VALIDITY AND RELIABILITY OF DETECTION OF DECEPTION
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP85B00263R000100240004-7
Release Decision:
RIFPUB
Original Classification:
K
Document Page Count:
33
Document Creation Date:
December 20, 2016
Sequence Number:
4
Case Number:
Content Type:
OPEN SOURCE
File:
Attachment | Size |
---|---|
CIA-RDP85B00263R000100240004-7.pdf | 2.63 MB |
Body:
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Validity
and Reliability of
Detection of
Deception
by
David C. Raskin, Ph.D.
Gordon H. Barland, Ph.D.
John A. Podlesny, M.A.
June 1978
National Institute of
Law Enforcement and Criminal Justice
Law Enforcement Assistance Administration
U. S. Department of justice
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
National Institute of Law Enforcement
and Criminal Justice
Blair G. Ewing, Acting Director
Law Enforcement Assistance Administration
James M. H. Gregg, Acting Administrator
This project was supported by Grant Number 75-NI-99-0001,
awarded to the University of Utah by the National Institute of
Law Enforcement and Criminal Justice, Law Enforcement
Assistance Administration, U. S. Department of justice, under
the Omnibus Crime Control and Safe Streets Act of 1968, as
amended. Points of view or opinions stated in this document
are those of the authors and do not necessarily represent the
official position or policies of the U. S. Department of Justice.
For sale by the Superintendent of Documents, U.S. Government Printing Office
Washington, D.C. 20402
Stock Number 027-000-00692_2
^-tT,n yew ~
M11
~
~lcf
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
CONTENTS
Page
I. METHODS ....................................................
1
A.
Laboratory Experiments ........................................
I
1. ExperimentI ..............................................
1
2. Experiment II .............................................
2
B.
Field Studies ..................................................
4
1. Reliability and validity with criminal suspects ...................
4
2. Effectiveness of physiological measures with criminal suspects ....
4
3. Evaluation of current practices by law enforcement and private
polygraph examiners ........................................
5
4. Accuracy of chart interpretation ..............................
5
5. The "friendly polygrapher" ..................................
6
6. Errors in examinations ......................................
6
II. RESULTS .....................................................
8
A.
Accuracy of Decisions ..........................................
8
1. Laboratory experiments .....................................
8
2. Field study ................................................
8
3. Psychopaths ...............................................
10
4. Other subject characteristics .................................
10
5. Type of crime .............................................
10
6. Behavior symptoms .. . ....................................
10
B.
Reliability of Chart Interpretation ................................
11
1. Experiment II .............................................
11
2. Criminal suspects study ............. ..................
11
3. Current practices by law enforcement and private examiners ......
11
4. Accuracy of chart interpretation ..............................
11
C.
Relative Effectiveness of Standard Measures .......................
11
1. Laboratory experiments .....................................
12
2. Field studies ..............................................
12
D.
Quantitative Analyses of Physiological Responses ..................
12
1. Respiration amplitude .......................................
12
2. Respiration cycle time ......................................
13
3. Skin conductance response amplitude .........................
13
4. Skin conductance response rise time ..........................
13
5. Skin conductance response recovery ..........................
13
6. Cardio responses ...........................................
13
7. Finger blood volume ........................................
13
8. Finger pulse amplitude .......................................
15
9. Heart rate changes .........................................
15
10. Cardio activity monitor .....................................
15
11. Skin potential responses .....................................
15
E.
Comparisons of Question Techniques .............................
20
1. Relative accuracy of control-question and guilty-knowledge tests ..
20
2. Comparison of Backster and Reid control questions .............
20
3. Effectiveness of a guilt-complex question ......................
20
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
F. The "Friendly Polygrapher.. .................................... 21
G. Errors in Examinations ......................................... 21
III. CONCLUSIONS AND RECOMMENDATIONS .................... 23
IV. REFERENCES ................................................. 26
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
ABSTRACT
This project was designed to provide information on the usefulness of polygraph
techniques in detecting truth and deception in criminal investigations. The report
describes the methodology of each of the eight experiments and studies conducted.
In addition to studies involving criminal suspects in a field situation, other studies
involved laboratory experiments with a mock-crime paradigm. The lab experiments
investigated aspects of the general problems of accuracy and reliability not easily
studied in the field situation. They also assessed the usefulness of a number of
physiological measures that had received little attention in previous scientific.
research. Studies also were undertaken to evaluate the commonly-held belief that
psychopaths can "beat the polygraph," the adequacy of current practices used by
field polygraphists, the usefulness of different question structures in polygraph
examinations, and the risks of different type's of errors in field applications. The
project results indicate that polygraph examinations using control-question or guilty-
knowledge tests are highly accurate.
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
1. METHODS
The eight separate phases of this project can be
divided into two categories. One category consists of
two laboratory experiments in which the subjects
were offered monetary incentives for participating in
a mock-crime and attempting to produce truthful
outcomes on the polygraph examination. Such stud-
ies have certain advantages. First, the laboratory
setting allows complete and certain determination of
ground (factual) truth. Therefore, the accuracy of
outcomes and other results can be assessed against
the certain knowledge of truthfulness or deception
on the part of the subject. Second, in a laboratory
situation it is possible to compare and evaluate
different question techniques (test structures) and
various physiological measures which may or may
not have been extensively employed in previous
research and application. Finally, it is possible to
investigate the influence of subject characteristics
such as psychopathy (sociopathy) in a controlled
situation.
The results of laboratory experiments are very
useful in making generalizations to the field situation
with criminal suspects. However, such inferences
should be made cautiously and tested by research in
the field setting. Furthermore, there are many ques-
tions concerning field practices which can be an-
swered only by studies of field applications. There-
fore, the remaining six studies involved investigations
which utilized data obtained from polygraph exami-
nations conducted on criminal suspects for real-life
purposes. Some of those examinations were con-
ducted at the University of Utah, and others were
provided by a variety of law enforcement and private
polygraph examiners.
A. Laboratory Experiments
The two laboratory experiments were similar in
design and procedures employed. Both utilized a
mock-crime situation in which the subjects were
informed about the nature of the crime. Half of the
subjects in each experiment were instructed to com-
mit the crime (guilty subjects), and the other half
were merely informed about the nature of the crime
(innocent subjects). Each subject was subsequently
administered a polygraph examination by an exam-
iner who had no knowledge concerning the guilt or
innocence of the subject.
All subjects had been instructed to deny having
committed the theft, and they were offered a cash
bonus if they could produce truthful results on the
polygraph test. The polygraph examiner (who was
trained and experienced in field polygraph tech-
niques) conducted a standard pretest interview with
each subject, administered the polygraph test, and
made his decision on the basis of numerical evalua-
tion of the polygraph charts (Raskin, 1975). The first
three charts for each subject were later subjected to
detailed quantitative analyses utilizing computer
techniques.
1. Experiment I. This research (Raskin, 1975) was
conducted at a provincial prison in British Columbia,
Canada with 48 male volunteers from the prison
population. All of the subjects were convicted felons,
and half of them had been clinically diagnosed as
psychopathic (sociopathic). The crime consisted of
stealing $20 from a drawer in a room which was off-
limits to inmates, and all subjects (guilty and inno-
cent) were instructed to deny the theft and attempt
to produce truthful results on the polygraph test. All
subjects who produced truthful polygraph charts
received a $20 bonus.
The polygraph test was a federal zone-comparison
control-question test (Barland & Raskin, 1975) con-
sisting of a number test followed by a minimum of
three charts. A typical question sequence was as
follows:
? (neutral) Were you born in Canda? Yes.
? (sacrifice relevant) Regarding that $20, do you
intend to answer truthfully each question about
that? Yes.
? (outside issue) Are you completely convinced I
will not ask a question that hasn't been re-
viewed? Yes.
? (control) Other than what you told me, before
you were 18 did you ever steal any money? No.
? (relevant) Did you take that $20? No.
? (control) Did you ever steal anything else from
someone who trusted you? No.
? (relevant) Did you take that $20 from the
drawer? No.
? (neutral) Is your last name ? Yes.
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
?(control) Have you ever taken anything of value
from an employer? No.
? (relevant) Do you have that $20 now? No.
According to the theory of control-question tests
(Backster, 1962; Podlesny & Raskin, in press; Reid
& Inbau, 1966) the subject will respond most to the
questions which are the greatest threat at that time.
For a guilty person the relevant questions are the
most threatening, and he will show larger reactions
to the relevant as compared to the control questions.
However, the innocent subject knows he is being
truthful to the relevant questions, and he should be
more concerned about the control questions. There-
fore, the innocent subject should produce larger
reactions to the control than to the relevant ques-
tions.
Skin conductance (SCR), thoracic and abdominal
respiration, plethysgiographic measures of finger
blood volume (FBV) and finger pulse amplitude
(FPA), heart rate (HR), and skin potential (SPR)
were measured on a Beckman Dynograph. The time
between the beginning of consectuve questions
ranged between 25 and 35 seconds. Following each
chart, the examiner asked the subject if any question
bothered him and if he would like to change the
wording of any questions. The subject's attention
was directed toward the control questions, which
were frequently modified following admissions or
expression of concern by the subject. If the results
did not seem to be conclusive after three charts, the
fourth chart was a silent answer test (Horvath &
Reid, 1972). Additional charts were sometimes ob-
tained up to a maximum of seven charts.
The charts were then scored numerically using the
three components of respiration, SCR, and cardio-
vascular (plethysmograph). If the total score was +6
or larger, the decision was truthful; if it was -6 or
smaller, the decision was deceptive; and scores
between ?6 were called inconclusive. All of the
physiological measures were then subjected to a
detailed quantitative analysis by a person who had
no knowledge of the outcome or experimental group
of any subject.
2. Experiment 11. This experiment (Podlesny et
al., 1976) was conducted at the University of Utah
using 60 male subjects recruited from the community
by newspaper advertisements. They were paid $5 for
participation and were offered a $10 bonus if they
produced truthful results. The crime consisted of
stealing a gold wedding ring from a drawer in a
secretary's desk on another floor of the building. All
subjects were informed that they were also suspected
of having stolen a watch. That was introduced to
allow the evaluation of the effectiveness of a guilt-
complex question as a type of control question
(Barland & Raskin, 1973; Lykken, 1974; Reid &
Inbau, 1966).
The testing procedures were the same as in
Experiment I with a few exceptions. For 20 subjects
the control questions were of the type which clearly
excludes the crime being investigated (Backster,
1969), and for 20 subjects the control questions
overlapped the relevent issue (Reid & Inbau, 1966).
A typical question sequence with Reid type control
questions was as follows:
? (neutral) Is your name
? Yes.
? (sacrifice relevant) Regarding the ring and watch,
do you intend to answer the questions about
them truthfully? Yes.
? (outside issue) Are you convinced I will only ask
questions on this test that you've already.
okayed? Yes.
? (Reid control) Have you ever stolen any
money? No.
? (relevant) Did you take that ring? No.
? (Reid control) Besides what you told me about,
have you ever taken anything of value? No.
? (relevant) Did you take that ring from that
desk? No.
? (guilt complex) Did you take that watch from
Room 702? No.
? (Reid control) Have you ever taken anything
from someone who trusted you? No.
? (relevant) Do you have that ring with you
now? No.
For subjects who received Backster type control
questions, the words "Have you ever" were re-
placed with wording which clearly excluded the
crime being investigated. That was done by specifly-
ing a certain time period such as "While you were in
high school." A typical Backster control question
was "Between the ages of 15 and 20 did you take
something of value?"
An additional 20 subjects were examined using the
guilty-knowledge test (Lykken, 1959). The rationale
for the guilty-knowledge test is that the absence or
presence of differential responsivity to items of
information known only to a guilty person provides
the basis for conclusions about truth or deception
concerning the crime. The subject was administered
a series of five charts each having six alternatives to
a different question concerning information related to
the crime. The items consisted of a set of equally
plausible alternatives, one of which was the correct
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Inc question sequence was as
follows:
Chart 1
Regarding the type of ring that may have been taken,
(I) Do you know if it was a sapphire class ring?
(2) Do you know if it was a pearl engagement ring?
(3) Do you know if it was a silver and turquoise ring?
?(4) Do you know if it was a gold wedding ring?
(5) Do you know if it was a ruby class ring?
(6) Do you know if it was a diamond engagement ring?
Chart 2
Regarding the floor of this building that the ring was hidden
(1) Do you know if it was the 1st floor?
(2) Do you know if it was the 12th floor?
(3) Do you know if it was the 6th floor?
(4) Do you know if it was the 4th floor?
?(5) Do you know if it was the 8th floor?
(6) Do you know if it was the 10th floor?
?
Chart 3
Regarding the number of the room that the ring was hidden
(1) Do you know if it was Room 800?
?(2) Do you know if it was Room 820?
(3) Do you know if it was Room 810?
(4) Do you know if it was Room 816?
(5) Do you know if it was Room 814?
(6) Do you know if it was Room 803?
Chart 4
Regarding the type of envelope that the ring was hidden in,
(1) Do you know if it was an inter-campus mail envelope?
(2) Do you know if it was a medium-sized manila envelope?
?(3) Do you know if it was a business-sized white envelope?
(4) Do you know if it was a small-sized manila envelope?
(5) Do you know if it was a small-sized white envelope?
(6) Do you know if it was a large-sized manila envelope?
nique was explained to each subject in that group,
and prior to each chart the question was reviewed,
but no alternatives were stated until the test was
administered except with Chart 4. Prior to that chart,
an example of each type of envelope was shown to
the subject and specifically named. Subjects were
instructed to answer "no" to each alternative on all
of the charts.
The probability that an innocent subject without
knowledge of the details of the crime would show
his largest reaction to a critical item is 1/5 for each
chart. Thus, only 6 of 100 innocent subjects would
produce their largest reactions to the critical item on
three or more charts. However, the probability of a
series of such large reactions from a subject with
guilty knowledge is very high. This guilty knowledge
procedure is designed to protect against false posi-
tives (an innocent person producing deceptive re-
sults) which Lykken (1974) claims are frequent
occurrences with control-question tests.
With both control-question and guilty-knowledge
tests, a minimum of 15 seconds elapsed between the
verbal response to the question or alternative item
and the beginning of the next question or item.
Control-question tests consisted of a number test and
a minimum of three charts. Responses were recorded
on a Beckman Dynograph and included thoracic
respiration, skin conductance (SCR), plethysmo-
graphic measures of finger pulse amplitude (FPA)
and finger blood volume (FBV), relative blood pres-
sure measured by a low-pressure (50-50 mmHg)
cardio cuff, skin potential (SPR), heart rate (HR)
measured on a second-by-second basis, and a dry
cardio activity monitor (CAM) placed on the palmar
tip of the second finger.
Following the last chart, the results were immedi-
ately evaluated to determine whether the subject was
truthful or deceptive. Control-question tests were
numerically evaluated in the same manner as Exper-
iment I using the SCR, respiration, relative blood
pressure (cardio), and plethysmograph measures. For
guilty-knowledge tests, the size of the SCR to each
critical item was measured to determine if it was the
largest of the responses to the five alternative items
for that question. If the response to the critical item
was largest on at least three of the five charts, the
subject was called deceptive. If he showed less than
three such responses, he was called truthful.
All of the charts were subsequently scored inde-
pendently by an examiner who had no contact with
the subjects and no knowledge of their guilt or
innocence. All of the analyses of numerical scores
reported here are based on the results of the inde-
Regarding the name
instructed to ask for,
(1) Do you know if it was Dr. Trumbull?
(2) Do you know if it was Dr. Tolman?
(3) Do you know if it was Dr. Heisse?
(4) Do you know if it was Dr. Jordan?
(5) Do you know if it was Dr. Calvin?
?(6) Do you know if it was Dr. Mitchell?
The first alternative was included to buffer initial
responding and was not scored. The remaining five
alternatives consisted of a critical item and four
noncritical items. The critical item was the correct
alternative, and noncritical items were all incorrect.
The critical items were positioned among the noncrit-
ical items in a pseudo-random order across charts.
In the above list, critical items are identified with an
asterisk. The rationale of the guilty-knowledge tech-
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
1 --j.. 1 i e contool-qU._:Ai,)il tests were
evaluated in the same manner as was originally
performed. For the guilty-knowledge tests, the rank-
ing system used by Lykken (1959) was employed. If
the critical item produced the largest SCR, it was
assigned a value of 2; if it was the second largest, it
received a 1; and ranks lower than second largest
was assigned a 0. The ranks for the critical items
were summed over the five charts. If the total was 6
or higher, the subject was called deceptive. If the
total was less than 6, the subject was called truthful.
All of the physiological measures were then sub-
jected to a detailed, quantitative analysis by persons
who had no knowledge of the outcome or experimen-
tal group for any subject.
B. Field Studies
A total of six different studies were completed
using polygraph examinations on criminal suspects.
Except for one source in the study described below
under the heading of "current field practice," all of
the examinations used a control-question technique.
Some of them were Backster zone-comparison tests
(Bailey & Rothblatt, 1970), some were federal zone-
comparison tests (Barland & Raskin, 1975), and
some were Reid control-question tests (Reid &
Inbau, 1966).
1. Reliability and validity with criminal suspects.
In spite of careful attempts to simulate field situa-
tions, there are a number of important differences
between most laboratory experiments and applica-
tion of detection of deception in the field situation
with criminal suspects. Those include profound dif-
ferences in the consequences of the outcome and the
resulting differences in subject motivation, differ-
ences in subject populations, the availability of
information about the case which may influence the
examiner and the subject, and frequent differences in
the type of techniques utilized and the training and
experience of field examiners as compared to the
typical laboratory researchers. In the two experi-
ments described above, many of those differences
were eliminated. However, it was not possible to
eliminate some of them.
This study (Barland & Raskin, 1976) was designed
to extend the results obtained by Bersh ( 1969) and to
overcome some of the limitations of that study by
using nonmilitary criminal suspects obtained by re-
ferral from both law enforcement and defense
sources. It also studied the relationship between
various personality, behavioral, socioeconomic,' and
crime categories and the results obtained on the
polygraph tests.
A total of 102 criminal suspects were examined at
the request of polio, defense attorneys, or prosecut-
ing attorneys; and 92 independent cases were se-
lected from those. All but one subject was tested on
field model polygraphs with a federal zone-compari-
son control-question technique, and the results were
evaluated using the numerical scoring procedures
described above. The charts were subsequently eval-
uated by an independent examiner who had no
knowledge of the case or the original outcome, and
the results reported here are based on those evalua-
tions. All tests included a minimum of three charts,
and numerical evaluations were based on the stand-
ard field measures of respiration, skin resistance, and
cardiovascular activity.
Background information was obtained from each
subject, and he or she responded to several scales
from the Minnesota Multiphasic Personality Inven-
tory (MMPI) including the Lie, K, Psychopathic
Deviancy, Hypochondriasis, and Depression scales.
The examiner also observed both spontaneous be-
havior cues and those elicited by specific questions
reported to be helpful in differentiating truthful and
deceptive persons (Horvath, 1973). On the basis of
that behavioral information, the examiner made two
covert predictions of.the outcome of the polygraph
test, the first immediately following the advisement
of rights and the second just prior to the administra-
tion of the polygraph test.
Three criteria were developed for assessing ground
truth in order to evaluate the accuracy of the
polygraph results. The first consisted of the inde-
pendent judgments of a 5-member panel of experts
composed of two criminal defense attorneys, two
criminal prosecuting attorneys, and a judge. In cases
where at least 3 of the 5 panel members agreed on a
decision of guilty or innocent, the judgment of the
panel was used as the criterion for ground truth. The
second criterion consisted of judicial outcomes in
which the polygraph results played no role and the
case was not dismissed for insufficient evidence. The
third consisted of a full confession or plea of guilty
to the original charge. The latter criterion was used
only for analyses to assess the effectiveness of the
three physiological components with guilty subjects.
2. Effectiveness of physiological measures with
criminal suspects. In order to test the findings from
laboratory Experiment II on a sample of subjects
tested on real-life criminal issues, cases were ob-
tained by referral from defense and prosecution
sources. A total of 26 subjects was tested on the
criminal charge involved in the case using the federal
zone-comparison control-question test.
All subjects were examined at the University of
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Utah laboratory using the same instrumentation
employed in Experiment II (above). From that group
of subjects, 14 were selected to analyze the effective-
ness of the non-standard physiological measures
evaluated in Experiment II. Since the standard
measures of respiration and skin conductance had
repeatedly been demonstrated to identify guilty and
innocent subjects in laboratory and field situations,
the subjects were separated into guilty and innocent
groups on the basis of the total numerical scores for
respiration and skin resistance responses. Only sub-
jects who produced decisive outcomes were utilized,
and 7 subjects were eliminated for analyses reducing
the innocent group to the same size (N=7) as the
guilty group. The mean total score on those measures
was -8.7 for the guilty group and +6.3 for the
innocent group. On the assumption that there was a
high degree of accuracy in assigning the subjects to
the proper group, it was then possible to evaluate
with a sample of criminal cases, all of the nonstand-
ard measures found to produce significant results in
Experiment II.
3. Evaluation ,Sf current practices by law enforce-
ment and private polygraph examiners. Although
there are several recent reports concerning the
accuracy and reliability of decisions made by field
polygraph examiners (Barland & Raskin, 1975; Hor-
vath, 1974; Horvath & Reid, 1971; Hunter & Ash,
1973; Slowik & Buckley, 1975), the sampling of
cases was highly selective in all but the Barland and
Raskin study. In the other four studies, only cases in
which the original examiner had made a definite
decision were utilized, and most of those were
selected from cases which had been verified by
confession of the guilty person. Those facts, plus the
laboratory nature of the Barland and Raskin experi-
ment, limit the representatives and generalizability of
the results to the typical field situation.
This study (Raskin & Barland, 1976) was designed
to overcome the above objections and to extend the
investigation to answer the questions of adequacy of
techniques employed by law enforcement and private
examiners, relative effectiveness with different crime
categories, and relative effectiveness of the standard
measures of respiration, skin resistance (galvanic
skin response), and relative blood pressure (cardio).
The procedure involved sampling approximately
60 recent cases from each of seven different loca-
tions, three law enforcement agencies and four well-
known polygraph firms. Cases were selected to
include examples from the crime categories of crimes
against people (homicide, assault), economic crimes
(robbery, burglary, theft), and sex and drug offenses.
Only tests performed on suspects were included in
the sample, and only one examination from any
specific case was included. Using those criteria, a
total of 419 examinations was obtained for analysis.
Each examination was independently evaluated by
Dr. Barland and Dr. Raskin. All of the standard
control-question charts were subjected to our stand-
ard numerical scoring procedure, and decisions were
made using the -*5 inclusive region for inconclusives.
The relevant-irrelevant tests obtained from one loca-
tion were evaluated subjectively since there is no
system for numerical scoring of such charts. Char-
acteristics such as chart quality, chart markings, case
information, question structure, source of referral,
length of pretest interview, and name of examiner
were noted for each examination. A total of 43
polygraph examiners was represented in the sample.
4. Accuracy of chart interpretation. Although there
has been a number of studies of reliability of chart
interpretation, the only reported studies which have
utilized numerical scoring procedures have been
conducted at the University of Utah (Barland &
Raskin, 1975, 1976; Podlesny, Raskin, & Barland,
1976). Those experiments have produced very high
rates of agreement on decisions (96-100%) based on
numerical scoring of charts. Those results together
with the high accuracy rates obtained with that
technique (Barland & Raskin, 1975; Podlesny, Ras-
kin, & Barland, 1976) seem to indicate that numerical
evaluation of polygraph charts might increase the
accuracy of decisions made on the basis of such
numerical scores. One purpose of this study (Raskin,
1976) was to investigate that possibility. Since poly-
graph examiners differ widely in their training and
experience with chart interpretation, a second pur-
pose of this study was to study the accuracy of chart
interpretation performed by polygraph examiners
with a variety of training, experience, and familiarity
with numerical scoring techniques.
Polygraph charts from 16 independent criminal
cases were selected from those obtained in a pre-
vious study (Barland & Raskin, 1976). Each exami-
nation consisted of three charts recorded on a field
model polygraph using the federal zone-comparison
control-question technique. All of the examinations
had been confirmed by confession of the guilty
person; 12 were from guilty subjects and four from
innocent subjects. The 16 sets of charts were inde-
pendently evaluated by 25 field polygraph examiners
from a variety of training backgrounds and experi-
ences. They were not informed about any aspect of
the case or the outcomes and were asked to render
a conclusion of truthful, deceptive, or inconclusive
,., 7
..... ?at a'.ri~.atie~ afifit.~liFi1e56...,i-A.,.., :'ir,~.$t. ~ nY~?~~n^ . ...,,
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
for each case. They were told to employ numerical
scoring if they had been trained in it and wished to
do so. Of the 25 examiners, 18 had at least one year
of experience, 13 had received formal training in
numerical scoring, but only 7 of those explicitly
scored the charts numerically.
5. The "friendly polygrapher." In a recent paper,
Orne (1975) stated that under certain conditions the
motivation of a guilty subject would be reduced to
the point that false negative errors would be greatly
increased. Since it is well known that motivation to
deceive and the threat of serious consequences of
detection are essential to successful detection of
deception, Orne speculated that polygraph examina-
tions conducted at the behest of defense attorneys
fail to meet those motivational requirements for
guilty subjects. He reasoned that a subject in such a
situation "knows that the results of the test if he is
found deceptive will not be used against him... As a
consequence, the client's fears about being detected
are greatly reduced (p. 114)." He also speculated
that the so-called "friendly polygrapher" employed
by the defense attorney will treat the subject differ-
ently than an "arms length" examiner such as a law
enforcement examiner or one working for the sub-
ject's employer. Orne concluded that such a situation
will make the guilty subject less detectable.
This study (Raskin, 1976) was designed to test
Orne's hypothesis using the results of polygraph
examinations conducted on a confidential basis for
defense attorneys and those conducted with explicit
knowledge and/or agreement with law enforcement
authorities or the subject's employer. The "friendly
polygrapher" hypothesis predicts that examinations
conducted confidentially for defense attorneys would
produce more truthful-appearing polygraph charts
and more truthful decisions than those performed
with the knowledge on the part of the subject that
the results would be reported to the law enforcement
authorities and/or his/her employer. However, the
theory and experience with control-question tests
would predict that such an effect would not occur
and the most to be expected would be an increase of
inconclusive results.
Three different samples of control-question exami-
nations were obtained. The first sample consisted of
all examinations of criminal suspects conducted dur-
ing a 1-year period by an experienced examiner for
law enforcement authorities or private attorneys. The
cases consisted of 106 examinations for law enforce-
ment authorities and 98 examinations for defense
attorneys. The outcomes of those examinations were
obtained in the form of the number of truthful,
deceptive, and inconclusive determinations.
The second sample consisted of control-question
examinations of criminal suspects obtained from two
private polygraph firms included in a previous study
(Raskin & Barland, 1976). A total of 19 different
examiners conducted the testing which consisted of
54 confidential examinations referred by defense
attorneys and 57 examinations performed with sub-
jects' knowledge that the results would be reported
to law enforcement authorities and/or his/her em-
ployer. Each set of polygraph charts was numerically
evaluated by Dr. Barland or Dr. Raskin prior to their
obtaining any information concerning the issue
tested, the source of referral, or the decision by the
original examiner. Since at least two charts were
obtained for each subject, the numerical score from
the first two charts comprised the data utilized in the
data analysis.
The third sample consisted of 27 control-question
examinations of criminal suspects conducted at the
University of Utah. Fourteen of the examinations
were performed on a confidential basis for defense
attorneys, and 13 were performed with the subject's
knowledge that the results would be reported to law
enforcement authorities. All examinations contained
a minimum of three charts, and the total numerical
scores for the first three charts were utilized in the
data analysis.
6. Fxmrs in examinations. In contrast to the con-
cern about false negative results expressed by Orne
(1975), another critic (Lykken, 1974) has focused a
great deal of attention on the problem of false
positive errors (a deceptive result obtained from a
truthful person). Lykken asserted that the control-
question technique cannot accurately identify inno-
cent suspects since he believes that it is impossible
to design control questions which will produce the
same level of responsiveness in innocent subjects as
is produced by the relevant questions with guilty
subjects. A number of scientifically conducted stud-
ies have investigated the accuracy of control-
question tests. Three of those are laboratory studies
conducted by Raskin and his associates (Barland &
Raskin, 1975; Podlesny, Raskin, & Barland, 1976;
Raskin, 1975). Of the errors obtained in those studies,
56% were false positives. Two recent field
experiments (Bersh, 1969; Barland & Raskin, 1976)
used the control-question technique with criminal
suspects and defined ground truth by means of the
judgments of a panel of experts. In the Bersh study,
44% of the errors were false positives, and all but one
of the errors in the Barland and Raskin study were of
the false positive type. Although the error rate is
relatively low (approximately 10%), there is evidence
that false positive errors may comprise a substantial
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
proportion of the errors obtained with criminal
suspects. Therefore, the major purpose of this study
(Raskin, 1976) was to attempt to determine some of
the factors which are associated with the occurrence
of such errors.
The type of error investigated was restricted to the
occurrence of inappropriate physiological responses
which occurred in the polygraph examination. Spe-
cifically, instances were sought in which other evi-
dence indicated innocence but the subject produced
a deceptive pattern on the polygraph charts. In all
cases more than one polygraph test had been con-
ducted on the suspect, and in some cases one or
more additional persons had received polygraph
examinations. In all but one case at least one of the
examinations of the suspect had been conducted at
the University of Utah.
In the 12 cases obtained for the study, three
criteria were used to justify the determination that an
error had been made. In two cases evidence obtained
subsequent to the polygraph test clearly proved that
an error had occurred, in nine cases opposing results
were obtained from two or more examinations on
the same subject, and in four cases conflicting results
were obtained from another person examined on the
same issue. In all cases the original deceptive result
with the suspects was confirmed by a numerical
evaluation by Dr. Barland or Dr. Raskin.
j $. j QQ? ~N,i 240004..;7.
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
II. RESULTS
A. Accuracy of Decisions
1. Laboratory experiments. The decisions made by
the polygraph examiner in Experiment I and Experi-
ment II are shown in Table 1. They are based on the
total numerical scores using the criterion of +6 or
larger for truthful outcomes and -6 or less for
deceptive outcomes. It can be seen that the accuracy
rates were quite high with a combined accuracy of
decisions which exceeded 90%. Approximately 10%
of the subjects yielded inconclusive results, and the
errors were almost equally distributed between false
positives and false negatives.
TABLE 1 Accuracy of Control-Question
Decisions and Types of Errors in Two Laboratory
Experiments
% %
% % False % False Incon- Correct
Correct Positive Negative clusive Decisions
Experiment 1 88 4 0 8 95
Experiment 11 80 2 8 10 89
In order to compare the effectiveness of control-
question tests in identifying quilty and innocent
subjects, the total numerical scores for the first three
charts were obtained for both types of subjects.
Those mean scores for innocent and guilty subjects
in both experiments are shown in Table 2. Statistical
analyses indicated that the technique was equally
effective in identifying innocent and guilty subjects.
TABLE 2 Total Numerical Scores for 3 Charts
Obtained in Two Laboratory Experiments
Experiment I Experimem II
Guilty - 11.1 -9.0
Innocent +9.4 +14.8
Since the criterion of requiring a score of at least
?6 in order to render a decision was developed from
the experiences of field examiners, the data from
these two experiments were combined to evaluate
the effectiveness of those cutoff points. Figure 1
presents the rates of correct decisions for guilty and
innocent subjects and the rates of inconclusives using
cutoffs for decisions which ranged from nonzero
scores up to scores exceeding ? 12 for all charts. It
can be seen that the accuracy of decisions reaches
an optimal level in the region of ?4, and inconclu-
sives are relatively low (9%) up to cutoffs of ?5.
Thus, it appears that the field practice of using
scores which fall outside an inconclusive region of
:t5 provides a good balance between accuracy and
rate of inconclusives. Furthermore, there seems to
be no compelling reason to alter that inconclusive
region.
2. Field study. The accuracy of polygraph exami-
nations with criminal suspects was evaluated using
the decisions based on the numerical scores obtained
from the independent chart interpretations. The
standard ?5 boundaries were used for the inconclu-
sive region. Those decisions were compared to the
combined judgments of the panel and also to the
judicial outcomes.
The results of the comparisons between the out-
come of the polygraph examinations and the deci-
sions based on agreement among at least a majority
of the panel are shown in Table 3. When both the
panel and the polygraph scores yielded a decision,
the polygraph outcome agreed with the majority
panel in 86% of the cases. More than half of the
suspects found truthful with the polygraph produced
inconclusive outcomes from the panel, and six of the
seven disagreements were false positives (deceptive
polygraph results on subjects considered innocent by
the panel).
TABLE 3 Comparison of Outcomes Based on
Independent Examiner's Numerical Score and
Panel Majority Decisions
Independent
Numerical
Panel Majority Decision
Evaluation
Guilty
Innocent
Inconclusive
Total
Deceptive
39
6
13
58
Truthful
1
5
7
13
Inconclusive
7
6
8
21
Total
47
17
28
92
The polygraph results were also compared to the
judicial outcomes which were considered conclusive
and were not influenced by the polygraph results.
Those results are presented in Table 4 and they
oG2 0 40044-7
_44f W
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
0.09 010i 0.0z
IN332JJd
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
indicated that there was 88% agreement between the
polygraph decisions and the judicial outcomes. All of
the disagreements (4) occurred on suspects who
produced deceptive polygraph charts and who were
acquitted in court.
TABLE 4 Comparison of Decisions Based on
Independent Examiner's Numerical Scores and
Independent Judicial Outcomes
Independent
Numerical
Evaluation
Guilty
Innocent
Total
Deceptive
27
4
31
Truthful
0
3
3
Inconclusive
6
1
7
Total
33
8
41
It is of interest to note that there was less than
complete agreement between the panel majority
decisions and the judicial outcomes. In the 35 cases
where a definite decision was made using both
criteria, there was 89% agreement between the two
criteria. Three of the four disagreements occurred
with persons judged guilty by the panel and acquitted
in court, and only three of the eight acquitted by the
judicial process were judged innocent by the panel.
Therefore, the extent of disagreement between poly-
graph outcomes and the two criteria for guilt and
innocence should be interpreted in light of the fact
that the two criteria were in less than perfect
agreement.
3. Psychopaths. Two types of data were obtained
to assess the effectiveness of examinations per-
formed on persons diagnosed psychopathic (socio-
pathic). The first consisted of the accuracy of deci-
sions obtained in Experiment I. The outcomes based
on numerical evaluation of those charts are presented
in Table 5. Among the 24 subjects who had been
diagnosed as psychopaths, decisions were 96% cor-
rect. The single error was a false positive, and not a
single guilty psychopath was able to produce a
truthful polygraph outcome. Although the polygraph
tests appeared to be slightly more effective with the
psychopaths than with the nonpsychopaths, there
was not a statistically significant difference between
the accuracy rates for the two groups.
TABLE 5 Examiner Decisions Based on Total
Numerical Scores in Experiment I
Correct Wrong Inconclusive
Psychopaths 23 1 0
Nonpsychopaths 19 1 4
The second set of data with regard to psychopaths
and polygraph outcomes was obtained from the field
study with criminal suspects. On the bases of their
MMPI scores, two groups were obtained from the 36
suspects judged to be guilty by the panel majority.
The 14 guilty suspects with the highest psychopath
scores were compared to the 12 guilty suspects with
the lowest psychopath scores. Using the total poly-
graph scores for the first three charts, the guilty
psychopaths had a mean score of -7.8, and the
guilty nonpsychopaths had a mean score of -7.1.
There was no significant difference between those
groups which indicated that the guilty psychopaths
were detected by the polygraph as readily as were
the guilty nonpsychopaths.
4. Other subject characteristics. The subjects from
the study of criminal suspects were also compared
on a number of biographical and personality varia-
bles. Those comparisons were made for sex, educa-
tion, number of previous arrests, religiousness,
previous polygraph tests, age, and the MMPI scores
for the he scale, K-scale, hypochondriasis scale, and
depression scale. There were no indications that any
of those variables were related to the polygraph
results.
5. Type of crime. The strength of polygraph
reactions exhibited by deceptive suspects was com-
pared for different crime categories. When the total
scores for the first three charts were compared for
suspects accused of sex crimes, drug crimes, crimes
of violence, and crimes of financial gain, there were
no discernible differences among the groups. A
similar analysis was performed to compare the
categories of sex crimes, drug crimes, and crimes
involving confrontation between criminal and victim,
and crimes without confrontation between the crimi-
nal and victim. Again, there were no discernible
differences among polygraph scores for deceptive
suspects divided into those categories. Thus, there
was no evidence that the type of crime affected the
strength of polygraph reactions among suspects
found deceptive on the polygraph test.
6. Behavior symptoms. The predictions based
upon the observation of behavior during the pretest
phase of the polygraph examinations of criminal
suspects were compared to the judgments of guilt
or innocence made by a majority of the panel. The
initial predictions agreed with the panel in 56% of
the cases, and the later predictions agreed with the
panel in 69% of the cases. Neither of those results
was significantly above chance, indicating that sys-
tematic observation of behavior during the pretest
phase of the polygraph examination was of no value
in determining truth or deception.
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
B. Reliability of Chart Interpretation
This project included four different assessments of
the reliability of chart interpretation. Two of those
involved comparisons between the decisions based
on numerical scoring by the Project Director and Co-
Director, the third compared decisions based on
numerical scoring by the Project Director and Co-
Director with the original decisions made by 43
examiners employed by law enforcement and private
firms, and the fourth involved a study of accuracy of
chart interpretation by 25 field examiners who eval-
uated the same set of 16 confirmed polygraph
examinations of criminal suspects. The results of
each is described below.
1. Experiment 11. The 40 sets of polygraph charts
obtained with the control question technique were
scored numerically by the original examiner (Dr.
Barland) and were independently scored by Dr.
Raskin. Both examiners made a definite decision on
36 of the 40 subjects, and they were in agreement on
100% of them.
2. Criminal suspects study. The outcomes based
on numerical scores by the original examiner (Dr.
Harland) and those based on the blind evaluaton of
the charts by Dr. Raskin are shown in Table 6. Both
examiners obtained the same categorization in 86 of
the 102 cases (84.3%) when inconclusives were
included. On cases in which both examiners made a
decision, they were in agreement 100% of the time.
The correlation between the numerical scores as-
signed by the two examiners was very high, r = .91.
TABLE 6 Comparison of the Original Examiner's
Chart Evaluation and the Blind Evaluation of the
Charts By an Independent Examiner
Original Examiner's Scores
Independent
Evaluation
Truthful
Deceptive
Inconclusive
Total
Tnithful
10
0
5
15
Deceptive
0
61
1
62
Inconclusive
1
9
15
25
Total
11
70
21
102
3. Current practices by law enforcement and
private examiners. The extent of agreement between
polygraph decisions by law enforcement and private
examiners and those based on independent numerical
evaluation by the Project Director and Co-Director is
presented in Table 7. Location G utilized the rele-
vant-irrelevant technique, and our evaluation of
those charts was subjective since numerical scoring
could not be utilized. The overall rate of agreement.
was 57% when inconclusives were included and 85%
when inconclusives were excluded. There was a
significant difference in agreement across the differ-
ent locations sampled, and our evaluations showed
significantly higher rate of agreement with decisions
made by law enforcement (92%) as compared to
private polygraph firms (79%). The results were
separated by type of crime, and there were no
significant differences in rate of agreement for crimes
against people, economic crimes, or sex and drug
crimes.
TABLE 7 Agreement Between Independent
Evaluations and Original Decisions at Each
Location
% Agreement
Including
Incondusives
Excluding
Inconclusives
I'blice
A
88.3%
98.0%
B
42.6%
95.8%
58.6%
82.1%
Combined
64.0%
92.1%
Private
D
60.0%
89.7%
E
43.3%
76.5%
F
55.0%
80.6%
G
50.8%
75.6%
Combined
52.2%
79.9%
4. Accuracy of chart interpretation. Of the 400
judgments made by the 25 polygraph examiners on
the set of 16 polygraph examinations, 79% were
correct decisions, 8% were errors, and 13% were
inconclusive. Excluding inconclusives, 90% of the
decisions were correct. Accuracy ranged from 53%
correct decisions for one examiner to 100% correct
decisions for nine examiners. The proportion of the
errors that were false positives (60.6!0) was more
than twice as high as would be expected by chance.
There was no significant difference in accuracy of
decisions for examiners with at least one year of
experience (92%) as compared to those with less
than one year of experience (89%). However, the
seven examiners who employed numerical scoring of
the charts were significantly more accurate in their
decisions (99%) than the 18 examiners who did not use
numerical scoring (88%). Furthermore, even among
examiners who had received formal training in nu-
merical scoring, the seven examiners who explicitly
employed numerical evaluation achieved significantly
higher accuracy of decisions (99%) than the six
examiners who knew how to numerically score
charts but did not explicitly employ the technique
(88%).
C. Relative Effectiveness of Standard
Measures
In four studies the standard field measures were
compared in terms of their relative effectiveness of
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
identifying truthful and deceptive subjects using
.numerical scoring of responses. That was accom-
plished in both laboratory experiments and two of
the field studies.
1. Laboratory experiments. In Experiment .I the
finger plethysmograph was used instead of a pressur-
ized cardio cuff, and the numerical scores for the
first three charts are shown in Table 8 for Experi-
ments I and II. In both experiments the skin
conductance measure provided the best discrimina-
tion between guilty and innocent subjects. In Exper-
iment I all three measures showed significant capac-
ity to identify both guilty and innocent subjects, and
the respiration component identified innocent sub-
jects better than it identified guilty subjects. In
Experiment II the plethysmograph measures pro-
duced better discrimination between guilty and inno-
cent subjects than did the cardio and respiration
measures. The cardio and respiration measures
showed significant identification of innocent but not
guilty subjects. It should be noted that respiration in
Experiment II was measured with a device different
from that typically employed in the laboratory or
field. Also, the cardio was measured using a low-
pressure cuff at an inflation pressure between 50 and
60 mmHg.
TABLE 8 Mean Numerical Scores for Each
Standard Component for the First 3 Charts in Two
Experiments
Experiment I
Experiment II
Guilty
Innocent
Guilty
Innocent
Respiration
-2.3
+4.3
+0.3
+1.8
Skin Conductance
-5.8
+3.8
-5.0
+5.6
Cardio
-
-
0.0
+3.6
Plethysmograph
-3.0
+1.3
-4.2
+3.1
2. Field studies. The three standard components
used in the study of accuracy with criminal suspects
were evaluated using the numerical scores on the
first three charts of 32 suspects who subsequently
made full confessions or pleaded guilty to the original
charge. Those numerical scores were assigned by the
independent evaluator who had no knowledge of the
confessions or pleas. All three of the components
yielded significant deceptive scores for those con-
finned guilty subjects, and the skin resistance (gal-
vanic skin response) measures produced significantly
better results than respiration or cardio measures.
Finally the three standard components were eval-
uated using the numerical evaluations of the charts
obtained from the six law enforcement and private
locations which used control-question tests. Two
types of data were utilized. First, our numerical
scores were tabulated for the 179 cases in which our
decisions agreed with the decisions made by the
original examiner. The mean scores for each compo-
nent for the 147 deceptive and 32 truthful decisions
are shown in Table 9. All measures contributed
significantly to both deceptive and truthful outcomes,
and the skin resistance measure was significantly
larger than respiration and cardio for deceptive
outcomes. The other analysis was performed on 56
deceptive results which were confirmed by confes-
sions or admissions. All three measures showed
significant identification of guilty subjects, and the
mean skin resistance score,(-5.8) was significantly
larger than that obtained for respiration (-2.9) or
cardio (-2.8).
TABLE 9 Mean Numerical Scores for Each Meas-
ure in Cases Where the Original Decision and the
Independent Numerical Evaluation Agreed
Respiration
Skin
Resistance
Candio-
vascular
Total
Deceptive (N = 147)
-3.8
-7.4
-4.7
-15.9
Truthful (N = 32)
+4.5
+4.8
+2.7
+12.0
D. Quantitative Analyses of Physiological
Responses
In order to identify the characteristics of the
various physiological measures which appear to be
useful in detecting truth and deception, detailed
quantitative analyses were performed on the poly-
graph recordings obtained in Experiments I and II
and on the 14 sets of charts using those measures
obtained from criminal suspects. Using control-ques-
tion tests, significant effects were demonstrated
when guilty subjects showed larger reactions to
relevant as compared to control questions and inno-
cent subjects showed larger reactions to control as
compared to relevant questions. Only those meas-
ures which showed some effectiveness in the labora-
tory were tested with the sample of criminal sus-
pects. Since measures of respiration and skin
conductance amplitude were used to categorize the
criminal suspects as truthful or deceptive, they were
not analyzed quantitatively.
1. Respiration amplitude. Experiment I measured
both thoracic and abdominal respiration. Both meas-
ures of respiration produced clear indications of
greater suppression in respiration amplitude following
relevant questions for guilty subjects and control
questions for innocent subjects. Furthermore, tho-
racic respiration showed an increase in amplitude
following relevant questions for innocent subjects. A
similar effect did not occur in abdominal respiration.
I 4 B4 ,OQ4-7 ,
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
No significant results were obtained with respiration
amplitude in Experiment II. However, the transducer
used in Experiment II appears to have been inade-
quate. Based on the results of Experiment I, it
appears that suppression of respiration is a clearly
effective indicator, and increase in respiration ampli-
tude should not be used as in index of deceptive
reaction but as a lack of reaction.
2. Respiration cycle time. There were significant
effects in respiration cycle time demonstrated in
Experiment I. The effects were accounted for by the
reactions of the innocent subjects who showed a
slowing in respiration following control questions and
a speeding of respiration following relevant ques-
tions. Respiration rate showed no significant effects
in Experiment II, but the poor transducer may have
prevented any significant findings. Based on the
results obtained, it appears that slowing of respiration
may be considered as a reaction, but speeding of
respiration should be viewed as a lack of reaction.
3. Skin conductance response amplitude. In both
Experiment I and Experiment II skin conductance
response amplitude was larger following relevant
questions for guilty subjects and following control
questions for innocent subjects. With guilty-knowl-
edge tests, guilty subjects produced larger responses
to critical items. Thus, skin conductance response
amplitude was found to be very effective.
4. Skin conductance response rise time. In Exper-
iment II the time required for skin conductance to
reach its maximum level from the beginning of a
response was found to be shorter to relevant ques-
tions for both guilty and innocent subjects. There-
fore, it was not useful in identifying deceptive
reactions.
5. Skin conductance response recovery. In Exper-
iment II the amount of time required for skin
conductance responses to return half of the distance
back toward their base level before the response
began was shown to be effective for two different
measurement methods. Basically, the results showed
that skin conductance responses which began after
the onset of a question recovered more slowly
following relevant questions for guilty subjects and
following control questions for innocent subjects.
However, those effects occurred only when Backster
type control questions were used and not with tests
employing Reid control questions. With guilty-knowl-
edge tests, there was some indication of slower
recovery of responses following critical items for
guilty subjects. Measurement of skin conductance
response recovery also produced significant effects
with the sample of criminal suspects. Deceptive
suspects showed slower recovery of responses to
relevant questions. Thus, the slowness with which a
skin conductance response returns to its pre-re-
sponse level may be given some consideration as an
index of response.
6. Cardio responses. The cardio responses were
measured in Experiment II with the low pressure
cuff inflated to a pressure between 50 and 60 mmHg.
The changes in diastolic pressure were similar and
somewhat better than those obtained with changes in
systolic pressure. Diastolic pressure was measured
on a second-by-second basis, and the changes in
diastolic pressure are shown in Figure 2 for the 14
seconds following the beginning of the questions.
The guilty subjects failed to show any differential
response to control and relevant questions. How-
ever, the innocent subjects showed an increase in
diastolic pressure following control questions and a
decrease in diastolic pressure following relevant
questions. The subsequent tests conducted with
criminal suspects utilized inflation pressures of ap-
proximately 70 mmHg which appear to yield better
reactions and more stable baselines. Using that
pressure with criminal suspects yielded significantly
greater increases in diastolic pressure following rele-
vant questions for deceptive suspects and following
control questions for truthful suspects. In addition,
significant decreases in diastolic pressure showed
greater decreases in diastolic pressure following
control questions, and truthful suspects showed
greater decreases following relevant questions. Since
no reliable changes in cardio pulse amplitude were
found in Experiment II, pulse amplitude changes
were not measured with the criminal suspects. How-
ever, the improved quality of recordings obtained
with 70 mmHg pressure showed some indications of
appropriate decreases in pulse amplitude. Based on
the obtained results with laboratory subjects and
criminal suspects, it seems safe to conclude that
increases in diastolic pressure are good indicators of
reactions and decreases in diastolic pressure should
be considered as lack of reaction.
7. Finger blood volume. Using a photoelectric
plethysmograph, significant changes in finger blood
volume were obtained in Experiments I and 11. In
both experiments guilty subjects showed greater
decreases in finger blood volume following relevant
questions, and innocent subjects showed larger reac-
tions following control questions. With the guilty-
knowledge test, guilty subjects showed significantly
larger reactions to critical items. Significant results
were obtained with the sample of criminal suspects,
but only the deceptive suspects showed the effects. In
Bl t? t40004-7
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
C
O o
m a a
v
a ~o
O
i
f 11
JD 49( A ~o ?
? IA
0
o .. a
N
F'yOO O (!7 ?.-- 0
% CD Cr- I- ELI IL 4-A #A
OIc 010e 0101 010 0.01- 0 0 - 0*0#-
(SIINfl 3AI1HIM) 13A31 Sfl1f1WI,LS38d W02ld NQI,LUTA3Q
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
IZ
>-I-/UL
I- .JUO
1-+MZZ
=CDZ--+
CD --~
J Z J Z
000cc
F uF-4 L.I
z J z J
OLOOU
L D L
11 Ii II II
1 1
411
0
0
O
ri
t'1 0'1 60 80 G'0
30f1ZI'1dwd Sfl'1f1WIZS3dd A NOI18OJ08d
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Experiment 11 the duration of responses was longer to
relevant questions for guilty subjects and to control
questions for innocent subjects. Duration effects were
nob obtained with the criminal suspects. On the basis of
the obtained results, it is clear that decrease in finger
blood volume is a good indication of reaction, and
duration of reaction may also be of some use.
8. Finger pulse amplitude. The amplitude of pulses
was obtained from the same plethysmograph used
for blood volume by recording with a short time
constant coupling. In Experiment I guilty subjects
showed significantly larger decreases in pulse ampli-
tude following relevant questions, but for innocent
subjects there was no difference in magnitude of
pulse amplitude changes following control and rele-
vant questions. In Experiment II the changes in
pulse amplitude were measured on a second-by-
second basis for the 14 seconds following the begin-
ning of the questions, and the results are shown in
Figure 3. The guilty subjects showed significantly
greater decreases in pulse amplitude following rele-
vant questions, and the innocent subjects showed
longer-lasting decreases following control questions.
With guilty-knowledge tests, significantly greater de-
creases occurred to critical items for guilty subjects.
The measurements of decrease in pulse amplitude in
the criminal suspects showed significantly larger
decreases in pulse amplitude following relevant ques-
tions for deceptive suspects, but no differences
between reactions to control and relevant questions
were obtained for truthful suspects. Based on the
obtained results, decrease in finger pulse amplitude
can be considered a useful index of reaction.
9. Heart rate changes. In Experiment I heart rate
in beats per minute was measured on a second-by-
second basis from prior to the beginning of the
questions through 20 seconds following question
onset. The changes in heart rate are shown in Figure
4. Guilty and innocent subjects showed an initial
increase in heart rate to control and relevant ques-
tions. Following that initial increase, heart rate
returned to previous levels except for guilty subjects
following their answer to relevant questions. They
showed a clear slowing of heart rate which was more
pronounced for the psychopathic group. In Experi-
ment 11 virtually identical results were obtained with
guilty subjects producing slowing of heart rate fol-
lowing their answers to relevant questions and inno-
cent subjects showing no differences in heart rate
responses to control and relevant questions. How-
ever, analysis of heart rate responses by the criminal
suspects failed to produce any differentiation be-
tween control and relevant questions. Although the
laboratory results appear to be encouraging with
regard to slowing of heart rate as an indicator of
reaction in guilty subjects, the results with criminal
suspects failed to produce a similar result. At this
time, the use of heart rate slowing as an index of
reaction should be viewed with caution.
10. Cardio activity monitor. In Experiment II the
systolic and diastolic levels obtained, from the cardio
activity monitor (CAM) were measured on a second-
by-second basis from just prior to the beginning of
the question through 14 seconds following question
onset. Only the changes in systolic level were
significant, and those results are presented in Figure
5. There was an early increase in systolic levels
which did not differentiate between control and
relevant questions for guilty or innocent subjects.
However, there was a subsequent decrease in sys-
tolic levels which appeared to be greater for guilty
subjects in response to relevant questions. A similar
but substantially stronger result was obtained with
the guilty knowledge test, and those data are pre-
sented in Figure 6. There was an early, non-differ-
ential increase in systolic level followed by a very
pronounced decrease in guilty subjects responding to
critical items. Those results are very similar to the
results obtained with finger blood volume and finger
pulse amplitude as measured by a photoelectric
plethysmograph. However, CAM measures of
changes in systolic levels failed to produce any
significant results with the sample of criminal sus-
pects. At this time, it appears that the CAM meas-
ures do not provide very much useful information,
and what is provided by the CAM can be better
obtained from a photoelectric plethysmograph.
11. Skin potential responses. The negative and
positive components of the skin potential response
were measured in Experiments I and II. In Experi-
ment I the significant results indicated that guilty
subjects gave larger negative and positive skin poten-
tial responses to relevant questions than to control
questions, but there was no differentiation for inno-
cent subjects. There was also a significant effect for
psychopathy which consisted of psychopaths produc-
ing positive skin potential responses which were
disproportionately larger to relevant questions than
those produced by nonpsychopaths. In Experiment
II significant effects were obtained with the control-
question technique only for negative skin potential,
and the results with mean negative skin potential
amplitude are presented in Table 10. They show that
with Backster control questions the guilty subjects
produced larger responses to relevant questions and
141-T 77T~1-
1 ~? 0 95 ~4. 04-7
0 / 1N
N
c 00
W 0 0
cD p ~, hi co
v O ? {$I 4`r
Q
Q
-I
I
,p
OD c ! -100
I I I I I I I I 1 1 1 1
W V N O N tt
11 I I I I I I I 1 I la
tD N 0 N IT
N P * 1N
OLD 0 c 00
C
o d
1 o ? -b~L -V
lca~j N
la-
(D tf N O N 'T t0 tt N O N Cf
i
0
z
0
(INd8) 3TV8 i8V3H NI 3JNVHO
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
O
t~ J
o
F N
O c
JZ F a
E-+ O O o (n ^
JZc> O ?&
OOEyL3 ^ o
w>Z
FLOC] 0 -
0[CJ Hd
O>->-W L] O a+
Z E'EyUU v 10
zz
L-1 D
J U' Q) N
II II II 11
.9
O
0'M- 0'01- 0'09-
(SIINf 3AIId'1321) '13A371 SIflf1WIIS3dd W0 J NDI.LUTA20
APQ ARM. 4, QOQ47.7,:
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
G N M
It 1 .-~
1= 1 W
pC7[~ ?. u
CLO
U 1 ti
It 11 0
RC W c z
y O +'
1~ W M u
R f I ~_
i ~ 1 u
mCf) i:01
J L
CC 0 1 -1
1" _ 0-+ U ? (n -
14 IY I--i b-I u
~i -J U co H w
mIp ol>- >-
kh ww ?
Z HE-IUU
zu E" 1-4 In u
%.j LL q a.+
Z1-~ZU ? p-
C) w H-
o ~+ V
E
z
(D CD--1--i N
Rlet ^ D ^ D n b
1
T I L O W
O OZ 0 *0 O.OZ O.1Oi 0 09
(SIINf 3AIZd71331 '13A371 SfT1f1WIJ.S32ld WOZIJ NOIIUTA30
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
the innocent subjects produced larger responses to
control questions. No differentiation was obtained
with `Reid control questions. The guilty-knowledge
tests also produced significantly larger negative and
positive skin potential responses by guilty subjects to
critical items. The measurement of skin potential
responses with criminal suspects showed larger am-
plitude of positive responses by deceptive suspects
to relevant questions, but no significant results were
obtained with negative skin potential. Thus, it ap-
pears that measures of skin potential may be of some
use as an index of deception, but they appear to be
inferior to other measures of electrodermal activity
such as skin conductance amplitude and recovery
time.
TABLE 10 Mean Negative Skin Potential Re-
sponse Amplitude (mV) with the Control-Question
Technique and Guilty -Knowledge Technique in Ex-
periment 11
Control Question
Control
Relevant
Guilty
1.0
1.7
Innocent
1.4
1.1
Guilty
1.1
1.2
Innocent
.9
.8
Guilty Knowledge
Guilty
Noncritical
1.3
Critical
2.5
Innocent
1.4
1.2
E. Comparisons of Question Techniques
Laboratory Experiment II was designed to allow
evaluations of several different approaches to ques-
tion structure employed in polygraph examinations.
Those features included comparisons of the effective-
ness control-question tests with that of guilty-knowl-
edge tests, evaluations of the relative effectiveness
of Backster and Reid types of control questions, and
the relative usefulness of a guilt-complex question
utilized as a control question. The results obtained
with each of those questions are described below.
1. Relative accuracy of control-question and
guilty-knowledge tests. The outcomes based on nu-
merical scoring of control-question and guilty-knowl-
edge tests are pmsented in Table 11. The rate of
accuracy of decisions using control-question tests
was 94% with Backster type control questions and
8317r, with Reid type control questions. Since the
difference in accuracy rate for those two types was
not significant, their combined accuracy rate was.
compared to that obtained with the guilty-knowledge
test. The accuracy of guilty-knowledge tests was
90% as compared to the combined decision accuracy
of 89% with control-question tests. The types of
errors which occurred consisted of false negative
errors in all but one subject who was legUvl will, II
control llllt?Nlhol ti?t.htdllu,.. '1'llirt, IItF lop-Bill tv?1-11
racy of decisions was virtually identical for contrul-
question and guilty-knowledge tests. In addition,
quantitative analyses of physiological responses pro-
duced a large number of significant results using both
control-question and guilty-knowledge tests as previ-
ously described.
TABLE 11 Accuracy of Decisions and Types of
Errors Using Backster and Reid Control Questions
and Guilty-Knowledge Tests in Experiment 11
%
Correct
% False
Positive
% False
Negative
Incon?
clusive
Correct
Decisions
Backster
85
0
5
10
94
Reid
75
5
10
10
83
Guilty-Knowledge
90
0
10
0
90
2. Comparison of Backster and Reid control
questions. As indicated above, the tests which uti-
lized Backster control questions produced a slightly
higher accuracy rate than those which utilized Reid
control questions, but the difference was not statisti-
cally significant. However, when the total numerical
scores for the first three charts were compared,
some differences were observed. The tests using
Backster control questions produced significant iden-
tification of innocent (mean score = + 13.6) and
guilty (mean score = -11.7) subjects, but the results
with Reid control questions were significant for
innocent (mean score = + 14.2) but not guilty (mean
score = -6.3) subjects. Quantitative analyses of
physiological responses also produced some results
which indicated a superiority for tests utilizing
Backster control questions. As previously described,
measures of skin conductance response recovery
times and amplitude of negative skin potential re-
sponses showed stronger reactions to relevant ques-
tions by guilty subjects and to control questions by
innocent subjects only with Backster control ques-
tions. The test which utilized Reid control questions
showed no discrimination for either of those meas-
ures. Thus, it appears that control questions which
are separated from the relevant issue by age or tin'._of occurrence have some advantages over control
questions which do not have those exclusionary
characteristics.
3. Effectiveness of a guilt-complex question. In
order to evaluate the usefulness of a guilt-complex
question as a control question, Experiment 11 in-
cluded a guilt-complex question as the eighth ques-
tion on each chart. The reactions to that question
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
were evaluated by using it as a control question for
the relevant question at position 7 and performing a
standard -numerical scoring for that question pair on
the first three charts. The scores obtained in that
manner were then compared to those obtained by a
numerical evaluation using the control question at
position 9 and the same relevant question. The
results of those evaluations are presented in Table
12. When the reactions to the standard control
question were compared to those produced by the
relevant question, the mean scores for guilty and
innocent subjects were of approximately the same
magnitude but in opposite directions, showing signif-
icant discrimination for guilty and innocent subjects.
The guilt-complex question produced a negative
mean score for guilty subjects but failed to produce
any difference for the innocent subjects. In addition,
the frequencies of scores in the wrong direction
(positive scores for guilty subjects and negative
scores for innocent subjects) were tabulated. With
the guilt-complex question, scores of "0" with guilty
subjects were considered to be in the wrong direction
since the theory of guilt-complex questions predicts
that only innocent subjects would produce scores of
"0." Using standard control questions, there were
three scores in the wrong direction with guilty
subjects and three scores in the wrong direction with
innocent subjects. With the guilt-complex question,
there were three such errors with guilty subjects and
nine with innocent subjects. Thus, it appears that the
standard control questions were clearly more effec-
tive than guilt-complex questions in identifying inno-
cent subjects.
TABLE 12. Mean Scores of Guilty and In-
nocent Subjects when a Control Question and the
Guilt-Complex Question were Compared with a
Relevant Question
Comparison Question
Guilty
Innocent
Control Guilt Complex
-4.2 -3.7
4.0 0.0
F. The "Friendly Polygrapher"
Three sets of data were obtained in order to
evaluate Orne's "friendly polygrapher" hypothesis.
The first sample showed that defense cases produced
78% truthful, 20% deceptive, and 2% inconclusive
outcomes. The law enforcement cases produced 76%
truthful, 20% deceptive, and 5% inconclusive out-
comes. Contrary to the "friendly polygrapher" hy-
pothesis, there was no difference in frequency of
truthful outcomes for defense and law enforcement
examinations conducted by the same examiner. The
second sample produced mean numerical scores of
-4.7 for defense cases and -2.0 for law enforce-
ment/employer cases. Although the difference be-
tween those means was not significant, it was in the
opposite direction from that predicted by the
"friendly polygrapher" hypothesis. The third sample
produced mean numerical scores of -10.4 for de-
fense cases and -0.7 for law enforcement cases. The
difference between those means was statistically
significant and in the opposite direction from that
predicted by the "friendly polygrapher" hypothesis.
Thus, the three samples of data obtained to test the
predictions from the "friendly polygrapher" hypoth-
esis not only failed to produce any evidence to
support that hypothesis, but some of the results
indicated effects which were totally contrary to
Orne's speculations.
G. Errors in Examinations
This project provided four different sources of
possible data concerning errors. The first was the
two laboratory experiments utilizing mock crimes.
The second was the field study of criminal suspects
using criteria of ground truth developed with a panel
of experts. The other two sources consisted of
independent evaluations by field examiners of con-
firmed cases provided by us and cases from our own
laboratory and those referred to us by other exam-
iners.
In the two laboratory experiments, there was a
total of eight errors in 108 subjects. They consisted
of three false positives and three false negatives with
control-question tests and two false negatives with
the guilty-knowledge test. The laboratory results
seem to indicate a-low rate of errors equally divided
among false positives and false negatives using
control question tests and only false negatives using
the guilty knowledge test.
The results obtained from examinations of criminal
suspects were somewhat different. Using the panel
criterion, there were six false positives and one false
negative with 92 suspects. When polygraph exam-
iners made 400 independent evaluations of polygraph
charts of confirmed cases from our laboratory, 20 of
the 33 errors were false positives. Only eight false
positives would be expected if examiners were
equally likely to make either type of error when
interpreting charts.
The last sample concerning errors yielded 12 cases
in which there was clear evidence of inappropriate
physiological responses on one of the polygraph
examinations. It appears that all but one of the cases
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
yielded false positive results on the first examination,
will orx.. suspect produced ,r false r:cg ttive result.
'110. 111(:14)udtJance of fals positives is not surprising
eirx:t: a guilty subject would be unlikely to insist that
a truthful outcome was erroneous. The one case of a
false negative appeared to be the result of delit raw
countermeasures which produced substantial respi-
ration reactions to control questions. A subsequent
examination at our laboratory confirmed the decep-
tion and the obvious use of countermeasures. Among
the false positive results, four were resolved by a
truthful result on a subsequent polygraph test which
followed a restructuring of some of the questions in
order to separate a related but irrelevant concern
expressed by the suspect. Such concerns were
typically incorporated into new control questions or
simply expressed by the subject as minor admissions
prior to the second test.
There were a number of characteristics which
seemed to be prevalent among the false positive
cases. In all but one case, the suspect had no
previous experience of being in serious difficulty
with regard to criminal activity. They were generally
well-educated (six had college degrees), had middle-
class values, and expressed strong concerns about
their reputations and their personal distress over
being charged with a criminal act. 'i hal pattern ut
characteristics is not typical of criminal suspects who
are given polygraph examinations. In five of the
cases, the suspect was examined twice by Dr. Raskin
or Dr. Barland and produced deceptive results on the
first test. After being informed of the outcome, each of
them described something which caused an emotional
reaction to the relevant questions. In three instances
minor, non-incriminating admissions were made, and
the second test produced truthful results with the
same relevant questions. In the other two cases, the
suspects described a feature of the situation which
caused them to experience an emotional reaction to
the relevant questions, e.g., the use of a certain name
or feelings of guilt or responsibility. When such
material was incorporated into control questions,
truthful results were obtained on the second test.
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
Ill. Conclusions and Recommendations
The results of this project clearly indicate that
polygraph examinations utilizing control-question or
guilty-knowledge tests are highly accurate. In light of
the available evidence from the laboratory and field,
it seems reasonable to conclude that the accuracy of
such tests is approximately 90% when they are
properly conducted and evaluated.
With regard to specific techniques, it appears that
the control-question test utilizing Backster type con-
trol questions is the most accurate test which is
suitable for a wide variety of criminal investigations.
Although the guilty-knowledge test also produced
high levels of accuracy, it is seldom possible to
utilize that type of test due to the nature of most
crimes or the lack of significant items of information
which would be known only to the guilty person and
the investigators. The relevant-irrelevant test pos-
sesses many weaknesses (Podlesny & Raskin, in
press) and should not be used as a substitute for
control-question tests.
The results of this project clearly indicate that
numerical scoring of polygraph charts produces
higher rates of accuracy and reliability of chart
interpretation than other methods of chart interpre-
tation. The basic scoring system taught by the U.S.
Army Military Police School seems to produce good
results using scores of ?5 inclusive to define the
inconclusive zone. However, the criteria for evaluat-
ing reactions may require some slight modifications
on the basis of the results obtained from the studies
of physiological responses performed on this project.
The results obtained with detailed analyses of
physiological measures support the continued use of
respiration, skin conductance (galvanic skin re-
sponse), and cardiovascular (relative blood pressure)
measures. Among all of the variables measured, the
skin conductance response (galvanic skin response)
was clearly superior in laboratory experiments and
with criminal suspects in field situations. In addition,
the results clearly indicated that a properly-designed
photoelectric plethysmograph would make a useful
addition to field polygraph instruments. Unfortu-
nately, those which have been marketed in the past
have not met the performance standards which are
easily attainable. The substitution of an electronic,
low-pressure blood pressure (cardio) device for the
mechanical, high-pressure system seems clearly ben-
eficial and desirable. The results obtained with re-
cordings made at pressures of 70 mmHg were quite
good, and the use of inflation pressures below
diastolic blood pressure have the clear advantages of
reducing discomfort to the subject and allowing a
slower rate (25-30 seconds) of question presentation.
Although some positive findings were obtained with
the cardio activity monitor (CAM) and measures of
skin potential and heart rate, the results do not seem
to warrant adding those measures to field polygraph
instruments at this time.
There were a number of findings which have
implications concerning the criteria for defining re-
actions and lack of reactions in different physiologi-
cal measures. With regard to respiration, the results
support the use of suppression of breathing, slowing
of rate, and apnea as indicators of reaction. Speeding
of breathing and increases in respiration amplitude
were found to be associated with truthfulness and
should be viewed as indicative of lack of reaction.
Although no quantitative analyses were made on
respiration baseline changes, their use in obtaining
accurate results with numerical scoring was consist-
ent with an interpretation of baseline arousal as a
reaction. However, detailed analyses should be done
to assess the usefulness of baseline arousal as an
indicator of reaction.
Measures of the recovery times of skin conduct-
ance responses clearly demonstrated that longer-
lasting responses are associated with reactions.
Therefore, additional weight may be given to scoring
skin conductance responses which show slower
recovery toward baseline levels. Since the short time
constant measurement technique which is employed
in the automatic mode of recording such responses
has the effect of eliminating information concerning
recovery time, skin conductance (galvanic skin re-
sponse) should always be recorded in the manual
(long time constant) mode. Also, the use of the
automatic mode can greatly distort the relative size
and shape of those responses, and the use of
automatic mode should be eliminated.
The findings with regard to cardio (relative blood
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
pressure) responses indicate that increases in dia-
stolic level represent good indications of reaction.
-Furthermore, decreases in diastolic level were found
to be associated with truthfulness and should not be
used as indicative of reactions. There was little
support for the use of decreased pulse amplitude as
an indicator, but the measurements with the low
pressure system may have greatly diminished the
possibility of finding such changes. Therefore, at this
time we do not recommend eliminating the use of
decrease in pulse amplitude as an indicator of
reaction.
The measures of finger pulse amplitude and finger
blood volume provided excellent results. It seems
clear that decreases in finger pulse amplitude and
finger blood volume are very useful indicators of
reaction. Since those measures are obtained by
recording with greatly different time constants, man-
ufacturers of field equipment should consider provid-
ing a photoelectric plethysmograph which offers a
choice between a very short time constant (.1
second) and a very long time constant (28 seconds).
With that option, the examiner could select the
setting which provides the more useful recording
with each subject.
The changes in physiological activity obtained on
this project were measured from the beginning of the
questions, and reactions were frequently observed
beginning shortly after the question started and prior
to the subject's answer. Although many field exam-
iners are of the opinion that reactions which begin
prior to the subject's answer are "listening reac-
tions" and should not be utilized in arriving at
decisions of truth or deception, the scientific evi-
dence supports the use of reactions which begin soon
after the start of the question. The use of such
reactions would be a problem only if the questions
had not been reviewed with the subject such as in
the current practice with relevant-irrelevant tests,
but such tests do not meet scientifically acceptable
standards.
The investigation of problems associated with
personality and psychopathy yielded a good deal of
useful information. The results were strongly contra-
dictory to the common belief that psychopaths
(sociopaths) can "beat the lie detector" (Barland &
Raskin, 1973). With convicted felons who were
diagnosed psychopathic, not a single guilty subject
was able to produce a truthful result. In fact, there
were some indications that psychopaths may be
somewhat easier to detect using polygraph examina-
tions. The results with criminal suspects supported
the position that deceptive psychopaths are as phys-
iologically reactive and as readily detected as non-
psychopaths. Thus, the fears that psychopathic
criminals are able to be successful in deception during
polygraph examinations can be dispelled. Perhaps the
greatest danger is that a clever and convincing
psychopath can talk a polygraph examiner into
believing him, even though the polygraph charts
indicate deception. Adequate training in chart
interpretation and numerical scoring should prevent
that from occurring.
With regard to a variety of personality, biographi-
cal, and circumstantial factors, the results failed to
show any relationship between those variables and
the polygraph outcomes. There were no differences
attributable to aspects of personality as measured by
the MMPI, age, sex, previous arrests or polygraph
examinations, educational attainment, or the type of
crime involved. Thus, in the absence of very low
intelligence or any incapacitating psychological or
physical illness, it seems reasonable to conclude that
polygraph examinations are effective with a wide
variety of individuals with respect to the broad range
of crimes typically investigated.
With regard to the risk of errors, the findings
provided information on important questions which
have been raised. The concept of the "friendly
polygrapher" (Orne, 1975) has been used as an
argument against the validity of polygraph examina-
tions conducted confidentially at the request of
defense attorneys (Dogin, 1974). The findings ob-
tained with three different samples of criminal cases
are contrary to the "friendly polygrapher" notion.
There appears to be no increased risk of false
negatives under such circumstances. If anything, the
results indicated a higher likelihood of deceptive
results on defense-conducted examinations. There-
fore, the results of defense-offered polygraph exami-
nations should be accorded careful consideration if
the examinations have utilized adequate techniques
employed by properly trained and competent exam-
iners.
The findings showed that in the criminal suspect
situation there may be a somewhat higher risk of
false positives rather than false negatives. In the
study with criminal suspects using the panel crite-
rion, all but one of the obtained errors were false
positives. In addition, the chart interpretation errors
by polygraph examiners showed a disproportionate
number of false positives decisions, which is consist-
ent with the report of Horvath (1974). Finally, among
the criminal cases in which we were able to obtain
fairly strong indications of an error, all but one
appeared to be false positives. In the latter study,
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
the increased risk of false positives occurred with
suspects who were generally well-educated, had no
prior history of criminal activity, and were very
concerned about their reputation or standing in the
community. When such subjects proclaim their in-
nocence after having produced deceptive reactions
on polygraph examinations, the examiner should
provide opportunities for the subject to explain why
he responded to the relevant question if he was
actually being truthful. If the subject provides an
explanation which could be incorporated into a
restructured set of questions, the examiner should
give serious consideration to administering a second
test. That procedure should be followed in all cases
where the deceptive results are strongly disputed by
the subject, but particular attention should be paid in
those cases where the subject fits the pattern de-
scribed above.
The usefulness of behavioral cues (Horvath, 1973;
Reid & Inbau, 1966) was investigated in the study of
accuracy with criminal suspects. The results were
not supportive of the claims that behavioral obser-
vations are effective in assessing truth and deception.
Similar results were obtained in Experiment II, and
they showed that decisions based on behavioral cues
produced more that 50% incorrect designations of
innocent subjects as deceptive. Unfortunately, many
examiners are taught to place great emphasis on
gestures, verbal behavior, and mannerisms in arriv-
ing at a decision. At this time the evidence does not
support such procedures, and examiners should
restrict their basis for decisions to the physiological
recordings on the polygraph charts.
On the basis of the existing scientific literature and
the findings of the research performed on this
project, the following conclusions and recommenda-
tions are made:
? Polygraph examinations can be highly accurate
in determining truth and deception regarding
specific issues in criminal investigations.
? In order to obtain maximum accuracy, poly-
graph examinations should employ control-ques-
tion tests or guilty-knowledge tests, when appli-
cable. Control questions should employ time or
age exclusions of the type utilized in the Backs-
ter control questions.
? Polygraph recordings should include respiration,
skin conductance (galvanic skin response), and
relative blood pressure (cardio cuff). Manufac-
turers of polygraph instruments are urged to
develop and market an adequate photoelectric
finger plethysmograph, and examiners are en-
couraged to use measures of finger pulse ampli-
tude and finger blood volume. The use of an
electronic low-pressure cardio cuff is encour-
aged, using inflation pressures of 70 mmHg.
Lower inflation pressures appear to produce less
satisfactory results.
? Examiners should be given formal training in
numerical scoring of polygraph charts. The nu-
merical system should be of the type utilized by
the U.S. Army and modified on the basis of the
findings described above. The use of a total
score of +6 or higher as a criterion for truthful-
ness and -6 or lower as a criterion for deception
is supported by this research.
? The results of control-question examinations
should always be determined by numerical eval-
uation of the polygraph charts. When important
investigative or judicial decisions may be influ-
enced by the results of such tests, an additional
numerical evaluation of the polygraph charts by
an independent examiner is recommended.
? If the above conditions are met, the results of
this research indicate an accuracy rate of ap-
proximately 90% with criminal suspects. In light
of that high rate of accuracy, the results of
polygraph examinations should be given careful
consideration in criminal investigations and judi-
cial proceedings.
Approved For Release 2007/06/05: CIA-RDP85BOO263ROO0100240004-7
IV. REFERENCES
Backster, C. Technique Fundamentals of the Tri-Zone Poly-
graph Test. New York: Backster Research Foundation, 1%9.
Bailey, F. L., & Rothblatt, H. B. Investigation and Preparation
of Criminal Cases, Federal and State. Rochester, New
York: The Lawyers Cooperative Publishing Company, 1970,
296-299.
Barland, G. H., & Raskin, D. C. Detection of deception. In W.
F. Prokasy & D. C. Raskin (Eds.), Electrodermal Activity in
Psychological Research. New York: Academic Press, 1973.
Barland, G. H., & Raskin, D. C. An evaluation of field
techniques in detection of deception. Psychophysiology,
1975,12,321-330.
Barland, G. H., & Raskin, D. C. Validity and Reliability of
Polygraph Examinations of Criminal Suspects. (Report No.
76-1, Contract 75-N1-99-0001, U.S. Department of Justice).
Salt Lake City, Utah: University of Utah, Department of
Psychology, March, 1976.
Bersh, P. J. A validiation study of polygraph examiner judge-
ments. Journal of Applied Psychology, 1969, 53, 399-403.
Dogin, H. S. Statement of Henry S. Dogin, Deputy Assistant
Attorney General, Criminal Division, Department of Justice.
In, The Use of Polygraphs and Similar Devices by Federal
Agencies (Hearings before a Subcommittee of the Committee
on Government Operations, House of Representatives,
Ninety-third Congress, Second Session, June 4 and 5, 1974).
Washington: U.S. Government Printing Office, 1974, 413-
417.
Horvath, F. S. Verbal and nonverbal clues to truth and
deception during polygraph examinations. Journal of Police
Science and Administration, 1973, 1, 138-152.
Horvath, F. S. The Accuracy and Reliability of Police Poly-
graphic ("Lie Detector") Examiners' Judgments of Truth
and Deception: The Effect of Selected Variables. Unpub-
lished doctoral dissertation, Michigan State University, 1974.
Horvath, F. S., & Reid, J. E. The reliability of polygraph
examiner diagnosis of truth and deception. The Journal of
Criminal Law, Criminology and Police Science, 1971, 62,
276-281.
Horvath, F. S., & Reid, J. E. The polygraph silent answer test.
Journal of Criminal Law and Criminology, 1972, 63, 285-
293.
Hunter, F. L., & Ash, P. The accuracy and consistency of
polygraph examiners' diagnoses.. Journal of Police Science
and Administration, 1973,1, 370-375.
Lykken, D. T. The GSR in the detection of guilt. Journal of
Applied Psychology, 1959,43, 385-388.
Lykken, D. T. Psychology and the lie detector industry.
American Psychologist, 1974,29,724-739.
Orne, M. T. Implications of laboratory research for the
detection of deception. In N. Ansley (Ed.), Legal Admissibil-
ity of the Polygraph. Springfield, Illinois: Thomas, 1975.
Podlesny, J. A., & Raskin, D. C. Physiological measures and
the detection of deception. Psychological Bulletin, in press.
Podlesny, J. A., Raskin, D. C., & Barland, G. H. Effectiveness
of Techniques and Physiological Measures in the Detection
of Deception. (Report No. 76-5, Contract 75-NI-99-0001,
U.S. Department of Justice). Salt Lake City, Utah: Univer-
sity of Utah, Department of Psychology, August, 1976.
Raskin, D. C. Psychopathy and Detection of Deception in a
Prison Population. (Report No. 75-1, Contract '75-N1-99-
0001, U.S. Department of Justice). Salt Lake City, Utah:
University of Utah, Department of Psychology, June, 1975.
Raskin, D. C. Reliability of Chart Interpretation and Sources
of Errors in Polygraph Examinations. (Report No. 76-3,
Contract 75-NI-99-0001, U.S. Department of Justice). Salt
Lake City, Utah: University of Utah Department of Psychol-
ogy, June, 1976.
Raskin, D. C., & Barland, G. H. An Evaluation of Polygraph
Techniques Currently Practiced by Law Enforcement and
Private Polygraph Examiners. (Report No. 76-2, Contract
75-NI-99-0001, U.S. Department of Justice). Salt Lake City,
Utah: University of Utah, Department of Psychology, April,
1976.
Reid, J. E., & Inbau, F. E. Truth and Deception: The
Polygraph ("Lie Detector") Technique. Baltimore: Williams
& Wilkins, 1%6.
Slowik, S. M., & Buckley, J. P. Relative accuracy of polygraph
examiner diagnosis of respiration, blood pressure, and GSR
recordings. Journal of Police Science and Administration,
1975,3,305-309.
A 4,0004-?7.
Approved For Release 2007/06/05: CIA-RDP85B00263R000100240004-7
U.S. DEPARTMENT OF JUSTICE
LAW ENFORCEMENT ASSISTANCE ADMINISTRATION
WASHINGTON, D.C. 20531,
OFFICIAL BUSINESS
PENALTY FOR PRIVATE USE, $300
POSTAGE AND FEES PAID
.S. DEPARTMENT OF JUSTICE