VOLUME 44 2015 NUMBER 2 Polygraph

Full text

VOLUME 44 2015 NUMBER 2
Pol
yg
raph
Contents
Published Semi-annually
© American Polygraph Association, 2015
P.O. Box 8037, Chattanooga, Tennessee 37414-0037
Laboratory Study of a Diagnostic Polygraph Technique in a Single
Sequence: a replication study
Rodolfo Prado, Carlos Grajales, Raymond Nelson
A Literature Review of Polygraph Countermeasures and the
Comparison Question Technique
Mark Handler, Charles Honts and Walt Goodson
A Book Review of Investigative Interview- The Essentials Edited by
Michel St-Yves, Published in 2014 by Carswell Publications
Mark Handler
Police Cadet Attrition and Training Performance Outcomes
Adam Park, James S. Herndon
Bonferonni and Šidák Corrections for Multiplicity Effects with
Subtotals Scores of Comparison Question Polygraph Tests
Raymond Nelson
Letter to the Editor Regarding article by Nelson and Handler
entitled Statistical Reference Distribution for Comparison Question
Polygraphs
James Allan Matte
Response to James Allan Matte Letter to the Editor
Raymond Nelson
Sexual History Disclosure and Sex Offender Recidivism
James Edward Konopasek , Raymond Nelson
A new paradigm for the experimental study of Malintent
Charles Honts
1
13
24
27
46
52
55
56
71

©American Polygraph Association, 2015
Polygraph
Editor-in-Chief: Mark Handler
E-mail: Editor@polygraph.org
Managing editor: Nayeli Hernandez
E-mail: polygraph.managing.editor@gmail.com
******
Associate Editors: Réjean Belley, Ben Blalock, Tyler Blondi, John Galianos, Don Grubin,
Maria Hartwig, Charles Honts, Matt Hicks, Scott Hoffman, Don Krapohl, Thomas Kuczek,
Mike Lynch, Ray Nelson, David Raskin, Stuart Senter, and Cholan V.
APA Officers for 2015-2016
President – Walt Goodson
E-mail: president@polygraph.org
Chairman – Raymond Nelson
E-mail: chair@polygraph.org
President-Elect – Patrick O'Burke
E-mail: presidentelect@polygraph.org
Director – George Baranowski
1912 E US Hwy 20, Suite 202
Michigan City, IN 46340
E-mail: directorbaranowski@polygraph.org
Vice-President, Private –
Gary F. Davis
E-mail: vp-private@polygraph.org
Director – Donnie Dutton
Vice-President, Government -
E-mail: directordutton@polygraph.org
Darryl Starks
E-mail: vp-government@polygraph.org
Director – Steve Duncan
E-mail: directorduncan@polygraph.org
Vice-President, Law Enforcement –
Daniel Violette
E-mail: vp-lawenforcement@polygraph.org
Director – Jamie McCloughan
E-mail: directormccloughan@polygraph.org
Treasurer - Chad Russell
Director – Barry Cushman
E-mail: treasurer@polygraph.org
109 Middle Street
Portland, ME 04101
E-mail: directorcushman@polygraph.org
General Counsel – Gordon L. Vaughan
E-Mail: generalcounsel@polygraph.org National Office Manager – Lisa Jacocks
Seminar Chair – Michael Gougler
Phone: 800-APA-8037; (423)892-3992
E-Mail: seminarchair@polygraph.org
E-mail: manager@polygraph.org
Subscription information: Polygraph is published semi-annually by the American Polygraph Association. Editorial
Address is Editor@polygraph.org. Subscription rates for 2015: One year $150.00 (Domestic), $180.00 (Foreign).
Change of address: APA National Office, P.O. Box 8037 Chattanooga, TN 37414-0037. THE PUBLICATION OF AN
ARTICLE IN POLYGRAPH DOES NOT CONSTITUTE AN OFFICIAL ENDORSEMENT BY THE AMERICAN
POLYGRAPH ASSOCIATION.

Instructions to Authors
Scope
The journal Polygraph publishes
articles about the psychophysiological
detection of deception, and related areas.
Authors are invited to submit manuscripts
of original research, literature reviews, legal
briefs, theoretical papers, instructional
pieces, case histories, book reviews, short
reports, and similar works. Special topics
will be considered on an individual basis. A
minimum standard for acceptance is that
the paper be of general interest to
practitioners, instructors and researchers of
polygraphy. From time to time there will be
a call for papers on specific topics.
Manuscript Submission
Manuscripts must be in English, and
may be submitted, along with a cover letter,
on electronic media (MS Word). The cover
letter should include a telephone number,
and e-mail address. All manuscripts will be
subject to a formal peer-review. Authors
may submit their manuscripts as an e-mail
attachment with the cover letter included in
the body of the e-mail to:
Editor@polygraph.org
As a condition of publication,
authors agree that all text, figures, or other
content in the submitted manuscript is
correctly cited, and that the work, all or in
part, is not under consideration for
publication elsewhere. Authors also agree to
give reasonable access to their data to APA
members upon written request.
Manuscript Organization and Style
All manuscripts must be complete,
balanced, and accurate. Authors should
follow guidelines in the Publications Manual
of the American Psychological Association.
The manual can be found in most public
and university libraries, or it can be ordered
from: American Psychological Association
Publications, 1200 17th Street, N.W.,
Washington, DC 20036, USA. Writers may
exercise some freedom of style, but they will
be held to a standard of clarity,
organization, and accuracy. Authors are
responsible for assuring their work includes
correct citations. Consistent with the
ethical standards of the discipline, the
American Polygraph Association considers
quotation of another’s work without proper
citation a grievous offense. The standard for
nomenclature shall be the Terminology
Reference for the Science of Psycho-
physiological Detection of Deception (2012)
which is available from the national office of
the American Polygraph Association. Legal
case citations should follow the West
system.
Manuscript Review
An Associate Editor will handle
papers, and the author may, at the
discretion of the Associate Editor,
communicate directly with him or her. For
all submissions, every effort will be made to
provide the author a review within 4 weeks
of receipt of manuscript. Articles submitted
for publication are evaluated according to
several criteria including significance of the
contribution to the polygraph field, clarity,
accuracy, and consistency.
Copyright
Authors submitting a paper to the
American Polygraph Association (APA) do so
with the understanding that the copyright
for the paper will be assigned to the
American Polygraph Association if the paper
is accepted for publication. The APA,
however, will not put any limitation on the
personal freedom of the author(s) to use
material contained in the paper in other
works, and request for republication will be
granted if the senior author approves.

1
Polygraph, 2015, 44 (2)
Diagnostic Polygraph Technique Replication Study
Laboratory Study of a Diagnostic Polygraph Technique in a Single Sequence:
a replication study
Rodolfo Prado
Carlos Grajales
Raymond Nelson
Abstract
A previous laboratory study of a diagnostic polygraph technique in a single sequence reported
results consistent with other validated techniques. This replication of our previous research included
163 new examinees and tested the effectiveness of an experimental single-sequence event-specic
diagnostic polygraph technique with two relevant questions and evaluated with the Empirical Scoring
System (Nelson et al., 2011). This replication experimental protocol had an unweighted accuracy
of 86.1%, an 11.1% inconclusive rate, a sensitivity of 83.5%, a specicity of 88.7%. Reliability via
Kappa’s statistic was 0.722. The distributions of truthful and deceptive scores were not signicantly
different between the two studies. These ndings are consistent with previous studies of event-
specic diagnostic polygraph techniques with two relevant questions. Results add further support to
the effectiveness of polygraph formats conducted in Spanish like those conducted in English.
Keywords: diagnostic exams, directed-lie comparison question, lie detection, polygraph
The previous project involved the study
of decision accuracy of an experimental single-
sequence diagnostic polygraph technique
(Prado, Grajales, & Nelson, 2015). Accuracy
of the experimental technique was 87%.
Inspection of the condence intervals reported
herein and by APA (2011) indicated that the
observed accuracy was consistent with the
previously reported normal range of accuracy
for diagnostic technique.
One of the limitations of that study
was the fact that some of the examiners had
only very recently completed their academic
polygraph training and had virtually no actual
eld experience. Also, the study presented a
large number of protocol violations resulting
in unusable examination data. Nearly 20% of
the examinations conducted could not be used
due to heavily artifacted data that could not
be interpreted and due to protocol violations
on the part of the examiners. We attribute this
to general inexperience on the part of many of
the examiners and to the unfamiliarity of the
examiners with an experimental test protocol
for which the examiners had not received
previous instruction or practice until the onset
of this project.
The experimental format in one
sequence did not outperform existing
polygraph diagnostic format in any way and a
conrmation of this false hypothesis is needed
with a replication study. Another interesting
and important aspect of the previous study
was that all participants and examiners – and
the rst and second authors – were native
Spanish speaking persons. Also, all of the
examinations were conducted in Spanish, in
Mexico City. This study is a replication of the
earlier study with Spanish speaking persons
in a different environment (in this case in
Honduras).
Finally, although the previous study
did not address the effectiveness of Directed Lie
Comparison (DLC) questions, it is noteworthy
that the study added support for the
assumption that accuracy and effectiveness
of Comparison Question Test (CQT) polygraph
techniques using DLC questions can
remain stable across language and cultural
differences. Results of other studies of this
experimental format, were needed to reach
any conclusions about dimensions of criterion
accuracy for this technique.
Back to Front Cover

2Polygraph, 2015, 44 (2)
Prado, Grajales, Nelson
Methods
The present research project was
designed to estimate the diagnostic accuracy
this specic technique has in an analog
laboratory setting. The study was performed
between August 15th and August 21, 2013
using a convenience sample population. Half
of the volunteers were cadets undergoing
training at the Military Academy in Honduras
and the other half were cadets training at the
Police Academy of Honduras.
All the volunteers of the study were in
between 18 and 23 years old and they all had
11 to 12 years of education. The volunteers
were at the 2nd and 3rd year at the academy
and almost 95% of them were males.
They all received an open invitation to
participate in the study. The participants were
told that:
- They could withdraw from the study at any
time without punishment.
- They could contact the researchers for
assistance if they experienced emotional
discomfort from the study.
- They should inform any future polygraph
examiner regarding their participation in this
research project and they can advise others
to contact the researchers if clarication is
needed.
Interested participants were taken
from the Military academy to the place where
the study was conducted. Participants were
required to:
1. be of legal age (18).
2. not to be under the inuence of alcohol
or drugs.
3. not to be excessively tired at the time
of the test and.
4. not to be suffering from hunger at the
time of the test.
Prior to volunteering, all subjects
received a consent sheet, informing them of
the use of an experimental polygraph format
and the requirements of their participation in
the activities. A total number of 163 volunteers
participated in the study.
The subjects were randomly assigned
a “status” by the Monitor of the study. Of the
163 original volunteers, 81 were assigned as
“innocent” and 81 as “guilty”. One case violated
the protocol, and was then disregarded from
the study.
The “guilty” status received instructions
in Spanish that are summarized as follows:
You have been chosen as a “guilty”
subject. You can decide if you don’t
want to continue in this study, but if you
will, go inside that room and take a knife
that you will nd on a table there. In
the same room you will nd “Chavita”.
Chavita is the doll that you have to stab
in repeated occasions until you damage
her. Stab “Chavita as many times as
you want, and make sure you do it well.
After that, you will be taken to the
polygraph room. The polygraph
examiner will tell you that the test is
about stabbing “Chavita”. You have to
deny any involvement in that activity
during the duration of the polygraph
evaluation. You have to lie when you
say that you have nothing to do with
that act.
If the Polygraph evaluation can´t tell
that you are guilty, and you obtain a
“not deceptive” result, your reward will
be one complete weekend out of the
Academy. If the Polygraph evaluation
tells that you are guilty, and you obtain
a “deceptive” result, your punishment
will be 15 days of arrest (not having
free days out of the academy).
The “innocent” status received
instructions that are summarized as follows:
You have been chosen as an “innocent”
subject. You can decide if you don’t
want to continue in this study, but if you
will, go sit outside this room.
You will be taken to the polygraph room.
The polygraph examiner will tell you
that the test is about stabbing “Chavita”.
You have to deny any involvement on
that activity during the duration of the
polygraph evaluation. You have to tell
the truth when you say that you have
nothing to do with that act.
If the Polygraph evaluation tells that
you are innocent, and you obtain a
“not deceptive” result, your reward will
Back to Front Cover

3
Polygraph, 2015, 44 (2)
Diagnostic Polygraph Technique Replication Study
be one complete weekend out of the
Academy. If the Polygraph evaluation
can´t tell that you are innocent, and
you obtain a “deceptive” result, your
punishment will be 15 days of arrest (not
having free days out of the academy).
Both groups also received a nal
instruction in which it was made very clear
that the reward or the punishment would
be announced at the end of the test. Finally,
all subjects were reminded that during the
polygraph examination they were to deny their
involvement in the false crime.
For the polygraph in-test, a Lafayette
model LX4000 polygraph was used to record
electrodermal activity (EDA), breathing
movement, cardiovascular activity and
voluntary activity. EDA was measured using
skin resistance measured by standard
Lafayette metal electrodes attached to the
medial phalanges of the rst and second ngers.
Thoracic and abdominal breathing movement
was measured by a standard Lafayette
pneumatic tube assembly. Cardiovascular
responses were recorded through the use of
a Lafayette blood pressure cuff set at 80 to 90
mm of pressure and placed on the subject´s
calf. A movement activity sensor pad was
placed on the subject’s seat.
The study was conducted at two
separate facilities. Facility “A” was where the
mock stabbing occurred. This was where the
Monitor received the volunteers and assigned
them their status. Once the subjects completed
their eld assignment, the “examiners
coordinator” escorted them to the examiner
who would conduct the test.
There were 25 independent examiners.
The examiners’ coordinator (coordinator) had
the examiners on a list from 1 to 25 in facility
“B” where the examiner had a communal
working area and 6 polygraph evaluation
rooms. The coordinator assigned the examinees
to an examiner in order of appearance, and
also assigned them a polygraph room. At the
end of the test the coordinator accompanied
the examinee to the waiting room in facility
“A”. Each exam was evaluated by both the
examiner and the quality control reviewer.
Quality Control examinations would result
in a test being considered Non-Valid due to
protocol violations that included:
- Physical illness or afiction in the examinee.
- Guilty subjects not appropriately denying or
confessing to the crime.
- Examiners not correctly following the DLDT
format.
- Interrupted or incomplete tests.
- Non-interpretable results.
Eleven examiners were Polygraph
examiners with 1 year of experience and
14 examiners were students who had just
graduated from their 10 week basic training
program in polygraph examination. The
quality control reviewer was the instructor
leading the class during the last week of the
training. An independent examiner with 20
years of experience conducted the “blind” Test
Data Analysis of the charts.
After the tests and the quality control
reviews were nished, the results (NDI,
DI, INC, protocol violation) of the test were
provided to the Monitor with the ground
truth status inside an envelope with the
name of the examinee. The coordinator sent
that envelope back to facility “A”, where the
Monitor opened the envelope and compared
the test result with the ground truth status.
Examinees whose test results were truthful
were rewarded with weekend time away from
the academy regardless of their guilt status.
No arrest or restriction consequences where
actually imposed on examinees who did not
produce truthful test results, though they were
informed of the potential for restriction prior
to their participation in the study activities.
Experimental Format
The test format included:
- Two neutral questions in
positions 2 and 8 and repeated
at positions 13 and 18
- One sacrice relevant question
in position 3
- Two relevant questions
in positions 5 and 7 (rst
presentation), repeated at
positions 10 and 12 (second
presentation), 15 and 17 (third
presentation), and nally in
positions 20 and 22 (fourth
presentation).
- Three DLC questions in
positions 4, 6 and 9 (rst
Back to Front Cover

4Polygraph, 2015, 44 (2)
Prado, Grajales, Nelson
presentation), repeated at 11, 14
and 16 (second presentation),
and nally at positions 19, 21
and 23 (third presentation).
Table 1. Questions presented (translated into English).
# ID Type Text Answer
1 X The test is about to begin, please do not move
and answer with yes or no to each question
2 1N1 N Are we in the year 2013? Yes
3SR SR Regarding the damage induced to Baby Chavita,
do you intend to answer truthfully each question
about that?
Yes
4 1C1 C Have you ever hurt to a loved one? No
51R1 R Today, did you stab Baby Chavita? No
6 1C2 C Have you ever done something you later
regretted?
No
7 1R2 R Today, you stabbed Baby Chavita? No
8 1N2 N Are we in Tegucigalpa City? Yes
9 1C3 C Have you ever been irresponsible with your
duties?
No
10 2R1 R Today, did you stab Baby Chavita? No
11 2C1 C Have you ever hurt a loved one? No
12 2R2 R Today, you stabbed Baby Chavita? No
13 2N1 N Are we in the year 2013? Yes
14 2C2 C Have you ever done something you regretted
later?
No
15 3R1 R Today, did you stab Baby Chavita? No
16 2C3 C Have you ever been irresponsible with your
duties?
No
17 3R2 R Today, you stabbed Baby Chavita? No
18 2N2 N Are we in Tegucigalpa City? Yes
19 3C1 C Have you ever hurt a loved one? No
20 4R1 R Today, did you stab Baby Chavita? No
21 3C2 C Have you ever done something you regretted
later?
No
22 4R2 R Today, you stabbed Baby Chavita? No
23 3C3 C Have you ever been irresponsible with your
duties?
No
24 XX The test is about to end, please don’t move until
I release the air in the cuff
The test questions, order and its type are
summarized in Table 1.
Back to Front Cover

5
Polygraph, 2015, 44 (2)
Diagnostic Polygraph Technique Replication Study
Table 2. Effective sample size in the study
SAMPLE SIZE
Effective Sample Size 162
Subjects assigned to the “Innocent” Status 81
Subjects assigned to the “Guilty” Status 81
Table 3 Inconclusive results and estimated condence intervals.
INCONCLUSIVE RATE
Number of Inconclusive Results 18
Number of Inconclusive Results (Within “Innocent”
Subjects)
10
Number of Inconclusive Results (Within “Guilty”
Subjects)
8
Total Inconclusive Rate
(Wilson’s Condence Interval)
11.111 %
(7.145%, 16.879%)
Inconclusive Rate (Within “Innocent” Subjects)
(Wilson’s Condence Interval)
12.345 %
(6.846%, 21.255%)
Inconclusive Rate (Within “Guilty” Subjects)
(Wilson’s Condence Interval)
9.876 %
(5.090%, 18.296%)
Results
Inconclusive results are shown in Table
3, along with the 95% condence intervals.
Condence intervals were obtained through
two different approaches. The rst is the
computationally simple approach following
Wilson, using a renement of the simple
asymptotic method. For the scoring, relevant
questions were compared always against
the comparison questions given immediately
before the crime relevant question, using the
ESS transformations, Two-Stage Decision
Rules and cut scores for a 2 relevant question
single issue test (Nelson et al., 2011).
From the 163 subjects that were rst
included in the study, only one case resulted
in some form of protocol violation and was
therefore excluded from the study calculations.
The excluded protocol violation case had no
signicant impact on the sample size. The
sample size that was subject to analysis after
the exclusions is summarized in Table 2.
Diagnostic accuracy was calculated
excluding all inconclusive cases resulting in a
sample size of 144 with, 71 programmed as
“Innocent” and 74 programmed as “Guilty”.
Diagnostic Accuracy Measures obtained by
the polygraph test are shown in Table 4.
Back to Front Cover

6Polygraph, 2015, 44 (2)
Prado, Grajales, Nelson
Table 4. Accuracy prole.
DIAGNOSTIC ACCURACY
Accuracy
(Wilson’s Condence Interval
86.111 %
(79.520%,
90.826%)
Sensitivity
(Wilson’s Condence Interval)
83.561 %
(73.429%,
90.339%)
Specicity
(Wilson’s Condence Interval)
88.732 %
(79.310%,
94.179%)
Error Rate
(Wilson’s Condence Interval)
13.889 %
(9.174%, 20.480%)
False Positives
(Wilson’s Condence Interval)
5.555 %
(2.842%, 10.579%)
False Negatives
(Wilson’s Condence Interval)
8.333 %
(4.831%, 14.001%)
Likelihood Ratio (+)
(Condence Intervals based on Risk Ratios)
7.42
(3.83 , 14.4)
Likelihood Ratio (-)
(Condence Intervals based on Risk Ratios)
0.185
(0.11 , 0.313)
Table 5. Diagnostic Reliability of the experimental format
DIAGNOSTIC RELIABILITY
Kappa Statistic
(Analytic Method Condence Interval)
0.722
(0.610 , 0.835)
Area Under ROC Curve
(Analytic Method Condence Interval)
0.861
(0.805 , 0.918)
Agreement 86.11%
Correlation 0.7235
This experimental format presented an
accuracy prole that is comparable to those
reported on the meta-analytic review (APA,
2011). Results show a respectable level of
precision in the test, with accuracy results
comparable and sometime exceeding those
of other polygraph techniques. Estimates of
diagnostic reliability obtained with the test are
shown in Table 5.
To provide methods for repeatability, a
cross-tabulation of the test results is shown
in Table 6. These numbers correspond to the
basis of accuracy and reliability calculations,
since inconclusive results are already
excluded.
Back to Front Cover

7
Polygraph, 2015, 44 (2)
Diagnostic Polygraph Technique Replication Study
Table 6. Cross-tabulation of classication performance, excluding Inconclusive cases.
PREDICTED CLASSIFICATION
Guilty Innocent TOTAL
STATUS Guilty 61 12 73
Innocent 8 63 71
TOTAL 69 75 TOTAL= 144
CASES
Results from the blind analysis of the data are shown in Table 7.
Table 7. Descriptive Statistics of Calculated Scores
CALCULATED SCORES
Arithmetic Mean of Scores 0.6234
Standard Deviation of Scores 7.279
Arithmetic Mean of Scores (Within Innocents) 5.135
Standard Deviation of Scores (Within Innocents) 5.442
Arithmetic Mean of Scores (Within Guilty) -3.889
Standard Deviation of Scores (Within Guilty) 5.983
Table 8. Comparison of Accuracy Proles among different techniques.
TECHNIQUE
Diagnostic
Acuracy
Criterion
DLDT/ESS Federal
You-
Phase/
ESS
Utah PLT
(Combined)/
UTAH
ZCT/ESS
Acuracy 86.111 % 90.4% 93.0% 92.1%
Sensitivity 83.561 % 84.5% 85.3% 81.7%
Specicity 88.732 % 75.7% 80.9% 84.6%
Comparison with other diagnostic
Techniques
Finally, the estimated prole accuracy
of the experimental format presented in this
research was compared with the mean results
reported in the meta-analytic review (APA,
2011) for diagnostic techniques, excluding
outliers. This comparison is shown in Table 8.
By using exact binomial tests it was
possible to verify statistically signicant
differences between the diagnostic measures
obtained with the experimental format and
other similar approaches already included in
the meta-analysis review.
The experimental format presented no
statistical difference in its estimated accuracy
to that estimated by the Federal You-Phase
technique (test’s p-value=0.088). There is
evidence of a slightly signicant difference in
accuracy between the experimental format and
the ZCT/ESS technique (test’s p-value=0.012).
Also there is a highly signicant difference
with the accuracy of the experimental format
technique (test’s p-value=0.002) and the Utah.
Back to Front Cover

8Polygraph, 2015, 44 (2)
Prado, Grajales, Nelson
Table 9. Criterion Accuracy Prole
Criterion Accuracy Prole
N Deceptive 73
N Truthful 71
Total N 144
Number Scorers 1
N of Deceptive Scores 69
N of Truthful Scores 75
Total Scores 144
Mean D -3.889
Std Dev D 5.983
Mean T 5.135
Std Dev T 5.442
Reliability – Kappa 0.722
Reliability – Agreement 0.861
Reliability – Correlation 0.723
Unweighted Average Accuracy 0.861
Unweighted Average Inconclusives 0.111
Sensitivity 0.835
Specicity 0.887
FN Errors 0.083
FP Errors 0.055
D INC 0.098
T INC 0.123
Likelihood Ratio (+) 7.42
Likelihood Ratio (-) 0.185
D CORRECT 0.8356
T CORRECT 0.8873
For sensitivity, there are not statistically
signicant differences concluding that there
is no evidence to assume that experimental
format technique’s sensitivity is lower than
for any of the other techniques (all test’s
p-values>0.60).
The experimental format specicity
results were signicantly higher than
the Federal You-Phase technique (test’s
p-value=0.008). There was no statistically
signicant difference compared with the
Utah PLT or the ZCT/ESS techniques
(test’s p-value=0.098 and p-value=0.411,
respectively).
Based on these three diagnostic
accuracy criteria, there is no evidence
to suggest that the experimental format
technique has different accuracy than that
of the Federal You-Phase. It is not different
from the ZCT/ESS, nor Utah PLT techniques
in terms of sensitivity and specicity. The only
compared techniques that provided evidence
of statistically better results was the Utah PLT
(Combined) and the ZCT/ESS techniques,
only due to a higher level of accuracy, since
the sensitivity and specicity are no different
than those of the experimental format. This
seems to indicate that both techniques may
only yield a marginal improvement over
experimental format.
Back to Front Cover

9
Polygraph, 2015, 44 (2)
Diagnostic Polygraph Technique Replication Study
Table 10. Comparison of the Effective sample size in the study.
SAMPLE SIZE
Effective Sample Size 161
Subjects with the “Innocent” Status 81
Subjects with the “Guilty” Status 80
Table 11. Comparison of the Inconclusive rates between two levels of presentations three
presentations (PRES) versus the use of three and four presentations of the test questions.
INCONCLUSIVE RATE
3 PRES Using up to 4
PRES
Number of Inconclusive Results 58 18
Number of Inconclusive Results (Within “Innocent”
Subjects)
31 10
Number of Inconclusive Results (Within “Guilty”
Subjects)
27 8
Total Inconclusive Rate 35.80 % 11.11 %
Inconclusive Rate (Within “Innocent” Subjects) 38.27% 12.34%
Inconclusive Rate (Within “Guilty” Subjects) 33.33 % 9.88 %
Comparison of results between different
number of presentations
We conducted a further analysis to
complement the results of the study. This section
presents the results of a series of statistical
comparisons investigating differences the
number of stimulus presentations may have
had in the test results.
The rst comparison was to test the
impact that the number of presentations had
in the inconclusive rates. These results are
presented in Table 11. It is worth remembering
that, after excluding the invalid case, 162
subjects were included in the sample, with 81
of these belonging to the “Innocent” group and
81 belonging to the “Guilty” group. See table
10.
The rst comparison was to test the impact
that the number of presentations had in the
inconclusive rates. These results are presented
in Table 11. It is worth remembering that,
after excluding the invalid case, 162 subjects
were included in the sample, with 81 of these
belonging to the “Innocent” group and 81
belonging to the “Guilty” group.
According to the results in the table above,
along with the results of a probability test on
the equality of proportions using a large-sample
statistic, there is a statistically signicant
decrease in the number of inconclusive
results when using up to four presentations,
compared to the numbers obtained with only
three presentations (test’s p-value<0.0001).
The difference in the inconclusive rate is
evident on both Innocent and Guilty Subjects
(for both cases, tests’ p-value ≈ 0.0001)
Back to Front Cover

10 Polygraph, 2015, 44 (2)
Prado, Grajales, Nelson
Table 12. Comparison of the Diagnostic Accuracy of the test.
DIAGNOSTIC ACCURACY
3 Presentations 4 Presentations
Accuracy 85.43 % 85.71 %
Sensitivity 92% 83.6%
Specicity 79.24 % 87.7%
Error Rate 14.56% 14.28%
False Positives 10.67% 6.34%
False Negatives 3.88% 7.93%
Likelihood Ratio (+) 4.43 6.79
Likelihood Ratio (-) 0.101 0.187
Table 13. Distributions of truthful and deceptive scores
DIAGNOSTIC RELIABILITY Experiment One Experiment Two
Arithmetic Mean of Scores 1.38 0.6234
Standard Deviation of Scores 6.934 7.279
Arithmetic Mean of Scores
(Within Innocents) 5.449 5.135
Standard Deviation of Scores
(Within Innocents) 5.545 5.442
Arithmetic Mean of Scores
(Within Guilty) -3.256 -3.889
Standard Deviation of Scores
(Within Guilty) 5.35 5.983
The results in Table 12 aim to verify whether
the use of different numbers of presentations
affect the diagnostic accuracy of the test in
any way. The numbers are shown below.
According to the results, there is no statistical
difference between the numbers obtained
with three presentations and with four
presentations (all test’s p-values>0.10), in any
of the accuracy measures presented in the
table above. This evidence indicates that the
difference is either negligible or too small to be
detected by our experiment.
Distribution of the scores
Finally, the results in Table 13
indicates that the distributions of truthful
and deceptive scores were not signicantly
different between this replication study and
the previous one (Prado, Grajales, & Nelson,
2015).
Back to Front Cover

11
Polygraph, 2015, 44 (2)
Diagnostic Polygraph Technique Replication Study
Conclusions
This replication study provides
additional evidence that a single sequence
technique did not out-perform traditional
diagnostic techniques. Hypothesized
advantages of a single sequence diagnostic
format, beginning with the potential for
increased test effectiveness that may result
from reducing a source of uncontrolled
response variance when starting and stopping
the recording when using traditional diagnostic
CQT formats cannot be conrmed by these
studies. There may be no real advantage of
single recording polygraph formats compared
with multiple chart formats.
Finally, although the previous study
did not address the effectiveness of DLC
questions, it is noteworthy that this study
adds support for previous nding in which
the accuracy and effectiveness of polygraph
evaluations conducted in Spanish are similar
as those conducted in English. The accuracy of
the polygraph evaluation remain stable across
language and cultural differences. This study
provides further support that DLC questions
are robust, even with “inexperienced”
examiners. We found no difference of accuracy
between examiners with or without experience.
Though it was not the goal of this
project we placed the cardio cuff on the lower
leg and we found that this generates similar
results as evaluations for which the cuff
was placed on the arm, though with a less
discomfort experience for the examinee.
Finally, we found that with
three presentations of each question,
the experimental format generated 8%
inconclusive results. With four presentations
it was reduced to 4%. Most of the inconclusive
results involved innocent examinees. The
4th presentation didn’t signicantly affect
accuracy, sensitivity and specicity of the test.
This study was limited in scope, and
intended only as an attempt to replicate
the results of an earlier study using this
experimental diagnostic polygraph format in
which multiple presentations of the question
stimuli are accomplished in a single recorded
sequence. This study did not compare the
effectiveness of DLC and PLC methods, and
did not compare the effectiveness of arm
cuff and leg cuff response data. This study
also made no attempt to dene or investigate
the psychological or physiological basis of
responses to polygraph stimuli and addressed
only a limited range of research questions
regarding the accuracy of categorical test
results and mean scores. These limitations
notwithstanding, we conclude that these
study results provide further support for the
effectiveness of the polygraph in general, and
further support for the effectiveness of DLC
polygraph formats with exams conducted
with native Spanish speaking persons.
Although there is no advantage to the use of
the experimental format compared with other
validated polygraph formats, we recommend
continued research and continued interest
in the potential for the development of a
further improved single sequence single issue
diagnostic polygraph format.
Funding for this study was provided by
the IPTC.
We want to thank to the Consejo
Nacional de Defensa y Seguridad de Honduras,
and to the Polygraph Unit of the Dirección
Nacional de Investigación e Inteligencia de
Honduras without whose participation this
study would not have been possible.
Back to Front Cover

12 Polygraph, 2015, 44 (2)
Prado, Grajales, Nelson
References
Abhyankar, Smruti & Kaur, Gursharn (2010). Condence interval for Binomial Proportions.
.Retrieved on January 5, 2015 at [http://www.isid.ac.in/~deepayan/SC2010/project-sub/
bootstrap_binomial_report.pdf].
American Polygraph Association. (2011). Meta-analytic survey of criterion accuracy of validated
polygraph techniques. Polygraph, 40, 194-305.
American Polygraph Association (2009a). Model Policy for Post-conviction Sex Offender Testing.
[Electronic version] Retrieved January 25, 2012, from http://www .polygraph.org
Ansley, N. (1998). The zone comparison test. Polygraph, 27, 108-122.
Barland, G. (1981). A validation and reliability study of counterintelligence screening tests. U.S.
Department of Defense, Security Support Batallion. 902nd Military Intelligence Group, Ft.
Mead, MD.
Bossuyt, Patrick et al. (2003). Towards complete and accurate reporting of studies of diagnostic
accuracy: the STARD initiative. BMJ 2003;326:41.1.
Kircher, J. C., Packard, T., Bell, B. G., & Bernhardt, P. C. (2010). Effects of prior demonstrations of
polygraph accuracy on outcomes of probable-lie and directed-lie polygraph tests. Polygraph
39, 22–67.
Kircher, J. C., Kristjiansson, S. D., Gardner, M. K., & Webb, A. (2005). Human and
co
m
p
u
t
e
r
decision-making
in the
psychophysiological detection
of deception. University of Utah.
Nelson, R. (2015). The scientic basis for polygraph testing. Polygraph, 44(1) 28-61.
Nelson, R., Handler, M., Adams, G., & Backster, C. (In press). Survey of reliability and criterion
validity of Backster numerical scores of You-Phase exams from conrmed eld investigations.
Polygraph.
Nelson, R. Handler, M. Shaw, P., Gougler, M., Blalock, B., Russell, C., Cushman, B., & Oelrich, M.
(2011). Using the Empirical Scoring System, Polygraph, 40 (2).
Newcombe, R (1998). Two-sided condence intervals for the single proportion: comparison of seven
methods. Statistics in Medicine 17, 857—872.
Prado, R., Grajales C., & Nelson R. (2015). Laboratory Study of Directed Lie Polygraphs with Spanish
Speaking Examinees. Polygraph. 44(1), 79-90.
R Development Core Team, (2008). R Foundation for Statistical Computing, Vienna, Austria. ISBN
3-900051-07-0, URL http://www.R-project.org.
Back to Front Cover

13
Polygraph, 2015, 44 (2)
Polygraph Countermeasures Literature Review
A Literature Review of Polygraph Countermeasures and the Comparison
Question Technique
Mark Handler and Charles Honts
Honts, Handler, & Hartwig, LLC.
Walt Goodson
Texas Department of Public Safety
Abstract
We reviewed the research of countermeasures effect on the comparison question technique. We pro-
vide a consolidation of countermeasure literature as well as an operational denition and taxonomy
of countermeasures. We surveyed the pertinent literature regarding the effectiveness and limitations
of certain countermeasure tactics. We offer evidence-based answers are to common countermea-
sures questions and make recommendations for reporting countermeasures.
Keywords: countermeasures, polygraph comparison question technique
Authors’ note: A portion of this document appeared previously in APA Magazine 48(2) under Handler, Honts & Blalock.
The authors are grateful to Pam Shaw and Don Krapohl for their thoughtful comments, suggestions and critical reviews of
earlier drafts of this manuscript. The thoughts, ideas, and recommendations do not necessarily reect those of the American
Polygraph Association or the Texas Department of Public Safety. Correspondence should be addressed to Mark Handler, APA
editor in chief. Contact: editor@polygraph.org
A Literature Review of Polygraph
Countermeasures and the
Comparison Question Technique
In order for a countermeasure to be ef-
fective in a Comparison Question Technique
(CQT), it must satisfy two requirements. First,
it must create a sufcient difference in the
polygraph measurements to comparison and
relevant questions to produce a truthful or
inconclusive outcome. Secondly, it must be
done covertly as to not be identied by the ex-
aminer, an observer, or any quality control re-
view. In considering what information would
be most helpful to examiners we provide evi-
dence-based answers to some important ques-
tions about countermeasures.
Our operational denition of “countermea-
sure”?
There have been a number of proposed
denitions from within and outside of the pro-
fession for the term countermeasure (CM). We
needed to operationally dene CM as it applies
to polygraph testing. For our purposes, we
considered a CM to be anything a test subject
does in an attempt to alter the test data so as to
produce a truthful (negative) test result. This
denition encompasses the truthful subjects
trying to ensure a True Negative (TN) result
and the deceptive subjects trying to produce
a False Negative (FN) outcome. One could os-
tensibly argue that all subjects engage in some
form of behavior to produce truthful outcomes
and are thus attempting CMs - the truthful
tell the truth and the deceptive lie, but we feel
these actions don’t t our denition for alter-
ing the test data. To alter means to change or
make different in a meaningful way.
What type of CMs do people use?
We followed Honts’ (1987) taxonomy
as it breaks down CMs into categories that
have been researched, though others have
produced different recommendations for CM
categorization (see Krapohl, 2009). In follow-
ing Honts (1987) we break CMs down into the
following categories;
1. General State CMs- actions intend-
ed to alter the subject’s psycholog-
ical state and/or measured physi-
Back to Front Cover

14 Polygraph, 2015, 44 (2)
Handler, Honts, Goodson
ological responses throughout the
entire examination. They include
such things as; drugs, relaxation,
or interfering agents. They are not
focused on any specic point in the
testing.
2. Specic Point CMs- as their name
suggests, these are actions the
subject takes at specic points in
the testing process. They can be
attempts to reduce responses to
relevant questions but are usual-
ly efforts to increase responses to
comparison questions. They can
be employed physically, mentally
or in combination.
3. Spontaneous CMs- these are CMs
that subjects report doing with-
out planning or forethought. A
number of laboratory studies de-
briefed subjects about efforts to
produce truthful outcomes. These
debriefs are the source of most of
our knowledge of spontaneous
CMs. Subjects report trying such
things as; relaxation, rationaliza-
tion, imagery, attempts to control
their breathing or heart rate, try-
ing to stay calm, biting their tongue
and pressing their toes at random
places.
4. Information CMs- people who know
they are going to take a polygraph
examination (both guilty and inno-
cent) often seek information about
polygraph techniques and CMs
from the internet or other sources.
This information-seeking can be
motivated by an attempt to satis-
fy curiosity, to try and hide decep-
tion, or in an effort to ensure that
truthfulness is obvious.
Given our operational denition and
taxonomy we sought to provide evidence-based
answers to some important questions about
CMs. Evidence-based answers and practices
concerning CMs are not simple. They have to
be based upon research and not on anecdote
or dogma. Evidence-based answers and prac-
tices have to be qualied by the limitations
of the research upon which they are based.
Those qualications depend on such things as
whether the subjects were coached or if they
received practice on an instrument. Who were
the subjects? Did the examiner use some sort
of activity sensor? The following is a summary
of some ndings from the peer-reviewed pub-
lished studies we examined for this paper.
1. Rovner (1986) is a rewrite of his
1979 doctoral dissertation in which
he trained subjects on the princi-
ples of CQT testing, including giv-
ing them pictorial examples of re-
actions. He called these the Info
group. He also gave the Info group
Specic Point CM training using a
variety of physical and mental CMs
known to produce reactions. He
had a second group called the Info
+ Practice group. He gave them the
same material but allowed them to
practice their CMs on a polygraph
before their real test. The accu-
racy of the results for the control
and the Info group was about 88%.
However, the Info + Practice group
accuracy results were about 62%.
He did not report using an activity
sensor or making attempts to iden-
tify CM subjects.
2. Dawson (1980) used Stanislavsky
trained actors to attempt Gener-
al State CMs in a mock-crime lab
study. They were not trained in
polygraph principles and they did
not receive practice. The CM group
actors were motivated to appear
innocent to display their superior
acting skills. CM deceptive sub-
jects said they used imagery and
prior memories as strategies. The
General State CM effects were in-
effective. Excluding inconclusive
results, all CM subjects were found
deceptive.
A side note, Dawson conducted an
interesting additional experiment.
He had subjects answer the poly-
graph questions in two ways, im-
mediately after the question and
delayed by 8 seconds. He measured
responses in three ways; after the
question without an answer, after
the immediate answer and after the
delayed answer. Based on numer-
ical scores he reported the follow-
ing. Immediate answers accuracy
was 75%correct, 12% incorrect,
and 12% inconclusive. The mea-
Back to Front Cover

15
Polygraph, 2015, 44 (2)
Polygraph Countermeasures Literature Review
surements following the question
but before the delayed answer pro-
duced 83% correct, 8% incorrect,
and 8% inconclusive. The mea-
surements taken after the delayed
answer resulted in 29% correct, 8%
incorrect, and 62% inconclusive.
3. Bradley & Ainsworth (1984) tested
General State CMs by using alco-
hol intoxication during a crime act
and also during polygraph testing.
They tested subjects with the CQT
and the CIT. We limit our discus-
sion to the CQT ndings. They mea-
sured heartrate, respiration and
electrodermal responses during a
mock-crime robbery and shooting.
They reported alcohol intoxication
during the crime decreased detect-
ability with electrodermal activity.
Intoxication during the testing was
ineffective.
4. Honts, Hodes & Raskin (1985) in
experiment 1 trained the CM group
on the principles of polygraph CQT
and Specic Point CMs. They
coached the subjects on phys-
ical (press toes to oor) and pain
(bite tongue) CMs. They did not
get any practice on an instrument.
They used a photoelectric plethys-
mograph instead of a cardio cuff.
They reported no signicant effects
for the CM group. They were un-
able to detect which subjects used
CMs by either direct observation or
reviewing the charts. They did not
use an activity sensor.
5. Honts, Hodes & Raskin (1985) in
experiment 2 trained the CM group
on the principles of polygraph CQT
and Specic Point CMs. There
they coached the subjects on phys-
ical (press toes to oor) and pain
(bite tongue) CMs. However, unlike
experiment 1, these CM subjects
got to practice on an instrument.
They used a standard cardio cuff
in this experiment. They reported
there was a 47% FN rate for the CM
group. Again they were unable to
detect which subjects used CMs by
either direct observation or review-
ing the charts. They did not use an
activity sensor.
6. Honts, Raskin, & Kircher (1987)
trained subjects in physical (press
toes to oor) and pain (bite tongue)
Specic Point CMs. They gave the
subjects training on CQT princi-
ples and coaching on when and
how to apply the CMs to the com-
parison questions. They did not
give any practice on an instrument.
They measured muscle movement
by electromyography (EMG) on the
subjects’ jaw and calf. They report-
ed no FNs with the guilty control
group who did not use any CMs.
They identied 78% of the truthful
subjects correctly. Seventy percent
of the guilty CM group produced
FN results. They identied 90% of
the CM subjects by EMG tracings.
7. Honts, Raskin, Kircher & Hodes
(1988) included 65 deceptive sub-
jects from four studies who were
debriefed about any use of Sponta-
neous CMs. Sixty percent (39/65)
of the deceptive subjects admitted
to Spontaneous CMs. The strat-
egies included: relaxation, ratio-
nalization, self-deception, disasso-
ciation, imagery, attempts to con-
trol breathing or heartrate, biting
tongue, attempts to control general
physiological responses and press-
ing toes to the oor. Blind scor-
ing accuracy was 80% correct, 3%
wrong and 17% inconclusive. The
use of Spontaneous CMs did not
affect the test results. Examiners
were unable to differentiate CM us-
ers. No activity sensor was used.
8. Raskin and Kircher (1990) trained
subjects in physical (muscle con-
traction) and mental (counting
backwards) Specic Point CMs.
They taught them the principles of
polygraph testing and when they
should employ the CMs. They
coached them and they had prac-
Back to Front Cover

16 Polygraph, 2015, 44 (2)
Handler, Honts, Goodson
tice on an instrument. They used a
seat activity sensor. The CMs pro-
duced about 50% FNs when scored
by the computer. All of the physi-
cal CMs were identied by review-
ing the seat sensor data.
They also trained a “relaxation”
group on the principles of poly-
graph. This group used autogenic
relaxation as a General State CM
throughout the entire examination.
The relaxation General State CMs
were not effective.
9. Honts, Raskin & Kircher (1994)
trained subjects in Specic Point
CMs and CQT polygraph principles.
Here they used physical (muscle
contraction), pain (bite tongue) and
mental (counting backwards) CMs
during the comparison questions.
The subjects were coached but
not given any practice on the poly-
graph instrument. They measured
EMG on the jaw and calf. The CM
group produced a 50% FN rate and
were not detected by direct obser-
vation or by looking at the poly-
graph charts. They were able to
identify the pain and physical CM
groups by EMG scores which were
signicantly larger than the other
groups.
10. O’Toole et al., (1994) was a par-
tial replication of the Bradley &
Ainsworth (1984) study on General
State CMs using alcohol. Here the
deceptive CM group were intoxi-
cated only during the mock-crime
theft. They measured skin conduc-
tance, respiration and nger pulse
amplitude. They did not provide
any information on polygraph prin-
ciples, they did not coach or prac-
tice any CMs with the CM subjects.
Alcohol intoxication during the
crime had no effect on polygraph
results.
11. Honts, Amato and Gordon (2001)
did not train any of their subjects
but debriefed them on what actions
they took to help pass their poly-
graph tests in a large lab study.
Overall 68% of the subjects report-
ed they attempted a Spontaneous
CM. Almost half of the truthful
subjects (46%) reported using at
least one Spontaneous CM. These
Spontaneous CMs included altered
breathing, mental and physical
strategies. The Spontaneous CMs
did not affect the deceptive scores
but it shifted the truthful scores in a
negative direction. In other words,
the truthful subjects who tried to
help increase their chance of a TN
result had less truthful scores than
those who did not attempt Sponta-
neous CMs. These subjects were
not coached, nor were they given
practice tests on an instrument.
There was no activity sensor used
in any of these cases.
Later, three federally certied in-
structors reviewed the charts in
an effort to identify the presence of
CMs. None of the three federally
trained instructors could identi-
fy the CM subjects at better than
chance levels by reviewing the test
data.
12. Oglivie & Dutton (2008) reanalyzed
unpublished data from a CM lab
experiment with and without the
activity sensor tracing. In this case
the activity sensors included, seat,
feet and arm sensors. The CM
polygraph subjects in these cases
received a copy of the internet pub-
lication of The Lie Behind the Lie
Detector (Maschke & Scalabrini,
2002) as a reference. This materi-
al has detailed descriptions of CQT
polygraph principles and CM strat-
egies. The CM subjects had to pass
a written examination on the CM
material before proceeding. They
used physical (toe curling) Specif-
ic Point CMs during the presenta-
tion of the comparison questions.
A subset of the CM group was al-
lowed to practice on an instrument
while receiving real-time feedback
from an experienced examiner.
Five experienced polygraph ex-
aminers reviewed the polygraph
charts in two conditions; with and
without the activity sensor data.
The examiners were asked; a)
to score the test data using a 7-po-
Back to Front Cover

17
Polygraph, 2015, 44 (2)
Polygraph Countermeasures Literature Review
sition federal scoring rules, and b)
using a 5-point Likert scale rate
the absence or presence of CMs.
Mean CM present/absent scores
were statistically signicant be-
tween the sensor and no-sensor
conditions. Without the activity
sensor, the CM present/absent
scores were near zero. Examiners
did not perform above chance lev-
els when relying on the traditional
polygraph channel data to iden-
tify CMs without the activity sen-
sor data. Adding the activity sen-
sor data made identifying the CM
group more likely.
Of note, the average guilty CM total
scores with the activity sensor was
similar to the truthful scores (they
were positive). The average guilty
CM total scores without the activ-
ity sensor was double the average
truthful scores. This suggests that
the Specic Point CM strategies
can create increased FN results.
13. Honts & Reavy (2015) did not train
any of their subjects but debriefed
them on what actions they took to
help pass their polygraph tests in
a large lab study. Overall 48% of
the subjects reported they attempt-
ed a Spontaneous CM. About 50%
of the subjects who received proba-
ble lie comparison questions (PLC)
reported Spontaneous CMs. For
the directed lie comparison ques-
tion variant (DLC) subjects, about
46% reported Spontaneous CMs.
Breaking it down by guilt, over-
all 78% of the deceptive group re-
ported Spontaneous CM attempts
(DLC=72%, PLC=83%). Within the
truthful group 18% overall report-
ed attempts at Spontaneous CMs
(DLC=20%, PLC=15%). None of the
differences were signicant.
The investigators did not make
any attempt to identify CMs either
through direct observation or data
review. Just like in Honts, Ama-
to, & Gordon (2001) the scores for
the truthful CM groups were less
positive. Once again this suggests
truthful subjects who use Sponta-
neous CMs hurt their chances of
passing the test.
14. Stephenson & Barry (1998) was not
a CQT study but is discussed here
due to the ndings. The subjects
were polygraph examiners so they
had extensive knowledge about
the CQT polygraph procedures
and CMs. They tested whether
they could identify when a subject
makes a “physical movement be-
tween the toes and shoulder” while
collecting polygraph test data. They
used an activity sensor mounted to
the front of the test subject’s chair.
They collected test data in a mock
acquaintance test during which the
examiner counted from 1 through
10. At some point the test subject
was to make a covert physical CM
attempt and remember the num-
ber at which they attempted the
CM. They stationed an “observer”
directly in front of the test subject
who watched for movement. The
overall CM detection rate for the
examiner observing the subject
was 9%. The “observer” identied
the Specic Point CMs in 36% in
of the cases. When they used the
movement sensor tracing, the ex-
perimenters identied 85% of the
CM attempts.
15. Honts & Alloway (2007) was a
constructive replication of Rovner
(1986) using the Test for Espionage
and Sabotage (TES). They gave
half of the truthful and deceptive
subjects the book The Lie Behind
the Lie Detector (Maschke & Scal-
abrini, 2000) to study for one week.
They found no signicant effect of
providing the Information CM ma-
terial on validity of the TES. Once
again, however, those subjects who
reported using CMs had signi-
cantly lower probabilities of truth-
ful scores. This included both the
truthful and the deceptive subjects.
Here we will provide some common CM ques-
tions along with evidence-based answers.
Back to Front Cover

18 Polygraph, 2015, 44 (2)
Handler, Honts, Goodson
1. Do both truthful and deceptive subjects
attempt CMs?
The simple answer is “Yes.” Research
shows that both truthful and deceptive sub-
jects report attempting Spontaneous CMs.
From study 9 (Honts, Raskin & Kircher 1994)
which was limited to deceptive subjects, 65%
of them attempted Spontaneous CMs. From
study 13 (Honts & Reavy, 2015) about half of
the subjects overall reported attempting Spon-
taneous CMs. A larger proportion of deceptive
subjects reported attempting Spontaneous
CMs but 18% of truthful subjects also report-
ed attempting Spontaneous CMs. From study
11 (Honts, Amato and Gordon, 2001) we see
about 68% overall and about 50% of truthful
subjects attempted Spontaneous CMs.
2. What type of CMs do subjects attempt?
From a number of studies above Spontaneous
CMs include a variety of reported strategies;
relaxation, rationalization, self-deception,
disassociation, imagery, attempts to control
breathing or heartrate, biting tongue, attempts
to control general physiological responses and
pressing toes to the oor. Specic Point CMs
generally included physical (press toes, curl
toes, etc.) or pain (biting tongue) and mental
(counting backwards) activities. Some Infor-
mation CM sources suggest such actions as
squeezing the anal sphincter (http://www.
polygraph.com/). More sophisticated advice
about examination behavior and chart record-
ing CMs is offered at https://antipolygraph.
org/ (Maschke & Scalabrini, 2005). Some ex-
aminees reported attempting a form of Gen-
eral State CMs when they describe attempts
at rationalization, relaxation, disassociation,
imagery, etc.
3. What type of CMs are effective at in-
creasing TN results, creating a FN re-
sult, or resulting in an inconclusive out-
comes- and to what degree?
Spontaneous CM produced no effects for
the deceptive subjects in terms of increased
TN or inconclusive outcomes, nor were there
reliable effects found in the numerical scores.
Deceptive subjects in study 15 shifted the
scores away from a truthful result. Sponta-
neous CMs by truthful subjects decreased
their chances of being found truthful. Infor-
mation CMs that lead to Spontaneous CMs
simply shifted truthful scores in the negative
direction (see study 15). General State CMs
have not been shown to be effective, see study
2 and 10. Study 3 reported some effect for in-
toxication during the mock crime act. Specic
Point CMs have been shown to be effective in
shifting differential response measurements
and increasing FN results (see studies 1, 4, 5,
6, 8, 9, & 12) following specic training, but
not just information. Specic Point CMs thus
seem to be most dangerous when coupled with
hands-on training and practice.
4. Do polygraph test subjects attempt CMs
more with Directed Lie Comparison
questions versus the Probable Lie vari-
ant?
This has not been shown by the relevant
research (see study 13).
5. Can examiners identify examinees us-
ing CMs at better than chance rates?
And does the addition of activity sen-
sors make a difference?
Without an activity sensor there are
no studies that support examiners can identi-
fy CMs at better than chance rates (see stud-
ies 4, 5, 7, 11, & 12). In fact, the research
indicates that when examiners try to identify
countermeasure they falsely accuse a sub-
stantial number (47% or more) of innocent
non-countermeasure users of using CMs
(study 5). With an activity sensor (or EMG)
polygraph examiners are able to signicantly
identify CM users (see studies 6, 8, 9, 12, &
14) who use CMs that required movement (for
example, pressing the toes to the oor.) Final-
ly, there is no evidence that current training in
countermeasure detection is effective. In fact
the alleged respiratory countermeasure signa-
tures caused by the countermeasure materials
produced by Williams (http://www.polygraph.
com/) have been shown to occur naturally in a
substantial number of actually innocent sub-
jects who were not using CMs (Honts & Craw-
ford, 2010).
6. How does using CMs affect the scores
of truthful and deceptive subjects?
Specic Point CMs increase FN outcomes
following training by producing signicant ef-
fects in all of the polygraph components de-
pending upon the countermeasure used (see
studies 1, 5, 6, 8, 9, & 12). It is unclear what
their effect would be for increasing TN out-
comes, though there is no reason to think they
would not be effective.
Spontaneous CMs don’t increase FN and
probably decrease TN results. Information
Back to Front Cover

19
Polygraph, 2015, 44 (2)
Polygraph Countermeasures Literature Review
CMs that lead to Spontaneous CMs would be
expected to have similar results. Spontaneous
CMs are extremely common with examinees
and there does not appear to be any evidence
that such CMs are effective. Therefore, as the
evidence seems to suggest, if the data simply
appears to be messy, and there is sufcient
uncontaminated data to conduct an analysis,
the scorer should attempt to analyze the un-
contaminated data, and a decision should be
rendered by the scorer if conclusive scores are
reached (ex. NDI/NSR, DI/SR). Examiners
should report when data quantity and quali-
ty are insufcient to complete a standardized
numerical evaluation. An example of report-
ing language is:
After assessing the quantity and quali-
ty of the test data collected in this exam-
ination, I determined that the test data
were of insufcient interpretable quan-
tity and/or quality as a result of nu-
merous artifacts to conduct a standard
numerical evaluation. In other words,
there was insufcient data to evaluate
in order to render a reliable decision on
this examination.
General State CMs are unlikely to cre-
ate a differential response between relevant
and control questions that would increase
TN or FN results. At worst they might be ex-
pected to cause an inconclusive result due to
mitigating the overall responsivity to all test
questions, but even increases in inconclusive
outcomes have never been demonstrated in a
published peer-reviewed study. An unpub-
lished study (Gatchel et al., 1983) tested the
General State CM effects of the beta-blocker
drug propranolol. The only signicant nd-
ing was an increase in accuracy with the in-
nocent. Study 3 reported no effect for alcohol
intoxication during a polygraph test. However,
as mentioned, they reported an effect for in-
toxication at the time of the crime. The repli-
cation of that study failed to nd an effect for
alcohol and FN results for intoxication at the
time of the crime (see study 10). In study 2 ex-
perienced actors try to produce FN results us-
ing General State CMs but produced no effect.
In summary the CM research base is
incomplete and additional research is needed.
However, the limited research shows trained
CMs are something that should concern ex-
aminers as under certain circumstances they
have produced substantial numbers of FN er-
rors. Moreover when trained deceptive sub-
jects use CMs, examiners have not shown
an ability to identify those subjects at better
than chance rates without some sort of ac-
tivity sensor (and then only for CMs that re-
quire physical movement). Regardless of any
alleged anecdotal successes at detecting CMs,
no research has shown that any examiner
can reliably detect CMs from simple pattern
recognition. In fact, as mentioned, research
has shown that the respiratory patterns that
are allegedly linked to some internet training
approaches occur naturally in the respiration
recordings of a substantial number of actually
innocent subjects (Honts & Crawford, 2010).
We realize a number of things that
might be CMs appear spontaneously among
truthful examinees. What may distinguish
these events from CMs, though, is the fre-
quency or the targeting of the behaviors. For
example, both truthful and deceptive examin-
ees move during polygraph tests. This does
not, in and of itself, mean that movements are
not useful in detecting CM attempts. Indeed,
research shows that movements can be strong
indicators in that regard. The mere presence
of hyperventilation, as another example, does
not conrm CMs, but if they persist despite ex-
aminer warnings or they seem to appear only
on one category of question, then they can be
useful indicators. Ultimately we hope further
research will help develop improved objec-
tive measures of anomalies among groups of
questions. Future CM detection efforts should
probably seek such an objective measurement
approach.
The research clearly shows that when
examiners do try to detect CMs they falsely ac-
cuse a substantial number of actually inno-
cent subjects. Examiners should be extremely
cautious about reporting CMs based on their
ability to intuit a subject has used CMs. Do-
ing so puts the innocent at risk. The upside to
this literature is that when deceptive subjects
engage in CMs that require movement they
can be reliably identied when examiners use
an activity sensor. Finally, there is no pub-
lished research that information provided by
internet CM websites is at all dangerous to the
validity of the CQT.
Back to Front Cover

20 Polygraph, 2015, 44 (2)
Handler, Honts, Goodson
References
Bradley, M.T. & Ainsworth, D., (1984). Alcohol and the psychophysiological detection of deception.
Psychophysiology 21, 63–71.
Dawson, M.E. (1980). Physiological detection of deception: Measurement of responses to questions
and answers during countermeasure maneuvers. Psychophysiology 17, 8–17.
Gatchel, R. J., Smith, J. E., Kaplan, N. M., et al. (1983). The effect of propranolol on polygraphic
detection of deception. Unpublished manuscript.
Honts, C.R. (1986). Countermeasures and the physiological detection of deception: a
psychophysiological analysis. Dissertation Abstracts International 47, 1761B.
Honts, C.R., Hodes, R.L. & Raskin, D.C. (1985). Effects of physical countermeasures on the
physiological detection of deception. Journal of Applied Psychology 70, 177–187.
Honts, C.R. (1987). Interpreting research on countermeasures and the physiological detection of
deception. Journal of Police Science and Administration 15, 204–209.
Honts, C.R., Raskin, D.C. & Kircher, J.C. (1987). Effects of physical countermeasures and their
electromyographic detection during polygraph tests for deception. Journal of Psychophysiology
1, 241–247.
Honts, C.R., Raskin, D.C., Kircher, J.C. & Hodes, R.L. (1988). Effects of spontaneous countermeasures
on the physiological detection of deception. Journal of Police Science and Administration 16,
91–94.
Honts, C.R., Raskin, D.C. & Kircher, J.C. (1994). Mental and physical countermeasures reduce the
accuracy of polygraph tests. Journal of Applied Psychology 79, 252–259.
Honts, C.R., Amato, S. & Gordon, A.K. (2001). Effects of spontaneous countermeasures used against
the comparison question test. Polygraph 30, 1–9.
Honts, C.R. & Alloway, W. (2007). Information does not affect the validity of a comparison question
test. Legal and Criminological Psychology 12, 311–312.
Honts, C.R. & Crawford, M. (2010). Polygraph countermeasures cannot be detected from respiratory
signatures: Government policy puts the innocent at risk. Paper presented at the American
Psychology-Law Society Meeting, Vancouver. 17–20 March.
Honts, C.R. & Reavy, R. (2015). The comparison question polygraph test: A contrast of methods and
scoring. Physiology & Behavior 143, 15-26.
Krapohl, D.H., (2009). A taxonomy of polygraph countermeasures. Polygraph 38, 89–105.
Maschke, G. W. & Scalabrini, G. J. (2002). The lie behind the lie detector. Internet https://
antipolygraph.org/lie-behind-the-lie-detector.pdf
Maschke, G. W. & Scalabrini, G. J. (2005). The lie behind the lie detector. Second Edition. Internet:
https://antipolygraph.org/lie-behind-the-lie-detector.pdf
Ogilvie, J. & Dutton, D. W. (2008) Improving the detection of physical countermeasures with chair
sensors. Polygraph 37(4), 136-148.
O’Toole, D., Yuille, J.C., Patrick, C.J. & Iacono, W.G. (1994). Alcohol and the physiological detection
of deception: arousal and memory inuences. Psychophysiology 31, 253–263.
Raskin, D. C. & Kircher J. C. (1990) Development of a computerized polygraph system and
physiological measures for detection of deception and countermeasures: A pilot study.
Back to Front Cover

21
Polygraph, 2015, 44 (2)
Polygraph Countermeasures Literature Review
Unpublished Manuscript.
Rovner, L.I., Raskin, D.C. & Kircher, J.C. (1979). Effects of information and practice on detection of
deception. Psychophysiology 16, 197–198 (abstract).
Rovner, L.I. (1986). The accuracy of physiological detection of deception for subjects with prior
knowledge. Polygraph 15, 1–39.
Stephenson, M., Barry, G. (1988). Use of a motion chair in the detection of physical countermeasures.
Polygraph, 17(1), 21-27.
Back to Front Cover

22 Polygraph, 2015, 44 (2)
Handler, Honts, Goodson
Table 1 – Breakdown of CM study ndings.
Study Test type Type of CM Training
Y/N
Coached/
Practice on
inst.
Y/N
Activity Sensor
Y/N
Findings reported
Rovner (1986) CQT Practice CM group
used physical and
mental CMs.
They got to practice
and received feedback.
Info group and
info + practice
group All
given extensive
training on
polygraph
principles and
CN strategies
Coached-
yes
Practice-
yes
No Accuracy of scoring:
Standard group = 87.5%; Info
group = 87.5%;
Info+practice group = 62.5%
Dawson (1980) CQT General State CMs No No No General State CMs had no
effect.
Bradley & Ainsworth
(1984)
Limited to
CQT part
General state CMs-
alcohol intox during
crime and during
polygraph
No No No No effect for intox during
testing. EDA responses were
reduced for intox during crime
scenario.
Honts, Hodes &
Raskin (1985)
Experiment number
1
CQT CM groups used:
1.Bite tongue or
2.muscle contraction
press toes to oor
All given
extensive
training on
polygraph
principles and
CN strategies
Coached-
yes
Practice-no
No
Also no BP
cuff used in
experiment 1
No signicant effects
for CMs found.
Unable to detect CM subjects
through chart interpretation or
observation.
Honts, Hodes &
Raskin (1985)
Experiment number
2
CQT CM groups used:
1.Bite tongue or
2.muscle
contraction press
toes to oor
All given
extensive
training on
polygraph
principles and
CN strategies
Coached –
Yes
Practice-
Yes
No 47% FN for CM groups.
Unable to detect CM subjects
through chart interpretation or
observation.
Honts, Raskin &
Kircher (1987)
CQT CM group used: bite
tongue & press toes to
oor
Yes CM group
trained on
polygraph
principles and
CM strategy
Coached-
yes
Practice-no
Yes EMG on the
jaw and leg.
No FN with Guilty control. 78%
innocent correct. 70% FNs. 90
% of CM subjects identied
with the EMG tracings.
Raskin & Kircher
(1990)
CQT CM group used:
Physical-muscle
contraction PLUS
Mental
Counting backwards
Relaxation group used:
mental imagery
Yes CM group
and relaxation
group trained
on polygraph
principles and
CMs. Practice
recording,
practice logs,
Coached-
yes
Practice-no
Yes seat sensor Computer:
CM group =50% FN.
Relax not effective.
All physical CM subjects
identied by activity sensor.
Honts, Raskin &
Kircher (1994)
CQT CM group used:
Physical-muscle
contraction Pain bite
tongue
Mental group:
Counting backwards
All given
extensive
training on
polygraph
principles and
CN strategies
Coached-
yes
Practice-no
EMG-jaw and
calf
FN= 50% for Mental & Phys.
CM
Unable to detect CMs by
observation or chart tracing
EMG measures for CM groups
were signicantly larger than
controls.
O’Toole et al. (1994) Limited to
CQT part
General state CMs-
alcohol intox during
crime
No No No No effect for intox during crime
Study Test type Type of CM Training
Y/N
Coached/
Practice on
inst.
Y/N
Activity Sensor
Y/N
Findings reported
Examiners may nd Table 1 a quick reference
for a consolidation of the CM study data. Note
Honts et al., 1988 is not included in the table
as those results were derived from included
studies.
Back to Front Cover

23
Polygraph, 2015, 44 (2)
Polygraph Countermeasures Literature Review
Study Test type Type of CM Training
Y/N
Coached/
Practice on
inst.
Y/N
Activity Sensor
Y/N
Findings reported
Honts, Amato &
Gordon (2001)
CQT All CMs were
spontaneous- Some
subjects used more
than one.
32%-breathing
76%-mental
10% physical
No Coached-
no
Practice-no
No 3 DoDPI instructors did not
identify presence of CMs above
chance level.
47% of CM present decisions
were on truthful subjects.
Oglivie & Dutton
(2008
CQT Physical CMs:
Toe curling
All given
extensive
training on
polygraph
principles and
CN strategies
Coached-
yes
Practice yes
for a subset
Compared with
and without seat,
feet, and arm
activity sensor
data.
15 Innocent. &
guilty control
and 38 CMs.
Charts printed
with and without
activity trace.
Five scorers looked at 68 cases
Mean activity sensor scores
signicant for CM group with
and without sensor.
Unable to determine presence of
CMs without the sensor data.
Honts & Reavy(2015) CQT spontaneous no no yes No attempt to identify CMs.
48% attempted CM
PLC=50%; DLC=46%
78% Guilty attempted
PLC 83%; DLC 72%
18% Innocent attempted
PLC 15%; DLC20%
Truthful scores less positive
when CMs attempted
Stephenson & Barry
(1988)
Examiner
counted
1-10 and
Subject made a
physical movement
between the toes and
shoulder
Info yes
because it was
an examiner.
Coached-
yes because
it was an
examiner.
Practice-
yes
Yes-used
Lafayette chair
with activity
sensor bar under
the front legs
Had an observer in front of
subject in addition to examiner.
CM detection rates were:
Examiner=9%; observer=36%;
movement sensor tracing= 85%
Honts & Alloway
(2007)
CQT Information and
Spontaneous CMs
No No Yes under the
chair legs
No effect on FN
Deceptive and Truthful CM
subjects’ probabilities scores
moved away from truthfulness.
Back to Front Cover

24 Polygraph, 2015, 44 (2)
Handler
A book review of Investigative Interviewing- The Essentials edited by Michel
St-Yves, published in 2014 by Carswell publications
Mark Handler
A topic intimately related to polygraph,
and yet often overlooked, is Investigative Inter-
viewing. Many polygraph examiners are unfa-
miliar with the concept. I hope by way of this
book review to introduce interested readers to
Investigative Interviewing. I can think of few
better ways to familiarize oneself with the es-
sentials of concept than by reading this book.
The practice recommendations from this book
will surely improve the quality of anyone’s in-
vestigative work, regardless of the milieu.
The book is for anyone who interviews
anyone else, but is especially appropriate for
polygraph examiners who are often in the
unique position of neutral fact-nder. Poly-
graph consumers and end-users look to poly-
graph to solve problems that for the moment
seem unclear. There is no better time to take
advantage of the essentials of Investigative
Interviewing than during a polygraph pretest
interview. Examiners are in a unique position
to establish an information gathering environ-
ment. Examinees can be cajoled into provid-
ing information that can be exculpatory or in-
culpatory. The interview setting can give them
rope to pull themselves out of their proverbial
hole - or hang themselves in the process.
I often hear examiners say they “seek
the truth”, but that really isn’t the rst step
in the process. We can’t get to the slippery
truth without facts to check out. We can’t get
the facts to check out if we don’t interview. In-
formation is the lifeblood of any investigation.
It provides direction; it can show attempts
at misdirection. Information helps conrm
what we know, disconrm what we thought
we knew, and help reveal what we don’t know.
Having the tools to best develop information is
essential- as the book points out.
Michel St-Yves is a Canadian forensic
psychologist who works with the police. He
is a friend and advocate for law enforcement
which is clearly reected in his work. His area
of expertise is in conducting investigative in-
terviews and teaching law enforcement to do
so. He gathered some of the world’s leading
experts on the subject and had them write
“how to” chapters for the book, geared towards
the practitioner. While many books from ac-
ademicians focus on theory, this is not one of
them. This is for the investigative interviewer
and it is especially relevant for polygraph ex-
aminers.
The book begins with a wonderful
primer on rapport by St-Yves. Investigators
are often taught the importance of establish-
ing and maintaining rapport. But what does
rapport look like? How do we get (and keep)
it? Why is it essential to the investigative in-
terview? The author tackles these, and many
other of the thorny questions about rapport.
In my opinion it would be worth buying the
book for this chapter alone. For without rap-
port, the interview is doomed.
The second chapter is an update on
the Cognitive Interview (CI), which should be
used in every polygraph pretest interview. Ed
Geiselman and Ronald Fisher developed the CI
around 1985 and published their rst book in
1992. In this chapter they describe updates,
improvements and ndings about the CI. The
CI is a general strategy for guided memory re-
trieval based on scientic knowledge of human
memory. The goal is to generate rich detail,
without contamination. Over 100 empirical
studies show a 25-50 % increase in detail over
a standard police type question and answer
interview.
In chapter three experts on child in-
terviewing provide recommendations for con-
ducting physical and sex abuse investigative
interviews. Cyr, Dion and Powell break down
Back to Front Cover

25
Polygraph, 2015, 44 (2)
A Book Review of Investigative Interview- The Essentials
and discuss best practices for the child inter-
view. They include; planning & preparation,
communication rules for obtaining an ac-
count, establishing and maintaining rapport,
memory limitation discussion and questioning
strategies. The authors give several example
interview protocols that have been scientical-
ly shown to work well with children. Finally
they remind us that children, including ad-
olescents, are not just “little adults” and we
need to modify our interviewing approaches to
maximize information gain with this popula-
tion.
Chapter four deals with eye witness
memory and identication. Hope and Sau-
er are cognitive psychologists whose focus
and expertise is human memory and deci-
sion making. We are asking examinees what
they remember so it is incredibly important to
have an understanding of the limits of human
memory. Likewise, we engage them in deci-
sion making process, tell or don’t tell, so we
should have a basic understanding of neuro-
economics. Much of this chapter focuses on
witness identication so it will be more useful
to examiners who also conduct police inves-
tigations that include witness identication.
The authors provide the current best practice
standards for conducting show-ups, line-ups
and photo identication. They also give sound
advice on presenting witness identication ev-
idence in court.
A short chapter on false memory by
James Ost follows and reminds us that we
have incredible power in the interview room
that can create false memories. Ost is a
false-memory expert who has published ex-
tensively on the subject. He provides a short
review that includes; what are false memo-
ries, how do they occur, what are some of the
mechanisms know to create them, what is the
evidence of their existence? Most importantly
he provides clues or indicators of concern that
a reported memory may be false. Much of this
relates to claims of physical abuse and sexu-
al abuse reports, which constitute the bulk of
many police polygraph examiner’s workload.
Chapter six is the heart of the book,
in my opinion. Written by Michel St-Yves and
Christian Meissner, two of the current leading
authorities on suspect confession and confes-
sion related concerns. They review; the impor-
tance of confession evidence, who confesses,
why people confess, the internal and external
pressures that precipitate confession, person-
ality factors that affect confession, and much
more. They break down “interrogation” into
its component parts that mirror the P.E.A.C.E.
model. They discuss important verication
and control practices to try to ensure the con-
fession is real, and not false. Many of us think
we know what to do, what not to do, and how
to do it. This chapter provides a benchmark
against which to see if you are following rec-
ommended practices.
Gisli Gudjonsson is one of the world’s
leading authorities on mental vulnerabilities
and false confessions. His chapter seven is a
comprehensive review of the current state of
knowledge on the subject. Mental vulnerabili-
ties are psychological states and traits that in-
crease a person’s risk of providing inaccurate
or unreliable information during an investiga-
tion. It includes; low intelligence, developmen-
tal disorders, personality disorders, high sug-
gestibility or compliance, and recall concerns
like memory distrust syndrome. Gudjonsson
reminds us that just because someone has
one of these, it does not invalidate their con-
fession. His concerns and recommendations
are prophylactic and protective. He reminds
us how important it is to assess for vulnera-
bilities ahead of time, if possible. He also re-
minds us to reect afterwards on whether the
subject had any mental vulnerabilities that
may have affected their statement or admis-
sion. Most polygraph examiners know that
people we encounter can seem overrepresent-
ed by the groups most concerning to Gudjons-
son. We can benet by paying heed to his con-
cerns and recommendations.
Aldert Vrij is a leading authority and
researcher on the science of detecting decep-
tion. He has authored several books and nu-
merous chapters and research articles on the
subject. In chapter nine, Vrij updates us on
“myths and opportunities in verbal and non-
verbal lie detection”. This chapter should be
a must read for every police ofcer, police re-
cruit, attorney, judge, and criminal justice
professor or professional. Vrij summarizes the
myths surrounding the unassisted human lie
detector hypothesis. He provides examples of
evidence-based practices that actually do sep-
arate truthful from deceptive subject, though
Back to Front Cover

26 Polygraph, 2015, 44 (2)
Handler
the differences are small. He closes out with a
best practice recommendation for conducting
an investigative interview.
The nal chapter of the book is co-writ-
ten by a number of experts in Investigative In-
terviewing training. They provide a framework
for effective police interview training. They
share their thoughts and experience on the
best way to train new (and old) police investiga-
tors in Investigative Interviewing. Most of the
authors have been involved in police training
development for many years and have helpful
insights on successful training strategies.
St-Yves skillfully closes the book with
his concluding thoughts on the past and fu-
ture of Investigative Interviewing. He reminds
us that Investigative Interviewing has moved
from the realm of art into the realm of science
and art. Learning from scientists in allied dis-
ciplines will only improve what we do. There is
a great deal of evidence from the lab and eld
that supports these recommendations. St-
Yves recaps the essentials of good communi-
cation skills and their importance when inter-
viewing witness, complainants and suspects.
He reminds us that all good interviews require
preparation and a mindset towards unbiased
information gathering.
As polygraph examiners it seems we
should be ethically bound to conduct investi-
gative interviews during our pretest interview.
We have a unique opportunity to gather infor-
mation before conducting any test. That infor-
mation can be exculpatory, inculpatory, or in-
form the investigation in some important way.
If we approach the pretest interview as an in-
formation gathering event we can increase the
information gained during the testing process.
Many interviewees (truthful and deceptive) will
provide information during an appropriate in-
vestigative interview. They simply have to be
given the opportunity. The book Investigative
Interviewing - The Essentials can open your
eyes to a world of improvement. I seldom rec-
ommend any book with such enthusiasm, but
this is a rare occasion.
The French novelist Marcel Proust
said, “The real voyage of discovery is not in
seeking new landscapes but in having new
eyes.” I hope this book helps you see inter-
viewing through new eyes. It certainly did so
for me.
Back to Front Cover

27
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
Police Cadet Attrition and Training Performance Outcomes
Adam Park
Texas Department of Public Safety
James S. Herndon
Private Practice, Orlando, Florida
Abstract
Research devoted to the predictive validity of criteria commonly used to screen police applicants
has received little attention. The need has increased for police agencies to evaluate their various
screening methodologies in the multiple-hurdles approach to police candidate selection. Grounded
in Schmidt and Hunter’s theory of general mental ability in job performance, this study examined
the predictive validity of candidates’ demographic proles and results of the pre-academy screening
polygraph to predict training outcomes and attrition. This quantitative study used logistical and
linear regression analysis to determine whether these variables were viable screening mechanisms
to predict attrition and training performance among police cadets at the Texas Department of Pub-
lic Safety. Each independent variable (age, prior military service, level of education, and polygraph
result) predicted cadet academy completion status (unsuccessful; successful). However, there was
no evidence to suggest that age, prior military service, or level of education predicted training per-
formance as measured by nal academy grade point average or score on the Texas Commission on
Law Enforcement Ofcer Standards and Education (TCLEOSE) exam. This study’s ndings relative
to each independent variable support contemporary police research by identifying potentially val-
ued characteristics of a successful police candidate. Additionally, these ndings could allow police
administrators to better implement training strategies that compliment agency goals; thus better
preparing candidates to protect society. By understanding the validity of these screening procedures
in candidate selection, police agencies could save time and money.
Keywords: cadet, attrition, polygraph, screening, demographic prole
Police Cadet Attrition and Training
Performance Outcomes
From a practical point of view, the
most valuable determinant in training or job
performance is the predictive validity of essen-
tial work functions (Schmidt & Hunter, 1998).
However, in an ever-changing world, it is be-
coming increasingly difcult to nd quality
personnel for the police profession (Henson,
Reyns, Klahm, & Frank, 2010). Additionally,
scholars continue to debate which variables
are the best predictors of quality personnel
(Henson, et al., 2010). This leaves a gap with
respect to which other variables consistently
predict the best cadet. Given the existing array
of selection criteria police agencies use in the
preemployment process (e.g., age, prior mili-
tary service, psychological and physical agility
testing), it is important to examine their rela-
tionship to police cadet performance.
For more than 90 years researchers
have studied psychological test data in the
Author Note
The views expressed in this paper do not necessarily reect those of the Texas Department of Public Safety. Correspondence
concerning this article should be addressed to Adam Park Ph.D., Cypress, TX 77429.
Contact: hyperion1975@aol.com
Back to Front Cover

28 Polygraph, 2015, 44 (2)
Park, Herndon
context of selecting law enforcement personnel
(e.g., Terman & Otis, 1917); yet there exists
some disagreement regarding the issue of best
practices in the “multiple hurdles” approach to
police cadet selection. Today, most candidates
for the position of police cadet must, as a part
of the screening process, participate in a back-
ground investigation and submit to psycho-
logical and polygraph tests (Fuss & Snowden,
2004). Contemporary screening methods also
generally include written tests, physical agil-
ity tests, and meeting certain demographic
thresholds (Decicco, 2000). These methods
have dual functions: to select the most com-
petent candidate for training as well as vet the
candidate to ensure that agency goals are met
and public safety is not compromised (Decic-
co, 2000).
Many organizations employ various
selection techniques in the preemployment
process, and often take action relative to the
weights given to each point of view (Society
for Industrial and Organizational Psychology,
2003). What this means for the cadet selection
process is that the police agency is often bur-
dened with the cost in time and money of us-
ing numerous techniques. Questions remain
about which predictors correlate best with
police cadet performance (White, 2010). The
majority of psychological screening for police
cadets in the United States uses personality
measures (Hancock & McClung, 1987). Re-
cent studies argue for the use of non-clinical
personality assessments for police cadet selec-
tion (Forero, Pujol, Olivares, & Pueyo, 2009);
while other studies suggest that demographics
and residency have little to do with overall ca-
det success (White, 2010).
Important to this discussion is a
strategy that effectively ties selection mecha-
nisms that are part of the demographic and
pre-academy screening data to cadet attrition
and training performance. This, it was sug-
gested, would provide a better bridge to ex-
isting disparities in police candidate selection
research. There is a gap in the empirical re-
search that connects the predictive validity of
each predictor in this study to cadet attrition
and training performance. Additionally, little
research has been devoted to using multiple
predictors to analyze police academy perfor-
mance; with current efforts leaning heavily
toward using psychological traits to predict
the job performance of a police ofcer (White,
2008). Noting the importance of nding plau-
sible links to the multiple predictors selected
in this study, it is appropriate to make a brief
introduction to the multiple hurdles approach
to general and police employee selection (e.g.,
application, demographic characteristics used
in the hiring process, interview, background
investigation, psychological testing, and integ-
rity testing). Additionally, a brief discussion of
the history of the Texas Department of Public
Safety (DPS) will provide the context for using
the demographic data in this study.
Literature Review
The Texas DPS uses a non-compen-
satory model-better known as a successive
or multiple hurdles approach to police cadet
selection, where the applicant must pass a
series of tests (Henson et al., 2010). For ex-
ample, failing the physical tness test elimi-
nates a candidate from further consideration;
the same is true of failing or performing below
established thresholds in any screening pro-
cess (medical, psychological, polygraph). In its
equation for selecting cadets, the Texas DPS
uses certain demographic criteria (minimum
age, prior military service, education). There
is no ceiling on age, but applicants must be at
least 20 years of age upon graduation from the
academy (Texas DPS, 2012). Applicants bring
a mixture of experiences to the preemployment
process. The Texas DPS addresses the issue
in a number of ways. While applicants for the
position of police cadet at the Texas DPS must
have earned at least 60 credit hours from a re-
gionally accredited college or university (Texas
DPS, 2012), they can substitute prior military
or law enforcement service for college educa-
tion, equivalent to the 60-hour requirement.
Police agencies use many screening
items in both the preemployment and training
academy process (White, 2010). Harris, Dwor-
kin, and Park (1990) examined the predictive
validity of numerous screening procedures
in the hiring process for large public compa-
nies. The three notable screening procedures
reported as most accurate in selecting future
superior workers were targeted interviews, ac-
complishment tests, and references. In con-
trast, physical ability tests, polygraph tests,
and genetic tests were deemed most inaccurate
in the selection of a superior worker; however,
failed drug and polygraph tests were two of the
most likely reasons employers gave that would
affect hiring decisions (Harris et al, 1990). In
order to provide a more broad understanding
of applicant selection processes, a portion of
the screening procedures utilized in previous
studies incorporating both private companies
and police agencies are included as predictors
in the present study. For example, acade-
Back to Front Cover

29
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
my tests, as well as the state licensing exam
were utilized as predictors for accomplishment
tests. Another study noted the work sample is
considered to have a high degree of validity be-
cause of its practical nature (Osoian, Zaharie,
& Lazar, 2011), while other employers look to
cognitive ability tests as a predictor of hiring
success (Lewis, Shimerda, & Graham, 1983).
Age as a Predictor of Training Performance
Perhaps no predictor in this study
has been underrepresented in empirical po-
lice cadet selection research than that of age.
There have been numerous attempts to ex-
plain age-related cognitive changes such as
common cause theory, the processing speed
theory, and the executive function hypothesis
(Luszcz & Byran, 1999); however, these efforts
have been difcult due to the various differ-
ences in statistical ndings which make gen-
eralizations about age across time arduous at
best (Clay et al., 2009). The specic focus in
this study was to examine age as a predictor of
cadet attrition and training performance us-
ing the theory of General Mental Ability GMA.
Therefore, in this study, GMA would be repre-
sented by the cadet’s c academic test results
in a 28- week police academy. Age has long
been judged a great variable in organization-
al contexts due to its common relationship
to the goals and objectives of many agencies
(Randhawa, 2007). Obvious correlations exist
utilizing age as a predictor of physical perfor-
mance in any setting; however, this study does
not examine physical tness as a predictor of
training performance.
It is appropriate to now examine pre-
vious studies to assist in establishing a back-
ground perspective of age’s relationship within
law enforcement training contexts. Chappell
(2008) used age, race/ethnicity, gender, mil-
itary experience, education, special position
(ranking ofcer in class), and type of acade-
my as independent variables to assess cadet
performance in a re-designed academy (com-
munity policing) versus a traditional academy
setting. A literature review revealed one study
that included age and training outcomes.
White (2008) examined a large metropolitan
police training class (n = 1,556) and found
that as a cadet’s age increases, their academy
performance decreases. The practical value of
age as it relates to police training appears to
be largely unknown, thus making it necessary
to examine this predictor’s current status in
police screening contexts.
Biographical data has been substan-
tially correlated with GMA (Schmidt, 1988).
Some scholars offer that police agencies prefer
younger applicants because they are less rigid
and more accepting of the community-based
policing model recently adopted by many po-
lice agencies (LaRose, Caldero, & Gutierezz,
2006). Aamodt’s (2004) review of 300 theses,
dissertations, journal articles, and conference
papers yielded several studies addressing
academy performance, but included age only
as a sample characteristic. However, recent re-
search covering broad spectrum occupations
(i.e. occupations other than public safety) as-
serts that workforce age is linked negatively to
quantitative organizational performance, but
positively associated with qualitative perfor-
mance (Gellner, Schneider, & Veen, 2011). In
other words, output appears to be correlated
with being younger, whereas an older work-
er shows to produce more quality in their
product. One viewpoint on age as it relates to
police performance posits a likely paradigm
shift from hiring younger police applicants,
to recruiting older prospects. For example,
some police departments have sought out sec-
ond-career applicants from other professions
not because of their previous law enforcement
experience, but due to their maturity and doc-
umented stability (Bennett & Hess, 2004).
Other studies examining age and work perfor-
mance report that demographics point to an
older workforce (Sharit, Czaja, Hernandez &
Yang, Y, 2004). As it has been suggested, the
reports in this area are dubious and provide
for a perplexing understanding of age and po-
lice training performance.
The Utility of Prior Military Service
Police training programs have tended
to follow militaristic patterns, which explains,
in part, rationales for police agencies hiring
former military members as patrol ofcers
(Birzer, 2003). Historically, there has been a
strong representation of individuals in law
enforcement with prior military experience
(Aamodt, 2004). Evidence of this can be seen
in the roots of many police units throughout
the United States. For example, police agen-
cies adopted quasi-military models in the early
1900s in efforts to eliminate corruption (Fogel-
son, 1977).
Despite strong academic opinions,
scholars continue the debate regarding the
practical value and contextual meaning of the
word “paramilitary”. Law enforcement agen-
cies traditionally have adopted a “paramil-
Back to Front Cover

30 Polygraph, 2015, 44 (2)
Park, Herndon
itary” model, despite a lack of empirical evi-
dence supporting the utility of a military struc-
ture within law enforcement contexts (Bittner,
1990; Franz & Jones, 1987). Some scholars
debunk the notion of police as a paramilitary
organization (Cowper, 2000), while others
agree the two share commonalities, but de-
bate about a common denition of paramili-
tary when applied to a law enforcement envi-
ronment (Jefferson, 1987, 1993; Waddington,
1993). Recent ndings on the nature of polic-
ing suggest there appears to be a convergence
of roles between police and military function
(Campbell & Campbell, 2010). Prior research
reveals positive correlations with police work
performance, but there remains a void with
respect to the usefulness of military service to
police training.
College as a Determinant of Success
It is becoming more common for appli-
cants who enter the police cadet selection pro-
cess to have some level of college education,
as studies reveal that 44% of police applicants
have attended at least 1 year of college (Ben-
nett & Hess, 2004). In contrast to the previous
discussion on military service, statistics reveal
that police applicants now have more college
education than military background (Bennett
& Hess). Using college experience as a pre-
dictor variable is an important component of
this study given its unknown predictive valid-
ity in police training contexts. Another reason
for the examination of college education as a
predictor lies with the fact that the literature
has produced mixed reviews with respect to
correlating formal education and training suc-
cess (Walker, 1994).
Criminal justice education has typically
focused on a distinct, three-level system (high
school diploma, associate’s degree, and bach-
elor’s degree) that in many ways ties into en-
try-level law enforcement jobs (Buerger, 2004).
As a practical matter, the previous statement
assists in establishing criteria for incorporat-
ing the three-level system (high school diplo-
ma, associate’s degree, and bachelor’s degree),
with the addition of some college experience,
as predictors. The practical value of having a
college degree in the police profession is still
largely misunderstood (Walker, 1994). One ra-
tionale for this uncertainly is grounded in the
thought that individuals make decisions rela-
tive to pursing higher education on the foun-
dation of a cost-benet analysis (Brand & Xie,
2010).
As early as 1916, August Vollmer, the
father of modern policing, underscored the
importance of education for ofcers (Guthrie,
2000). The Wickersham Commission (1937)
and the President’s Commission on Law En-
forcement and the Administration of Jus-
tice (1967) highlighted the signicance of a
post-secondary education for police ofcers
(Bennett & Hess, 2004).
Although the previous discussion not-
ed that more applicants come to the screen-
ing process with some level of college educa-
tion, studies show that this is certainly not
an expectation (Capsambelis, 2004). A study
from the Bureau of Justice Statistics not-
ed that nationally only 1% of police agencies
required a 4-year degree; whereas only 6%
required some college, and 8% mandated a
2-year degree (Paoline & Terrill, 2007). Advo-
cates of the college experience argue that ed-
ucated ofcers produce better reports, receive
fewer complaints, and produce a better over-
all work product (Baker, 1995; Carter et al.,
1989; Trautman, 1986; Vodicka, 1994). Oth-
er scholars stated that the college experience
provides a better opportunity for an individu-
al to mature, offers a broader base of general
knowledge, and enhances verbal and commu-
nication skills (Armstrong & Polk, 2002). Tak-
ing this evidence into account, administrators
would perhaps suggest the college experience
a valuable criterion for cadet selection.
The TCLEOSE Exam
In 1965 the Texas State Legislature
created the Texas Commission on Ofcer
Standards and Education (TCLEOSE) to es-
tablish standards for peace ofcers (TCLE-
OSE, 1997). The Basic Peace Ofcer Course
Format consists of 618 hours of academics
related to entry-level policing and is a test of
job content knowledge (TCLEOSE, 2008). The
TCLEOSE Exam is comprised of 250 multiple
choice questions that address: (a) Texas Penal
Code; (b) Texas Code of Criminal Procedure;
(c) The Texas Constitution; (d) Texas Trafc
Law; (e) Drug Questions; (f) Police approaches
to family violence and mental health; and (g)
Civil Law (TCLEOSE, 2010). Applicants for the
position of police cadet, who are not already
peace ofcers, must rst pass the TCLEOSE
Exam before becoming commissioned troopers
(Texas DPS, 2012). This study incorporated
only datasets from subjects who had not pre-
viously taken the TCLEOSE exam.
A comprehensive literature review re-
vealed that great disparity exists in the realm
of research related to the validity of state po-
lice licensing examinations. Although not a
civil service test, the TCLEOSE exam serves to
Back to Front Cover

31
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
assess the cadet’s knowledge of general police
aptitude. Schroeder (1973) studied the validity
of the entrance examination for the position
of patrolman under the guidelines established
by the Equal Employment Opportunity Com-
mission (EOEC) and found exam scores were
positively related to performance. A review of
literature relating to the validity of the TCLE-
OSE exam yielded no results. However, it was
discovered Peace Ofcer Standards and Train-
ing (POST) commissions are located in every
state so as to set minimum requirements for
entry-level law enforcement positions (Bennett
& Hess, 2000).
The Polygraph
Although empirical studies fail to in-
vestigate polygraph’s predictive validity for-
training performance in the screening process
for ofcers, literature suggested the method of
preemployment polygraph is only increasing
(Krapohl, 2002). Additionally, there is strong
evidence that integrity tests have practical ap-
plication when paired against cognitive ability
(Schmidt & Hunter, 1998). For example, GMA
has produced more incremental validity regard-
ing the prediction of private employee training
performance than any other measured study
to date (Schmidt & Hunter, 1998). Meesig and
Horvath (1995) reported that 99% of large and
90% of small law enforcement agencies re-
quire the use of a polygraph as a condition of
employment for sworn positions. The reliabili-
ty of a candidate’s truthfulness is of high value
to police administrators throughout the police
preemployment process. A meta-analysis of
integrity test validities found that preemploy-
ment tests of honesty can predict certain orga-
nizational disruptive behaviors (Ones, Viswes-
varan, & Schmidt, 1993). Extending on this
idea, the prevalence of general preemployment
polygraph screening appears to be on the rise.
The use of the polygraph is prohibited in most
private sector arenas because of the Employ-
ee Polygraph Protection Act of 1988 (EPPA)
(Decicco, 2000). Platform for the inclusion of
polygraph results in this study lies in the fact
that virtually nothing is known about the ex-
isting relationship between polygraph results
in preemployment settings and training per-
formance in police cadets or ofcers. This is
the rst known study to incorporate polygraph
examinations as a predictor of police cadet at-
trition and cadet performance.
One study (Ho, 2001) was found to
have utilized polygraph results as one predic-
tor for assessing the effect of a psychologist’s
recommendations for hiring. Ho (2001) used
linear regression to examine the effects of in-
dependent predictors (demographics, gender,
age, prior military service, self-reported drug
usage, and prior encounters law enforcement)
on each dependent measure. Past research has
even examined the correlation between civilian
preemployment tests and future employee be-
havior. Although much controversy still exists
surrounding the validity of polygraph, many
police departments today use this instrument
as a tool for veracity in the ofcer selection
process (e.g., Ben-Shakhar & Furedy, 1990;
Lykken, 1981; Saxe, 1994). This would imply,
in a purely non-systematic way that police ad-
ministrators have found a certain value to the
requirement of a polygraph as a condition of
employment. More research is needed in the
area of preemployment polygraph to facilitate
knowledge with respect to its utility in these
settings.
Theoretical Base
The theoretical base for this study was
grounded in the framework of predicting oc-
cupational performance and the validity of
paired combinations of general mental abili-
ty (GMA); dened as the outcome of GPA and
scores from the TCLEOSE exam. This theory
was introduced in 1904 by C. Spearman and,
as Schmidt and Hunter (2004) noted, is often
used for predicting occupational performance:
“GMA predicts both occupational level attained
and performance within one’s chosen occupa-
tion and does so better than any other abil-
ity, trait, or disposition” (p. 162). According
to Schmidt and Hunter (1998) the percentage
of validity for preemployment personnel mea-
sures (i.e., integrity tests, biographical data
measures, and years of education) increase
together. The thrust of this study was to ex-
amine this theory by assessing GMA’s ability
to predict training performance.
There are two reasons for using GMA:
(a) it has the highest level of validity in per-
sonnel selection and the lowest cost in terms
of monetary measurement (Schmidt & Hunter,
1998) and; (b) it repeatedly provides the best
evidence of validity among other measures
(Hunter, 1986; Hunter & Schmidt, 1996; Ree
& Earles, 1992; Schmidt & Hunter, 1981).
This evidence makes GMA a viable avenue
for future research on selecting cadets. In
the present study, grade point average (GPA)
and TCLEOSE exam scores were reported as
a measure of GMA (e.g. cadet GPA at end of
academy and TCLEOSE score). From a the-
oretical perspective, the utilization of GMA,
Back to Front Cover

32 Polygraph, 2015, 44 (2)
Park, Herndon
along with multiple predictors (i.e. age, prior
military service, and education) was the best
choice, considering the similarity of items used
in Schmidt and Hunter’s (1998) work and the
screening methods used by the Texas DPS in
the preemployment process.
The theory of GMA embraces general
intelligence and specic aptitudes and abili-
ties; it then shows important differences be-
tween groups (Schmidt & Hunter, 1998). This
was important considering the proposed re-
search questions in this study. Finally, ele-
ments that comprise GMA have been exam-
ined in both military and police occupations,
thus supporting the idea of using the theory
to predict training performance in a police
academy. Grade Point Average (GPA) and the
Texas Commission on Law Enforcement Of-
cer Standards and Education (TCLEOSE)
exam score were considered outcome or cri-
terion variables since they were generated by
the applicant throughout the police academy
training.
The purpose of this quantitative study
was to examine whether the two sets of vari-
ables, demographic proles and pre-academy
polygraph screening results, were signicant
predictors of police cadet attrition and train-
ing performance at the Texas DPS. It incor-
porated data from both the preemployment
process and the training academy. Many
studies have generalized various independent
and dependent variables in attempts to cor-
relate demographic data with training and job
performance (Aamodt, 2004). Some studies
have used biographical data with test data,
while others have mixed interview data with
archived data. However, no study was found
in the literature that examined the predictive
validity of age, education, academic perfor-
mance, and polygraph results to predict cadet
attrition and training performance.
Demographic variables are aspects
of the applicant they bring into the selection
process. For example, demographic variables
in this study were age, prior military service,
and level of education. A fourth predictor, pre-
employment polygraph results, was analyzed
to assess the polygraph’s ability to predict at-
trition as well as training performance. The
polygraph result was incorporated as a predic-
tor to determine its usefulness for predicting
attrition rates among police candidates.
All candidates were administered a
polygraph examination before admission into
the academy. As per DPS policy, candidates
determined to be either inconclusive or de-
ceptive were interviewed and administered a
second “break-out” examination. This process
is established to assist the candidate in clear-
ing inconsequential issues related to hiring,
or to obtain additional information that may
disqualify the candidate from the hiring pro-
cess. Candidates receiving a second deceptive
or inconclusive result were disqualied from
the process. Those candidates who were ad-
ministered a second break-out polygraph and
scoring no deception were admitted to the
academy. Because a portion of the candidates
in the rst inconclusive and deceptive group
were ultimately cleared and admitted to the
academy, they were separated from the rst
no deception rst polygraph test group. This
was considered by the researcher to be of val-
ue in determining the group’s predictability in
training success as opposed to the group who
received no deception on the rst examination.
It was hypothesized that younger ca-
dets who possessed a college degree and prior
military experience would perform better in a
training academy environment as measured
by GPA and the TCLEOSE exam score, as op-
posed to older cadets who possessed no formal
college education or prior military service. Fi-
nally, it was hypothesized that cadets entering
the academy with a polygraph result of “no de-
ception indicated” on their rst polygraph ex-
amination had a better chance of completing
the academy, as opposed to those cadets hav-
ing either a rst examination result of “incon-
clusive” or “deception indicated”; thus requir -
ing a subsequent “break-out” examination.
Hypotheses:
H1
1: The demographics of cadet age,
military experience, and level of education are
predictive of academy completion for cadets
who participated in the 2008 Texas Depart-
ment of Public Safety Training Academy.
H1
2: The demographics of cadet age,
military experience, and level of education are
predictive of overall training performance for
cadets who participated in the 2008 Texas De-
partment of Public Safety Training Academy.
H1
3: Preemployment polygraph results
of “No Deception Indicated” are predictive of
academy completion for cadets who partici-
pated in the 2008 Texas Department of Public
Safety Training Academy.
Back to Front Cover

33
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
Methodology
This quantitative study used a non-ex-
perimental, descriptive design. Cadet demo-
graphics and pre-academy polygraph screen-
ing results were captured from Texas DPS re-
cords. Categorizing these predictor variables
was consistent with Aamodt’s (2004) research
which examined each variable used in this
study, other than polygraph test results, to
police ofcer training and job performance.
The demographics of the population under ex-
amination provided for a largely unied sam-
ple. This study included stratication of the
population; incorporating true characteristics
(e.g., White/Hispanic males) representative of
the entire sample (Fowler, 2002). Specically,
195 subjects were examined to nd if demo-
graphic proles and pre-academy polygraph
screening results were signicant predictors
of attrition and training performance by the
selected population. Eligibility for group clas-
sication included: (a) Each participant was
to be an applicant for the position of police
cadet at the Texas DPS for the 2008 acade-
my; (b) The participant must have completed
demographic information which documented
age, any prior military service, and level of ed-
ucation as criteria for selection; (c) The partic-
ipant had to numerically score no deception
indicated, inconclusive, or deception indicat-
ed on a rst or second attempt of the Texas
Department of Public Safety Modied General
Question Technique (DPSMGQT); and (d) The
participant must have either failed to complete
training after beginning the academy, or grad-
uated from the Texas DPS academy with nu-
merical scores for grade point average (GPA),
as well as participated in a rst attempt of the
TCLEOSE exam, which provided a numerical
score. Cadets are allowed to take the examina-
tion multiple times; therefore, the researcher
included only numerical scores from the rst
attempt of the TCLEOSE exam so as to better
compliment true test-taking ability. Prior po-
lice ofcers who held a TCLEOSE license at
the time of training were not included in any
dataset; as they were exempt from the state
examination.
Instrumentation
Demographic data was derived from
the preemployment polygraph questionnaire
form (HR-39). Outcome variables were collect-
ed by obtaining scores from each cadet’s cu-
mulative grade point average (GPA) and score
on the Texas Commission on Law Enforce-
ment Ofcer Standards and Education (TCLE-
OSE) exam. Finally, pre-academy screening
polygraph results were gathered from data
produced on the Texas DPS Modied Gener-
al Question Technique (DPSMGQT). As previ-
ously discussed, the TCLEOSE exam is com-
prised of 250 multiple choice questions and
covers the topics of: (a) Texas Penal Code; (b)
Texas Code of Criminal Procedure; (c) Texas
Constitution; (d) Texas Trafc Law; (e) Drug
Questions; (f) Police approaches to family vi-
olence and mental health and; (g) Civil Law
(TCLEOSE, 2012). The polygraph instrument
used for the collection of all physiological data
was the Axciton Five-Channel Computerized
Polygraph System. Channels within polygraph
contexts refer to the individual components
attached to the examinee. The data gathered
from polygraph examinations administered in
this study was exclusively archived. Those ex-
aminations were conducted at two locations:
(a) The Texas DPS Headquarters in Austin,
Texas- Physical Address 5805 North Lamar
Boulevard, and (b) Building M at the Head-
quarters Complex in Austin, Texas.
Analysis
This study was driven by past meth-
odology, as a number of studies have incor-
porated regression analysis in attempts to as-
sess cadet training outcomes (Aamodt, 2004;
Guller, 2003; Jacobs & Solomon, 1977; Sand-
ers, 2008; Waugh, 1996), though no known
study existed that incorporated the multiple
screening criteria in this proposal in attempts
to predict cadet attrition and training perfor-
mance. Research consistently shows that re-
gression analysis is the most appropriate sta-
tistical treatment with respect to predicting
police academy performance, as it permits the
relationship between variables to be inspect-
ed due to the correlation of variables (Bern-
stein, Schoenfeld, & Costello, 1982; Waugh,
1996). The core of the analysis was linear and
logistical regression for all three hypotheses.
Regression analyses are used to predict the
relationship between predictor variables and
outcome or criterion variables (Tabachnick
& Fidell, 2001). More specically, regression
analyses are used to study the relationship of
a dependent variable y to two or more inde-
pendent variables, using a regression model
that is represented by the equation y = ß0 +
ß1x1 + ß2x2 + ɛ. This model also includes ɛ, or
an error variable, which is a random variable
that refers to the variability in y that the listed
independent variables do not account for.
Training performance (GPA and TCLE-
OSE exam score) and cadet attrition were
treated as two separate analyses. Addition-
Back to Front Cover

34 Polygraph, 2015, 44 (2)
Park, Herndon
ally, GPA and TCLEOSE results were not re-
ported for a cadet that failed to complete the
academy. The rationale behind this method-
ology was based on the premise that a cadet
would be required to graduate in order to cap-
ture a cumulative GPA and qualify to sit for
the state licensing exam. Within this context,
a cadet selected to the academy that ultimate-
ly resigned for undisclosed reasons or failed
out was included in the polygraph dataset,
as this supported the hypothesis for cadet at-
trition. Cadets that were expelled or resigned
were treated as analogous.
Results
Among the 195 study participants, 190
(97.4%) were male and 5 (2.6%) were female.
The ethnic distribution was 132 (67.7%) White;
56 (28.7%) Hispanic; and 7 (3.6%) Black. The
education distribution was 48 (24.6%) high
school graduate or possessing a GED; 44
(22.6%) with some college; 40 (20.5%) pos-
sessing an associate’s degree, and; 63 (32.3%)
holding a bachelor’s degree. The average (and
standard deviation) age was 27.8 (6.4) and the
range was 20 to 52. There were 121 (62.1%)
study participants that had no prior military
service and 74 (37.9%) with prior military ser-
vice. A total of 150 (76.9%) study participants
completed the academy and 45 (23.1%) failed
to complete the academy. The average (and
standard deviation) academy GPA was 92.1
(3.35) and the range was 83.1 to 97.6. The av-
erage (and standard deviation) TCLEOSE exam
score was 80.2 (6.2) and the range was 67 to
93. Polygraph decisions were used as the mea-
sure for inter-rater reliability. Cohen’s Kappa
statistic K to measure inter-rater reliability of
the polygraph test (n = 10 polygraphs; 2 rat-
ers) was .40 (moderate agreement). A total of
121 (62.1%) study participants had a poly-
graph result of “No deception indicated” and
74 (37.9%) “Inconclusive or deception indicat-
ed” (see Tables 1 - 3).
Table 1 Descriptive Statistics for Ethnic Distribution
Frequency %
Valid
percent Cumulative %
White 132 67.7 67.7 67.7
Hispanic 56 28.7 28.7 96.4
Black 73.6 3.6 100.0
Male
Female
Total
190
5
195
97.4
2.6
100.0
97 . 4
2.6
100.0
Note: Among the 195 study participants, 190 (97.4%) were male and 5 (2.6%) were female. The eth-
nic distribution was 132 (67.7%) Caucasian; 56 (28.7%) Hispanic; and 7 (3.6%) Black.
Back to Front Cover

35
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
Table 2 Descriptive Statistics for Age by Academy Completion Status
Academy
Completion Status NMean SD Minimum MaximumValid Missing
Unsuccessful 45 0 29.67 6.779 21 52
Successful 150 0 27.19 6.147 20 52
Note: The average age was signicantly smaller for those who completed the academy compared to
those who failed the academy. The average (and standard deviation) age was 29.7 (6.8) versus 27.2
(6.1) for those who failed the academy and those who completed the academy, respectively, t(193) =
2.31; p = .022. This nding was consistent with the ndings from hypotheses 1 and 3 where older
age was associated with lower odds of completing the academy.
Table 3 Frequency Distribution of Study Participant’s Polygraph Result
Frequency Percent
No Deception Indicated 121 62.1
Inconclusive or
Deception Indicated
74 37.9
Total 195 100.0
Note. Ten polygraph examinations were scored by two examiners. Their scores were then utilized to
calculate inter rater reliability using the K statistic.
Table 4 Stepwise Multiple Logistic Regression Analysis of Academy Completion versus Age,
Prior Military Service, and Level of Education
b SE Wald df p-value OR e95% C.I. for OR
Lower Upper
Model aAGE b-.078 .028 7.617 1 .006 .925 .875 .978
PMS c1.232 .423 8.485 1 .004 3.429 1.497 7.857
EDU1 d-1.017 .400 6.448 1 .011 .362 .165 .793
Constant 3.289 .829 15.729 1 .000 26.816
Note. The most important predictor of completion status was prior military service, age, and
nally level of education. The three independent variables collectively explained 10.4% of the
variance in completion status.
Hypothesis 1 was tested using step-
wise multiple logistic regression analysis. The
dependent variable was academy completion
status (successful; unsuccessful). As Table
4 illustrates, all three independent variables
were found to be statistically signicant. This
means that prior military service (p = .004),
age (p = .006), and level of education (p = .011)
provided independent information in predict-
ing academy completion status. That is, the
three independent variables explained inde-
pendent variance in academy completion sta-
tus. The Nagelkerke R-Square Statistic asso-
ciated with prior military experience was .05;
.054 for age, and; .045 for level of education.
Back to Front Cover

36 Polygraph, 2015, 44 (2)
Park, Herndon
Table 5 Stepwise Multiple Linear Regression Analysis of Academy Grade Point Average a versus
Age, Prior Military Service, and Level of Education
Model a
Unstandardized
Coefcients
Standardized
Coefcients t p-valueb SE Beta
(Constant) 91.729 1.351 67.884 .000
AGE b.002 .047 .004 .048 .962
PMS c-.261 .639 -.039 -.409 .683
EDU1 d.156 .860 .018 .181 .856
EDU2 e.043 .861 .005 .050 .960
EDU3 f1.062 .805 .151 1.320 .189
Note. It was found that neither age, prior military service, or education level were predictive of acad-
emy GPA. Age = cadet age (measured on a continuous measurement scale in years); PMS =prior
military service; EDU = level of education; b = Estimated values of raw (unstandardized) regression
coefcients; SE = standard error.
Hypothesis 2 was tested using step-
wise multiple linear regression analysis. There
were two separate measures of training per-
formance (dependent variables): (1) Academy
GPA, and; (2) TCLEOSE exam score. There-
fore, this analysis was repeated for each of the
two measures. For the rst regression analy-
sis, the dependent variable was the academy
GPA. The independent variables entered into
the stepwise model selection procedure were
age (measured on a continuous measurement
scale in years), prior military experience (0 =
No; 1 = Yes), and; level of education. As was
done in testing hypothesis 1, education was
re-coded into dummy variables prior to con-
ducting the analysis. None of the independent
variables met criteria for entry into the model
(i.e. p < .05). As Table 5 illustrates, the null
hypothesis was not rejected and it was con-
cluded that neither age, prior military service,
or education level were predictive of academy
GPA. The p-values for each of the independent
variables were .96 for age; .68 for prior mili-
tary service; .86 for EDU1 (high school diplo-
ma); .96 for EDU2 (some college) and; .19 for
EDU3 (college degree).
For the second regression analysis, the
dependent variable was the academy TCLE-
OSE exam score. The independent variables
entered into the stepwise model selection pro-
cedure were age (measured on a continuous
measurement scale in years), prior military
experience (0 = No; 1 = Yes), and; level of edu-
cation. As was done previously, level of educa-
tion was re-coded into dummy variables prior
to conducting the analysis. Table 6 shows that
only EDU2 (some college experience) met cri-
teria for entry into the model, p = .023. The
null hypothesis was not rejected and it was
concluded that age, prior military service, and
level of education do not explain indepen-
dent variance in TCLEOSE exam scores. The
equation of the model is: TCLEOSE = 80.82
-2.82*EDU2, where TCLEOSE = the average
Texas Commission on Law Enforcement Of-
cer Standards and Level of Education Score;
EDU2 = level of education (0 = Not Associate’s
degree; 1 = Associate’s degree).
Back to Front Cover

37
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
Table 6 Stepwise Multiple Linear Regression Analysis of TCLEOSE a Exam Scores versus Age,
Prior Military Service, and Level of Education
Model a
Unstandardized
Coefcients
Standardized
Coefcients t p-valueb SE Beta
(Constant) 80.822 .566 142.788 <.001
EDU2 b-2.822 1.225 -.186 -2.303 .023
Note. Only EDU2 (some college experience) met criteria for entry into the model,
p = .023. The null hypothesis was not rejected and it was concluded that age, prior
military service, and level of education do not explain independent variance in
TCLEOSE exam scores. b = Estimated values of raw (unstandardized) regression
coefcients; SE = standard error; t = sample value of the t-test statistic; p-value =
probability value.
Hypothesis 3 was tested using sim-
ple logistic regression analysis. The depen-
dent variable was academy completion status
(successful; unsuccessful). The independent
variable was the initial polygraph test result
(“No deception indicated”; Inconclusive or De-
ception Indicated”). The null hypothesis was
rejected and it was concluded that polygraph
test results are predictive of academy comple-
tion (see Table 4). The Nagelkerke R Square
statistic was .044, which means that the poly-
graph test results explained only 4.4% of the
total variance in academy completion status.
Considering that the results for testing
Hypothesis 1 showed that age, prior military
service and level of education were predictive
of academy completion status, it was of inter-
est to determine if the polygraph test result ex-
plained additional variation in academy com-
pletion status, above and beyond the variance
explained by the three demographic variables.
A stepwise multiple logistic regression anal-
ysis was performed in order to address this.
The independent variables entered into the
stepwise model selection procedure were age,
prior military experience, initial polygraph test
result, and level of education. All four inde-
pendent variables were statistically signi-
cant. This means that polygraph test result (p
= .011), age (p = .003), prior military service
(p = .003), and education level (p = .011) pro-
vided independent information in predicting
academy completion status. That is, the four
independent variables explained independent
variance in academy completion status. The
Nagelkerke R Square statistic was .054 for
age; .050 for prior military service; and .048
for polygraph test result, and; .043 for level
of education. Thus, the most important pre-
dictor of completion status was age, followed
by prior military service, polygraph test result,
and level of education. The four independent
variables collectively explained 19.5% of the
variance in completion status.
Back to Front Cover

38 Polygraph, 2015, 44 (2)
Park, Herndon
Figure 1 illustrates the age disparity between successful and unsuccessful cadets.
Back to Front Cover

39
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
Exploratory Analysis
The results of testing Hypotheses 1
and 3 showed that age, prior military service,
level of education, and polygraph test results
contributed independent information in pre-
dicting academy completion status. In order
to further explore the relationships between
the independent and dependent variables, bi-
variate analyses were conducted. A two-sam-
ple t-test was used to compare the average age
between those who did, and did not, complete
the academy. Chi-square tests were used to
evaluate the relationships between prior mil-
itary service, education level, and polygraph
test results, and the dependent variable, acad-
emy completion status.
The average age was signicantly
smaller for those who completed the academy
compared to those who failed the academy.
Figure 1 is an error bar chart that shows the
average and 95% condence interval for the
average age by academy completion status.
The gure gives strong evidence that those
who completed the academy tend to be young-
er on average, compared to those who fail the
academy. The prole for Figure 1 illustrates
the 95% condence interval for age, as well
as distributes the two categories of cadet age
by completion status. Figure 1 clearly reveals
disparity between the two status groups. The
average (and standard deviation) age was 29.7
(6.8) versus 27.2 (6.1) for those who failed the
academy and those who completed the acade-
my, respectively, t(193) = 2.31; p = .022. This
nding was consistent with the ndings from
hypotheses 1 and 3 where older age was asso-
ciated with lower odds of completing the acad-
emy.
A chi-square test was performed in
order to determine if there was an associa-
tionbetween level of education and academy
completion status. There was not a statisti-
cally signicant difference in the percentage
of cadets that completed the academy among
the four education groups, X² (3) = 5.85; p =
.12. Although not statistically signicant, the
largest standardized residual (in absolute val-
ue) was 1.8, which shows the “some college”
group contributed the most to the magnitude
of the Chi-Square statistic. In particular, the
percentage of cadets with “some college” that
completed the academy (63.6%) was less than
the percentage that completed the academy
among the other 3 education groups (79.2%
to 82.5%). This nding was consistent with
the results from hypotheses 1 and 3 where the
“some college” group was found to have low-
er odds of completing the academy compared
to the other education levels. Specically, age,
prior military status and polygraph test re-
sults explained some of the variance in acad-
emy completion status. The residual variance
(the variance left over) could be better attribut-
ed to other factors (i.e. level of education).
A chi-square test was performed in or -
der to determine if there was an association-
between polygraph test results and academy
completion status. There was a statistically
signicantly smaller percentage of cadets that
completed the academy with initial “Inconclu-
sive or deception indicated” polygraph results,
compared to those with a polygraph test result
of “no deception indicated”. The number (and
percentage) of cadets that completed the acad-
emy was 100 (82.6%) versus 50 (67.6%) for the
“No deception indicated” and “Inconclusive or
deception indicated” groups, respectively, X ²
(1) = 5.88; p = .015. This nding was consis-
tent with the ndings from Hypotheses 1 and
3 where a polygraph test result of “inconclu-
sive or deception indicated” was associated
with lower odds of completing the academy.
Conclusion
There is an increasing body of re-
search within police personnel selection; how-
ever, gaps remain as to what predictors tend
to make the best cadet. Extending on the
aforementioned statement, future research
should consider how the process of training
translates into making a better police ofcer.
Physical tness should be examined from the
perspective of its impact to cadet attrition.
Within this context, age would, on the sur-
face, appear to hold some validity when con-
sidering high rates of cadet fallout. Anecdotal
evidence suggests younger cadets would have
less difculty than older cadets in completing
the physical challenges associated with state
police academies. Research should consider
peripheral analyses that incorporate academ-
ics and level of education at the time of at-
trition. Including this analysis might provide
independent information as to which char-
acteristics enhance cadet successfulness, as
well as provide answers to the role of general
intelligence in cadet attrition. Military experi-
ence has been shown to better prepare indi-
viduals for police academy settings (Campbell
& Campbell, 2010). However, studies must be
conducted that examine the military occupa-
tional specialty’s (MOS) role in training (i.e.,
identifying which MOS better prepares an in-
dividual for a police academy). Additionally,
it would be of interest to know which military
Back to Front Cover

40 Polygraph, 2015, 44 (2)
Park, Herndon
service (i.e. Air Force, Army, Marine Corps, or
Navy) produces the best police cadet. Isolating
these variables could assist military personnel
in their transition out of the service and into
the United States workforce. Police adminis-
trators could nd such data useful in future
recruiting endeavors. Meta-analysis studies
should be conducted that expose the true util-
ity of preemployment polygraph screening.
Research supports that age, prior mil-
itary experience, and level of education have,
to some degree, been shown as viable factors
that predict cadet and future police ofcer
performance (Aamodt, 2004; Peterson, 2002).
Research also documents that GMA better
predicts training and job performance than
any other measure (Schmidt & Hunter, 1998),
however, the current study did not support
GMA (dened by test-taking ability) in predict-
ing cadet training performance. Taking these
ideas into context, it might be postulated that
external stakeholders (society) would be better
served by an ofcer who ts a certain demo-
graphic prole (i.e. age, military experience,
level of education).
More effort is needed in the areas of
research dedicated to general intelligence and
polygraph results as they pertain to cadet per-
formance. Empirical research documents the
increasing utilization of polygraph testing in
these settings. However, research fails to cap-
ture the essence of why government agencies
place such trust in an instrument that is con-
tinually scrutinized for its controversy. Lon-
gitudinal studies that examine multiple poly-
graph testing techniques are needed to formu-
late hypotheses that either support or refute
its usefulness in screening for the best cadet.
Conducting research in these areas may tie to-
gether mechanisms that better predict cadet
attrition and training performance; thus pro-
ducing a better police ofcer.
Police agencies continue to evolve and
scholars must produce research that is fruit-
ful in the area of police personnel selection.
More research is needed in the areas of effec-
tive measures of GMA and polygraph results
as they relate to police personnel selection.
By examining current hiring proce-
dures, police agencies stand a better chance
to effectively implement strategies that com-
pliment both agency and societal goals. Ulti-
mately, this means incorporating hiring stan-
dards that are not only fair and legal, but
remain competitive in order to facilitate the
best possible outcome. From a police orga-
nizational perspective, a positive outcome is
identifying hiring procedures that effectively
capture the essence of what society demands;
a competent ofcer that can protect his or her
citizens. This, in turn, might provide police ex-
ecutives with insight as to the qualities they
desire in a future ofcer. From a public safety
standpoint, society is the benefactor by having
the best qualied ofcers protecting their com-
munities in an ever-changing world.
Back to Front Cover

41
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
References
Aamodt, M. G. (2004). Law enforcement selection: Research summaries. Washington, DC.: Police
Executive Summaries.
Aamodt, M. G., & Flink, W. (2001). Relationship between educational level and cadet performance in
a police academy. Applied HRM Research, (6), 1, 75-76.
Armstrong, D., & Polk, O. E. (2002). In W. W. Bennett, & K. M. Hess, (2004). Management and
Supervision in Law Enforcement (4th ed.). Thompson Wadsworth: Belmont, CA.
Baker, S. A. (1995). Police academy training: Are we teaching recruits what they need to know?
Policing, (21), 1, 54-79.
Bennett, W. W., & Hess, K.M. (2004). Management and Supervision in Law Enforcement (4th ed.).
Belmont, CA: Thompson Wadsworth.
Ben-Shakhar, G., & Furedy, J. J. (1990). Theories and applications in the detection of deception. A
psychophysiological and international perspective (pp. 146-169) New York: Springer-Verlag.
Bernstein, I. H., Schoenfeld, L. S., & Costello, R. M. (1982). Truncated component regression,
multicollinearity and the MMPI’s use in a police ofcer selection setting. Multivariate
Behavioral Research, (17) 1, 99-116.
Bittner, E. (1990). Aspects of Police Work. Northeastern University Press, Boston: MA.
Brand, J., & Xie, Y. (2010). Who benefits most from college? Evidence for negative se l e cti o n
in heterogeneous economic returns to higher education. American Sociological Review, (75),
2, 273-302.
Buerger, M. (2004). Educating and training the future police ofcer. FBI Law Enforcement
Bulletin, (73), 1, 26.
Campbell, D. J., & Campbell, K. M. (2010). Soldiers as police ofcers/Police ofcers as soldiers: Role
evolution and revolution in the United States. Armed Forces and Society, (36), 2, 327-350.
Capsambelis, C. (2004). Effective recruitment to attract the ideal law enforcement officer candidate.
Sheriff, (56), 4, 34-71.
Carter, D. L., Sapp, A.D. & Stephens, D. W. (1989). Police academy training: Are we teaching
recruits what they need to know? Policing, (1), 21: 54.
Chappell, T. A. (2008). Police academy training: Comparing across curricula. Policing, (31), 1, 36-56.
Clay, O. J., Edwards, J. D., Ross, L. A., Okonkwo, O., Wadley, V. G., Roth, D. L., & Ball, K. K. (2009).
Visual function and cognitive speed of processing mediate age-related decline in memory span
and uid intelligence. Journal of Aging and Health, (21), 4, 547-566.
Cowper, T. J. (2000). The myth of the “military model” of leadership in law enforcement. Police
Quartely, (3), 3, 228-246.
Back to Front Cover

42 Polygraph, 2015, 44 (2)
Park, Herndon
Decicco, D. (2000). Police ofcer candidate assessment and selection. FBI Law Enforcement Bulletin.
(69), 12, 1-6.
Fogleson, R. M. (1977). Predicting the effects of military service experience on stressful occupational
events in police ofcers. Policing: An
International Journal of Police Strategies & Management,
(25), 3, 602-618.
Forero, C. G., Pujol, D. G., Olivares, A. M., & Pueyo, A. A. (2009). A longitudinal model for predicting
performance of police ofcers using personality and behavioral data. Criminal Justice and
Behavior, (36), 6, 591-606.
Fowler, F. J. (2002). In Creswell, J. W. (2009). Research design: Qualitative, quantitative, and mixed
methods Approaches (3rd ed.). Thousand Oaks, CA: Sage Publications.
Franz, V., & Jones, D. M. (1987). Predicting the effects of military service experience on stressful
occupational events in police ofcers. Policing: An International Journal of Police Strategies
& Management, (25), 3, 602-618.
Fuss, T., & Snowden, L. (2004). Importance of background investigations. Law & Order, (52), 3, 58-
63.
Gellner, U.B., Schneider, M.R., & Veen, S. (2011). Effect of workforce age on quantitative and
qualitative organizational performance: Conceptual framework and case study evidence.
Organizational Studies, (32), 8, 1103-1121.
Gottfredson, L. S. (1996). The triumph of techniques over purpose revisited: Evaluating police ofcer
selection. Review of Public Personnel Administration, (21), 3: 219-236.
Guller, M. (2003). Predicting Performance of Law Enforcement Personnel Using the Candidate and
Ofcer Personnel Survey and Other Psychological Measures. (Doctoral dissertation), South
Orange, NJ: Seton Hall University.
Guthrie, E. (2000). Higher learning and police training. Law & Order, (48), 12, p. 124.
Hancock, B. W., & McClung, C. (1987). Psychological testing and the selection of police ofcers: A
national survey. Criminal Justice and Behavior, (30), 5, 511-537.
Harris, M., Dworkin, J., Park, J. (1990). Preemployment screening procedures: How human resource
managers perceive them. Journal of Business and Psychology, (4) 3, 279-292.
Henson, B., Reyns, B. W., Klahm, C. F. IV., & Frank, J. (2010). Do good recruits make good cops?
Problems predicting and measuring academy and street-level success. Police Quarterly, (13),
1, 5-26.
Ho, T. (2001). The interrelationships of psychological testing, psychologist’s recommendations, and
police departments’ recruitment decisions. Police Quarterly, (4), 3, 318-342.
Hughes, F., & Andre, L. (2007). Problem ofcer variables and early-warning systems. The Police
Chief, (74), 10. Alexandria, Virginia: International Association of Chiefs of Police.
Back to Front Cover

43
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
Hunter, J. E., Schmidt, F. L. (1986). The validity and utility of selection methods in personnel
psychology: Practical and theoretical implications of 85 years of research ndings.
Psychological Bulletin, (124) 2, 262-274.
Krapohl, D. (2002). In Kleiner, M. (2002). Handbook of Polygraph Testing. Academic Press: London,
UK.
Jacobs, R., Solomon, T. (1977). Strategies for enhancing the prediction of job performance from job
satisfaction. Journal of Applied Psychology, (62), 4, 417- 421.
Jefferson, T. (1987; 1993). Predicting the effects of military service experience on stressful
occupational events in police ofcers. Policing: An International Journal of Police Strategies
& Management, (25), 3, 602-618.
LaRose, A. P., Caldero, M. A., & Gutierrez, I. G. (2006). Individual values of Mexico’s new centurions:
Will police recruits implement community-based changes? Journal of Contemporary Criminal
Justice, (22), 4: 286-302.
Lewis, T.D., Shimerda, T.A. & Graham, G. (1983). What the academic advisor needs to know about
job placement. Journal of Accounting Education, (1), 2, 135-142.
Lykken, D. T. (1981). A Tremor in the Blood: Uses and abuses of the lie detector. New York: McGraw-
Hill.
Luszcz, M. A., & Bryan, J. (1999). Visual function and cognitive speed of processing mediate age-
related decline in memory span and uid intelligence. Journal of Aging and Health, (21), 4:
547-566.
Meesig, R., & Horvath, F. (1995). National survey of practices, policies and evaluative comments
on the use of preemployment polygraph screening in police agencies in the United States.
Polygraph, (24), 2, 57-136.
Ones, D., Viswesvaran, C., Schmidt, F. (1993). Comprehensive meta-analysis of integrity test validities:
Findings and implications for personnel selection and theories of job performance. Journal of
Applied Psychology, (78) 4, 679-703.
Osoian, C., Zaharie, M., & Lazar, I. (2011). Does ownership matter? Employee selection practices in
private and public sectors. Transylvanian Review of Administrative Sciences, (7), 33,
218-232.
Paoline, III Eugene A., Terrill, W. (2007). Police education, experience, and the use of force. Criminal
Justice and Behavior, (34), 2, 179-196.
Patterson, G. (2002). Predicting the effects of military service experience on stressful occupational
events in police ofcers. Policing: An International Journal of Police Strategies & Management,
(25), 3, 602-618.
Randhawa, G. (2007). Work performance and its correlates: An empirical study. Vision: The Journal
of Business Perspective. (11), 1: 47-55.
Back to Front Cover

44 Polygraph, 2015, 44 (2)
Park, Herndon
Ree, M. J., J. A. Earles., & M. Teachout, M. S. (1994). What matters most? The perceived importance
of ability and personality for hiring decisions. Cornell Hospitality Quarterly, (52), 2, 94-101.
Sanders, B. A. (2008). Using personality traits to predict police ofcer performance. Policing: An
International Journal of Police Strategies & Management, (31), 1, 129-147.
Schmidt, F.L., & Hunter, J.E. (1998). The validity and utility of selection methods in personnel
psychology: Practical and theoretical implications of 85 years of research ndings. Psychological
Bulletin, (124), 2, 262-274.
Schmidt, F. L. & Hunter, J.E. (2004). General Mental Ability in the world of work: Occupational
attainment and job performance. Journal of Personality and Social Psychology, (86), 1, 162-
173.
Schroeder, D. J. (1973). A study of the validity of the entrance examination for the position of
patrolman under the guidelines established the Equal Opportunity Employment Commission.
(Unpublished master’s thesis) New York, NY: John Jay College.
Sharit, J., Czaja, S. J., Hernandez, M., & Yang, Y. (2004). Investigating issues in aging and work
performance using a customer service task simulator. Proceedings of the Human Factors and
Ergonomics Society Annual Meeting, (48), 2, 238-242.
Society for Industrial and Organizational Psychology (2003). Principles for the Validation and use of
Personnel Selection Procedures. Retrieved from www.siop.org
Tabachnick, B. G., & Fidell, L. S. (2001). Using Multivariate Statistics (5th ed.). Boston, MA:
Pearson.
Terman, L., & Otis, A. (1917). Predictive Validity of the MMPI-2 Psy-5 Scales and facets for law
enforcement ofcer employment outcomes. Criminal Justice and Behavior, (37), 2, 217-238.
Texas Department of Public Safety (2012). Retrieved from www.txdps.state.tx.us.
Texas Commisson on Law Enforcement Standards and Education (1997). Texas Peace Ofcer
Job Task Analysis Report. Retrieved from http://www.tcleose.state.tx.us/
Texas Commisson on Law Enforcement Standards and Education (2008). Sunset Advisory Committee
Staff Report. Retrieved from http://www.tcleose.state.tx.us/
Texas Commisson on Law Enforcement Standards and Education (2010). Basic Licensing Exam
Results First Attempt Pass Rate 3 Year Summary (FY07 – FY09). Retrieved from http://www.
tcleose.state.tx.us/
Texas Commisson on Law Enforcement Standards and Education (2012). Retrieved from h t t p : / /
www.tcleose.state.tx.us/
Texas Department of Public Safety (2012). Retrieved from www.txdps.state.tx.us.
Trautman, N. E. (1986). Police academy training: Are we teaching recruits what they need to know?
Policing, (21), 1: 54.
Back to Front Cover

45
Polygraph, 2015, 44 (2)
Police Cadet Attrition and Training Performance Outcomes
Vodicka, A.T. (1994). Police academy training: Are we teaching recruits what they need to know?
Policing, (21), 1: 54.
Waddington, P.A.J. (1993). Predicting the effects of military service experience on stressful
occupational events in police ofcers. Policing: An International Journal of Police Strategies
& Management, (25), 3, 602-618.
Walker, R. B. (1994). Police academy training: Are we teaching recruits what they need to know?
Policing, (21), 1: 54.
Waugh, L. (1996). Police Recruit Selection-Predictors of Academy Performance. City, State: The
Criminal Justice Commission.
White, M. D. (2008). Identifying good cops early: Predicting recruit performance in the academy.
Police Quarterly, (11), 1, 27-49.
White, M.D. (2010). Do good recruits make good cops? Problems predicting and measuring academy
and street-level success. Police Quarterly, (13), 1, 5-26.
Back to Front Cover

46 Polygraph, 2015, 44 (2)
Nelson
Bonferroni and Šidák Corrections for Multiplicity Effects
with Subtotal Scores of Comparison Question Polygraph Tests
Raymond Nelson
Abstract
The problem of multiple statistical comparisons is discussed as it applies to the use of subtotal
scores of comparison question polygraph tests. Multiplicity phenomena, including ination of alpha
when any of a set of multiple subtotal scores are used to make deceptive classications, and dea-
tion of alpha when all of a set of multiple subtotals are used to make truthful classications of test
results. Common statistical corrections, including the Bonferroni correction and Šidák correction
are described. Mathematical examples are provided to illustrate the application of these statistical
corrections to the comparison question polygraph test.
Keywords: scoring, test data analysis, polygraph techniques, normative data, reference distributions,
statistical signicance, Bonferroni, Šidák, multiple comparisons, inated alpha, deated-alpha, mul-
tiplicity
Introduction
Multiplicity effects, also known as the
problem of multiple comparisons (McDonald,
1996; Miller, 1981), are well known to scien-
tists, researchers, statisticians and other pro-
fessionals whose work involves the evaluation
of data as a basis for classication and infer-
ence. These effects have also been referred to
as the “look elsewhere” effect (White, 2011),
because of the impulse or desire to contin-
ue to look elsewhere when we do not initial-
ly nd what we are looking for. In the context
of scientic research and testing, multiplicity
effects, and the impulse to keep looking else-
where until we nd what we are looking for,
can be thought of as a manifestation of a con-
rmation bias described by Nickerson (1998).
We ignore results when we are unsatised and
continue searching until we nd a result with
which we are satised.
A card-playing analogy can be useful to
better understand the practical implications:
image a poker player who deals himself a hand
of cards with the goal of doing so repeatedly
until he gets a Royal Flush. Probability theory
tells us that with a sufcient number of trials,
the odds will accumulate to a sufciently high
level that we are likely to eventually observe its
occurrence. But, assuming a fair and unbiased
deck of cards, it will be a mistake to attempt
to infer that the deck of cards has any spe-
cial characteristics or that the player has any
unique attributes that caused the Royal Flush
to occur. Instead, the occurrence of the Roy-
al Flush is simply a function of continuing to
look elsewhere (in subsequent hands of cards)
for its occurrence. Similarly, looking repeated-
ly at any scientic dataset can confound our
attempts to make realistic and accurate in-
ferences about signicance or meaning when
we eventually observe what we are looking for.
More specically, multiplicity effects are the
compounding of error probabilities. They can
result in a loss of accuracy or precision and
corresponding increase in classication error.
Discussion
Multiplicity effects play a role in com-
parison question polygraph examinations
when using subtotal scores to classify the re-
sults as deceptive or truthful. Subtotal scores
for individual relevant questions have been
shown to be an effective basis for deceptive
classications when the grand total score is
inconclusive (Senter, 2002; Senter & Dollins,
2003; 2008). But polygraph techniques that
make use of grand total scores have consis-
tently produced higher accuracy rates than
techniques for which decisions are based sole-
ly on subtotal scores (APA, 2011). Subtotal
scores have been the traditional basis with
which to classify the results of multiple is-
sue screening exams (Department of Defense,
2006a, 2006b) when hand scoring. The grand
total hand scores are not traditionally used in
multiple issue screening tests.
Use of polygraph subtotal scores as a
Back to Front Cover

47
Polygraph, 2015, 44 (2)
Bonferroni and Šidák Corrections for Multiplicity Effects
basis for statistical classication and infer -
ence will introduce known and predictable
mathematical and statistical increases to the
probability of error unless corrections are ap-
plied. These effects occur because every im-
perfect and non-deterministic test result is a
probabilistic result. There is always some as-
sociated probability that the result is correct
and incorrect. The test error estimate will be
an aggregation of the errors for all probabilis-
tic results used to classify the test result.
Ination of alpha for deceptive results of
event-specic diagnostic polygraphs.
Use of subtotal scores in event-specic
examinations, for which one classication will
be made at the level of the test as a whole, in-
troduces multiplicity into the statistical model.
It amounts to the practice of making multiple
statistical decisions regarding a single classi-
cation. When making multiple probabilistic
judgements regarding a single target incident
or allegation, for which any deceptive subto-
tal result will result in the classication of the
examination as deceptive, the resulting prob-
ability of error is the cumulative or additive
probabilities of error for all subtotal probabili-
ty scores. In the case of an event-specic diag-
nostic polygraph with three relevant questions
(RQs) and alpha = .05, the total error probabil-
ity can be determined by summing the alpha
levels for all RQs (.05 + .05 + .05 = .15). Cal-
culations indicate a potential for a 15% error
rate even though the test is conducted with al-
pha = .05, with the goal of constraining errors
to a rate less than 5%. This has sometimes
been referred to as the problem of inated al-
pha because of the predictable increase in test
errors. Left unmanaged, inated alpha can
result in a false positive error rate that is po-
tentially several times greater than that which
was intended or anticipated. So while the goal
was to constrain false positive errors to 5%,
the practice of using subtotals increased that
false positive error rate to about 15%.
Bonferroni correction.
Fortunately, the problem of inated
alpha is only mildly vexing and is quite eas-
ily rectied through the use of a simple sta-
tistical correction - the Bonferroni correction
(Abdi, 2007), named for famous Italian stat-
istician Carlo Emilio Bonferroni (1892-1960).
The Bonferroni correction is calculated by di-
viding the desired alpha level by the number
of statistical decisions. The number of statis-
tical decisions is equal to the number of sub-
total scores which is the same as the number
of RQs. The resulting corrected alpha level is
referred to as the Bonferroni corrected alpha.
For event-specic diagnostic polygraph
with three RQs and a desired alpha of .05 we
divide .05 by three (alpha = .05 / 3 RQs = al-
pha .0167 per RQ). It will be necessary to use
the Bonferroni corrected alpha = .0167 for
each of the three subtotal scores. When these
per question error probabilities accumulate
(.0167 + .0167 + .0167 = .05) the total cumu-
lative margin of error for the test will be alpha
= .05. The error estimate will be constrained
to within the desired range of less than 5%.
Event-specic diagnostic exams with two RQs
will require the use of Bonferroni corrected al-
pha = .025. This is because alpha = .05 / 2
RQs = .025 per RQ, and this will accumulate
to .025 + .025 = .05. Similarly, event-specic
diagnostic exams with four RQs will use Bon-
ferroni corrected alpha = .05 / 4 = .0125 per
RQ, which will accumulate to .0125 + .0125 +
.0125 + .0125 = .05. Because subtotal scores
are not used to make truthful classications
for event-specic diagnostic exams, no statis-
tical correction is needed for truthful classi-
cation for these type of exams.
Deation of alpha for truthful results of multi-
ple-issue screening exams.
Multiple-issue screening exams make
use of subtotal scores for both deceptive and
truthful classications. This is accomplished
with the “any or all” rubric which states that
any subtotal result that is deceptive will be
sufcient to classify the exam result as de-
ceptive, whereas all subtotal results must
indicate truth-telling in order to classify the
overall exam result as truthful. Results are in-
conclusive whenever one or more of the sub-
total scores are not statistically signicant for
truth-telling and none of the subtotal scores is
statistically signicant for deception. As with
event-specic diagnostic exams the test error
statistic for a multiple-issue screening poly-
graph is a function of the number of subto-
tal scores (the number of relevant questions).
This phenomena applies to all forms of testing
that involve multiple statistical comparisons.
With event-specic diagnostic exams
all relevant questions describe details relat-
ed to a single allegation or incident. Relevant
questions for multiple-issue screening poly-
graphs will describe different behavioral issues
with a strong assumption of independence.
The independence assumption is not premised
Back to Front Cover

48 Polygraph, 2015, 44 (2)
Nelson
solely on the use of different action verbs or
semantic content for each relevant question.
It also involves the assumption that an exam-
inee could engage in one or more behaviors
while conceivably remaining completely unin-
volved in other behaviors. (This independence
assumption is not used with event-specic di-
agnostic polygraph for which the all relevant
questions describe aspects of a single allega-
tion or incident.)
The independence assumption is said
to be a strong assumption because, in real-
ity, although the target behaviors might be
assumed to be independent or unaffected by
one another, even though the examinee´s re-
sponses to multiple issue polygraph stimulus
questions are not completely independent. Re-
sponses to different target stimuli can affect
one another within an exam. This is because
all responses to multi-issue screening stimuli
have an important source of shared variance -
the examinee. The fact that responses are not
completely independent appears to be the ba-
sis of the need for the “any or all” rubric and
for traditional prohibitions against attempting
to make both deceptive and truthful classica-
tions within a single examination. Because the
any or all rubric does not allow both truthful
and deceptive results, it will eliminate the po-
tential to observe both false positive and false
negative errors within a single examination.
Instead, observed testing errors will be in the
form of either false-positive or false-negative
errors, for which we can constrain their occur-
rence to desired levels.
Because the target issues for multi-
ple-issue screening exams are treated inde-
pendently, there is no great concern that we
are subjecting a single target issue to multi-
ple statistical decisions. Screening tests are
intended to identify possible problems that
can be subsequently evaluated in more thor-
ough detail, and test sensitivity is therefore
an important concern. Statistical corrections
are not used when there is a potentially costly
loss of sensitivity that would reduce the test
effectiveness (McDonald, 2009). For these
reasons, Bonferroni correction is not used to
make deceptive classications for multiple-is-
sue screening polygraphs. Deceptive classi-
cations of multiple-issue screening polygraphs
are made with the uncorrected alpha bound-
ary.
Multiplicity plays an important role
in truthful classications for multiple-issue
screening polygraphs, but in a slightly dif-
ferent way. Truthful classications are made
when the observed data differ at a statistically
signicant level from the statistical reference
distributions for deceptive cases. Alpha for
truthful classications therefore represents
the tolerance for risk or error that a decep-
tive person may be classied as truthful in a
multiple issue screening test (a false-negative
error). Quite obviously, most deceptive per-
sons can be expected to produce deceptive test
scores, and the proportion of deceptive per-
sons that produce a test question score that
is statistically signicant for truth-telling (i.e.,
differs at a statistically signicant level from
the normative reference distributions for de-
ceptive cases) is expected to be observed at the
dened alpha level (.05). Perhaps equally obvi-
ous is the fact that the proportion of deceptive
persons that produce two statistically signi-
cant truthful scores in a test with two relevant
questions will be lower than the proportion of
deceptive persons who produce only one sta-
tistically signicant truthful score. Similarly,
the proportion of deceptive persons who pro-
duce three out of three truthful scores, or four
out of four truthful scores, can be expected to
be even lower. This phenomena can be thought
of as the deation of alpha that occurs as a
result of the requirement that the examinee
pass all questions in order to pass the test.
Deation of alpha will result in a reduction of
the observed false-negative error rate to some-
thing predictably lower than the established
alpha tolerance for error.
Deation of alpha will reduce testing
errors for deceptive classications, but will
also have an effect on truthful classications.
The requirement that all subtotal scores are
statistically signicant for truth-telling will
effectively provide the truthful examinee with
multiple opportunities to not produce a statis-
tically signicant truthful score. This is a sim-
ple feature of the fact that all tests are proba-
bilistic and not deterministic, and that proba-
bilities can be cumulative under these circum-
stances. For this reason, the requirement for
statistically signicant truthful scores for all
subtotals can be expected to cause a substan-
tial ination of inconclusive results for truthful
persons, along with a corresponding substan-
tial reduction of test specicity for truth-tell-
ing - unless a statistical correction is used.
Šidák correction.
The preferred statistical correction
for truthful classications of multiple-issue
screening polygraphs is not the Bonferroni
Back to Front Cover

49
Polygraph, 2015, 44 (2)
Bonferroni and Šidák Corrections for Multiplicity Effects
correction but is instead a related procedure
called the Šidák correction (Abdi, 2007, Šidák,
1967). The Šidák correction is named for
Zbyněk Šidák (1933-1999), a renowned Czech
statistician. It is an exact version of the sim-
ple Bonferroni correction that is better suited
to the context of multiple independent classi-
cations. Calculation of the Šidák correction
is thus: 1-(1-alpha)number-of-decisions. The Šidák
correction is the mathematical compliment of
the compliment of the alpha raised to number
of decisions. As with the previously described
Bonferroni correction, the number of deci-
sions is equal to the number of subtotal scores
which is also equal to the number of relevant
questions.
The normal form of the Šidák correc-
tion is used to calculate the ination of alpha.
But we are concerned with the deation of
alpha, so it will be the inverse of the Šidák
correction that is used calculate this deation.
The inverse of the Šidák correction is calculat-
ed using the following equation: 1-(1-alpha)1/
number-of-decisions. The inverse Šidák is the mathe-
matical compliment of the compliment of the
alpha raised to the inverse of the number of
decisions.
To demonstrate the application of the
Šidák correction to adjust or correct the al-
pha boundary for the number of relevant
questions, consider the following example: a
multiple issue polygraph with 4 relevant ques-
tions for which alpha = .05 will give the fol-
lowing uncorrected, deated, alpha level: 1-(1-
.05)1/4 = .0127. That means that instead of
constraining false negatives to our desired 5%
we actually constrain them to 1.27%. This will
result in a corresponding increase in incon-
clusive truthful cases. Correcting for this will
involve rst calculating the corrected alpha
boundary using the normal form: 1-(1-.05)4
= .1854. Use of the Šidák corrected alpha =
.1854, will give the following: 1-(1-.1854)1/4 =
.05. This will preserve the test specicity to
truth-telling for multiple-issue screening poly-
graphs at acceptably high levels, while reduc-
ing the occurrence of inconclusive results for
truthful persons. It will also constrain the oc-
currence of false-negative test errors to rates
that are within the tolerance level expressed
by the alpha = .05 level.
In practice, statistical corrections can
be applied to either the alpha boundary or to
the p-values using either the normal or in-
verse forms. However, correction of p-values
can only be accomplished after conducting
and scoring an examination, whereas correc-
tion of alpha boundaries can be accomplished
prior to the conduct of an examination. This
is accomplished by using the subtotal score
with a p-value at or below the corrected alpha
as a decision threshold.
Conclusion
All scientic testing is a process of
classication and inference. Classication, in
this case, refers to the formulation of a simple
categorical test result. Inference is the process
of calculating a statistical or probabilistic es-
timate of the likelihood that an error has oc-
curred. In a more abstract sense, the purpose
of scientic testing is to evaluate and quan-
tify an amorphous phenomena that cannot
be subjected to simple and perfect determin-
istic observation or to direct physical/linear
measurement. Deterministic observation re-
quires the existence of some phenomena that
is uniquely and perfectly associated with the
thing we want to evaluate. This would be the-
oretically perfect, and would also obviate the
need for testing. Physical measurement, in
contrast, is near perfect, though still subject
to mechanical measurement error, and would
require two things: 1) a physical substance to
measure, and 2) a well-dened unit of mea-
surement. Scientic tests are inherently prob-
abilistic - they are neither deterministic nor
an actual physical measurement. Scientic
tests are not expected to be perfect. They are
expected to quantify the probabilistic mar-
gin of uncertainty surrounding a conclusion.
Good scientic tests will do this in manner
such that the predicted proportions of testing
errors concurs reasonably with the observed
evidence of testing errors. Multiplicity effects
have a potentially serious impact on the accu-
racy of test error estimates. The use of statis-
tical corrections can be an important part of
the validity and effectiveness of a test method.
Two core ideas underlie all scientic
tests and experiments. The rst core idea is
that all scientic conclusions or hypotheses
are relative to some alternative. Professionals
who make scientic conclusions are expect-
ed to articulate the alternatives and to use
probability theory to weigh the evidence. The
second core idea is that all conclusions and
hypotheses must be stated as statistical or
probabilistic hypotheses in order to be quanti-
able. Conclusions or hypotheses that cannot
be stated as statistical hypotheses cannot be
measured or tested, and are therefore not sci-
Back to Front Cover

50 Polygraph, 2015, 44 (2)
Nelson
entic. Unscientic ideas that portend to be
scientic can be said to be pseudoscience.
Related to the need for testable hy-
potheses is the need to make a priori decla-
rations about the tolerance for error and re-
quired alpha level for statistical signicance.
Field practitioners generally do not themselves
decide on alpha boundaries or numerical cut-
scores - these are most often a matter of agen-
cy policy and are developed around the needs
specic to the risk management context. Field
practitioners themselves are also not expected
to calculate statistical formulae themselves.
Instead, they commonly use published statis-
tical reference tables for which calculations
have been previously computed for all possible
test results.
If the polygraph test is merely a tool to
amplify or enhance an interrogation or inter-
view, then examiners need not ever account
for or explain the test results. If this were the
case they need not even score the test, and
certainly need not learn about probability
theory and statistical phenomena. Similarly,
polygraph examiners will never be expected to
account for or explain a test result if a confes-
sion is obtained for every deceptive test result
without fail. If the information from the pretest
and posttest discussions are the sole purpose
of the polygraph test then there would be no
need to ever provide a test result. If, however,
there is ever a need to explain a test result or
account for the level of certainty or uncertain-
ty that should be attributed to a test result,
examiners might be obligated to numerically
score and statistically quantify the test result.
Examiners who are unprepared to do this will
be vulnerable to professional embarrassment,
either due to an inability to provide evidence
based computations of the expected test preci-
sion and error rate, or due to frustration when
it is eventually discovered that polygraph re-
sults are probabilistic and imperfect despite a
feigned attitude of certainty.
Examiners who are prepared to ac-
count for test results using the basic princi-
ples and concepts of statistics and probabili-
ty and theory will be better prepared to make
favorable professional impressions while dis-
cussing test results without the sense of inse-
curity that stems from naive expectations for
deterministic perfection from a probabilistic
test. Although there will always be practical
value in in the information that can be ob-
tained from the polygraph pretest and posttest
interviews, test results without realistic com-
putations of statistical error estimates will, in
the end, be of no real value.
Ultimately, all test scores, including
both grand total and subtotal scores of com-
parison question polygraph tests, will have an
associated probability of error. Probability the-
ory informs us that error rates are predictably
cumulative whenever we attempt to make mul-
tiple statistical comparisons within a single
test or experiment. While mildly concerning,
the predictability of multiplicity phenomena
means that we can also apply the principles
of probability theory to statistically correct for
multiplicity effects - if we understand the prin-
ciples of probability. While very simple calcu-
lations such as the Bonferroni correction can
be easily managed in eld settings, eld practi-
tioners should be relieved of complex calcula-
tions such as the Šidák correction through the
inclusion of statistically corrected information
in published normative reference tables. Use
of computer algorithms can also accomplish
the application of these statistical corrections
with automated reliability. Although many re-
searchers, statisticians and scientists will pre-
fer to use omnibus statistical methods such as
ANOVA and other methods to simultaneously
test multiple statistical hypothesis without the
introduction of multiplicity effects, Bonferroni
correction and Šidák correction are two clas-
sical solutions to the well know problems of
multiplicity. They are well suited to the analy-
sis and interpretation of comparison question
polygraph test results.
Back to Front Cover

51
Polygraph, 2015, 44 (2)
Bonferroni and Šidák Corrections for Multiplicity Effects
References
Abdi, H. (2007). Bonferroni and Šidák corrections for multiple comparisons. In N.J. Salkind (Ed.), Encyclopedia of
Measurement and Statistics. Sage.
Department of Defense (2006a). Federal psychophysiological detection of deception examiner handbook. Reprinted in
Polygraph, 40 (1), 2-66.
Department of Defense (2006b). Test data analysis: DoDPI numerical evaluation scoring system. [Retrieved from http://
www.antipolygraph.org on 3-31-2007].
McDonald, J. H. 2009. Handbook of Biological Statistics 2nd ed.. Sparky House Publishing, Baltimore, Maryland.
Miller, R. G. (1981). Simultaneous Statistical Inference 2nd Ed. Springer Verlag New York.
Nickerson, R. S. (June 1998). “Conrmation Bias: A Ubiquitous Phenomenon in Many Guises”. Review of General
Psychology 2 (2): 175–220.
Šidàk, Z. (1967). Rectangular condence region for the means of multivariate normal distributions. Journal of the American
Statistical Association, 62, 626–633.
White, L. A. (August 12, 2011). Word of the Week: Look Elsewhere Effect. Sanford National Accelerator Laboratory.
Back to Front Cover

52 Polygraph, 2015, 44 (2)
Matte
Letter to the Editor Regarding article by Nelson and Handler entitled
Statistical Reference Distribution for Comparison Question Polygraphs.
James Allan Matte
Dear Editor:
This letter pertains to Appendix P, Matte Quadri-Track Zone Comparison Technique, of article
entitled Statistical Reference Distributions for Comparison Question Polygraphs by Raymond Nelson
and Mark Handler, Polygraph, Volume 44, Nr. 1, 2015.
In Footnote #9, Nelson and Handler, referring to the 2011 APA meta-analytic survey, stated
“Studies supporting this technique have been described as substantially methodologically awed, and
it is considered unlikely that the reported accuracy rates will be achieved in eld settings.” The three
eld studies validating the Quadri-Track ZCT were in eld settings (Matte, Reuss 1989b; Mangan,
Armitage, Adams 2008a; Shurany, Stein, Brand 2009), and the studies were not substantially awed
as indicated in this author’s critique (Matte 2012). In fact, the aforesaid eld studies met the most
stringent requirements set forth in the Guiding Principles and Benchmarks for the Conduct of Validity
Studies of Psychophysiological Veracity Examinations Using the Polygraph (Matte 2010), requiring a
minimum sample of 50 conrmed cases (Matte 122, Mangan 140, Shurany 57). Conversely, the
APA meta-analytic survey listed four studies that used sample cases from 20 to 30 cases validating
their respective evidentiary techniques. One of them, the Nelson, Handler, Blalock, Cushman 2012
eld study with a sample of 22 cases (Polygraph, In Press) has not been published as of 6 January
2015 (R. Nelson, personal communication 6 January 2015). Sample size has a direct relationship
to the applicability of the study’s results to the general population. As explained in detail in the
aforementioned Guiding Principles and Benchmarks, several important elements present in eld
studies are lacking in laboratory studies, which is beyond the scope of this Letter to the Editor which
APA now limits to 400 words, one table and 10 references.
In Footnote #9 Nelson, et al stated “published procedures for this technique involve the
average total score per chart instead of the more common grand total score.” This statement is
inaccurate as reected in diagram below and several published articles and studies listed in the
unabridged 2000 word Letter-to-the-Editor published on website at www.mattepolygraph.com under
heading of Publications by James Allan Matte.
Back to Front Cover

53
Polygraph, 2015, 44 (2)
Letter to the Editor Regarding article by Nelson and Handler
Back to Front Cover

54 Polygraph, 2015, 44 (2)
Matte
References
American Polygraph Association. (2011). Meta-analytic survey of criterion accuracy of validated
polygraph techniques. Polygraph, 40(4): 193-305.
Mangan, D. J., Armitage, T. E., Adams, G. C. (2008a). A eld study on the validity of the Quadri-
Track Zone Comparison Technique. Physiology & Behavior, 95, 17-23.
Matte, J. A. (2012). Critique of Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph
Techniques. European Polygraph, 6, 1(19): 19-44.
Matte, J. A., Reuss, R. M. (1989b). A Field Validation Study on the Quadri-Zone Comparison
Technique. Polygraph, 18(4): 187-202.
Nelson, R. Handler, M. (2015). Statistical Reference Distributions for Comparison Question Polygraph.
Polygraph, 44(1): 91-114.
Back to Front Cover

55
Polygraph, 2015, 44 (2)
Letter to the Editor Response to James Allan Matte
Response to James Allan Matte Letter to the Editor
Raymond Nelson
Appendix P in Nelson and Handler (2015) is calculated from the statistics published on page
98 in Matte and Reuss (1989), which recommends cut-scores of -5 and +3 per chart and includes
an instruction to average the scores for all charts. We know of no peer-reviewed publication that
recommends the use of any other distribution or cutscores. Mr. Matte´s suggested cutscores for 2, 3
and 4 charts are simply multiples of these cutscores - disregarding that standard deviations are not
subject to linear addition and multiplication. Instead of attempting to rectify Mr. Matte´s statistical
and scientic inconsistencies we elected to republish his recommendations. Our concerns about
methodological issues were included in footnotes.
Mr. Matte´s citation of a self-publication cannot be taken to re-dene how scientic research
and statistical analysis actually work. In particular, Mr. Matte´s assertions about sampling and
generalizability are wrong. Sample size does not affect the generalizability of scientic conclusions;
sampling method does. Sample size does affect statistical power – the ability to nd a signicant effect
– and this will be important when investigating small effect sizes. Polygraph research is commonly
seeking large effect sizes - large improvements over chance - for which smaller sampling sizes are
often adequate. It is not surprising that Mr. Matte´s research sample, consisting of examinations
conducted or supervised by himself (using his eponymously named technique), included no error
cases and ~100% classication accuracy. We hypothesize that reliance on confessions as a sampling
method may have systematically excluded both false-positive and false-negative error cases, for
which a confession is not likely to be obtained, from the reported study samples.
If Mr. Matte is correct in his conclusion of ~100% accuracy, then the problem of perfect lie
detection has been solved. No further research is necessary, and there is nothing more we need to
learn. If Mr. Matte is incorrect, if ~100% polygraph accuracy cannot be achieved by most examiners,
most of the time, with most examinees, then Mr. Matte´s conclusions would appear to tell us little, if
anything, about what to expect in reality.
We thank Mr. Matte for letting us know his feelings, but we remain in disagreement with his
assertions and conclusions. We apologize for subjecting the readership to another round of argument
and controversy in a matter for which there will be no benet to the profession.
Back to Front Cover

56 Polygraph, 2015, 44 (2)
Konopasek, Nelson
Sexual History Disclosure and Sex Offender Recidivism
James Edward Konopasek, Ph.D.
Colorado State University – Global Campus
Raymond Nelson, MA, NCC
Private Practice
Abstract
This research examines the extent to which non-deceptive sexual history polygraph test results
are associated with treatment outcomes and sexual recidivism. Specically, the project examined
correlations among several independent variables (i.e., sexual history polygraph results, risk
level/score, age at which a non-deceptive sexual history polygraph was achieved, achievement
of a non-deceptive sexual history polygraph result within six months of treatment onset, sexual
deviance, psychopathy, and denial) and the dependent variables of treatment completion status
and sexual recidivism. A cohort of 170 convicted sexual offenders was evaluated for a period of ve
years following completion of treatment or discharge from supervision. Analysis revealed that the
achievement of a non-deceptive sexual history polygraph result was moderately associated with
completion of treatment (rφ = .328, p < .001). Two variables, achievement of a non-deceptive sexual
history polygraph result within six months of treatment onset (rφ = -.152, p = .047), and age under
35 at the time of a non-deceptive sexual history polygraph (rφ= .167, p = .029) were shown to be
correlated with sexual recidivism. This research provides preliminary evidence that non-deceptive
sexual history polygraph results are associated with favorable treatment and recidivism outcomes.
Keywords: sex offender, rehabilitation, recidivism, Static-99R, polygraph, disclosure
Sexual History Disclosure and Sex
Offender Recidivism
Sexual history polygraph testing,
is an adjunct component of approximately
two thirds (67%) of the adult outpatient sex
offender treatment programs in the U.S.
(McGrath, Cumming, Burchard, Zeoli &
Ellerby, 2010). Sexual history polygraph
examinations are conducted in an effort to
motivate full, accurate, and timely disclosure
of sexually deviant behavior (Ahlmeyer, Heil,
McKee & English, 2000; Association for the
Treatment of Sexual Abusers, 2004; Emerick
& Dutton, 1993; O’Connell, 1998). Polygraph
testing is intended to encourage offenders to
fully disclose their history of sexual offending
behaviors. Treatment providers who make
use of sexual history polygraph testing do so
with the goal of the identication of paraphilic
and crossover offense behaviors so they can
formulate accurate and effective treatment
planning to facilitate the cessation of continued
offenses (Ahlmeyer, Heil, McKee & English,
2000; Bourke & Hernandez, 2009; O’Connell,
1998; Wilcox & Sosnowski, 2005). For more
information on the use of polygraph testing in
sex offender management see (English, 1998;
English, Jones, Pasini-Hill, Patrick & Cooley-
Towell, 2000; Grubin, 2008; Hindman & Peters,
2001; Levenson, 2009). Skeptics of polygraph
testing have pointed to unanswered questions
regarding the contribution of polygraph testing
to observable and measurable outcomes (Ben-
Shakhar, 2008; National Research Council,
2003; Rosky, 2013). Despite the existence of
controversy, polygraph testing has become
a recognizable component of sex offender
supervision and treatment programs.
Author's Note
Correspondence Address: PO Box 178, Felton, PA 17322
Email: james.konopasek@csuglobal.edu
Back to Front Cover

57
Polygraph, 2015, 44 (2)
Sexual History Disclosure and Recidivism
The Importance of Timely, Honest,
Disclosure
Farber (2003) and Farber and Hall
(2003) discussed the importance of disclosure
in psychotherapy and how this process
includes inherent challenges. Obstacles to
timely, honest, disclosure of sexual problems
include feelings of shame, guilt and fear among
clients--which may contribute to deliberate
secret-keeping, things left unsaid in therapy
sessions, and what is termed the untold story
relating to clinically relevant history. With
general psychotherapy patients, the following
have been described as the most prevalent
items not disclosed: sexual and body-oriented
experiences, sexual feelings and fantasies
toward the therapist, interest in pornography,
bathroom habits, experiences and feelings
toward masturbation, loss of virginity and
delity (Farber & Hall, 2002). Nearly all of
these topics will also apply to the treatment of
sex offenders.
Recent research in sex offender
treatment has shown that measurement of
the quantity and seriousness of clinically
relevant disclosures (CRDs), obtained through
polygraph testing, may be useful to identify
and adjust treatment targets (Gannon,
Wood, Pina, Tyler, Baroux & Vasquez, 2014).
Gannon, et al., quantied CRDs in four
categories: thoughts, feelings and attitudes
(e.g., abusive fantasies and desires); sexual
behavior (e.g., use of pornography); historical
information (e.g., admitting unknown offense
behavior); and changes in circumstance/risk-
behaviors (e.g., increased access to children).
They reported that sex offenders who were
subject to polygraph testing made CRDs in
572 sessions versus 320 CRDs for controls not
subject to polygraph testing, but also found
that the seriousness ratings of disclosures did
not differ across the two groups. Gannon et al.
cautiously described the evidence showing that
user satisfaction benets were expressed by
treatment and supervision professionals who
made use of maintenance/compliance testing
with sex offenders in the United Kingdom.
Although less is known, at the present
time, about the benets of the sexual history
polygraph, there is general and emerging
evidence that timely self-disclosure of problem
behaviors is viewed by professionals as a
favorable indicator that may be a factor
in effective risk assessment, treatment
planning and supervision in the community.
How disclosure is measured varies, and the
relevance of disclosure to treatment and
recidivism outcomes remains elusive. This
research aims to ll some of this gap in
knowledge.
Accuracy and Validity of Sexual History
Polygraph Testing
Polygraph examinations have been
used to measure the veracity of sex offender
disclosure for over 30 years (Abrams, 1991),
beginning in the early 1980s. The accuracy
of such testing has been questioned.
Despite continued controversy, there
is a substantial and growing literature
supporting the polygraph as capable of
discriminating deception and truth-telling
at rates signicantly greater than chance;
and publishing the validity and reliability of
various polygraph testing techniques, with
both favorable and unfavorable ndings
(Furedy, 1996; Honts, 1996; Honts & Alloway,
2007; Iacano, 2008; Iacano & Lykken, 1997;
Krapohl, 2006; National Research Council,
2003; Offe & Offe, 2007; Nelson, Handler,
Krapohl, Gougler, Shaw & Bierman, 2011).
Research and publication has also addressed
threats to polygraph accuracy (Honts, Hodes &
Raskin, 1985; Honts, Raskin & Kircher, 1994;
National Research Council, 2003; Patrick &
Iacano, 1989).
Most relevant to this study is research
on the accuracy of two types of Comparison
Question Technique (CQT) exams (i.e., the Air
Force Modied General Question Technique –
AFMGQT, and the Directed Lie Screening Test -
DLST). The AFMGQT and DLST are considered
by polygraph examiners to be equally well
suited for multiple issue screening polygraphs
in the post-conviction and other settings. Both
techniques involve the same instrumentation
and sensors. Differences between the two
formats primarily involve the procedural rules
for recording of several presentations of the
test stimulus questions.
The scientic basis of the CQT is
that difference in the pattern of response to
relevant and comparison test stimuli can be
observed as a function of deception or truth-
telling in response to the relevant stimuli.
Used in the context of a sexual history
polygraph, the premise of the CQT is that
an examinee who withholds information
about his or her sexual history will produce
physiological reactions that are loaded onto
sexual history questions, whereas an examinee
who is truthful will produce responses that
are loaded on comparison stimuli. Senter,
Weatherman, Krapohl, and Horvath (2010)
referred to this phenomenon as differential
salience. Polygraph results can be analyzed for
Back to Front Cover

58 Polygraph, 2015, 44 (2)
Konopasek, Nelson
their statistical signicance, and can also be
described categorically. Polygraph examiners
have adopted the categorical terms Signicant
Reaction and No Signicant Reaction, though
some results are described using the more
traditional categorical terms Deception
Indicated, or No Deception Indicated that
remain in more common use in diagnostic
polygraph contexts. These categorical labels
are the contextual analog for the more
abstracted terms Positive and Negative as
used in other scientic testing context.
During recent years, there has been
an increase in published polygraph research
on the validity of the wide variety of polygraph
testing techniques (Handler, Nelson & Blalock,
2008; Krapohl, 2006; Nelson, et al., 2011).
Nelson, et al., (2011) summarized the results
of 14 studies involving 1,008 cases and 31
different scores, and described the accuracy
of polygraph techniques scored with an
assumption of independent criterion variance,
such as those used in post-conviction sex
offender testing, as proving a mean unweighted
accuracy rate of .850, with a 95% condence
range from .773 to .926.
Polygraph testing, applied to the
sexual history disclosure polygraph, will
address whether an offender has truthfully
disclosed the details of his or her history of
sexual offense behaviors. Because there is no
known allegation or incident that is the target
of the exam or investigation, that exams are
screening tests, also referred to as exploratory
exams, and investigative techniques (Handler,
Nelson & Blalock, 2008), and are not intended
to become the sole or central basis for decision
and action in the same way manner as a
diagnostic testing context.
Relevant questions for sexual history
screening polygraphs can address several
types of sex offending behavior, including:
sexual offenses against underage children
since becoming an adult, sexual contact
with relatives and family members, forced
sexual contact/violent sexual offenses, and
sexual contact with persons who were asleep
or unconscious. Relevant questions can also
address sexual behaviors that can signal
problems involving sexual compulsivity or
sexual preoccupation, such as voyeurism,
exhibitionism, public masturbation, stalking
behaviors, theft or use of underwear/
undergarments or personal property for
masturbation or sexual arousal, child
pornography, and other sexual behaviors that
may indicate problems with sexual deviancy.
Sexual History Polygraph Testing in the
Sex Offender Management Context
Several studies describe the increase
in disclosure of information (Abrams, 1991;
Ahlmeyer, et al., 2000; English et al., 2000;
Grubin, Madsen, Parsons, Sosnowski &
Warberg, 2004; Gannon, et al., 2014; Heil et
al., 2003; Kokish, Levenson & Blasingame,
2005; O’Connell, 1998; Wilcox, Sosnowski &
Middleton, 1999) that results from polygraph
testing, and that may contribute to improved
risk assessment, treatment planning, and case
management. Fewer studies have addressed
the relationship between polygraph testing
and recidivism. McGrath, Cumming, Hoke
and Bonn-Miller (2007) and Cook (2011) are
the only studies we located that address this
relationship.
McGrath, et al. (2007), using a
comparative matched-pairs design, studied
recidivism incidence among treated offenders
who were subject to maintenance polygraph
testing regarding compliance with the
supervision and treatment program. One
hundred four (104) polygraph cases were
matched (relative to completion status, risk
level and offense severity) with 104 treated
sex offenders who did not undergo polygraph
testing. Although no relationship to sexual
recidivism was found, the polygraph group
showed lower violent recidivism rates than the
non-polygraph group, 2.9% versus 11.5%.
Cook (2011) reported results using logistic
regression and a modied version of the
Static-99 risk assessment instrument that
included additional variables gleaned from
sexual history polygraph examination (SHPE)
reports. Those results showed that data
extracted from SHPE reports on 93 convicted
sex offenders, including early onset of sex
offending prior to age 13, more than 2 admitted
paraphilias, and passing/failing the SHPE did
not account for any statistically signicant
change in the odds of violent, including sexual,
recidivism. The present study is similar to
McGrath, et al. (2007) and Cook (2011) in that
it is aimed at increasing the body of knowledge
on polygraph-assisted interventions related to
treatment outcome and sex offense recidivism.
Purpose and Method
The purpose of this study is to explore
correlations among variables related to test
results from sexual history polygraph testing,
treatment outcome, and sexual recidivism
among convicted sex offenders. This study
also examines relationships between study
Back to Front Cover

59
Polygraph, 2015, 44 (2)
Sexual History Disclosure and Recidivism
variables and several other literature-
derived variables that have been shown to
be correlated with sexual recidivism. Those
variables include: age, sex offender type,
Static-99R risk score/level, denial, sexual
deviance, and psychopathy. Relationships
among the study variables were analyzed using
bivariate analyses utilizing the Phi Coefcient
for dichotomous variables and Point Biserial
Correlation.
Assumptions
A key assumption in this project is that
the CQT polygraph screening examinations
conducted on individuals in this sample are
similar enough in design and administration,
and criterion accuracy rates are similar to those
described by Nelson et al. (2011), specically
the AFMGQT. Polygraph examination reports
utilized in this study indicated that CQTs were
administered in all cases and that relevant
sexual history questions were constructed
around the target issues described previously
in this report. However, the exact nature of
comparison questions, scoring rules, and
decision cut-off scores are unknown.
Procedures
Data were extracted from hard-
copy les stored in program archives and
from the computerized database (Microsoft
Access) of an outpatient treatment program.
Consultation during the design phase of this
project (O’Connell, 2011) resulted in the use of
independent third-party research assistants
to ensure that the principal author remained
blind as to the identity of recidivist identity
in order to alleviate the potential for bias
or conict of interest. A research assistant
constructed three research data sets: 1)
a small 8-item identied data set for the
purposes of requesting and capturing criminal
history recidivism data from criminal history
gatekeeper agencies, 2) a 64-item identied
data set containing all variable data, and 3)
a 52-item unidentied data set. Only the
unidentied data set was shared with this
researcher. The unidentied data set was
exported to a SPSS database le for analysis.
Data was captured for the dependent
variable, recidivism, by a research assistant
who conducted public record criminal history
searches in accordance with the gatekeeper
agency’s condentiality, terms of use,
storage, retention and destruction guidelines
(LexisNexis Accurint, 2011; Oregon Judicial
Department--Justice Information Network--
OJIN, 2009; Washington State Patrol--WATCH,
2011; Washington State Institute for Public
Policy, 2009).
Test results were extracted from
archived polygraph examination reports and
pretest questionnaires, along with information
on whether a non-deceptive SHPE result was
achieved within six months of treatment
onset. These data were entered into the
larger, 64-item, database. Criminal history
recidivism data was then matched for each
case by the research assistant. Archived paper
records included psychosexual evaluations,
presentence reports, polygraph examination
reports and treatment progress reports. Static-
99R risk computation forms were completed
by the principal author, who received training
from the Justice Institute of British Columbia
(JIBC, 2011) on the scoring of the Static-99R
in accordance with published coding rules
(Harris, Phenix, Hanson & Thornton, 2003;
Helmus, Babchishin, Hanson, & Thornton,
2009) during 2011. These data were entered
into the identied database, after which
the research assistant entered matching
recidivism data into the dataset. Due to time,
personnel and training budget constraints of
this self-funded project, third party Static-
99R scorers were not used, and inter-rater
reliability statistics were not obtained.
Sample
The study involved a convenience
sample of adult male sex offenders (N = 170)
mandated by the courts and parole/probation
agencies to be evaluated and treated at four
outpatient sex offender treatment programs
between 1994 and year-end 2004. The sample
is a small fraction of sex offenders supervised,
treated and polygraph tested in the Oregon and
Washington correctional treatment systems
over the same time period, and so these results
may not be generalizable. The sample cases
were delimited by the availability of data for
every adult male that was evaluated, treated,
and polygraph tested at program ofces
during those years, as a result of the principal
author´s role as the former outpatient program
director.
Descriptive Statistics
The sample cases consisted of 170
men, the majority of whom were Caucasian
(97.6%). The mean age of offenders, at the time
Back to Front Cover

60 Polygraph, 2015, 44 (2)
Konopasek, Nelson
Table 1 Frequency of types of sexual offenses
Sex Offender Type
Child Molester Rapist Non-contact Other Total
n = 110 n = 33 n = 20 n = 7 N = 170
Table 2 Frequency distribution of Static-99R risk scores and risk levels
Static-99R Score Frequency Percent Risk Level
-3 n = 3 1.8 Low
-2 n = 3 1.8 Low
-1 n = 7 4.1 Low
0 n = 14 8.2 Low
1 n = 24 14.1 Low
2 n = 36 21.2 Moderate-
Low
3 n = 35 20.6 Moderate-
Low
4 n = 26 15.3 Moderate-
High
5 n = 12 7.1 Moderate-
High
6 n = 6 3.5 High
7 n = 2 1.2 High
8 n = 1 0.6 High
9 n = 1 0.6 High
Total 100.0%
The sample cases consisted of low
risk offenders (n = 51, 30.0%); moderate-low
risk offenders (n = 72, 42.4%); moderate-high
risk offenders (n = 37, 21.8%) and high risk
offenders (n = 10, 5.9%) as classied by the
Static-99R (Hanson and Thornton, 2000;
Hanson, 2005). Table 2 shows the frequencies,
percentages and Static-99R risk levels.
The mean Static-99R score was 2.41,
the median 2.0, standard deviation 2.057,
with a range of -3 to +9. Not surprisingly, the
majority of offenders were classied in the
moderate risk categories. Relatively few of the
sample cases were in the low and high risk
categories.
Other independent control variables
were also analyzed, including the presence
of substantial denial, sexual deviance and
psychopathy. Denial of the instant offense
details and/or problematic sexual behavior at
intake was observed and documented by the
evaluator or therapist in 57.1% (n = 97)
of polygraph testing was 36.6 years, the
median age is 34.5 years, standard deviation
of 14.56 years and range of 18 to 80 years.
Table 1 shows the types of sexual offenses.
Back to Front Cover

61
Polygraph, 2015, 44 (2)
Sexual History Disclosure and Recidivism
of the cases. Sexual deviancy, as indicated by
self-reported paraphilias during presentence
investigation or psycho-sexual evaluation
reports, or through penile plethysmograph
testing, was a factor for 48.8% (n = 83) of the
individual offenders. Antisocial personality
disorder or psychopathy was diagnosed by a
psychologist or other licensed mental health
professional and described in the presentence
or psycho-sexual evaluation reports for 12.9%
(n = 22) of the offenders.
Sample variables related to the this
study were also analyzed, including treatment
outcome, disclosure of sexual offense history,
and whether a non-deceptive sexual history
polygraph was completed within six months of
treatment onset. Non-deceptive sexual history
polygraph results were achieved by 67.1%, (n
= 114) of the 170 sample cases. For 47.1% (n
= 80) cases the non-deceptive sexual history
polygraph was completed within six months of
treatment onset. Case records indicated that
32.9% (n = 56) did not achieve a non-deceptive
sexual history polygraph examination result.
Offenders who completed treatment comprised
41.8% (n = 71) of the sample. A small portion of
the sample 2.9% (n = 5) received no treatment
due to the recommendations contained in the
presentence and psychosexual evaluation
reports. Some of the offenders did not
complete treatment, including those who were
terminated for non-compliance 17.1% (n = 29),
dropouts 5.3% (n = 9) and discharges due to
jurisdiction or funding changes 26.5% (n =
45). Some cases 5.3% (n = 9) were transferred
to other treatment programs, and 1.2% (n =
2) deceased. Some cases were counted in
multiple categories.
Findings
Table 3 shows the recidivism offenses
committed by 39 recidivists, including
failure to register/report, sexual abuse, child
molestation, and forcible rape. Eleven (6.5% of
the total sample) recidivists perpetrated new
sex crimes within 5 years of discharge from
treatment program discharge. Twenty eight
(22.9% of the total sample) recidivists failed
to register or report, a status sexual offenses.
Within the group of status reoffenders, 50% (n
= 14) also committed new non-sexual crimes,
including non-sexual assault (n = 6), felony
possession of controlled substance (n = 3), and
theft of varying degrees (n = 5).
Table 3 Frequency of recidivism during a 5-year follow-up period
Sexual Recidivism Offense Frequency
Child Molestation offense n = 1
Dealing Child Pornography n = 1
Encouraging Child Sexual Abuse n = 1
Failure to Register/Report (FTR) n = 28
Incest offense - Sexual Exploitation of Minor n = 1
Rape n = 2
Child Sexual Abuse/Assault n = 5
_______________________________________________________________________________
Total 5-Year Recidivists (including FTR) n = 39 (22.9% of N = 170)
5-Year Sexual Recidivists (excluding FTR) n = 11 (6.5% of N = 170)
_______________________________________________________________________________
Results of bivariate analyses are shown
in Table 5. That analysis showed statistically
signicant relationships between sexual
recidivism and the variables of non-deceptive
sexual history polygraph result (rφ = .167, p =
.029) and age under 35 at time of non-deceptive
sexual history polygraph result (rφ= -.152, p =
.047). That analysis also showed a statistically
signicant relationship between successful
completion of sex offense specic treatment
and a non-deceptive sexual history polygraph
result at any time during treatment (rφ= .328, p
< .001). There was also a statistically signicant
negative correlation between sexual deviancy
Back to Front Cover

62 Polygraph, 2015, 44 (2)
Konopasek, Nelson
Table 4 Correlation of study variables to treatment completion and sexual recidivism
________________________________________________________________________
Variable Treatment Completion Sexual Recidivism
Value / Sig. Level Value / Sig level
Age under 35 at non-deceptive SHPE -.095 / .217 .167* / .029
Denial -.057 / .461 .035 / .649
Non-deceptive SHPE within 6 months .077 / .317 -.152* / .047
Non-deceptive SHPE .328* / .000 -.070 / .361
Psychopathy .041 / .592 -.101 / .186
Sexual Recidivism .004 / .959 --------------
Sexual Deviance -.168* / .028 .126 / .101
Static 99R Risk Score -.129 / .095 .134 / .082
Treatment Completion ----------- .004 / .959
Total Cases N = 170
*Signicant at p < .05
Non-deceptive SHPE results were
correlated with successful completion of
treatment (rφ= .382, p < .001), but the
correlation with sexual recidivism was
not statistically signicant. There was a
statistically signicant negative relationship
between a non-deceptive SHPE within six
months of treatment onset and sexual
recidivism (rφ= -.152, p = .047) at the .05 level.
Interestingly, age under 35 at time of non-
deceptive SHPE was also signicant for sexual
recidivism at the .05 level (rφ = .167, p = .029).
Sexual deviancy was negatively correlated
with completion of treatment (rφ = -.168, p =
.028). No adjustment or correction was made
to theses reported p-values.
Table 5 shows the case frequencies
for the study variables, and indicates that
nine of 11 sexual recidivist cases (81.8%) did
not complete a non-deceptive sexual history
polygraph within six months of treatment
onset, and this relationship was statistically
signicant (rφ = -.152, p = .047). Results also
show that 81% of sexual recidivists were
under age 35 at the of the SHPE, and that the
relationship between age under 35 at the time
of a non-deceptive sexual history polygraph
examination and sexual recidivism was
statistically signicant (rφ = .167, p = .029).
Cases from the two moderate risk groups, as
determined by Static-99R risk scores, were
most represented in the recidivist category
(63.6%), though the relationship risk and
recidivism was not statistically signicant (rpb
= .134, p = .082) for this sample.
and treatment completion (r φ= -.168, p =
.028). Other study variables, including denial,
were not statistically signicantly correlated
with treatment completion or recidivism,
as indicated in Table 4. Only the Static-99R
variable was approaching a signicant level (rφ
= .134, p = .082) relative to sexual recidivism.
Back to Front Cover

63
Polygraph, 2015, 44 (2)
Sexual History Disclosure and Recidivism
Table 5 Frequency of sexual recidivism by study variables
Non-deceptive SHPE within 6 months of treatment onset
No Yes Total
Sexual Recidivism No 81 78 159
Yes 9 2 11
Age at time of non-deceptive SHPE
<35 >35
Sexual Recidivism No 83 76 159
Yes 9 2 11
Static 99-R Risk Level
Low Med Low Med High High Total
Sexual Recidivism No 49 68 34 8 159
Yes 2 4 3 2 11
Table 6 shows the crosstab frequencies
for each independent variable and whether a
non-deceptive SHPE was completed within six
months of treatment onset. Forty-eight (60%)
of the 80 offenders who completed a non-
deceptive SHPE within six months of treatment
onset were under age 35, which is statistically
signicant (rφ = 0.189, p = .014). Denial was
a factor for 69 of 90 cases for which a non-
deceptive SHPE was not completed within six
months of treatment onset, while denial was a
factor for only 35 of 80 (35%) of the offenders
who completed a non-deceptive SHPE within
six months of treatment onset (rφ = - .420, p
= .000). No statistical correction was used for
these reported p-values. Correlations were not
statistically signicant for the relationship
between non-deceptive SHPE results
and ve other study variables, including
recommendations for continued treatment at
the time of discharge, sexual deviancy, anti-
social personality or psychopathy, and Static-
99R risk level.
Back to Front Cover

64 Polygraph, 2015, 44 (2)
Konopasek, Nelson
Table 6 Distribution for non-deceptive sexual history polygraph within six months of treatment
onset
Age at time of non-deceptive SHPE
<35 >35 Total
Non-deceptive SHPE within 6
months of treatment onset
No 37 53 90
Yes 48* 32 80
Treatment Status at Discharge – More Tx Needed
Yes No Total
Non-deceptive SHPE within 6
months of treatment onset
No 50 40 90
Yes 41 39 80
Presence of Sexual Deviance
Yes No Total
Non-deceptive SHPE within 6
months of treatment onset
No 42 48 90
Yes 41 39 80
Presence of Psychopathy – APD
Yes No Total
Non-deceptive SHPE within 6
months of treatment onset
No 14 76 90
Yes 872 80
Presence of Denial at Intake
Yes No Total
Non-deceptive SHPE within 6
months of treatment onset
No 69 21 90
Yes 28* 52 80
Static 99-R Risk Level
Low M-Low M-High High Total
Non-deceptive SHPE within 6
months of treatment onset
No 31 32 22 5 90
Yes 20 40 15 5 80
*Signicant at p < .05
Back to Front Cover

65
Polygraph, 2015, 44 (2)
Sexual History Disclosure and Recidivism
Discussion
This project was a correlation
study, involving N = 170 adult males who
were convicted of sexual offenses, designed to
investigate the relationships between treatment
and recidivism outcomes and other variables,
including the completion of a non-deceptive
sexual history polygraph examination, under
age 35 at time of non-deceptive sexual
history polygraph examination, anti-social
personality disorder or psychopathy, sexual
deviancy, and denial of offense details at the
time of treatment intake. Bivariate analysis
suggests that some study variables are worthy
of further interest and continued investigation
of their value towards the prediction of sexual
recidivism and treatment completion. Two
variables were signicantly correlated with
treatment completion, non-deceptive sexual
history polygraph examination results (rφ=
.328, p < .001) and sexual deviancy (rφ =
-.168, p = .028), though the relationships
of these and the sexual recidivism outcome
were not statistically signicant. Other sexual
history polygraph variables were signicantly
correlated with sexual recidivism: completion
of non-deceptive polygraph examination
results within six months of treatment onset (rφ
= -.152, p = .049) and age under 35 at the time
of a non-deceptive sexual history polygraph
exam (rφ = .167, p = .029). Although non-
deceptive sexual history polygraph result was
associated with lower sexual recidivism, non-
deceptive sexual history polygraph results by
offenders under age 35 were associated with
higher sexual recidivism.
In a very preliminary way, these results
indicate a statistical relationship between
sexual history polygraph examination results
and sexual recidivism outcomes. These results
suggest the possibility of an interaction
between age and motivation for disclosure
and the outcome of sexual recidivism might
be an interesting area for further study.
However, data were not analyzed for younger
age adults who completed a non-deceptive
sexual history polygraph within six months
of treatment onset. Although the non-random
sample is somewhat small and the stability or
generalizability of these results is presently
unknown, this bivariate analysis suggests
the possibility that treatment outcomes and
5-year sexual recidivism rates may be partially
informed by timeliness of non-deceptive
sexual history polygraph test results. Further
study is needed to better understand the value
or meaning of non-deceptive sexual history
polygraph results when evaluating motivation
and treatment progress.
These results must be interpreted
cautiously because sexual history polygraph
is a complex process with a number of
dimensions. Non-deceptive sexual history
polygraph results indicate that all an offender
produced statistically signicant truthful
numerical scores for all investigation target
questions. Test results are neither deterministic
nor a direct physical measurement of the
amorphous social construct of deception.
Instead, test results are a probabilistic
computation of the margin of error or level of
condence that can reasonably assigned to a
categorical conclusion of deception or truth-
telling when comparing the numerical scores
of validated physiological discriminators with
statistical reference distributions for deceptive
and truthful persons.
Target questions for sexual history
polygraph exams are selected for their
actuarial, operational, or clinical relevance
to risk assessment, risk management and
treatment goals. Target issues for sexual
history polygraph testing are presently
unstandardized, though the basic requirements
for polygraph questions are that relevant target
questions describe a behavioral issue that can
be answered either yes or no, and for which
the examinee will know whether the answer is
truthful or deceptive. Relevant questions will
often include this following: sexual offenses
against underage children since becoming an
adult, sexual contact with relatives and family
members, forced sexual contact/violent sexual
offenses, and sexual contact with persons
who were asleep or unconscious. Other target
questions can be used to investigate sexual
behaviors that can signal problems involving
sexual compulsivity or sexual preoccupation,
including: voyeurism, exhibitionism, public
masturbation, stalking behaviors, theft or
use of underwear/undergarments or personal
property for masturbation or sexual arousal,
child pornography, and any other sexual
behaviors that may indicate problems with
sexual deviancy.
Achievement of a non-deceptive sexual
history polygraph result necessitates that
the offender rst admit the allegation of the
instant offense, as there are potential error
hazards associated with attempting to screen
for unknown sexual assault behaviors while
denying a alleged sexual offense. Without
strong evidence that denial of the instant
offense is factually truthful, any attempt to
Back to Front Cover

66 Polygraph, 2015, 44 (2)
Konopasek, Nelson
screen for unreported sexual offenses might
amount to a form of collusion with the offender.
Preparation for sexual history polygraph
testing will involve reviewing conceptual
vocabulary terms that describe sexual abuse
behaviors, along with operational denitions
that dene and describe and those behaviors.
Preparation will also involve a personal review
of one’s history of sexual behavior with the
goal of identifying behaviors that were abusive
or unlawful and those that were within normal
limits. Preparation of and review for sexual
history testing is a clinical process for which
the details will be inform and be informed by
the larger clinical treatment picture, including
corresponding mental health and personal
trauma issues, in addition to the nature and
extent of the individual’s history of sexual
behaviors.
Effective preparation for sexual
history polygraph testing may be partially
a function of the quality of professional and
therapeutic rapport between the offender and
the treatment and supervision professionals.
This preparation may also be a function of the
treatment cohort group, family system, and
social support network. Ultimately, effective
preparation will contribute to structured and
organized review of the sexual history during
the sexual history polygraph pretest interview.
An important aspect of the sexual
history polygraph is that offenders are not
compelled to disclose identifying information
regarding their sexual assaults. Instead they
are permitted to withhold information about
jurisdiction, exact name or the exact nature of
the relationship. This is viewed as unfortunate
by some, but is neither intended to devalue
the impact of sexual on the personhood of
sexual abuse victims, nor to endorse the
rights of the offender as more important, but
is necessary to facilitate the disclosure of more
complete information without compelling such
disclosure that it results in legal vulnerability
related individual rights.
Sexual history polygraph testing is a
multidimensional process that may be affected
by a number of factors. Sexual history polygraph
testing, as a component of sex offender
treatment and management, especially in
cognitive behavioral programs that encourage
the reduction of cognitive distortions and the
increase of personal responsibility, may be
an indicator of motivation for learning and
change. Polygraph testing may be used by
some programs as tool for assessing readiness
for activities and privileges such as social/
public events, safety and activity plans, and
family reunication. Finally, the achievement
of non-deceptive sexual history polygraph
results may also be a function of program
expectations, for which both overt and subtle
message can either reinforce or undermine
attitudes and perceptions about the meaning
and value of the disclosure process.
In this small study, age under 35 at
the time of a non-deceptive sexual history
polygraph was found to be correlated with the
outcome of sexual recidivism at a minimal
statistically signicant level. As indicated in
Table 6, 60% of the offenders who completed a
non-deceptive sexual history polygraph within
six months of treatment onset were under age
35, 35% of those were in denial at intake, and
75% were in the two lowest Static-99R risk
categories. These results suggest the potential
that some difference may exist for the under
age 35 group, which also appears to be over-
represented in this sample. Unfortunately, for
this study, we did not capture information
about the details of the offenders’ reported
histories of sexual offenses.
Limitations
The most signicant limitations to this
study relate to the type of sample, sample size
and project design. This study was conducted
on a convenience sample of 170 adult male
offenders referred to outpatient treatment
programs in the Pacic Northwest region of
the United States between 1994 and 2004-
-and who were administered at least one
sexual history polygraph examination. This
study was designed to be an investigatory,
correlational survey aimed at evaluating
whether the variable of non-deceptive sexual
history polygraph results are associated with
sexual recidivism. This research was not an
experimental study and so it did not include
a non-polygraphed control group. Causal
inferences are not possible based on these
results, and the generalizability of these
ndings is presently unknown.
Interaction effects were not evaluated
in this correlation study, as might have been
accomplished with data coded in a manner
that would support a multivariate analysis.
Multivariate analysis would have also relieved
concerns about the effects of multiplicity in
a survey study of signicant relationships
between numbers of variables. Another,
previously mentioned, limitation to the design
of this project has to do with the absence of
Back to Front Cover

67
Polygraph, 2015, 44 (2)
Sexual History Disclosure and Recidivism
data on the nature and scope of the reported
sexual offense behaviors, both prior to and
during the polygraph testing process.
Recommendations
Future studies should include this
type of data into the design and analysis,
perhaps using level of detail similar to that
outlined in Pratley and Goodman-Delahunty
(2011), which includes incidents of abuse,
duration of offending (in days), frequency
of offending, number of locations, range of
abusive acts committed, and intrusiveness of
abuse. Caution should always be exercised
in the area of professional expectations for
disclosure of information, always recognizing
the impossibility of the notion of full disclosure
or the expectation that professionals can
somehow know everything or every detail
of an offender’s history of sexually abusive
behavior. Instead, it remains within the realm
of realistic possibility that non-deceptive
sexual history polygraph results signify only
the probability that an offender has reported
the major behavioral detail as described by the
test stimuli, and that there may always remain
additional under-reported detail regarding
behavioral and interactional aspects of sexual
abuse behaviors.
For future research, the evaluation of
a larger and more randomly selected sample
will be important. Comparison of recidivism
outcomes for a polygraph cohort and non-
polygraph cohort may also be informative of
the value of polygraph testing. Future studies
should attempt to capture more ne-grained
data and information about the length of
time necessary to achieve a non-deceptive
sexual history polygraph test result, and
the contribution of CRDs as a recidivism
predication variable in regression or other
form of prediction model. Other research
questions might include qualitative aspects of
how supervision and treatment professionals
know, or measure, the extent to which their
client is being honest in CRDs, and what
additional information can be accessed by
supervision and treatment professionals to
more effectively predict treatment and sexual
recidivism outcomes.
Replication of this study is
recommended, including the comparison of
these results with polygraph cohorts and
non-polygraph cohorts for whom reported
sexual history is veried in some other
manner. Future studies should attempt to
further investigate the possibility that causal
relationships may exist between motivation for
disclosure and outcomes for both treatment
and sexual recidivism. It remains possible that
the observed results are an anomaly resulting
from programatic or sampling factors.
At present, the potential value for sexual
history polygraph testing appears worthy of
a recommendation for continued interest and
continued study.
Back to Front Cover

68 Polygraph, 2015, 44 (2)
Konopasek, Nelson
References
Abrams, S. (1991). The use of polygraphy with sex offenders. Sexual Abuse: A Journal of Research
and Treatment, 4(4), 239-263. doi:10.1177/107906329100400401
Ahlmeyer, S., Heil, P., McKee, B., & English, K. (2000). The impact of polygraphy on admissions
of victims and offenses in adult sexual offenders. Sexual Abuse: A Journal of Research and
Treatment, 12(2), 123-138.
American Polygraph Association. (2007). Model policy for post-conviction sex offender testing.
Polygraph, 36(3).
Association for the Treatment of Sexual Abusers (ATSA). (2004). Practice standards and guidelines
for the evaluation, treatment and management of adult male sexual abusers. Retrieved April 8,
2009, from https://www.atsa.com/pubSoT.html
Ben-Shakhar, G. (2008). The case against the use of polygraph examinations to monitor
post-conviction sex offenders. Legal & Criminological Psychology, 13(2), 191-207.
doi:10.1348/135532508X298577.
Bourke, M. L. & Hernandez, A. E. (2009). The ‘Butner Study’ Redux: A report of the incidence of
hands-on child victimization by child pornography offenders. Journal of Family Violence, 24,
183-191.
Cook, R. (2011). Predicting recidivism of the convicted sexual offender using the polygraph and
Static-99. (Doctoral Dissertation). Retrieved from UMI Dissertation Publishing, ProQuest,
(UMI Number: 3445252).
Emerick, R.L. & Dutton, W.A. (1993). The effect of polygraphy on the self-report of adolescent
sex offenders: Implications for risk assessment. Sexual Abuse: A Journal of Research and
Treatment, 6, 83-103.
English, K. (1998). The containment approach: An aggressive strategy for the community management
of adult sex offenders. Psychology, Public Policy and Law, 4(1/2), 218-235.
English, K., Jones, L., Pasini-Hill, D., Patrick, D., & Cooley-Towell, S. (2000). The value of polygraph
testing in sex offender management. research report submitted to the National Institute of
Justice. Denver, CO: Denver Colorado Department of Public Safety, Division of Criminal
Justice, Ofce of Research and Statistics.
Farber, B.A. (2003). Patient self-disclosure: A review of the research. Journal of Clinical Psychology,
59 (5), 589-600.
Farber, B.A. & Hall, D. (2002). Disclosure to therapists: What is and is not discussed in psychotherapy.
Journal of Clinical Psychology, 58 (4), p 359-370.
Furedy, J. J. (1996). The North American polygraph and psychophysiology: Disinterested,
uninterested, and interested perspectives. International Journal of Psychophysiology, 21(2-3),
97-105. doi:DOI: 10.1016/0167-8760(96)00003-7
Gannon, T.A., Wood, J.L., Pina, A., Tyler, N., Barnoux, M.F.L., & Vasquez, E.A. (2014). An evaluation
of mandatory polygraph testing for sexual offenders in the United Kingdom. Sexual Abuse: A
Journal of Research and Treatment, 26 (2) 178-203.
Grubin, D. (2008). The case for polygraph testing sex offenders. Legal & Criminological Psychology,
13, 177-189. doi:10.1348/135532508X295165
Grubin, D., Madsen, L., Parsons, S., Sosnowski, D., & Warberg, B. (2004). Prospective study of the
impact of polygraphy on high-risk behaviors in adult sex offenders. United States: Retrieved
from National Criminal Justice Reference Service Abstracts.
Back to Front Cover

69
Polygraph, 2015, 44 (2)
Sexual History Disclosure and Recidivism
Handler, M., Nelson, R., & Blalock, B. (2008). A focused polygraph technique for PCSOT and law
enforcement screening programs. Polygraph, 37(2), 100-111.
Harris, A., Phenix, A., Hanson, R. K. & Thornton, D. (2003). Static-99 Coding Rules: Revised 2003,
the Static-99R. Ottawa, ON: Solicitor General of Canada. Printed by the Justice Institute of
British Columbia, Corrections and Community Justice Division.
Heil, P., Ahlmeyer, S. & Simons, D. (2003). Crossover sexual offenses. Sexual Abuse: A Journal of
Research and Treatment, 15(4).
Helmus, L., Babchishin, K. M., Hanson, R. K. & Thornton, D. (2009). Static-99R: Revised age weights.
Retrieved January 10, 2011 from: www.static99.org
Hindman, J., & Peters, J. (2001). Polygraph testing leads to better understanding adult and juvenile
sex offenders. Federal Probation, 65(3), N. PAG.
Honts, C. R. (2004). The psychophysiological detection of deception. In P. A. Granhag, & L. A.
Stromwall (Eds.), The detection of deception in forensic contexts (pp. 103-123). Cambridge,
England: Cambridge University Press.
Honts, C. R. (1994). Psychophysiological detection of deception. Current Directions in Psychological
Science (Wiley-Blackwell), 3(3), 77-82. doi:10.1111/1467-8721.ep10770427.
Honts, C. R. (1996). Criterion development and validity of the CQT in eld application. Journal of
General Psychology, 123(4), 309.
Honts, C. R., & Alloway, W. R. (2007). Information does not affect the validity of a comparison question
test. Legal & Criminological Psychology, 12(2), 311-320. doi:10.1348/135532506X123770
Honts, C. R., Hodes, R. L., & Raskin, D. C. (1985). Effects of physical countermeasures on the
physiological detection of deception. Journal of Applied Psychology, 70(1), 177-187. doi:DOI:
10.1037/0021-9010.70.1.177
Honts, C. R., Raskin, D. C., & Kircher, J. C. (1994). Mental and physical countermeasures reduce
the accuracy of polygraph tests. Journal of Applied Psychology, 79(2), 252-259.
Iacono, W. G. (2008). Accuracy of polygraph techniques: Problems using confessions to determine
ground truth. Physiology & Behavior, 95(1-2), 24-26. doi: DOI: 10.1016/j.physbeh.2008.06.001
Iacono, W. G., & Lykken, D. T. (1997). The validity of the lie detector: Two surveys of scientic
opinion. Journal of Applied Psychology, 82(3), 426-433.
Justice Institute of British Columbia – JIBC. (2011). Static-99R: Sex Offender Risk Assessment.
Retrieved January 10, 2011 from http://www.jibc.ca/course/soap105
Kokish, R., Levenson, J.S., & Blasingame, G. D. (2005). Post-conviction sex offender polygraph
examination: Client-reported perceptions of utility and accuracy. Sexual Abuse: A Journal of
Research and Treatment, 17(2).
Krapohl, D. (2006). Validated polygraph techniques. Polygraph, 35(6), 149.
Levenson, J.S. (2009). Sex offender polygraph examination: An evidence-based case management
tool for social workers. Journal of Evidence-based Social Work, 6: 361-375.
LexisNexis Accurint (2011). Public record criminal history records search program description.
Retrieved May 9, 2011 from: http://www.accurint.com
Marshall, W.L., Thornton, D., Marshall, L.E., Fernandez, Y., & Mann, R. (2001). Treatment of sexual
offenders who are in categorical denial: A pilot project. Sexual Abuse: A Journal of Research
& Treatment, 13, 205-215.
Back to Front Cover

70 Polygraph, 2015, 44 (2)
Konopasek, Nelson
McGrath, R.J., Cumming, G.F., Burchard, B.L., Zeoli, S., & Ellerby, L. (2010). Current practices
and emerging trends in sexual abuser management: The Safer Society 2009 North American
Survey. Brandon, VT: Safer Society Press.
McGrath, R. J., Cumming, G., Hoke, S. E., & Bonn-Miller. (2007). Outcomes in a community
sex offender treatment program: A comparison between polygraphed and matched non-
polygraphed offenders. Sexual Abuse: A Journal of Research and Treatment, 19(4), 381.
National Research Council of the National Academies of Science, Committee to Review the Scientic
Evidence on Polygraph (Ed.). (2003). The polygraph and lie detection (D.O. Sciences ed.).
Washington DC: The National Academies Press.
Nelson, R, Handler, M., Krapohl, D., Gougler, M., Shaw, P. & Bierman, L. – Ad-hoc Committee
on Validated Techniques – American Polygraph Association (2011). Meta-analytic survey of
criterion accuracy of validated polygraph techniques. Polygraph 40 (4).
O’Connell, M. A. (1998). Using polygraph testing to assess deviant sexual history of sex offenders.
(Doctoral Dissertation), University of Washington, Seattle, WA. Retrieved from Dissertation
Abstracts International (49).
O’Connell, M.A. (2011). Recommendations for research methodology to avoid conict of interest,
dual relationships and vested interest: The utilization of independent third party research
assistants. Personal communication on February 26, 2011.
Offe, H., & Offe, S. (2007). The comparison question test: Does it work and if so how? Law and
Human Behavior, 31(3), 291.
Oregon Judicial Department. (2009). Oregon judicial information network [OJIN], access to Oregon
cases on the internet. Retrieved October 3, 2010, from http://courts.oregon.gov/OJD/
OnlineServices/OJIN/index.page
Patrick, C.J. & Iacono, W.G. (1989). Psychopathy, threat and polygraph test accuracy. Journal of
Applied Psychology, 74, 347-355.
Pratley, J. & Goodman-Delahunty, J. (2011). Increased self-disclosure of offending by intrafamilial
child sex offenders. Sexual Abuse in Australia and New Zealand, 3 (1) 10-22.
Rosky, J.W. (2013). The (f)utility of post-conviction polygraph testing. Sexual Abuse: A Journal of
Research and Treatment, 25(3), 259-281.
Senter, S., Weatherman, D., Krapohl, D., & Horvath, F. (2010). Psychological set or differential
salience: A proposal for reconciling theory and terminology in polygraph testing. Polygraph,
39(2), 109-117.
Tabachnick, B. C., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston, MA: Pearson
Education.
Washington State Institute for Public Policy, Director Roxanne Leib. (2009). Recidivism study technical
assistance – obtaining criminal history data. Retrieved December 1, 2009 from http://www.
wsipp.wa.gov/default.asp
Washington State Patrol (2011), Washington Access to Criminal History – WATCH, Retrieved May 1,
2011 from: https://fortress.wa.gov/wsp/watch.
Wilcox, D. T., Sosnowski, D., & Middleton, D. (1999). The use of the polygraph in the community
supervision of sex offenders. Probation Journal, 46, 234-240.
Wilcox, D. T., & Sosnowski, D. E. (2005). Polygraph examination of British sexual offenders: A pilot
study on sexual history disclosure testing. Journal of Sexual Aggression, 11(1), 3-25. doi:10.
1080/13552600410001667797
Back to Front Cover

71
Polygraph, 2015, 44 (2)
A New Paradigm for The Experimental Study of Malintent
A new paradigm for the experimental study of Malintent
Charles R. Honts
Psychology Department
Boise State University
Abstract
A new laboratory paradigm for the study of credibility assessment and deception concerning
malintent was tested. Malintent may be a distinct concept from the traditional deception for past
action that has been the subject of numerous studies. Sixty participants were either innocent
or were given information and malintent to commit a theft. Participants were then screened for
malintent using a modied Test for Espionage and Sabotage (TES). The TES was scored with the
Kircher and Raskin (1988) discriminant analysis algorithm and produced better than chance results
that were comparable to the results for the TES in a forensic deception detection setting. (Honts &
Alloway, 2007). This new paradigm provides an experimental framework for exploring the concept of
malintent and efforts to detect it.
Key Words: malintent, deception detection, national security screening, portals
Corresponding Author:
Charles R. Honts
Psychology Department
Boise State University
1910 University Drive MS-1715
Boise ID 83716-5664 USA
chonts@boisestate.edu
Voice: 208.867.2027
FAX: 208.426.4386
A large body of research indicates
that unassisted individuals, including trained
forensic and security professionals are only
slightly better than chance at detection
deception (Hartwig & Bond, 2011). In the
post 9-11 environment, the United States
Government responded to the problem of
assessing credibility at portals in several ways
involving new research and the application
of new techniques to the eld setting (Honts
& Hartwig, 2014). Unfortunately, all of those
efforts are scientically problematic because
of poor methodology, and because they fail to
address the conceptual differences between
assessing credibility for intent to perform
bad acts and assessing credibility for the
commission of past acts (Honts & Hartwig).
Polygraph examiners who do employment
screening examination in law enforcement
and national security face a similar credibility
assessment problem as that faced at portals.
Individuals may present themselves for
polygraph screening with the malintent to
perform bad actions if they are hired, but they
may not, at the time of the polygraph screening
have actually committed any bad acts. It is
not clear how polygraph screening tests would
perform under such circumstances.
Honts and Hartwig (2014) note that
the deceptive context for assessing malintent
differs in critical ways from assessing
credibility concerning a past act. Most of the
research conducted on credibility assessment
has focused on the problem of detecting
deception concerning statements about acts
that took place in the past while the study of
deception for intent has received little attention
(Granhag, 2010). Most research is typied by
asking questions about some event that the
person either did or did not participate in the
the past. A typical study from the past event
deception literature would have a mock crime
where some participants stole money and some
Back to Front Cover

72 Polygraph, 2015, 44 (2)
Honts
did not (see Honts & Reavy, 2015 for a recent
example). In such a setting the guilty person
has the episodic memories associated with
the criminal act and a concern for deception
detection. Concern about deception detection
then results in a variety of sequelae that
involve masking the deception, monitoring
the receiver and associated physiological and
behavioral responses (Vrij & Gannis, 2014).
For the innocent person, the concern that her
or his truthful statements will not be believed is
energized by the same potential consequences
and sequelae as those faced by the guilty.
The deceptive context for assessing
malintent at a portal or at an employment
screening session is quite different (Honts &
Hartwig, 2014). At the portal, or in employment
screening situation, the innocent person is
not the accused in a criminal investigation,
and it seems doubtful that many innocent
people approaching border or transportation
portals, or employment screening, feel
anything near the equivalent of the emotional
response felt by a falsely accused criminal
suspect. However, Honts and Hartwig also
note that the truthful person at a portal may
well feel anxious about the general process of
screening. Nevertheless, it seems likely that
for most innocent individuals, approaching a
portal is a necessary inconvenience that may
cause the innocent aggravation and minor
anxiety, but little concern of true jeopardy
(Honts & Hartwig). In an employment screening
situation, a person without malintent
seems unlikely to have much concern about
questions concerning malintent. Certainly the
concern and jeopardy for truthful individuals
in an employment screening situation is much
less than that of a falsely accused suspect in a
criminal investigation.
For those intending to do bad acts if
granted access, the deceptive context may
also be different from the traditional situation
(Honts & Hartwig, 2014). Those intending to do
bad acts, may not have, as yet, have committed
any bad acts, nor may they know specically
what their future bad acts might be. Those with
malintent want to pass the portal, or obtain
the job, so that he or she can do bad acts in
the future. The person with malintent may or
may not have false credentials, but if it is their
intention to do bad acts in the future, that is
the central nature of their deception and the
focus of a relevant credibility assessment. To
date, little research has addressed credibility
assessment for malintent in the deceptive
context presented by portals or employment
screening. It is simply not clear whether or
not research done in a criminal/investigative
context will generalize to a malintent situation.
There is a body of research concerning
deception detection for intent. Typical of this
research was a study by Vrij, Leal, Mann, and
Granhag (2011). In Vrij, et al., participants were
asked to pretend they were part of a mission to
collect a package from a specied location and
then deliver it somewhere else. Participants
were told that they might be stopped by
agents and were given a code exchange that
would identify an agent as friendly or hostile.
Participant were instructed to be truthful with
friendly agents and to lie about their intent
and mission to the hostile agents. While on the
way to complete the mission participants were
stopped and interviewed by either a friendly or
a hostile agent. After completing the mission
participants were stopped a second time and
interviewed by a friendly or hostile agent. Data
extracted from the interviews indicated that
after the completed event were more markers
of plausibility and they contained more details
than the interviews about intention. However,
when undergraduate college students were
asked to evaluate transcripts of the interviews
no signicant effects of intention versus
completed event were found.
Although Vrij et al., tested the detection
of intent, I would argue that they did not
test malintent because in both the truthful
and deceptive conditions every participant
had exactly the same knowledge. While this
created a powerful experiment design, it is not
representative of conditions in the eld. The
lack of any specic motivation associated with
deception detection is also troubling. Kircher,
Horowitz and Raskin (1988) reported a meta-
analysis that found explicit motivation was
an important variable in predicting laboratory
accuracy rates with polygraph tests. Some
other studies that tested intent, but without
malintent or specic motivation associated
with deception detection were reported by
Meijer, Verschuere, and Merckelbach (2010),
Sooniste, Granhag, Knieps and Vrij (2013),
and Vrij, Granhag, Mann, and Leal, (2011).
The paradigm described in this study
was an attempt to model the critical aspects of
malintent detection in a laboratory paradigm.
For purely pragmatic reasons we decided to use
psychophysiological deception detection (PDD)
methods and technology. There is a substantial
literature on the use of PDD to assess credibility
in both forensic and screening settings
(Raskin & Kircher, 2014). Although clearly not
perfect, PDD has consistently shown better
Back to Front Cover

73
Polygraph, 2015, 44 (2)
A New Paradigm for The Experimental Study of Malintent
than chance performance in both forensic and
screening settings and provides a substantial
amount of information gain over unassisted
deception across a wide range of base rates
(Honts & Schweinle, 2009). Although PDD in
its present form is clearly not applicable to
airport portals, it is used in national security,
and law enforcement employment screening
situations. Moreover, the substantial database
of deception detection in traditional settings
provides a stable reference for initial efforts to
detect malintent.
Method
Participants
Participants were 60 college students
enrolled in General Psychology classes. The
participants received course credit for their
participation and as part of the manipulation
described below some participants had
the possibility to win $14 in movie passes.
Participants average age was 20.7 years, SD
= 3.16, and 37 (62%) of the participants were
men.
Apparatus
The apparatus was the same as that
used in Honts and Alloway (2007). Physiological
data were collected with a Stoelting commercial
polygraph instrument running version 3.2 of
the Computerized Polygraph System (CPS)
software (Kircher & Raskin, 2002). Respiration
data were collected from Pneumotrace sensors
placed over the upper chest and the abdomen.
Skin conductance was collected from two Ag-
AgCl electrodes placed on the palmar surface
of the distal phalanx of the rst and third nger
of the participant’s left hand. Relative blood
pressure was recorded from an inated cuff
placed around the participant’s upper right
arm. Vasomotor activity was recorded from the
palmar surface of the participant’s left thumb.
Movement was monitored with a sensor placed
under the legs of the participant’s chair. The
CPS software was used to edit artifacts from
data and evaluate the data.
Procedure
Participants were recruited through
the General Psychology course signup
software. They were responding to an ad for a
research project where participants might be
asked to lie and then take a polygraph test.
On arrival, participants watched a video about
their rights as research participants and were
then given an informed consent form to review
and sign. The video and form indicated to the
participants that they could withdraw from
participation at any time and still received their
participation credit in General Psychology.
After agreeing to continue, the research
assistant instructed the participants chose an
envelope from a box. The research assistant
then left the participant alone in the room.
Participants thus opened that envelope in
private. The envelope contained instructions
to watch a second video and a password for
that video. Two participation videos dened
the two conditions of the study. The Innocent
Participant video provided the following
instructions: Some people approaching the
portal polygraph examination were given
the combination to the safe in Room A and
those people have the intention of opening
the safe and stealing something valuable
within. However, as an innocent person you
do not know the combination to the safe and
you have no intention to steal anything. You
will be given a polygraph test about knowing
the combination to the safe, and about your
intention to open the safe in Room A. Maintain
your innocence and lack of knowledge during
the polygraph and if you pass your polygraph
you will be admitted to Room A. There you will be
immediately debriefed and your participation
will be over. However, if you fail the polygraph
you will be taken to Room B where you will
watch a 30-minute educational lm (TV Eye,
1983) and then you will be debriefed.
The Malintent Participant video (with
the same person as shown in the Innocent
Video) provided the following instructions:
Some of the participants are going to be given
the combination to the safe in Room A. You are
one of those persons. The combination to the
safe is 1 0 6 A. You will be given a polygraph
about your knowledge of the combination and
about your intention to rob the safe. If you are
successful in passing the polygraph, you will
be admitted to Room A where you can open
the safe and take movie passes worth $14. You
will be allowed to keep those movie passes. To
pass the polygraph you must maintain your
lack of knowledge and intention through the
entire examination. If you fail the polygraph
you will be taken to Room B where you will
watch a 30-minute educational video and then
you will be debriefed. Failing the polygraph
means you will never have the chance to take
the movie passes.
Using methods similar to those used
Back to Front Cover

74 Polygraph, 2015, 44 (2)
Honts
by Honts and Alloway (2007) all Participants
were then tested with a standard eld
polygraph instrument and they were given a
complete polygraph using a modied version
of Test for Espionage and Sabotage (TES;
Department of Defense Polygraph Institute
Research Division Staff, 1999). The TES was
chosen so that the results of our malintent
polygraph could be directly compared to
the results Honts and Alloway. The relevant
questions of this examination were: Do you
know the combination to the safe in Room
A? Do you have a plan to open the Safe in
Room A? If you are admitted to Room A,
will you attempt to open the safe? Are you
taking part in a plan to take something
from the safe in Room A? Physiological data
were collected from three repetitions of each
relevant question. Following data collection
the examiner used the discriminant analysis
classication algorithm included in the CPS
software (Kircher and Raskin, 1988; 2002)
to classify the participant as truthful or
deceptive. Depending upon the classication
the participant was then taken to the room
associated with their outcome. Participants
who passed their examinations were given
the movie passes and those who produced
deceptive results on their examinations
were asked to watch the educational lm.
Participants were then debriefed by a research
assistant and were fully informed about the
design of the experiment. Participant questions
were answered.
Results
The discriminant analysis procedure
used in this study (Kircher & Raskin, 1988);
provided as part of the CPS software) produced
an a posteriori probability of truthfulness
(p|T) that can be considered by itself for its
direct informative value or used against a
cut score for classication. Those p|T values
were then tested to see if deception detection
was possible in this setting. An independent
groups t-test of the p|T values indicated
that signicant detection was obtained,
t(58) = 3.26, p = 0.002. The p|T for Innocent
participants (M = 0.69, SD = 0.37) was higher
than for Malintent participants (M = 0.39, SD
= 0.34). The correlation between the guilt
criterion and the p|T values was 0.394, p <
.01. These results are similar to those reported
by Honts and Alloway (2007) for the TES in a
forensic setting (Innocent M = 0.72 and Guilty
M = 0.40). If decisions were made so that p|T
> .5 were classied as truthful and p|T values
< .5 were deceptive, 70% of the malintent and
67% of the innocent individuals were classied
correctly. That classication was signicantly
above chance, χ2 = 8.01 (1), p = .004, Kendall’s
tau-b = .37, p = .002, and was again similar
to the performance of the TES in Honts and
Alloway’s (2007) forensic paradigm.
Discussion
The present results provide a proof
of concept for a new paradigm to assess
malintent. Malintent participants approached
a screening task not having committed a
transgression in the past, but with knowledge
and intent to commit a transgression if they
were able to pass the screening. Innocent
participants approached the screening
task without malintent or malintent related
information and were motivated to pass the
screening to avoid delay. A modied version
of the U. S. Government’s Test for Espionage
and Sabotage, a psychophysiological
deception detection test, was used as the
credibility assessment tool. The TES was
able to discriminate malintent from innocent
participants at better than chance levels. The
performance of the TES in this malintent
paradigm was much better than that
reported for unassisted individuals (Hartwig,
Granhag, & Luke, 2014) and was comparable
to the performance of the TES in a forensic
laboratory paradigm (Honts & Alloway, 2007).
Unfortunately, the nature of the equipment
and the time necessary for administration of
the TES make it an unreasonable candidate
for use a high volume portals, although it is
currently used for employment screening
in national security screening settings in
the United States (Department of Defense
Polygraph Institute Research Division Staff,
1998).
Nevertheless, these results validate
this paradigm as a way to establish a basic
malintent paradigm. The paradigm is easy to
implement and should be easily adaptable to
a variety of manipulations that would allow
for the explication of the malintent construct.
Research is urgently needed to dene the limits
and nature of the malintent concept. Although
already being widely attempted in the eld
(Honts & Hartwig, 2014), such basic research
would seem to be absolutely necessary before
legitimately applying techniques to detect
malintent in the eld.
Back to Front Cover

75
Polygraph, 2015, 44 (2)
A New Paradigm for The Experimental Study of Malintent
References
Department of Defense Polygraph Institute Research Division Staff. (1998). Psychophysiological
detection of deception accuracy rates obtained using the test for espionage and sabotage.
Polygraph, 27, 68–73.
Granhag, P. A. (2010). On the psycho-legal study of true and false intentions: Dangerous waters and some
stepping stones. Open Criminology Journal, 3(2), 3743. doi:10.2174/1874917801003020037
Hartwig, M., & Bond, C.F. (2011). Why do lie-catchers fail? A lens model meta-analysis of human lie
judgments. Psychological Bulletin, 137, 643–659.
Hartwig, M., Granhag, P. A., & Luke, T. (2014). Strategic use of evidence during investigative
interviews: The state of the science. In, Raskin, D. C., Honts, C. R., & Kircher, J. C. Credibility
assessment: Scientic research and applications:. (pp. 1-36 ). Oxford, UK: Academic Press.
http://dx.doi.org/10.1016/B978-0-12-394433-7.00001-4
Honts, C. R. & Alloway, W. (2007). Information does not affect the validity of a comparison question
test. Legal And Criminological Psychology, 12, 311-312. (Available online in 2006)
Honts, C. R. & Hartwig, M. (2014). Credibility assessment at portals. In, Raskin, D. C., Honts, C.
R., & Kircher, J. C. Credibility assessment: Scientic research and applications:. (pp. 37-61 ).
Oxford, UK: Academic Press. http://dx.doi.org/10.1016/B978-0-12-394433-7.00002-6
Honts, C. R., & Reavy, R., (2015). The comparison question polygraph test: A contrast of methods
and scoring. Physiology and Behavior, 143, 15-26. Published online 24 February 2015,
doi:10.1016/j.physbeh.2015.02.028
Honts, C. R., & Schweinle, W. (2009). Information gain of psychophysiological detection of deception
in forensic and screening settings. Applied Psychophysiology and Biofeedback, 34, 161-172.
(Available online July 2009)
Kircher, J.C., Horowitz, S.W., Raskin, D.C., 1988. Meta-analysis of mock crime studies of the control
question polygraph technique. Law and Human Behavior, 12, 79–90
Kircher, J. C., & Raskin, D. C. (1988). Human versus computerized evaluations of polygraph data in
a laboratory setting. Journal of Applied Psychology, 73, 291-302.
Meijer, E. H., Verschuere, B., & Merckelbach, H. (2010). Detecting criminal intent with the concealed
information test. The Open Criminology Journal, 3, 44-47.
Raskin, D. C., & Kircher, J. C. (2014). Validity of polygraph techniques and decision methods. In,
Raskin, D. C., Honts, C. R., & Kircher, J. C. Credibility assessment: Scientic research and
applications:. (pp. 63-129). Oxford, UK: Academic Press. http://dx.doi.org/10.1016/B978-0-12-
394433-7.00003-8
Sooniste, T., Granhag, P. A., Knieps, M., & Vrij, A. (2013). True and false intentions: asking about the
past to detect lies about the future. Psychology, Crime & Law, 19, 673-685.
TV Eye (1983). Telling the truth. Episode in TV Eye series, London, UK: Thames Color Productions.
Vrij, A., & Gannis, G. (2014). Theories in deception and lie detection. In, Raskin, D. C., Honts, C. R.,
& Kircher, J. C. Credibility assessment: Scientic research and applications:. (pp. 301-374).
Oxford, UK: Academic Press. http://dx.doi.org/10.1016/B978-0-12-394433-7.00007-5
Vrij, A., Granhag, P. A., Mann, S. A., & Leal, S. (2011). Lying about ying: The rst experiment to
detect false intent. Psychology, Crime & Law, 17, 611-620.
Back to Front Cover

76 Polygraph, 2015, 44 (2)
Honts
Vrij, A., Leal, S., Mann, S. A., & Granhag, P. A. (2011). A comparison between lying about intentions
and past activities: Verbal cues and detection accuracy. Applied Cognitive Psychology, 25,
212–218. http://dx.doi.org/10.1002/acp. 1665.
Author Note
The author would like to thank the following individuals for their work in collecting the data in this
study: Scott McBride, Flavia Pittman, James Pittman, Adela Anderson, and Ashley Christiansen.
Correspondence concerning this paper should be sent to the author at: Psychology Department,
Boise State University, 1910 University Drive, Boise, ID 83725-1715 or chonts@boisestate.edu
Some of the results presented here were also given as a poster presented at the annual meeting of
the Association for Psychological Science in Chicago: Honts, C. R., Pittman, F. A., Pittman, J. V.,
McBride, S. T., Anderson, A. B., & Christiansen, A. K., (2008, May). A New Paradigm for the Study of
Deception Detection at Portals.
Back to Front Cover