DETECTING DECEPTWE COMMUNICATION

FROM VERBAL VISUAL AND
PARALINGUISTICCUES .

' AN 'EXPLORATORY EXPERIMEN? f», .1- : r,

E'bSDf film the Deggen 3f PAID
NEH STATE uI‘l VERSE}? _
- ,OHN EMA-0 r536 AA ‘
1378 I

Tv—."VI

 
    
    

  

, LIBRARY "
Michigan Straw
I University

This is to certify that the

thesis entitled

DETECTING DECEPTIVE COMMUNICATION FROM VERBAL,
VISUAL AND PARALINGUISTIC CUES:

AN EXPLORATORY EXPERIMENT
presented by

John Edward Hocking
has been accepted towards fulfillment

of the requirements for

Ph . D . degree in Communication

 

Major professor

Date April 234 1976

   
   

a! 80'!
BINDERY

Unit” a, moans
mum”? a :cmnl

 

do / a") 7

ABSTRACT
DETECTING DECEPTIVE COMMUNICATION FROM VERBAL, VISUAL AND

PARALINGUISTIC CUES: AN EXPLORATORY EXPERIMENT

BY

John Edward Hocking

This experiment explored how differences in the
verbal and nonverbal behaviors available to persons influenced
their ability to detect deception. Earlier research suggested
that observers would be more accurate in judging veracity when
the judgments were based on observations of persons' bodies
than when based on observations of persons' heads (Ekman &
Friesen, 1974). These persons were making true or false state—
ments about their personal emotional feelings as they watched
stressful films. The purpose of the present experiment was
to extend and refine this research by: (l) examining observer
accuracy in a situation in which persons made true or false
statements about a factual event, in addition to replicating
the situation in which they concealed their emotional respon-
ses; and (2) to present observers with additional categories
of information, both in isolation and in combinations, and to
compare any resulting accuracy differences.

Samples of both lying and "truthing" behavior were

videotaped under conditions of high saliency for 16 subjects.

John Edward Hocking
These tapes included eight subjects who were making true or
false statements about the factual content of an event they
had seen and eight subjects were making true or false state-
ments about their emotional feelings as they watched stress-
ful slides. These tapes were shown to 730 male and female
undergraduate college student observers under 14 experimental
conditions. Observers saw either a close-up shot of the
subject's head, a shot of the head and body with the head
blocked from view, or a shot of the head and body. They
viewed the tapes in either a color or a black-and-white tele-
vision format. They viewed the tapes with the audio compo-
nent of the 16 subjects' behavior either present or absent.
These variables created a three (head-only/body-only/head-
and-body) by two (color/black-anddwhite) by two (audio-and-
visual/visual-only) factorial design. Audio-only and
transcript-only conditions brought the total number of cells
in the design to 14. Observers dichotomously judged the
veracity of each of the 16 subjects and also indicated their
degree of confidence in the accuracy of each judgment.

For the segments in which subjects made true or
false statements about a factual event, results indicated
greater accuracy for head-only observers than for either the
body-only or the head-and-body observers, who did not differ
from each other. Audio-and-visual observers were more
accurate than were observers who based their judgments on

visual information only. Transcript-only and audio-only

John Edward Hocking
accuracy scores did not differ although both were greater
than the visual only accuracy scores.

For the segments in which subjects made true or false
statements about their emotional feelings, body-only
observers were more accurate than head-only observers with those
in the head-and-body condition falling in between. There was
no accuracy difference between the visual-and-audio observers
and the visual-only observers.

For the head-only conditions of both segments there
was a trend in favor of color resulting in higher accuracy
scores than black-and-white. Also for both segments, observers
who received audio and visual information were more confident
of their judgments of veracity than those who received only
visual information. Head-only observers were more confident
than head-and-body observers, while body-only observers were
the least confident. The transcript and audio-only observers
were more confident than were the visual-only observers.

The Ekman and Friesen (1974) finding was thus repli-
cated when the deceptive communication involved statements
about emotional feelings but not when it involved statements
about a factual event. Implications and qualifications are
discussed. Future research is suggested.

Ekman, P. & Friesen, W. Detecting deception from the body

and face. Journal of Personality and Social Psychology,
1974, 29, 288-298.

DETECTING DECEPTIVE COMMUNICATION FROM VERBAL, VISUAL AND

PARALINGUISTIC CUES: AN EXPLORATORY EXPERIMENT

BY

John Edward Hocking

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY

Department of Communication

1976

Accepted by the faculty of the Department of
Communication, College of Communication Arts, Michigan
State University, in partial fulfillment of the require-

ments for the Doctor of Philosophy degree.

Jemfiw

Directoribf Dissertation

Guidance Committee: , Chairman

 

 

41 @I’Dfiiireg
M 3/23

ACKNOWLE DGMEN TS

Dr. Gerald Miller provided firm intellectual, moral,
and financial support for this project. It has been a
privilege to observe and work with this outstanding scholar
during my three years at Michigan State. Thanks GR!

Dr. Edward L. Fink went way beyond the call of duty
in contributing to whatever excellence may be perceived in
.this paper. I am grateful for his selfless efforts. Drs.
Erwin Bettinghaus, David Ralph, and Michael Dowdle rounded
out my guidance and dissertation committees. They have been
very helpful on this project, as well as throughout my gradu-
ate career at Michigan State.

Joyce Bauchner was a trusted research colleague and
friend throughout this project. Joyce, Edmond Kaminski,
and Mark Steinberg made up the research team which ran this
study. It could not have been completed without the loyal
efforts of these fine people.

My friend and mentor, Dr. Cal Hylton, had little to
do with this dissertation. He is, however, very much
responsible for its existence. His influence on my career,

and my life, is pervasive.

ii

Finally, my parents, Julia and Ernest Hocking,
deserve the biggest thanks of all. The love and devotion
of these wonderful people is the most valuable thing I have.
It would be impossible to overstate my gratitude.

I have been very, very luckytx>have been associated
with each of these people. To them, to the many instructors
and students who assisted and participated in this research,
and to my friends at Michigan State and Georgia, my sincere

appreciation and thanks.

iii

LIST OF

LIST OF

CHAPTER

I.

II.

III.

IV.

APPENDIX A.

APPENDIX B.

APPENDIX C.

TABLE OF CONTENTS

TABLE S O O O O O O O O O O O O

FIGURES O O O O O O O O O O 0

INTRODUCTION . . . . . . . . .

The Problem . . . . . . . .
Some Definitions . . . . . .
The Deception Literature . .
Summary and Research Design

“THOD O O O O O O O O O O O 0

Overview . . . . . . . . . .
Creating the Stimulus Tapes
Collecting the Data . . . .

MSULTS O O O C O O O O O O .

Tape . . . . . . . . . . . .
Accuracy . . . . . . . . . .
Confidence . . . . . . . . .

DI SCUSS ION O O O O O O O O O 0

Accuracy . . . . . . . .
Confidence . . . . . . .
Problems with the Study
Future Research . . . .
Summary . . . . . . . .

Letter . . . . . . . .

iv

Subject Participation Invitation

Interview Format Review Sheet

Subjects Post-Interview Questionnaire

PAGE

vii

11
29

39
39
59
67
69
75
86
97
97
109
112
118
123
128
129

130

TABLE OF CONTENTS (continued)

PAGE
APPENDIX D. Guidelines for Subjects' Debriefing . . 131
APPENDIX E. Observer Questionnaire . . . . . . . . 132
APPENDIX F. Observer's Results Letter . . . . . . . 140
APPENDIX G. Analysis of Variance for Factual and
Emotional Accuracy Using the Trans-
formed Scores . . . . . . . . . . . . 142
APPENDIX H. Correlations Between Confidence and
Accuracy for Both the Factual and
Emotional Segments . . . . . . . . . 144

EFEMNCES O O O O O O O O O O O O O O O O O O O O C 145

TABLE

1.

10.

11.

12.

LIST OF TABLES

Method Used to Create Lying and Truthful
Stimuli Materials in All Studies
ReViewed O O C O O O O O O O O O O O O O 0

Sample Size in Each Cell of the Design . . .

Subject Sex, Actual Veracity, Number of
Judgments of "True" and "False," and Mean
Accuracy and Confidence Scores for Both
Factual and Emotional Segments on Each
Tape, and Summed . . . . . . . . . . . . .

Cell Mean Accuracy Scores, Variances, and
Marginal Means for the Factual Segments .

ANOVA for Factual Accuracy . . . . . . . . .

Cell Mean Accuracy Scores, Variances, and
Marginal Means for the Emotional
segments 0 I I O O O O O O O O O O O O O 0

ANOVA for Emotional Accuracy . . . . . . . .

Cell Means, Sample Sizes, and Marginal
Means for the Two-Way Interaction Between
Shot and Tape on Factual Accuracy . . . .

Cell Means, Sample Sizes, and Marginal
Means for the Three-Way Interaction of _‘
Visaudio, Shot and Tape on Factual
Accuracy . . . . . . . . . . . . . . . . .

Cell Means, Sample Sizes, and Marginal
Means for the Two-Way Interaction Between
Shot and Tape on Emotional Accuracy . . .

Cell Means, Sample Sizes, and Marginal
Means for the Two-Way Interaction Between
Visaudio and Tape on Factual Accuracy . .

Mean Confidence Scores, Variances, and
Marginal Means for the Factual Segments .

vi

PAGE

17

60

70

71
72

73
74

76

77

80

87

LIST OF TABLES (continued)

TABLE
l3.
14.

15.

16.

17.
18.
19.

20.

PAGE
ANOVA for Factual Confidence . . . . . . . . . 88

Cell Mean Confidence Scores, Variances, and
Marginal Means for the Emotional Segments . 89

ANOVA for Emotional Confidence . . . . . . . . 90

Comparison of the Accuracy Means and Sample
Sizes for Honest and Deceptive Samples of
the Ekman and Friesen (1974) Experiment
with Those From the Present Experiment . . . 126

ANOVA for Factual Accuracy . . . . . . . . . . 142
ANOVA for Emotional Accuracy . . . . . . . . . 143
Correlations Between Accuracy and Confidence
and Probability Levels in Each Cell of the
Design for the Factual Segments . . . . . . 144
Correlations Between Accuracy and Confidence,

and Probability Levels in Each Cell of the
Design for the Emotional Segments . . . . . 145

vii

FIGURE

LIST OF FIGURES

PAGE

The Experimental Design . . . . . . . . . . 33
Room Arrangement During the Stimulus Taping 52
Two-Way Interactions Between Shot and Tape

on Factual Accuracy . . . . . . . . . . . 76
Three-Way Interaction Between Visaudio, Shot

and Tape on Factual Accuracy . . . . . . 78
Two-Way Interaction Between Shot and Tape on

Emotional Accuracy . . . . . . . . . . . 80
Two-Way Interaction Between Visaudio and

Tape on Factual Accuracy . . . . . . . . 82
Three-Way Interaction Between Visaudio, Shot

and Tape on Factual Accuracy . . . . . . 83

viii

Chapter I

INTRODUCTION

‘Lying during social interaction is relatively common
in our culture. The lies may vary in seriousness from the
most innocuous compliment ("white lies") to falsehoods about
unambiguous matters of fact which have important consequences
for both liar and victim. Write Wolk and Henley (1970),
"EVERYONE lies. And the person who denies that he lies is
the most egregious liar of all." (p. 1).

Given the ubiquitous nature of deceptive communica-
tion and the important role it can play in the development,
maintenance, and termination of human relationships, it is
not surprising that people have long been concerned, even
fascinated, with detecting lying. Prescriptive advice about
how to detect lying has existed from as early as 900 B.C.
when a papyrus Vedas described the behavior of liars?

He does not answer questions, or they are evasive
answers; he speaks nonsense, rubs the great toe
along the ground, and shivers; his face is dis-
colored; he rubs the roots of his hair with his
fingers; . . . (Trovillo, 1939, p. 849)
K From the time of Christ, through the middle ages,
deception was detected with the method of the ordeal in which

the accused liar would have to survive various forms of cruel

1

2
torture to prove his innocence. Even today, with suspicions
about "shifty-eyed" criminals, there remains a considerable
folklore about the behavior of liars.

Considering the pervasiveness of deceptive communi—
cation, the historical interest, and the potential importance
of detecting deception, we might expect an extensive body of
extant knowledge about this phenomenon. On the contrary,
little is known about lying behaviors; what they are; how
(and if) they can be detected. Knapp, Hart, and Dennis (1974)
recently noted that, "At this point in time, almost any sys-
tematic study of deception must be labeled exploratory"

(p. 16).)

Research using mechanical devices to identify physio-
logical differences between liars and "truthers" began as
early as 1897 (Lee, 1953). This research, motivated primarily
by law enforcement needs, has continued to the present day and
provides ample documentation that the physiological responses
of individuals who are lying usually differ from those who are
telling the truth (Cutrow, Parks, Lucas and Thomas, 1972).
However, mechanical detection of lying has many obvious limi—
tations and is applicable only under highly controlled con-
ditions. Social science has made a belated entry into study-
ing the detection of lying and has focused on observable
behaviors which relate to deception. The research reported
in this p352: attempts to add to this small but growing body
of literature on the detection of deceptive communication

from observable behaviors: Specifically, the experiment

3

reported here was designed to answer the question: How do If

R
l

to persons influence their ability to detect deception? «w

differences in the verbal and nonverbal behaviors available

The remainder of this chapter describes the problem
in more detail, explicates some key terms, and critically
reviews previous deception literature. The chapter concludes
with an overview of the experimental design and a discussion
of why and how detection accuracy could vary across the

experimental conditions.

The Problem

 

Little is known about what observable behaviors are
associated with lying. Even less is known about the extent
to which observer sensitivity to any behavioral differences
between liars and "truthers" allows accurate discrimination
between the two. Paul Ekman and Wallace Friesen recently A
began investigating these questions.

I Ekman and Friesen (1969a) develop a theory of the
relationship between the nonverbal information "sending
capacity" of parts of the body and nonverbal clues to decep-
tion. The face, they argue, has a much greater sending
capacity than other parts of the body because the facial
muscles allow for a larger number of discriminable stimuli
patterns. The feet and legs, on the other hand, are the
worst information senders, the number of discernible stimuli
patterns which can be emitted being far less. The hands are

intermediate between the face and feet or legs.) The degree

4

of visibility that the parts of the body have to others in
social interaction, say Ekman and Friesen (1969), correspond
to their sending capacity. The face is typically more
visible during interaction than the hands which in turn are
more visible than the feet or legs. "External feedback," they
argue, parallels the sending capacity and degree of visibility
of various parts of the body. They define external feedback
as responses by other peOple to the focal individual's non-
verbal behavior. Most external feedback occurs in response to
the face. Individuals are more willing to hold others
(responsible for, and comment on, information which comes from
the face. Less external feedback is directed at the hands and
even less at the feet. "Internal feedback," that is, the
individual's conscious awareness of his nonverbal behaviors
from various parts of the body, also parallels sending
capacity, visibility, and external feedback. "People have the
greatest internal feedback about their face, next most about
their hands and least about their legs and feet" (Ekman and
Friesen, 1969a, p. 96).

(Since peOple receive less external feedback about
information which is conveyed from the feet and legs and
since they monitor the behavior of their feet and legs less
closely, Ekman and Friesen (1969a) reason that the feet and
legs will be the greatest source of deception clues, with the
hands next, and the face least likely to provide information
on which judgments of deception may be accurately based.)

Simply put, people are more aware of their face than they are

5

of the rest of their body. They are thus better able to con-
trol their facial behavior during lying. Based on this
reasoning, Ekman and Friesen (1969a) hypothesize that obser-
vers should be able to detect deceptive communication more
accurately when they observe the body than when they observe
the head. I

( Ekman and Friesen (1974) tested this hypothesis by
showing videotapes of subjects who were lying or telling the
truth in response to questions about their feelings as they
watched pleasant and stressful films. Observers were shown
videotapes of the subjects with either a close-up shot of the
head only or a full shot of the head and body with the head
portion blocked from View.) Audio portions of the subjects'
behavior were not included on these tapes. Results supported
the prediction that observers will be more accurate in
detecting deception when they view the body only than when
they View the head only. Ekman and Friesen (1974) took the
most obvious first step in terms of partialling categories of
information on which judgments of deceptionmay be based.
Their study, however, has left many questions unanswered.

What categories of information provide the best cues

on which to judge veracity when the lying involves a factual
event rather than the concealment of emotional responses? The
Ekman and Friesen (1974) experiment deals only with subjects’
false statements about their personal feelings as they watch
stressful or pleasant films. Cues which were emitted by

subjects may have been in response to these films themselves

6

and not a result of their false verbal statements about their
feelings. It is unknown if the Ekman and Friesen (1974) head-
body finding would replicate in other kinds of dishonest
interactions.

How accurate would observers be if they were provided
with both the head and body for observations? If the head
had been available in Ekman and Friesen's "body only" con-
dition, it is likely that observers would have watched it at
least some of the time, since it is natural to look at some-
one's head when they are talking. This, in turn, might have
distracted viewers from paying attention to deception cues
which were being emitted by the body. In "real life" decep-
tion situations the head is obviously available for observa-
tion. It is unknown what categories of information observers
use to accurately infer deception when the head is available.

To what extent does verbal and paralinguistic informa-
tion provide cues on which accurate judgments of veracity are
based? The audio portion of the lying and truthful behavior
of subjects in the Ekman and Friesen (1974) experiment was
omitted. The audio portion contains both the verbal content
of what is said and paralinguistic cues, which are also
audible. Several researchers (e.g., Mehrabian, 1971; Knapp,
Hart, and Dennis, 1974) have conducted content analyses of the
behavior of lying and truthing individuals and report differ-
ences in the verbal content of their speech. Differences 3
between the paralinguistic speech characteristics of lying I

i
I

and truthing individuals have also been demonstrated (e.g.,

7
English, 1926; Mehrabian, 1971). Research which examines the
behavioral correlates of lying does not directly indicate
what behaviors observers are using to make accurate inferences
of veracity. Differences between the behavior of liars and
"truthers" are useful to the observer attempting to detect
lying only if the behavior is noticed and correctly attributed
to the veracity of the source. Conceivably there could be
many observer "cues" indicative of deception which were either
missed entirely by observers, or noticed and not correctly
identified as being associated with lying. Research which
investigates behavioral correlates of lying, while identifying

potential categories of behavioral data from which accurate

 

inferences about veracity are made, does not directly provide

information about what behavioral cues actually are used by

 

observers. In short, it is unknown how accurate observers
could be in detecting deception if they were provided with the
verbal content of lying and truthful persons. It is also
unknown how accurate they could be if they were provided both
the verbal content and the paralinguistic characteristics of
subjects' behavior, as would be the case if they were exposed
to deceptive and truthful behavior in an audio only format.
More importantly, it is unknown how these categories of
information would interact with visual information in allowing
accurate inferences of veracity. Knapp et a1. (1974) suggest
that inconsistencies between various categories of information
may provide the best cues for making accurate inferences about

lying.

\.

/‘

\

)
l

8

In most deception situations, the observer has the
opportunity to both see and hear the suspected deceiver. The
only way to determine what cues observers actually use to
make accurate judgments of veracity is to provide categories
of information to observers both in isolation and in combina-
tion. The experiment reported in this paper was designed to
do exactly this. Its purposes are to refine and extend the
work of Ekman and Friesen by: (l) presenting observers with
additional categories of information, both in isolation and in
combinations, and to compare the resulting accuracy differen-
ces; and (2) to examine observer accuracy in a situation
involving false statements about a factual event, in addition
to replicating the situation used by Ekman and Friesen (1974).
It is necessarily "exploratory" because theories of the
detection of deception are insufficiently developed and
research findings sufficiently scarce to make predictions
about accuracy based on these additional categories of infor-
mation possible. The deception literature does, however,
suggest specific cues on which accurate judgments may be
based. These cues fall within three categories of "observable
behavior": verbal, visual, and paralinguistic. These cate-
gories, as well as other relevant concepts, will not be

defined.

Some Definitions

 

It will be useful from the outset to define some

important terms and to delimit explicitly the areas of

9

concern in this paper. Deceptive communication, or lying, as
used here, means the purposeful transfer of symbols with the
intent to create beliefs in the receiver which are counter to
fact as perceived by the source/deceiver. Unintentional lies
(misstatements), or lies which are a result of misperceptions
of reality by the source as agreed upon by "objective" obser-
verstfiage-thus not within our interest.
/W7I The factors which contribute to the accurate identi-
fication of deception in everyday encounters are many.
Interest here is on observable behaviors which are perceived

(.by the receiver. Situational or context variables will not

,” be considered. This is not to deny their importance. Circum-

m - . .h...

 

stances such as the receiver's knowledge of the past truthful-
ness of the source or perceptions of benefits to be accrued
by the source from lying probably contribute to the detection
of deception. Suspicion of lying by the source may even be a
prerequisite for receiver identification of lying.1 For this
reason observer suspicion will be assumed as a constant.
However, a thorough examination of the role of situation in
detecting deception is beyond the scope of this paper.

The phrase "observable behavior" is broad and poten-
tially ambiguous. Observable behavior, as used here, refers
to two categories of acts: verbal and nonverbal. verbal
behavior simply refers to the content or meaning which the

linguistic symbols themselves elicit in the receiver. Non-

.verbal behavior is more difficult to define and the particu-

 

lar areas of interest within a broad class of acts falling

10
under the rubric "nonverbal communication" need to be explic-
itly identified.

Researchers have created a variety of nonverbal come
munication category systems. Knapp (1972) attempts to synthe-
size various systems and comes up with seven categories.2 Of
those, research examining the detection of deception from
observable behaviors has focused on two: body motion and/A

./7

paralanguage. These two categories of nonverbal communication
Iwill also be the primary concern in this paper.

Ekman and Friesen (1969b) suggest a five category
system for analyzing body movement: (1) emblems, which have
direct verbal meaning--e.g., the gestures used to represent
"A—OK" or the extended thumb hitch-hiking symbol; (2) illus-
trators, which serve to illustrate what is said verbally--
e.g., pointing when giving directions or depicting spacial
arrangements with the hands; (3) affect displays, which are
primarily facial expressions of emotion--e.g., anger or fear;
(4) regulators, which maintain and control the back-and-forth
flow of speaking--e.g., looking up at the other person when
it is her or his turn to speak; (5) adapters, which are move-
ments containing fragments of behavior that once had an
instrumental purpose--e.g., wiping off one's mouth during
conversation or indicating anxiety by rapid hand movements.
This seems to be a catch-all category and includes most
movement which does not fit clearly into any of the other

categories. Body movement provides information to observers

through the visual channel only, and the term "visual cues"

11
will be used to refer to body movements on which judgments
of deception are based.

"Paralanguage deals with how something is said, not
what is said. It deals with the range of nonverbal vocal
“cues surrounding common Speech behavior" (Knapp, 1972, p. 7).
Paralinguistic cues are vocal cues which are audible to the
receiver but which by themselves do not have linguistic mean-
ing. Paralanguage consists of "voice set" and "nonverbal
vocalizations" (Eisenberg and Smith, 1971). "voice set"
includes such characteristics of the voice such as volume,
pitch, rate, and rhythm. "Nonverbal vocalizations" include
pauses, audible yawns, belches, and sounds such as "uh-huh,"
"ah," and "mmm."

I

’” Taken together, verbal cues, visual cues, and para-
Ilinguistic cues constitute the definition of "observable
; behavior" to be used in this paper.
One final distinction: The term "subject" will be
used to denote the individual who is attempting deception

and "observer" will refer to the individual who attempts to

detect the deception.

The Deception Literature

 

Researchers who have studied deceptive communication
have approached this topic from two perspectives. One line
of research has attempted to identify behavioral correlates
of lying by examining differences in the observable communica-

tion behaviors of lying and truthing individuals (English,

12
1926; Berrien and Huntington, 1943; Ekman and Friesen, 1969a;
Matarazzo, Wiens, Jackson and Manaugh, 1970; Mehrabian, 1971;
Horvath, 1973; Knapp, Hart and Dennis, 1974; McClintock and
Hunt, 1975).

A second line of research has examined the extent to
which untrained observers can accurately detect deceptive
communication (Fay and Middleton, 1941; Hildreth, 1953; Maier
and Thurber, 1968; Shulman, 1973; Ekman and Friesen, 1974).
Both of these research approaches require the creation of
samples of lying and truthful behavior. Most of the methods
commonly used to create lying and truthful stimuli have weak-
nesses which affect interpretation of the results the particu-

lar method yields.

Creating Stimulus Materials

 

A common method used to generate stimulus materials
has been to have individuals either pro or counterattitudin-
ally advocate. Knapp et_al., (1974), for example, in the most
comprehensive content analysis of lying and truthful behavior
conducted to date, used this method. Counterattitudinal
advocacy simply involves advocating a position counter to
one's attitude and Knapp st 31. (1974) use this situation to
generate their sample of deceptive behavior and employed pro-
attitudinal advocacy to get their sample of truthful behavior.
This method suffers from three shortcomings. First, if the
advocate is simply stating the advantages of a position with

which he disagrees, he may not be lying. Since it is quite

13

possible to hold one position and recognize some advantages
in a counter position, the advocate may sincerely believe
that the advantages are real. Second, there is evidence that
when individuals counterattitudinally advocate they sometimes
change their attitude to conform more closely with their
verbal behavior (cf. Miller and Burgoon, 1973). Although it
is unclear when the "change" takes place, it may conceivably
occur before the end of the speech, thus making at least some
of the deception stimuli material proattitudinal and truthful.
Some researchers suggest that the effects of counterattitudi-
nal advocacy can best be conceptualized as attitude formation
rather than attitude change. Bem and McConnell (1970) demon-
strate that unless an advocate's attitude is explicitly called
to his or her attention and made salient before s/he advocates,
slhe may be unable to accurately recall his or her initial
position. The final shortcoming of this method is that when
individuals counterattitudinally advocate, they may purposely
generate observable cues to "tip-off" their audience to the
fact that they do not "really" believe what they are saying.
These problems make the relationship between the observable
behaviors of individuals who are counterattitudinally advo-
cating and those who are lying problematic. While studies
using this method to generate stimulus materials may be sug-
gestive of areas to examine, confidence in the results must
be qualified.

A second method has been to have subjects role play

lying or truthful behavior. Maier and Thurber (1968), in a

14

study examining the accuracy of observers in detecting decep-
tion, used this method. When role-playing, the "lying"
behavior is not inconsistent with any matters of fact known
to the subject; rather, the subject pretends that he is lying.
In this situation, subjects may generate behavioral cues which
they believe to be consistent with the behavior of a person
who was actually lying in that situation. Whether subjects
"know" how real life liars behave is unknown. Role playing
as a research method has been severely attacked for this
reason (cf. Freedman, 1969). Mehrabian (1972) content analyzed
the behavior of individuals who either engaged in counteratti-
tudinal advocacy or role-play lying and found several differ-
ences. For example, he reports less frequent head nodding by
role-playing liars. At best, it is unclear how the behavior of
role-playing liars relates to actual liars and this is impor-
tant in evaluating the results of studies using this method
to create stimulus materials.

A third method used by researchers to generate lying
and truthful stimuli has been to have subjects lie or tell
the truth in situations where detection has no consequences
for the subject. Berrien and Huntington (1941), for example,
planted dimes in a classroom and told subjects where to find
them. The subject was then asked whether s/he had taken the
dime from the room. Subjects were told that if they were
detected they would have to give the dime back. Research
using instruments to detect deception provides ample evidence

that subjects' perceptions of the consequences of being

15
detected affect the likelihood of detection (Gustafson and
Orne, 1963). Davis (1961) suggests a "punishment theory"
which states that the greater the consequences of being
detected, the greater the physiological response during
lying and therefore the greater the likelihood of detection.
If the theory also holds for observable behaviors, liars and
truthers should behave differently only when there were per-
ceived consequences resulting from detection. The more

serious the consequences, the more the behaviors should

. I"- ‘u—.—

differ. Several studies which used trivial lies with no con—
sequences resulting from detection, however, did find differ-
ences between liars and truthers. An alternative explanation
for these differences is the demand characteristics of the
research situation. If the subjects are aware that deception
is being studied and that there are no consequences if they
are detected, they might help observers by purposely genera-
ting cues which they believe to be indicative of deception. !
The applicability of findings based on lying in situations /
where detection has trivial consequences to situations
involving more serious lies is unknown.

A final group of studies has generated stimulus
materials in situations where the lies are both real and
important. Greatest confidence may be placed in the results
of these studies. To properly evaluate the results of studies
examining deceptive behavior, it is important to note the

method used in creating the stimulus presented to observers.

16
Figure 1 classifies each study to be reviewed according to

the method used.3

Research Findings

 

If a category of information provided to observers is
to have an effect on the accuracy of their judgments, behavior
available for observation in that category must co-vary in
some way with veracity. Findings from studies which examine
differences between the behavior of liars and truthers will
now be presented within the framework of the three categories
of observable behavior defined above.

This review will isolate generalizations (if any are
possible) about differential behavior of liars and truthers
with respect to that category of observable behavior. It
should again be noted that this type of research identifies
potential cues only. Differences between the behavior of
liars and truthers within a particular category do not neces-
sarily mean that observers will be more accurate when they
are provided with this category for observation. To identify
veracity they need to bgth notice the cue and correctly

associate it with either lying or truthing.

Behavioral correlates of lying: visual cues.

 

Head--The Ekman and Friesen (1969a) theory of leakage
and clues to deception indicates that the head will be the
area of the body least likely to emit cues from which lying

can be accurately inferred. Research on head behavior,

17
Table 1

Method Used to Create Lying and Truthful Stimuli
Materials in All Studies Reviewed

 

 

 

Counterattitudinal Advocacy» Role Playing
Hildreth (1953) Maier (1965)
Mehrabian (1971) Maier and Janzen (1967)

Experiment I

Experiment 11* Maier and Thurber (1968)

Knapp, Hart and Dennis (1974)

Trivial or No Perceived Conse-
quences Resulting for the

 

 

Subject from Detection Serious Consequences
English (1920) Ekman and Friesen (1969)
Marston (1920) Mehrabian (1971)

Fay and Middleton (1941) ExPerlment III

Berrien and Huntington (1943) Shulman (1973)

Matarazzo, Wiens, Jackson Ekman and Friesen (1974)

and Manough (1970)

Cutrow, Parks, Lucas, and
Thomas (1972)

Motley (1974)
McClintock and Hunt (1975)

Mehrabian (1971)
Experiment II

 

*

Mehrabian (1971, Experiment II) manipulated the method used
to induce lying and truthful stimuli including both role
playing and counterattitudinal advocacy.

18
however, suggests there may be some differences between lying
and truthful subjects.

The most common area of the head used as the dependent
variable in research on deception has been the eyes.
McClintock and Hunt (1975) had subjects both truthfully
answer questions about pleasant, unpleasant and neutral
topics, and lie in answer to questions on a topic of high
importance to the subject. There were no consequences for
being detected. They found more eye contact when subjects
were lying than when they were truthfully answering questions
about a pleasant topic. There were no differences in eye con-
tact between the deception condition and conditions in which
subjects answered questions about either unpleasant or neutral
topics. Knapp st 31. (1974) found the duration of eye contact
units to be longer when subjects were proattitudinally rather
than counterattitudinally advocating. The number of separate
eye contact units did not differ in the two conditions.
Matarazzo e£_§l. (1970) had subjects lie and tell the truth
during an interview. Subjects rehearsed their lies ahead of
time and there were no perceived consequences resulting from
detection. No relationship between eye contact with the
interviewer and veracity was found. These studies provide no
basis for generalizations about eye contact and veracity.

One study found more eye contact for liars (McClintock and
Hunt, 1975), one found more for truthers (Knapp st 21., 1974),

and one study found no relationship (Matarrazo, et a1. 1970).

M.

I
1!
‘4»

(I)
~a

RF

We

'D'C

cl:

8

th

19

None of these studies used realistic lying situations with
important perceived consequences resulting from detection.

Two other kinds of eye behavior have been examined:
pupillary change and blinking. Berrian and Huntington (1943)
examined eye pupil change during lying and truthful beha—
vior. They found sudden changes in pupil stability more
often when subjects were lying. Berrian and Huntington
(1943) used a mechanical "short-focus telescope" to measure
pupil change. Since they do not indicate how sensitive this
device was relative to what could be observed without a
mechanical device, it is unclear if pupil change during lying
could be noticed with the naked eye. Obviously an observer
would at least have to be in close proximity to notice pupil
change. Cutrow st 31. (1972) examined eye blinks during
deception. They created the perception of moderately serious
consequences resulting from detection by telling subjects
that "persons of superior intelligence and maturity" could
usually avoid detection and that if their lies were not
detected they would receive "extra participation credit."
They found fewer eye blinks when subjects were lying than
when they were telling the truth. Eye blink latency, defined
as the length of time between a question and the first blink,
did not differ as a function of veracity. Cutrow et 31.
(1972) used a mechanical device to measure blinking. Obser-
vers would also have to be very close to subjects to notice

eye blinks.

20

Several studies have examined smiles and overall
facial affect display as dependent variables in deception
contexts. Mehrabian (1971) conducted three experiments to
examine the behavior of deceitful communicators and used
"facial pleasantness" as one of his dependent variables.
Experiments I and II used counter and proattitudinal advocacy
to generate lying and truthful behavior for analysis and
there were minor perceived consequences resulting from detec-
tion. Experiment II also had role played lying and truthful
conditions. In Experiment III a lying situation was created
which probably had important perceived consequences for the
subjects. They were induced by a confederate to cheat on a
test and were later questioned about how they had done so
well. Mehrabian (1971, Experiment I) found higher overall
"facial pleasantness" for lying males than for truthful males.
No differences were found for females. In Experiment II,
subjects who were counterattitudinally advocating displayed
greater overall pleasantness than did subjects who were pro-
attitudinally advocating. No consistent pattern was found
in Experiment III.

McClintock and Hunt (1975) and Knapp st 31. (1974)
both investigated the smiling behavior of liars and truthers.
Smiles would be subsumed under Mehrabian's (1971) category,
"facial pleasantness." McClintock and Hunt (1975) found lying
subjects smiled less than truthful subjects, regardless of the
pleasantness of the topic about which they were being truthful.

Knapp gt El- (1974) found no difference on smiles between pro

21
and counterattitudinal advocates although they do report a
slight trend consistent with Mehrabian's findings; i.e.,
counterattitudinal advocates smiled more than proattitudinal
advocates. These findings provide insufficient justification
for generalizations about differences between liars and
truthers with respect to smiling and overall facial pleasant-
ness.

Hands and Arms--According to Ekman and Friesen
(1969a) the hands and arms should emit more cues indicative
of deception than the face. Unfortunately, inconsistencies
in the operationalizations used to examine hand and arm move-
ment in deception contexts make comparisons among the studies
difficult.

Ekman and Friesen (1972) report some preliminary
analyses of the stimuli used in their 1974 study. This study
probably created quite high perceived consequences of detec-
tion by telling subjects that success in their chosen pro-
fession was related to the ability to lie successfully. They
found that subjects used fewer hand illustrators when lying
than when truthing. Liars used more hand shrug emblems and
more face play self-adapters than truthers. Knapp st 31.
(1974) lumped both illustrators and emblems under "gestural
duration" and found no difference between liars and truthers.
However, this finding may have been consistent with the Ekman
and Friesen (1972) result if the latter authors had not dis-
tinguished between the two forms of hand movement. Knapp et

31. (1974) included Ekman and Friesen's (1972) "self-adapter"

22
category under a broader category, "adapter duration." Both
studies report more of these behaviors in deception conditions,
but since Knapp 33 31.'s (1974) category is broader, it is
unknown if differences in the face play self-adapter account
for the consistency in the findings, or some other movement
within their broader category "adapter duration."

McClintock and Hunt (1975) found no differences in
"gestures" which presumably includes both hand shrug emblems
and illustrators. They did find more "self-manipulation" in
their deception condition than in any of the truthful con-
ditions. "Self-manipulations" as operationalized by McClintock
and Hunt (1975) would be broader than, but would include, Ekman
and Friesen's (1972) "face-play self-adapter." It would also
include other adapter movements in which the hand and arms
came into contact with other parts of the body. "Self-
manipulation" would be contained in the Knapp 31_31. category
"gestural duration." The issue is further complicated by the
fact that all three studies used different methods to create
lying and truthful stimuli. Mehrabian (1971) was unable to
find a consistent relationship between "self-manipulations"
and veracity. Nevertheless, the results of the Ekman and
Friesen (1972), McClintock and Hunt (1975), and Knapp 31 31.
(1974) studies might justify a tentative generalization that
liars use self-adapters more than truthers.

Legs and Feet--Ekman and Friesen (1969a) indicate
that more leakage cues should be emitted by the legs and feet

than from other parts of the body. Knapp 3E 31. found no

23
differences between liars and truthers with respect to leg
movements. Mehrabian (1971, Experiment I) found more foot
movement by truthers than by liars. He found no differences
for leg movement, and in Experiments II and III found no dif-
ferences for either foot or leg movements.

Several studies have examined posture shifts. Ekman
and Friesen (1974) offer a preliminary finding that liars
engage in more posture shifts than do truthers. McClintock
and Hunt (1975) support this result, finding more posture
shifts in the deception condition than when subjects truth-
fully answered questions about neutral or pleasant topics.
However, these studies probably provide insufficient basis
for a generalization that the body will be a greater source of
leakage cues than will the head. Findings from research which
examines the paralinguistic characteristics of speech in

deception contexts will now be presented.

Behavioral correlates of lying: p3ralinguistic cues.
Voice set--Only two studies have examined voice set
in deception settings. Mehrabian (1971) measured speech
volume of lying and truthful speakers and found no differences
in any of his experiments. Motley (1974), using a Kay sono-
graph instrument which displays sound spectograms, was able
to correctly identify veracity in 24 of 36 subjects. Obser-
vers, unaided by such an instrument, were unable to exceed
chance levels of accuracy. There were no perceived consequen-

ces for the subject resulting from detection.

24

Mehrabian (1971) found speaking rate to be faster
for liars than for truthers in Experiment III but not in
Experiments I or II. Knapp 3E 31. (1974) did not calculate
rate but a rate estimate is obtainable by dividing the total
words used by message duration. This yields a slight and
probably non-significant trend the opposite of the Mehrabian
(1971) result. That is, truthers talked at a slightly higher
rate than liars. Other voice set variables such as pitch,
range, rhythm, and resonance, have not been investigated.

Nonverbal vocalizations—-Mehrabian (1971, Experiment
I) found more speech errors in the deceitful condition than
in the truthful condition, but this finding was not replica—
ted in Experiments II or III. Knapp 3£_31. (1974) report a
non-significant trend in the same direction. Neither study
makes clear how speech errors were measured.

Knapp 33_31, (1974) found no differences in pauses
used by truthers and liars. It is unclear, however, how
pauses are measured. Matarazzo 33 31. (1970) found that
lying and truthful subjects were equally likely to interrupt
the person who was interviewing them.

Four studies have investigated "reaction time
latency" and deception. Reaction time latency is the time
lapse from when a question is asked to the beginning of the
answer. Early researchers (Marston, 1920) thought that liars
would require more time to respond because they had to think
longer about the answer. Marston (1920) conducted a word

association experiment in which subjects were instructed

25
either to follow directions or to respond in an opposite
manner. No consistent relationship between lying and reaction
latency was found. Goldstein (1923) found that liars took
longer to begin responding to questions than did truthers.
English (1926) failed to replicate this finding. All three
studies had no perceived consequences resulting from detec-
tion. Matarazzo (1970) found truthful and deceitful subjects
responded to questions with equal speed. Based on these
studies, no generalizations about the paralinguistic charac-

teristics of truthers and liars are justified.

Behavioral correlates of lying: verbal cues.

 

Knapp 3£_31. (1974) undertook a detailed content
analysis of the verbal behavior of liars and truthers. They
analyzed twenty categories of verbal behavior and report
differences on ten of them at beyond the .01 level (one-
tailed). These categories, and the direction of the relation-
ship of liars vs. truthers, are: confidence ratio, liars less
than truthers; different words, liars less; factual state-
ments, liars less; self-experience, liars less; past referen-
ces, liars less; probes, liars more; self-interest, liars
less; other references, liars, more; disparaging statements,
liars, more; and total words, liars less. Message length was
also measured by Matarazzo 33 31. (1970) and Mehrabian (1971),
but operationalized as time duration instead of total words.
Mehrabian (1971) in Experiment II found role playing male

truthers talked longer than role playing liars. The

26
difference was not found for the pro and counterattitudinal
advocacy conditions of the experiment or for Experiments II
or III.

Since none of these studies created stimuli in situ-
ations with definite perceived consequences for the liar
resulting from detection, there is no firm basis for generali-
zations about the relationship between veracity and verbal
content. Research which has focused on the detection of

deception from observable cues will now be examined.

Detecting deception. The second major approach used

 

by social scientists to study deceptive communication has been
to examine the extent to which observers, unaided by mechani-
cal devices, can accurately detect lying. These studies pro-
vide information about the particular behavioral differences
that discriminate truthers and liars 931y_if the categories of
behavior which are presented to observers are manipulated. If
observers are given identical information, it is possible to
make statements about their accuracy in detecting deception,
but it is not known on what basis the judgments were made.

Fay and Middleton (1941) had subjects answer ques-
tions about their personal characteristics over a public
address System. There were no perceived consequences for
being detected. They found correct judgments of veracity in
55.6% of the trials. Hildreth (1953) found that observers
were unable to distinguish between pro and counterattitudinal

advocates at beyond a chance level. He, too, did not create

27
conditions of perceived consequences resulting from detec-
tion for his subjects. Maier (1965) found that role playing
liars and truthers were accurately detected at a better than
chance level. Again there were no consequences for detec-
tion. Shulman (1973) created a realistic lying situation
modeled after Mehrabian's (1971, Experiment III) technique
in which subjects were induced to cheat on a task and then
asked how they had done so well. His observers were unable
to detect deception at beyond chance levels.

Maier and Janzen (1967) attempted to identify the
specific cues that lead to accurate judgments by asking
observers to list the observable behaviors which resulted in
particular judgments of veracity. They found no relation-
ship between accuracy and the reasons cited for judgments.
Observers may not have known the basis for their judgments and
thus may have generated a list of post hoc reasons to
rationalize their decision about any given individual.

Two studies have manipulated the categories of lying
and truthful behavior made available to observers and come
pared accuracy scores. Maier and Thurber (1968) had subjects
role play lying and truthing behavior in an interview situa-
tion. Subjects' answers to questions were provided to
observers in three different formats. One group watched the
interviews live (watchers). A second group heard a tape
recording of the interviews (listeners) and the final group
read a written transcript of the stimuli material (readers).

Results indicated that listeners and readers were more

28
accurate (77.0% and 77.3%, respectively) than watchers
(58.3%). The authors reason that the visual cues available
to the watchers may have distracted them from noticing the
verbal (and paralinguistic) cues and thus reduced their
accuracy scores.

Ekman and Friesen (1974) conducted the only experi-
ment which manipulated the categories of behavior presented
to observers and created lying and truthful stimuli in a situ-
ation with important perceived consequences resulting from
detection. Student nurses were told that success in lying
was associated with professional success. They were shown
silent films which were either pleasant or unpleasant and
asked to describe their feelings while watching them. Immedi-;
ately before the film began they were instructed to answer
either honestly or untruthfully. Observers were shown silent
videotapes of these answers which contained either a close-up
shot of the head, or a shot of the head and body with the
head blocked from view. Half of the observers first viewed
a truthful sample of each subject's behavior. In the con-
ditions where no truthful "familiarity" segment was initially
presented, subjects who viewed the body only were no more
accurate than those who viewed the head only. However, when
the familiarity segments were included, subjects were more
accurate when they saw the body only than when theysaw the
head only (56.8% vs. 45.4%).

The Ekman and Friesen procedures for generating

honest and deceptive behavior do have one shortcoming. The

29
unpleasant films they used were extremely stressful. The
cues which they emitted may have indicated extreme emotional
stress not directly associated with lying.. Specifically,
the cues could have been indicative of the subjects' failure
to control their emotional re3ponse to the films themselves,
and not a result of their verbal statements about their feel-
ings. The behaviors which subjects emitted could have
occurred even if they were not questioned about their feel-
ings as they watched the films. The experiment reported here
attempts to improve external validity by having subjects
answer questions about a factual event, in addition to
answering questions about their feelings while viewing stress-

ful materials.

Summary and Research Design

The research on visual, paralinguistic, and verbal

correlates of lying and truthing behavior offers little in

 

terms of identifying specific cues on which accurate judg-
ments of deception may be based. However, each of these
categories include at least several studies which do report
differences between the behavior of liars and truthers.
Visual, paralinguistic, and verbal behavior all contain the
potential for accurate inferences of veracity.

The studies on detecting deception demonstrate that

 

untrained observers, unaided by mechanical devices, are able
to detect deception more accurately in some circumstances

than in others. The studies by Maier and Thurber (1968) and

30
Ekman and Friesen (1974) suffer from procedural weaknesses
that limit their contribution to our understanding of the
behavioral basis for accurate inferences of deception. How-
ever, both studies suggest useful approaches to the problem
and demonstrate that it is possible to manipulate categories
of behavior presented to observers and to observe systematic
differences in accuracy. The present study attempts to refine
and extend the work of these researchers by presenting
observers with categories of observable behavior in isolation,
and in combinations, and comparing the resulting accuracy of
judgments of veracity.

Visual information will be operationalized in this
experiment as "shot," which represents physical areas of the
bodies of subjects made available to observers, and "colorbw,"
which represents whether observers are exposed to the stimuli
in a color or a black and white format. The shot variable
will have three levels: head only; body only; and head and
body. Head only and body only conditions will allow the
direct replication of Ekman and Friesen (1974) while the
head and body condition will permit an examination of
accuracy in a situation more closely approximating the visual
information available to observers in "real life" deception
contexts.

The colorbw variable has two levels: color, and
black and white. The Ekman and Friesen (1974) experiment
used black and white videotapes, yet visual stimuli are per-

ceived in color in face-to-face situations. Certain potential

31
visual correlates of deception, such as facial flush or per-
spiration, may be noticed more readily in color than in
black and white. Including this format variable also
increases the potential applicability of the results. For
example, considerable recent discussion has dealt with appli-
cations of video tape technology in the legal system (Bermant
and Jacoubovitch, 1975; Doret, 1974; McCrystal, 1972; Morrill,
1971), and the use of videotape in court proceedings is
increasing. Accurate detection of lying is important for
equitable juror decisions. The type of television format
(color or black and white) could affect jurors' ability to
detect lying. Thus, including this variable creates the
possibility of generating findings with important applications
in the court system.

Observer accuracy based on paralinguistic information
will be examined in several ways. A transcript only con-
dition, with all paralinguistic information removed, will
Operationally define the verbal content of the subject's
responses. Comparisons between this condition and an audio
only condition will allow an examination of the extent to
which paralinguistic cues alone contribute to accuracy.
Observers in the Ekman and Friesen (1974) experiment did not
hear the audio portion of the tapes. Half of the observers
in this experiment, who base their judgments solely on visual
information, will also not have the audio information avail-
able. The other half will both see and hear the behavior of

the subjects.

32

Those conditions which include visual information
(i.e., all conditions except audio-only and transcript-
only) create a 3 (shot: head-only/body-only/head-and-body)
by 2 (colorbw: color/black and white) by 2 (Visaudio:
visual-only/visual-and-audio) factorial design. The audio-
only and transcript-only conditions bring the total number
of cells in the design to 14. The design is graphically dis-
played in Figure 1.

We will now turn to a discussion of how and why
observer accuracy in identifying subject veracity could vary

across levels of each independent variable.

Shot

 

Head only vs. body only. If Ekman and Friesen (1969a)

 

are correct in theorizing that the body provides the greatest
source of leakage and clues to deception, their finding that
observers are more accurate when judgments of veracity are
based on the body-only rather than on the head-only should be
replicated. Their 1969 paper makes a distinction between

deception clues and leakage. Deception clues tip off an

 

observer that deception is in progress but do not reveal the
nature of the concealed information. Leakage indicates what
information is being withheld. Leakage would thus only be
available in situations where affective responses were being
concealed. If the content of the verbal statements were
about complex factual events, rather than merely statements

about emotional feelings, it would be impossible for nonverbal

33

 

 

 

 

.cmwmmo Houcmsflummxm was

 

 

 

 

.H ounmflm

 

 

 

 

 

 

 

 

 

 

Mano umfluomccua haco Geode
onwaa
can
Rowan
HoHou
Mano Hmomw> haco Hmomw> mace Hmamw>
accmfl> one acsmw> new anoma> one
owned owosd Geese
atom can one: Mano woom Mano omwm

34
cues to convey accurately the nature of the withheld infor-
mation. Thus, leakage would not be possible and judgments
of veracity would have to be based on deception clues. If,
as has been suggested above, the Ekman and Friesen (1974)
head-body finding was a result of subjects' body movements
caused by their general discomfort as they watched stressful
films, the body may not be the best source of information
about lying when the content of the false statements is
factual rather than emotional. In short, the original find-
ing could have been the result of leakage, i.e., the failure
of subjects to control their emotional feelings. Thus, while
we might expect to replicate Ekman and Friesen's head-and-
body finding with lying and truthful stimulus materials simi-
lar to theirs, it is less clear what will happen when the
stimuli are based on statements about concrete matters of
fact. Consequently no directional prediction will be offered
about the accuracy of observers who view the head-only versus

those who view the body-only.

Head-and-body vs. head-only and/or bogyronly. Obser-
vers who view the head-and-body obviously have more available
information than those who view either the head-only or the
body-only. Whether the head or the body is the greatest
source of cues indicative of deception, observers in the head-
333 body conditions have those cues available, and they
should have the greatest chance to identify veracity accu-

rately. Also if, as Knapp 33 31, have suggested,

35
inconsistencies between categories of behavior result in the
identification of veracity, observers in the head-and-body
condition would have the best opportunity to notice such
"out of sync" behaviors between the head and body. For these
reasons, they could be 3333 accurate.

On the other hand, if the greater information sending
capacity of the head results in more deception cues emitted
from this area, the head-and-body condition could result in
less accuracy, because less head detail would be available for
observers in the head-and-body condition than for those in the
head-only condition. Also, the head might distract viewers
from noticing body cues or vice-versa and less accuracy might
result. For these reasons, observers in the head-and—body
conditions could be 1333 accurate than those in either the
body-only or head—only conditions. Since arguments can be
marshaled in favor of each condition resulting in higher
observer accuracy scores, and since previous research provides
no clear-cut basis for a directional prediction, none will be

offered.

Colorbw

Research on the effects of viewing color versus black
and white television formats indicates there may be greater
infbrmation retention of peripheral material when the presen-
tation is in color (Katzman, 1971; Katzman and Nyenhuis, 1971).
As was pointed out above, several potential visual correlates

of deception, such as facial flush and perspiration, may be

36

noticed more readily in color than in black and white, par-
ticularly in the head-only condition of the shot variable.
This greater richness and variety of color cues could
result in higher observer accuracy when the stimuli are
viewed in color. Also color may increase observer interest
in the stimuli, causing them to pay closer attention and to
be more sensitive to noticing cues indicative of deception.

On the other hand, a color presentation might dis-
tract observers from noticing cues indicative of deception.
For example, brightly colored clothing might receive more
observer attention in a color format than in a monochromatic
format. Since there is no clear-cut basis for a directional

prediction, none will be offered.

Visaudio (Paralinguistic)

 

Observers in the transcript-only condition will make
their judgments based on the verbal content of the subjects'
behavior. In addition to this same verbal content, audio-
only observers will also have paralinguistic information
available. Such paralinguistic information could provide
cues indicative of deception which, in turn, could result in
greater accuracy. However, paralinguistic cues could also
distract observers from scrutinizing carefully the verbal
content for any inconsistencies which would allow the accu-
rate identification of veracity.

A similar situation exists in comparing the visual-

only with the visual and audio conditions. The audio band

37
could provide cues from which accurate judgments could be
made. Visual and audio observers would also have the oppor-
tunity to notice any inconsistencies between the verbal and
nonverbal behavior of subjects which were associated with
deception. The audio band, on the other hand, could distract
observers from noticing visual cues signaling deception, or
vice-versa. Here too, then, as with the other variables,
there is no solid foundation on which to base directional
predictions and none are offered. Thus, this experiment is
an exploration of how the verbal and nonverbal behaviors
available to observers influence their ability to detect

deception.

38

Footnotes

lShulman (1973) tested this proposition and failed
to find any support for the notion that forewarning the
observers that lying might occur increased the accuracy of
judgments of veracity.

anapp's entire category system for organizing
various kinds of nonverbal communication includes physical
characteristics, touching behavior, proxemics, artifacts,
and environmental factors, in addition to body movement and
paralanguage to be defined below.

3The consequences of detection for the subject is,
of course, only one side of the coin. Observers motivation
to accurately judge veracity probably also affects accuracy.
Unfortunately most of the studies provide no information
about observer motivation and consequently none will be pre-
sented here.

Chapter II
METHOD

Overview

Samples of both lying and truthing communication
behavior were videotaped under conditions of high saliency
for 23 subjects. These tapes were edited to include eight
subjects who were making true or false statements about the
factual content of an event they had seen and eight subjects
who were making true or false statements about their emo-
tional feelings as they watched pleasant and stressful slides.
Two versions of each tape were made, one of which was the
exact inverse of the other; that is, if a particular subject
were shown lying on one tape, that same subject was always
shown truthing on the other and vice-versa. One or the other
of these two tapes was shown to 730 college student observers
under 14 different conditions thus creating a total of 28
cells in the design. Observers saw either a close-up of the
subjects' head-only, a shot of the body-only, or a shot of
the head-and-body. They viewed these tapes in either a color
or a black and white format. Half of the observers viewed
the tapes without the audio channel present and half were

allowed to hear the audio. Half the observers viewed one

39

40

version of the tape and half viewed the other. These vari-
ables created a 3 (head-only/body-only/head-and-body) by 2
(color/black and white) by 2 (visual-only/audio-and-visual)
by 2 (Tape I/Tape II) factorial design. Audio-only and
transcript-only conditions (for both Tapes I and II) brought
the total number of cells in the design to 28. Measures of
observer confidence in their judgments were also obtained.

Procedurally, then, this research was conducted in
two phases. First the stimulus materials, which consisted of
videotapes of subjects engaging in lying and truthful beha-
vior, were created. These videotapes were then shown to
groups of observers under each of the manipulated experimen-
tal conditions. Presentation of the research methods in this

chapter follows this same logical order.

Creatingythe Stimulus T3pes

 

Saliengy

It was important to create a situation in which
successful lying was difficult. If the behavior of lying sub-
jects were no different than when they were telling the truth
it would be impossible to interpret observer accuracy scores
across the experimental conditions. Put another way, if
there were no cues which were indicative of deception,
observers would probably not be able to differentiate between
lying and truthful subjects at different levels of accuracy
in the various conditions; and even if they did, it would be

impossible to ascertain why. Thus, many of the procedures

41

used to create the stimulus materials were designed so that
it would be difficult for subjects to lie without being
detected.

Ekman and Friesen (1969a) discuss the conditions
under which lying will be maximally difficult. First the
deception should be salient fcr both the source (subject) and
receiver (observer). "Saliency" refers to the extent to which }/
the deceiver is consciously concerned about the deception, 3

I
and to the extent to which it is important to the deceiver to i

,./
avoid detection. The deceiver should be highly ego-involved
with his or her success in deceiving the observer. Under
these conditions, Ekman and Friesen (1969a) argue, cues giving
away the deception are most likely to occur. This argument
is supported by Davis' (1961) theory discussed earlier and NI)
Gustafson and Orne's (1963) finding that the greater the per-
ceived consequences resulting from detection, the greater the }
physiological response during lying. Deception will also be ,1)
difficult, Ekman and Friesen (1969a) suggest, when the sub-
ject is in the role of both deceiver and detector while the
observer is only in the role of detector. In this situation
the subject wants to both deceive the observer and monitor
the behavior of the observer to determine the success or his
or her deception.

An effort was made to create both roles for the sub-
jects. Saliency was created in a manner similar to that used

by Ekman and Friesen (1974). Subjects were 19 male and four

female senior criminal justice majors at Michigan State

42
University, all of whom planned careers in law enforcement.
They were sent a letter by the Director of the School of
Criminal Justice requesting that they participate in a
research project designed to ”identify certain personal
characteristics of individuals which may contribute to their
successful performances as police officers" (see Appendix A).
When the subjects arrived to participate in the study they
were told that the research was attempting to develop
improved screening tests for prospective police officers;
that earlier research had shown that those who could lie suc-
cessfully were better officers than those who could not; that
the Criminal Justice Department was very interested in how
well they, in particular, did on this task; and that feedback
about their performance would be provided the department and
could affect such things as their letters of recommendation
for securing positions as police officers. They were
questioned during the taping by a detective from the East
Lansing Police Department whose purpose was to add credi-
bility to this cover story. In an effort to place the sub-
ject in the dual role of detector and deceiver, the subjects
were also told to monitor the officers' reactions to deter-

mine if they were successfully deceiving him.

Stimulus Content

 

The interviews, during which the stimulus tapes were
made, consisted of four segments. First, subjects were

asked five questions about personal characteristics, all of

43
which they answered truthfully. As was noted earlier, Ekman
and Friesen (1974) were able to support their head vs. body
accuracy prediction only when they provided the observers
with a sample of honest behavior first. It may be necessary
to have some familiarity with the behavior of an individual
to be able to accurately identify lying. The truthful seg-
ment was provided to increase the likelihood that observers
would be able to detect lying and thus the manipulated cate-
gories of information would be more likely to produce syste-
matic differences.

I The second segment involved questions about the fac-
tual content of a videotape subjects had seen before being
interviewed. They were shown one of two versions of this
videotape. Both versions showed the sentencing of a criminal
who had been tried and found guilty of murder. The individual
who was being sentenced reacted on one tape very violently.
He swung at his attorney, attacked the prosecutor and
bailiff, and had to be forcibly removed from the courtroom.
In the other version of this tape, the individual was very
passive in hearing his sentence. He listened to the judge
politely and was quite docile as he was led from the court-
room. Subjects were instructed to lie to the first three of
the five questions they were asked about the content of this
videotape and to tell the truth to the last two questions,
or vice versa. Subjects who were instructed to lie were also
told what to say, so as to cut down on the variability of the

content of the lying responses. If subjects had been able to

44
make up their own lies, the content of those who were tell-
ing the truth would have been very consistent across sub- I
jects, while the content of those who were lying would have
had great variability. The variability factor probably
would spuriously increase observer accuracy.

The last of the five questions during this second
segment was open ended and simply asked for a description of
the defendant's reaction to being sentenced. Subjects typi-
cally took considerably longer to answer this question than
the other four. On this question, the answer which lying
subjects were instructed to give was consistent with the ver-
sion of the tape which they did not see. In other words,
those who saw the violent version were told to say that the
defendant responded in a very docile manner while those who
saw the nonviolent version were told to say that he reacted
violently. Thus, the same answer could be true for one sub-
ject and false for the other. To further out down on the
variability of the answers to this fifth question, subjects
were given a slip of paper on which phrases were written
summarizing the reaction of the defendant on the version of
the tape which they did not see. This was designed to help
them recall the form of their lying response. This second
segment of the interview is henceforth labelled the "factual"
segment of lying or truthful behavior.

The last two segments of the interviews consisted of
questions about subjects' feelings or emotions as they

watched color slides on a television screen in front of them.

45
One sequence of slides was very pleasant and showed scenes
with boats, lakes, children playing, and so on. The un-
pleasant sequence of slides showed explicit close-up pic-
tures of people who had been seriously burned. Subjects
were told that immediately before each sequence of slides
began, they would see a slide with either the word "lie" or
"true" on it. If it said ”true" they were to answer truth-
fully about their feelings. If it said "lie" they were to
respond.with answers which were the opposite of their feel-
ings. They were falsely told that some subjects would see
two sequences of pleasant slides, some two sequences of un-
pleasant slides, and some one of each. They did not know
which sequence would be prefaced by the word "true" and which
with "lie." In reality, the word "true" always prefaced the
pleasant sequence, and the word "lie” the unpleasant sequence.
Thus the content of their answers was always that they were
experiencing pleasant feelings, regardless of which slides
they were viewing.1 The order in which the sequence of
slides was shown was randomized across subjects. These
segments of the interview will subsequently be labelled the
"emotional" segments.

In summary, tapes were made of each subject: (1) A
telling the truth in answer to questions about their personal
characteristics; (2) lying and telling the truth about the
factual content of the videotape they had seen; (3) lying

about their feelings as they watched unpleasant slides; (4)

46
telling the truth about their feelings as they watched

pleasant slides.

Procedures

 

Subjects were contacted by phone two to three days
after receiving the letter from the Director of the School
of Criminal Justice requesting that they participate in the
research. The caller identified himself as being associated
with the Department of Communication and asked the subject
if he or she had received the letter. The caller then briefly
repeated the content of the letter and asked the subject if
s/he could come in for one hour to participate several days
later.

Thirty-five potential subjects were sent letters. Of
these, 28 were successfully reached by phone. Twenty-six
individuals agreed to participate. They were scheduled to
arrive every half hour from two in the afternoon running into
the evening hours over a two-day span.

When a subject arrived s/he was greeted by the
author (g1) and thanked for coming. The following is an
accurate paraphrase of the cover story which was given them:

Hello, I'm John Hocking and I'm with the Department
of Communication. Thank you for coming. I know

that Mr. Brandsteader was very anxious to have you
participate.

This research is being funded by the RANN division

of the National Science Foundation. RANN stands for
"research applied to national needs." This project
has been going on at several universities and police
agencies for about three years now. Essentially

this research is an extension of some earlier research

47

which was conducted by the New York police training
and performance group and research which was done

by the RANN corporation on police background charac-
teristics and performance. As the letter from Mr.
Brandsteader indicated, we're interested in identi-
fying certain personal characteristics of individuals
which may contribute to performance as successful
members of the law enforcement profession.

As I'm sure you're aware, there has been consider-
able dissatisfaction with present techniques for
screening prospective police officers. Presently
paper and pencil tests are the primary means of
screening applicants. Some individuals turn out

to be poor officers even though they score well on
the tests. Others, who might make good officers,
are in some cases eliminated from consideration by
these paper and pencil tests. The research you'll
be participating in today is designed to identify
specific behaviors which could be used as predictors
of successful job performance. Tests on these
behaviors could then be used to supplement the tra-
ditional paper and pencil tests now used.

 

Your background records were reviewed by the crimi-
nal justice department and they selected you for
participation in this study. What are your career
goals? (Pause for response.) Ah, I can see why the
department is so interested in your performance on
this research.

One of the consistent findings of the earlier research
in this area is that successful members of the law
enforcement profession are able to lie successfully
under a variety of circumstances. For example, when
interrogating someone, it might be necessary to give
the impression that you know more than you do. There
are other examples, too. When you're dealing with
people under stress circumstances, it's pretty impor-
tant to respond to them in an appropriate way, even
if you feel differently. You might need to appear
calm and in control when you don't feel that way at
all. In short, members of the law enforcement pro-
fession need to be pretty good actors at times.

Well, the research that we're doing here today
involves giving you the Opportunity to attempt to
successfully lie under several different circum-
stances. The earlier research has shown that indi-
viduals who later proved to do well in their jobs as
police officers have been extremely successful at
this task. I guess you can see why Mr. Brandsteader
was so anxious to find out how certain people in his
department performed in this research.

48

Subjects were then told that they would be inter-
viewed by a police officer from the East Lansing Police
Department and the format of the interview was explained.
After being shown the tape on which the factual segment of
the interview was to be based, they were instructed about
which questions to answer truthfully and untruthfully. Sub-
jects were also given the information on which to base their
lying responses. They were told the general information the
factual questions would be seeking, but were not told
specifically how the questions would be worded. After they
understood the format of the factual questions they were told
about the emotional segments, and were given the opportunity
to decline to view the burn victims' slides with no penalty.
All subjects agreed to view those slides. Any questions they
had were answered in a manner consistent with the cover story.
After El was sure each subject understood what was required
of him/her, he took the subject to the nearby television
studio and introduced him/her to the police detective. Dur-
ing the walk to the studio, it was explained rather vaguely
that the answers would be videotaped for later analysis. The
prebriefing took an average of about 20 minutes.

The detective seated the subject and then sat down in
his own chair which was about 15 feet directly in front of
the subject. He then reinforced each of the major points
which had been made by El during the prebriefing cover story,
indicating that there was dissatisfaction with current

screening tests for prospective officers and that research

49

had indicated that good liars made good officers. He
described how in his own work a frequent strategy used dur-
ing interrogation was to exaggerate the amount of information
he had about a crime in order to elicit information from the
suspect. Only to the extent that he was successful in con-
vincing the suspect that he really had the information was
this a good strategy. The officer explained that he was
involved in the research because the East Lansing Police
Department was interested in adopting the new screening pro-
cedure. The last thing the officer told the subjects before
beginning the questioning was that they should do their very
best because the E.L.P.D. and the Criminal Justice Depart-
ment were both "very interested" in their performance.

While the officer was talking with the subject, 3

1
gave a form to E in the control room of the television studio

2

which summarized the instructions the subject had been given
about when to lie and when to tell the truth during the
factual questions. This form also indicated which sequence
of slides the subject was to view first during the emotional
questioning. As the interview was conducted, 32 monitored
the subjects' answers and noted any deviations from the
instructions (see Appendix B).

Subjects were videotaped with two color cameras. One
had a close-up shot of the subject's head while the other had
a full shot of the head and body, including the feet. Conse-

quently, less head detail was observable in the head-and-

body shot than in the head-only shot. The head-and-body

50
shot also showed the microphone and microphone stand. The
background for both camera shots was a light blue curtain
which was about eight feet (2.5 meters) behind the subject,
thus appearing slightly out of focus. While subjects wore
colors and styles of clothing which varied considerably, none
were dressed in an unusually loud or gaudy manner. The sub-
ject was seated in a moderately comfortable chair. Two Sony
8600, half-inch, reel-to-reel videotape recorders were used
to record the subjects' answers. A professional video tech-
nician ran all recording equipment and adjusted the close-up
camera for each subject. He also controlled the film chain
Operating in the control room through which the slides were
shown to the subject on a television monitor. Both cameras
were about 25 feet away from the subject and sitting next to
each other, thus keeping the angle of the two shots relatively
constant. The close-up picture was achieved with a telephoto
lens. This camera placement resulted in camera shot angles
from slightly to the left of where the officer was sitting.
Thus in looking at the officer as they answered questions,
subjects were looking slightly to the left of the cameras.
The television monitor on which the slides appeared during
the emotional segments was sitting on an eight-inch high
platform in front of the officer's desk and slightly to his
right. The subject thus had to look to the officer's left
and slightly down to watch the slides. The monitor was 12-
13 feet (about four meters) from the subject. The officer

could not see the monitor from where he was sitting and

51

consequently from the subject's point of view (and in

reality) he did not know which slides were being shown when.

Figure 2 graphically displays this room arrangement. Sub—

jects were then asked the 22 questions in the following

order.

11.
12.

13.

14.

Truthful Segpence

 

What is your name?

What year in school are you?

What are you majoring in?

Where are you from?

Did you happen to go to the football game with Ohio
State a couple of weeks ago? What did you think?

Factual Sequence

 

What was the crime which Mr. Bostick was convicted of?
Did Mr. Bostick say he was sorry for his crime?

Did the judge say that Mr. Bostick has the right to
appeal his conviction?

What was Mr. Bostick wearing?
What was Mr. Bostick's reaction to being told that all

he needed to do to appeal his conviction was to sign a
form?

Slide Sequence 1

 

What kinds of feelings are you having right now?
What kind of mood do these slides create?

What other experiences have you had which convey the
same feelings as these slides?

What are your feelings now that the slides are over?

52

color
cameras

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l

l

l

l

I _

' officer's

' chair

I 8" high
: platform
: desk

I

I

I T.V.

I monitor

i

l

I

l

| Subject's

' curtains chair wall
I

l

l

h----------------- --------—

Figure 2. Room arrangement during the stimulus taping

53
15. Are you telling me the truth?

16. Do you think I believe you?

Slide Sequence 2

 

17. What kinds of feelings are you having right now?
18. What kind of mood do these slides create?

19. What other experiences have you had which convey the
same feelings as these slides?

20. What are your feelings now that the slides are over?
21. Are you telling me the truth?

22. Do you think I believe you?

The entire interview sequence took about five to
seven minutes. After the interview was over, the subject was

thanked by the police officer and introduced to 3. who

3
debriefed the subject.

Debriefing

 

Subjects were taken to a nearby snack bar by 33.
Before beginning the debriefing, they were given a short one—
page questionnaire to fill out (see Appendix C). The primary
purpose of this questionnaire was to find out how the sub-
ject actually felt when s/he watched the slides of burn
victims. It was possible that some subjects might have said
they felt pleasant when watching these slides and been tell-
ing the truth. An examination of responses to an open ended
question indicated that all subjects selected for inclusion
on the emotional segments of the stimulus tapes did indicate

feeling unpleasant while watching the burn slides. This

54
questionnaire also asked subjects how successful they felt
they had been at deceiving the officer during both of their
lying sequences.

During the debriefing, subjects were told that the
Criminal Justice Department would 333 be receiving informa-
tion about their performance (see Appendix D). They were
asked for their permission to show their tape to undergradu-
ates at Michigan State the following term and to a group of
adults from the Lansing area. They were asked not to discuss
the research with other criminal justice majors who might be
participating, until after the second day of running. They
were told that they would soon be learning the full details

of the research.

Editing the Tapes

 

Two subjects failed to show up for participation, one
subject refused to participate after the prebriefing had been
completed, and the tapes of two subjects were lost because of
equipment failure. Taping procedures were completed for 19
male and four female subjects. This resulted in approximately
four hours of lying and truthful behavior on tape-~two hours
of the head shot and two hours of the head-body shot. These
tapes had to be edited into master stimulus tapes of no
longer than 25-30 minutes to make data collection in a 50
minute class possible. It would have been desirable to use
longer segments of each subject and to use samples of all 23

subjects. This would have necessitated collecting data

55

outside of class. The large number of conditions necssita-
ted a very large number of observers (see page 60 below).
Unfortunately it would have been impossible to obtain this
many observers outside of class.

Whatever the observable cues from which inferences
of veracity are made, it is safe to assume that not everyone
emits them equally. Fay and Middleton (1941), for example,
found that subjects were judged accurately by as few as 50.9%
of the observers and by as many as 62.8%. Some subjects in
the present experiment probably emitted many cues which would
likely be perceived by observers as indicative of lying both
while they were lying and truthing. If by chance a lying seg-
ment were selected from such persons and included on a tape
for judgment, observers' accuracy scores would probably be
spuriously inflated. This would not be because they were able
to discriminate between the lying and truthing behavior of the
persons, but rather because they always looked like they were
lying. Other people may not give off many revealing cues;
these persons would probably look like they were telling the
truth all the time. If a truthful segment of these persons'
behavior were included on a tape, accuracy scores would also
be increased; once again not because of the observers'
ability to detect veracity, but because of an idiosyncrasy of
those people's behavior. Analogously, if opposite segment of
behavior were selected, observer accuracy scores would
probably be spuriously deflated. An effort was made to solve

this problem by creating two tapes, one of which was the

56
inverse of the other. Thus if a truthful segment was selec-
ted at random for inclusion on Tape 1, a lying segment from
the same person would be included on Tape 2, and vice-versa.
This procedure resulted in an additional experimental manipu-
lation thus doubling the number of conditions in the design
from 14 to 28.

Whether a segment of any given subject's behavior was
included from the factual or the emotional portion of their
interview ‘was assigned at random except in cases where the
subject's failure to follow instructions resulted in only the
factual or the emotional segment being available. Eight
truthful and eight lying segments were assigned at random to
Tape 1. Of these, four of each were factual and four of each
were emotional. Tape 2 was created by taking the opposite
behavior from the segment (factual or emotional) for each sub-
ject.

There were several subjects who failed to provide both
a lying and a truthful segment for either the factual or
emotional portions of the interview. One of these individuals
was selected for inclusion as the practice example (see below)
for observers, and the rest were not included on the tapes.
This left a total of 16 subjects, 13 male and three female.

The format of each master stimulus tape was identi- ’
cal. Each subject was shown twice. The truthful segment
appeared first and was fdllowed by the "test" segment which J
was either truthful or not. Before each of these segments,"

an announcement was made on all tapes which told observers ’

57
which segment they were about to see (truthful or test) and ,
which subject was about to appear. After the test segment,r
the announcer stated, "That was the test segment for subject
number 1 (or 2, . . . , or 16), please mark your question-
naires." There was then a 10 second pause before the truth-

ful segment for the next subject was announced.

 

Before the 16 subjects who were to be judged appeared
on the tape, a practice example was presented in the same
format as the rest of the tape. It was, however, labelled
the "practice example." I

The physical editing was done by a professional video
technician. Editing was accomplished by recording from the
original tape onto master stimulus tapes in the proper
sequence. The announcements which accompanied each segment
were recorded onto the master tapes at the same time. Since
it was necessary that the head only tapes have exactly the
same segments of the subjects' behavior as the head and body
tapes (and vice-versa), this was accomplished by making note
of the verbal content of the audio portion of the beginning
and end of each segment. It was thus possible to record
this segment of behavior onto the master tape made for the
other camera angle by closely monitoring the verbal content
of the tape.

Four tapes which included the audio portions of the
questions and answers were made: head only (Tape 1), head
and body (Tape 1), head only (Tape 2), head and body (Tape 2).

The final number of master stimulus tapes was doubled by the

58
tapes which contained the visual behavior only. These were
created by making a copy of each of the above four tapes
with the audio recording jack removed during the segments
of the subjects' behavior. Thus, the tapes to be used in
the visual-only conditions were identical to those in the
video and audio conditions, except for the absence of the
audio. The announcements of segments and subject numbers
were on all tapes.

All tapes were in color. In the black and white con-
ditions they were played by using black and white television
monitors. Tape 1 was 27 minutes and 15 seconds long, while
Tape 2 was 26 minutes and 20 seconds long.

The transcript condition for both tape versions was
made from the audio conditions. It included the verbal
answers only. All paralinguistic cues such as 'ahs," "mmm," ’
and other audible sounds which do have corresponding verbal
symbols were removed. This was done so that the comparisons
between the transcript-only conditions and the conditions
which included audio would partial out only the verbal con-
tent.

The audio-only condition was created by playing the
audio and visual tapes with the video portion of the tele-
vision monitors adjusted so that there was a black screen.
Thus the sound quality was identical to the sound which was

played in the video and audio conditions.

59

Collecting the Data

 

Observers

 

The observers who judged the veracity of the sub-
jects appearing on the stimulus tapes were 730 undergraduate
male and female students enrolled in 18 introductory communi-
cation classes during the spring term of 1975 at Michigan
State University. The two classrooms in which all classes
met each had room dividers which allowed the class to be
broken into two groups. This was done by having students
with even student numbers on one side of the divider and
those with odd student numbers on the other. This usually
resulted in nearly equal numbers of observers in both groups.
On those occasions when this was not the case, observers were
arbitrarily moved to create equal sized groups. Thus there
were 36 separate intact groups which were randomly assigned
to the 28 experimental conditions. The only constraint on
this random assignment was that two conditions which con-
tained the audio band were always run at the same time,
because the room dividers were not completely sound-proof.
The 28 largest groups were each assigned to one condition.
The remaining eight smallest groups were run in the same con-
dition as the eight smallest groups within the 28. This was
done to create as large and evenly distributed sample size
across conditions as possible. Table 2 presents the sample

size obtained in each of the 28 cells.

60

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hm HH «N HH
oamH
ma H mm H
man I haco stmfi>
omm n Hmsmfi> pom oavs< haco umfiuomcous haao owv=<
Maw moN mam
mofl woa .mm mod Rafi MOH
mqa mu wN NN mm «N «N HH mugs:
Hon - can
xomam
mma nm «N mu HN ma mm H
NmH «N MN Nu an em om HH
HM” HOHOU
mmH mm mm mm on an mN H
sHao Hmamfi> sfiao Hmsmfi> mace Hmsmu> mama
Hmsmw> pom Hmomfi> paw Hmsmfi> vow
owvs< owvs< oaoac
scan 8 came xaao seam sane emu:

cwwmon mnu mo HHwo zoom aw onwm madamm

 

 

N mHQAH

61

Questionnaire

 

The questionnaire on which observers indicated their
judgments (see Appendix E) had three pages of instructions
which explained that the research was examining people's
ability to detect lying and described in some detail the cir-
cumstances under which the stimulus tapes had been created.
For example, it was explained that the first eight peOple
were answering questions about a videotape they had seen and
the last eight were answering questions about slides that
they were viewing. Observers were told that it was very
important to each of the subjects that they successfully
deceive the police officer who was interviewing them. The
presentational format for the various segments on the tape
was also explained. To motivate the observers to pay close
attention to the tapes, they were told that if they put
their name and summer mailing address in the place provided,
they would be sent their personal accuracy score. It was
emphasized that participation was voluntary and they should
do so only if they were willing to do their very best to
detect whether or not a subject was lying.

An effort was made to avoid observers' attempts to
make an equal number of judgments of truthing and lying by
explicitly telling them that the particular tape which they
saw might contain mostly truthful segments, mostly lying
segments, or about equal numbers of each. Observers were
told to make each judgment independent of their other judg-

ments. They were also told there was no relationship

62
between the length of a segment and whether the subject in
that segment was telling the truth or lying, and that the
same answers could be truthful for some subjects and untruth-
ful for others.

Observers made judgments dichotomously for each sub-
ject. Following each judgment, observers indicated how
confident they were of their judgment on an ll-point scale.
There were 16 such sets of measures, corresponding to the 16
test segments on each stimulus tape. Each observer was thus
required to make 16 separate judgments of veracity and 16
assessments of their degree of confidence.

The questionnaire also collected information about
how successful observers perceived themselves to be at lying,
how interested they were in participating in the experiment,
and whether they were sitting in a good position to observe
the tape. Standard demographic information was also

gathered.

Procedures

Two Sony color television monitors were set up (and
turned on) on one side of the classroom and two black and
white monitors were on the other, prior to the arrival of the
students. Both monitors were run off the same videotape
recorder and thus showed identical pictures. When the body-
only visual conditions were run, the upper third of the

monitors were blocked from view with a piece of cardboard.

63

Several minutes after the class was scheduled to
begin the instructor introduced g1 as a graduate student in
the Department of Communication who was conducting some
interesting research. El then briefly explained that the
research was being funded by the National Science Foundation
and was examining individuals' ability to detect lying. It
was explained that some students had been interviewed by a
police officer the previous fall and had been instructed to
lie at certain times and to tell the truth at others.
Observers were told that if they chose to participate, they
would be watching tapes which were made of these individuals
and attempting to detect lying.

After this brief introduction, the observers were
divided into two groups as described above, and the room
divider was closed. E2 and an assistant were with one group
of observers and El and an assistant were with the other.
The questionnaires were then handed out. Each questionnaire
had previously been coded by condition to reduce the possi-
bility of subsequent confusion. While the questionnaires
were being handed out, those observers who appeared to be in
a bad position for viewing the tape were asked to move to a
better spot.

The instructions on each questionnaire were read
aloud and observers were instructed to follow along. After
the instructions were read, the tape was started and the
observers watched the practice example. When the practice

example was completed the tape was turned off and observers

64
were asked if they had any questions. When all questions
were answered and the Es were satisfied that all observers
understood what was required of them, the tape was started.
The last thing said before the tape began was a reemphasis
of the importance of making independent judgments without
consulting with other observers. Students who came into
class late after the instructions had been read were allowed
to sit in the room during the experiment but did not fill
out a questionnaire.

After the tape was completed, subjects were asked to
finish filling out the questionnaire. The questionnaires
were then collected, the room dividers opened, and the
observers debriefed.

The experiment was run over an eight day period,
beginning on one Friday and concluding the following Friday

(April ll-18, 1975).

Debriefing

 

In the 50 minute classes, of which there were 11
(out of 18), there was only time for an extremely short
debriefing. This consisted of reaffirming that there was
no deception used during this phase of the research and
explaining the other conditions in the experiment. Obser-
vers were assured that they would be receiving the promised
information about their accuracy scores if they had reques-
ted it and thanked for participation. In the longer classes

a more detailed debriefing was presented if the instructor

65
was willing to allow the additional time. This longer
debriefing included showing samples of tapes from other con-
ditions, describing procedures through which the stimulus
tapes were made, and a more thorough explanation of the
various questions that the research was designed to answer.
The Ekman and Friesen head-body hypothesis was not explicitly
mentioned for fear of observers talking about the study with
other students who were yet to participate. Except for this,
no effort was made to keep details of the research secret.
There was also no effort to reduce interclass discussions

about the experiment.

Results Letter

 

A letter briefly summarizing the experimental design,
the major findings, and individual personal accuracy scores
was sent to those observers who had requested this informa-
tion. This letter was sent about four months after the data
were collected (see Appendix F). The importance of the
observers' contribution to the research was emphasized and

they were again thanked for participation.

66

Footnotes

1It may have been desirable to have had subjects
also respond.with lying answers as they watched the pleasant
slides and truthing answers as they watched the burn slides.
Ekman and Friesen (1974), however, only showed segments to
observers of subjects truthing during pleasant films and
lying during unpleasant films. In order to replicate them
as closely as possible, the same procedure was followed here.

CHAPTER III

RESULTS

The primary dependentvariable in this experiment
was observer accuracy in identifying subject veracity. Each
observer made 16 accuracy judgments, eight of subjects engaged
in lying or truthing behavior about the factual content of
the videotapes they had seen and eight of subjects making
true or false statements about their emotional feelings as
they watched pleasant and stressful slides. The accuracy
scores for each observer were obtained by separately adding
the number of correct judgments within both Segments and
dividing by eight. Thus, each observer's level of accuracy
is represented by two proportions: one for the factual
segments and one for the emotional segments. Scores conse-
quently have a possible range of O to 1.0, with .50 repre-
senting four correct judgments out of the eight attempted.1

Observers' degree of confidence in their accuracy
judgments was also treated separately for the factual and
emotional segments. These responses were added and divided

by eight thus giving confidence scores a range of O to 10.

Identical analyses were performed on each of the four

67

68

dependent variables: factual accuracy; emotional accuracy;
factual confidence; and emotional confidence.2

Four-way analyses of variance were performed on the
data from the 24 cells which fit within the completely
crossed factorial design.3 Comparisons between the audio-
only and transcript-only conditions, and other comparisons,
were performed with E tests. These tests used the MSW from
the overall analysis when cells included in the comparison
were from the overall analysis of variance. The MSW used
when the transcript-only and audio-only conditions were
included in comparisons were calculated by averaging the
variance within each of these cells into the MSN from the
overall analysis of variance. Any comparison which exceeded

the .05 level two-tailed was considered significant.4

 

Tap§_is not a theoretical variable in this experi—
ment. The purpose of creating two versions of the stimulus
was to have both a lying and truthing sample of each sub-
ject's behavior judged by observers. It was assumed that
observers would make more judgments of "true" (or "false")
for some subjects than for others, regardless of the sub-
ject's actual veracity. Observer accuracy in judging each
subject's veracity across both levels of the tape variable
will now be presented, followed by a presentation of the
results for both accuracy dependent variables. The chapter
concludes with the presentation of the results for obser—

vers' degree of confidence in their accuracy judgments.

69
Tgpg

Table 3 presents the number of judgments of "true" and
"false" made of each subject on each tape, summing across all
14 conditions. This table also displays the mean accuracy and
confidence scores for each subject on each tape, as well as
the summed scores from both tapes. Examination of this table
shows that some subjects were judged to be truthing by a
majority of observers both when they were truthing and when
they were lying, and vice versa. For example, subject 13 was
judged to be lying by 74.3% of the observers when he was
truthing and by 69.1% when he was actually lying. Accuracy was
thus much higher when he was lying (.691) than when he was
truthing (.257). Overall, the accuracy with which this subject
was judged was .474, which probably represents a better
accuracy estimate than if only a truthful or a lying sample of
behavior had been included on the stimulus tape. Further
examination of Table 3 reveals similar examples for many of
the 16 subjects.

Table 4 presents the means and variances for factual
accuracy in all cells; Table 5 presents the analysis of vari-
ance for the 24 cells which fit within the factorial design;
and Tables 6 and 7 present this information for emotional
accuracy. Examination of Table 5 indicates a significant main
effect for Tape (2 < .001) on factual accuracy. The tape I
mean for all 14 cells was .567 while the tape II mean was
.512. For factual accuracy, tape also enters into interactions

with both visaudio and shot. These interactions will be

'70

 

cannon

 

 

 

 

 

 

 

 

 

 

«00. wo0w 0wwn ow.0 one. ow.0 «we. conw wwcw wn.0 wwn. nonw nwnw w-aowualu
0H0. eon wwc nw.n w0n. ww.0 n00. n0w wow H 0w.0 «00. nnw «cw u :. 0w
wo0. wnc onw no.0 «00. oo.n ccn. nww own a no.0 nnn. now o0w a z nw
non. nwc won no.n non. wn.n now. new con u ww.0 who. own won a 2 0w
wnw. now wwn w0.0 one. ow.0 wn0. wow wow u nn.0 wnw. no onw a : nw
woe. wnw wwn 00.0 o0n. w0.0 awn. onw oow h 0w.0 can. wow www a : ww
woe. cow nnc oo.0 onn. wa.n 0n0. nnw can a no.0 n00. nnw new m : an
awn. own non 00.0 won. nw.0 woe. wen now a «0.0 own. nnw oew h x on
non. 00w ~00 00.0 0w0. no.0 cwn. now 00w 9 wn.0 own. now nww u r a
3.. .33 2: 8.» a. a.“ .2. 32 9.: 8.. Sn. :2 23 :38
waauunu
woe. 0ow one no.0 one. wn.0 nan. and www a no.n non. nnw nww u z 0
nan. nwc won nw.n nnn. 00.n n0w. onw wn u 00.n n00. w0w oww a x w
w0n. n0w 000 ow.0 0wn. oo.0 occ. 00w 0am H ow.0 wnw. now n0w u x 0
nan. 00w ~00 0w.0 can. on.0 now. wow new a no.0 woo. can can a u n
nn0. o00 w0w nn.0 wn0. 0o.0 own. 00w now a 00.0 own. 09w on a n 0
own. nww nnc no.0 0o0. no.0 one. awn own H no.0 www. cow 0ww u x n
nnn. onw one n0.n wnc. wn.n wn0. www nww w o0.n. wen. nnn new a z w
non. wnc now 00.n nnn. n0.n 0n0. nww www a ow.n n00. wow 0ww h u w
noun «our canon ago: one: can: can: wank conch huh one: new: cans «swan nth nun .
dough wanna .ucoo nonuaou< 0:60 noousuu< 0:00 nunusuu¢
ago-us... guinea. 3.2.6
unuuoh 00 .oz 00 .oz
8.5. 58 S 2...... H 25

 

vol-aw can .oaua no.» so nuns-mum Hosanna-fl can Hosanna Anon now
nouoon oocokuaoo 0c. nuauauoc can: an. =.ouwnh= as. :oaua: no nus-luvan we wanna: .huwuuuo> w-auu< .nou uuonasm

n owaah

7].

\

oen. I coal ocean
.,w0e. I undo quuwp

won. I One:- 0 Anson»

 

nwo.
n0n.

 

nwo.
n50.

 

 

 

ance caea<

 

nwo.
nw0.

 

Owe.
nwo.

 

 

 

HH

HH

nwso guano-aqua

onus

neck

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.oouw- Oneida you Ao0 cacao w ownuh com "ouoze
amen. nae. wan.
o r .. .1 Girl 0 o r. o I. x . , o a.“
one mmw\\ wee nnn one nmn.. Hangman:
one. eno. eno. wwo. nno. wwo. oeo. we ww
0nn w0n o0e wnn ooe nwn m «awn:
own. van
nno. was. «we. use. «no. use. w. noose
con man. «an. nos. con. man. can. m H
0wo. nno. non. nwo. wwo. ono. -
can. a ma
non. wan. 00e. w0n. nne. wo0. m
nnn. ~38
I O O O O O .
onn. nwo oeo eno nwo eno nwo w H
onn. w00. nne. nnn. nen. nw0. m
usual nwso mason» nwco nuauw> nwco woaaw onus
wouwuuux woomw> vco wasow> van wuaaw> van
ow0=< cwva< oat:
neon a use: wane wean wane on»: 1_
yoga

 

cacao-$00 w-suo-u osu uou sumo: wchuuuz van .uoucdwuu> .aououm nouusuu< coo: wwoo

e onaafi

72

Table 5

ANOVA for Factual Accuracy

 

 

 

Sum of Mean Probability
Source of Variation Squares DF Square F Level
Main Effects 2.889 5 .578 18.527 .001
Colorbw .040 1 .040 1.296 .254
Visaudio 2.033 1 2.033 65.189 .001
Tape .563 1 .563 18.068 .001
Shot .342 2 .171 5.487 .005
2—way Interactions .914 9 .102 3.258 .001
Colorbw . .029 1 .029 .940 .999
Colorbw tape .009 1 .009 .275 .999
Colorbw shot .031 2 .015 .492 .999
Visaudio tape .169 l .169 5.428 .019
Visaudio shot .002 2 .001 .036 .999
Tape shot .632 2 .316 10.131 .001
3-way Interactions .322 7 .046 1.473 .173
Colorbw visaudio tape .001 l .001 .021 .999
Colorvw visaudio shot .104 2 .052 1.674 .186
Colorbw tape shot .003 2 .001 .044 .999
Visaudio tape shot .219 2 .110 3.519 .029
4-way Interactions .046 2 .023 .737 .999
Colorbw visaudio tape .046 2 .023 .737 .999
shot
Residual 18.711 600 .031
Total 22.882 623 .037

 

Multiple £2 = .126

'73

won. I came cacao

 

 

.oouwo owns-o now Ao0 cacao w own-a can ”ouozc

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

doe. u oaco ~.:.os . woe. own. owe.
”an. u cows. 0 guano»
O O O O O O 3':
Hoe non own. ”on one coo Hanson-x
“do. 0 O O I O I
no . owe owe woo woo owe one .
ewn. «we . . . . . N HH
one one no. ooo use. see m as
ouch .IIIu «:3
ooe. can
one H «on. one. owo. coo. owo. owo. ENO. w. H xenon
awe “on. gun. coo. «on. was. non. m
sane cause . . . . . .
00¢. coo one one one one cue w. an
woo. woo. one. goo. one. one. m
owo. mama nonou
own. no one. woo. owo. moo. owo. owo. .
non. N H
eons moo. ooo. non. non. moo. nno. m
owe a onus: once anamoo oaco ”mamas ooao aaaooo coop
«on. Hosanna: Huauoo can gasses can Haaooo on.
cause cause can:
:8 3288:. ones a Box .38 oven :8 32. H

«cued-Bum wccoducla can now

 

cans: wocwuuax van .cuocowuoo .couoom nuauauud and: wwoo

0 onnnh

ANOVA for Emotional Accuracy

74

Table 7

 

 

 

Sum of Mean Probability
Source of Variation Squares DF Square F Level
Main Effects 1.002 5 .200 8.901 .001
Colorbw .018 1 .018 .802 .999
Visaudio .020 l .020 .902 .999
Tape .743 l .743 32.995 .001
Shot .227 2 .114 5.044 .007
2-way Interactions .328 9 .036 1.618 .106
Colorbw visaudio .011 1 .011 .476 .999
Colorbw tape .006 1 .006 .282 .999
Colorbw shot .027 2 .013 .590 .999
Visaudio tape .000 l .000 .005 .999
Visaudio shot .077 2 .039 1.718 .178
Tape shot .205 2 .103 4.557 .011
3dway Interactions .173 7 .025 1.100 .361
Colorbw visaudio tape .003 l .003 .152 .999
Colorbw visaudio shot .023 2 .012 .521 .999
Colorbw tape shot .027 2 .014 .609 .999
Visaudio tape shot .119 2 .059 2.636 .071
4dway Interations .001 2 .000 .020 .999
Colorbw visaudio tape .001 2 .000 .020 .999
shot
Residual 13.511 600 .023
Total 15.015 623 .024

 

Multiple 5? - .067

75

described when the results of these theoretical variables
are presented.

Examination of Table 7 indicates a significant emo-
tional accuracy main effect for tape (p < .001). The tape I
mean for all 14 cells was .524 while this figure for tape II
was .470. The tape variable also enters into an interaction
with shot. The nature of this interaction will be described
in the context of describing the effect of shot. Clearly,
then, the rationale for including two versions of the stimu-
lus tape is supported by the results. The results for the
primary dependent variable in this experiment, observer

accuracy, will now be presented.

Accuracy

 

§h9t_

Factual. Table 5 shows a significant main effect for
shot (2 < .005). Inspection of the marginal means in Table 4 4
indicates that those observers who viewed the body-only had a ,
mean of .497 which was less accurate than both head-only
observers, who had a mean of .537 (E'= 2.31, df = 413, E.< /
.05), and head-and-body observers who had a mean of .545 (E
= 2.80, gf= 421, p < .02). Table 5 also indicates a shot
by tape interaction. Table 8 presents the cell means which
resulted in this interaction while Figure 3 displays the
relationship graphically. Examination of Table 8 coupled
with subsequent t tests shows that both the main effect for

shot and the interaction between shot and tape were a result

Mean Accuracy

76
Table 8

Cell Means, Sample Sizes, and Marginal Means for the Tw04Way
Interaction Between Shot and Tape on Factual Accuracy

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Shot
Tape Head Body Head & Body Marginals
E .569 .484 .607
I n 110 103 119 '553
i .504 .510 .483
II n 100 102 98 '499
Marginals .537 .497 .545
.60
.55
.50 "\.‘
.45 _________ Tape I
--------Tape II
4 Main Effect
° 4 A
o I z I

head body head
only only 8
body

Figure 3. Two-Way Interaction Between Shot and Tape on
Factual Accuracy

77

of the cell mean from tape I. Within tape I, the body-only
condition, with a mean of .484, was significantly lower than
both the head-only mean of .567 (3': 3.59, Q: = 218,;p S
.001) and the head-and-body mean of .607 (E.= 5.13, d: = 213,
p ‘<.001). The head-only condition does not differ signifi-
cantly from the head-and-body condition (t = 1.58, df = 213).

Within tape II, an examination of Table 8 indicates
that the cell mean in the body-only condition is largest with
a mean of .510, but this does not differ significantly from
either the head-and-body mean of .483 (t_= .84, d: = 198) or
the head-only mean of .504 (t = .24, g: = 204).

Table 5 also shows a significant interaction between
shot, tape, and visaudio. The cell means resulting in this
effect are shown in Table 9, and this interaction is graphic-

ally displayed in Figure 4.

Table 9

Cell Means, Sample Sizes, and Marginal Means for the Three4Way
Interaction of Visaudio, Shot and Tape on Factual Accuracy

 

 

 

 

Shot
Head Only Body Only Head & Body
And. & Visual Aud. & Visual Audi & Visual
Tape Visual Only Visual Only Visual Only Marginals
I x .598 .540 .551 .418 .641 .574 .554
n 53 57 51 52 57 62

 

.588 .419 .556 .463 .580 .386
II . .499

XI

 

Marginals .593 .480 .554 .441 .611 .480 .527

78

 

 

 

 

 

 

 

 

 

 

 

Tape I
.6
1
\
\
.55 X
\\
>.
H \
3 ,5 1,
U i
<1 ‘ ,
\
8 \V
4’
.4
.4 ‘ r ‘
0
head body head
only only &
body
Tape 11

 

\

 

 

Mean Accuracy
0.

 

 

 

 

 

 

 

visual 6 audio

u-=-=--visual only

.m.----amain effect

.45 /
.4
o I
head body head
only only &
body
Figure 4. Three-way Interaction Between Visaudio, Shot and Tape

on Factual Accuracy*

*
These same relationships are displayed from the visaudio per-
spective in Figure 7, page 83.

79
The Ekman and Friesen (1974) experiment dealt with
the shot variable in a visual—only context. The head-only
visual-only cell mean of .480 did not differ significantly

from the body-only mean of .441 (E_= 1.58, g; = 202).

Emotional. Examination of Table 7 indicates that

 

there is a significant main effect for shot (p’< .007).
Inspection of the marginal means in Table 6 indicates that
subjects in the body-only condition, with a mean of .520,
were more accurate than subjects who viewed either the head-
only (.475) or the head-and—body (.497). Table 7 also indi- -1
cates a significant shot by tape interaction. Table 10
displays the cell means which resulted in this interaction
while Figure 5 displays this relationship graphically. Sub-
sequent 3 tests showed that the main effect for shot was not
general across conditions but was limited to tape I, where
the head-only condition mean of .483 was significantly less
than both the body-only mean of .572 (E_= .431, df = 214, E.<
.001) and the head-and-body mean of .541 (E.= 2.81, g; = 231,
E.‘ .005). The body-only mean was not significantly larger
than the head-and-body mean (5 = 1.52, df = 222). Within
tape II the body-only conditions mean of .467 did not differ
significantly from either the head-and-body mean of .452 (E
= .70, d£_= 200) or the head-and-body mean of .466 (t = .04,
_£_= 201).

Examining those cells in Table 6 which replicate

Ekman and Friesen (1974), the head-only visual-only cells had

80

Table 10

Cell Means, Sample Sizes, and Marginal Means for the Two#Way
Interaction Between Shot and Tape on Emotional Accuracy

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Shot
Tape Head Body Head & Body Marginals
; .483 .572 .541 532
I n 110 103 119
i .466 .467 .452 462
II n 100 102 98
marginals .475 .520 .497

.60
>‘ .55 ‘\
U
0 ,5q ’r’ ‘\x
‘2 15'
g '—
g .45 Tape I

-__. - - ---Main effect
Ta e 11
.40 ‘1 p
o -
head body head
only only &
body

Figure 5. Two-Way Interaction Between Shot and Tape on

Emotional Accuracy

81
a mean of .455 which was significantly smaller than the body-
only visual-only mean of .528 (E’= 3.42, g; = 202, p < .002).
Also, within the visual-only conditions, the head-and-body
mean of .491 does not differ significantly from either the
head-only (E.= 1.74, g; = 216) or the body-only (t = 1.74,
g: = 206).

Visaudio

Factual. Table 5 indicates that there is a main
effect for visaudio (p’< .001). An Examination of Table 4
indicates that for all 12 pairs of cells, observers who were -"
exposed to both the visual and audio information were more ’
accurate than those whose judgments were based on visual -/
information only. The mean for all 12 audio and visual cells
was .585 while the mean for the 12 visual-only cells was
.467.

Table 5 also indicates a visaudio by tape inter-
action. Examination Table 11 and Figure 6, which display
this interaction, indicates that the effect of the visaudio
variable was more pronounced for tape III. Visaudio also
enters into a threedway interaction with shot and tape.
Figure 7 graphically displays this interaction from the per-

spective of the visaudio variable.

Emotional. Examination of Table 7 indicates that

 

for emotional accuracy, no main effects were found for vis-

audio, nor did this variable enter into interactions.

82
Table 11
Cell Means, Sample Sizes, and Marginal Means for the

Two-Way Interaction Between Visaudio and Tape
on Factual Accuracy

 

 

 

 

 

Tape Visual-Audio Visual Only Marginals
I x '.596 .511 .554
n 161 171
II x .575 .423 .499
159 141
Marginals .585 .467
.60

 

 

.50 \
.45 \ Tape I

Main effect
Tape II

 

 

Mean Accuracy

 

 

 

 

 

.40
0 1 ’7
visual visual
& only
only

Figure 6. Two-Way Interaction Between Visaudio and Tape on
Factual Accuracy

83

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Tape I
.60
/ /
I
’I
.55 r3
I
>.
U
o
‘5
o .50
0
.¢
c
o
o
z .45
/ —-—head 6 body
.40 ----- head only
7
0 __ body only
visual visual
only 6
audio
Tape II
. 0
.55
>»
U
o
*5
0 ~50
0
<2
a
m
o
z:
.45
.40
o ’ 1
visual visual
only 6
audio

Figure 7. Three-Way Interaction Between Visaudio, Shot, and Tape
on Factual Accuracy

84

Color/Black and
White (Colorbw)

 

Factual. An examination of Table 5 indicates that
no main effects or interactions were found for colorbw for

factual accuracy.

Emotional. An examination of Table 7 indicates that

 

no main effects or interactions were found for colorbw for

emotional accuracy.

Transcript-Only
and Audio-Only

 

 

Factual. Observers in the transcript-only condition
had a mean accuracy score of .625 (for both tape I and II)
while observers in the audio-only condition had a mean of
.618. This difference is not significant (E.= .22, g: = 184). 1
The visual-only mean accuracy score of .467 for all 11
12 visual cells was significantly less than the audio-only ‘”““
mean of .618 (E = 5.24, d£_= 351, p_< .001). The audio-only >-"
condition also’resulted in higher observer accuracy than all
three visual-only levels of the shot variable. Both the head-
and-body and the head-only conditions had means of .480 (E =
4.51, g£|= 146 and 147, respectively, p < .001), while the
body-only condition had a mean of .441 (E,= 5.67, §£.= 134,
E < .001) .
The visual and audio mean factual accuracy score
across all 12 conditions was .585. This value does not
differ significantly from the audio-only mean of .618 (t_=

1.14, g£_= 360). Within the visual and audio conditions,

85
only the body-only condition, with a mean of .553 differed
significantly from the audio-only mean of .618 (E = 2.12,
QE = 148, p_<:.05). The head-only, visual and audio con-
ditions had a mean of .593 (E,= .81, QE = 141) and the
head-and-body, visual and audio conditions had a mean of

.610 (1; = .26, g; = 146).

Emotional. Observers in the transcript-only condi-

 

tions had a mean accuracy score of .522, while observers in
the audio-only conditions had a mean of .473. This difference
was not significant (E’= 1.71, QE'= 85).

The visual-only mean accuracy score of .491 for all
12 cells was not significantly different from the audio-only
mean of .473. Within the visual-only conditions, only the
body mean of .528 was significantly greater than the audio-
only mean (E_= 2.09, EE’= 135, p_< .05). The head-only mean
of .455 (E.= .69, QE_= 145) and the head-and-body mean of .491
(E_= .70, EE’= 149) did not differ significantly from the
audio-only mean of .473.

The visual-and-audio mean emotional accuracy for all
12 cells was .503, which does not differ significantly from
the audio-only mean (E.= 1.22, EE'= 359). All three levels
of the shot variable within the visual and audio conditions
also failed to differ significantly from the audio-only mean.
Head-only had a mean of .494 (E_= .81, EE = 141), body-only
had a mean of .511 (E = 1.47, EE = 147), and the head-and-body

86
condition had a mean of .503 (E_= 1.16, QE = 147). Results

for observer confidence will now be presented.

Confidence

 

Table 12 presents the means and variances for fac-
tual confidence in all cells; Table 13 presents the analysis
of variance for the 24 cells which fit within the factorial
design; and Tables 14 and 15 present this information for

emotional confidence.

Tag
Factual. Examination of Table 13 reveals that, for

factual confidence, there were no significant main effects or

5

interpretable interactions for tape.

Emotional. Examination of Table 15 indicates a sig-

 

nificant main effect for tape (2 < .025). Tape I had a mean
for all 14 cells of 6.38 while tape II had a mean of 6.12.
Tape did not enter into any interpretable interactions with

the other variables.

Shot

 

Factual. An examination of Table 13 reveals that
there was no significant main effect for shot, nor does shot
enter into interpretable interactions with the other variables.
However, a comparison between the head-only mean of 6.13
with the body-only mean of 5.77 yields a significant E of
2.36 (QE = 413, p < .05).

87

0o.n I anal each»
00.n I Anne nesses

oo.0 1 H6000» 0 6H000

 

0w.w
ww.0

 

oe.w
wa.n

 

 

 

oHoo oHo:<

 

n0.n
an.n

 

on.w
wn.0

 

 

 

Jl

HH

«one

an
coon

wwao uawuuocuua

.oouwc «wale- uou Ao0 cacao w annoy com Hauozc

 

 

 

 

 

 

 

 

 

 

 

 

 

 

oo.0. oo.0 HH.0
. . . O . . 8's
oH 0 Ho 0 on m 0H 0 00 n on 0 HaoHouqx
oH.H oo.H oo.H o~.H 00.o Ho.H .
00.0 oo.0 No.0 oo.0 H0.o oo.0 oo.0 Ha HH
1 «0H0:
om.n can
oo.~ H0.H 00.» HH.~ oo.H oo.H H. soaHo
0o 0 oo.0 oo.0 HH.H HH.0 oo.0 0H.0 m H
oo.o Ho.H o0.H Ho.H 00.0 0“." .
oo.n w HH
00.0 0o.o Ho.0 oo.0 oo.0 H0.0 m
oo.0 06Hoo
H0.H Ho.H HH.H Ho.~ 0o.H H0.H .
0o.0 H H
00.0 00.0 oo.0 Ho.o Ho.n 00.0 m
oHoo H30; :8 H25; HHS H30; 25.
nuanw> vac wcauw> use woaowb . ecu
oHoo< 6Ho=< 600:0
oooo 0 on»: oHoo oooo oHao on»:
0000

 

NH munch

«cued-non ncauoch ago new coco: woodman: to. .neuaewue> .uouoom oucokucoo one:

88

Table 13

ANOVA for Factual Confidence

 

 

 

Sum of Mean Probability
Source of Variation Squares DF Square F Level
Main Effects 52.482 5 10.496 4.419 .001
Colorbw 1.691 1 1.691 .712 .999
Visaudio 39.308 1 39.308 16.549 .001
Tape 2.682 1 2.682 1.129 .288
Shot 10.535 2 5.267 2.218 .108
Zdway Interactions 13.069 9 1.452 .611 .999
Colorbw visaudio 3.559 1 3.559 1.498 .219
Colorbw tape .035 1 .035 .015 .999
Colorbw shot 3.407 2 1.703 .717 .999
Visaudio tape .934 l .934 .393 .999
Visaudio shot 4.619 2 2.310 .972 .999
Tape shot .034 2 .017 .007 .999
3-way Interactions 9.702 7 1.386 .584 .999
Colorbw visaudio tape .029 1 .029 .012 .999
Colorbw visaudio shot 6.082 2 3.041 1.280 .278
Colorbw tape shot 2.830 2 1.415 .596 .999
Visaudio tape shot 1.068 2 .534 .225 .999
4dway Interactions 47.316 2 23.658 9.960 .001
Colorbw visaudio tape 47.316 2 23.658 9.960 .001
shot
Residual 1425.140 600 2.375
Total 1547.708 623 2.484

 

Multiple £2 - .034

£39

nw.0 I soul vain»
oo.n . HHao HoooHo

oo.0 - HuooHo 0 6H0:-

 

ww.w
e0.0

 

en.w
nn.0

 

 

 

HHoo oHoo<

 

ow.n
0a.n

 

ew.e
ne.0

 

 

 

an

an

nHoo uaHuuoccuu

each

cook

I

55
O

.oouwe own-cc How Ao0 cacao w awash com .Houozc

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

HH.0 oH.0 oo.0
Ho.0 Ho.0 00.0 oo.0 Ho.0 Ho.0 .0.»-
H0505.
Ho.0 0o.H Ho.H 0n H oo.H 0H.H oo.H H. HH
Ho n 00 0 Ho 0 oo 0 00 0 0o 0 m ongn
can
HH.0 oH.H Ho.H HH.H 00.H 0H.H Ho.H . H HouHo
HH.0 oH.n oo.0 oo.0 oo.0 oH.0 m
0 O O O O 0 .
HH.0 on H 00 H H0 0 HH H HH H on H H HH
HH.n H0.0 00.0 HH.0 Ho.0 Ho.0 m
06H60
0 O O O O O .
Ho.0 on H 0o H oH H oH H o0 H on H H H
Ho.0 Ho.0 0H.0 0H.0 HH.0 HH.H m
nwoo Hoauw> undo wasoH> nwoo wuacw each
Honow> 0cm waaow> can woaow> can
6Ho=< oHo:< 6H0:
oooo 0 on»: oHoo oooo HHoo one:
0600

 

oucuauoa HIuOHHOIH can sou use»: weowuua: can uuucoHuo> .cuuoom ooaoewucoo can: wwoo

en awash

90

Table 15

ANOVA for Emotional Confidence

 

 

 

Probability
Source of Variation Squares DF Square F Level
Main Effects 98.141 5 19.628 8.255 .001
Colorbw 2.068 1 2.068 .870 .999
Visaudio 69.560 1 69.560 29.254 .001
Tape 11.964 1 11.964 5.031 .024
Shot 18.199 2 9.099 3.827 .022
2~way Interactions 20.041 9 2.227 .936 .999
Colorbw visaudio 1.120 1 1.120 .471 .999
Colorbw tape .030 l .030 .013 .999
Colorbw shot 4.789 2 2.395 1.007 .367
Visaudio tape 6.743 1 6.743 2.836 .089
Visaudio shot 7.034 2 3.517 1.479 .227
Tape shot .780 2 .390 .164 .999
3~way Interactions 16.933 7 2.419 1.017 .418
Colorbw visaudio tape 1.076 1 1.076 .452 .999
Colorbw visaudio shot 6.388 2 3.194 1.343 .261
Colorbw tape shot 1.225 ' 2 .613 .258 .999
Visaudio tape shot 8.859 2 4.429 1.863 .154
4dway Interactions 23.338 2 11.669 4.908 .008
Colorbw visaudio tape 23.338 2 11.669 4.908 .008
shot
Residual 1433.793 603 2.378
Total 1592.246 626 2.544

 

Multiple E? = .062

91

Emotional. Examination of Table 15 indicates a sig-

 

nificant effect for shot (p_< .025). The marginal means in

Table 14, coupled with subsequent E tests reveals the source
of this effect. The head-only mean of 6.50 is significantly
larger than both the body-only mean of 6.10 (E = 2.64, EE_=

413, p_< .01) and the head-and-body mean of 6.12 (E’= 2.55,

QE = 425, p_< .025).

Visaudio

Factual. Observers who based their judgments on both
audio and visual information indicated a mean confidence
level of 6.20 in their accuracy judgments, while those who
had visual information only had a mean confidence of 5.68.
Examination of Table 13 indicates that this difference is
significant (p_< .001). Visaudio did not enter into any sig-

nificant interpretable interaction with the other variables.

Emotional. Observers in the audio and visual con-

 

ditions reported a mean emotional confidence level of 6.58
while those who received visual information only had a mean
of 5.89. Table 15 indicates that this main effect was sig-
nificant (p,< .001). Visaudio did not enter into any sig-
nificant interpretable interactions with the other variables.

Color/Black and
White (colorbw)

 

 

Factual. Examination of Table 13 indicates no main
effect or interpretable interactions for colorbw on factual

confidence.

92
Emotional. Examination of Table 15 indicates no sig-
nificant main effect on interpretable interactions for colorbw
on emotional confidence.

Transcript-Only
and Audio-Only

 

Factual. Observers in the transcript-only conditions
had a mean confidence score of 6.07, while observers in the
audio-only condition had a mean of 6.02. This difference is
not significant (E_= .22, QE = 84).

The visual-only mean confidence score of 5.68 for all
12 visual conditions did not differ significantly from the
audio-only mean of 6.02 (E.= 1.31, §£.= 350). Within the
visual-only conditions, just the body-only mean of 5.38 was
significantly less than the audio-only mean (E’= 2.31, EE =
134, p_< .05).

The visual and audio mean confidence score across all
12 conditions was 6.20, which does not differ significantly
from the audio-only mean of 6.02 (E_= .72, EE = 358). With-
in the visual and audio conditions, none of the shot means

differed significantly from the audio-only mean.

Emotional. Observers in the transcript-only con-

 

ditions had a mean confidence score of 6.20 while observers
in the audio-only conditions had a mean of 6.49. This dif—
ference is not significant (E_= .75, EE,= 85).

The visual-only mean confidence score of 5.89 for

all 12 cells is significantly less than the audio-only mean

93
of 6.49 (E.= 2.36, QE’= 350, p_< .02). Comparing the audio-
only condition mean of 6.49 with the visual-only conditions,
shows both the body-only mean of 5.84 (E = 2.35, QE,= 134,
p_< .025) and the head-and-body mean of 5.82 (E = 2.47, EE =
147, p < .02) to reflect less observer confidence.

The visual and audio mean confidence score across
all 12 conditions was of 6.58 which does not differ signifi-
cantly from the audio-only mean of 6.49 (E’= .35, EE_= 358),
nor does the audio-only mean differ from the three means from
the visual and audio conditions of the shot variable. The
head-only condition had a mean of 6.97 (E’= 1.75, EE = 141);
the body-only had a mean of 6.35 (E’= .52, EE_= 147); while
the head-and-body visual and audio mean was 6.41 (E,= .29,
QE_= 146).

The correlation between accuracy and confidence for
the factual segments was .063, while for the emotional seg-
ments it was .061. Due to the large samples on which these
overall correlations are based, both are significant (2 <
.05). The correlations between accuracy and confidence for
each cell in the design for both the factual and emotional

segments appear in Appendix H.

94

Footnotes

1The analysis of variance assumes that error vari-
ances within treatment conditions are equal. Hays (1963,
p. 379) points out that this assumption can be violated
without serious consequences if the sample size in each con-
dition is equal. An examination of Table 2 reveals that,
while the sample sizes in each cell are close to being equal,
they are not exactly so. The resulting problem is poten-
tially serious since the basic unit of analysis for both the
factual and emotional accuracy are proportions. Winer
(1971, pp. 399-400) points out that in this situation, the
individual cell means and variances are not independent. As
the particular cell mean deviates from .50, the cellksvari-
ance tends to increase. Thus only in the situation in which
the overall null hypothesis were true (there were no treat-
ment effects and all cell means were equal) would the homo-
geneity assumption likely be met. Winer (1971) recommends
stabilizing the variances by transforming the raw proportions
into a value two times the arcsine of the square root of
their original value. Analyses of variance on the accuracy
scores were performed both on the raw proportions and on
these scores after the recommended transformation had been
performed. Both of these analyses yielded almost identical
results. Since using the transformed scores results in
interpretation difficulties, the results to be presented here
are from the analysis of variance performed on the raw pro-
portions. The same analyses for the transformed scores appear
in Appendix G.

2Since the dependent variables were obtained from the
same observers and are thus correlated in the same unknown
way, the significance tests reported are not independent of
one another and no exact probability that at least one of
them will exceed a particular critical level on the null
hypothesis can be calculated (see Kerlinger and Pedhazur,
1973, p. 352). An alternative procedure to the one used here
would be to have conducted a multivariate analysis of vari-
ance (MANOVA) which would have allowed comparison on the
dependent variables taken as a set. This was not done because
while the dependent variables are not statistically inde-
pendent of one another, they are conceptually distinct. For
example, a major purpose of this experiment was to examine
emotional versus factual accuracy across levels of the vari-
ous independent variables and MANOVA would not directly
address such comparisons.

3The analyses of variance were performed using sub-
program "ANOVA" from version 6.0 of the Statistical Package
for the Socigl Sciences (Nie, Hull, Jenkins, Steinbrenner,
and Bent, 1975). When a factorial analysis of variance is

 

95

performed on data with unequal cell sizes, as in the case of
this experiment, the effects of the independent variables

are not orthogonal (i.e., joint effects will exist). The
technically correct solution to this problem would be to
perform a series of Anovas, varying the order in which the
independent variables were entered. A series of Anova

source tables would result, each of which would be slightly
different depending on which variables were entered first
and consequently which variable was assigned the joint effects.
The SPSS Anova program performs an unweighted analysis of the
cell means which results in just one source table. This
method is an approximation technique in which each cell is
treated as if it contains only one observation. An estimate
of the extent to which this approximation technique deviates
from what would have been attained if cell sizes had truly
been equal is obtainable by comparing the true total sums of
squares with the summed sums of squares from each source of
variance in the analysis used in the approximation technique.
(The true sums of squares is found by multiplying the vari-
ance of the total sample by sample size.) The true SS for
factual accurary is 22.88, while the SS in the analysIs per-
formed is 23.43. For emotional accurac the figures are
15.37 and 15.02. For factual confidence and emotional confi-
dence, respectively, the figures are: 1575.17 and 1547.71;
1608.84 and 1592.25.

These comparsions indicate that the source tables to
be presented here are nearly identical with what would have
been reported if the sample sizes had been equal for every
cell.

4Inasmuch as this experiment is exploratory, it is
desirable to make some comparisons which were not specifically
planned. Scheffe (1953) has suggested a test that is appro-
priate for making any and all comparisons that may be sug-
gested by the values of the cell means themselves with a
probability of 1 minus alpha that Ell statements concerning
the significance of these comparsions will be true. This
procedure involves performing a standard E test (using the
MS from the overall Anova) but requiring a critical E equal
to or greater than the square root of the number of cells in
the experiment minus one, times the critical E required for
significance from the original Anova. The critical E_for an
alpha of .05 in this experiment (df = 27,600) is about 1.6.
This yields a critical E for Scheffe's test of about 6.4. For
an alpha of .01 the E required is about 7.0. Scheffe's test,
then, would require an extremely large difference in cell
means to yield a significant result since there are so many
possible comparisons against which to protect an overall
experiment-wise alpha. Using such a conservative test.would
probably result in failing to detect many "true" relation-
ships which exist in the population. Consequently, the pro-
cedure which will be followed here will be to perform the E

96

tests and report the resulting E values and the two-tailed
probability. Any comparison which exceeds the .05 level
two-tailed will be considered significant. The reader should
be alerted to the fact that this procedure does not provide
experimentdwise error protection and it makes the probability
of committing Type I error on any given test considerably
higher than if Scheffe's procedure were strictly followed.

5For both factual and emotional confidence, four-way
interactions were found. These interactions were extremely
difficult to describe verbally and consequently such a
description will not be attempted. The nature of these
interactions are represented numerically by the 24 cell means
in Table 12 and 14. These interactions will be discussed in
the next chapter.

Chapter IV

DISCUSSION

The purpose of this chapter is to interpret the
results of the experiment, to discuss the practical and theo—
retical implications, to evaluate the procedures, and to sug-

gest future research.

Accuragy

Shot

 

The rationale for this exploratory experiment draws
heavily on the work of Ekman and Friesen (1969a; 1974). A
major purpose was to replicate their (1974) reSearch using
samples of lying and truthing behavior from another kind of
dishonest interaction in addition to replicating under con-
ditions approximating their original procedures. For emo-
tional accuracy, the body-only condition resulted in signi-
ficantly higher accuracy scores than did the head-only
condition. Examination of the marginal and cell means in
Table 4 indicates that while this finding exists as a main
effect, it is the result of the accuracy scores in the
visual-only conditions of tape I. It is perplexing that *
this effect is only found for tape I. A reasonable possi-

bility is that only some subjects emitted more body than
97

98

head leakage when they were exposed to the stressful slides.
By chance these individuals may have ended up lying on tape I
and truthing on tape II, and consequently, accuracy would be
higher on tape I.

For emotional accuracy, then, the results here pro- !
vide relatively unambiguous support for the Ekman and Friesen 3
(1974) head-body finding.1 For factual accuracy a different
situation exists. The head-only condition observers were
significantly more accurate than were the body-only observers.“*
Within the visual-only conditions, this difference approached
but failed to reach significance (E = 1.58, EE = 202, p_<
.10). This finding makes questionable the generalizability
of the Ekman and Friesen (1974) head-body finding. The body ‘
may be a better source of cues which are indicative of decep- 1
tion than the head only when the deception involves the con- A
cealment of emotional responses. When the lying and truthing t

behavior involve statements about factual events, greater '5

accuracy resulted from judgments which were based on the 1

head only. It is thus likely that the head is the best 7
\

source of information in these situations. Probably few

body cues were emitted by subjects who were lying or telling k

the truth about the factual events. I
For emotional accuracy, if the body provides the

greatest source of cues for detecting deception concerning

reports of emotional feelings (as the results of both the i

Ekman and Friesen [1974] study and this experiment suggest),;

then the presence of the head cOuld distract observers from 4

99
noticing these cues. It is natural to watch the head of
someone who is talking, rather than concentrating exclusively ‘
on the body as observers in the body-only condition were *
forced to do. Observers in the head-and-body condition,
with a mean of .497, were less accurate than those in the
body-only condition, who had a mean of .520. This difference
approaches but fails to reach significance both as a main
effect (E_= 1.56, EE_= 419, p_< .15) and within the visual-
only conditions, where the head-and-body mean was .491 and
the body-only mean was .528 (2.: 1.74, QE = 206, p_< .10).

For factual accuracy, the body apparently did not pro- 1
vide the best source of information about veracity, since
observers were more accurate in the head-only condition. The
accuracy level in the head-and-body condition (.545) was
almost identical to the head-only (.537) as a main effect,
and within the visual-only conditions these means were identi-
cal (.480). If observers who were shown the head-and-body
were distracted from watching the body, this distraction
would not detrimentally affect their accuracy score unless
the body were a good source of information about truthfulness.
In fact, it would be advantageous for the observer to be dis-
tracted from paying attention to an area of the body which
was not emitting cues indicative of deception. Observers in x!
the head-and-body condition could observe less head detail 4
than their counterparts in the head-only condition, yet their A'

accuracy scores were just as high. Several possibilities

could account for this finding. Accurate judgments could be

100
based upon gross head movements, rather than head detail, and
if so, both the head-only and the head-and-body observers
may have been able to notice these movements. Another possi-
bility is that the body may have provided enough information
about deception to counteract the information lost by head-
and-body observers because of the unavailability of head
detail. A final possibility is that inconsistencies between
the nonverbal behaviors being emitted by the head and the
body, which would have only been available to the head-and-
body observers} may have enhanced their accuracy scores. The
design of this experiment does not provide evidence bearing
on these explanations.

It is clear that emotional accuracy was higher when
judgments were based on the body-only than when based on the
head-only, whereas for factual accuracy an opposite result
was found. The head-and-body accuracy scores are consistent‘
with the interpretation that for emotional accuracy the best
source of cues is the body, but that for factual accuracy the
body EE£.§2 is not the best source of information. The
overall pattern of results on the shot variable can be taken
as a potentially important refinement and extension of the
Ekman and Friesen head-body hypothesis about leakage and
clues to deception.2

Before turning to a discussion of the other variables
in the study, two qualifications about the meaning of these
results need emphasis. First, there is a temptation to con-

clude that the head-body results presented here provide

101
strong evidence that the body is the best source of informa-
tion about emotional deception, while the head is the best
source of information about factual deception. Although
there is some basis for this conclusion, another possibility
cannot be completely discounted. As was noted earlier, cues
indicative of deception are useful to observers only if they

are both noticed and correctly attributed. Observers'

 

beliefs and stereotypes about how liars and truthers behave
may be grossly inaccurate. Thus, for factual accuracy, the
body-only observers may have been less accurate than their
head-only counterparts because some bodily cues actually

indicative of deception were incorrectly attributed to truth-

 

ing, and/or vice versa. Thus, the results of this experiment
do not necessarily mean that the "best" source of cues denot-
ing deception are those areas where accuracy is highest. If
observers believed that many body movements indicated lying
and few body movements indicated truthing, and the Opposite
were true, accuracy scores for body-only observers would be
low even though cues indicative of deception were plentiful.
Conversely, fewer cues might exist in the head-only condition,
but these cues might be more consistent with observers'
beliefs about the behavior of lying individuals, and conse-
quently, relatively higher accuracy scores would be likely.

A second qualification concerns the familiarity
sample of subjects' behavior which was presented to observers.
In this segment, observers knew that subjects were always

telling the truth. Thus, it is not clear if observers saw

102
more cues indicative of deception in one condition than in
another, 2E if the deception behavior deviated from the
truthful familiarity sample more in some conditions than in
others. In other words, observers may not have known if
specific cues were indicative of deception or not, but if
a particular behavior were different in the test segment than
in the familiarity sample, they may have correctly inferred
that this difference was a result of lying.

Several factors suggest that this latter possibility
is at least partially accounting for the results. If obser-
vers were able to identify cues indicative of veracity more
accurately from some parts of the body than from others,
Ekman and Friesen (1974) should have been able to obtain the
predicted head-body differences even when no familiarity
sample were provided. As was noted in Chapter I, they could
not. Furthermore, the inconsistencies and contradictory find-
ings in previous studies employing content analyses of lying
and truthing behavior add support to this interpretation.
Perhaps lying behavior is highly idiosyncratic. If so, theflmm
best way to detect lying would probably be to compare samples
of known truthful behavior with samples of doubtful veracity.
Behavior which deviated from known truthing behavior would *’
then result in an inference of lying. To the extent that
this reasoning is correct, the results of this study do not
provide evidence about which categories of behavior contain

specific cues indicative of veracity, but rather which

1,-..—

103
categories contain behavior which deviates the most from

truthing to lying.

Colorbw

No accuracy differences were observed for the color-
bw variable. Several features of the overall pattern of
results, however, dictate caution in concluding that obser-
vers were equally accurate when watching the tapes in color
and in black and white. The colorbw variable would be most
likely to make a difference in the head-only condition where
observers were viewing a close-up of the subject's head.
Within this condition, all eight pairs of cell means (four
pairs for each dependent variable) were higher in color.
Within the other levels of the shot variable there is no con-
sistent pattern of either color or black-anddwhite resulting
in different levels of accuracy. For factual accuracy, com-
paring the head-only color mean of .550 with the black and
white mean of .523 yields a non-significant E’to 1.12 (EE =
208). A similar finding results from the same comparison
for emotional accuracy, where the color mean was .493 and
the black-and-white mean was .457 (E = 1.73, EE = 206, p <
.10). Although not technically appropriate, combining the
two accuracy dependent variables results in a color head-only
mean of .521 and a black and white mean of .490. This com-
parison is significant (E'= 1.98, EE = 416, p_< .05). This
finding, plus the consistency of the pattern of color result-

ing in greater accuracy than black and white in the head-only

104
conditions, suggests that for these conditions color may
improve observers' ability to detect deception.

Whether there are color cues such as facial flush
which result in greater accuracy in the color conditions or
whether observer interest or motivation is higher in color,
which in turn results in greater accuracy, is unknown. Data
bearing on this latter possibility exist in the form of
observer responses to a questionnaire item which asked the
extent to which they found the task of attempting to identify
the veracity of the subjects interesting. Responses to this
question indicate no relationship between format and obser-
vers' reports of their interest in the task. For the head-
only conditions, black-anddwhite observers indicated a mean
interest of 6.19 while color observers had a mean of 6.16 (E
= .07, 92.: 210).

Part of the rationale for including the colorbw
variable was to increase the potential applicability of the
findings, specifically to the court system. If, in the
future, trials or portions of trials are presented to jurors
on television, the results of this experiment provide some
evidence that jurors may be able to detect lying more accu-

rately in a color than in a monochromatic format.

Paralinguistic Cues

 

Perhaps the most perplexing findings in this experi-
ment are the results associated with the paralinguistic cues.

For both factual and emotional accuracy, observers who were

105
exposed to subjects' behavior in an audio-only format did not
differ significantly from those in the transcript-only con-
ditions, although for emotional accuracy, transcript-only
observers approached being significantly more accurate than
those in the audio-only conditions (E’= 1.71, EE = 85, p <
.10). Several possibilities could account for this lack of
differences. Perhaps there were no paralinguistic correlates
of lying which permitted accurate inferences by observers. A
second possibility is that the paralinguistic characteristics
associated with the audio format distracted observers from
noticing aspects of the verbal content which enhanced the
accuracy of those in the transcript conditions. A final possi-
bility is that the paralinguistic characteristics of the
behavior confused observers. For example, in some cases they
may have inferred erroneously that nonfluencies were a result
of lying when, in fact they were merely a result of the sub-
jects' general nervousness as they were being interviewed by
the police officer. The situation in which the stimulus
tapes were made was designed to create high saliency and ego-
involvement on the part of subjects. The effort to
heighten saliency and involvement may have been too suc-
cessful; that is, many subjects were visibly nervous during
their interviews both when they were lying and when they
were truthing. To some extent observers in the audio-only
condition may have correctly inferred veracity based on para-
linguistic information, but errors resulting from subjects'

general anxiety level may have reduced their accuracy scores.

106

The strongest finding in the entire experiment was
that, for factual accuracy, observers in the visual and audio
conditions were more accurate than those in the visual-only
conditions. Several factors mitigate against concluding 7
that the paralinguistic cues are providing observers with the
best information from which to infer veracity. First, deSpite
efforts to reduce the variability of the content of the sub-
jects' lying responses, there was still more variability than
when they were truthing. Also, subjects in some cases pro-
vided more detailed responses when truthing than when lying.
The fact that the factual accuracy for the transcript-only;
observers were so high (.625) suggests that the verbal content
itself did contain information from.which to judge veracity
accurately. This content, of course, was not available to
visual-only observers. Furthermore, an error was made in
creating the factual accuracy questions which the police
officer asked the subjects. One question asked if Mr. Bostick,
the person they had seen on the videotape before their inter-
view, had the right to appeal his conviction. The last
question in the factual segment asked what Mr. Bostick's
reaction was to being told that all he had to do to appeal
his conviction was to sign a form. Thus, this question sug-
gested the answer to the earlier question. No subject
appeared on either tape answering both of these questions.
However, it is possible that some observers noticed this

inconsistency between the questions and correctly inferred

107
that any subject who said that Mr. Bostick did not have the
right to appeal his conviction was lying.

No such problem with either the questions or the
variability of the answers exists for emotional accuracy and
no main effect or interactions were found for visaudio.
Whatever information the audio band contains which was rele-
vant to the identification of veracity apparently was not
useful to observers. The audio band may have distracted
observers from noticing visual cues, or vice versa. Or what-
ever aid the paralinguistic information provided observers
may have been cancelled by inaccurate judgments resulting
from misinterpreting these cues. In any event, visual-only
observers were just as accurate as visual and audio observers.

Comparing the audio-only conditions with the visual-
only conditions results in a non-significant trend in favor
of greater emotional accuracy for the audio-only observers.
The highest visual-only accuracy within the shot variable was
found in the body-only condition. This mean of .528 was sig-
nificantly higher than the audio-only means of .473 (E =
2.09, QE = 135, p.< .05). It is unknown if there were more
cues indicative of deception emitted by the body than by the
voice, or if perhaps observers' beliefs about which cues
were indicative of lying were more accurate for the body than

for the voice.

108

Emotional Versus
Factual Accuragy

 

 

Summing across all 28 cells in the experiment, face- I

tual accuracy was considerably higher than emotional accuracy
(.541 vs. .499). It should not be concluded from this com-
parison, however, that more cues indicative of deception
existed for the factual segments or that these cues were more
consistent with observers' beliefs about which cues are asso-
ciated with deception than cues which were emitted by sub-
jects during the emotional segments. These factors mentioned
which contributed to the audio and visual observers having
higher factual accuracy scores than visual-only observers are
also likely to be accounting for the overall factual accuracy
scores being higher than the emotional accuracy scores. The

visual—only factual accuracy mean of .467 was actually lower

.4

than the emotional accuracy mean of .491. It seems clear that

the verbal content of the audio and visual conditions is also
accounting for factual accuracy being higher than emotional
accuracy.

A further problem which makes comparisons between the
factual and emotional accuracy scores difficult is a possible
order effect. The factual segments were always presented
first. Consequently, subjects may have become bored with the
task by the time the emotional segments appeared, and thus
have been more accurate for the factual segments or alterna-
tively, they could have become better at making the judgments

with practice and thus could have been more accurate for the

109
emotional segments. Randomizing the order of presentation
would probably have been desirable but this would only have
been possible if observers could have been run in smaller
intact groups or individually. It should be noted that
since the order of presentation was a constant, it would not
influence comparisons between the experimental conditions

unless order interacted with the manipulated variables.

Confidence

 

For both factual and emotional confidence there was
a highly significant main effect for visaudio (p_< .001 for
each). Observers who both heard and viewed subjects' beha-
vior were more confident in the accuracy of their judgments
than their counterparts in the visual-only conditions. The
audio-only observers were significantly more confident for
the emotional segments than either the body-only or the
head-and-body observers in the visual-only conditions. A
similar pattern of results was found for the factual segments
except only the body-only mean within the visual-only cells
was significantly less than the audio-only mean. Apparently
the presence of the audio band increases observers' degree
of confidence in their judgments and observers are more con-
fident when they have just audio information than when they
have just visual information, except when the visual informa-
tion is in the form of a close-up of the head for the
emotional segments, and when the head is available in any

form for the factual segments.

110

For emotional confidence, there is a significant
(E.< .025) main effect for shot with the head-only conditions
having a significantly higher mean than either the head-and—
body or the body~only. No similar main effect exists for
factual confidence; however, a E_test indicates that obser-
vers in the head-only condition were significantly more con-
fident than were those in the body-only condition (E_= 2.38,
ggp= 413, pl< .02). Thus the overall pattern of results for
shot on both confidence dependent variables indicate that
observers are most confident when judgments are based on a
close-up of the head, next most confident with the head and
body, and least confident when judging the veracity of a head-
less body.

A color vs. a black-andrwhite format apparently has
no effect on observer confidence. Even combining the two
dependent variables and examining the head-only level of the
shot variable fails to yield a significant color vs. black-
and-white comparison (E_= 1.16). It should be noted that
there is a slight overall trend for greater confidence result-
ing from the color conditions.

A particularly difficult finding to interpret is the
fact that significant four-way interactions were found for
both factual and emotional confidence. The procedure for
marking the scales on which observers indicated their degree
of confidence in their accuracy judgments was not explained
on the questionnaire in detail. Consequently in several con—

ditions, observers asked questions about how to mark these

lll

measures. Deviations in the nature of these questions and
in the manner in which they were answered by the experimen-
ters may have created systematic biases towards indicating
more or less confidence in some cells than in others. Inas-
much as it is difficult to imagine theoretical reasons for
such complex interactions, it seems plausible that some
random factors were contributing to confidence ratings. The
lack of specificity in the directions for filling out the
confidence scales may have contributed to this. More detailed
instructions probably should have been provided on the ques-
tionnaire.

The relationship between accuracy and confidence was
a very weak one. Within most of the cells of the design for
both the emotional and factual segments, the correlations do
not exceed what would be expected by chance (see Appendix H).
Overall, including all 730 observers, there were modest cor-
relations of .063 for the emotional segments and .061 for the
factual segments. Because of the large sample size, these
correlations are significant. These small correlations are
consistent with the results of the only other study which has
examined the relationship between accuracy and confidence
(Matarazzo 22.213! 1970). This study failed to find any
relationship between these variables.

Observers probably have beliefs about which categories
of behavior provide the best information about deception. The
weak association between accuracy and confidence may be evi-

dence that those beliefs are largely wrong. When observers

112
thought they were being provided with a category of behavior
from which cues to deception were likely to be emitted, they
were more confident of their judgments than when other cate-
gories of behavior were presented. For example, the eyes
have been thought to provide accurate information about decep-
tion. When observers were provided with a close-up view of
the eyes, as in the head-only conditions, they indicated
more confidence than when the eyes were either not seen at
all, as in the body-only conditions, or were difficult to
see, as in the head and body conditions. Thus eye observa-
bility may have at least partially accounted for the confi-
dence ratings varying across levels of the shot variable. If
eye behavior were unrelated to veracity, this would contribute

to low correlations between accuracy and confidence.

Problems with the Stugy

 

The purpose of this section is to describe weaknesses
in this experiment and to suggest ways in which they might
have been corrected. This presentation will be divided into
two sections: weakness associated with creating the stimulus
tapes; and weaknesses associated with the data collection pro—

cedures.

Creating the Tapes

 

The decision to use criminal justice majors as sub-
jects stemmed from the need to create the perception that
important consequences would result from detection and

because of the need to replicate the Ekman and Friesen (1974)

113
procedures as closely as possible. Criminal justice is con-
cerned with lie detection, and it is possible that many of
the subjects who participated had more knowledge about lie
detection than the average person. This knowledge could
have affected the manner in which these particular subjects
approached the task of attempting to deceive the officer.
Using this subject population obviously limits the confidence
with which the findings reported here may be generalized to
other lying and truthing individuals. There is no way of
knowing, at this point, the extent to which these subjects
may have had knowledge about lying and its detection or the
extent to which this knowledge did affect their own behavior
during their interviews. It probably would have been desir-
able to have attempted to systematically gather information
from subjects about these possibilities.

The attempt to create the perception in subjects that
serious negative consequences for them would result if they
were detected was apparently quite successful. In general,
subjects were quite anxious about their performance on this
task. Unfortunately, this anxiety may have been observably
manifested both when subjects were lying and truthing. Thus
the overall accuracy scores may have been reduced by the
erroneous inference by observers that these indications of
nervousness were a result of lying when they were actually
indicative of nervousness resulting from the situation itself.
The two color cameras, the use of a television studio, the

police detective and other characteristics of the taping

114
environment probably contributed to this. Furthermore, the
familiarity segments which were shown of each subject before
observers saw the "test" segments, were the first segments
taped. Thus they were taped when any general anxiety sub-
jects were experiencing was likely to be at a maximum. It
probably would have been better to have taped the truthful
"baseline" segments at a latter point in the interview to
increase the likelihood that subjects would appear relaxed
during this segment.

It has been implied here that the attempt to create
the belief in subjects that serious consequences would result
for them if they were detected was highly successful. The
only evidence for this, however, is the subjective impres-
sions of the researchers. No subjects indicated suspicion
about the cover story at any point and all appeared to take
their task very seriously. This subjective impression should
have been verified with a more systematic assessment of sub-
jects' beliefs about their participation. An item on the
subjects' post-interview questionnaire should have asked
them about their beliefs about the consequences resulting
for them from detection.

The police detective who interviewed the subjects
asked the questions in a non-threatening way. It may have
been desirable for him to take a more adversarial role; that
is, to have verbally aggressed against the subjects to try

and upset them. Lying subjects may have become more upset

115
than truthful subjects and more cues to deception may have

existed, thus providing a better test of the hypotheses.

Data Collection

 

One major problem associated with data collection
was the use of intact groups. This procedure created two
related problems; statistical, which are relatively minor;
and rival hypotheses for the results, which are serious.

The most important assumption of all inferential sta-
tistical techniques is that the sample is random and that the
observations are assigned to particular experimental condi-
tions independently. Using intact groups violates this
assumption with the result that the degrees of freedom in the
analyses of variance are inflated. Instead of having a total
of 623 total degrees of freedom, there would be 23. Each
cell would necessarily be treated as if it had just one
observation. Consequently, no within group variance could
exist and the error term with which to test the main effect
hypotheses would be the mean square associated with the
interaction term. This procedure requires the assumption
that there will be no interaction and thus the interaction
MS is assumed to be error. The F's required for significance
would be considerably larger for all tests. Overall, this
procedure, while statistically accurate, would be conserva-
tive and especially inappropriate given the exploratory
nature of this experiment.

Of a more serious nature is the threat to the inter-

nal validity of the experiment that the use of intact groups

116
creates. Each intact group had a short history (about 2-3
weeks) together as members of an introductory communication
class. In these classes, communication research is some-
times discussed. These discussions could affect the par-
ticipants' attitudes toward research which, in turn, could
have influenced their motivation and interest in the task.
It is further possible that class discussions of nonverbal
communication could have influenced the observers' abilities
to detect deceptive communication. If observers had been
assigned at random, these possibilities would have been much
less serious because it would be likely that observers with
a common history would be represented about equally in each
of the cells of the design.

A further problem*was that the treatments themselves
were, for the most part, administered to all observers with-
in a cell at the same time. The existence of "intra-
audience" effects has been well documented (cf., Hocking,
Margreiter,& Hylon, £2.E£S§§)° Individuals' evaluations of
messages and sources are influenced by the evaluations and
observable responses of those around them when the message is
received. Observers in this experiment were explicitly told
to make their own judgments of subjects' veracity and not to
talk at any point while the tape was being played. In spite
of these precautions, some observers occasionally made
comments which could have affected the responses of other
observers. In at least one group it was necessary for E1 to

ask for such comments to cease.

117

Each of these factors is a rival hypothesis for the
results of this experiment. So, for example, it is possible
that the group which had the highest factual accuracy score
of .682 (head-and-body, audio and visual, color, tape I) had
attitudes and knowledge which were unique to them as a group
which increased their accuracy relative to other groups. Or
the group which had the lowest emotional accuracy score of
.417 (head-only, visual-only, black and white, tape 1) may
have had (a) particularly influential class member(s) emitting
nonverbal, and perhaps even verbal, responses which caused
other observers to error in judging veracity. This problem
would have been solved if observers were assigned indepen-
dently at random to the 28 cells of the design and exposed
to the stimulus in isolation (i.e., one at a time).

It should be noted that if the use of intact groups
were seriously influencing the results of this experiment, it
would be unlikely that the main effects which have been
reported would be as systematic as they are. It is fortu-
nately improbable that the use of intact groups had a major
affect on accuracy judgments. The point, however, is not
that this procedure did or did not have an affect on the
results, but rather that the affect is unknown. If a com-
pletely random observer assignment procedure had been used
and if observers had been run in iSolation, such an unknown
would not exist.

Observers were run over an eight-day period. Thus

it is likely that many of the observers who had already

118

participated interacted with others who had not yet partici-
pated. If these interactions included discussions of this
research, it is possible that later participants' knowledge
of the procedures influenced their performance. Since the 28
cells were assigned (almost) at random to the particular
classes and days, it is not likely that this factor could
have influenced the results systematically.

Each of these problems represents clear weaknesses in
this research. Confidence in the results needs to be tempered

accordingly.

Future Research

 

A reoccurring theme throughout this paper has been the
relationship between detection accuracy and the amount of
information contained within a particular presentational for-
mat. It was pointed out that more information within a level
of an independent variable could either increase accuracy by
providing additional cues from which observers could correctly
infer veracity, or decrease accuracy by distracting observers
from noticing cues to deception. Each level of each variable
in this experiment might be classified according to its total
amount of information. For example, the visual-and-audio
conditions contain more information than do the visual-only
conditions; the head-and-body conditions contain more infor-
mation than the body-only conditions; and the color conditions
contain more information than the black-anddwhite. If these

levels could be quantified along a single dimension of "total

119
information" for each subject, the relationship between
accuracy and information could be more precisely examined.
It might be found, for example, that the relationship is
curvilinear: additional information increasing accuracy to
a point, but decreasing it beyond this point.

Probably no variable in the Communication discipline
has received as much research attention as "credibility." The
present study also has relevance for this variable. Situ-
ations in which a dichotomous judgment of veracity are made
by observers are probably rare. Situations in which the rela-
tive credibility of a source is judged are quite common.
These behaviors from which observers in the present study
infered lying, would probably also lower observers' percep-
tions of credibility in situations where a dichotomous judg-
ment of veracity was unlikely. An interesting research
approach might be to present the stimulus tapes from the
present study to observers and ask them to assess the credi-
bility of each, then relate these evaluations to judgments
of veracity made by comparable observers.

In detecting deception in everyday life, the extent
to which observers rely on observable behaviors is unknown.
Before an inference of lying.is made, certain situational
variables probably need to create the suspicion of deception.

The most obvious situational factor would be the observers' .

0...
FM“-
“w‘

perceptions of gains to be made by the source by presenting

v". w .-

particular verbal statements. A criminal suspect obviously

has something to gain by denying the crime, thus the police

120
officer is suspicious of the denial. The subordinate has
gains to be realized by flattering the superior, thus the
superior is suspicious of the flattery. The same subject
behaviors emitted in different situations would be likely
to result in different observer judgments. For example,
speaking non-fluencies in one situation might result in an
inference of lying, in another of nervousness, and in another
of low credibility. The cause to which the behaviors were
attributed by observers would probably affect the type of
evaluation they would make. Attribution theory may provide a
useful perspective from which to address the role of situation
Iquin detecting deception. This approach would involve manipu-
lating situational characteristics to affect the cause to
which subject behaviors were attributed, and examining the
effect on judgments of veracity.

Nonverbal communication has received extensive
research attention in recent years. The importance of the
nonverbal component of communication messages has been empha-
sized by numerous writers to the point that such proclama- ‘
tions have become accepted truisms. Mehrabian (1968) has
gone so far as to say that 93% of the impact of messages
comes from the nonverbal dimension. The results of the
present experiment suggest that for accurately detecting
deceptive communication, the nonverbal component is lEEE
important than the verbal content. For both the factual and
emotional segments, accuracy was higher in the transcript-

only conditions which contained only the verbal content, than

121
in conditions which contained either visual or paralinguistic
information. These results are consistent with the findings
of Maier and Thurber (1968) that readers were more accurate
than were either listeners or watchers in detecting deception.
If, as these results suggest, there are circumstances in i
which nonverbal behaviors can distract observers from accu-
rately decoding the verbal meaning in a message, there may be
an important implication for the court system. Proponents of
the use of videotape have suggested that videotaped records
of trial proceedings would be superior to the traditional
transcript. One advantage, they say, would be that during
deliberations jurors could review testimony on television,
thus being able to examine the nonverbal behavior of the wit-
ness in addition to having the verbal content. These results
suggest that the nonverbal dimension of the behavior could
distract the jurors from carefully scrutinizing the verbal
material. Research examining this possibility in a legal
context would be of value.

Ekman and Friesen (1974) make the obvious suggestion
that their results need to be replicated with additional sub-
jects and observers. This is true particularly with respect
to subjects, since both Ekman and Friesen (1974) and the
present experiment used subjects from very homogeneous popu-
lations. Replications using stimuli from other kinds of
dishonest interactions would also be of value.

Individual idiosyncracies may be contributing to the

confusion associated with the results of those studies

122
employing content analyses. A line of research suggested by
the results of this experiment would involve a modification
of the content analytic approach. If, as has been suggested
above, there are few specific lying behaviors which are
general to many deceivers, it may be that categories of

behavioral deviations from lying to truthing are more general.

 

For example, Knapp EE.E£- (1974) found no differences between I
the number of "eye contact units" between liars and truthers. ”
_Perhaps some lying individuals have more of those units than
when they are truthing, and others less. The category itself
still could be a good source of information if deviations

within the category for 232E individual were examined, rather

than summing the number of behaviors across individuals. Such

a lying to truthing deviation content analytic study on the

 

stimulus tapes from the present experiment is intended. This
is not to suggest that content analyses looking for specific
behaviors across individuals should be abandoned. Possibly
when these analyses are conducted using stimulus materials
created in situations which have non-trivial consequences

for the subjects, more consistency in the findings will
result.

As discussed earlier, it is impossible to infer from
the results of this experiment which categories of observable
behavior contain the best cues to deception since observers'
beliefs about the behavioral correlates of veracity are
influencing their accuracy. I have been able to find no study

in the deception literature which examines beliefs about how

123
liars behave. Maier and Janzen (1967) asked observers to
list reasons for their judgments of lying or truthing and
found no relationship between accuracy and the reasons given.
As the authors point out, observers may not have known why
they were making particular judgments and generated reasons

in a pgst-hoc fashion to justify their decisions. Research

 

examining beliefs about what observable behaviors are associ-
ated with lying would probably be of value. The materials
for one such study exist within the present experiment and I
intend to pursue this question. By combining information
about the total number of judgments of true and false for each
subject in each condition, with a content analysis of each
subject's behavior, it should be possible to identify the
specific behaviors from.which observers are inferring veracity.
The data generated by this experiment will also allow
several questions related to other aspects of deceptive com-
munication to be addressed. Demographic information was
collected from all observers. It is thus possible to examine
both the degree of detection accuracy and confidence in
judgments by various subgroups within the overall sample. It
will be possible to identify the characteristics of these

observers who were most adept at judging veracity.

SummaEy

The following is a summary of the major results.

1. The Ekman and Friesen (1974) finding that obser-

vers who based judgments on the body-only were more accurate

124
than those who based their judgments on the head-only was
replicated for the emotional segments. For the factual seg-
ments, a strong trend in the opposite direction was
observed. These results suggest that the original finding
partially resulted from subjects' emotional responses to the

stimulus films themselves.

2. The accuracy scores for the head-and-body con-
ditions suggest that the head distracts observers from
closely viewing the body. For the emotional segments, where
the body apparently provided the best information for making
inferences about deception, the presence of the heads thus
reduced accuracy. For the factual segments, where the head
apparently provided the best information about veracity, the

presence of the head tended to increase accuracy.

3. Some evidence indicates that within the head-only
conditions of the shot variable, accuracy is higher in color
than in black and white. The colorbw variable had no effect
on accuracy for the other conditions of the shot variable.
Although it is not clear whether this effect within the head-
only conditions is a result of color cues EE£.§E or a result
of greater observer interest in observing the color format,
it appears that observers were equally interested in viewing

both formats.

4. For the factual segments, audio information
resulted in greater accuracy than visual information. The

fact that the transcript-only factual accuracy scores were

125
also high suggests that it was the verbal content of the
behavior, rather than paralinguistic information, which

resulted in this increased accuracy.

5. For the emotional segments, there was no rela-
tionship between visaudio and accuracy. Overall observers
in the visual—only conditions were not more accurate than
audio-only observers. However, within the body-only visual-
only cells, the mean accuracy score was significantly larger

than the audio-only mean.

6. For both the factual and emotional segments,
observers who received audio and visual information were sig—
nificantly more confident in their accuracy judgments than
those who received visual information only. Head-only
observers were more confident than head and body observers,

while body-only observers were the least confident.

126

Footnotes

1Ekman and Friesen originally hypothesized that
observers in the body-only condition would be more accurate
than those in the head-only condition only when they were
observing lying samples of the subject's behavior. Their
results, however, suggest that the head-body finding might
be more general than they predicted, since there was a
marked, though not significant, trend in the same direction
when truthful behavior was being judged. A distinction
between accuracy for lying and truthing subjects has not
been made here for this reason. The results of this experi-
ment also suggest that, for emotional accuracy, the body
resulted in greater observer accuracy than did the head.both
when subjects were lying and truthing. For factual accuracy,
the head and body were about equal when truthing behavior was
being judged, and the head resulted in more accuracy than did
the body when lying behavior was being judged. To compare
the Ekman and Friesen (1974) results with those obtained
here, it is necessary to examine the accuracy scores sepa-
rately for lying and truthing samples of subject's behavior.
The following table does this, allowing a direct comparison
in those cells which directly replicate Ekman and Friesen
(1974) for both honest and deceptive segments of the sub-
jects' behavior. That is, the means in the present experi-
ment from the visual—only, head-only and body-only cells,
collapsing on tape and colorbw, are shown.

Table 16

Conpsrison of the Accuracy Means and Sample Sizes* for Honest and
Deceptive Samples of the Ekman and Friesen (1974) Experiment
with Those From the Present Experiment

 

 

 

 

 

 

Heed 304!

Ekman gggking Ekman Hocking

(Emo- (Eno- (Fsc- (Emo- (Eno- (Fec-

tionsl) tional) tual) tional) tional) tual)

. i .431 .416 .387 .501 .457 .400
3““ ' n 48 107 107 55 96 96
i .477 .497 .573 .635 .602 .483

Deceptive

n 48 107 107 55 96 96

 

*All observers judged both honest and deceptive segments. Thus. in this
table the some observers, within each experiment, made all head judgments,
while s second group made all body judgments. Each observer node four
Judgments in each cell of this table.

127

2It should again be noted that Ekman and Friesen

(1974) used only female subjects. The present experiment
used three female and 13 male subjects. Although several
experiments have reported that females display more non-
verbal facial expressions in non-deception contexts than
do males (Buck, Miller, & Caul, 1974; Buck, Savin, Caul,
and Miller, 1972), the fact that Ekman and Friesen's (1974)
head-body finding was replicated for emotional accuracy
suggests that the sex of the subjects did not account for
the failure to replicate this finding for factual accuracy.

APPENDICES

APPENDIX A

Subject Participation Invitation Letter

128
MICHIGAN STATE UNIVERSITY

 

COLLEGE OF SOCIAL SCIENCE ' SCHOOL OF CRIMINAL jUSTICE EAST LANSING ° MICHIGAN ' 48824

November 12, 19?“

Dear Criminal Justice Student:

The School of Criminal Justice in cooperation with the Department
of Communication is conducting a research project which will attempt
to identify certain personal characteristics of individuals which
may contribute to their successful performance as police officers.
This project compliments such research as the New York Police Training
and Performance Study, the RAND Study of Police Background Characteristics
and Performance, and the Systems Training Analysis of Requirements (STAR)
Study. The results of this study will influence decisions regarding
police personnel development programming. '

we are particularly desirous of obtaining your participation in the
study because of your status in the School of Criminal Justice. Some 30
students haye been selected from those currently enrolled in CJ 335 and
CJ 499 because of their expression of a particular interest in pursuing
a career in the police profession. It is important to the research design
to solicit participation from those currently enrolled in a course
addressing police issues and preblems and.who additionally intend to
pursue a career in the field.

Your participation in the study will involve Just one hour of your
time selected to be convenient for you. Within the next few days you.will
be contacted by a representative of the Department of Communication who
will provide additional information.

we appreciate your cooperation in this important research effort and
look forward to your participation.

Yours sincerely,

i 27%4112 51,;

A. F. Brandstatter
Director

AFBmka

APPENDIX B

Interview Format Review Sheet

129

APPENDIX B

Subject number Name

 

 

Supposed Actually
to be was

Q 1. What was the crime which Mr.
Bostick was convicted of?

Q 2. Did Mr. Bostick say he was
sorry for his crime?

Q 3. Did the judge say that Mr.
Bostick had the right to
appeal his conviction?

Q 4. What was Mr. Bostick wearing?

Q 5. What was Mr. Bostick's reac-
tions to being told that all
he needed to do to appeal his
conviction was to sign a
form?

(violent or non-violent)

Which slides did the subject see
first? (stress or pleasant)

 

Comments:

APPENDIX C

Subjects Post-Interview Questionnaire

IJ.‘II

 

 

130

APPENDIX C

Michigan State University
Department of Communication
1. How successful do you feel you were in lying when you were questioned
about the video-tape you saw before being interviewed?

extremely extremely
successful 1 2 3 4 S 6 7 8 9 10 unsuccessful

 

Comments:

2. How successful do you feel you were when you lied about the slides

 

you saw?

extremely extremely

successful unsuccessful
Comments:

3. How did you really feel when you saw the slides of burn victims?

Name

 

Age

 

Sex

 

Career Goals

 

 

APPENDIX D

Guidelines for Subjects' Debriefing

131

APPENDIX D

Give them the questionnaire.

Most of what we told you today was true. However, we implied that the
department of criminal justice was going to receive information about
your performance on this task. This is false. In order to attempt to
make this lying situation as involving as possible, we wanted you to
think that your performance was somehow going to affect your law enforce-
ment career. We are going to provide gg_information of gny_kind to the
department of criminal justice about you. Nor will your name be associ-
ated in any way with data which this research provides or in any pro-
fessional report of our findings.

The department of criminal justice is very interested in this research,
as we said, and we are going to provide your name to Mr. Brandstedder
indicating that you did participate (if you want). We really do appre-
ciate your help.

We would like to show some short excerpts from some of the people we.
are interviewing here to some freshmen and sophomore students in the
communication department next term and also to some adults in the area.
We will ask them to try and identify the truthful responses. If we
select some of your answers for this purpose, we will confidentially see
to it that you get feedback as to how well they are able to do this in
your case. We need to make it clear, however, that you may ask to have
the tapes of your responses erased and we will do this, and it will in
no way reflect negatively on you. Will you allow us to use your

tapes? Thank you. We really do appreciate your help.

We also have a favor to ask of you. Many students in criminal justice
are participating in this exercise. So that others may have the oppor-
tunity to participate under realistic circumstances, it is important
that they §2£_know that the criminal justice department will not
receive any information about their performance. So please do not

talk to them about this research until after Thursday night.

If you have any questions, we'd rather not answer them until after
Thursday night. You will be learning the full details of all aspects
of this research soon.

APPENDIX E

Observer Questionnaire

132
APPENDIX E

Michigan State University
Department of Communication

We are conducting research which is funded by the National Science
Foundation and is designed to examine the extent to which peOple can accu-
rately detect deceptive communication, that is, lying. Earlier this year
we created a situation in which some students who are planning careers in
law enforcement were interviewed by a police officer. Each of the students
had been instructed to lie to the officer at certain times and at other
times to tell the truth. The situation we created was such that it was
extremely important to the students that they successfully deceive the
officer. They did not want the officer to be able to accurately detect
their deception.

The students were videotaped during their interviews and today we
would like you to watch these tapes and do your best to accurately detect
when they were lying and when they were telling the truth. Participation
in this research is voluntary. If you do choose to participate you will be
seeing a total of 16 peOple engaging lying or "truthing" behaviors. We will
provide feedback about how accurate you are in detecting the deception to
everyone who wants this information. If you would like to receive your
personal accuracy score, please put your name and a summer mailing address
here:

Name

 

Address

 

 

Instructor's Name

 

I133

For each of the 16 people you will first see a short segment (about
30 seconds long) during which they always tell the truth. During this seg-
ment they were asked their name, major, year in school, hometown, and
whether they had seen a recent football game. Immediately after this truth-
ful segment, you will see another segment of the same person. This second
segment will be the test segment and will show the person either lying or
telling the truth. It is your task to determine which.

The first eight people you will see will be answering questions about
a videotape they had seen. They saw one of several versions of this tape.
On all versions a man is shown being sentenced to jail for a crime he had
committed, but the nature of the crime, what the man is wearing, and his
reactions to being sentenced are different. Those who lie about what was on
the tape they saw were instructed to give answers which were consistent with
a version of the tape which they did not see. Thus, the same answer could
be truthful for one person and a lie for another.

The last eight people you will see will be lying or telling the
truth about feelings they were experiencing as they watched some color
slides on a television screen. They saw two separate sequences of slides.
One set was very pleasant and included scenes such as lakes, boats, and
children playing. The other sequence of slides showed individuals who had
been seriously burned. As they watched these slides and immediately after
they were over, they were asked questions about the mood or feelings which
these slides created. They were instructed to say that they were having
pleasant and happy feelings in answer to questions about both sets of
slides. Of course, when they were watching the burn slides, their answers

were lies.

134

The length of the segments you will see will vary and is unrelated
to whether the person in the segment is telling the truth of lying. There
will be a ten second pause between each person to give you time to mark
this questionnaire indicating whether you think the person was lying or
telling the truth during the test segment. Remember the first segment you
will see of each person will always be truthful. Also, indicate how con-
fident you are in your judgment by checking the space on the scale which
most nearly reflects your degree of confidence in your answer.

The tape which you see might contain mostly deceptive segments, or
mostly truthful segments, or about equal numbers of each. Please try to
judge each sample without regard to your previous judgments. Once the tape
begins, please do not talk to anyone. Make your own judgments. Your
answers are anonymous and will never be associated with your name.- If you
do choose to participate please do your very best to accurately detect when
an individual is being truthful and when he is lying. Thank you very much
for your help. Your answers are very important to us and we appreciate your
cooperation.

We are first going to do a_practice example.

 

Example: Was the person in the practice example
lying _____ truthful
How confident are you in your judgment of whether this
person was lying or telling the truth?
not at all extremely

confident confident
0 l 2 3 4 5 6 7 8 9 10

 

135

Remember, the first eight peOple you will see are lying or telling

the truth about the videotape they had seen.

Segment 1
I think this person is: lying ____ truthful ____
How confident are you in your judgment?
not at all extremely
confident confident
O l 2 3 4 5 6 7 8 9 10
Segment 2
I think this person is: lying ___ truthful ___
How confident are you in your judgment?
not at all extremely

confident confident
O l 2 3 4 5 6 7 8 9 10

 

Segment 3
I think this person is: lying ____ truthful____
How confident are you in your judgment?
not at all extremely

confident confident
0 l 2 3 4 5 6 7 8 9 10

 

Segment 4
I think this person is: lying truthful

How confident are you in your judgment?

not at all
confident

extremely
confident

 

10

Segment

Segment

Segment

Segment

Segment

136

5

I think this person is: lying truthful

How confident are you in your judgment?

not at all extremely

confident confident
O l 2 3 4 S 6 7 8 9 10

6

I think this person is: lying truthful

How confident are you in your judgment?

not at all extremely

confident confident
0 l 2 3 4 5 6 7 8 9 10

7

I think this person is: lying truthful

How confident are you in your judgment?

not at all extremely

confident confident
O l 2 3 4 5 6 7 8 9 10

8

I think this person is: lying truthful

How confident are you in your judgment?

not at all extremely

confident confident
0 1 2 3 4 5 6 7 8 9 10

 

9

I think this person is: lying ____ truthful ____

How confident are you in your judgment?

not at all extremely

confident confident
O 1 2 3 4 5 6 7 8 9 10

 

Segment

Segment

Segment

Segment

Segment

137
10

I think this person is: lying truthful

How confident are you in your judgment?

 

 

 

 

 

not at all
confident
0 1 2 3 4 5 6 7 8 9 10
11
I think this person is: lying truthful
How confident are you in your judgment?
not at all
confident
O l 2 3 4 5 6 7 8 9 10
12
I think this person is: lying truthful
How confident are you in your judgment?
not at all
confident
0 l 2 3 4 5 6 7 8 9 10
13
I think this person is: lying truthful
How confident are you in your judgment?
not at all
confident
0 l 2 3 4 5 6 7 8 9 10
14
I think this person is: lying truthful
How confident are you in your judgment?
not at all
confident
0 l 2 3 4 5 6 7 8 9 10

extremely
confident

extremely
confident

extremely
confident

extremely
confident

extremely
confident

.138
Segment 15
I think this person is: lying ___ truthful____
How confident are you in your judgment?
not at all extremely
confident confident
0 l 2 3 4 5 6 7 8 9 10
Segment 16
I think this person is: lying____ truthful ___
How confident are you in your judgment?
not at all extremely

confident confident
0 1 2 3 4 5 6 7 8 9 10

How good a liar do you consider yourself to be? That is, when you want to,
are you able to successfully deceive:

your friends: never always
0 1 2 3 4 5 6 7 8 9 10

strangers: never always

 

How interesting or boring did you find the task of trying to detect deception?

extremely extremely
boring interesting

0 1 2 3 4 5 6 7 8 9 10

 

How reasonable a use of class time was this experiment?

very very
reasonable unreasonable

 

Were you sitting in a good position to observe this tape?

no, I couldn't see yes, I could
it well at all 0 l 2 3 4 5 6 7 8 9 10 see it very
well

139

We would now like to get some information about you.

Age

 

Sex

 

Grade point average

 

Race

 

Are you an American citizen?

Marital status

 

If you can tell us, on what basis did you make your judgments about whether

a person was lying or telling the truth?

APPENDIX F

Observer's Results Letter

140.
MICHIGAN STATE UNIVERSITY

 

College of Communication Arts East Lansing, Michigan 48824

Department of Commieat ion

August 27, 1975

Dear

This letter is in regard to the research on detecting deceptive commun-
ication which you participated in during your commmication class last
spring. The purpose of this study was to find out what categories of obser-
vable behavior people use to make accurate judgments of lying. You watched
a videotape which showed 16 individuals either lying or telling the truth
and attempted to accurately judge which was which. These videotapes were
shown to a total of 730 communication students under 14 different conditions.
Some groups observed a close-up of the head of the persons on the tape,
others watched the head and body, and others saw the body only. Some oh-
servers both heard and viewed the individuals on the tapes, while others
made their judgments based on the visual information only. Some students
saw these tapes in color and others saw them in black and white. We put
these three variables (Head only/Body only/Head and Body; Visual and Audio!
Visual only; Color/Black and White) together in all possible combinations,
thus creating 12 separate conditions. That is, there were different groups
of students who watched these tapes under each coublnation of these condi-
tions. So, for example, one group of observers viewed a black and white
tape of only the body of those on the tapes, and also heard what was being
said. Another group saw a close-up of the head in color, but did not have
the sound, and so on. Finally, one group heard what the individuals being
judged were saying but did not see them and one group read a written trans-
cript of what they were saying. This resulted in the total of 14 different
experimental conditions .

We have now compared the accuracy scores of groups of observers who
saw the tapes under each of these conditions. In general the results indi-
cated that those who saw the head and body were slightly more accurate than
those who saw either the head or body only, who were about equally accurate.
Observers who were provided with both the audio and visual information were
quite a bit more accurate than those who were exposed to the visual material
only. Those who saw the tapes in color were slightly more accurate than
those who say them in black and white. The accuracy score in each condition
is shown below. Fifty percent represented what would be expected just by
chance.

141.

 

 

 

 

 

 

 

 

 

Head Only Body Only Head 5 Body
Audio 8 Visual Audio 8 Visual f Audio 5 Visual
Visual -Only Visual Only Visual Only
Color 55.8% 48.4% 53.7% 48.4% 57.4% 48.5%
31“" and 52 .6% 44.9% 53.0% 48.6% su.o% u9.o%
White
Audio Only = 54.5% Transcript Only = 57.3%
Your personal accuracy score was out of the 16 judgments you made.

This task was very difficult and most peopIe did not do better than what
would be expected by chance (8 out of 16).

Your participation in this research project is very important for a
variety of reasons. First, when the data you provided us are analysed fully
and written up for publication, our knowledge of the behavioral basis for
accurate judgments of lying will have been advanced. This individual study,
by itself, makes only a small contribution to our understanding of human
communication behavior and its effects. However, other researchers interest-
ed in this same area will be able to examine our procedures and results and
be able to extend and refine the knowledge rained. In this way, as more
studies are completed, the accumulation of knowledge grows steadily and sys-
tematically.

Second, there are practical applications of this study. It was paid for
by a National Science Foundation grant which is looking at the effects of
using videotape during court proceedings. If, in the future, trials are re-
corded on videotape and shown to juries on television monitors, it will be
important to know under what conditions jurors would have the best chance to
detect lying by witnesses.

Third, the researchers have benefited from your participation in this
research. It is necessary to conduct major research projects to complete
graduate studies in all Social Science disciplines, including Communication.
Your participation has been important to us personally for this reason.

Finally, you personally may have benefited from this experience. By par-
ticipating in this experiment, by listening to the description of the research
which was provided during the class period in which you participated, and by
reading this letter, you may have learned a little bit about how Social
Science research is conducted. Hopefully this project has been an educational
experience for you. Also, much of the content of Social Science courses
comes from research projects like this one. Thus, future communication stu-
dents may benefit from your participation just as you have likely benefited
from the contributions of past participants.

In short, your participation in this experiment was extremely valuable.
We sincerely appreciate your help and.would like to thank you very much.

Sincerely,
" I ‘\ ' II:

\ A I’. - _-.‘ "l‘. ’ "‘1'
John Hocking‘
Graduate Assistant

APPENDIX G
Tables 17 and 18: Analysis of Variance for Factual and

Emotional Accuracy Using the Transformed Scores

2142
Table 17

ANOVA for Factual Accuracy* (using the transformed scores)

 

 

 

Sum of Mean Probability
Squares DF Square F Level
Main effects 14.423 5 2.885 18.309 .001
Colorbw .222 l .222 1.411 .233
Visaudio 10.099 1 10.099 64.103 .001
Tape 2.701 1 2.701 17.144 .001
Shot 1.907 2 .954 6.053 .003
Zdway Interactions 4.240 9 .471 2.990 .002
Colorbw visaudio .138 l .138 .874 .999
Colorbw tape .001 l .001 .009 .999
Colorbw shot .111 2 .056 .352 .999
Visaudio tape .692 1 .692 4.390 .034
Visaudio shot .074 2 .037 .233 .999
Tape shot ‘ 3.077 2 1.538 9.764 .001
3~way Interactions 1.611 7 .230 1.461 .178
Colorbw visaudio tape .026 1 .026 .165 .999
Colorbw visaudio shot .572 2 .286 1.814 .162
Colorbw tape shot .017 2 .008 .052 .999
Visaudio tape shot 1.031 2 .516 3.273 .037
4~way Interactions .341 2 .170 1.082 .340
C°l°rbw Visaudi° tape .341 2 .170 1.082 .340
shot
Residual 95.790 608 .158
Total 116.405 631 .184

 

 

*
transformation formula: a = 2 (Idarcsine x
multiple 5? - .124

1423
Table 18

ANOVA for Emotional Accuracy (using the transformed scores)

 

 

 

Sum of Mean Probability
Squares DF Square F Level
Main Effects 4.321 5 .864 8.228 .001
Colorbw .056 l .056 .530 .999
Visaudio .154 1 .154 1.469 .224
Tape 3.148 1 3.148 29.975 .001
Shot 1.028 2 .514 4.895 .008
2-way Interactions 1.699 9 .189 1.797 .065
Colorbw visaudio .130 1 .130 1.239 .265
Colorbw tape .002 1 .002 .017 .999
Colorbw shot .193 2 .096 .917 .999
Visaudio tape .016 l .016 .148 .999
Visaudio shot .452 2 .226 2.154 .115
Tape shot .862 2 .431 4.104 .017
3-way Interactions .741 7 .106 1.008 .425
Colorbw visaudio tape .003 1 .003 .027 .999
Colorbw visaudio shot .205 2 .103 .977 .999
Colorbw tape shot .033 2 .017 .159 .999
Visaudio tape shot .525 2 .263 2.501 .081
4-way Interactions .054 2 .027 .255 .999
Colorbw visaudio tape .054 2 .027 .255 .999
shot
Residual 63.854 608 .105
Total 70.669 631 .112

 

Multiple £2 - .061

APPENDIX H
Tables 19 and 20: Correlations Between Confidence and

Accuracy for Both the Factual and Emotional Segments

144

 

sao.na
emm.

 

 

asa.ua
em~.-

 

 

HH
some

mono oae=<

 

eno.-a
awn.

 

 

mme.ua
eno.

 

 

HH

cosy

aaco unauomoeue

Haoo comm ow mHo>sA muwaaooooum pom .ooooouwsou pom momusoo< scanner msouumaouuoo

 

 

 

 

 

 

 

 

 

 

moo. was. mac.
~eo.- «mo.+ mmo.+ ~ae.+ soo.- aeo.+
mac. mma.ua mas.na Noe.na mae.ua eo~.ua mmo.ua
mam. Nae. eno. «so. ~aa.- men. as users
nwo. a
use. mo~.ua ew~.ua mom.ua meo.ua «am.ua eao.na semen
nee. nae. moa.- «as. owo. em~.; H
ano.ua oe~.ua nno.ua sou.ua am~.ua mme.ua
oao omm.- are. own. was. «an. ace. Ha
duo. uoaou
oao.ua mem.ua mm~.ua ms~.ua ~cm.ua aa~.nd
m~o.- aa~.- Noe. aea.u one. eno.- «as. H
aaao .e=< a sane .e=< a sane .ea<
seams» anamo> assess assess seams» assess
scan was ease keno seam sane seam

 

 

 

 

 

musoawom Hosuoom one now swamon sou mo

ma manna

1145

 

mm~.ua
NmH.-

 

HH~.ua
cow.

 

 

 

HH

aHao oHea<

 

~h~.I
find.

 

HHs.aa
use.

 

 

 

HH

coca

mama

aaso nowuomsnua

 

 

 

 

 

 

 

 

 

oao. moH. nno.-
ace. oao. omH. msH. HNH.- «Ho.
nee. _wao.ua HHs.ua mam.ua «mo.ua OHe.na nno.-
can. nso.- amo. Now. see. Ham.- HH moans
mac. a
was..a nes.ua eno.-a am~.ua maH.ua am~.ua H eomHm
“Ho . nno. moo. Hen. asH.- sH~.- aNH.-
can oun— oum can—nu cum slum
moo. has Hem amH HAN «NN Ham HH
«No.- who. New. mNH. omH.- mHH.
00H. Hoaoo
DHa.ua amo.ua asm.ua oso.aa ASH.ua ~Ho.u
aeH aso.- mam. owo. can. oeH.- _ams. H
aHeo .eaa a aHeo .e=< a aHao .e=<
HasmH> HesmH> HmamH> HmamHa, HmamH> HmsmHs
seen can ease aHeo seam aHeo ems:

 

 

 

 

 

madmawom Hmsouuoam ecu you daemon use no

Huoo zoom ca mHo>oA muwawoeooum can .ooooowmsoo use monsooo< sooauom msowueaouuoo

ON manna

REFERENCES

REFERENCES

Bem, D., and McConnell, H. Testing the self-perception
explanation of dissonance phenomena: on the salience
of pre-manipulation attitudes. Journal of Personality
and Social Psychology, 1970, $21 23-31.

 

Bermant, G. and Jacoubovitch, M. Fish out of water: A brief
overview of social and psychological concerns about
videotaped trials. The Hastings Law Review, 1975, 26,
999-1011.

Berrien, F. and Huntington, G. An exploratory study of
pupillary responses during deception. Journal of
Experimental Psychology, 1943, 32, 443-449.

 

Buck, R., Savin, V., Miller, R., and Caul, W. Communication
of affect through facial expression in humans. Journal
of Personality and Social Psychology, 1972, 33, 362-
371.

Buck, R., Miller, R., and Caul, W. Sex, personality, and
physiological variables in the communication of affect
via facial expression. Journal of Personality and
Social Psychology, 1974,—20,_587F596.

 

Cutrow, R., Parks, A., Lucas, N., and Thomas, K. The objec-
tive use of multiple physiological indices in the
detection of deception. Psych0physiology, 1972, 9,
578-588.

Davis, R. Physiological responses as a means of evaluating
information. In A. Biderman and H. Zimmer (eds.), The
Manipulation of Human Behavior. New York: Wiley, I32;
168,_1961.

Doret, D. Trial by videotape--Can justice be seen to be
done? Temple Law Quarterly, 1974, 11, 228-268.

Eisenberg, A., and Smith, R. Nonverbal Communication. New
York: Bobbs-Merrill, 1971.

Ekman, P. and Friesen, W. Nonverbal leakage and clues to
deception. Psychiatry, 1969a, 33, 88-106.

146

147

Ekman, P. and Friesen, W. The repertoire of nonverbal
behavior: categories, origins, usage, and coding.
Semiotica, 1969b, 1, 49-98.

 

Ekman, P. and Friesen, W. Hand movements. Journal of Come

 

 

Ekman, P. and Friesen, W. Detecting deception from the
body and face. Journal of Personality and Social

 

English, H. Reaction-time symptoms of deception. American
Journal of Psychology, 1926, 31, 428-429.

 

Fay, P. J. and Middleton, W. C. ‘The ability to judge truth-
telling or lying, from the voice as transmitted over a
public address system. Journal of General Psychology,
1941, g5, 211-215.

Freedman, J. Role playing: psychology by consensus. Journal
of Personality and Social Psychology, 1969, 13, 107-114.

 

Goldstein, E. R. Reaction times and the consciousness of
deception. American Journal of Psychology, 1923, 34,
562-581.

Gustafson, L. and Orne, M. Effects of heightened motivation
on the detection of deception. Journal of Applied Psy-
chology, 1963, 11, 408—411.

 

Hays, W. Statistics. New York: Holt, Rinehart, & Winston,
1963.

 

Hildreth, R. An experimental study of audiences' ability to
distinguish between sincere and insincere speeches.
Unpublished doctoral dissertation, University of
Southern California, 1953.

Hocking, J., Margreiter, D., and Hylton, C. Intra-audience
effects: A field test. Human Communication Research,

in press.

HOrvath, F. Verbal and nonverbal clues to truth and decep-
tion during polygraph examinations. Journal of Police
Science and Administration, 1973, 1, 1386152.

 

Katzman, N. I. Violence and color television: What children
of different ages learn. Unpublished manuscript, Depart-
ment of Communication, Michigan State University, 1971.

Katzman, N. I. and Nyenhuis, J. Color versus black and white
effects on learning, opinion, and attention. Unpublished
manuscript, Department of Communication, Michigan State
University, 1971.

148

Kerlinger, F. and Pedhazur, E. Multiple Reggession in
Behaviora; Research. New York: Holt, Rifiehart and
Winston, Inc., 1973.

 

 

Knapp, M. Nonverbal Communication in Human Interaction.
New Yofk: Holt, Rinehart and Winston, Inc., 1972.

Knapp, M., Hart, R., and Dennis, H. An exploration of
deception as a communication construct. Human Communi-
cation Research, 1974, 3, 15-29.

 

 

Lee, C. The InstrumentalTDetection of Deception: The Lie
Test. Springfield, Illinois: Thomas, 1953.

Maier, N. and Thurber, J. Accuracy of judgments of decep-
tion when an interview is watched, heard, and read.
Personnel Psychology, 1968, 33, 23-30.

Maier, N. Sensitivity to attempts at deception in an inter-
view situation. Personnel Psychology, 1965, 33, 55-65.

Maier, N. and Janzen, J. The reliability of reasons used
in making judgments of honesty and dishonesty. Percep-
tual and Motor Skills, 1967, 33, 141-151.

 

Matarazzo, J., Wiens, A., Jackson, R., and Manaugh, T.
Interviewer speech behavior under conditions of engo-
genously-present,and exogenously-induced motivation
states. Journal of Clinical Psychology, 1970, 33, 141-
148.

 

Marston, W. Reaction-time symptoms of deception. Journal
of Experimental Psychology, 1920, 3, 72-87.

 

McClintock, C. and Hunt, R. Nonverbal indicators of affect
and deception in an interview setting. Journal of
Applied Psychology, 1975, 3, 54-67.

 

 

McCrystal, J. Ohio's first videotape trial. The Ohio Bar,
1972' 15., 1-40

 

Mehrabian, A. Communication without words. Psychology

Mehrabian, A. Nonverbal betrayal of feelings. Journal of
Egperimental Research in Personality, 1971, 3, 64-75.

 

Miller, G. and Burgoon, M. New Techniques of Persuasion.
New York: Harper and Row, 1973.

 

Morrill, A. Trial Diplomacy. Chicago: Court Practice
Institute, 1971.

 

149

Motley, M. Accoustic correlates of lies. Western Speech,
1974, 31, 81-87.

 

Nie, 19., Hull, C., Jenkins, J., Steinbrenner, K., and
Bent, D. Stagistical Packa e for the Social Sciences.
New York: McGraw-Hill, l .

 

 

Scheffe, H. A method for judging all contrasts in the
analysis of variance. Biometrika, 1953, 33, 87-104.

 

Shulman, G. An experimental study of the effects of
receiver sex, communicator sex, and.warning on the
ability of receivers to detect deceptive communicators.
Unpublished master's thesis, Purdue University, 1973.

Trovillo, P. A history of lie detection. Journal of
Criminal Law and Criminology, 1939, 33,848-881.

 

 

Winer, B. Statistical Pringiples in Experimental Design.
New York: McGraw-Hill Co., 1971.

Wolk, R. and Henley, A. The Right to Lie. New York: Wyden,
1970.

 

HICH

1an SIM: UNIV. LIBRARIES
1|WIIWIIWNIHIIIUIWIWIIWHIWIIHIHHI
3100230527

3129