
terminated solely as a result of their background investigations, 3 made no
admissions and appeared truthful on the polygraph test, while the other 5 made
minor admissions or were unfit for testing. Altogether, 8.10% of all employees in
the study were terminated, 78.3% on the basis of information from the polygraph
interview, 9.6% on the basis of information from the background investigation, and
12.1% from a composite of the two. In addition to the 83 persons who were
terminated, 248 persons covered by this study resigned voluntarily, for reasons
unrelated to security, leaving 693, or 67.67% of this group remaining with the
Agency.
The clearance of 76 employees was delayed pending receipt of background
investigations. Six of these persons were later terminated on the basis of
information provided by the background investigation, while in the other 70 cases
no information was developed which was a bar to clearance. These persons had been
held for the completion of a background investigation because they were unfit for
testing, because they refused to volunteer for a polygraph test, or because the
iniformation resulting from the polygraph interview was not sufficient to enable
Declassified and Approved For Release 2013/04/16 :_CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
analysts to reach a security determination.
In the case of 2,902 applicants for employment, 41 persons refused to volunteer
for a polygraph interview, leaving 2,861 who were examined with the polygraph. The
tests of 1,251 applicants indicated no deception and the applicants gave no
information requiring security evaluation. The remainder of the group of
applicants, 1,651 persons (56.9%), required further security determination. Of
these, 696 (24.3%) were disapproved for clearance on the basis of their admissions
during polygraph examination. The 610 (21.3%) who gave minor admissions were
evaluated as satsifactory for clearance. Of those 2,902 who applied for
employment, 1,765 (61%) entered on duty. Some of the remaining 345 applicants were
denied employment for reasons not related to security, others declined proffered
employment.
There were 19 persons who were approved for hire on the basis of the
satisfactory evaluation of their polygraph interview and a National Agency Check
who were later terminated. Eleven of these persons had made no admissions and
appeared truthful on their polygraph interview. In 3 cases the individual made
minor admissions which were evaluated satisfactorily, and 2 cases involved those
who had refused to volunteer for a polygrah test, but were nonetheless hired. A
composite of the information from the background investigation and the polygraph
test formed the basis for the termination of 2 persons, and the remaining
individual was terminated on the basis of information developed during the
polygraph interview of another applicant. The 17 persons on whom derogatory
information was developed despite clear polygraph test results and no admissions or
only minor admissions represent a false negative error rate of 0.6% in the testing
of 2,902 applicants.
The use of the polygraph interview resulted in a considerable monetary saving
in the security clearance of employees. In the group under study there were 696
applicants who were disapproved for hire and 65 employees who were terminated
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
immediately on the basis of information derived from the polygraph interview.
Calculated on the basis of a nine-month clearance period at the GS-4 level, this
group of individuals would have cost the government $1,814,322 in salary alone
while awaiting the completion of a background investigation. This figure is low,
since the average salary of individuals awaiting clearances in the agency was above
the GS-4 level and the average background time was more than nine months.
The study concluded that the background investigation is at its best in areas
which are a matter of record, such as employment, education, naturalization,
criminal history, etc., although even in these areas more pertinent information was
developed through polygraph interviews. On the other hand, in personal matters not
of record, the polygraph interview is far more reliable as a producer of
information for security determination. In the polygraph interview this
information is obtained directly from the subject. It does not, as with the
background investigation, come from informants whose motiviation and reliability
are often unknown, and who, in a few cases where derogatory information is
concerned, are vindictive.
The authors of the 1953 study observed that the primary justification for the
use of the polygraph interview in personnel screening is not its financial savings
to the government, but from the fact that a combination of a complete background
investigation and polygraph interview provides more informaton for.security
determination than can be obtained from any other methods yet devised. Neither the
polygraph interview nor the background investigation may be considered complete in
itself. Each supplements the other; each improves the validity of security
determinations.
1983 - Study of Polygraph/Background
In 1983, the National Security Agency conducted a pilot study comparing the
information obtained from polygraph interviews where the examiner did not have any
information from a background investigation, with background investigations where
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4 -
none of the polygraph developed information reached the investigators. These "dual
track" cases are unusual in that the investigation started well befcre the
polygraph examination was conducted. Even so, some of the dual track cases had to
be discarded from the study because information developed during the polygraph
intervew was totally disqualifying and the Defense Investigative Service was
advised to stop the investigation before it was complete. Other cases were deleted
from the study because information from the polygraph interview was given to DIS as
leads for further investigation, because DIS had not yet closed their case.
The subjects whose files were reviewed in this study had already been evaluated
as having acceptable information on their Personnel Security Questionnaires, and
had received a polygraph interview sometime between 1 October 1982 and 31 January
1983. The polygraph interview and Special Background Investigation (SBI) were
conducted independently, without any exchange of information. From a total of 248
randomly selected dual tracks cases, 194 cases were selected as meeting the
criterion of complete independence, required by this study.
The results of the study revealed that 78% of all information used for
evaluation was obtained from polygraph reports. The remaining 22% of information
used for evaluation came from the Special Background Investigations.
In 113 cases (58%), the polygraph report was of particular value to the
clearance adjudicator because it was the only investigative method to develop
information for evaluation. By comparison, the SBI developed the only information
in two cases (1%).
In addition, there were 22 cases (11%) in which the polygraph interview and the
background investigation produced dissimilar but useful information, and 14 cases
(7%) in which they developed similar information. There were 43 cases (22%) in
which neither source produced useful information. Although polygraph interviews
produced far more information than the background investigation, the contributions
of the latter were vital, and often on matters not developed or even covered in
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
the polygraph interview. For example, credit Ls discussed in a full scope
polygraph screening examination, but it is not the subject of a specific polygraph
question because it is not practical to develop a dichotomous issue. Similarly,
character deficiencies, alcoholism, a poor work record, and many other topics are
often better learned from the observation of others, or from official records. In
addition, the background investigation provides positive "whole person"
information. The obvious conclusion was that both sources of informtion are
productive, each from different sources, and both vital to the process of
evaluation for clearance and access.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
r
1974 - 1979 Polygraph Results from Cleared Military Personnel
A study was made of admissions made by applicants who were currently or
recently in the military service who were cleared for SCI access and were in a
cryptologic service. This table includes the 2,426 who applied in the fiscal years
1974 through 1979.
Admissions During Polygraph Interviews, FY 74 - 79
from 2,426 Military Applicants
Topics Number of Persons making
Percentage
Admissions
Espionage
13
0.6%
Communist, Fascist
or Terrorist Activity
25
1.0%
Divulgence of Classified
Information: -
To family
135
5.6%
To others
132
5.4%
Secret Foreign Contacts
1
.04%
Soviet Bloc Travel
70
2.9%
Soviet Bloc Contacts
60
2.5%
Drugs
Marijuana
1654
68.2%
Marijuana over 100
times
159
6.6%
Uppers, Downer,
etc.
243
10.0%
Cocaine and Heroin
65
2.7%
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A -
? Crimes
Misdemeanors
322
13.3%
Felonies
68
3.3%
Alcohol Problems
Past
25
1.0%
Present
4
0.2%
Deliberate Falsification
of the Security Form
152
6.3%
Serious Credit Problems
57
2.3%
Psychological Problems
Treatment
107
4.4%
Suicide Attempted
14
0.5%
In regard to the espionage admissions, most knew of or had strong reason to
suspect others of engaging in espionage against the United States but had not .
reported it. A few were personally involved in espionage. See examples in this
chapter.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
1981 Polygraph Results from Cleared Military Personnel
A study was made of the 318 applicant cases for FY 81 in which the applicant
was currently or recently in the military service in a cryptologic assignment with
SCI access. In every case they had been the subject of a Special Background
Investigation at some time in their career, and some had update investigations.
All these investigative results had been evaluated as satisfactory for clearance
and access. In some cases, the background investigation was in the security file
available to the examiner. In the majority of cases it was not.
The table below lists only that information obtained during the polygraph
examination which was not in the security file. The tables exclude all information
obtained during background investigation, special investigations, and security
interviews.
Admissions During Polygraph Interviews, FY 81
Topic
from 318 Military Applicants
Percentage
Number of Persons making
Admissions
.Espion4ge
0
0
Communist, Fascist or
Terrorist Activity
1
0.3%
Divulgence of Classified
Information:
To family
13
4.1%
To others
10
3.1%
Secret Foreign Contacts
0
0
Soviet Bloc Travel
15
4.7%
Soviet Bloc Contacts
6
1.9%
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
Drugs
Marijuana
141
44.3%
Marijuana over 100
times
16
5.0%
Uppers, Downers,
etc.
31
9.7%
Cocaine & Heroin
12
3.8%
Crimes
Misdemeanors
38
11.9%
Felonies
12
3.8%
Alcohol Problems
Past
7
2.2%
Present
1
0.3%
Deliberate Falsification
of Security Form
7
2.2%
Serious Credit Problems
14
4.4%
Psychological Problems
Treatment
14
4.4%
Suicide Attempted
5
1.6%
Not all of the admissions listed above were disqualifying. The table does not
disclose amounts or recency.
Note that the first major topics are similar to the topics usually included
among the questions in counterintelligence polygraph examinations used for special
access determinations.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved ForRelease2013/04/16 : CIA-RDP07-01280R000400400001-7
1983 - Study of Interviews/Polygraph
One of the additional techniques sometimes used in processing people for
classified access, particularly SCI, is a thorough security interview. For several
years, NSA has conducted security interviews of incoming military personnel
assigned to NSA. The interviewers are experienced Special Agents of NSA who are
specifically trained for this work. The scope of their interview is very similar
to that of the polygraph examinations given by NSA to civilian applicants for
clearance and access.
A 1981 study at NSA reviewed the information obtained on 13 topics that are
covered in both the security interview and the polygraph examination. A random
sample of 100 security files containing military security interviews was taken from
the list of those military persons who were interviewed between March 1981 and July
1983 on arrival at NSA. Another random sample of 100 cases for that period was
taken from the security files in which there was a polygraph examination of a
military assignee to NSA who took the examination as part of his processing for
civilian employment at NSA.
In 49% of the cases, the interviewers obtained admissions from already cleared
incoming military personnel which required security adjudication. In all, there
were 87 specific admissions. The subject matter included ten of the thirteen
topics covered in both the interview and the polygraph examination.
The polygraph examiners obtained admissions requiring security adjudication in
82% of the examinations of the already cleared military personnel who were seeking
civilian employment. There were a total of 111 specific admissions. The
admissions covered nine of the mutual topics.
A second analysis compared the security interview conducted on entry and the
subsequent polygraph examination of those 100 military persons who applied for a
civilian position at NSA. The security interview predated the polygraph
examination by a range of a few months to three years. The polygraph examiner, of
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A ?
course, had access to the interview report prior to the examination. What is
interesting is that in these 100 randomly selected cases the examiner got
information which predated the security interviews, on nine of the thirteen mutual
topics. For example, in 31 cases (31%), the examiner obtained information on drug
involvement that predated the security interview.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
USE OF POLYGRAPH IN ARMY PERSONNEL SECURITY CASES
The U.S. Army has no general personnel screening program in which polygraph is
utilized; however, the polygraph is utilized on a selective case-by-case basis.
When the U.S. Army Central Clearance Facility (adjudications) receives .a completed
Background Investigation from the Defense Investigative Service in which there are
areas of adverse information which remain unresolved, a polygraph examination may
be requested to aid in the resolution. The polygraph examination is offered to the
subject on a voluntary basis to aid in clarifying the areas not fully clarified by
the investigation.
From February 1982 to September 1983, 180 such cases were referred for
polygraph examination. Of the 180 cases, 67 subjects declined to take the
examination. The following statistics summarize the results of those 113 cases in
which the subject agreed to undergo a polygraph examination:
Overall Results
Nondeceptive Deceptive Inconclusive No Opinion
28(25%) 81(72%) 2(1.5%) 2(1.5%)
Where there was Deception Indicated (DI), relevant admissions were obtained in
71 cases (or 88%). The following is a breakdown by area of the DI admissions:
Loyalty Drugs Sex Crimes Other Criminal Acts
16(20%) 40(49%) 11(14%) 14(17%)
Loyalty cases involve unreported contacts with foreign intelligence, unreported
travel to denied areas, unreported contacts with persons within denied areas,
association wtih communist/subversive organizations, and hostage situations.
Drug cases involve the use, possession, distribution, purchase or sale of
illegal drugs.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Sex cases involved homosexuality, child molestation, and other aberrant sexual
practices.
Other criminal cases involved allegations of criminal activity other than drug
involvement or sex crimes, e.g. theft, embezzlement, fraud, etc.
Clearance Actions in Polygraph Cases
Deceptive
Nondeceptive
Clearance Granted:
15
(24%)
24
(89%)
Clearance Denied:
36
(76%)
3
(11%)
Action was still pending in 19 DI cases and 2 nondeceptive (MDI) cases. Some
of the DI cases were cleared up by minor confessions. The 3 MDI cases were denied
access based on information developed during the polygraph examination.
The above figures indicate that adjudications were not based solely on
polygraph results, but on the overall investigative file.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
FrELLIGENCE, COUNTERINTELLIGENCE AND SECURITY
History
The instrumental detection of deception was first used in an investigation for
the Army when a Sect t Code Book was stolen from the Surgeon General's safe in
1917. The thief was
confronted. Followe
the act of passing t
trained a group of r
counterintelligence
use (Marston 1938).
Beginning in the
facility for screeni
company, the program
under contract were
World War II at Caw
274 German prisoners
leadership positions
experienced examiner
communist sympathy,
Gestapo, SS, or SA,
senior police positi
There were 8 inconcl
examiners obtained s
SA by two and SS by
organizer.
1978).
detected among the 70 suspects tested, but he was not
from Washington, D.C. to New York City, he was apprehended in
e code book to a German agent. During World War I the Army
ychologists at Camp Greenleaf in lie detection techniques for
urposes, but the Armistice was signed before they were put to
1940's, the polygraph was used at the Oak Ridge atomic
g of employees. Operated entirely under contract by a private
continued until 1951 when the security responsibilfties then
aken over by the government (Trovillo 1951). At the end of
Wetherill, New Jersey, the Army used the polygraph to screen
3f war who had been screened and selected for police
in the post?war German Government. A team of seven -
using RI Technique asked questions about Nazi party sympathy,
lens to engage in sabotage or subversion, activity for the
ad serious crmes. The results cleared 156 Germans (57%) for
as.
sive cases (3%) and 110 (40%) were not recommended. The
missions to Nazi Party membership by 24 candidates, and to
ae. Three were communists, of which one was a party
Among tt Nazis was the Party Treasurer from 1.933 to 1938 (Linehan
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
In 1947 the newly formed Central Intelligence Agency began using the polygraph
in support of operations and for specific investigations. They also began to use
the polygraph in screening. By 1950, CIA was screening all applicants.
In May 1951, the Armed Forces Security Agency, predecessor of the National
Security Agency, began using the polygraph to expedite the clearance processing of
more than a thousand employees who had been hired but could not be cleared until
the background investigation had been completed. Because of the Korean War, those
investigations were taking from nine to eighteen months to complete. Accordingly,
AFSA gave top secret access on an interim basis to employees based on a
satisfactory polygraph examination and a National Agency check. Soon thereafter,
AFSA-NSA began screening local applicants before they were hired and by 1953 was
giving polygraph examinations to all applicants as a condition of employment.
During the years 1951 to 1983 a number of other special activities in the
Department of Defense have used the polygraph for clearance and access ?
determinations, and for operational purposes. Some of these polygraph operations
were for specific projects of a limited duration, others were for long terms.
Acting on recommendations of a DoD Select Panel on Personnel Security, Frank C.
Carlucci, Deputy Secretary of Defense, issued a Memorandum on 6 August 1982 which
established an aperiodic counterintelligence polygraph program in the Department of
Defense for those person who have SCI access. Mr. Carlucci also restored the
Reinvestigation Program in DoD. The Carlucci polygraph program was implemented
only in the National Security Agency. Other elements of DOD needed a revision in
the DOD polygraph regulation before they could begin. The polygraph regulation was
under revision when Congress expressed an interest in the proposed expansion of the
polygraph in the Department of Defense. On March 11, 1983, the President issued
National Security Decision Directive Number 84 "Safeguarding National Security
Information," which was separate from the DoD action. Among the many security
measures in NSDD 84 is the requirement to revise regulations and policies so that
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
employees may be required to take a polygraph examination in the course of
investigations of unauthorized disclosures of classified information.
Congress will hold additional hearings, on NSDD84 and the polygraph, and has
passed legislation which prohibits The Department of Defense from expanding the use
of the polygraph beyond that existing on 5 August 1982 (the day before the Carlucci
Memorandum), until 15 April 1984. NSA is exempt from this moratorium. The
moratorium is to provide Congress with time for hearings and, if necessary,
subsequent legistative action.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Specific Investigations of Espionage
In many cases, espionage activity is initially detected by other means, and the
role of the polygraph examination is to find the guilty persons among those
suspects developed by investigation. Not infrequently, the examiner will get
admissions that are useful in furthering the investigation, or get confessions.
Polygraph examinations were important to the successful prosecutions of Barnett,
Kampiles, and Helmich. All received lengthy sentences for violation of the
espionage statutes. And in each of these cases the polygraph examination was
important to counterintelligence-analysis and damage assessment.
Other espionage cases have been detected through screening operations conducted
by the Department of Defense and the Central Intelligence Agency. Here, again, the
examiners have obtained information in sufficient detail to initiate
investigations, and in some cases have obtained confessions. It is important to
note that in the screening situation, the applicant initiates the contact by
applying for assignment, employment, or some form of access. The applicants know
beforehand that they will be given a polygraph examination. The decision to take
the examination and risk exposure must be a difficult one to make. There are
several known cases in which agents have refused an assignment to seek employment
where the classified access required a polygraph examination. This deterrent
effect is illustrated by the refusal of Christopher Boyce to accept a job with the
CIA as a courier between the United States and sites in Australia. He feared that
the examination would reveal his espionage activities. He refused again for the
same reason on a later occasion when his Soviet handler suggested he apply for work
at CIA (Lindsey 1979). Likewise, Barnett opted not to apply for a staff position
at CIA or NSA because he feared the polygraph test.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Prime Minister Thatcher (1982), in her statement on the report of the Security
Commission about the Prime Case, said that "The Commission concludes that the
polygraph is the only measure of which it could be said with any confidence that it
would have protected GCHQ from Prime's treachery, either because it would have
deterred him from applying to join or would have exposed him in the course of
examination. In view of this and of the extreme gravity of the damage caused by
Prime, the Government accepts the Commission's recommendation that a full and
thorough pilot scheme should be carried out. The Commission recognizes that a
polygraph examination would be seen by some as an unwarranted invasion of their
privacy. But we are dealing with matters of the highest national security, and
those who have access to the nation's most sensitive secrets must expect to be
subject to the most rigorous vetting procedures."
Examples of Polygraph Cases ? Espionage
A U.S. directed informant with access to the intelligence and security
functions of a foreign power reported on an unidentified agent inside a highly
sensitive U.S. communication center, and said that foreign government officials
were reading U.S. classified communiques almost as soon as they were put on the
wire. These communiques dealt with highly sensitive activities involving the
Defense and State Departments, and the White House. Investigation led to five
suspects. All were interviewed with sworn statements executed denying any activity
on behalf of the foreign government. Investigation produced no additional
indication of espionage activity. All five suspects agreed to undergo examination
on the polygraph. Four of the five were determined by the examiner innocent of
involvement. The remaining individual was reported as deceptive and subsequently
confessed that he was, in fact, in the employ of the foreign power. He admitted
furnishing far more information to the foreign power than was reported by the U.S.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
informant. Much of the information he admitted providing to the foreign power was
subsequently confirmed by the U.S. directed informant.
A U.S. military member was arrested after selling classified material to
undercover federal agents following documented contact with the Soviets. The
military member denied any additional contacts or any additional violations. He
was administered a polygraph examination in which showed deception. He
subsequently admitted to additional Soviet contacts and possession of additional
classified information at his residence.
A U.S. controlled agent allowed himself to be recruited by a foreign
intelligence service in 1978. From 1978 until 1982, the agent was met periodically
by a foreign case officer, but never in the United States. *In April 1982, the
foreign intelligence service agreed to have a courier meet the agent in the United
States. The courier was arrested when the meeting took place. He was tried and
convicted. Throughout his interrogation following his :arrest, he did not identify
anyone else in the U.S. who might be engaged in espionage against the U.S. The
convicted courier subsequently agreed to undergo a polygraph examination. The
examination resulted in the courier furnishing the names of several persons in the
U.S. known or suspected by him to be espionage agents for the foreign power he
represented.
A military service received information from another agency that a military
member was involved in espionage with a hostile intelligence agency. Subsequent
investigation included a polygraph examination where the member admitted being
involved in espionage against the United States for over a year.
A military service received information that an officer was involved in
espionage. Subsequent investigation disclosed the suspect was a missile launch
officer assigned to a western base. During one of his many trips he was arrested.
Through his defense counsel, arrangements were made to grant immunity if the
officer would disclose all compromises of classified information and pass a
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
polygraph examination to confirm the admissions. He was then thoroughly debriefed
by counterintelligence personnel and said he had told everything. However, he made
many additional significant admtssions to the polygraph examiner before ultimately
passing the polygraph test.
A foreign civilian who was formerly employed by U.S. Forces, was reportedly
providing U.S. documents to a foreign intelligence service. That individual's
daughter was married to a member of the U.S. military, and she was also implicated
in the espionage activity. During an investigation of these allegations, three
sources of information were examined by the polygraph and were determined to be
truthful. Those examination .results significantly aided in the development of the
investigation. The host country police arrested the foreign civilian and his
daughter for espionage. The American member of the military forces was interviewed
and denied any knowledge of his wife or father-in-law being involved in espionage.
He agreed to undergo a 'polygraph examinaton. .The examination resulted in the
member admitting that he knew his wife and her father were involved in espionage,
and that he did not report that fact, even though he knew he was required to,
I
because he wanted to protect his wife.
Espionage Detected During Screening Examinations
An applicant reacted to questions about involvement in intelligence for a
Foreign government, espionage activity, intent to commit espionage against the
United States, the accuracy of his security forms, and other related questions.
After the tests, he admitted that his forms were not correct when he listed his
academic affiliation during two long tours abroad, amounting to a total of eight
years. He admitted that he was, during those years, a scientific advisor to the
chief of a foreign military intelligence service. Although he admitted to the
examiner that he might pass classified information to that service, he claimed that
he had not done so. He had a Secret U.S. clearance before, between tours, and
after his work abroad for that agency. He admitted that he was still in contact
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
with the foreign intelligence agency, and the subsequent test charts indicated
consistent significant reactions to questions about giving or selling U.S.
classified information to that service.
An Army Sergeant who had access to cryptologic information applied for a
civilian position. During the polygraph examination, he reacted to various
relevant questions. In the post-test interview, he admitted to various petty
crimes and miscellaneous wrongdoing. The polygraph examiner noted continued
specific reactions to relevant questions and when the Sergeant was reexamined
several weeks later, the same situation continued. His access was withdrawn and an
investigation opened. While that investigation was still in progress, he was found
dead in his automobile. It was subsequently determined that he had been engaged in
espionage on behalf of the Soviet Uhion.
A contractor employee seeking clearance and access reacted to polygraph
questions concerning espionage. She told the examiner that her former husband was
currently engaging in espionage against the United States for a foreign power.
This had been going on for several years while her spouse held various positions in
the government and with defense contractors requiring classified access. She gave
very specific detPils on his espionage activities.
An applicant for employment reacted to questions about committing espionage,
clandestine contacts, and related questions. He admitted that if employed, he
would sell classified information to a foreign intelligence service if he could get
enough money for it to live comfortably. The applicant had access to classified
information while in the service. He continued to react to counterintelligence
questions after his admissions.
An applicant described his various radical and Marxist connnections including
residence with a British Communist Party (CP) member while they both were students
at a British university. The applicant visited the Soviet Union twice, both times
with British student tours. He became particularly friendly with a young woman who
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
described herself as a student and part-time INTOURIST guide. She was his tour
guide on the second trip. Applicant later wrote his tour guide telling her he had
applied for employment with U.S. intelligence agencies. The Soviet tour guide came
to the U.S., supposedly to visit relatives; contacted the applicant; then arranged
to meet the applicant at an airport the day after his polygraph examination. He
continued to react to counterintelligence questions during his polygraph testing.
An applicant said he had lived with a foreign national, who he said was an
intelligence agent of a foreign power, while he was stationed in Berlin with U.S.
forces. The applicant provided information about her activities including
observation of a transmitter/receiver which was apparently for clandestine use. He
did not report her activity to his superiors, despite the fact that she worked for
U.S. military finance, because she paid for their apartment and he enjoyed living
with her.
/-
An applicant for employment who had lived abroad- for several years overseas
\
admitted that she was cultivated for several months by the host country
intelligence agency and finally offered a position in which she was to travel to
another country under a false identity, but as a U.S. citizen, then obtain
employment in the target country (not the U.S.). She said that after two days of
thoughtful deliberation, she declined. She had not previously reported this to
U.S. authorities, and did so during the polygraph test only after reacting to
counterintelligence questions.
An applicant who was about to retire from military service reacted to
questions about intending to commit espionage against the United States. The
applicant then described several visits to the Soviet Embassy to make arrangements
to defect to the Soviet Union. The Soviet officials took copies of his documents,
got extensive biographical information and when they learned of his pending
applications for employment with U.S. intelligence agencies, encouraged him to stay
in the U.S.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
An applicant reacted to questions about knowing othtIrs engaging in espionage,
intent on his part to engage in espionage, and to questions about the truthfulness
of the information on his security forms. He admitted that he had been terminated
from employment with a U.S. intelligence service after a year of training but
before he was assigned to a position. After being fired he engaged in free-lance
journalist work in the U.S. and abroad, with an Asian. During their work, the
Asian told him he was an intelligence agent for his country. Although the
applicant identified the agent by name, his location in the U.S., and his
intelligence service, he would not give many details of their joint activities.
After these admissions, he continued to react to questions about intending to
commit espionage against the United States and the accuracy of his security forms.
An applicant admitted that a Russian, who stayed at his home for several days,
asked for a tour of potential military targets in New Jersey. The Russian was
atlegedly with a Soviet delegation from the USSR Ministry of Health. The Russian
took numerous pictures while the subject drove him to all the major bridges and
petroleum plants in New Jersey, and to a nuclear power plant. The following year,
he allowed two Soviet citizens from the Ministry of Health to stay in his home
while they made tours during the daytime. :They were vague about where they had
been. The applicant provided additional information about other U.S. citizens who
were assisting these Soviets in their non-medical activities.
One applicant, a frequent traveler to a Communist-bloc country, told the
examiner of being drugged, and while drugged, an attempt was made to get him to
engage in a homosexual act. Later, he was directly asked to work for the
intelligence service of that country against the United States. He said he
refused, but he did not report the entrapment attempt or the offer to any U.S.
official.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Suitability Statistics
When considering for clearance and access those persons who have not yet had
the opportunity to demonstrate their loyalty and trustworthiness, it is necessary
to make judgments from their general conduct. Initial clearance is also the
appropriate point at which undesirable people should be barred from access to
classified information.
The extent and nature of admissions to criminal activity, including drugs, made
by NSA applicants, has beep set forth elsewhere in this report. Some other aspects
of the admissions made during screening should also be reported .
Falsification of the Statement of Personal History or Personnel Security
Questionnaire is reported only when it is deliberate, and the omission or false
entry is significant. Among the 20,511 applicant polygraph examinations conducted
at NSA from 1 October 1974 to 28 February 1979, there were 1,573 persons whp
falsified their security forms. That was 7.77 of the total number of persons
be
examined. Omissions included criminal arrests and convictions, employment where
their work was unsatisfactory or they were terminated, military service, relatives,
subversive organizations, citizenship, education, foreign travel and numerous other
matters. In many cases, but not all, an accurate record would have given
information that would have been a bar to clearance and access, or to employment,
or both.
Mental and nervous disorders were described by 1,306 applicants (6.4%). When
appropriate, medical releases were obtained for use by those who conduct the
background investigation. In addition, 297 (1.5%) admitted they had attempted
suicide on at least one occasion. Because these applicants were also interviewed
at NSA by a qualified clinical psychologist, an appropriate evaluation and
prognosis was possible.
There were 2,820 persons (13.8%) who disclosed close contact with foreign
nationals. Of particular interest were the 857 persons (4.2%) who admitted close
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
contact with Soviet bloc citizens, whose identities were not listed on the security
forms. Many of these were contacts were made during foreign language training
which developed into friendships, with contacts continuing through personal
meetings and correspondence.
The examination includes questions about unauthorized divulgence. There were
2,153 persons in the 1974-1979 study who had prior access to SCI level material.
Of those, 126 (5.9%) divulged classified information to members of their family and
129 (6.0%) disclosed classified information to others. The admissions, always
checked by a classification authority, ranged from a one-time disclosure of
confidential information to a member of the family to very serious compromises of
extremely sensitive operations to outsiders, including foreign nationals. There
were cases where the divulgence was not only serious, but frequent, and sometimes
to total strangers. There were a few cases where persons who did not have official
access to classified information heard it or saw it, and knowing it was classified,
passed it on to other unauthorized persons. The most flagrant case of this kind
was the son of a federal employee who was stationed abroad. The son had casual
access to his father's office, and was able to get improper access to highly
classified material in the office without others noticing. He then dislcosed this
information to the sons and daughters of foreign diplomats and other foreign
nationals in order to impress them. The disclosures were serious, but he was never
detected while doing this. Because some people do gain improper access to
classified information, all applicants at NSA are asked about unauthorized
disclosures.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
SURVEYS OF EXAMINEES
1982 Air Force Survey
In 1.982 the Air Force began a pilot polygraph project to conduct
conunterintelligence/security polygraph examinations of personnel who have access
to extremely sensitive and highly classified USAF projects. After the project
began, AFOSI began a survey of individuals who had voluntarily taken an examination
in connection with that access. The questionnaire was answered in the reception
room immediately following the examination out of the presence of the examiner. No
identifying data was included with the replies. The 1,355 individuals responding
to the survey gave their answers, totaled below in percentages, to each of the six
questions. The "other" response includes answers such as "don't know" and no
answers.
a. Were you offended, embarrassed, humiliated
YES
NO
OTHER
or degraded in any way during any part of the examination?
2.0%
98.0%
0 '
b. Do you think the examination was unfair in any way?
c.. Was there any objectioneJle or unwarranted invasion
of your privacy during the examination?
d. Do you feel counterintelligence security polygraph
examinations enhance the security of your work environment?
e. If you changed jobs and your new employer required
a counterintellignece security polygraph examination, would
you undergo the examination?
.9%
.7%
94.9%
98.2%
99.0%
99.1%
4.1%
1.5%
.1%
.2%
1.0%
.3%
F. If an act, of sabotage of espionage occurred and
you were asked to take a polygraph examination to find the
person who committed these acts, would you take a polygraph
examination to resolved the matter?
?
99.2%
.7%
.1%
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
There are only two other surveys of employees who have taken polygraph
examinations for security clearance and access, or have taken counterintelligence
examinations.
1951 Survey of Atomic Energy Commission Employees
In 1951, Professor Paul V. Trovillo, a psychologist from Arizona State
University, was employed to survey the use of the polygraph by a private.
contractor, Russell Chatham, Incorporated, at Oak Ridge, Tennessee. His survey
covered the use of the polygraph on employees at the Atomic Engergy Commission
facilities who had sensitive access. All of the employees had a full field
background investigation conducted by the FBI. The period covered by the survey
was 17 February. 1946 to 10 April 1951. (It is noted that none of the persons
involved in the extensive Soviet atomic espionage network which operated during
those years were among those who took polygraph examinations at Oak Ridge,
Tennessee. Whether or:not the polygraph deterred Soviet intellignece from an
attempt to penetrate the Oak Ridge facility is unknown.) Trovillo reported that
approximately 12,000 polygraph examinations were conducted in that period,
including applicants for clearance by employees of contractors, and applicants
seeking initial employment which required a Q clearance. In addition to the
initial polygraph examination, each employee was subject to additional
examinations. Depending on the sensitivity of access, some employees were
reexamined every three or six months.
Trovillo personally interviewed a representative sample of employees who were
in the retesting program. The employees included engineers, scientists, clerks,
and chemical workers, including their union President. Of the 59 interviewed by
Trovillo, only two (3.4%) expressed any resentment toward the polygraph tests
(Trovillo 1947).
1956 Survey of NSA Applicants
A survey was taken of all applicants for employment at NSA who were processed
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
in January 1956. Of the 522 applicants who were surveyed, 361 were male, 362 were
unmarried, 186 had a baccalaureate degree in a liberal arts field, and 206 had
degree in a scientific field. The survey covered these statements, with the
responses indicated in percentages:
a. The polygraph examiner impressed me as being professionally competent and
well trained.
1. strongly agree
.58%
2.
agree
36%
agree
94%
3.
undecided
4%
undecided
4%
4.
disagree
1%
disagree
2%
5.
strongly disagree
1%
b. In my opinion, the polygraph as used by NSA is important to protect the
security of the Agency.
1.
strongly agree
38%
2.
agree
36%
agree
74%
3.
undecided
19%
undecided
19%
4.
disagree
5%
disagree
7%
5.
strongly disagree
2%
c. If for future reasons, I were asked to volunteer for another polygraph
examination, I would:
1. be willing 91%
2. undecided 8%
3. refuse 1%
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Surveys in Law Enforcement and Commerce
The Washoe County Sheriff's Department (Reno, Nevada) surveyed 85
applicants after their ereemployment polygraph examinations. The results were
tabulated (Putnam 1978) according to the responses to these questions:
1. Were you in any manner embarrassed, humiliated, or degraded by any
part of the polygraph examination process? No: 92.9% (n=79), Yes: 7.1%
(n=6).
2. In your opinion, was there any objectionable or unwarranted
invasion of your privacy during the conduct of the polygraph examination?
No: 100% (na85).
3. Should you be hired, do you believe you will be more secure and
comfortable in your work knowing the polygraph is used in personnel
evaluation? Yes: 97.6% (na83), No: 1.2% (nal), no opinion: 1.2% (nal).
Several very similar surveys of subjects of commercial screeningi have been
made in recent years. Dr.. B. A. Silverberg of Canada conducted three surveys
of the subjects of polygraph examinations in 1980 asking the same questions
used earlier in surveys by Ash (1973) and Buckley (1980). The questions were
used again recently by Phannenstill (1983). A table shows the results of the
six surveys. The answers which are other than the one listed for each item
are mostly the opposite, but there are also a few unclear replies, or no
answers at all to that item, in the remainder.
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
L.
-E
SURVEYS OF EXAMINEES OF COMMERCIAL TESTING
% (Number) d 0
Answered Buffalo & m
. 40
NO Chicago Chicago & Denver Western N.Y. Toronto Toronto Milwaukee m
to These Ash, 1973 Buckley, 1980 Silverberg, 1980 Silverberg, 1980a Silverberg, 1980b Phannenstill, 1983 w
w
Questions
1. Do you think
the test was
unfair in any
way?
2. Did the test
or any part
of it offend
you?
3. Do you think
the test was
an invasion
of your
privacy?
ANSWERED YES
4. If the
occasion
arose would
you take a
test like
this as an
applicant for
a job?
5. If a loss
occurred
at your
company
and you
were asked
to cooperate
by taking the
Lest to help
find the
person who
caused the ,
loss, would
you?
n=241
n=270 .
n=102
n=217
n=115 '
n=220
86%
90%
100%
99%
99%
89%
(208)
(244)
(102)
(216)
(114)
'
(195)
9,1%
87%
97%
98%
98%
87%
(220)
(235)
(99)
(212)
(113)
(191)
83%
77%
98%
98%
98%
79%
(200)
(209)
(100)
(213)
(113)
(173),
96%
94%
93%
95%.
93%
94%
(232)
(253)
(95)
(207)
(108)
(207)
97%
97%
96%
96%
97% '
96%
(233)
(263)
(98)
(209)
(111)
(210)
lq
CD
L-1-0000170017000108Z1-0-L0dCll-V10
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
QUALITY CONTROL
Selection and Training
The quality control program within the Department of Defense begins with a
detailed selection process of applicants for polygraph training. By regulation all
applicants must be trained investigators with a minimum of two years Federal
Government investigative experience and possess a baccalaureate degree. Applicants
must also be a U.S. citizen, 25 years of age who have been the subject of a Special
Background Investigation. The culminating application procedure has the qualified
applicant undergo a suitability and familarization polygraph examination prior to
final selection. Polygraph training is always on a voluntary basis. Most
important, agencies are careful to select only those agents who have demonstrated
mature investigative and personal attitudes. Examiners must be able to conduct low
key interviews as well as intense interrogations. They must be highly professional
investigators who get along well with subordinates and superiors, as well as
suspects, witnesses and victims.
The basic examiner training is accomplished at the Army Polygraph School, Fort
McClellan, Alabama. This course is 12 weeks in length. The examiner learns
various test techniques and receives instruction in psychology, physiology,
pharmacology, semantics, and polygraph instrumentation procedures. Each trainee
conducts at least of 45 to 50 practical exercise examinations prior to graduation.
After graduation each examiner serves a six to twelve month internship under the
direct supervision by a certified examiner of his respective agency.*
*NSA examiners take another two weeks of formal instruction in
counterintelligence screening techniques before beginning their internship. NSA
always includes students and experienced examiners from other DoD agencies in their
course whenever requested; and on a space available basis, NSA trains examiners
from other federal agencies.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
In 1965 DoD initiated research to establish the validity of polygraph
techniques and sought a means of monitoring field polygraph operations by having a
qualified polygraph examiner conduct a "blind" analysis of the polygraph charts and
allied documentation of another examiner. After a comprehensive study, it was
concluded that blind analysis was a valid concept which would increase the quality
of polygraph examinations. This concept led to the polygraph quality control
programs now existing throughout the DoD. Under the quality control concept,
supervisory polygraph quality control examiners, generally located at the
headquarters level, review every polygraph examination conducted within their
respective agencies. The review begins with a "blind" analysis of all polygrams
collected by the field examiner. Once the "blind" analysis is performed, the
reviewer theri checks all related documents for correctness, completeness and
technical accuracy. The specific areas reviewed are pretest interview, test
construction, question formulation, chart patterns, chart markings, timing of
questions, post-test interrogation, length of examination, and the polygraph
report. Any notes, examiner comments and written admissions are analyzed and
compared to the information contained in the formal report. The formal report is
also checked for grammatical correctness and format prior to final approval.
At NSA, every case is completely tape recorded. In all cases where the subject
of the interview has made admissions which will be reported and will require
evaluation prior to granting clearance and access, the examiner reviews with the
subject, in precise detail, the contents of his notes. While doing that, he makes
an additional recording with a cassette recorder in full view of the subject, and
has the subject audibly acknowledge the accuracy of his notes on each topic as it
is covered. When the supervisor reads the examiner's report he must listen to this .
summary tape and certify on the report that the accuracy of the report has been
verified by listening to the recording.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
sx,
I.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
All Air Force counterintelligence/security examinations are audio and video
recorded. Similarly, review of the audio and video tapes of these examinations by
polygraph supervisors provides a detailed quality control check. In criminal
cases, and specific issue cases, it is customary to take a witnessed, signed, sworn
statement covering all admissions/confessions. A written statement containing a
subject's admission/confession is documented in the final report as well as any
refusal to undargo further questioning or polygraph testing.
Each agency keeps independent statistics and oher personal information on the
?
professional attributes, training qualifications and performance of each examiner.
If an examiner's proficiency is found to be falling below accepted standards,
supervisor determines if the problem is the result of an atypical series of cases,
or a change in the examiner's methodology. Methodology changes are discussed with
the examiner and, when appropriate, corrective action is taken.
Advanced training of At least one week or more is required once every two years
for all DoD polygraph examiners. In addition, examiners generally receive several
days. of advanced training throughout each year by attending state and national
polygraph association meetings and seminars. Advanced training is an integral part
of the quality control program within the Federal Government as it assures the
continuing professional development of each examiner.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
US ARMY POLYGRAPH TRAINING FACILITY
The first use of the instrumental detection of deception was in World War I to
recover a stolen code book. The U. S. Army began using the modern polygraph during
World War II to screen German prisoners of war (POW) for post?war police
assignments. Use of the polygraph spread quickly into criminal and
counterintelligence investigations due to the success of the polygraph in resolving
the POW cases.. Some Army agents received individual instruction in polygraph
techniques, and in 1948, the Army began sending special agents to the new Keeler
Institute, in Chicago, Illinois, the only polygraph school in existence at the
time. The Army's training needs soon exceeded the capacity of the Keeler
Institute. Therefore, in 1951 the Army established its own polygraph school and
attached it to the Military Police School at Fort Gordon, Georgia. Initially, the
school was eight weeks in length and was used exclusively by the military services
and the U.S. Secret Service.
In 1964, as a result of the Moss Committee Hearings about the use of polygraph
in the federal government, the need for more extensive training was recognized and
the school was lengthened to 12 weeks. In 1969, the school was lengthened to 14
weeks in order to accommodate training requirements by military and federal
intelligence agencies in the personnel security screening arena. As use of the
polygraph in the federal government increased, all federal investigative agencies,
except the CIA, began training their examiner personnel at the Army school. Three
agencies gave up their own training courses to take advantage of the excellent Army
training.
In 1975, the Military Police School and the Polygraph School were moved to Fort
McClellan, Alabama. During 1980, to accommodate the increased training
requirements of the other federal agencies, the school was condensed to 12 weeks,
permitting an increase of sessions from three to four classes per year.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Since the school's founding in 1951, the-U.S. Army has trained over 2,000
military and federal civilian investigators to be polygraph examiners. Numerous
training requests are received each year from state and municipal police agencies.
While these requests cannot be accommodated in the basic training course,
occasionally spaces are provided to state and municipal examiners in the advanced
training course. The Defense Department plans to expand the Army Polygraph
training facility to handle a maximum 24 students per class. Once expanded, the
Army may be able to be more responsive to state and municipal police agency
training requests.
There are two programs of instruction at the U.S. Army Polygraph School. The
primary program of instruction is the Basic Polygraph Examiner Training Course,
which is now twelve weeks in length and is conducted four times a year. Each class
consists of 12 students and the student to instructor ratio is 2 to 1. This ratio
provides the student with many hours of personal 'contact with qualified Instructors
during the learning process. Further, each student conducts over 40 polygraph
examinations using a variety of testing techniques. This exposure to simulated
testing and the instruction in multiple testipg techniques has established the Army
school as the leader in polygraph training.
The second program of instruction at the school is the Advanced Polygraph
Examiner Course, which is three weeks in length. It provides experienced examiners
with training in new instrumentation procedures and sophisticated testing
techniques. These lessons are then incorporated in comprehensive practical
exercises.
By regulation all tool) examiners must attend at least one advanced training
course every other year. In practice, all examiners are involved in several short
courses, seminars, and lectures every year, with a full course of a week or more at
least every other year. Many get a week or more of formal training every year.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Advanced courses used by DoD examiners are quite varied. Most commonly used are:
U. S. Army Advanced Course 3 weeks
Federal Interagency Polygraph Seminar 1 week
National Security Agency Advanced -
Polygraph Screening Techniques Course 2 weeks
Federal Bureau of Investigation
Advanced Polygraph Studies Program 4 weeks
Delta College Polygraph Workshop 1 week
University of North Carolina Polygraph
Colloquium 1 week
University of Utah Polygraph Workshop 1 week
American Polygraph Association
Annual Seminar 1 week
In addition, most DoD examiners are active in a regional or state polygraph
association. These associations hold one or more weekend seminars each year, and a
few conduct lengthy training courses. Several of the courses above and some of the
state associations seminars offer college credits or continuing educatiou units.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
PRIMARY POLYGRAPH TESTING TECHNIQUES
The test formats taught at the Army Polygraph Training Facility include the
- following:
Zone of Comparison Technique (ZCT): The ZCT was developed by Cleve Backster in
1960. The ZCT is a modification of the control question concept developed in the
1940's and introduced two new safeguards against errors and inconclusive results.
The safeguards are referred to as the sacrifice relevant and symptomatic
questions. The sacrifice relevant question allows for dissipation of excessive
general nervous tension or undue-anxiety prior to the asking of the primary
relevant questions. The symptomatic question provides a means to identify outside
issues extraneous to the test issue(s) which may be of overwhelming significance to
the examinee. Identification of such issues aids the examiner in structuring a
reliable test that is not degraded by these external issues. The symptomatic
question also helps the examiner to properly word the relevant questions.
Backster's ZCT was adopted by the Army in 1961. It is the technique of preference
for resolving criminal issues which are limited or easily defined. The ZCT, as
used by the Army, permits the asking of questions pertaining to two primary issues
in the body of the test. It also has a "SKY" phase utilized on one or two charts
which helps determine the degree of the examinee's criminal knowledge, if any, of
the primary relevant issue. Because of its demonstrated accuracy (Bersh 1969), it
has become one of the most popular and widely used techniques in government law
enforcement.
Modified General Question Technique (MGQT): The MGQT was adapted by the Army
in 1956 from the 1953 Reid control question test. It is the technique of
preference in criminal investigation testing where there are multiple issues within
a singular case to be resolved, as it allows the examiner to address primary
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
involvement, secondary involvement, guilty knowledge and evidence connecting
issues. The MGQT is one of the more popular examination formats in government law
enforcement because of its flexibility. In addition to being able to address
multiple issues, examiners can shift the relevant questions on
by collecting a mixed series questioning sequence. The Army's
control questions which are separated from the relevant issues
one or more charts
MGQT utilizes
by time and/or place.
Relevant/Irrelevant (Rh): The Rh I techniques were developed in the 1920's and
1930's, and the Keeler version became the first standardized polygraph question
technique. Rh I techniques were used exclusively within the U. S. Government until
1956, when a control question technique was introduced. Rh I techniques are used
extensively in intelligence screening examinations and occasionally in criminal
investigations. Rh I techniques in the government employ the use of irrelevant -
questions interspersed among relevant questions, plus additional special purpose
questions, including one or more control questions. The R/I technique is the
technique of preference for people who decline to answer broad crime?related
questions necessary for the development of control questions as they are used in
ZCT and MGQT, as well as these persons who are familiar with criminal polygraph
testing procedures.
Peak of Tension (POT): The POT was standardized by Leonarde Reeler in the
1930's and is used in conjunction with virtually every type of testing technique
known to polygraph.
examinee's
detected.
knowledge
This test
Essentially, the POT is a guilty knowledge test where the
of a particular item within the overall testing issue can be
is frequently used to find weapons, bodies, property, etc. In
can be a known POT, wherein the examiner and the guilty subject know the key item
and the innocent person does not, or a searching POT, where the key item is only
known by the guilty subject and not by the examiner.
Counterintelligence Screening Test (CIST): The CIST is a general question test
used to cover multiple issues pertaining to intelligence or counterintelligence
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
activity. If an examinee shows response to relevant questions during the course of
an examination, the issues to which the examinee showed responses will be
completely explored through the use of other polygraph techniques or through
interrogation. Each relevant question will be asked at least three times.
Other Techniques: The techniques described above are among the techniques
utilized in the federal government. There are a number of other techniques used in
the law enforcement and commercial polygraph field. Among the more popular are
those developed by John E. Reid, Richard 0. Arther, and Lynn Marcy. Although some
civilian schools teach these techniques, there are other civilian schools which
teach the Army techniques, or teach both Army and other techniques. Government
examiners are given familiarization lectures on the more common other techniques so
that they will understand them; but they are not expected to use them.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Chapter 4 Summaries of Research
.?
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Field Abstracts
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Gordon H. Barland and David C. Raskin. "Validity and Reliability of Polygraph
Examinations of Criminal Suspects." Report No. 76-1, Contract 75?NI-99-0001.
National Institute of Law Enforcement and Crimirial Justice; Department of
Psychology, University of Utah, Salt Lake City, Utah, March 30, 1976.
Procedure
One hundred?two criminal suspects (92 independent cases) were subjects in a
study to determine the validity and reliability of the polygraph in a field
setting. Most of the subjects were under arrest at the time of the examinations.
No victims, witnesses, or persons involved in civil suits were included.
? Eighty?four males and 18 females (age range, 15 to 54 years; education range 3
to 16 years) received polygraph tests where abdominal or thoracic respiration, skin
resistance, and cardiovascular activity were recorded. Location of test and type
of instrument varied with the subjects. It is noted that 34 subjects were tested
. with a polygraph in poor condition. Experience of the polygraph examiner is not
discussed.
During a pretest interview the examiner obtained a background history,
administered several scales from the MMPI, determined what the subject knew about
the crime, and formulated the test questions with the subject. The federal
modification of Backster zone?comparison control question test was used in 101
cases. One case used the U.S. Army 1.966 version of the Reid technique. A minimum
of three charts (10-13 questions per chart) were run; if no decision was made,
additional charts were run up to a maximum of six.charts. Charts were numerically
evaluated at time of test by scoring each of three pairs of relevant and control
questions from +3 to ?3 based-on magnitude of the differences. Truthful was
indicated by a total test score of +6 or greater, deception by ?6 or lower, and
inconclusive between +5 and ?5. In some cases the examiner's decision was based on
more information than the numerical evaluation of record.
Each case was independently reviewed by a panel with no knowledge of the
polygraph result. Evidence on each case was collected by untrained investigators
and summarized by technicians with no knowledge of the case. All available
documentation was included except polygraph and judicial outcome. If a subject
confessed as a result of the polygraph test, the confession was reported but the
source of the confession was not. If the subject had pleaded guilty, the
circumstances of the guilty plea were included. The panel consisted of two
prosecuting attorneys, one judge, and two defense attorneys. Panelists were to
disregard legal technicalities and rules of evidence, and judge guilt and innocence
based on the facts. Each member independently reviewed each case and made one of
five possible decisions; (1) deflnitely innocent; (2) probably innocent; (3)
undecided; (4) probably guilty; (5) definitely guilty. These decisions were
converted to numbers on a 5? point scale by the examiners. Scores were summed for
each case and total panel score was compared to polygraph outcome.
The following table demonstrates the agreement between panel decision and
examiner decision when +1 or greater panel score re011ected innocence, ?1 or_Lower
panel score reflected guilt, and a 0 panel score reflected inconclusive. The ten
examinations that were not independent were eliminated in this analysis.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Panel Decision
Guilty
Innocent
Inconclusive
Deceptive
47
14
5'
Examiner Decision
Truthful
3
6
4
Inconclusive
6
5
2
Omitting inconclusives there was agreement in 76% of the cases. Increasing the
panel score required for agreement increased rate of agreement (also note there
would be an increase in inconclusives) and it fluctuated around 87%.
The polygraph examiner's judgment also was compared with various other types of
criteria. The following results were obtained when panel decisions were based on a
majority of the panelists making a decision in the same direction.
Panel Decision
Guilty Innocent Inconclusive
Deceptive
43
9
14
Examiner Decision
Truthful
O.
5
8
Incoqclusive
4
3
6
When panel resuLts were compared to blind numerical evaluation of charts similar
results were obtained;
Polygraph examiner's decisions also were compared to judicial outcome in 41
cases where the jury did not know a polygraph test had taken place and are as
follows:
Examiner Decision
Judicial Outcome
Guilty Innocent
Deceptive 30 6
Truthful- 0 1
Inconclusive 3 1
Results were similar with blind numerical chart analysis.
The relationship between judicial outcome and panel majority decision is in the
following table: -
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Akiva Ben-Ishai. "Some Remarks on Polygraph Research." Paper presented at the
Ninth Annual Meeting of the American Embassy of Polygraph Examiners, Chicago,
Illinois, August 1962.
Procedure 1
A psychology student, assigned to the Israeli Police Department, determined the
relationship between polygraph examiner decision and case outcome for those cases
whose results had been determined by other means. It should be noted that the
description of the procedure is incomplete.
Results
The psychology student's evaluation of guilt or innocence from the records was
consistent with the polygraph results in 94% of the cases. It is not possible to
determine from the presented data the percentage of guilty subjects and the
percentage of innocent subjects correctly classified.
Procedure 2
The student was given all the charts from ten criminal cases in which the guilt
or innocence had been definitely -established by independent means. The student
measured the amplitude of skin resistance responses to each of the control
questions and to each of the relevant questions. He then averaged all the
responses to the control questions and averaged all the responses to the relevant
questions. If the average amplitude of the control question responses was greater
than the average amplitude of the relevant question, responses, the subject was said
to be nondeceptive. If the average amplitude of the responses to the relevant
questions was greater than that of the control questions, the subject was said to
be deceptive. The chart analysis did not include respiration or cardiovascular
recordings, and was limited to this unusual method of averaging electrodermal
responses for analysis.
Results
All ten subjects were correctly classified as guilty or innocent.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Philip J. Bersh. "A Validation Study of Polygraph Examiner Judgments."
Journal of Applied Psychology 53 (1969): 399-403.
This was a report of research for the Department of Defense Joint Services
Group on Lie Detection Research, United States Army Behavioral Science Research
Laboratory, Washington, D.C., April 1968.
Procedure
The lie detection judgments of polygraph examiners in criminal investigations
conducted by the military services were validated against unanimous guilt-innocence
decisions by a panel of four Judge Advocate General (JAG) attorneys. The panel of
lawyers had access to the complete investigation file, except that all references
to the polygraph examinations were removed. Cases were selected at random from
those conducted from 1963 to 1966 with some restrictions. Cases involving
confession were ruled out because some polygraph examiners could have made their
judgment of deception after the subject had confessed. No cases where the examiner
reported inconclusive results were included. Cases were selected so that there
were about an equal number of General Question tests and Zone of Comparison tests,
and with each type, an equal number of judgments of deception indicated and no
deception indicated. Examiners did have access to a subject's case file, but the
authors maintain the file was not complete and that polygraph tests are only given
when there is real doubt about a person's guilt or innocence. The polygraph
examiners' judgments were subjective decisions regarding subjects' guilt or
innocence.
A total of 326 case files were submitted to the panel. Each member of the
panel was initially required to judge whether a file contained sufficient evidence
to warrant a decision of guilt or innocence. Files with inadequate evidence Were
eliminated from further consideration, and 157 remained in which all four decisions
were unanimous. The attorneys were given explicit instructions to disregard all
legal technicalities and to judge each case solely on the evidence contained in the
file.
Results
Results are presented in the following table (cases are those involving
unanimous panel judgements):
Panel Judgment
General Question Test Zone of Comparison Test
Guilty. Innocent Guilty
Innocent
Examiner
Guilty
31
4
34
3
1741377E
Innocent
1
32
4
48
The examiner and the panel agreed in 92% of the cases. When polygraph examiner
judgment was compared with panel judgment for cases where a majority of the panel
agreed on guilt or innocence (instead of unanimous decisions), there was agreement
by the examiner hind the panel in 75% of the cases.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
M. E. Bitterman and F. L. Marcuse. "Cardiovascular Responses of Innocent
Persons to Criminal Interrogation." American Journal of Psychology 60 (1947)
407-412.
Procedure
A theft of $100.00 took place in one.of the rooms of a campus dormitory at
Cornell University. It was recognized that the theft could have been committed
either by one of the men in the dormitory or an outsider. Two professors of
psychology who had no polygraph training but did have some books and articles on
the topic, conducted polygraph examinations of the 81 men who lived in the
dormitory, as an aid to the police investigation. Respiration and cardiovascular
activity were recorded. They were unable to devise a method to evaluate the
pneumograph pattern. The first stage of testing used a fixed sequence
relevant-irrelevant technique of seven questions. It opened with two irrelevant
questions, then a relevant question, an irrelevant question, and three relevant
questions. The last relevant question was a general one, asking the subject if he
had answered all the questions truthfully. The whole pretest interview and one
chart took ten minutes.
Results
Subjects were placed in 1 of 4 categories based on their responses to relevant
and irrelevant questions (reliability data on this classification scheme ?is
presented): 1) Insignificant response to all items (N=31); 2) Moderate reactions
to all items (n=23); 3) Extensive reactions to all items (n=20); and 4) Larger
reactions to relevant than irrelevant items (n..7). Subjects in groups 3 .and 4 were
retested from .l to 5 times using relevant-irrelevant tests; with retest their
patterns became similar to those of groups 1 and 2. Supplementary peak of tension
tests indicated nothing significant. Thus, the authors concluded that none of the
men had committed the theft. Their conclusion was supported by other evidence (not
specified) which indicated that a person who was unavailable for examination was
guilty. Thus, this study classified 100% of the innocent subjects correctly.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Avital Ginton, Netzer Daie, Eitan Elad, and Gershon Ben-Shakhar. "A Method for
Evaluating the Use of the Polygraph in a Real-Life Situation." Journal of Applied
Psychology (1982) 67(2): 131-137.
Procedure
Twenty-one males (mean age 29.2 years) from the Israeli police participated in
a study involving detection of deception in a real-life situation. Subjects were
administered pencil and paper tests presented as part of a police course. The
answer sheet for one of the tests had a hidden chemical page that recorded what the
subject had written. The chemical page was removed, the answer sheet returned, ani
subjects were asked to score-their own tests. Seven subjects actually changed
their answers as they were scoring the test. This was determined by comparing the
scored answer sheets with the answers indicated on the chemical page.
After several days subjects were told they were suspected of cheating and were
offered an opportunity to take a polygraph test. It was made clear that their
future could depend on test outcome. All 21 subjects initially agreed to the test;
however, 1 guilty subject did not arrive for the test, 3 guilty subjects confessed,
and 1 guilty and 1 innocent subject refused to take the test. Thus, 13 innocent
subjects and 2 guilty subjects took the polygraph examination.
Testing was done at police laboratories by 7 interrogators with at least
one and one-half years experience as professional polygraphers. Respiration,
electrodermal activity, and cardiovascular activity were recorded. Each
interrogation involved 3 examiners - 1 interrogator, 1 who later blindly evaluated
the charts, and 1 who observed from another room. A control-question test with 9
questions, 2 relevant questions, and 2 control questions was administered. There
were 3 repetitions of the 9 questions and a card test after the first series. The
polygraphers did not know subject status or the proportion of guilty subjects. Two
months later charts were blindly re-evaluated by 8 examiners (5 of the original
examiners and 3 additional examiners). Decisions of deception and no deception
were based on both global subjective judgments and a numerical field scoring
technique. The numerical procedure involved scoring each relevant and control pair
from +3 to -3 and summing across responses and charts. Scores between +5 and -5
were considered inconclusive. Global subjective judgments were based on different
data depending on the examiner's role in each case. The blind evaluators used the
polygraph charts, the observer used subject behavior, and the interrogator used
both the charts and behavior.
Results
The following table presents results of subjective judgments of the 3 examiners
initially involved in the polygraph testing of the 15 subjects:
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Interrogator
Actual
Innocent
Guilty
Guilty
2
2
Innocent
0
11
Inconclusive
0
0
Predicted
Blind Evaluator
Guilty
1
3
Innocent
1
7
Inconclusive
0
3
Observer (behavior only)
Guilty
0
2
Innocent
2
11
The original interrogator who based his decision on polygraph charts and behavior
was most accurate, correctly classifying 100% of the guilty subjects and 84.6% of
innocent subjects.
Results for numerical scoring are presented in the following table:
Predicted
Actual
Guilty Innocent
Interrogator
Guilty 1 1
Innocent 0 6
Inconclusive I 6
Blind Evaluator
Guilty 1 1
Innocent 0 5
Inconclusive 1 7
Numerical scoring increased the numer of inconclusives, but did not change the ?
accuracy of the interrogator if inconclusives are omitted. The.blind evlluator did
have higher accuracy with numerical scoring (omitting inconclusives). The authors
note that the blind evaluator was under considerable time pressure'for the
subjective scoring and suggest that this factor may have influenced accuracy.
The 8 polygraphers who reanalyzed the data did so using both subjective and
numerical scoring. There were 16 decisions for each type of scoring for guilty
subjects (2 guilty subjects, 8 examiners) and 104 decisions for each type of
scoring for innocent subjects (13 innocent subjects, 8 examiners). Results are in
the following table:
Predicted
Actual
Guilty Innocent
Subjective
Guilty 15 18
Innocent 1 84
Inconclusive 0 2
Numerical
Guilty 11 8
Innocent 0 38
Inconclusive 5 58
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Eugene C. Edel and Jacob Jacoby. "Examiner Reliability in Polygraph Chart
Analysis: Identification of Physiological Responses." Journal of Applied
Psychology (1975) 60(5): 632-634.
Procedure
Forty randomly-selected polygraph cases from preemployment screening
examinations conducted by a federal agency were used to determine reliability of
polygraph chart analysis. There were 10 polygraph examiners ranging in actual
interview experience from three months to ten years. Each examiner was the
examiner in 4 of the cases and was a blind rater (with no knowledge of interview
information) in 8 cases. Thus, each case was analyzed by 1 examiner and 2 blind
raters.
The forty cases contained 2,530 relevant, irrelevant, control, and
overall-truth questions. The relevant-irrelevant technique was used and several
topics were covered in each examination. Examiners were required to state whether
or not there was a significant physiological response in the cardiosphygmograph,
electrodermal, and respiration channels to each questions, requiring 7,590
decisions. The federal agency which conducted these examinations does not report
deception or no deception in screening cases. It only reports either that there
are consistent, significant physiological reactions to specific relevant questions,
or there are no reactions to relevent questions. Agreements between examiners and
raters were based on the responses following each question in each of three
recording channels. Thus, if both judges said there were significant physiological
reactions to the question in the electrodermal channel, and no reactions in
cardiovascular and respiratory channels, this would constitute three agreements.
s. Results
The percentage agreement between examiner and both raters was 96% for
cardiovascular, 95% for electrodermal, and 96% for respiration, with the average
agreement across physiological channels 96%. Percentage agreement between raters
was 96% for cardiovascular, 91% for electrodermal, and 96% for respiration with an
average agreement across channels 94%.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Eitan Elaad and Esther Schahar. "Polygraph Field Validity," in Israel Nachshon
(ed.) "Scientific Interrogation it Criminal Investigation." Selected papers
presented at the First National Conference on Scientific Interrogation in Criminal
Investigation held at Bar-Ilan University, Ramat-Gan, Israel, 3-4 November 1976.
Reprinted in a Special Issue of Crime and Social Deviance 6(1) (Spring-Summer
1978): 4-5.
Procedure
Of all the cases conducted by the polygraph examiners in the Scientific
Interrogation Unit of the Israel Police during the years 1973 and 1974, 184 were
confirmed by conviction or confession. The classification of these cases as ?
deceptive or truthful was determined by the utilization of the polygraph.
Results
Of the 184 cases that were confirmed, there were 10 polygraph results that were
inconclusive (18.4 percent). There were 6 errors (3.3 percent), 5 false negatives
and,1 false positive. Eliminating the inconclusive reports, and assuming the
conviction or confession was correct when there was disagreement, polygraph tests
correctly classified 96.6% of the guilty subjects. The authors note that polygraph
results could have influenced the confessions and convictions, the sample was not
random, and false confessions and convictions could occur.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4_
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Robert H. Edwards. "A Survey: Reliability of Polygraph Examinations Conducted by
Virginia Polygraph Examiners." Bureau of Forensic Science, Department of General
Services, Commonwealth of Virginia, July 31, 1981, Richmond, Virginia, 62 pp.
Reprinted in Polygraph 10(4) (December 1981): 229-272.
Procedure
The study, conducted for the Commonwealth of Virginia, involved a survey of
fully licensed Virginia polygraph examiners. Twenty?four examiners of the Virginia
State Police and 38 examiners with local law enforcement agencies responded (77.5%
of the total law enforcement sample). The response rate for private firms was so
low it will not be reported.
The survey questionnaire had six questions including questions about the total
number of cases, the conclusion of each case (truthful, deceptive, inconclusive),
the number of cases verified as truthful and verified as deceptive, and the number
of those cases in which later information indicated the examiner was in error
(false negatives, false positives). Only specific issue examinations were included.
Results
The 62 law enforcement examiners included in the survey conducted 2,433
specific issue examinations in 1980. Of these, 1,113 (45.7 percent) were reported
as truthful, 983.(40.4 percent) were reported as deceptive, and 337 (13.9 percent)
were reported as inconclusive.- Excluding the inconclusives, 53.1 percent were
reported truthful and 46.9 percent were reported deceptive.
The following table illustrates the polygraph examiners' determinations and the
dispositions of the 959 cases that were verified by means such as conviction,
acquittal, confessions, etc.
Case Disposition
Guilty Innocent
Examiner's Guilty 587 9
Decision Innocent 7 356
Thus, 98% of the cases reported by the examiner as truthful were verified as
truthful and 98.5% of the cases reported by the examiner as deceptive were verified
as deceptive. It should be noted that case disposition was not necessarily
independent of the polygraphexaminer's decision.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
t_
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Frank Horvath. "The Effect of Selected Variables on Interpretation of
Polygraph Records." Journal of Applied Psychology 62 (2) (1977):127-136.
Procedure
Ten field-trained polygraph examiners from law enforcement agencies (5 having
more than 3 years experience and 5 having less than 3 years experience including
some interns) judged a sample of polygraph records from 112 criminal suspects from
independent cases investigated from 1969-1972. The records were sampled in stages
so that there were an equal number of verified and unverified cases, truthful and
deceptive subjects, and crimes against people and crimes against property. Thus,
there were 8 categories and 14 sets of records in each category. Verified cases
were those where a confession had been obtained and the examiners had correctly
identified the subject. Unverified cases were those where there was no knowledge
of ground truth. The truthful versus deceptive category was based on polygraph
examiner judgment.
The examinations used a standard-control question technique with some
modifications. All examinations included, in order, a pretest interview, a control
question test chart, a stimulation test chart, and another control question test
chart. In 48 of the 112 examinations there were additional test charts. Examiners
were to decide overall subject guilt or innocence, rate on a 6-point scale
confidence in their judgment, and rate on a 5-point scale the ease of interpreting
each physiological measure. Electrodermal activity, cardiovascular activity, and
abdominal respiration were recorded. Twenty-six subjects also had thoracic
respiration recorded.
Results
Of the 1120 truth-deception judgments, 707 (63.1%) were correct and 15 (.137.)
were inconclusive. An analysis of variance with within factors of verification
(verified-unverified), truthfulness (truthful-deceptive), and crime type
(person-property), and between factor experience (high-low) was computed. The
dependent measure was percentage of correct judgments by each examiner in each
category. There was no significant effect for experience. There were significant
effects for the other factors including a significant verificatiou X truthfulness X
crime type interaction. The following tables demonstrate the percentage
distribution in each category:
Person Crime
Verified Cases Unverified Cases
Deceptive Truthful Deceptive Truthful
Examiner Deceptive 78.6 50.0 80.0 62.1
Judgment Truthful 21.4 50.0 20.0 37.9
Property Crime
. Verified Cases Unverified Cases
Deceptive Truthful Deceptive Truthful
Examiner Deceptive 75.7 47.9 64.3 33.6
, Judgment Truthful 24.3 52.1 35.7 66.4
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Comparisons also were made between records with only the basic set of three
charts and those with more than the basic set. 52.5% in the former category and
71.1% in the later category were correctly classified. Consistency among examiners
was .89 for verified and .85 for unverified records based on Hoyes intraclass
reliability coefficient. Data on examiner confidence and ease of interpreting
record sets is also reported.
It should be-noted that, in this study, all cases were from criminal suspects.
Thus, innocant subjects would have obtained some knowledge about the crime.
Additionally, the initial polygraph examination was based on an overall assessment
of guilt or innocence and any scoring criteria used were not reported.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Frank S. Horvath and John E. Reid. "The Reliability of Polygraph Examiner
Diagnosis of Truth and Deception." Journal of Criminal Law, Criminology and Police
Science 62(2) (1971): 276-281.
Procedure
Forty sets of polygraph records were selected from the files of John E. Reid
and Associates. Twenty sets were from verified guilty subjects and 20 sets were
from verified innocent subjects. The charts contained 164 relevant questions of
which 81 were confirmed as deceptive answers and 83 were confirmed as truthful
answers. The forty sets of records wele from twenty-five cases involving theft,
sexual misconduct, sabotage, bribery, and criminal damage to property. The
polygraph records contained no sets that were very obvious and no sets that were
considered uninterpretable. Thoracic and abdominal respiration, blood
pressure-pulse rate, muscular movements, and electrodermal activity were recorded.
Examinations used the standard Reid Control-Question Technique.
Ten Reid examiners reanalyzed the records. They were not familiar with the
cases and they agreed not to discuss any of the charts among themselves while the
research was in progress. Seven examiners had over a year of experience; three had
only four to six months of experience and were still in their internship training
program. The examiners did not know the number of truthful and deceptive sets of
records. they were not allowed to call any set inconclusive, but they could call
individual questions inconclusive.
Results
The ten examiners correctly classified 85% of the guilty cases and 90.5% of the
innocent cases. Experienced and inexperienced examiners differed in percentage of
correct classifications. The seven expertenced examiners correctly classified from
85% to 97.5% of the cases, with an average of 91.4%. The three inexperienced
examiners correctly classified from 77.5% to 90%, with an average of 79.2%.
Analysis of'Lindividual questions revealed that, for truthful questions, 82.7%
were correctly classified, 9.6% were incorrectly classified, and 7.7% were
inconclusive. For deceptive questions, 81.4% were correctly classified, 12.5% were
incorrectly classified, and 6.2% were inconculsive. Omitting inconclusives, 89.6%
of the truthful questions and 86.7% of the deceptive questions were correctly
classified. Experienced examiners had higher correct. classification rates for
individual questions than inexperienced examiners. It should be noted that it is
not known from the presented data whether examiners tend to misclassify the same
people.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Fred L. Hunter and Philip Ash. "The Accuracy and Consistency of Polygraph
'Examiners' Diagnosis." Journal of Police Science and Administration 1(3) (1973):
370-375.
Procedure
Twenty cases from the files of John E. Reid & Associates were selected of which
10 were verified truthful and 10 were verified as untruthful. All the charts were
of a control question technique, recorded on a four-channel polygraph (cardio,
electrodermal, thoracic and abdominal respiration). The 20 sets of charts had 88
relevant questions of which 41 were verified as untruthful answers and 47 were
verified as truthful.
Seven examiners from John E. Reid & Associates analyzed each record on two
occasions separated by at least a three-month interval. Examiners did not know
they were analyzing the same records on the two occasions. Examiners analyzed the
records by making a subjective judgment of truth or deception based on comparison
of control and relevant questions: Six examiners had more than one year of
experience and one examiner had four and a half months of experience (was still an
intern).
Results
Examiners correctly classified 86% of the truthful cases and 87% of the
deceptive cases. The.results were the same for the first analysis and the second
analysis. Omitting inconclusives, 86% of the truthful cases and 88% of the
deceptive cases (occasions combined) were correctly classifed.
Occasion-to-occasion agreement for'the'same examiner on the individual cases ranged
from 75% to 90% (average 85%).
The seven examiners correctly classified 81% of the truthful questions and 83%
of the deceptive questions (occasions combined). Omitting inconclusives, 85% of
the truthful. questions and 88% of the deceptive questions were correctly
classified. Occasion-to-occasion agreement for the same examiner ranged from 67% ?
to 91% (average 81%).
It should be noted that it is not known whether examiners tended to misclassify
the same people, and that, in the analyses, each misclassification by each examiner
is treated as an error. The authors note that the poorest results were obtained
from the least-experienced examiner.
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Verne W. Lyon, Deception Tests with Juvenile Delinquents. Journal of Genetic
Psychology 48(3) (1936): 494-497
Procedure
One hundred cases were selected at random from the polygraph files of the
Institute for Juvenile Research in Chicago. All of the polygraph charts were of
juveniles of which 81 were boys, and 19 were girls, over age 15, who were suspected
of crimes but had not confessed, and positive corroboration by other means had not
yet been established.
The instrument made a record of respiration and circulatory activity. The
technique was unique, but is in the general category of relevant-irrelevant.
During the first recording no questions were asked. "During the second recording
there were three indifferent questions, free from any complex situations, and so
framed that the truth of the answers can be checked at once." The critical
questions followed, uninterrupted by any indifferent stimuli. All questions were
phrased to be answered "yes" or "no."
The institute attempted to follow up on these 100 cases to see if their results
were corroborated by positive evidence or confession. There were no inconclusive
cases among those reviewed.
Results
Of the 100 polygraph tests, 20 were called "clear", or truthful; 80 cases were
reported as "disturbed records," that were "presumptive of deception." Of the 20
"clear" or truthful cases, only 7 (35 percent) could be follo4ed up. All 7 cases
were verified as innocent by factual evidence. Of the 80 "disturbed records" or
deceptive cases, 33 (41 percent) of the subjects later confessed, many of them
immediately after the test. Thus, the polygraph examiner correctly classified 100%
of the verified cases.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4 Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Jesse Orlansky. An Assessment of Lie Detection Capability, Institute for
Defense Analyses, July 31, 1962. Report UBG 62-641
Procedure
The United States Army Military Police prepared a report for the assessment in
which they reported on the follow up of 3,153 cases in which the examiner reported
no deception indicated."
Results
There were 3,080 (97.7%) "instances in which results of examination in this
category were in accord with results of other investigative techniques."
There were 73 (2.3%) "instances in this category in which contrary results were
obtained through investigation or interrogation."
The extent to which there may be other errors is unknown. Those found
nondeceptive, and not otherwise found guilty by court or appear guilty by contrary
evidence, do not complain.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4? Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Robert B. Peters. "A Survey of Polygraph Evidence in Criminal Trials." The
American Bar Association Journal 68 (February 1982): 162-165.
Procedure
The author, a polygraph examiner at the Wisconsin State Crime Laboratory,
surveyed the results of stipulated polygraph examinations conducted at the state
crime laboratory during the 33-month period from 1976 to 1979. Stipulation means
that the prosecution, defense and judge agree to admit results of the polygraph
examination into evidence, with the examiner testifying and subject to
cross-examination. Actually, the examiner testified in only 11 of the cases.
Initially, there were 220 cases. Twenty-five were eliminated because they were
inconclusive (not clearly indicative of truth or deception). Twenty-three cases
were eliminated because no opinion could be given (e.g. subject unfit, made
statements that resolved the issue of the test). The disposition of the 89 cases
the examiner classified as truthful and the 83 cases classified as deceptive was
determined.
Polygraph tests used standard control-question procedures. Respiration,
electrodermal activity, and cardiovascular activity were recorded. Charts were
numerically scored with a score of +6 or more considered truthful, -6 or less
deceptive, and between -5 and +5 inconclusive.
Results
The following table shows the polygraph examiner's aecision versus the
disposition of the case (i.e. whether the outcome of the case reflected
truthfulness or deception by the examinee):
Examiner's
Decision
Deceptive,
Not Deceptive
Disposition of Case
Guilty Innocent
62 12
1 88
Eighty-five of the 89 cases classified truthful by the polygraph examiner
involved criminal suspects. One of the suspects was convicted at trial and 84
suspects had charges dropped. Four of the truthful cases involved alleged victims
or witnesses. In all 4 cases, the accused was convicted at trial making these
subjects truthful.
Of the 83 cases classified as deceptive by the examiner, 2 claimed to be
victims of crimes; the rest were criminal suspects. Of the suspects, 51 pleading
guilty, 1 was acquitted, 3 had charges dropped with conditions (not included in
table), 11 had charges dropped, 6 cases were pending (not included in the table),
and 9 were convicted. In the cases involving the 2 alleged victims, charges
against the accused were dropped at trial making these subjects deceptive.
The following table illustrates the outcome of cases that went to trial (the
examiner testified in only 11 of the cases).
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Examiner Testifying Examiner Not Testif, jag
Convicted Acquitted Convicted Acquitted
Examiner's
Deceptive
9
1
9
0
Decision
Truthful
1
0
0
0
As a measure of validity, the Peters survey must be tempered by the realization .
that the results of the polygraph examination had a direc: influence on those who
made decisions to prosecute or drop charges. However, it was not the sole basis
for those decisions. as is evidenced by the decision to prosecute one who was called
truthful. Polygraph results were also influential in bolstering the testimony of
witnesses and victims in the view of those who decided to use them in court,
despite the fact that the polygraph examiner did not testify in their behalf.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
David C. Raskin. "Reliability of Chart Interpretation and Sources of Errors in
Polygraph Examinations." Report No. 76-3 Contract 75-NI-99-0001. National
Institute of Law Enforcement and Criminal Justice, Law Enforcement Assistance
Administration, U.S. Department of Justice, Department of Psychology, University of
Utah, Salt Lake City, Utah, June 7, 1976.
Procedure I
Sixteen sets of polygraph records from a previous study (Berland and Raskin,
1976) were used. Twelve were from confirmed guilty subjects and four were from
confirmed innocent subjects. All examinations had used the federal modification of
Backster zoni-comparison technique consisting of three or more charts of ten
questions each. Physiological responses were recorded on a field polygraph.
Twenty-five field polygraph examiners with different types of training and
experience (18 with at least one year experience; 13 with formal training in
numerical scoring) rescored the records. 'Seven examiners explicitly used numerical
scoring.
Results
Of the 400 judgments, 78.8% were correct, 8.2% were incorrect, and 13% were
inconclusive. Excluding inconclusives, there were 90.5% correct decisions. There
were 33 errors - 20 false positives and 13 false negatives. (It is not possible
from the presented data to determine the percentage correct for guilty and innocent
subjects separately).
The number of correct decisions did not differ with length of examiner
experience (more than one year versus less than one year). The ten examiners who
had attended school emphasizing numerical scoring had higher accuracy than the
fifteen examiners who attended other schools (97.1% versus 86.9%). The seven
examiners explicitly using numerical scoring had a significantly higher accuracy
than the eighteen who did not (98.9% versus 87.9%). Comparison among examiners
trained in numerical scoring revealed that the seven examiners explicitly using
numerical scoring had higher accuracy than the six not explicitly using numerical
scoring (98.9% versus 88.5%).
Procedure 2 and Results
Different samples of polygraph records were analyzed to determine whether
decisions of polygraph examiners depended on the person or agency involved in case
referral. These analyses were computed to test the "friendly polygrapher"
proposition of Orne (1975). Orne suggested that successful detection depends on
motivation to deceive and threat of consequences. Thus, subjects referred for
polygraph examinations by defense attorneys should be less easy to detect since
they know results will not be used if they are found deceptive. The first sample
consisted of 204 criminal cases from an experienced examiner who used
control-question techniques. Ninety-eight cases were examined for defense counsel
and 106 were examined for law enforcement authorities. Of the defense cases, 77.6%
were classified truthful, 20.4% deceptive, and 2% inconclusive. Of the law
enforcement cases, 75.5% were truthful, 19.8% deceptive, and 4.7% inconclusive.
Thus, the two groups did not differ in type of decision.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A second sample (from Raskin and Barland, 1976) involved cases from two private
polygraph firms. Nineteen examiners used control-question tests with criminal
suspects. Fifty-four cases were cases for defense counsel and 57 were cases for
law enforcement authorities or the subject's employer. Comparison of numerical
scores for at least two charts of each set revealed no significant differences by
type of case.
A third sample involved control-question tests conducted by the author on
criminal suspects. Fourteen cases were from defense counsel and 13 were from law
enforcement officials. Each case involved at least three *charts and were
numerically scored. Defense counsel cases were significantly more deceptive than
law enforcement cases. Thus, there was no support for the "friendly polygrapher"
hypothesis.
Procedure 3 and Results
Twelve cases involving conflicting polygraph examinations are analyzed to
determine reasons for polygraph error. Descriptive data are presented.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Stanley M. Slowick and Joseph P. Buckley, "Relative Accuracy of Polygraph Examiner
Diagnosis of Respiration, Blood Pressure, and GSR Recordings." Journal of Police
Science and Administration 3 (3) (1975): 305-309.
Procedure
Polygraph charts from 15 truthful and 15 untruthful subjects whose cases had
been independently vertfied and supported the original examiners conclusions were
randomly selected from the laboratory case files of John E. Reid & Associates. The
cases involved theft, industrial sabotage, drug abuse, and rape. Seven experienced
staff examiners (average experience 3.8 years) participated in the study in which
they were instructed to render decisions on the subject's veracity of 141 relevant
questions (71 verified truthful and 70 verified untruthful) and 30 case issues by
reviewing the complete charts. The examiners were not given the test questions or
any other case information, and were instructed not to discuss their cases.
Examiners also made decisions about the 30 cases and 141 relevant questions
analyzing each physiological measure independently. This was done by analyzing
each physiological measure on a different occasion separated by three month
intervals and masking two channels while the third was analyzed. Abdominal or
thoracic respiration, blood pressure/pulse rater and electrodermal activity were
the physiological measures.
Results
Chart review as a composite of three physiological measures:demonstrated that
the examiners correctly classified 91% of the truthful cases and 84% of the
untruthful cases. Omitting inconclusives, 93% of the truthful and 85% of the
untruthful cases were correctly classified. The individual question analysis
revealed that 88% of the truthful questions and 73% of the untruthful questions
were correctly classified. Omitting inconclusives, these figures are 93% and 80%,
respectively.
The analysis by independent physiological measure at three month intervals
revealed slightly lower correct classification rates. Differences between
physiological measures in overall classification rates were small. It should be
noted that it is not possible to determine from the presented data if examiners
tended to misclassify the same subjects. The presented data consider each
misclassification by each examiner an error.
Declassified and Approved For Release 2013/04/16 : CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Douglas E. Wicklander and Fred L. Hunter. The Influence of Auxiliary Sources
of Information in Polygraph Diagnosis." Journal of Police Science and
Administration 3(4) (1975): 405-409.
Procedure
Twenty sets of polygraph records selected from the files of John E. Reid &
Associates, of which 10 sets were those of verified truthful subjects and 10 sets
of verified untruthful subjects. The sets contained 89 relevant questions, 43
verified deceptive and 46 verified truthful. The instruments used were four
channel Stoelting polygraphs (recording thoracic and abdominal respiration, blood
pressure-pulse rate, and electrodermal activity) and the technique was Reid
Control-Question Technique. Six experienced polygraph examiners on the Reid staff
(average experience = 4 years) read each set of polygraph records on two different
occasions, with a two-month lapse to eliminate recall. Each examiner had to call
each set and each question truthful or deceptive without discussing his opinions
with others. The judgment was subjective based on comparison of relevant and
control questions.
Prior to the first review of charts the examiner was told only the type of case
involved. Prior to the second review of the charts the examiner was given a
synopsis of the case history, subject data, verbal and nonverbal behavior symptoms
routinely noted by the original examiner, and the relevant test questions. Of
interest was whether this extra information would influence chart analysis.
Results
? . Examiners correctly classified 87% of the truthful cases and.90% of the .
deceptive cases on the first chart review. Omitting inconclusives, 95% of the
truthful and 92% of the deceptive cases were correctly classified.
At the second review 87% of the truthful cases and 98% of the deceptive cases
were correctly classified. Omitting inconclusives these figures are 91% and 98%,
respectively. There were fewer inconclusive judgments at the second review.
On the eirst review of the individual question analysis, 80% of the truthful
questions and 86% of the deceptive questions were correctly classified (omitting
inconclusives, 91% and 91%, respectively). On the second review, 82% of the
truthful questions and 95% of the deceptive questions were correctly classified
(omitting inconclusives, 92% and 97% respectively). It should be noted that it
cannnot be determined from presented data whether examiners tended to misclassify
the same people. Each misclassification is treated as an error.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Jan Widacki. "Analiza Przestanek Diagnozowania W. Badanich Poligraficznych."
Uniwersytetu Slaskiego, Katowice, 1982 ("The Analysis of Diagnostic Premises in
Polygraph Examinations.")
Procedure
All of the polygraph cases conducted on criminal suspects by the Department of
Criminalistics of the Silesian University (Poland) in the years 1978-1979 which
were verified by subsequent legal proceedings were analyzed. Sixteen people were
judged guilty by both the polygraph examiner and subsequent legal proceedings.
Twenty-two subjects were judged innocent by the examiner and subsequent legal
proceedings.
Results
Control question tests were analyzed by the Backster numerical scoring method.
In 91.6% of the cases, guilty and innocent subjects were differentiated (separate
percentages are not given for guilty and innocent).
With peak of tension tests, 80% of the guilty persons reacted to the relevant
questions. The authors say that 60% of the innocent subjects did not react to the
relevant questions and 40% showed only minor reactions to relevant questions.
Visual biofeedback on peak of tension tests enhanced the ability to discriminate
guilty subjects from innocent subjects. It should be noted that the authors give
incomplete details about a number of aspects of the study.
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
? Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Laboratory Abstracts
Control Question Techniques
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Gordon H. Berland (Prepared By). "A Validity and Reliability Study of
Counterintelligence Screening Test." Security Support Battalion, 902d Military
Intelligence Group, Fort George G. Meade, Maryland, May 12, 1981.
Procedure
Fifty-six U.S. Army Employees (38 military, 18 civilian; 40 men, 16 women; age
range 21 to 55 years; educational level 12 to 17 years) were subjects in a mock
screening situation to determine polygraph accuracy with the Counterintelligence
Screening Tesz with directed lie questions. This test is derived from the federal
version of the zone comparison test. The test contains 13 questions - five
relevant, four directed lies, and two symptomatic. Polygraph tests were
administered by three polygraph examiners trained at the U.S. Army Military Police
School polygraph course, certified by the Department of the Army with three, six,
and nine years of polygraph experience. Procedures were used to standardize
testing procedure among the examiners. Five different models of field polygraphs
were used. Respiration, skin resistance, and relative blood pressure were recorded.
Questions involved subject's data of birth, place of birth, education,
residences, and employment. All subjects initially filled out a biographical data
sheet with truthful information. The information was confirmed by background
investigations of each subject. Subjects then filled out new biographical data
sheets (some giving false information) to give to the examiner. It was randomly
determined which subjects would answer all questions truthfully and which would lie
to one question. The question subjects lied about varied. Deceptive subjects were
told they would receive $20 if they appeared truthful on the polygraph test. The
rationale of lie detection was explained to all subjects. After a pretest
interview, subjects received a stimulus test. The biographical data sheet was then
reviewed. The polygraph test consisted of a minimum of three presentations of 13
questions. If the examiner was uncertain of the subject's status at the end of
three charts, additional charts were run with a maximum of six. If the examiner
was still uncertain after six charts, the subject was called inconclusive. The
examiner was not allowed to interrogate the subject.
Three methods were used to evaluate the charts:
1. Zone method - Each relevant question was evaluated against the larger
control question in its zone for each channel and scored from +3 to -3 as taught at
the U.S. Army Police School Polygraph course. Scores for each relevant question
were summed across charts. The answer to a question was designated as truthful if
the score to a question was +3 or greater, deception if the score was -3 or lower,
and inconclusive if the score was between +3 and -3.
2. Greatest control - The method used the same procedure as the zone method
except that the five relevant questions were evaluated against the one control
question on the chart with the largest overall reaction.
3. Relevant-irrelevant method - There was no numerical scoring and each
relevant question was evaluated in terms of size and consistency of response
without reference to control questions. For all methods, if the largest response
on the chart occurred to one of the symptomatic questions, the subject was
classified inconclusive.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4.
Results
The first set of. analyses used as data the examiners' classification of the
subjects as deceptive or not deceptive irrespective of whether the examiner had
identified accurately the deceptive question. The following results were obtained
with the zone method:
Actual
Deception No Deception
Deception. 21 5
Predicted No Deception 5 16
Inconclusive 4 5
Omitting inconclusives, 81% of the guilty and 76% of the innocent were correctly
classified.
The following results were obtained with the greatest control method:
Actual
Deception No Deception
. Deception 15 4
Predicted No Deception 7 20
? Inconclusive 8 2
'
Omitting inconclusives, 68% of the guilty and 83% of the innocent were correctly,
classified.
The following results were obtained with the relevant/irrelevant method:
Actual
Deception No Deception
Deception 24 - 6
Predicted No Deception 4 19
Inconclusive 2 1
Omitting inconclusives, 86% of the guilty and 76% of the innocent were correctly
classified. All three methods were able to identify nondeceptive subjects
significantly greater than chance. Zone comparison and relevant/irrelevant were
able to identify non?deceptive subjects greater than chance.
Results also were analyzed based on individual questions. There were 250
truthful questions and 30 deceptive questions. Nondeceptive questions were
identified significantly greater than chance by all methods (range 91% to 97%
correct classifications, omitting inconclusives). The relevant/irrelevant method
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
was the only method able to identify deceptive questions significantly greater than
chance (range 54% to 69% correct classifications, omitting inconclusives). Data
also are provided on the stimulus test. Subjects received more than one chart on
this test if the examiner's first or second choice was not the one seledted by the
subject. Classification accuracy (global assessment of deception/no deception) was
significantly related to number of charts on the stimulus test:
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Gordon H. Berland and David C. Raskin. "An Evaluation of Field Techniques in
Detection of-Deception." Psychophysiology, 12(3)(1975): 321-330.
Procedure
Seventy-two college students (sex not specified) were subjects in an experiment
to evaluate detection of deception field techniques. Subjects were randomly
assigned to guilty and innocent conditions. Guilty subjects were required to steal
$10 from a desk drawer and could keep the money if they deceived the examiner.
Innocent subjects were told the details of the crime and that failure to appear
innocent reflected lack of cooperation and they may not get course credit for
participating.
After a pretest interview involving collection of background information
explanation of the rationale of lie detection, and at least two reviews of the
questions on the test, subjects received three presentations of ten questions. A
control - question technique (the Federal Government modification of the zone
comparison polygraph test) was used. Between charts one and two, a card test was
administered with some subjects getting positive feedback, others negative
feedback, and others no feedback. The author noted that this manipulation had no
effect and did not present the results. The polygraph examiner (experience not
discussed'), blind to subject status, scored the tests immediately by semi-objective
scoring using the U.S. Army Military Police School criteria. Each of the
control/relevant question pairs was scored ftom -3 to +3. A score of +5 or greater
was considered truthful, -5-or lower guilty, and between +4 inconclusive. The
polygraph records also were scored by five additional examiners, blind to subject
status with intensive experience who had graduated from the U.S. Army Military
Police School. An objective quantitative scoring also was conducted by people
unaware of the study and unfamiliar with field techniques using more physiological
measures.
Cardiovascular activity, skin resistance, and thoracic or abdominal respiration
were recorded with a field model polygraph. The authors note that the recording of
skin resistance and cardiovascular activity were not of the quality obtained from
newer field polygraphs.
Results
The following results were obtained with the semi-objective scoring procedure:
Actual
Guilty . Innocent
Guilty 23 6
3 15
Predicted
Innocent
Of the seventy-two subjects, 53% were correctly classified, 12% incorrectly
classified, and 35% inconclusives. (The inconclusives are omitted from the above
table because the authors did not report their actual status). Omitting
inconclusives, 81% were correctly categorized. Analysis of variance on the
semi-objective scores indicated that all physiological measures discriminated
between guilt and innocence but skin resistance was superior.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
There was considerable agreement between the five additional examiners and the
original examiner. Percentage of correct classifications ranged from 79% to 86%
(mean = E1.7%) for the six examiners. Thus, the blind examiners were similar to
the original examiner. Correlations between pairing of examiners for total
semi-objective numerical scores ranged from .78 to .95 (mean = .86). Of the 1,080
pairwise comparison, the examiners agreed on 66.3%. There were only 25 instances
(2.3%) where the examiners reached opposite conclusions.
Data on the effect of changing the inconclusive range and the objective
quantitative analysis also are provided.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
M. T. Bradley and Michel Pierre Janisse. "Accuracy Demonstrations, Threat, and
the Detection of Deception: Cardiovascular, Electrodermal, and Pupillary
Measures." Psychophysiology (1981)_18(3): 307-315.
Procedure
One hundred ninety-two male college students were subjects in a study to
determine the effect of several variables on detection of deception. Guilty
subjects participated in a mock crime involving stealing and hiding an envelope
containing $1.00. All subject were told that they were accused of the crime and
were to deny involvement. One-half of the innocent and one-half of the guilty -
subjects were told they would receive a painful electric shock (the shock was not
administered) if they were judged guilty.
Pupil size, heart rate, and skin resistance were recorded. The polygraph
examiner was blind to subject status. The experience of the polygraph examiner is
not discussed. Subjects were given both a control-question test and a guilty
knowledge test about the crime (tests were counterbalanced). Prior to the
polygraph test, the questions on the control-question test were reviewed with the
subject. Subjects were allowed to discuss the questions and eliminate ambiguities
about the contxol questions. The control-question-test consisted of nine questions
(three relevant pairs) presented three times. Questions on the guilty knowledge
test were not reviewed with the subject prior to the polygraph test. This test
consisted of four, five-item question sequences plus a buffer question for each
sequence; the test was presented once to the subject.
Before the polygraph test, a series of rigged card tests with three trials were
tested. Subjects were given false feedback about the number of times the card they
had selected had been detected. Subjects were falsely told that from 0 to 3 cards
had been detected.
Each control-relevant pair from the control-question test was scored 1, 0, or
-1 and summed across questions. A score of +2 or greater reflected innocence, -2
or less reflected guilt; scores between +2 and -2 were classified as inconclusive.
On the guilty knowledge test, responses to each of the four sequences were scored
2, 1, or 0 based on the relative magnitude of the response to the critical item.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Results
The following table demonstrates frequency of detection for each physiological
measure on each test:
Actual
Control Question
Skin Resistance
Test
Test
Guilty Knowledge
Guilty
Innocent
Guilty Innocent
Guilty
58
9
57
11
Innocent
13
56
39
85
Inconclusive
25
31
Predicted
'Pupil Size Change
Guilty
33
30
32
19
Innocent
21
25
64
77
Inconclusive
42
41
Heart Rate Change
Guilty
34
19
43
17
Innocent
20
32
53
79
Inconclusive
42
45
Skin resistance was the most efficacious measure in separating guilty and innocent
people. With skin resistance, 82% of the guilty subjects and 86% of the innocent
subjects were correctly detected (excludig inconclusives) on the control-question
test. With the guilty knowledge test, 59% of the guilty subjects and 89% of the
Innocent subjects were correctly classified using skin resistance. Detection
accuracy results for a combination of measures are not presented.
Analysis of variance on the numerical scores demonstrated that, for the
control-question test, punishment influenced the change in heart rate measure and
manipulated effectiveness was directly related to detection detectability with skin
resistance. In the guilty knowledge test, punishment did not affect any of the
dependent measures, but there were significant effects for manipulated
effectiveness with skin resistance.
The authors suggest that the detection accuracy with the control-question test
and guilty knowledge test should not be directly compared because they used a weak
version of the guilty knowledge test (few items of information) and they used less
than optimal procedures for the control-question test. That is, for the
control-question test, they did not repeat trials for inconclusive subjects, they
did not combine physiological measures for detection and their semi-objective
scoring system had a limited range.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
-
Robert J. Gatchel, John E. Smith, and Norman M. Kaplan. "The Effect of
Propranolol on Polygraphic Detection of Deception." (1983) Unpublished Manuscript,
University of Texas Health Sciences Center.
Procedure'
Twenty-eight males aged 20 to 40 years were paid t15 for participating in a
study to determine the effect of propranolol on detection of deception in a mock
crime situation. Subjects were randomly assigned to one of four groups (seven
subjects per group); (1) Guilty - propranolol, (2) Guilty - no propranolol; (3)
Innocent - propranolol, and (4) Innocent - no propranolol. The guilty subjects
were required to steal and then hide a piece of jewelry from an office in one
hour. Innocent subjects were told that a mock crime had taken place but were not
told details. All subjects were told to deny involvement in the crime. Both
innocent and guilty subjects could receive a t10 bonus for appearibg truthful on
the test. Drug subject received a 40 mg. tablet of propranolol one to two hours
before the polygraph test.
Blood pressure, electrodermal activity, and abdominal and thoracic respiration
were recorded. Two professional polygraph examiners (each testing 14 subjects),
blind to subject's assigned group status, administered a nine-item control question
polygraph test. Detection was based on pairs of control-relevant questions (the
last six questions). Each pair was scored (by the examiner) from +3 to -3 based on
the size of the relative difference. Respiration was scored based on the clearest
channel. Scores were summed across pairs and channels. Subjects with score +6 or
higher were designated truthful, 76 or lower were designated as deceptive, and
inconclusive between +5. Examiners also rated on a seven-point scale whether they
thought the subject had taken a drug.
Results
Results are listed in the following table:
Actual
Guilty: Innocent: Guilty: Innocent:
No Drug No Drug Drug Drug
Guilty
3
0
4
0
Predicted
Innocent
1
4
0
7
Inconclusive
3
3
3
0
-
Analysis of variance revealed a significant guilt-innocence effect, but no
significant drug effects or interactions. Examiners were correct about subject's
drug status in 71% of the cases; however, they were not very confident about their
ratings. The authors note that their study has more inconclusives than other
studies. They suggest that this reflects conservative behavior on the part of the
examiners since the examiner knew that drugs were being administered.
? Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
David L. Hammond. "The Responding of Normals, Alcoholics, and Psychopaths in a
Laboratory Lie-Detection Experiment." Unpublished Doctoral Dissertation,
California School of Professional Psychology, 1980.
Procedure
Normal college students (11 innocent and 10 guilty), alcoholics (10 innocent
and 10 guilty) and psychopaths (9 innocent and 12 guilty) participated in the
study. All subjects were male, aged 21 to 55 years, and paid $7.00 for
participating. Guilty subjects participated in a mock crime involving thefts of
$10.00 and were told not to confess. Instructions were presented by tape and were
designed to motivate the subjects. Innocent subjects listened to the same tape
and, thus, knew the general nature of the crime. All subjects could receive an
additional $10.00 for a truthful outcome on the test. Alcoholics had previous
treatment for alcoholism and met certain diagnostic criterion. Psychopaths had a
prior prison record and met other diagnostic criteria. Respiration, electrodermal
activity, and cardiovascular activity were recorded. Inexperienced examiners from
the Backster School of Lie Detection in the fifth and sixth weeks of their
seven-week training session conducted the tests. The Backster "You Phase"
control-question procedure was used. The test was scored with the zone-comparison
method. Testing involved a minimum of three charts, but in most cases two charts
were used for analysis. Numerical scores greater than 8 were considered truthful,
less than 8 were considered deceptive, and between +8 and -8 were inconclusive. An
expert examiner also independently evaluated the polygraph charts.
Results
The following results were obtained for the student examiners:
Actual
Guilty
Guilty
Innocent
Normal
6
3
Alcoholic
9
1
Psychopath
8
2
Predicted
Innocent '
By Student
Normal
1
4
Examiners
Alcoholic
0
5
Psychopath
0
3
Inconclusive
Normal
3
4
Alcoholic
1
4
Psychopath
4
/
4
Omitting inconclusives, 86% of the normal guilty, 100% of the alcoholic guilty, and
100% of the psychopathic guilty were correctly classified by the student
examiners. Omitting inconclusives, 57% of the normal innocent, 83% of the
alcoholic innocent, and 60% of the psychopathic innocent were correctly
classified.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
The following results were obtained for the expert examiner:
Guilty
Actual
Innocent
Guilty
Normal
4
2
Alcoholic
8
0
Psychopath
9
0
Predicted.
Innocent
By Expert
Normal
0
0
Examiner
Alcoholic
0
2
PSychopath
0
2
Inconclusive
Normal
6
9
Alcoholic
2
8
Psychopath
3
7
Omitting inconclusives, 100% of the guilty subjects in each condition were
correctly classified by the expert examiner. Omitting inconclusives, the expert
examiner correctly classified 0% of the normal innocents, and 100% of the alcoholic
and psychopathic innocents. Data also are presented using other inconclusive
ranges.
Analysis of variance on the numerical scores revealed no significant effects
for examiner experience, personality, or physiological measures. Student
examiners' numerical scores and expert examiner's numerical scores were correlated
.64 for the innocent condition and .61 for the guilty condition.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Charles Robert Honts. "The Effects of Simple Physical Countermeasures on the
Physiological Detection of Deception." Unpublished Master's Thesis, Virginia
Polytechnic Institute and State University, Blacksburg, Virginia, May, 1982. [Also
presented at the Society for Psychophysiological Research, Minneapolis, Minnesota,
October 1982, with Robert L. Hodes.]
Procedure
Forty-eight students (males and females, mean age 18.8 years) were participants
in a study to determine-the effect of physical countermeasures on the detection of
deception, using a control question test. Subject's were randomly assigned to one
of four groups (12 subjects per group); (1) Innocent, (2) Guilty Controls - no
countermeasure, (3) Guilty - pair countermeasure, and (4) Guilty - muscle tension
countermeasure. All guilty subjects participated in a mock crime which involved
stealing an examination. The guilty control subjects were told that it may be
possible to deceive the polygraph examiner. Guilty-pain countermeasure subjects
were given training on biting the tongue at particular points during the control
question test and were told to relax during relevant questions. Guilty-muscle
tension subjects received training on pressing their toes against the floor and
tensing their muscles at critical points during the test and also were told to
relax during relevant questions. Innocent subjects knew the general nature of the
mock crime, but were naive to its details. They were told that if they were honest
the examiner would find them truthful. All subjects were told they would get extra
points applied to their final course grades if they produced a truthful outcome on
the polygraph.
All subjects received two experimental sessions. During the first session,
subjects were assigned to a group and received training if they were to use
countermeasures. The second session (within one week of the first) was the
polygraph exam. All guilty subjects were encouraged to appear innocent to receive
bonus points. Subjects were cautioned not to be detected using countermeasures or
they would lose bonus points.
Thoracic or abdominal respiration, skin resistance, and cardiovascular activity
were recorded. A licensed examiner with five years of field experience trained at
the Backster School of Lie Detection, administered the polygraph tests. The
examiner did not know the guilt/innocence status of the subjects, but did know the
proportions of guilt and innocent subjects and did know that muscle or pair
countermeasures would be used. Subjects received four presentations of eight
questions.
Immediately after the polygraph test, the examiner was required to make an
innocent/guilt/inconclusive decision based on limited chart analysis (less than
five minutes) and to judge whether countermeasures were used. Two weeks after the
last subject was tested, the examiner scored the charts by semi-objective methods
and judged whether countermeasures were used from the physiological data. Data
were sent to an experienced evaluator for a blind evaluation using the
semi-objective method. An objective, quantitative analysis of the data also
occurred. The semi-objective analysis used the?Backster Spot Analysis Technique in
which pairs of responses were rated from -3 to +3 (based on the relative magnitude
of relevant and associated control question) and summed. Post examination
debriefings indicated that five subjects in the countermeasure groups failed to
employ the designated
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
countermeasures and three subjects in the guilty-control group tried to use
countermeasures. Data were analyzed with regard to subject's debriefing reports.
? Results
The original examiner was not able to detect countermeasure usage either from a
subject's behavior or from the polygraph charts. The blind rescorer also could not
detect countermeasure usage.
The ability of the orignal examiner to correctly classify subjects based on a
limited chart analysis is reflected in the following table (numbers in parentheses
are those subjects that did not comply with the group's instruction).
? Actual
Guilty: Control Innocent .Guilty,: Pain Guilty: Muscle
Guilty 5(1) 4 5(3) 7(2)
Predicted Innocent ? 1 6 1 1
Inconclusive 3(2) 2 3 2
Results for semi-objective analysis (+12 or greater indicating innocence, -12
or lower deception, and between +12 and -12 inconclusive based on the three most
productive charts) are in the following table:
Actual
Guilty: Control Innocent Guilty: Pain Guilty: Muscle
Guilty 6(1) 2 5(3) 6
Predicted Innocent 0 4 1 0
Inconclusive 3(2) 6 3 4(2)
Excluding inconclusives, the examiner was correct 89% of the time. Countermeasures
did not affect the results of either the limited chart analysis or the
semi-objective analysis. Analysis of variance for semi-objective scores from four
charts (with the five noncompliant subjects is the countermeasure groups removed)
also did not reveal a countermeasure effect. Objective quantitative analysis of
each physiological measure found significant countermeasure effects for finger
pulse amplitude, the results for respiration amplitude approached significance.
Semi-objective scores of the original examiner were significantly correlated (r
= .88) with those of the blind evaluator. When there was a decision of truthful or
deception, there was 100% agreement between the raters.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Raskin, D. C., and Hare, R. D. Psychopathy and Detection of Deception In a
Prison Population. Psychophysiology, 15 (1978); 126-136
Procedure
Twenty-four psychopaths (mean age = 23.1 years) and 24 non-psychopaths (mean
age = 26.5 years), from an inmate population, participated in the study. Twelve
psychopaths and 12 non-psychopaths were innocent and 12 psychopaths and 12
non-psychopaths were guilty of a mock crime involving the theft of $20 from an
adjacent room. Subjects were then given a polygraph test about the crime and could
keep the $20 if their test performance was successful. Guilty subjects were to
deny involvement in the crime; innocent subjects were told about the crime and
instructed to tell the truth. Psychopaths were selected by an experienced
researcher based on an assessment of their behavior over a long period. The
conception of psychopathy was consistent with that of Clecklez (1964).
Subjects were administered a control question polygraph test by an examiner
trained in field polygraph techniques blind to subject status. Control questions
were developed from a pretest interview. Relevant questions were constant for all
subjects; control questions differed. Prior to the 10 questions, subjects were
administered a stimulation test to demonstrate technique effectiveness. The
polygraph test consisted of three charts; if results were not obvious, additional
charts were run to a maximum of seven. In such situations, the fourth chart used a
silent answer procedure. The examiner continually stressed the importance of being
truthful and directed the attention of innocent subjects to control questions and
guilty subjects to relevant questions.
Thoracic and abdominal respiration, skin conductance, skin potential, heart
rate, finger blood volume, and finger pulse amplitude were recorded. Data were
evaluated in two ways. The first was by the examiner using responses previous
research had demonstrated significantly discriminated between guilt and innocence.
Each control and relevant question pair were compared, scored from +3 to -3,
and summed; +6 or higher indicated innocence; -6 or lower indicated guilt. Results
using all charts for each subject are presented in the following table?
Actual
Guilty Innocent
Guilty - (Psychopath) 12 1
(Nonpsychopath) 9 1
Examiner Innocent - (Psychopath) 0 11
Decision (Nonpsychopath) 0 10
Inconclusive - (Psychopath) 0 0
(Nonpsychopath) 3
Thus, 88% were correctly classified, 4% incorrectly classified, and 8% were
inconclusive. Omitting inconclusives 96% were correctly classified. Psychopaths
and nonpsychopaths did not significantly differ in classification accuracy.
Comparison of classification based only on three charts per subject with
classification using additional charts when necessary demonstrated a reduction in
inconclusives with more charts, but similar correct classification rates. Optimal
boundary was suggested to be in the range of +2 to +4. Analysis of variance and
post-hoc comparisons on numerical scores revealed that skin conductance was the
best guilt/innocence discriminator and the cardiovascular measures the poorest,
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A
although all measures significantly discriminated between the groups. Respiration
was better at identifying innocent subjects than guilty subjects. Effects of
psychopathy on the individual measures were minimal. Results of a more
quantitative analysis using additional physiological measures scored by a
technician with no knowledge of the experiment also are reported. Characteristics
of individual physiological measures associated with guilt and innocence are
discussed. -
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
L. I. Rovner, D. C. Raskin and J. C. Kircher. "Effects of Information and Practice
on Detection of Deception." Presented at the Society for Psychophysiological
Research, Madison, 1978.
Procedure
Seventy-two males participated in a stuey to determine the effect of
information and practice on detection of deception. The 36 guilty subjects
participated in a mock crime (stealing a ring). The 36 innocent subjects were told
a theft had occurred. Prior to their polygraph test, subjects were divided into
the following groups with 12 innocent and 12 guilty subjects in each group: (1)
STD - Subjects merely waited in a room before their polygraph test, (2) rffF0 -
Subjects were given information about the nature of the polygraph and how to appear
innocent, and (3) ram + Practice - Subjects were given the same information as
group two, and two polygraph practice sessions with feedback about their
performance. Examiners, (not professional polygraph examiners), blind to a
subject's guilt or innocence, administered a control question polygraph test.
Subjects were paid $7.50 for participating in the study and il0 if the outcome of
their test was truthful. All subjects were to dehy involvement in the theft. Skin
conductance, blood pressure, respiration, and digital vasomotor activity were
recorded; there were three charts per subject. Responses were scored with a
numerical scoring system and an inconclusive range of +5. Results are based on the
scoring of a second examiner blind to test results and subject status.
Results
In both the STD and INFO groups, 88% were correctly classified, 4% were
incorrectly classified, and 8% were inconclusive. Omitting inconclusives 95% were
correctly classified. The one incorrect classification in the STD group and the 1
incorrect classification in the INFO group were false positives. The actual guilt
or innocence of inconclusive subjects was not reported. INFO + Practice reduced
the nrmber of correct classifications and increased the number of false positives
and negatives. That is, 62.5% were correctly classified, 25% were incorrectly
classified (three false positives and three false negatives), and 12.5% were
inconclusive. Analysis of variance and post-hoc comparisons revealed that the
innocent subjects in the INFO + Practice group had significantly lower scores than
the other two groups, whereas the three guilty groups did not significantly
differ. The authors note that the lack of time and similarity between practice and
the actual polygraph test may be important to observation of the effect.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
William M. Waid, Emily Carota Orne, and Martin T. Orne. "Selective Memory for
Social Information, Alertness, and Physiological Arousal in the Detection of
Deception." Journal of Applied Psychology (1981) 66(2): 224-232.
Procedure
Thirty-four designated innocent and 40 designated guilty subjects (males-mean
age 18 to 28 years) participated in a study to determine the relationship between
detection of deception and memory of questions and words involved in the polygraph
test. Guilty subjects overlearned 6 code words during timed interpolated tasks.
Innocent subjects performed the same tasks but learned no words. InnocInt subjects
were told it was often difficult to prove one's innocence on a polygraph test.
Guilty subjects were told that highly mature and intelligent people could deceive
the examiner. Care was taken to involve innocent subjects in the procedures.
Subjects were tested by a professional examiner blind to subject status; skin
conductance, blood pressure, and respiration were recorded. Subjects were
administered three guilty person tests and a guilty knowledge test. The guilty
person tests consisted of 9 relevant questions, 8 control questions, and 5
irrelevant questions. A card test was administered between the first and second
tests and subjects were. questioned about their answers. In the guilty knowledge
test, subjects were asked if each of 24 words had special meaning. The list was
presented twice. Subjects were then asked to write down as many words as they
could remember from the guilty knowledge test and as many questions as possible
from the guilty person test.
Results
Different methods of quantifying the data were used to determine detection
accuracy for the guilty person tests. .One method averaged response to relevant and
control questions. A subject was considered deceptive if average response to the
relevant questions was larger than average response to control questions. The
following results were obtained when skin conductance was used.
Predicted
Guilty
Innocent
Actual
Guilty Innocent
29 8
11 26
72.5% of the guilty subjects and 76.5% of the innocent subjects were correctly
classified.
When the detection criterion was showing a larger skin conductance response to
relevant than control questions on 7 or more questions, the following results were
obtained:
Predicted
Actual
Guilty Innocent
Guilty 22 8
Innocent 18 26
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
4
Fifty-five per cent of the guilty subjects and 76.5% of the innocent subjects were
correctly detected. Detection accuracy was much lower with respiration and with
blood pressure.
There were results for memory of test items. Analysis of variance indicated
that there were larger skin conductance responses to relevant and control questions
that were subsequently recalled than to relevant and control questions not
subsequently recalled on the first and second guilty person tests. On the third
guilty person test subsequently recalled control questions elicited larger
responses than control questions not subsequently recalled, but this was not true
for relevant questions: On certain tests, more guilty subjects were called
deceptive when skin conductance responses were larger to relevant than control
questions, and more innocent subjects were called not deceptive when responses were
larger to control questions than to relevant questions.
There was less of a relationship between recall of items and detection on the
guilty knowledge test than with the guilty person test. There were no
relationships between recall and blood pressure or respiration. Detection accuracy
with the guilty knowledge test was not reported.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
William M. Waid, Martin T. Orne, and Stuart K. Wilson. "Effects of Level of
Socialization on Electrodermal Detection of Deception." Psychophysiology 16 (1)
(1979): 15-22.
Procedure
Thirty male college students (aged 18 to 28 years ) participated in an
experiment to determine the effect of socialization on detection of deception. The
fifteen subjects designated guilty overlearned 6 "code words" during a one hour
procedure of time interpolated tasks. The fifteen innocent subjects performed the
same tasks but learned no "code words." The experimenters emphasize that care was
taken to involve both the innocent and guilty subjects in the task. Guilty
subjects were told that highly intelligent, mature individuals could escape'
detection.
A professional polygraph examiner with no knowlege of a subject's status or the
code words administered several types of polygraph tests. The six "code words"
differed for the subjects to prevent bias in the examiner. Skin conductance was
recorded. The following tests were administered (always in the same order) after a
pretest interview to obtain personal history and a review of the relevant questions
on the first test: a) Guilty Person Test - Subject was asked three questions about
his guilt or innocence and two interpolated control questions. The subject was
questioned about his answers, presented a stimulus test, and then a second Guilty
Person Test. The first and second Guilty Person Tests differed in that the former
used unreviewed control questions, whereas the later reviewed control questions
with the subject; b) Peak-of-Tension Test - Subjects were asked how many correct
code words existed. Two series of numbers were presented one with the sequence in
ascending order and one with the .sequence in descending order; c) Guilty Knowledge
Test - Subjects were asked if each of 24 words had special meaning to them. Six of
the words were code words. This list was repeated four times.
Subject then completed the socialization scale of the California Psychological
Inventory and other questionnaires. Criteria for detection differed with type of
test. On the Guilty Person Tests, subjects were classified as deceptive if the
response to any of the three critical questions was larger than the largest
response to the control questions. Deception was indicated on Peak-of-Tension if
the largest response of the series was to the critical number (omitting the first
number). For the Guilty Knowledge test a code word was considered detected if it
elicited a larger response than the three code words in the same category.
Subjects having code words detected seven or more times were considered deceptive.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Results
Results are in the following table:
Guilty Person 1
Actual
Innocent
Guilty
Guilty
12
3
Innocent
3
12
Guilty Person 2
Predicted
Guilty
12
3
Innocent
3
12
Peak of Tension
Guilty
7
3
Innocent
8
12
Guilty Knowledge
Guilty
8
1
Innocent
7
14
There were significant effects for socialization. Decpetive subjects not
detected scored aignificantly lower on the socialization scale than detected
deceptives for most of the tests. Among guilty subjects socialization was
significantly positively correlated with conductance response magnitude (mean of
four tests). Misclassified innocents had higher socialization scores than other
innocents. Conductance response magnitude was significantly positively correlated
with socialization in innocent subjects. It should be noted that the effectiveness
of the different polygraph tests can not be determined in this study, since the
tests were administered in the same order for all subjects.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
:R
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Jan Widacki and Frank Horvath. "An.Experimental Investigation of the Relative
Validity and Utility of the Polygraph Technique and Three Other Common Methods of
Criminal Identification." Journal of Forensic Science 23 (3) (July 1978): 596-601.
Procedure
Subjects were 80 student volunteers (42 males, 38 females; aged 19 to 24 years)
at Jagolian University (Poland). Each subject provided the expetimenter with
hand-writing specimens, fingerprints and full-face photographs. Subjects were
divided into 20 groups of four each. In each group there were three innocent
persons and one perpetrator. Each perpetrator was involved in an independent, but
similar case. The perpetrator was instructed to take an envelope to a specific
building and give the envelope to a specific person (whom they did not know).
After the envelope was delivered, the perpetrator received a parcel containing a
cosmetic of small value and signed a receipt form with a fictitious name. They
were to try to disguise their handwriting.
All subjects were then given a polygraph examination by an examiner blind to
their guilt or innocence. A four-channel Lafayette polygraph and Reid Control
Question procedures were used. Respiration, cardiovascular activity and
electrodermal activity were recorded. Examiners made subjective decisions as to a
subjects guilt or innocence. Examiners knew that each group of four subjects was
involved in the same case and tested all subjects in a group before selecting the
perpetrator. The experience of the polygraph examiner is not discussed.
Independent of the polygraph test, three other types of evidence were collected
including handwriting analysis of the receipt by a handwriting expert, fingerprint
analysis (from fingerprints obtained from the envelope and instruction sheet) by a
fingerprint expert, and eyewitness identification by the person receiving the
envelope (those people had prior knowledge that they would be required to m4e the
identification). The handwriting expert, the eyewitness, and the fingerprintl
expert were to select the perpetrator from each group of four.
Results will be presented by cases instead of individuals since the data from
subjects involved in each case was not independent.
Correct
Error
Inconclusive
Acct?ar-EL
Polygraph
18
1
1
95%
Handwriting
17
1
2
94%
Eyewitness
7
4
9
64%
Fingerprints
4
0
16
100%
Fingerprinting was the most accurate method, but yielded many inconclusive
results. Polygraph and-handwriting had high accuracy and few inconclusives.
Eyewitness identification was neither accurate nor useful. It should be noted that
the fingerprints obtained in many cases were not of high quality. The authors note
that the forced choice procedure used in the study would be expected to increase
the detection rates and caution against overgeneralization.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Laboratory Abstracts
Relevant-Irrelevant Techniques
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Richard H. Blum and William Osterloh. "The Polygraph Examination as a Means
for Detecting Truth and Falsehood in Stories Presented by Police Informants."
Journal of Criminal Law, Criminology and Police Science 59 (1968): 133-137.
Procedure
A total of 17 male and 3 female informants who had given information on
criminal cases to local or federal agencies were selected and paid to engage in the
experiment. Great care was taken to protect their identities and the fact they
were cooperating with the police, as their lives would otherwise be in danger.
Some of the stories that the informants told the examiner were completely true
(n=9), and some were completely false (n=4). Some were partly true and partly
false (n=7). (This is typical of informants, and similar to the problem in
preemployment screening.) The topics about which the informants gave information
covered a wide variety of crimes. The true stores were those which had been
provided to their police handler on a previous occasion which had stood the
scrutiny of investigation. A false story was one jointly invented by the officer
and the informant, but compatible with the informant's ordinary role and
opportunities, and containing credible information. True stores with false items
were true stores told in the past (which had been confirmed by investigation) in
which some critical items had been made false such as the name of the offender, the
place where the goods were hidden or fenced, etc. Each story was first written,
then rehearsed with the police handler. The examiner had no idea as to which
examinee was telling the truth, which was lying all together, and which was only
partly truthful. He also was blind as to the ratio of truth telling and lying.
Seventeen stories contained 5 relevant questions; the remaining 3 stories contained
6, 7 and 8 relevant questions.
-Three professional polygraph examiners used a standard three-channel Stoelting
polygraph and the relevant-irrelevant technique. They were forced to make a
decision on all of the 106 relevant questions from 20 subjects. Inconclusive
decisions and reexaminations were not permitted as might occur in regular
practice. Subjects were instructed not to make admissions during the polygraph
examination. The physiological measures recorded are not discussed. Examiners
made a subjective judgment of truth or deception on each critical item of each
story.
Results
Examiners were able to identify stories as totally true, totally false, or
stories containing some false information (irrespective of the particular question)
as illustrated in the following table:
Totally False
ACTUAL
Totally True Partially False
Totally False
4
0
0
Predicted
Totally True
0
9
0
Partially False
0
0
7
However, the examiners were not always able to correctly classify individual
relevant questions of informants telling partially true stories. They were
incorrect about 4 questions involving 3 subjects. Thus, 102 of 106 questions were
correctly classified.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Eileen J. Correa and Henry E. Adams. "The Validity of the Preemployment
Polygraph Examination and the Effects of Motivation." Polygraph 10 (3) (September
1981): 143-155. (Also-Eileen Joyce Israel. Unpublished Doctoral Dissertation,
University of Georgia, Athens, Georgia, 1977.)
Procedure
Twenty male and 20 female undergraduates at the University of Georgia
participated in a laboratory study to determine the validity of a preemployment
polygraph test. An equal number of males and females were randomly assigned to
four groups: (1) lying with motivation, (2) lying without motivation, (3) truthful
with motivation, and (4) truthful without motivation. .
Each subject received two sessions.
subjects filled out a Preemployment Data
lie questions were identified. Three of
lies) were included in case subjects had
Preemployment Data Sheet. These control
the initial screening session.
During the initial screening session,
Sheet with an experimenter; nine potential
the nine potential lie questions (control
initially, lied while filling out the
lies involved activities occurring during
Subjects in the lying groups were to lie to the nine questions; subjects in the
truthful groups were to answer all questions truthfully. Subjects in the motivated
groups were told that the best liar in the lying group and the best truth teller in
the truthful group would get a $25 reward and that the award selection Would be
based on the polygraph charts.
The polygraph test used.a relevant/irrelevant technique. The test consisted of
three series of questions about the Preemployment Data Sheet. Bach series
contained three potential lie questions (one control lie in each series). Each
question from the Preemployment Data Sheet either preceded or followed an .
irrelevant norm question. Between charts the examiner questioned subjects about
any responses that appeared to be larger to the relevant questions than the
irrelevant questions. These questions were rephrased and inserted at the beginning
of the next series. The examiner (experience not discussed ) made a subjective
decision of lying or truthful for each subject and for each question. The examiner
was blind as to subject status. Respiration (thermister inside nostril), heart
rate, and skin resistance were recorded.
Results
The following table presents the data on the examiner's overall judgement of
the subject:
Predicted
Lying
Truthful
Actual
.Lying Truthful
Motivated Unmotivated Motivated Unmotivated
10 10
The examiner correctly classified all of the subjects.
10 10
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
The individual lie questions also were analyzed. Three of the nine lies were
control lies; the remaining six lies were from the Preemployment Data Sheet. All
of the control lies were correctly identified; there were some errors in the
identification of the other lies as indicated in the following table:
Correct
Incorrect
Control Lie
Lying-Motivated
30
0
Questions
Lying-Unmotivated
30
0
Data Sheet
Lying-Motivated
49 (82%)
11 (18%)
Lies
Lying-Unmotivated
41 (68%)
19 (32%)
The difference
between the motivated and unmotivated groups was not significant.
Objective numerical analyses of the data also were computed. There was no
overall motivation effect, but there were some interactions between motivation and
other independent variables. Lying responses were accompanied by increases in skin
conductance level and large decreases in heart rate, whereas truthful responses
were accompanied by decreases in. skin conductance level and small decreases or
increases in heart rate.
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Laboratory Abstracts
Guilty Knowledge
and ,
Peak of Tension
Techniques
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Kristen D. Balloun and David S. Holmes. "Effects of Repeated Examinations on
the Ability to Detect Guilt with a Polygraphic Examination: A Laboratory
Experience with a Real Crime." Journal of Applied Psychology (1979) 64(3):
316-322.
Procedure
Eighteen male college students with high Pd (psychopathic deviate scores on the
Minnesota Multiphasic Personality Inventory) and 16 subjects with low Pd scores
participated in a study to determine the effect of repeated examinations on
detection of deception. Each subject was tested with two confederates. Subjects
were told that the purpose of the experiment was to determine whether they did
better on a bogus intelligence-information test when working alone or in groups,
and that they would work in groups. Subjects were told it was important to do well
and if their performance was poor they would have to meet with a panel of
psychologists. When the experimenter left the two confederates urged the subject
to cheat. One of the confederates said a friend had previously taken the test and
produced a copy of the answer sheet. About 50% of the subjects actually cheated.
Subjects were taken to the laboratory and told that it had been learned that
some had cheated. They were administered the guilty knowledge test after the test
procedures were explained. The test consisted of two series of questions, each
series containing 5 relevant questions, and each of those 5 relevants had 5
alternatives. The relevant questions contained information that only guilty
subjects could have known from the answer sheet provided by the confederate. The
first alternative to each question was a buffer. Heart rate, skin resistance, and
finger pulse volume were recorded. If the critical alternative to each question
elicited the largest response it was scored 2; if the .critical item elicited the .
second largest response it was. scored 1; all other possibilities were scored 0.
Scores were summed acrbss questions. Subjects with scores greater than 5.5 were
called cheaters; those with scores less than 5.5 were called noncheaters.
Results
Results for skin resistance are in the following table:
Actual
Predicted
Series 1 Series 2
Cheaters 'Noncheaters Cheaters Noncheaters
Cheaters 11 2 3 1
Noncheaters 7 14 15 15
On series 1, 61% of the cheaters and 87.5% of the noncheaters were correctly
identified. On series 2, 17% of the cheaters and 94% of the noncheaters were
correctly identified. There was a significant reduction in the detection of
cheaters from series 1 to series 2.
Analysis of variance on the scores for skin resistance revealed significantly
higher scores on test I than test 2 and significantly higher sores for cheaters
than noncheaters. There were no significant effects or interactions for Pd
(psychopathic deviance). Heart rate and finger pulse volume were not effective in
detecting cheating.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?t
P.O. Davidson. "Validity of the Guilty - Knowledge Technique: The Effects of
Motivation," Journal of Applied Psychology 52(1) (1968): 62-65.
Procedure
Forty-eight college students (sex not specified) were divided into 12 groups to
determine the effects of motivation on the detection of deception. In each group
there were 3 thieves and 1 innocent subject. Each thief was to attempt to steal a
sealed envelope containing a voucher with .a value up to 150.00 from a victim. The
innocent subject had no knowledge that a crime was to be committed. There were 12
victims and 12 crimes, one for each group.
The study was designed to maximize the involvement of the participants. The
thieves worked individually and personally planned the details of their crime.
They were told that there were others after the same victim and the first one
successful would get the envelope. The voucher could be cashed if the thief
deceived the polygraph examiner. There also were to be prizes for the most
ingeniously devised crimes whether successful or not. Motivation was varied by
size of the voucher. Ralf of the victims had vouchers ranging from $25.00 to
150.00 while the other half had vouchers worth 10.10 to $1.00. The thieves did not
know the value of the voucher held by their victim unless the envelope was
obtained. Each thief in each group thought obtaining the envelope was possible.
In reality, the situation for each group was structured so that one thief tried but
failed, one had no opportunity to attempt the crime, and one who committed the
crime. Only the thieves that completed the crime were considered guilty. Thus,
there were 12 guilty an 36 innocent subjects.
Subjects were administered a polygraph test using the guilty-knowledge
technique.' The experience of the polygraph examiner was not discussed. Each test
consisted of 6 multiple-choice questions each with 5 alternatives about the crime.
Subjects were to listen to all questions and make no verbal response. The first
alternative on each question was not scored. The victims had supplied the
examiners with some of the details of the crimes, but the examiner did not know the
correct answer to each question or who was the criminal. The criminal was the only
one who could recognize the correct answer to each question.
Electrodermal activity, respiration, and heart rate were recorded, but only
electrodermal activity was scored. Scoring was done without knowledge of the
correct answer. Each subject's electrodermal responses to each question were
ranked in order of amplitude. If the largest response was to a relevant
alternative the question was scored 2. If the second largest response occurred to
the relevant question it was scored I. Other possibilities were scored O.
Subjects with a score of 6 or less were considered innocent. Subjects with a score
greater than 6 were considered guilty.
The following results were obtained:
4 Predicted
ACTUAL
Guilty Innocent
Guilty (high value) 6 0
Guilty (low value) 5 0
Innocent 1 36
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
1.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
The difference between motivation groups was not significant. The authors
noted that their study used a very involved mock .crime with some of the
uncertainties of real crimes.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
A
Miroslav Dufek, Department of Criminalistics Legal Faculty, Charles University,
Prague. "A Contribution on the Problem of Polygraph Examinations." Czechoslovak
Criminalistics (February 1969). (Tr. from Czech.)
Procedure
Several experiments were conducted investigating detection of deception with
guilty information type tests. The authors state they worked with equipment from
the Alvar firm and recorded electroencephalograms, electromyograms, electrodermal
activity, etc.
Procedure 1
In the first experiment 30 individuals were told to write a random number from
1 to 10 on a piece of paper which they retained during the test. Chance was 10
percent. The experimenter utilized a searching peak of tension test in which the
opening words were always (Czech, the equivalent of) "Did you write the
number...?" The answer was "No."
Results
Using a searching peak of
correct number with certainty
the remaining five cases, the
one of which was correct. In
determined.
tension test, the examiner was able to pick the
in 25 of the examinations (83 percent). In three of .
examiner could not distinguish between two numbers,
two of the examinations, the number could not be
Procedure 2
The second experiment involved 20 men who examined in detail one of six objects
placed before them, then concealed from the examiner the name of the object they
handled (a lighter, cigarette, pencil, pen, eraser, or ashtray). The object was to
be selected at random from the group. The examiner was blind to the item
selected. Chance was 17 percent.
Results
Using a searching peak of tension test, in which the subject did not know the
sequence of questions, the examiner detected the correct object in 18 of 20
examinations (90 percent). There were no reactions on which to make a judgment for
two other subjects, an inconclusive rate of 10 percent.
Procedure 3
Five men and five women were told to select someone with whom they were closely
associated with their personal life, an association of an intimate nature. With
the help of an assistant, another 10 first names were selected which had no
emotional meaning to the subject. The examiner was blind as to the meaningful
name. During the searching peak examination, the subjects denied a relationship
with all of the names. Chance was 9 percent.
yalidity,
The examiner determined the correct name in all ten cases (100 percent).
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
JO'
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Procedure 4
Experiment 4 used the same procedures as Experiment 2. The difference was that
in this series of peak of tension examinations, the test individual knew in advance
the exact order in which the objects would be presented. The purpose was. to
develope progressive stress, as the correct object was approached, and relaxation
following the correct item. Twenty men were tested.
Validity
In these experiments the correct item was selected in 17 of the 20 examinations
(85 percent), while three of the examinations did not yield clear reactions to a
single item. Instead, in these three cases, there were responses to two of the
items, of which one of them was correct.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Jana Kronbergerova and Miroslav Dufek. "Experimental Experiences With the Use
of Polygraph." Socialist Legality (May-June 1969). (Tr. from Czech.)
Procedure
Ten men (aged 21 to 47 years) were instructed to hide a weapon previously used
in a crime of violence in any one of 18 locations. In a subsequent polygraph
examination, questions were structured so that it would be possible to get a
general specification (such as the building)- as well as the floor number and the
room of the hidden weapon. The instrument recorded respiration, electrodermal
responses, and blood pressure. The first and last questions were control questions
(i.e. places that were not possible locations); the remaining questions involved
possible hiding places. The number of questions asked was not stated, and it was
reported that none of the questions were repeated. Subjects were required to deny
the real location. Chance was six percent for detecting the complete location.
The method of scoring the polygraph tests is not reported. Examiners are said to
be experienced and educated in psychology, physiology and other fields. The
authors give insufficient detail about how the polygraph test was conducted.
The experimenters were able to determine the exact location of the weapon in
nine of the ten cases. In the tenth, they were able to determine the building and
floor, but not the correct room. The examinee reacted more strongly to another
room than he did to the room in which he hid the weapon. In retrospect it appeared
that he had a strong personal reason to react to the other room.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
'I
David T. Lykken. "The GSR' in the Detection of Guilt." Journal of Applied
Psychology 43 (6) (1959): 385-388.
Procedure
Forty-nine male college students were ass4gned at random to four groups. Those
in Group 1 (n=13) enacted two mock crimes in random sequence, a "murder" and a
"theft." Both crimes required some rather detailed play-acting. Those in Group 2
(n=12) enacted only the murder, and those in Group 3 (n=12) only the theft. Those
in Group 4 (n=12) were not exposed to either crime. Each subject was told that he
was to be questioned in relation to two crimes. The subject was told that if the
experimenter (experience is not discussed) felt that the subject's electrodermal
response indicated guilt to any question, the subject would receive an electric
shock. An unpleasant shock was then demonstrated. In fact, shocks were given
during testing of all the subjects to heighten general. anxiety. Each test
consisted of two lists of six multiple choice questions - one for the murder and
one for the theft. The experimenter read the question and then read each of
several alternative answers, with adequate time between them for GSR recovery. The
six questions related to details of the crimes known only to the perpetrators
(guilty knowledge). The subject was not to reply to any questions. The position
of the critical item in each list was varied from one subject to another, and the
examiner was in the blind as to which item was critical in each of the lists and
was blind to the subject's group. Electrodermal responses to the alternatives in
each question were ranked based on amplitude. If the largest response occurred to
the relevant alternative it was scored 2; if the second largest response occurred
to the relevant alternative a score of 1 was given; otherwise, a score of 0 was
given. Scores were summed across questions. A subject with a score of 6 or less
.mras classified innocent; a subject with scores over 6 was classified guilty.
Results
Four subjects from Group land one subject from Group 2 were misclassified as
to group. There were six errors An the 98 interrogations (each subject had two
interrogations, one for each crime); all of these errors involved classifying a
subject innocent when he was really guilty. The following table was constructed
from Lykken's presented data:
Actual
Guilty of Guilty -of
Two Crimes One Crime Innocent
Guilty of Two Crimes
9
0
0
Predicted
Guilty of One Crime
3
23
0
Innocent of Both Crimes
1
1
12
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
David T. Lykken. "The Validity of the Guilty Knowledge Technique: The Effects
of Faking." Journal of Applied Psychology 44(4) (1960): 258-262.
Procedure
Twenty subjects (males and females) participated in a study on the validity of
the guilty knowledge technique. Each subject initially completed a 25-item
questionnaire requiring biographical information. Subjects were subsequently given
a test of guilty knowledge using the 25 questions each with 6 alternatives. One
alternative pertained to the subject; 4 alternatives to each question pertained to
other subjects. The first alternative was a randomly selected item ard was not
scored to reduce the effects of orienting. The examiner (experience not discussed)
attempted to identify the subject's set of questionnaire responses.
Electrodermal activity was measured. Each subject was given a 15-minute
lecture on electrodermal activity, lie detection in general, and the principle of
the guilty knowledge method in particular. Each subject was allowed to practice
(with feedback) producing voluntary electrodermal responses by various methods.
The group was told of test format, cautioned not to attempt to defeat the test by
inhibiting responses, and advised that the best way to defeat the test would be to
produce electrodermal responses of various amplitudes to the innocent alternatives
as in random a pattern as possible. Each subject was offered a $10.00 prize if
they could defeat the test.
Electrodermal responses to the 5 alternatives (omitting the first item) were
ranked based on amplitude. Thus, there was a distribution of ranks associated with
each of the 5. questionnaires used for each subject. Each distribution was then
given a .score by giving a'point every time a particular rank occurred more than .5
times. A distribution in which all 25 responses had the same rank would be scored
20. A distribution with each rank occurring 5 times would be scored O. When no
measurable electrodermal response occurred an item was not ranked. The scores for
each distribution were compared.- Patterns of nonrandom responding (higher scores)
were considered indicative of guilty knowledge.
Results
The questionnaire belonging to each subject was correctly identified in all 20
cases. The author notes that this study uses a more comprehensive interrogation
than his previous study (Lykken, 1959).
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Kazuo Ohnishi, Katsunori Matsuno, Masana Arasuna and Akhiro Suzuki. "The
Objective Analysis of Physiological Indices in the Field of Detection of
Deception." Reports of the National Institute of Police Science 29(3) (August
1976): 181-188. Abstract in English.
Procedure
A searching peak of tension test to detect the correct number from six numbers
(unknown to the examiher) was conducted during the polygraph examination of 50
suspects in criminal cases in the laboratory of the Osaka Prefecture Police
Headquarters. The charts were analyzed numerically at the National Institute of
Police. Science applying a criterion of rank scores. Chance on each test was 17
. percent. The location of the number in the list was unknown to the blind reviewer
and no dummy number was inserted at the beginning of each list.
Results
Applying only the rank score of .breathingsuppression, the detection rate was
46 percent. Applying only the rank score of electrodermal responses, the correct
detection of the critical number was 72 percent. Detection with each individual
measure was significantly greater than chance. When the two indices were combined,
the detection rate was 92 percent.
4
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Robert M. Stern, John P. Breen, Takami Watanabe, and Bradley S. 'Perry. "Effect
of Feedback of Physiological Information on Responses to Innocent Associations and
Guilty Knowledge." Journal of Applied Psychology, 66 (6) (1981): 677-681.
Procedure
Experiment 1 involved 24 males and 24 females to determine the effect of
feedbick on innocent associations. Subjects selected a geometric figure from five
possible figures and wrote down their social security number and four additional
social security numbers from a provided list. Sixteen subjects received no
feedback, 16 received auditory heart rate feedback, and 16 auditory electrodermal
feedback while being asked four series of questions. In two series, subjects were
asked if they had picked each of five geometric forms. In another two series,
subjects were asked if each social security number was theirs. Subjects were
required to respond with no to all questions. The first question of a series was a
buffer question to reduce orienting. Detection was based on the electrodermal data
with the response to each of the five questions in each series scored from one to
five based on magnitude. The mean rank to each question of the two social security
number series and the mean rank to each question of the two geometric form series
was then computed.
The proportion of subjects showing the largest electrodermal response to the
relevant stimulus (their social security number and the selected geometric form)
significantly varied with feedback and type of stimulus and is as follows:
(1) No.feedback - form (5/16), number (9/16); (2) Heart rate feedback - form
(8/16), number (11/16); (3) Electrodermal feedback - form (11/16), number (12/16).
Thus, subjects receiving electrodermal feedback were easiest to detect and social
security numbers were easier to detect than geometric forms.
In Experiment 2, 26 males and 26 females were subjects in a mock crime
situation. All subjects played the role of an American Embassy hostage in Iran:
The 26 designated guilty subjects received written instructions (which they were to
read three times) about their involvement in a plot to murder the Ayatollah. The
26 innocent subjects received instructions that had some of the same relevant words
as those of guilty subjects but involved no murder plot. Thirteen innocent
subjects and 13 guilty subjects received auditory electrodermal feedback. 'The
remaining subjects received no feedback. The guilty knowledge technique was used.
Ten questions, each with five alternatives, were presented to the subject with the
first alternative serving as a buffer and not scored. Four of the ten questions
were control questions. Subjects were to make no response to the questions.
Responses were scored by the method of Lykken (1959).
Results
Results are demonstrated in the following table:
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
Predicted
Actual
Guilty Innocent
Guilty (No Feedback) 12
Guilty (Feedback) 13 1
Innocent (No Feedback) 1 11
Innocent (Feedback) 0 12
Thus, 96% of the guilty and 88.5% of the innocents were correctly classified.
Electrodermal feedback did not significantly affect detection accuracy.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7 .
William M. Waid, Emily Carota Orne, Mary R. Cook, and Martin T. Orne. "Effects
of Attention as Indexed by Subsequent Memory On Electrodermal Detection of
Information." Journal of Applied'Psychology. (1978) 63(6): 728-733.
Experiment 1
Procedure
Twenty-nine designated guilty subjects and 11 designated innocent subjects
(males) participated in a study to determine the effect of attention on detection
of deception using a guilty knowledge test. Guilty subjects learned six code words
during a series of tasks. Innocent subjects completed a questionnaire during this
period. Guilty subjects were told that highly intelligent and mature individuals
could deceive the examiner. Care was taken to involve innocent .subjects in the
procedure. All subjects were told that the way to appear innocent was not to
respond during the test. Subjects were tested by an experimenter blind to subject
status. Subjects were presented twice a list of 24 words. There were 6 categories
of words with each category containing one of the learned code words. A word was
considered detected if the code word elicited a larger response than the remaining
3 words in the category. For innocent subjects, one of the four words in each
category was arbitrarily designated as the critical word. One category of four
words was eliminated from the analysis. Skin resistance was measured. After the
polygraph test, subjects were given a list of 115 words and asked to indicate the
words heard during the polygraph test.
Results
When the criterion for detection was 3 or more words (chance = 2.5 words), the
following results were obtained:
Predicted
Guilty
Innocent
Actual
Guilty Innocent
23 3
6 8
Seventy-nine percent of the guilty subjects and 73% of the innocent subjects
were detected. Guilty subjects had significantly more critical words detected than
innocent subjects.
Guilty and innocent subjects did not differ in percentage of words recalled.
Detection in innocent and guilty groups did not appear related to number of words
recalled although there were some trends.
Experiment 2 4
Procedure
Procedure was the same as Experiment 1 with some minor differences. Eighteen
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
guilty and 10 innocent subjects (males) were used. The list of 24 words was
presented 5 times. Subjects were tested in groups of 3 to 7 by an examiner in
another room. Subjects received the stimuli through headphones and indicated their
answers by moving their head. The answer was observed through a one?way window.
Subjects were blindfolded so they could not see the other subjects and were told
they would receive a shock at the end of the test proportional in strength to the
number of detected lies. After the test, subjects were to write down as many of
the words involved in the test-they could recall. A code word was considered
detected if it elicited a larger response than the other three codewords in its
category. One category of 4 words was removed from all analysis.
Results
Subjects were called guilty if seven or more critical words were detected
(chance = 6 words). The following results were obtained:
Predicted
Actual
Guilty' Innocent
Guilty 11 1
Innocent . 7 9
Sixty?One per cent of the guilty subjects and 90 percent of the innocent subjects
were correctly detected.
Guilty subjects had significantly more critical words detected than innocent
subjects and guilty subjects recalled significantly more words than innocent
subjects. Among guilty subjects correlations between number of words recalled and
number of words detected was .48. Detected guilty subjects recalled significantly
more words than not detected guilty subjects.
Experiment 3
Procedure
Procedure was the same as Experiment I except for minor differences. Fifteen
innocent and 15 guilty subjects (males) were tested. The list contained 36 words
with 6 words in each category; 1 word in each category was the code word. The list
was presented 4 times. One category of 6 words was eliminated from all analyses.
Skin conductance was measured. As in Experiment 2, subjects were to write down as
many words they could recall after the test.
Results
The following results were obtained when the criterion for detection was 6
words (chance = 5 words):
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Predicted
Guilty
Innocent
Actual
Guilty Innocent
11 3
4 12
Seventy-three percent of the guilty subjects and 80% of the innocent subjects
were detected. Guilty subjects had significantly more code words detected than
innocent subjects but there was no difference between guilty and innocents in
number of words recalled. Among guilty subjects, number of words recalled was
correlated .41 with number of words detected. Detected guilty subjects recalled
significantly more words than not detected guilty subjects.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
William M. Waid, Emily C. Orne, Mary R. Cook, and Martin T. Orne. "Meprobamate
Reduces Accuracy of Physiological Detection of Deception." Science, 212 (1981):
71-73.
Procedure
Male college students (aged 18 to 24 years) were paid t2 an hour as subjects in
an experiment to determine the effect of meprobamate on detection of deception.
Subjects designated guilty overlearned six words during timed, interpretated tasks
prior to the polygraph tert. Innocent subjects performed the same tasks but
learned no words. Guilty subjects were told that it was possible to deceive the
examiner, whereas innocent subjects were told it is often difficult to prove one's
innocence. Guilty subjects were randomly assigned to one of three groups (11 per
group). One group was told they were taking a tranquilizer which would help them
avoid detection and were given 400 mg. meprobamate. Another group was given a
placebo. The third guilty group received nothing.
After 30 minutes, a second experimenter, blind to subject status, administered
a polygraph test using the guilty knowledge technique. The question list consisted
of 24 words, four in each of six categories. One of four words in each category
was a word the guilty subject memorized. The first word of the list was a dummy
word. The list was presented twice. The examiner also completed a rating scale
about his perception of subject's drug status.
Electrodermal response amplitude, smallest inspiration, and change in relative
blood pressure were recorded and scored by technicians who had not observed the
test. A, subject was classified guilty if his response to the relevant word in each
set of four produced a larger response than the remaining three words on five or
more sets. In order to analyze data from innocent subjects, one word in each set
of four was randomly designated as the critical word.
The following results for each measure were obtained:
Electrodermal
Guilty:
Actual
Guilty:
Guilty:
Innocent
No Pill
Placebo
Meprobamate
Guilty
9
0
8
3
Innocent
2
11
3
8
Cardiovascular
Predicted
Guilty
0
1
2
0
Innocent
11.
10
9
11
Respiratory
Guilty
4
1
3
5
Innocent
7
10
8
6
Correct classification was reduced significantly in the meprobamate group.
Electrodermal activity was a better discriminator between guilty and innocent than
dhe other measures.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
The authors note that results were not due to reduced electrodermal
responsiveness in the meprobamate group since there were no differences between the
groups in average number of critical words eliciting a measurable response and no
differences in average, response to all questions. Examiners were not able to
accurately judge whether qr not subjects had received meprobamate.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
Kazunoba Yamaoka and Akihiro Suzuki. Studies on Skin-Blood Flow as an Index of
Lie Detection. Polygraph 9(4) (December 1980): 232-237*.
Procedure
Seven females and 6 males-(aged 18 to 29 years) were subjects in a study to
determine the effectiveness of skin-blood flow as an index of lie detection.
Subjects selected 1 of 5 playing cards. When the subject was asked about each card
during a subsequent polygraph test he was to reply no to all questions. The
examiner was blind to the selected card; experience of the examiner is not
discussed. Subjects were presented 2 series of questions about each card; each
series contained a different question order. The response to the first question
was eliminated. Skin potential, skin resistance, and skin-blood flow were
recorded. Responses to each question for each system were ranked from 1 to 5 with
the largest response receiving a rank of I. The following results were obtained:
Number of Number of
Successful Detections Unsuccessful Detections
Skin-Blood Flow
5
8
Skin Resistance
2
11
Skin Potential
10
3
Skin potential was the only response measure able to detect the selected card
significantly greater than chance levels.
*Previously published in Reports of the National Institute of Police Science
26 (1973): 206-209. 'Text in Japanese.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
References Cited
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
?
References Cited
Most of the references cited in the text will be found in the abstracts
section. Those references which are not in the abstracts are listed below:
Ash, Philip. "Survey of Attitudes on the Polygraph." Polygraph 2(3)
(September 1973): 200-223.
Barland, Gordon H. "On the Accuracy of the Polygraph; an Evaluative Review of
Lykken's Tremor in the Blood." Polygraph 11(3) (September 1982): 258-272.
Beijk, J. "Experimental and Procedural Influences on Differential
Electrodermal Activity." Psychophysiology 17(3) (1980): 274-278.
? Carlucci, Frank. Memorandum for Secretaries of Military Departments, Chairman,
Joint Chiefs of Staff, Directors of Defense Agencies. Subject: "Degradation of
Operational Readiness/Mission accomplishment Due to Personnel Security
Investigative Shortfalls." 6 August 1982.
"A Comparison of Information Obtained during Military Interview Program
Interviews with that Obtained during Military Conversion Polygraph Interviews."
unpublished manuscript, National Security Agency, 1983.
Dawson, Michael E. "The Delayed Answer Test (DAT) and the Effects of ?
Countermeasures." Paper presented at the Annual Meeting of the American Polygraph
Association, Las Vegas, August, 1977.
Dawson, Michael E. "Physiological Detection of Deception; Measurement-of
Responses to Questions and Answers During Countermeasure Maneuvers."
Psychophysiology 17(1) (January 1980): 8-17.
Engel, B. T. "Stimulus-Response and Individual-Response Specificity."
Archives of General Psychiatry 2(1960): 305-313.
Garwood, M.; Engel, B. T.; and R. Capriotti. "Autonomic Nervous System
Functioning and Aging: Response Specificity." Psychophysiology 19(4)(1982):
378-385.
Gustafson, Lawrence A. and Orne, Martin T. "Effects of Heightened Motivation
on the Detection of Deception." Journal of Applied Psychology 47(1963): 408-411.
Gustafson, Lawrence A. and Orne, Martin T. "The Effect of 'Lying 'in 'Lie
Detection' Studies." Paper read at the 35th Annual Meeting of the Eastern
Psychological Association, April 17, 1964, Philadelphia, Pennsylvania.
Gustafson, Lawrence A. and Orne, Martin T. "The Effects of Verbal Responses on
the Laboratory Detection of Deception." Psychophysiology 2(1)(1965): 10-13.
Heckel, R. V.; Brokaw, J. R.; Sulzberg, H. C. and Wiggins, S. L. "Polygraph,
Variations in Reactivity Between Delusional, Non-Delusional, and Control Groups in
a 'Crime' Situation." Journal of Criminal Law, Criminology and Police Science
53(3)(1962): 380-383.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
r
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Kugelmass, Sol and Lieblich, Israel. "Relation Between Ethnic Origin and GSR
Reactivity in Psychophysiological Dectection." Journal of Applied Psychology
52(2)(1968): 158-162.
Kugelmass, Sol; Lieblich, Israel; Ben-Ishai, Akiva; Opatowski, Abraham and
Kaplan, Maier. "Experimental Evaluation of Galvanic Skin Response and Blood
Pressure Change Indices During Criminal Interrogation." Journal of Criminal Law,
Criminology and Police Science.
Lieblich, Israel. "Manipulation of Contrast Between Differential GSR Responses
Through the use of Ordered Tasks of Information Detection." Psychophysiology
6(1)(1969): 70-77.
Lieblich, Israel; Kugelmass, Sol; and Ben-Shakar, Gershon. "Efficiency of GSR
Detection of Information as a Function of Stimulus Set Size." Psychophysiology
6(5)(1970): 601-608.
Lieblich, Israel; Naftali, Gideon; Shmueli, Joseph and Kugelmass, Sol.
"Efficiency of GSR Detection of Information with Repeated Presentation of Series of
Stimuli in Two Motivational States." Journal of Applied Psychology 59(1)(1974):
113-115.
Lindsey, Robert. The Falcon and the Snowman: A True Story of Friendship and
'Espionage. New York: Simon & Schuster, 1979.
Linehan, John G. "An Aspect of World War II Use of the Polygraph." Polygraph
7(3)(September 1978): 233-'239.
Lykken, David T. "The Detection of Deception." Psychological Bulletin
86(1)(1979): 47-53.
Lykken, David T. "The Psychopath and the Lie Detector." Psychophysiology
15(2)(1978): 137-142.
Lykken, David T. A Tremor in the Blood, Uses and Abuses of the Lie Detector.
New York: McGraw-Hill, 1980.
MacNitt, Reginald D. "In Defense of the Electrodermal Response and Cardiac
Amplitude as Measures of Deception." Journal of Criminal Law, Criminology and
Police Science 33(1942): 266-275.
_Marston, William Moulton. The Lie Detector. Test. New York: R. R. Smith, 1938.
Miyake, Yoichi. "A Study of Skin Resistance Response, Photoplythsmographic
Vasomoter Response and Eye Movement as Indices of Lie Detection." Reports of the
National Institute of Police Science 31(2)(May 1978): 18-24.
Ohkawa, Hisatsugi. "Comparison of Physiological Response of 'yes', 'no', and
'mute' Conditions in Peak of Tension Test." Research Materials No. 21, Polygraph
Reports, National Institute of Police Science (1963): 1-4.
Phannenstill, Richard J. "The Polygraph Passes the Test." Security Management
27(8)(August 1983): 58-60.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7
Podlesny, John A. and Raskin, David C. "Effectiveness of Techniques and
Physiological Measures in the Detection of Deception." Psychophysiology 15(4)(July
1978): 344-359.
Putnam, Richard L. "Polygraph Screening of Police Applicants: Necessity or
Abuse?" Polvaraph 7(4)(December, 1978): 257-262.
"Security Interviews with the Aid of a Polygraph Compared to the Special
Background Investigation as Sources of Information." Unpublished manusctipt,
National Security Agency, 1983.
Silverberg, Ben A. "Attitudes of Job Applicants Toward the Arther Security
Clearance Examination." Journal of Polygraph Science I4(6)(May-June, 1980): 1-4.
Silverberg, len A. "Attitudes of Job Applicants and Employees Toward the
Polygraph." Polygraph 9(3)(September, 1980): 162-169.
Silverberg, Ben A. "Employee 4ttitudes Toward Periodic Polygraph
Examination." Journal of Polygraph Science 15(4)(January-February, 1981): 1-4.
Statistical Study on the Use of the Polygraph as a Personnel Security Screening
Aid." Unpublished manuscript, National Security Agency, 1953.
Thatcher, Margaret. "Text of Prime Minister's Statement: Security Commission
Report on the Prime Case." 1982.
Trovillo, Paul V. "Report on a Survey of. Methods of Operation and
Accomplishments of Russell Chatham, Inc. Polygraph Program at Oak Ridge,
.Tennessee." Unpublished manuscript, April 14, 1951. 43 pp.
Declassified and Approved For Release 2013/04/16: CIA-RDP07-01280R000400400001-7