Concealed Information Test with Combination of ERP Recording and Autonomic Measurements
The concealed information test (CIT) is based on a comparison between the subject’s physiological responses to a probe and irrelevant items in order to detect concealed information. The main purpose of our study was to investigate the CIT accuracy enhancement related to a combination of recording...
Gespeichert in:
| Datum: | 2013 |
|---|---|
| Hauptverfasser: | , |
| Format: | Artikel |
| Sprache: | English |
| Veröffentlicht: |
Інститут фізіології ім. О.О. Богомольця НАН України
2013
|
| Schriftenreihe: | Нейрофизиология |
| Online Zugang: | http://dspace.nbuv.gov.ua/handle/123456789/148094 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Zitieren: | Concealed Information Test with Combination of ERP Recording and Autonomic Measurements / E.D. Farahani, M.H.A. Moradi // Нейрофизиология. — 2013. — Т. 45, № 3. — С. 252-263. — Бібліогр.: 38 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraine| id |
irk-123456789-148094 |
|---|---|
| record_format |
dspace |
| spelling |
irk-123456789-1480942019-02-17T01:26:30Z Concealed Information Test with Combination of ERP Recording and Autonomic Measurements Farahani, E.D. Moradi, M.H.A. The concealed information test (CIT) is based on a comparison between the subject’s physiological responses to a probe and irrelevant items in order to detect concealed information. The main purpose of our study was to investigate the CIT accuracy enhancement related to a combination of recording of event-related potentials (ERPs) and autonomic measurements. We tried to maximally liken the experimental conditions to real ones by the use of a criminal context in the “mock crime” instruction and real innocent subjects instead of hypothetical ones. Fifty-two subjects volunteered and performed just one of the innocent or guilty scenarios. The CIT was designed in five blocks with short interstimulus intervals. In each block, stimuli were presented in the 7th-order balanced sequence. In addition to EEG phenomena, the heart rate, skin conductance responses (SCRs), respiratory activity, and finger plethysmogram were recorded. Statistical analyses showed that there was a significant difference between standardized difference scores of the guilty and innocent groups in both ERP and autonomic measures. The SCR did not achieve the expected results reported in standard autonomic-based CIT studies. A review of the two classification methods showed that the combination of ERP and autonomic measurements enhances the CIT accuracy. The best classification accuracy obtained by the aid of linear discriminant analysis (LDA) was 90.9%. It seems that using a criminal context in the “mock crime” instruction and the rewardpunishment system made subjects more attentive and involved in the experiment; therefore, the accuracy was improved compared with that in similar studies Тест із прихованою інформацією (ТПІ) базується на порівнянні фізіологічних реакцій суб’єкта в зондуючих і нейтральних ситуаціях, спрямованому на виявлення такої інформації. У нашій роботі ми визначали точність результатів ТПІ в умовах, коли цей тест поєднували з відведенням пов’язаних з подією потенціалів (ППП) і вимірюванням вегетативних показників. Особливістю нашого дослідження було максимальне наближення експериментальних умов до реальних за допомогою використання кримінального контексту в інструкції „макетування злочину” та „реально невинуватих” суб’єктів замість „гіпотетично невинуватих”. 52 волонтери виконували один із сценаріїв „винуватий/невинуватий”. ТПІ вміщував п’ять блоків з короткими міжстимульними інтервалами. У кожному з блоків стимули пред’являлись у балансованій послідовності сьомого порядку. Крім ЕЕГ-активності, реєстрували частоту пульсу, зміни шкірної провідності (ШП), дихальну активність і плетизмограму пальців. Статистичний аналіз показав, що між стандартизованими оцінками відмінностей характеристик як ППП, так і вегетативних показників у „винуватій” та „невинуватій” групах виявлялися вірогідні відмінності. При виявленні змін ШП очікувані результати, описані для стандартних результатів ТПІ, що базуються на вимірювання вегетативних показників, не досягалися. Порівняння двох класифікаційних методик показало, що поєднання відведення ППП і вегетативних вимірювань підвищує точність результатів ТПІ. Найбільша точність класифікації, отриманої із застосуванням лінійного дискримінантного аналізу, складала 90.9 %. Скоріш за все, використання кримінального контексту в інструкції „макетування злочину” і системи преміювання/покарання забезпечувало більший рівень уваги тестованих та їх більше залучення в експеримент, що й підвищувало точність тестування порівняно з такою в аналогічних дослідженнях 2013 Article Concealed Information Test with Combination of ERP Recording and Autonomic Measurements / E.D. Farahani, M.H.A. Moradi // Нейрофизиология. — 2013. — Т. 45, № 3. — С. 252-263. — Бібліогр.: 38 назв. — англ. 0028-2561 http://dspace.nbuv.gov.ua/handle/123456789/148094 159.9.019.4:612.821 en Нейрофизиология Інститут фізіології ім. О.О. Богомольця НАН України |
| institution |
Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| collection |
DSpace DC |
| language |
English |
| description |
The concealed information test (CIT) is based on a comparison between the subject’s
physiological responses to a probe and irrelevant items in order to detect concealed
information. The main purpose of our study was to investigate the CIT accuracy enhancement
related to a combination of recording of event-related potentials (ERPs) and autonomic
measurements. We tried to maximally liken the experimental conditions to real ones by the
use of a criminal context in the “mock crime” instruction and real innocent subjects instead
of hypothetical ones. Fifty-two subjects volunteered and performed just one of the innocent
or guilty scenarios. The CIT was designed in five blocks with short interstimulus intervals.
In each block, stimuli were presented in the 7th-order balanced sequence. In addition to
EEG phenomena, the heart rate, skin conductance responses (SCRs), respiratory activity, and
finger plethysmogram were recorded. Statistical analyses showed that there was a significant
difference between standardized difference scores of the guilty and innocent groups in both
ERP and autonomic measures. The SCR did not achieve the expected results reported in
standard autonomic-based CIT studies. A review of the two classification methods showed
that the combination of ERP and autonomic measurements enhances the CIT accuracy. The
best classification accuracy obtained by the aid of linear discriminant analysis (LDA) was
90.9%. It seems that using a criminal context in the “mock crime” instruction and the rewardpunishment system made subjects more attentive and involved in the experiment; therefore,
the accuracy was improved compared with that in similar studies |
| format |
Article |
| author |
Farahani, E.D. Moradi, M.H.A. |
| spellingShingle |
Farahani, E.D. Moradi, M.H.A. Concealed Information Test with Combination of ERP Recording and Autonomic Measurements Нейрофизиология |
| author_facet |
Farahani, E.D. Moradi, M.H.A. |
| author_sort |
Farahani, E.D. |
| title |
Concealed Information Test with Combination of ERP Recording and Autonomic Measurements |
| title_short |
Concealed Information Test with Combination of ERP Recording and Autonomic Measurements |
| title_full |
Concealed Information Test with Combination of ERP Recording and Autonomic Measurements |
| title_fullStr |
Concealed Information Test with Combination of ERP Recording and Autonomic Measurements |
| title_full_unstemmed |
Concealed Information Test with Combination of ERP Recording and Autonomic Measurements |
| title_sort |
concealed information test with combination of erp recording and autonomic measurements |
| publisher |
Інститут фізіології ім. О.О. Богомольця НАН України |
| publishDate |
2013 |
| url |
http://dspace.nbuv.gov.ua/handle/123456789/148094 |
| citation_txt |
Concealed Information Test with Combination of ERP Recording and Autonomic Measurements / E.D. Farahani, M.H.A. Moradi // Нейрофизиология. — 2013. — Т. 45, № 3. — С. 252-263. — Бібліогр.: 38 назв. — англ. |
| series |
Нейрофизиология |
| work_keys_str_mv |
AT farahanied concealedinformationtestwithcombinationoferprecordingandautonomicmeasurements AT moradimha concealedinformationtestwithcombinationoferprecordingandautonomicmeasurements |
| first_indexed |
2025-07-12T18:18:05Z |
| last_indexed |
2025-07-12T18:18:05Z |
| _version_ |
1837466168222285824 |
| fulltext |
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3252
UDC 159.9.019.4:612.821
E. D. FARAHANI1 and M. H. MORADI1
A CONCEALED INFORMATION TEST WITH COMBINATION OF ERP RECORDING
AND AUTONOMIC MEASUREMENTS
Received October 31, 2012.
The concealed information test (CIT) is based on a comparison between the subject’s
physiological responses to a probe and irrelevant items in order to detect concealed
information. The main purpose of our study was to investigate the CIT accuracy enhancement
related to a combination of recording of event-related potentials (ERPs) and autonomic
measurements. We tried to maximally liken the experimental conditions to real ones by the
use of a criminal context in the “mock crime” instruction and real innocent subjects instead
of hypothetical ones. Fifty-two subjects volunteered and performed just one of the innocent
or guilty scenarios. The CIT was designed in five blocks with short interstimulus intervals.
In each block, stimuli were presented in the 7th-order balanced sequence. In addition to
EEG phenomena, the heart rate, skin conductance responses (SCRs), respiratory activity, and
finger plethysmogram were recorded. Statistical analyses showed that there was a significant
difference between standardized difference scores of the guilty and innocent groups in both
ERP and autonomic measures. The SCR did not achieve the expected results reported in
standard autonomic-based CIT studies. A review of the two classification methods showed
that the combination of ERP and autonomic measurements enhances the CIT accuracy. The
best classification accuracy obtained by the aid of linear discriminant analysis (LDA) was
90.9%. It seems that using a criminal context in the “mock crime” instruction and the reward-
punishment system made subjects more attentive and involved in the experiment; therefore,
the accuracy was improved compared with that in similar studies.
Keywords: concealed information test, event-related potentials, autonomic responses,
logistic regression model, linear discriminant analysis.
1 Biomedical Engineering Faculty, Amirkabir University of Technology,
Tehran, Iran.
Correspondence should be addressed to M. H. Moradi
(e-mail: mhmoradi@aut.ac.ir).
INTRODUCTION
Attempts to detect concealed information using
recording of physiological indices have a rather
long history. Initial studies in this matter refer to
understanding the relationship between the heart
rate (HR) and deceptively denying knowledge.
Further studies (such as Lombroso in the late
19th century and Marston in 1917) were carried out
to find new deception signs [1-3]. The most common
physiological measures in polygraph systems are
parameters of respiration, cardiovascular measures,
and electrodermal responses, which mainly reflect
functions of the autonomic nervous system (ANS).
In recent years, various approaches were introduced
for psychophysiological detection of deception, such
as studying brain functions in a deception procedure
using functional brain imaging and also recording and
investigation of brain potentials [4, 5]. Event-related
brain potentials (ERPs) were widely studied and
demonstrated more satisfactory results [6]. The P300
wave is the most important component of ERPs, which
is recently used in most studies. In some previous
ERP studies, the P300 amplitude was reported to be
a reliable index for detection of deception [7, 8]. In
recent studies, a new approach was introduced in
order to improve the results of detection of deception
by combining the measurements of ANS and СNS
functioning.
The concealed information test (CIT), also referred
to as the guilty knowledge test [9], is an effective
method of psychophysiological detection of concealed
information on crime [10]. In this method, differential
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3 253
A CONCEALED INFORMATION TEST WITH COMBINATION OF ERP RECORDING
physiological responses to specific items are surveyed
[9]. One of these items (probe) corresponds to the
aspects of the crime that are under investigation, and
the other one is an irrelevant item. The irrelevant
and probe items are repeatedly presented in a certain
sequence. An innocent subject without any knowledge
of the crime demonstrates similar physiological
responses to both items, whereas a guilty subject who
deceptively denies his deed-related knowledge shows
different physiological responses to these items [9].
A CNS response that mirrors cognitive processing
and a peripheral response that mainly reflects a
function of the ANS might complement each other
more effectively. However, a combination of ERP
recording and autonomic measures within the same
experiment generally entails some difficulties [11].
First, short interstimulus intervals (ISIs) should be
used in ERP-based tests because it was shown that
the P300 amplitude is affected by ISI values [12].
At the same time, long ISIs (20-30 sec) are used in
autonomic-based CITs to provide an adequate recovery
time. Measuring the skin conductance response (SCR)
within a short-ISI paradigm results in overlapping
responses that are difficult to quantify independently
[11, 13]. Second, EEG evaluation based on a single
trial is not very reliable due to a rather low signal-
to-noise ratio. Therefore, large numbers of stimuli
are presented in most ERP studies in order to obtain
an adequate number of valid ERPs per condition
[11]. In contrast, the autonomic-based CIT uses
smaller numbers of stimulus presentations because
the amplitude of autonomic responses becomes
critically small [13], while the SCR is known to
intensely habituate when a large number of stimuli
are presented [11].
However, a few recent studies have combined
autonomic and ERP measurements in the detection
of concealed information [11]. Studies in this field
are divided into two categories. The first category
includes studies where short ISIs and large numbers
of stimulus presentations were used [11, 14]. The
second category includes studies with long ISIs and
limited numbers of stimulus presentations [15, 16].
In the second category, the studies, except that by
Matsuda et al. [17], did not achieve a typical probe
vs irrelevant difference for the P300 amplitude. The
authors believed that using long ISIs was the reason
for this phenomenon and suggested that further studies
should evaluate shorter ISIs as a solution [16]. In the
first category, studies gained incremental validity
from combined measurements. Due to using short
ISIs and large numbers of stimulus presentations,
the discrimination and correct classification rate of
autonomic measures in these studies remained below
those in most autonomic-based CIT studies.
The purpose of our study was to evaluate the CIT
accuracy enhancement with the aid of combination of
ERP recording and autonomic measures. In this study,
short ISIs and large numbers of stimulus presentations
are primarily selected as the paradigm. On the other
hand, some arrangements should be considered to
handle the side effect of short ISIs in autonomic
measurements. The second purpose of the study was
to make the experimental conditions closer to real
conditions. This purpose has been achieved in two
ways, first by designing a “mock crime” scenario
with a criminal context, and, second, by designing an
innocent scenario and involving real innocent subjects
instead of hypothetical ones.
METHODS
Participants. Fifty-two healthy students of the
Biomedical Engineering Faculty in the Amirkabir
University of Technology (40 men and 12 women;
mean age 22.5 ± 3.5 years; all right-handed; all
had normal or corrected vision) participated in this
experiment. They were rewarded with a gold coin
(value around US $12) after the experiment.
Design and Procedure. After the subjects had
given their informed consent, they were asked to
select one of two envelopes containing the instruction
of a “guilty or innocent” scenario. It should be taken
into account that both of envelopes contained the
same instruction, but this instruction was referred to
the guilty scenario for half of the subjects, while the
scenario was innocent for another half.
After selection of the scenario, the experimenter left
the laboratory, and the subjects read the instruction
in order to perform the respective actions. The guilty
scenario consisted of stealing a gold coin (value
around US $12, hidden in a wallet) and a cell phone
from a personal locker in the laboratory. In order to
open the personal locker, subjects had to find a key
hidden in a cupboard in the kitchen. In the innocent
scenario, the task involved washing dirty cups put in
the kitchen sink. At the end of both scenarios, subjects
were requested to go to the lobby and wait for the
experimenter to come. After 7 min, the experimenter
approached the subjects and informed them that “a
crime has been committed, and you are one of the
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3254
E. D. FARAHANI and M. H. MORADI
suspects; if you succeed in taking the test, you will
obtain a gold coin.” Following this, the experimenter
escorted the subjects back to the laboratory, where the
main test was performed. Because the subjects did
not know that the two envelopes contained the same
instruction, and they were trying to get the gold coin,
it was expected that the subjects were under stress.
These arrangements likened the mock crime scenario
to real condition and prevented the subjects from being
inattentive.
In this study, a variant of the CIT consisting the five
blocks was used. The presented pictures in each block
belong to one category of the details of the crime scene
(i.e., coins, keys, cell phones, wallets, and lockers).
Each block contained seven types of the stimuli (one
target, one probe, four irrelevant, and a null event).
Each stimulus was presented seven times except for
the null stimulus that was presented six times. The
sequence of the presented stimuli was based on a
pseudorandom sequence called M-sequence that is a
balanced-order sequence [18]. Fourteen stimuli were
added to the sequence of 48 ones, creating a “history”
and “future” for the initial and last stimuli. Thus, the
stimuli were presented in a pseudorandom series of
62 stimuli in each block. These stimuli were presented
with ISIs varying from 2.3 to 2.7 sec. One-minute-
long rest between blocks were set. The designed CIT
is illustrated in Fig. 1. A 17-color screen at a distance
of 90 cm from the subject was used for presentation of
the stimuli.
To overcome the overlap problem (in particular,
overlapping SCRs) in our study, we used the 7th-order
M-sequence for each block of the CIT. For this purpose,
a null stimulus was added to the standard stimulus set
(Fig. 1). The characteristic of the M-sequences is that
each type of the stimulus has an identical bias. This
means that, although the average response for each
type is biased by responses to previous stimuli, this
bias is identical for every stimulus type [14]. So, the
responses can be compared without worrying about
an overlap problem. As an example of the 3rd-order
M-sequence, see Fig. 2. This sequence has three types
of stimuli. Each stimulus was repeated three times,
except for C (a null stimulus was added to stimulus
set) that was repeated two times. Figures 2b and 2c
show the bias on the A and B, respectively. Due to the
characteristic of M-sequence, both A and B have an
identical bias.
The subjects were randomly divided into guilty and
innocent groups. At least 26 subjects were assigned
to the guilty group and performed the guilty scenario.
The other subjects were assigned to the innocent group
and performed the corresponding scenario. The items
presented to the subjects of both groups were identical.
A few test results were removed due to misdoing of the
protocol or inappropriately recorded signals. Finally,
23 guilty subjects and 21 innocent subjects were
chosen for use the subsequent data analyses.
CIT
Block 1 Block 2 Block 3 Block 4 Block 5
28 irrelevant,
7 control,
7 probe, 6 null.
Interstimulus inter
val 2.5 ± 0.2 sec
2.5 ± 0.2 sec
irr2 irr2 irr1 irr3irr4 control probe null
Resting
time
Resting
time
Resting
time
Resting
time
F i g. 1. The designed CIT. Control, probe, irr1, irr2, irr3, and irr4 are control stimulus, probe stimulus, and four irrelevant stimuli.
Р и с. 1. Схема організації тесту з прихованою інформацією.
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3 255
A CONCEALED INFORMATION TEST WITH COMBINATION OF ERP RECORDING
In the first step of the test, EEG electrodes,
electrooculogram (EOG) electrodes, and leads for
the peripheral measurements were attached, and the
subjects were instructed to attend to all presented items
and acknowledge recognition of the target picture by
right clicking the mouse, while left clicking for all
pictures meant that they were unable to recognize.
Physiological Recording. The procedure of
physiological recording took place in a silent
environment (laboratory) and in the absence of other
people.
The EEG data were recorded using eight active
Ag/AgCl electrodes with the gUSBamp system (G.Tec,
Austria). Electrodes were placed at Fz, Cz, Pz, Oz,
C4, C3, P4, and P3 sites according to the international
10-20 system and referenced to an electrode at the left
earlobe. For controlling eye movements, vertical and
horizontal EOGs were recorded. The EEG and EOG
data were digitized at 256 sec–1 and filtered online
using a 0.1-30 Hz bandpass and a 50-Hz notch filter.
The skin conductance changes were recorded by two
electrodes via an isolated amplifier (MLT 116F and
FE116, respectively; ADInstrument, Australia) with
low-voltage 75-Hz alternating current. Electrodes
were placed on the volar side of the middle phalanges
of the index and fourth fingers of the left hand.
The finger plethysmogram signal was recorded using
an infrared system in a spring clip (MLT1020F) via
an isolated amplifier (ML110; both by ADInstrument,
Australia) from the middle fingertip of the left hand.
The thoracic and abdominal respiratory activities
were recorded using two piezo respiratory belt
transducers (MLT1132, ADInstrument, Australia)
generating a voltage when there is a change in
the thoracic or abdominal circumference due to
respiration. All peripheral signals were digitized with
a sampling rate of 103 sec–1.
Data Analysis. After filtering the signals, we
separated each continuous record into single sweeps
according to the known onset times of the stimulus
presentation. The EOG data were checked for blink
artifacts by visual inspection, and sweeps with such
artifacts were removed. The ERPs for each type of the
stimuli (probe, or target, or irrelevant) were separately
extracted by averaging between related single sweeps.
The P300 peak-to-peak amplitude was measured. In
this measurement, a maximally positive segment
average of 100 msec was searched within a window
from 400 to 900 msec after the stimulus. The midpoint
of the maximum positivity defined the P300 latency.
After that, the algorithm searched for the maximum
100-msec-long negativity within the window from the
P300 latency to the end of the sweep. The difference
between the maximum positivity and negativity
defined the peak-to-peak measure [19].
The SCR is one of the slow responses of the
ANS. It has an onset latency and a rise time of 1 to
3 sec and a half-recovery time of up to 10 sec [20].
When stimuli are presented with short intervals, the
current response is influenced by previous responses.
In other words, the responses overlap. To overcome
the problem of overlapping in SCRs in our study, the
stimuli were presented in a 7th-order M-sequence,
as was mentioned above. The SCR was assessed
based on the averaging method proposed by Meijer
[14]. The epochs were extracted from –1 to 20 sec
relative to the stimulus onset and baseline corrected
at the sample preceding the stimulus onset. Within
each block, these epochs were averaged per stimulus
type. Since no picture was presented at presentation
of the null stimulus, an estimate of the bias produced
by the response to the previous stimuli was obtained
by averaging on this event [14]. Thus, by subtracting
it from other responses, we can assume that they are
unbiased. So, the average of the null stimulus was
subtracted from the average of each stimulus type in
each block.
The respiration line length (RLL) is a useful
a)
b)
c)
History Sequence Future
The 1 st A is preceded by
The 1 st B is preceded by
The 2 nd A is preceded by
The 2 nd B is preceded by
The 3 rd A is preceded by
The 3 rd B is preceded by
Bias on stimulus A
Bias on stimulus B
F i g. 2. The third-order M-sequence. a) Sequence of the stimuli,
b) bias on stimulus “A,” and c) bias on stimulus “B” are illustrated.
Р и с. 2. М-послідовність третього порядку.
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3256
E. D. FARAHANI and M. H. MORADI
measurement for detection of deception that integrates
information on the frequency and depth of respiration.
In our study, the RLL was automatically computed over
a time window from 600 to 3000 msec post-stimulus
onset; the method was derived from Timm [21, 22].
The data from abdominal and thoracic channels were
averaged.
The phasic heart rate (pHR) was calculated based
on the HR. The HR was defined based on the R-R
intervals of the ECG signal. The AC component of the
photoplethysmogram (PPG) pulse is synchronous with
the heart beat and, therefore, can also be a source of
HR information [23]. In our study, PPG peaks were
automatically detected based on an adaptive threshold
method [24]. Peak-to-peak intervals were transformed
into the HR and real-time scaled [25]. The HR during
the last second before the stimulus onset served as a
prestimulus baseline. The pHR values were defined
by subtracting this baseline value from each second-
per-second poststimulus value [11]. For extracting the
trial-wise information of pHR, the mean change in the
HR within 3 sec after the stimulus onset, compared
with the prestimulus baseline, was calculated [26].
In order to eliminate individual differences in
the responsiveness, physiological and behavioral
measures should be standardized [27]. Z-transformed
values were calculated for each subject and each
block. All probe and irrelevant trials of one block
(not including 14 trials of “history” and “future” in
each block) were used for calculation of individual
means and standard deviations [14]. In the detection
of the deception procedure, the difference between
responses to probe and irrelevant items is a basic
indicator. Thus, difference scores were calculated as
proposed by Gamer et al. [28]. In this method, the
difference score is defined as the difference between
the mean of the standardized probe trials and the mean
of all standardized irrelevant trials within each block.
Afterwards, the mean of five blocks was computed
as an overall index of the differential responsiveness
in each physiological or behavioral measure. These
values were used in subsequent statistical analyses.
Statist ical Analysis . The cross-correlat ion
coefficient is a reliable feature that is studied in the
detection of deception by means of a bootstrapped
correlation difference (BCD) method. The BCD
answers the question: “Are the cross-correlation
coefficients between ERP responses to probe and target
stimuli significantly greater than the corresponding
cross-correlation of responses to probe and irrelevant
stimuli?” If so, the subject is found to be guilty [5].
The statistical technique of bootstrapping [29] shows
the statistical significance of this hypothesis. In our
study, the BCD method was applied to the artifact-
free single sweeps as proposed by Abootalebi et al.
[5]. The output parameter of the BCD method (ND-0)
means that the probe response is more different from
the irrelevant and more similar to the target response;
thus, this subject is more likely to be guilty, and vice
versa. The ND-0 value determined for each subject was
used in the subsequent statistical analyses.
The standardized difference scores of autonomic
and ERP responses were compared between the
guilty and innocent groups using the ANOVA test.
The significance level for the assessment of main and
interaction effects was set to 0.05. The Cohen’s d was
calculated as an estimate of the effect size [30, 31].
Classification. To achieve applicable aspects of this
study, it is necessary that the subjects be classified
into two groups, innocent and guilty. In order to find
an optimized combination of the ERP and autonomic
measures, the discrimination performance of each
measure and combination of measures were evaluated
using two methods, a binomial logistic regression
model and linear discriminant analysis.
The logistic regression model is used extensively
in medical and social science fields as a basic method
for describing the relationship between a response
variable and one or more explanatory variables. The
goal of an analysis using the logistic regression model
is to find the best fitting and most parsimonious,
yet biologically reasonable, model to describe the
relationship between an outcome (dependent, or
response) variable and a set of independent (predictor,
or explanatory) variables [32].
Linear d iscr iminant analys is (LDA) is a
commonly used technique for data classification and
dimensionality reduction. This method maximizes
the ratio of the between-class variance to the within-
class variance in any particular data set, thereby
guaranteeing maximal separability [33, 34]. The
aim of LDA (also known as Fisher’s LDA) is to use
hyperplanes to separate the data representing different
classes. For a two-class problem, the class of a feature
vector depends on which side of the hyperplane the
vector is [35]. This technique is characterized by
very low computational requirements and generally
provides good results, which makes it suitable for
many pattern recognition problems.
The performance of the logistic regression model
and LDA was estimated using the leave-one-out cross-
validation method. Each subject, once excluded from
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3 257
A CONCEALED INFORMATION TEST WITH COMBINATION OF ERP RECORDING
the original data, was set as test data, and the other
subjects were used as training data. In each iteration,
the classifier was trained for using training subject data
and their real labels. After that, the held-out subject
was classified as guilty or innocent with the trained
classifier. Finally, the accuracy was calculated based
on the classification result of the held-out subjects.
RESULTS
The means and s.d. of standardized difference scores
of autonomic and ERP responses in the guilty and
innocent groups are summarized in Table 1.
Event-Related Potentials. All statistical analyses
were performed on the Pz channel, where the P300
amplitude is typically the largest. Figure 3 shows
grand means of the ERP waveforms for probe, target,
and irrelevant stimuli in the guilty and innocent
groups (A and B, respectively) for 1,000 msec after
the stimulus. As was expected, a large positivity was
elicited by the target stimuli but not by the irrelevant
stimuli. In the guilty group (A), probe responses
demonstrated similarity with target ones, while some
1000
1000
–4
–6
–2
–4
0
–2
2
0
4
2
2
2
3
1
1 3
6
4
8
msec
msec
µVµV
6
8
10
12
A B
200
200
300
300
400
400
500
500
600
600
700
700
800
800
900
900
1000
1000
F i g. 3. Grand means of the ERP waveforms for irrelevant stimuli (1), target stimuli (2) and probe stimuli (3) in the guilty (A) and innocent
(B) group. Abscissa) Time, msec; ordinate) amplitude, mV.
Р и с. 3. Усереднені пов’язані з подією потенціали при пред’явленні іррелевантних (1), цільових (2) та зондуючих (3) стимулів у
групах „винуватих” (А) та „невинуватих” (В) тестованих суб’єктів.
TABLE 1. Means and Standard Deviations (s.d.) of the Standardized Difference Score of Autonomic and ERP Measures in the Guilty
and Innocent Groups
Т а б л и ц я 1. Середні значення та стандартні відхилення стандартизованих різниць оцінок вегетативних показників та
параметрів пов’язаних з подією потенціалів у групах „винуватих” та „невинуватих” суб’єктів
Measures
Guilty group Innocent group
means s.d. means s.d.
P300 0.60 1.20 –0.05 0.81
BCD 75.65 22.91 23.38 16.41
pHR –0.18 0.24 0.08 0.21
SCR 0.43 0.86 –0.24 0.84
RLL –0.13 0.41 0.08 0.15
Footnote. P300 is the most important component of ERP; BCD is the bootstrapped correlation difference; PHR is the phasic heart rate;
SCR is the skin conductance response, and RLL is the respiration line length.
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3258
E. D. FARAHANI and M. H. MORADI
striking similarities between the probe and irrelevant
responses were observed in the innocent group (B).
The standardization procedure of P300 amplitude
was so different from the other measures. As was
explained, the ERP was extracted from all single
sweeps of each type of stimulus. Thus, since the
within-block standardization was impossible, the
within-subject standardization was calculated based
on the mean and s.d. of the P300 amplitude of probe
and irrelevant ERPs. ANOVA for the P300 amplitude
data showed that this parameter observed in the
guilty group was significantly greater than that in the
innocent group (F = 4.32; P < 0.043; d = 0.62).
As was explained, the percentage of the probe
responses, which was more similar to target than
to irrelevant ones, was computed as the BCD. In
the BCD, before computing the cross-correlation
coefficients, a time window was applied to single
sweeps between 300 and 900 msec after stimulus,
and correlation of the sweeps was only noticed in
this time-limited interval, because we expected that
the P300 would appear exclusively in this region
[5]. Only the correlation coefficient at lag = 0 was
considered. Since there is no individual difference in
the BCD measure, there is no need for standardization.
The ANOVA test between two groups showed that the
BCD measure was significantly greater in the guilty
group (F = 74.39, P < 0.001, d = 2.60). Since the
difference between the guilty and innocent groups in
the BCD was much more significant than that in the
P300 amplitude, only the BCD was evaluated as the
ERP measure in the subsequent analyses.
Autonomic Responses. To examine the statistical
distribution of autonomic responses in the guilty and
innocent groups, the box plot of the standardized
difference score of these responses was used (Fig. 4.).
ANOVA for the pHR data showed that these values
in the innocent group were significantly greater than
those in the guilty group (F = 15.04; P < 0.001;
d = –1.17; Fig. 4A).
ANOVA for the RLL data showed that these values
in the innocent group were significantly greater
than those in the guilty group (F = 5.13; P = 0.028;
d = –0.68; Fig. 4B). As can be seen in this figure,
the box plots are very compact, and there is good
discrimination between the guilty and innocent groups.
Assessment of the SCR using ANOVA showed
that the respective values in the guilty group were
significantly greater than those in the innocent group
(F = 6.95; P = 0.011; d = 0.79; Fig. 4C).
Logistic Regression Model. To compare the
discrimination performance of the ERP and the
autonomic measures, different combinations of these
measures were evaluated. Subjects were classified
as guilty or innocent based on a criterion P > 0.5 for
classification as guilty. Figure 5 shows the correct-
classification rates based on the leave-one-out cross-
validation method using BCD, SCR, pHR, RLL, and
other combinations of measures as predictors.
Evaluation of different independent measures shows
that BCD gives the best performance with 88.63%
correct-classification rate, and RLL with 79.54% is the
best autonomic measure.
As can be seen in Fig. 5 and Table 2, the best
correct-classification rate of different combinations
of measures is 88.63%, which was obtained in the
Innocent Innocent InnocentGuilty Guilty Guilty
–1.5
–1.0
–0.6
–0.8
1.0–0.5
–0.4 –0.5
–0.2
0.5
0
0
0
0.2
1.0
0.5
0.4
A B C
2.0
1.5
1.5
1.0
F i g. 4. Box plot of the standardized difference scores of autonomic measures in the guilty and innocent groups. A) Phasic heart rate (pHR),
B) respiration line length (RLL), and C) skin conductance response (SCR).
Р и с. 4. Бокс-діаграми стандартизованих різниць бальних оцінок вегетативних показників у групах „винуватих” та „невинуватих”
суб’єктів.
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3 259
A CONCEALED INFORMATION TEST WITH COMBINATION OF ERP RECORDING
tree combinations (pHR + BCD, RLL + BCD, and
RLL + SCR + pHR + BCD). None of the combination
of the ERP and the autonomic measure yielded the
incremental validity.
Linear Discriminant Analysis (LDA). Figure 6
shows the correct-classification rate of LDA based
on the leave-one-out cross-validation method using
BCD, SCR, pHR, RLL, and other combinations of
the measures as features. As can be seen in Fig. 6 and
Table 2, evaluation of different independent measures
shows that the BCD gives the best performance with
an 88.63% correct-classification rate, and the RLL at
79.54% is the best autonomic measure.
Evaluation of different combinations of the measures
using LDA shows that the best correct-classification
rate is 90.9% (Table 2). According to Fig. 6, the
incremental validity can be seen in two cases with a
90.9% correct-classification rate (pHR + SCR + BCD
and pHR + SCR + RLL + BCD).
DISCUSSION
In this study, a variant of the CIT based on
the simultaneous use of EEG and autonomic
measurements was designed. Recent publications
emphasized the need for studying the combination of
EEG and autonomic measurements in the detection of
TABLE 2. Best Correct-Classification Rates (Accuracy) of a Logistic Regression Model and Linear Discriminant Analysis Using
Event-Related Potential (ERP) Measures, Autonomic Measures, and Combination of the Measures
Т а б л и ц я 2. Найкращі рівні коректної класифікації (точності) для моделі логістичної регресії та лінійного
дискримінантного аналізу з використанням вимірів параметрів пов’язаних з подією потенціалів, вегетативних показників
та комбінації таких показників
Methods ERP measure, % Autonomic measures, % Combination of measures, %
Logistic regression model 88.63 79.54 88.63
Linear discriminant analysis 88.63 79.54 90.9
BCD pHR
pHR+
BCD
SCR
SCR+
BCD
SCR+
pHR
SCR+
pHR+
BCD
RLL
RLL+
BCD
RLL+
pHR
RLL+
pHR+
BCD
RLL+
SCR
RLL+
SCR+
BCD
RLL+
SCR+
pHR
RLL+
SCR+
pHR+
BCD
accuracy (%) 88.63 75.0 88.63 68.18 84.09 70.45 86.36 79.54 88.63 75.0 86.36 70.45 86.36 70.45 88.63
0
10
20
30
40
50
60
70
80
90
100
%
F i g. 5. Correct-classification rates (accuracy) for classification of the subjects as guilty or innocent using a logistic regression model
based on the leave-one-out cross-validation method. Bootstrapped correlation difference (BCD), phasic heart rate (pHR), skin conductance
response (SCR), respiration line length (RLL), and other combinations of the measures are used as predictors.
Р и с. 5. Рівні коректної класифікації (точності) при поділі тестованих cуб’єктів на „винуватих” та „невинуватих” з використанням
моделі логістичної регресії.
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3260
E. D. FARAHANI and M. H. MORADI
deception [11, 15-17]. The purpose of our study was
to evaluate the CIT accuracy enhancement caused by
the combination of ERP and autonomic measurements.
The second purpose was likening of the experimental
conditions to real conditions. So, the CIT was tested
on subjects who performed under the innocent and
guilty scenarios.
Statistical analyses (summarized in Table 1) showed
that both in the ERP and autonomic measures, there
is a significant difference between the standardized
difference scores in the guilty and innocent groups.
Due to the use of short ISIs, this significant difference
was predictable for the ERP measure, while the
significant difference for the autonomic measures
means that the considered arrangements allowing us to
solve the problems of habituation and overlapping have
been successful. These arrangements included the five-
block CIT, variable ISIs, and stimulus presentation in
the balanced order (using M-sequence series). The
five-block CIT and variable ISIs were utilized to solve
the problem of habituation; stimulus presentation in
the balanced order was used to overcome overlapping
in the SCRs.
The difference between innocent and guilty groups
in the BCD is significantly greater than that in the P300
amplitude. Figure 3 shows that there are some striking
similarities between the probe and irrelevant responses
in the innocent group in contrast to the guilty group.
This result is consistent with our expectation and also
with the previous ERP-based CIT studies [6, 8, 36].
Statist ical analyses of autonomic measures
show that there are significant differences between
standardized difference score in the guilty and
innocent groups. More detailed examination shows
that the pHR has the best efficiency among autonomic
measures, while most autonomic-based studies have
introduced the SCR as the best measurement for
detection of deception. The lower significance of the
SCR might be due to overlapping caused by short ISIs.
Although the characteristic of M-sequence is used to
solve the problem of overlapping SCRs, it seems that
it is necessary to use much more powerful methods for
decomposing overlapping responses, e.g., such as the
method proposed by Lim et al. [37].
Two types of classification methods were employed.
Furthermore, all possible combinations of features
were classified in order to examine their interactions.
In both methods of classification, the best accuracy
was achieved from the BCD and RLL. The incremental
validity from the combination of measures was
obtained using the LDA classification method,
while no combination of the brain and autonomic
measures yielded the incremental validity in logistic
regression. The best results of different classification
methods in different measures are summarized in
Table 2. The absence of incremental validity in the
BCD pHR
pHR+
BCD
SCR
SCR+
BCD
SCR+
pHR
SCR+
pHR+
BCD
RLL
RLL+
BCD
RLL+
pHR
RLL+
pHR+
BCD
RLL+
SCR
RLL+
SCR+
BCD
RLL+
SCR+
pHR
RLL+
SCR+
pHR+
BCD
accuracy (%) 88.63 75.0 86.36 70.45 86.36 70.45 90.9 79.54 88.63 77.27 86.36 68.18 86.36 70.45 90.9
0
10
20
30
40
50
60
70
80
90
100
%
F i g. 6. Correct-classification rates (accuracy) for classification of the subjects as guilty or innocent using linear discriminant analysis
based on the leave-one-out cross-validation method. Bootstrapped correlation difference (BCD), phasic heart rate (pHR), skin conductance
response (SCR), respiration line length (RLL), and other combinations of the measures are used as features.
Р и с. 6. Рівні коректної класифікації (точності) при поділі тестованих cуб’єктів на „винуватих” та „невинуватих” з використанням
лінійного дискримінантного аналізу.
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3 261
A CONCEALED INFORMATION TEST WITH COMBINATION OF ERP RECORDING
logistic regression method might be due to the type
of this classification technique. Different methods
of classification according to the assumptions and
the rules used in their design might show different
performances. This issue was observed in a review
of classifiers in this study. Thus, the hypothesis of
accuracy enhancement of the CIT combined with the
ERP and autonomic measures can be validated.
Matsuda et al. [15] designed a CIT with simultaneous
measurements of autonomic and brain signals. The
cited authors used auditory stimulation with long ISIs
(22 sec). The use of long ISIs causes no overlap
between sequential autonomic responses, and these
measures showed a significant difference between
critical and noncritical items. The P300 amplitude did
not show significant differences, which is probably due
to the use of long ISIs and a low number of stimulus
presentations. Gamer and Berti [16] performed a
similar study sometime later. They designed a CIT with
the combined measurement and ISIs of 7 to 9 sec and
tried to improve the P300 amplitude discrimination
using an increase in the frequency (and number) of
stimulus presentations, but their results were similar
to those reported by Matsuda et al. [15]. As was
already mentioned, using short ISIs and considering
some arrangements to solve the problems of autonomic
measures, both the ERP and autonomic measures
showed a significant difference in our study.
In a CIT study with multimodal measurements,
Ambach et al. [11] used short ISIs (3.0-3.5 sec).
As compared with similar studies, the cited authors
reported that there is a significant difference between
the probe and irrelevant items in the ERP and
autonomic measures but with a smaller effect size.
These researchers discussed the absence of a criminal
context in the mock crime instruction and complete
omission of answers, which led to the diminished
involvement and attention of subjects as a possible
reason for rather small effect sizes. They reported a
0.738 correct-classification rate by only P300 and
a 0.829 correct-classification rate with combining
P300 and SCR based on the logistic regression model
between the guilty and hypothetical innocents. In
our study, a higher effect size and a higher correct-
classification rate were obtained. Most likely, the
reason for this might be the subject’s greater attention
and stronger involvement in the experiment. The use
of target stimuli and the answering of all presented
items by the subject (via mouse click) prevented
reducing the subject’s attention. Furthermore, the use
of a criminal context in the mock crime instruction and
the reward-punishment system (winning or losing the
gold coin) have made the subjects more actively pay
attention and get involved in the experiment.
Thus, our study showed that, both in the ERP and
autonomic measures, a significant difference between
the standardized difference scores of the guilty and
innocent groups can be achieved using short ISIs
and large numbers of stimulus presentations with
the consideration of some arrangements. These
arrangements include several blocks and variable ISIs
to solve the problem of habituation, and also stimulus
presentation in the balanced order (using M-sequence
series) to solve the overlapping problem of the SCR.
Furthermore, the criminal context in the mock crime
instruction and reward-punishment system were used
to make subjects to increase attention and get more
involved in the experiment.
Finally, the hypothesis of accuracy enhancement
of the CIT combined with the ERP and autonomic
measures was confirmed by a review of two
classification methods. To reach such a result, we
tried to make experimental conditions closer to real
conditions as much as possible using the criminal
context in the mock crime instruction and real innocent
subjects instead of hypothetical innocent ones.
Our own data show that the expediency of further
studies is obvious. First, slightly longer ISIs should
be used in order to elicit greater autonomic responses
and to allow experimenters to use longer scoring
intervals. Second, the method of decomposing the SCR
in paradigms with short ISIs should be used [37, 38].
As a concluding suggestion, the use of more powerful
classification methods, such as a support vector
machine, might be useful for further studies.
All tested subjects were volunteers; they were informed
in detail on the pattern of the experiment and gave informed
consent.
The authors, E. D. Farahani and M. H. Moradi, declare that they
have no conflict of interests.
Acknowledgment. This research was supported by the Iran
National Science Foundation (INSF). The authors would like to
thank them and also all those who took part in this study.
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3262
E. D. FARAHANI and M. H. MORADI
Е. Д. Фарахані1, M. Мораді1
ТЕСТ ІЗ ПРИХОВАНОЮ ІНФОРМАЦІЄЮ „ВИНУВА-
ТИЙ/НЕВИНУВАТИЙ” У ПОЄДНАННІ З РЕЄСТРАЦІЄЮ
ПОВ’ЯЗАНИХ З ПОДІЄЮ ПОТЕНЦІАЛІВ І
ВИМІРЮВАННЯМ ВЕГЕТАТИВНИХ ПОКАЗНИКІВ
1 Технологічний університет Аміркабір, Тегеран (Іран).
Р е з ю м е
Тест із прихованою інформацією (ТПІ) базується на порів-
нянні фізіологічних реакцій суб’єкта в зондуючих і ней-
тральних ситуаціях, спрямованому на виявлення такої
інформації. У нашій роботі ми визначали точність резуль-
татів ТПІ в умовах, коли цей тест поєднували з відведен-
ням пов’язаних з подією потенціалів (ППП) і вимірюванням
вегетативних показників. Особливістю нашого досліджен-
ня було максимальне наближення експериментальних умов
до реальних за допомогою використання кримінального
контексту в інструкції „макетування злочину” та „реаль-
но невинуватих” суб’єктів замість „гіпотетично невинува-
тих”. 52 волонтери виконували один із сценаріїв „винува-
тий/невинуватий”. ТПІ вміщував п’ять блоків з короткими
міжстимульними інтервалами. У кожному з блоків стиму-
ли пред’являлись у балансованій послідовності сьомого
порядку. Крім ЕЕГ-активності, реєстрували частоту пуль-
су, зміни шкірної провідності (ШП), дихальну активність
і плетизмограму пальців. Статистичний аналіз показав, що
між стандартизованими оцінками відмінностей характерис-
тик як ППП, так і вегетативних показників у „винуватій”
та „невинуватій” групах виявлялися вірогідні відміннос-
ті. При виявленні змін ШП очікувані результати, описані
для стандартних результатів ТПІ, що базуються на вимірю-
вання вегетативних показників, не досягалися. Порівняння
двох класифікаційних методик показало, що поєднання від-
ведення ППП і вегетативних вимірювань підвищує точність
результатів ТПІ. Найбільша точність класифікації, отрима-
ної із застосуванням лінійного дискримінантного аналізу,
складала 90.9 %. Скоріш за все, використання кримінально-
го контексту в інструкції „макетування злочину” і системи
преміювання/покарання забезпечувало більший рівень ува-
ги тестованих та їх більше залучення в експеримент, що й
підвищувало точність тестування порівняно з такою в ана-
логічних дослідженнях.
REFERENCES
1. D. T. Lykken, A Tremor in the Blood: Uses and Abuses of the
Lie Detector, McGraw-Hill Book Comp., New York (1981).
2. J . A. Matte, Forensic Psychophysiology Using the
Polygraph: Scientific Truth Verification – Lie Detection,
J. A. M. Publ., New York (1996).
3. E. B. Ford, “Lie detection: Historical, neuropsychiatric and
legal dimensions,” Int. J. Law. Psychiat., 29, 159-177 (2006).
4. National Research Council, The Polygraph and Lie Detection,
Committee to Review the Scientific Evidence on the
Polygraph, Division of Behavioral and Social Sciences and
Education, The Nat. Acad. Press, Washington (2003).
5. V. Abootalebi, M. H. Moradi, and M. A. Khalilzadeh, “A
comparison of methods for ERP assessment in a P300-based
GKT,” Int. J. Psychophysiol., 62, 309-320 (2006).
6. J. P. Rosenfeld, “Event-related potentials in the detection of
deception, malingering, and false memories,” in: Handbook
of Polygraph Testing , M. Kleiner (ed.), Acad. Press,
New York (2002), pp. 265-286.
7. J. P. Rosenfeld, B. Cantwell, V. T. Nasman, et al., “A modified,
event-related potential-based guilty knowledge test,” Int.
J. Neurosci., 42, 157-161 (1988).
8. L. A. Farwell and E. Donchin, “The truth will out: inter-
rogative polygraphy (lie detection) with event-related
potentials,” Psychophysiology, 28, 531-547 (1991).
9. D. T. Lykken, “The GSR in the detection of guilt,” J. Appl.
Psychol., 43, 385-388 (1959).
10. G. Ben-Shakhar and E. Elaad, “The val id i ty of
psychophysiological detection of information with the Guilty
Knowledge Test: a meta-analytic review,” J. Appl. Psychol.,
88, 131-151 (2003).
11. W. Ambach, R. Stark, M. Peper, and D. Vaitl, “A concealed
information test with multimodal measurement,” Int.
J. Psychophysiol., 75, 258-267 (2010).
12. J. Polich, “P300, probability, and interstimulus interval,”
Psychophysiology, 27, No. 4, 396-403 (1990).
13. R. J. Barry, S. Feldmann, E. Gordon, and K. I. Cocker,
“Elicitation and habituation of the electrodermal orienting
response in a short interstimulus interval paradigm,” Int.
J. Psychophysiol., 15, 247-253 (1993).
14. E. H. Meijer , Psychophysiology and the Detection of
Deception: Promises and Perils, Unpublished Doctoral
Dissertation, Maastricht Univ., the Netherlands (2008).
15. I. Matsuda, H. Nittono, A. Hirota, et al., “Event-related brain
potentials during the standard autonomic-based concealed
information test,” Int. J. Psychophysiol., 74, 58-68 (2009).
16. M. Gamer and S. Berti, “Task relevance and recognition
of concealed information have different influence on
electrodermal activity and event-related brain potentials,”
Psychophysiology, 47, 355-364 (2010).
17. I. Matsuda, H. Nittono, and T. Ogawa, “Event-related potentials
increase the discrimination performance of the autonomic-
based concealed information test,” Psychophysiology, 48,
1701-1710 (2011).
18. G. T. Buracas and G. M. Boynton, “Efficient design of event-
related fMRI experiments using M-sequences,” Neuroimage,
16, 801-813 (2002).
19. M. Soskins, J. P. Rosenfeld, and T. Niendamm, “Peak-to-
peak measurement of P300 recorded at 0.3 Hz high pass filter
settings in intraindividual diagnosis: complex vs. simple
paradigms,” Int. J. Psychophysiol., 40, 173-180 (2001).
20. M. E. Dawson, A. M. Schell, and D. L. Filion, “The
electrodermal system,” in: Handbook of Psychophysiology,
J. T. Cacioppo, L. G. Tassinary, and G. G. Berntson (eds.),
Cambridge Univ. Press (2000), pp. 200-223.
21. H. W. Timm, “Effect of altered outcome expectancies
stemming from placebo and feedback treatments on the
validity of the guilty knowledge technique,” J. Appl.
Psychol., 67, 391-400 (1982).
22. H. W. Timm, “Analyzing deception from respiration patterns,”
J. Police Sci. Adm., 10, 47-51 (1982).
NEUROPHYSIOLOGY / НЕЙРОФИЗИОЛОГИЯ.—2013.—T. 45, № 3 263
A CONCEALED INFORMATION TEST WITH COMBINATION OF ERP RECORDING
23. J. Allen, “Photoplethysmography and its application in
clinical physiological measurement,” Physiol. Meas., 28,
R1-R39 (2007).
24. H. S. Shin, C. Lee, and M. Lee, “Adaptive threshold
method for the peak detection of photoplethysmographic
waveform,” Comput. Biol. Med., 39, 1145-1152 (2009).
25. M. Velden and C. Wӧlk, “Depicting cardiac activity over real
time: a proposal for standardization,” J. Psychophysiol., 1,
173-175 (1987).
26. M. T. Bradley and M. P. Janisse, “Accuracy demonstrations,
threat, and the detection of deception: cardiovascular,
electrodermal, and pupillary measures,” Psychophysiology, 18,
307-315 (1981).
27. G. Ben-Shakhar, “Standardization within individuals: a
simple method to neutralize individual differences in skin
conductance,” Psychophysiology, 22, 292-299 (1985).
28. M. Gamer, H. G. Rill, G. Vossel, and H. W. Gӧdert,
“Psychophysiological and vocal measures in the detection of
guilty knowledge,” Int. J. Psychophysiol., 60, 76-87 (2006).
29. S. Wasserman and U. Bockenholt , “Bootstrapping:
applications to psychophysiology,” Psychophysiology, 26,
208-221 (1989).
30. J. Cohen, Statistical Power Analysis for the Behavioral
Sciences, Lawrence Erlbaum Associates, Inc., Hillsdale (1988).
31. R. L. Rosnow and R. Rosenthal, “Computing contrasts, effect
sizes, and counternulls on other people’s published data:
general procedures for research consumers,” Psychol. Methods,
1, 331-340 (1996).
32. D. W. Hosmer and S. Lemeshow, Applied Logistic Regression,
John Wiley & Sons, Hoboken (2000).
33. K. Fukunaga, Statistical Pattern Recognition, Acad. Press,
New York (1990).
34. R. O. Duda, P. E. Hart, and D. G. Stork, Pattern Recognition,
Wiley-Interscience, New York (2001).
35. V. Abootalebi, M. H. Moradi, and M. A. Khalilzadeh, “A
new approach for EEG feature extraction in P300-based
lie detection,” Comput. Methods Programs Biomed., 94,
No. 1, 48-57 (2009).
36. J. Allen, W. G. Iacono, and K. D. Danielson, “The identification
of concealed memories using the event-related potential and
implicit behavioral measures: a methodology for prediction
in the face of individual differences,” Psychophysiology, 29,
504-522 (1992).
37. C. L. Lim, C. Rennie, R. J. Barry, at al., “Decomposing
skin conductance into tonic and phasic components,” Int.
J. Psychophysiol., 25, 97-109 (1997).
38. W. Ambach, R. Stark, M. Peper, and D. Vaitl, “Separating
deceptive and orienting components in a Concealed
Information Test,” Int. J. Psychophysiol., 70, 95-104 (2008).
|