MARKETING DECEPTION: BRAND IDENTIFICATION AND SEARCH,
EXPERIENCE, AND CREDENCE CHARACTERISTICS AS MODERATORS OF
TRUTH-BIAS AND DETECTION ACCURACY
By
Kim Blaine Serota

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Communication
2011

ABSTRACT
MARKETING DECEPTION: BRAND IDENTIFICATION AND SEARCH,
EXPERIENCE, AND CREDENCE CHARACTERISTICS AS MODERATORS OF
TRUTH-BIAS AND DETECTION ACCURACY
By
Kim Blaine Serota
In marketing communication, as in interpersonal communication, there is a
presumption that people can detect deception as it is occurring and, therefore, protect
themselves from the deceptive intent of the message. The Park-Levine probability model
(Park & Levine, 2001) posits that in interpersonal situations the veracity judgment is a
function of the receiver’s truth-bias and the base rate of untruthful messages evaluated.
The original model was supported by empirical testing. The study presented here extends
the Park-Levine model in two ways. First, it provides a conceptual replication using
marketing claims such as those found in advertising or other marketing communications
to support the model. Second, it shows that truth-bias toward marketing claims and,
subsequently, accuracy of detection is moderated by the presence or absence of a brand
and by information search characteristics that determine whether or not the claim can be
verified prior to purchase. Results demonstrate that the model can be generalized to noninterpersonal situations and that factors influencing accuracy need to recognize the
interaction between truth-bias and base rates in order to be meaningful interpreted.

Copyright by
KIM BLAINE SEROTA
2011

To Karen; this is as much hers as it is mine.

iv

ACKNOWLEDGMENTS

My graduate education began in the fall of 1972 and after 39 years this is no
longer just a dissertation, it is a life’s journey. There are many people I must
acknowledge and thank whose paths I have crossed along the way.
Let me begin by thanking my current committee for all of their guidance and
encouragement. Frank Boster chaired the committee and has been both a friend and
mentor; Frank started his doctoral program in Communication at MSU the year after I
did; obviously, he had a better sense of how to get from point A to point B. Tim Levine
was de facto co-chair and should be so acknowledged; Tim is a great teacher and a fine
colleague and he has helped to restore my academic curiosity. Chuck Atkin has the honor
(dubious or otherwise) of sitting on both this and my original doctoral committee; as
Chair of the Department of Communication, it was Chuck who encouraged me to return
and finish what I began many years ago. Joe Chartkoff of the Department of
Anthropology is a first-rate scholar and a good friend who provided considerable
perspective when it was most needed. I could go on for pages about these four
individuals; they have told me they wish I wouldn’t.
I want to acknowledge my original doctoral committee for supporting my first
pass at the degree. As chair, the shamanistic Joe Woelfel managed to guide me when I
had not yet found my compass. Ed Fink, Erwin Bettinghaus, Harold Spaeth of the
Department of Political Science, and the aforementioned Chuck Atkin were patient
supporters; the journey would not have begun, let alone have concluded, if they had not
allowed me to take the first steps.

v

Special thanks go to George Barnett, colleague, friend, and fellow traveler; his
simple advice got me over the last hurdles: “Just do it!”
Thanks to Mary Ellen Vaydik and Barbara Ward for the seeds of inspiration.
Appreciation goes to my first round of MSU professors, especially Randy Harrison and
Bill Crano. The graduate student cohort of the early 1970’s was remarkable; many of my
fellow students from that period today define the field of communication. A number of
them were influential to setting my course: Joe Cappella, Jim Danowski, Mac Parks, Rolf
Wigand, and Walter Gantz. Meg Brophy and Felipe Korzenny were two whose paths
have continued to cross mine in later lives. Thanks also to my Arizona State colleagues
Bill Arnold and Dean Hewes.
I am very grateful to the current faculty and my fellow students in the Department
of Communication at MSU. They have been supportive and encouraging of my quest. I
must especially thank faculty members Heesun Park (who taught the first class I took
after a 30 year hiatus), John Sherry, Bill Donohue, Ron Tamborini, Joe Walther, and
Maria Lapinski. Special thanks go to Sandi Smith; when I started to wander, she pointed
the way back to the path. My fellow graduate students deserve special acknowledgment,
their energy and enthusiasm was contagious; many thanks go to Chris Carpenter, Kyle
Andrews, Mike Kotowski, Allison Shaw, Hillary Shulman, and Rachel Kim (who are
also all co-authors). My deepest appreciation goes to Marge Barkman; I don’t think any
of us would graduate without her.
The road from and back to MSU was long and has included a number of
significant stops along the way. The people I encountered have helped to shape my
journey. Some are still on the road with me today. Many thanks go to Frank Ward and

vi

Mark Konkel of PACE; they eased my transition from academia into the world of
business. At Ford Motor Company, Tom Gorcyca, Henry Wolpert, and Paul Malboeuf
proved to be good friends and great people from whom to learn. Volkswagen of America
gave me a professional family and a place to grow; Dave Huyett, Alan Thompson (and
Alan’s wife, Marion), Mary Ann Wilson, Laura McKelvey, Laure Morrison, Jim Stefan,
and Margret Gillard were good companions; Danke sehr.
The next pathway took me to Allison Fisher, where I was blessed with colleagues
and clients who kept alive my desire to learn and teach, and to discover new ideas. Many
thanks go to colleagues Doug Scott, David Garfield, Dick Williams, Bob Quigley, Jim
Mulcrone, and Dorthea Bartz. Thanks also go to some great people I had the opportunity
to work with and learn from, most especially Bob Baxter and Matt Lewis at MercedesBenz; and Dan Knudsen, Sondra Paxton, and John MacDonald at National.
To my colleagues and students at Oakland University, I want to express my
warmest thanks for keeping me on track. Ravi Parameswaran was a key supporter of my
return to complete the degree, Janell Townsend provided immeasurable guidance, and I
relished the day I could answer “yes” to John Kim’s constant, “Have you finished your
dissertation yet?” A most special thank you goes to my friend and supporter Mukesh
Bhargava; some journeys cannot be completed in just one lifetime. Thank you to a very
important group of students, Lauren Began, Jon Flynn, Jessica Reese, and Bill Shearer,
who worked with me in the early stages to create the foundation for my dissertation
research. I also want to thank one of my very first Oakland University students, Craig
Stinson, for proving that teaching and caring about your students matters.
Appreciation goes to Jeff Molway and the folks at Synovate in Chicago for

vii

collecting my dissertation data; thank goodness for the practioners of 21st century
marketing research.
No successful journey is undertaken without the accompaniment of good personal
friends. I have been blessed with the help and encouragement of many. Those who have
trekked the farthest with me are mentioned here: Sabra Wood, Todd Estroff, Darra-Lee
Walker, Jerry Dunklee, Dan Gorrell, Jacquie Meredith, Paul Gurwitz and Susan Metzger,
Kendall Smith, and Kerry Chartkoff. Thank you all.
I would like to express my love and gratitude to my wife Karen, who was been
with me on this journey from the very start, and to my sons, KC and Kevin, who are old
enough now to be on to their own journeys, though they have never ceased to support and
encourage the continuation of mine. Thanks go to my brother Kip; his accomplishments
inspired me. Finally, I must acknowledge the love and support that I received over the
years from my in-laws Peggy and John Nickell, my aunt Edith Steinberg, and most
especially my parents Billie and Bob Serota; I wish that they could be here still to enjoy
the moment.

viii

TABLE OF CONTENTS

LIST OF TABLES............................................................................................................. xi
LIST OF FIGURES .......................................................................................................... xii
CHAPTER 1: INTRODUCTION ........................................................................................1
Background: The Marketing Communication Context............................................2
Objective ..................................................................................................................5
Detecting Deception.................................................................................................7
Direct Observation Approaches...................................................................7
Indirect Evidence Approach ........................................................................9
The Park-Levine Probability Model ..........................................................10
Extending TBT to Marketing Messages ................................................................12
Branding and Product Classification as Sources of Detection Information ..........12
Branding.....................................................................................................12
Product Classification ................................................................................14
Hypotheses.............................................................................................................21
CHAPTER 2: METHOD ...................................................................................................26
Main Experiment ...................................................................................................26
Subjects ......................................................................................................26
Design ........................................................................................................27
Procedure and Measures ............................................................................28
Creation of Marketing Claims Induction ...............................................................29
Search/Experience/Credence Classification ..............................................30
Final Claim Generation..............................................................................33
CHAPTER 3: RESULTS...................................................................................................36
Hypotheses 1 and 2: Assuming Equal Truthful and Not Truthful Claims.............36
Hypotheses 3 and 4: Overall Accuracy Predictions...............................................38
Hypotheses 5-7: Moderator Influences on Truth-bias ...........................................41
Hypotheses 8-10: Moderator Influences on Judgment Accuracy ..........................45
CHAPTER 4: DISCUSSION.............................................................................................56
Conceptual Replication..........................................................................................57
Moderation of Marketing Claims...........................................................................58
Moderation of Truth-Bias ..........................................................................60
Effect on Accuracy ....................................................................................62
Limitations .............................................................................................................66
Implications............................................................................................................69
Validation of the Park-Levine Model ........................................................69
Extending the Park-Levine Model to Include Moderators ........................70
Future Research .........................................................................................70

ix

Marketing Applications .............................................................................71
Conclusion .............................................................................................................72
APPENDICES ...................................................................................................................74
Appendix A: Questionnaire ...................................................................................75
Appendix B: Claims...............................................................................................76
Appendix C: Ideation Stage 1 – Stimulus Paper....................................................82
Appendix D: Ideation Stage 2 – Coding ................................................................87
Appendix E: Example Creative Brief ....................................................................93
REFERENCES ..................................................................................................................94

x

LIST OF TABLES

Table 1

Observed and Expected Accuracy Rates by Number of
Truthful Claims (Total Sample).................................................................40

Table 2

Average Truth-bias for Six Moderator Conditions, Branding by
SEC Characteristic .....................................................................................43

Table 3

Two-way ANOVA for Truth-bias with a Comparison of
Branding versus SEC Categories as Measured and Branding versus
SEC Categories Combined (S versus EC) .................................................44

Table 4

Two-way ANOVA for Accuracy with Branding versus Number of
Truthful Claims..........................................................................................46

Table 5

Two-way ANOVA for Accuracy with SEC Category versus
Number of Truthful Claims .......................................................................48

Table 6

Three-way ANOVA for Accuracy with Branding versus
SEC Category versus Number of Truthful Claims ....................................51

Table 7

Regression Equations and Observed Intercepts for the Relationship
between Accuracy and Number of Truthful Claims by the Six
Combinations of Moderator Variables.......................................................52

Table 8

Pair-wise Comparison of Simple Slopes and Intercepts for all
Moderator Combinations ...........................................................................53

xi

LIST OF FIGURES

Figure 1

Three factor design with 54 independent groups; (A) branding by
(B) SEC characteristics by (C) number of truthful claims
(among 8 claims evaluated) .......................................................................28

Figure 2

Observed accuracy as a function of actual veracity (number of
truthful claims out of eight presented). The linear fit is y = 0.0167x
2
2
+ 0.5213, R = 0.78; the curvilinear (polynomial) fit is y = -0.0033x
2

+ 0.0043x + 0.4903, R = 0.94...................................................................39
Figure 3

Observed accuracy as a function of actual veracity (number of
truthful claims out of eight presented). The linear fit is y = .0234x
2
+ .4827, R = 0.94......................................................................................47

Figure 4

Observed accuracy as a function of actual veracity (number of
truthful claims out of eight presented). The linear fit is y = .0093x
2
2
+ .5636, R = 0.31; the curvilinear (polynomial) fit is y = -.0052x
2

+ .0513x + .5147, R = 0.81.......................................................................47
Figure 5

Observed accuracy as a function of actual veracity (number of
truthful claims out of eight presented). The linear fit is y = .405x
2
2
+ .389, R = 0.74; the curvilinear (polynomial) fit is y = -.00972x
2

+ .1177x + .298, R = 0.95.........................................................................49
Figure 6

Observed accuracy as a function of actual veracity (number of
truthful claims out of eight presented). The linear fit is y = .00002x
2
+ .562, R = 0.00........................................................................................50

Figure 7

Observed accuracy as a function of actual veracity (number of
truthful claims out of eight presented). The linear fit is y = .011x
2
+ .6059, R = 0.66......................................................................................50

xii

CHAPTER 1: INTRODUCTION
The question of how receivers detect deceptive messages has been explored
across a variety of disciplines and contexts. In social psychology and communication,
there has been a pervasive assumption that accuracy is tied to individual differences and
the abilities of receivers. In most cases there is also a presumption that the veracity of the
message is readily knowable. Theory varies according to whether the locus of detection is
principally psychophysical or is transactional with the sender. An emerging alternative to
prevailing theory proposes that the truth is often not readily knowable and that those
judging message veracity rely on heuristics, primarily truth-bias, that rarely improve
accuracy by much better than chance. In other contexts involving applied
communications and mass media, there is little specific theory regarding detection of
deceptive messages; instead recognizing deception is assumed to be one of many
informational outcomes of the receiver’s evaluation of messages received.
The purpose of this study is to test the impact of truth-bias on the accuracy of
veracity judgments of non-interpersonal messages, specifically claims made in the
context of marketing communications. Two important factors influencing the evaluation
of marketing claims are inferences receivers make from branding information and the
perceived ability to verify the claim, which may be inferred from the product form or its
attributes. This study will examine the potential for these two types of information to
moderate truth-bias and influence judgment accuracy. In this introduction, literature is
reviewed and several hypotheses are presented. In the following sections a method for
testing these hypotheses is described, and results of the study are presented, reviewed,
and discussed. This report concludes with a discussion of study limitations and

1

implications for further research. In summary, this study provides a conceptual
replication of existing research testing the influence of truth-bias on deception detection
accuracy and extends that research from interpersonal communication to mediated
promotional messages and assesses additional factors that can alter the accuracy of
veracity judgments.
Background: The Marketing Communication Context
In the realm of marketing, some communication is intended to persuade by
presenting the producer’s goods and services in ways that will achieve favorable
outcomes for the producer. Although marketing is an exchange that is normatively
undertaken with the goal of creating value for both the producer and the consumer, the
producer often seeks to maximize value by obtaining the highest possible return at the
lowest possible cost. The consumer also seeks to maximize value; value is achieved by
obtaining desired goods, services, and ideas, also at the lowest possible cost. When
conducted ethically, both parties are rewarded by a fair and equitable exchange. Supply
and demand factors determine what both parties will consider equitable.
However, except when a monopoly exists, the producer is faced with competition.
In order to maximize value, the producer’s goal is to create competitive advantage by
having the best price or by being differentiated with the best product and/or the best time
and location of delivery (Day, 1984; Porter, 1985). The producer will also strive for the
most compelling promotional message in order to achieve perceived preference for these
differentiating characteristics (Parasuraman, 1997; Woodruff, 1997). When the cost to
achieve competitive advantage exceeds the cost at which the producer can meet
normative or personal ethical standards (e.g., acceptable levels of integrity, fairness, trust,

2

respect and empathy; Murphy, 1999, 2002) or when goals are not being met, there is
increased incentive for lying, deception, and other forms of unethical behavior (Lewicki,
1983; Lewicki & Robinson, 1998; Schweitzer, Ordóñez, & Douma, 2004). Alternatively,
when the bar of ethical standards is set too low for utilitarian or other reasons, as in the
famous cases of Nestlé’s infant formula marketing in Africa (Robin & Reidenbach, 1987;
Sethi, 1994) or the Enron energy and financial scandal (Healy & Palepu, 2003),
marketing deception and other unethical practices may be treated as routine. And finally,
deception may result from the intersection of greed and immorality (Cooke, 1997, 2004).
Marketing messages are purposefully planned and executed often with the intent
to persuade consumers to think, feel, and act in ways that are favorable to the producer.
From a marketing science perspective, the focal question in the study of deceptive
advertising and marketing persuasion most frequently has been whether or not a
promotional message was appropriately comprehended (Jacoby & Hoyer, 1987) and, as a
result, was the consumer (receiver) harmed (Burke, DeSarbo, Oliver, & Robertson, 1988;
Boush, Friestad, & Wright, 2009). Although there are ancillary issues such as whether or
not consumers have pre-existing false beliefs that may be reinforced through or in spite of
marketing messages (Russo, 1976; Russo, Metcalf, & Stevens, 1981), the contemporary
focus of marketing deception research has been the effect of the message on the receiver.
Three factors account for this approach. First, marketing scholars are most concerned
with whether or not marketing communications are effective and influence buying
behavior. The introduction of false or misleading information can create unfair
advantages that often lead to unjustifiably favorable attitudes, increased product interest,
and more sales for the deceptive marketer. For the deceived consumer the result is a

3

failure to have their needs met at the desired value level. Second, the marketing
profession is concerned because deception by some can distort perceptions of the roles
and practices of all marketers. The third important factor that drives the marketing
science approach is the evolution of legal and regulatory perspectives. Commercial
speech, unlike interpersonal speech, is highly regulated (Preston, 1977; Boush, et al.,
2009). Much of the research in the area of marketing deception has been conducted in
order to inform and structure the regulatory processes that are intended to protect
competitors and consumers from unfair advertising. In the United States, the Federal
Trade Commission (FTC) and industry organizations such as the Better Business Bureau
(BBB) and the American Association of Advertising Agencies (AAAA) rely on rules that
are fundamentally derived by observing either (a) the act of deceiving or (b) the effect of
advertising on consumers (Gardner, 1975). As has been demonstrated by the relatively
small number of successful attempts to prosecute deception by major advertisers, it is
extremely difficult to prove intent (Preston, 1977).
As a result, marketer scholars have tended to define deception in ways that reflect
a legalistic view. Prior to the formation of regulatory agencies and interest groups, the
response to marketing deception was fundamentally the warning philosophy of caveat
emptor, or “buyer beware.” In its earliest incarnation, the FTC was focused not on
consumers, but on protecting competition and marketing practioners; and in this context
the existence of marketing deception was identified by focusing on the message.
According to an early 20th Century FTC definition, when “an objectively ascertainable
material fact is presented falsely, is ambiguous, or is misleading” the presenting message
was viewed as evidence of the sender’s act of deception (Gardner, 1975). More recently,

4

the FTC perspective has shifted to one of protecting consumers. Gardner offered a
conceptualization that shifted the evidence of deception from the message to the
behavioral outcome, “If an advertisement (or advertising campaign) leaves the consumer
with an impression(s) and/or belief(s) different from what would normally be expected if
the consumer had reasonable knowledge, and that impression(s) and/or belief(s) is
factually untrue or potentially misleading, then deception is said to exist. (p. 42)”
Subsequently, marketing scientists have shifted their focus from judging message
veracity to the less difficult issue of establishing when the consumer has been harmed,
regardless of intent.
While the approach taken by marketing scientists, industry oversight groups, and
government regulators serves a useful purpose, the effect of this approach has been to
deflect attention away from a theoretically meaningful understanding of deception
detection processes in the commercial environment. There is an implicit assumption that
agencies need to regulate marketing communications because consumers cannot judge
the veracity of marketing messages for themselves. While this may be correct it is an
assumption made in the absence of empirical evidence and ignores the point that the
outcomes that drive the regulatory process might be moderated by the varying abilities of
consumers to detect deception in different situations.
Objective
The purpose of this study is to examine one aspect of the marketing deception
process as a first step toward developing a more comprehensive theory of deception in
mediated promotional communication. Specifically, this study addresses the ability of the
consumer to know whether or not she or he has been deceived by marketing messages.

5

Furthermore, the study hypothesizes that the accuracy of detection is moderated by
factors promoting the inference of truthfulness and the potential for verifying the veracity
of the promotional message.
It is reasonable to assume that consumers, given a choice, would prefer to avoid
false information in order to make accurately informed purchase decisions. While the
regulatory and monitoring functions provide some measure of protection and assurance,
the first line of defense would ideally be the consumer’s own perceptions and
interpretations. This raises the fundamental research question of this dissertation: Can
consumers distinguish truths from lies in marketing communications, thereby knowing
the veracity of the message, and gain the ability to avoid being deceived? In order to
address this question it is necessary to assess their accuracy in making veracity
judgments. This issue is a bounded subset of the larger communication question: can
human beings make accurate veracity judgments?
A related question has to do with what resources are available to the consumer as
a marketing lie detector. A well-established element of marketing theory is that a
consumer’s ability to verify product quality and characteristics varies with the kinds of
goods and services being promoted (Zeithaml, 1981). Some product information is easily
obtained (i.e., pricing, color, size) while in other cases the consumer is left to make
purchase decisions based on limited information (e.g. restaurant reviews, which may or
may not conform to the consumer’s tastes) or no information at all; thus the consumer is
faced with uncertainty and potentially higher levels of suspicion or skepticism. Similar to
several of the theorems in Berger’s uncertainty reduction theory (Berger & Calabrese,
1975; Berger & Gudykunst, 1991), an increase in communicative activity should reduce

6

consumer uncertainty, but when the levels of uncertainty remain high preference (liking)
may be weak and other forms of negativity such as suspicion are likely to occur. In
marketing this has been exacerbated by the inherent lack of reciprocity in mass media
communication. This is changing with the increasing use of internet and phone based
media as well as integrated marketing communication (IMC) strategies that focus on
relationships and the creation of a marketing dialogue (Duncan, 2005; Blakeman, 2007).
But the messages emanating from producers are still predominantly one-way in nature.
Detecting Deception
From the interpersonal communication perspective, there are four major theories
of deception detection that can be collapsed into two competing paradigms that differ
with regard to when the deception is actually detected. The first holds that deception can
be detected at the time the deception occurs by observation of nonverbal cues and/or
patterns of behavior that reveal the sender is engaged in deception (Ekman & Friesen,
1969a, 1969b). The second view is that receivers are handicapped by their own biases
and information processing capabilities and therefore, except in a few limited
circumstances, cannot accurately detect deception as it is occurring (Levine, 2009;
Levine et al., in press). Furthermore, according to the second view, people cannot
confirm or negate the veracity of a message without the availability of information from
sources or experiences external to the message interaction being observed.
Direct observation approaches. The prevailing paradigm in the deception
literature is based in Ekman and Friesen’s (1969b) fundamentally psycho-physiological
concept of leakage. The idea of leakage is that deception is an emotionally-charged
activity and the affects of emotion are largely uncontrolled. Furthermore, it is believed

7

that interpretative rules can be assigned to the facial display of emotions and other
nonverbal gestures or behaviors (Goffman, 1963). And despite serious concerns about the
ecological validity of research used to establish interpretations of nonverbal
communication (Argyle, 1972), the notion that nonverbal cues are associated with
leaking emotions has provided the foundation for an extensive body of deception
detection research (Knapp, 2008). The leakage approach was extended by Zuckerman,
DePaulo, and Rosenthal’s (1981) four-factor theory. This conceptualization, which
blends the physiological and psychological, states that truths and lies are differentiated by
four internal states: emotions, arousal, cognitive effort, and over control. Zuckerman et
al. (1981) associate these states with various nonverbal behaviors. When an individual is
consciously telling a lie he or she is more likely to experience strong emotions and a
heightened state of arousal, increase their cognitive effort, and attempt to control
nonverbal displays. The resulting nonverbal behaviors provide cues that a receiver may
use to detect that a deception is occurring. A third approach to deception detection falls
within the prevailing paradigm of direct observation theories, but differs with regard to
the fundamental complexity of the process. Developed by Buller and Burgoon, the
interpersonal detection theory (IDT; Buller & Burgoon, 1996; Burgoon & Buller, 2008)
takes a transactional approach to detection (cf. DePaulo, Ansfield, & Bell, 2006 and
Burgoon & Levine, 2009). In this approach, detection accuracy is “related to context
interactivity, receiver truth-biases, receiver familiarity, receiver decoding skills, sender
encoding skills, and sender deviations from expected patterns. (Burgoon & Levine, 2009;
p. 228)” DePaulo et al. have criticized Buller and Burgoon for merely giving a timeline to
the sequencing of deceptive interactions. At the very least, the number of independent

8

variables and the degree of variability within each makes it almost impossible to falsify
IDT, if it is in fact a theory.
Indirect evidence approach. The fourth major theoretical approach is Levine’s
(2009) truth-bias theory (TBT). TBT argues that receivers cannot accurately detect
deception because their biases inhibit their ability to make accurate judgments and
because the information necessary to overcome these biases is not usually present when
the deceptive message is being delivered. Levine argues:
(1) Senders deceive for a reason. This “principle of veracity” holds that there is a
moral asymmetry in which truthful statements are preferable to lies unless a
sender has a reason to lie; lies require justification that the truth does not (Bok,
1999).
(2) Receivers are truth-biased. Gilbert (1991) provides evidence that believing is the
default method of processing new information and disbelief requires active
processing. This leads to the “veracity effect,” which states that truths and lies
will be judged with unequal accuracy (Levine, Park, & McCornack, 1999). As a
consequence, measuring the accuracy of deception detection will be a function of
the truth-lie base rate, the ratio of truths to lies in the condition being judged for
veracity (Park & Levine, 2001; Levine, Kim, Park, & Hughes, 2006).
(3) Receivers perceive that senders deceive for a reason. This perception is referred to
as “projected motive” (Levine, Kim, & Blair, 2010) and is supported by research
showing that the patterns associated with truth-bias are altered when receivers
have a reason to be suspicious or information that causes them to believe the
sender has a reason to deceive (Fein & Hilton, 1994).

9

(4) Receivers do not accurately detect deception as it is occurring, unless they have a
reason to suspect deception (e.g. they infer motive) and the ground truth is known
in advance. This outcome is the result of the first three arguments, which suggest
that detection ability, or more correctly, inability, is a function of receiver
variables rather than sender cues and behaviors, as well as a fourth factor, the
absence of sufficient information in the deceptive message or accompanying
behaviors to allow the message veracity to be accurately determined. Park,
Levine, McCornack, Morrison, and Ferrari (2002) show that most lies are
detected after the fact as a result of third party information, physical evidence, and
confessions. At-the-time verbal and nonverbal behaviors are rarely the source of
detection, although they may increase suspicion, which in turn can lead to
confrontation and confession.
The Park-Levine probability model. An important demonstration of truth-bias
theory is provided by the Park-Levine probability model (Park & Levine, 2001) and
subsequent empirical testing (Levine, Kim, Park, & Hughes, 2006). Levine, Park, and
McCornack (1999) had questioned the conclusions of previous detection studies because
the results of most research fell into a fairly narrow range of accuracy. Subsequent metaanalysis by Bond & DePaulo (2006) re-affirmed this observation, showing that across
206 detection studies and 24,483 judges accuracy in most studies was within +/- 10% of
the average 54% accuracy rate. The common theme in most of these studies was that, as
part of the experimental design, ground truth was controlled by presenting subjects with a
50-50 mix of truths and lies.
Park and Levine proposed that detection accuracy depends on two factors:

10

conditional probability and base rate. The first factor states that accuracy is a conditional
probability; that is, the probability that a message is judged truthful given that the
message is truthful will be different from the probability that a message is judged a lie
given that the message is in fact a lie. The second factor is the probability that the
message is a truth or lie. A key assumption is that in reality the truth-lie base rate (the
veracity of a source’s message) is unknowable, but in deception detection experiments,
the veracity of source messages is under the control of the researcher. Levine et al. (1999)
found that the conditional probabilities for truth and lie judgments are not the same but
the ranges of these probabilities are fairly consistent across studies using an equal number
of truth and lie messages. By varying the base rate (changing the mix of truth and lies),
truth and lie accuracy can be altered.
An empirical test of the Park-Levine model (Levine et al., 2006) provides
evidence that varying the base rate does significantly alter the accuracy rate. In the
experiment, subjects judged a series of high-stakes lies. The subjects were divided into
treatment groups with each group judging eight videotaped interviews in which the
interviewee was either lying or telling the truth. Each treatment group viewed a different
mix of messages with the veracity base rate varying from eight truths to eight lies (and
each combination in between). Results showed that truth accuracy is higher than lie
accuracy and varying the base rate had a substantial impact on overall accuracy; as the
proportion of true messages increases, mean accuracy improves. The observed results fit
the predicted results with the latter falling within a 95% confidence interval of the
observed values. Levine et al. conclude that “chance explains deception detection rates
well.”

11

Extending TBT to marketing messages. TBT argues that deception is purposeful
and goal-oriented, that people are vulnerable to being deceived because they are
predisposed to trust new information, and even when they have reason to not trust a
sender, the message judge lacks sufficient information for veracity to be accurately
judged. Although TBT has been developed in the context of interpersonal interaction, the
principles of truth-bias theory can be applied in communication contexts where the
receiver is not directly interacting with the sender. In marketing there are interpersonal
interactions such as direct selling, but the majority of marketing communications fall
outside the realm of interpersonal communication. Based on TBT trust should be the
receiver default when marketing messages are presented to consumers (and in most cases
of marketing communication, as in most cases of interpersonal communication, trust is
probably justified). But consumers may be wary of new information; and mediated
marketing messages, like interpersonal messages, will provide insufficient information to
allow the consumer to verify the truthfulness of the claims in the message. In consumer
psychology, a number of studies have established that consumers can distinguish
persuasive intent from other kinds of message content (cf. Campbell & Kirmani, 2008,
for a review and critique) but there has been little evidence to show that consumers can
distinguish honest from dishonest persuasion. Before using the information in a
marketing message to make a purchase decision the consumer must either choose to trust
the claim or seek additional information in order to accurately assess the veracity of that
marketing claim.
Branding and Product Classification as Sources of Detection Information
Branding. Park et al. (2002) provide a clear indication that in most cases it takes

12

more than evaluating the verbal message or coincident nonverbal behaviors to detect
deception. To effectively determine whether or not deception has occurred contextual
information is required (Blair, Levine, & Shaw, 2010). In marketing, contextual
information often takes the form of brand reputation. Aaker (1991) defines brand as a
“distinguishing name and/or symbol (such as a logo, trademark, or package design)
intended to identify the goods or services of either one seller or a group of sellers, and to
differentiate those goods or services from those of competitors. (p. 7)” At a minimum a
brand identifies the source of the product and helps to protect the consumer and the
producer for products that appear identical but originate from a different source. In
contemporary usage a brand is a source of differentiation that helps distinguish product
based on characteristics other than price (Olson, 1977). It communicates what the brand
stands for and accentuates quality over price (Kirmani & Wright, 1989; Rao & Monroe,
1989). The fair assumption is that good brands want to protect their equity; therefore they
will not intentionally undertake marketing actions that will damage their reputation. This
is manifest in consumer trust for the brand (Aaker). This trust element of brand equity is
grounded in the cognitive functions of branding: awareness, associational schema, quality
perceptions and loyalty intentions, a view supported by Keller’s (1993) conceptual model
of consumer-based brand equity and at least implied by Srivastava, Alpert, and Shocker’s
(1984) product-based approach to identifying market structure (cf. Serota & Bhargava,
2010, for a discussion of the relationship between market structure, brand architecture,
and branding effects). Alternatively, Erdem and Swait (1998) view the role of brands as
fundamentally informative; brands have assets (the value ascribed to the brands) that
consumers use as signals of product quality, product positions, product attributes, and

13

product claim credibility. Like Levine’s theoretical concern that the veracity of
interpersonal messages is not readily knowable, Erdem and Swait stress that product
messages are inherently fraught with imperfect and asymmetrical information; therefore,
brands provide important cues, or signals, that increase credibility and reduce perceived
risk.
Product Classification. Some products or their qualities and characteristics may
be more susceptible to false or misleading claims than others. A structure for classifying
products based on information resources that consumers might use to detect marketing
deception has been suggested by information economics, an emergent component of
modern micro-economic theory (Akerlof, 1970; Stigler, 1961). The classification system
consists of three components: search, experience, and credence (SEC). The relevance of
the SEC approach to markets and marketing has been established through empirical
examination of the role of information in advertising, signaling, and information search
(Nelson, 1970, 1974; Spence, 1973).
The term search refers to the acquisition of information with which the consumer
may objectively evaluate products and on which they may base their purchase decision.
In 1961, George Stigler’s article, “The Economics of Information” published in The
Journal of Political Economy, proposed that the search for information is an important
aspect of economic activity and furthermore the nature and availability of information is
critical to the determination or ascertainment of market price. In 2001 Stigler shared the
Nobel Prize in Economics with A. Michael Spence and George Akerlof for this work,
which collectively became know as the analysis of markets with asymmetrical
information. Akerlof (1970) explains that asymmetrical information means buyers and

14

sellers do not have equal information when they enter into a transaction and the extent to
which that information is unequal will affect the outcome of the transaction. In general,
the seller has more control over information about the product, and the exercise of this
control will influence the economic power of both parties. Within this framework, seller
control means that the seller may provide full disclosure of information, withhold some
information, withhold all information, or create false information. Marketing messages
are a form of this disclosure; when the message is anything other than fully transparent,
the seller is engaged, to a lesser or greater degree, in deception. The extent to which the
buyer is able or willing to search for information that will allow the message to be
verified is a contributing factor in detecting and overcoming, or preventing, marketing
deception.
Stigler (1961) focused primarily on the price of goods, noting that the dispersion
of prices is a measure of ignorance in the market. The market price is the price the buyer
ultimately pays for the product and Stigler proposes that prices vary frequently and, with
few exceptions, no one can know all the prices that sellers of a given product are quoting
at a given time. Therefore, to ascertain the most favorable price, buyers must search to
find various price quotes. Stigler goes on to weigh the cost of search against its value in
reducing the price to the buyer. If the prospect of significantly reducing the price is high,
more searches are warranted; if the prospect is low, the value of the search activity is also
low and the buyer will engage in limited search activity.
Stigler (1961) also observes that this measure is biased because the products are
never homogeneous; the products can vary by the services the seller performs, by the
range of products stocked, and (though Stigler explicitly avoids discussing it) the quality

15

of the products. In response to Stigler’s heterogeneity concern, Nelson (1970, 1974)
examined the role of quality and the difficulty of obtaining information about the
performance characteristics of the product under various conditions of use. Nelson argues
that the utility of quality would vary more than the utility of price because of the greater
difficulty of obtaining information about quality and performance. Nelson further argues
that one solution to this problem is that some buyers rely on their post-purchase
experience with the product instead of searching for new information. Some products
(especially those that combine goods and services or are primarily services) do not afford
the opportunity to search in advance and determine which has the best quality. In these
cases, the buyer may choose to use the product and assign a posterior utility to the
purchase which in turn informs future decision making. Like prior knowledge that is used
to judge message veracity in the detection of interpersonal deception, experience can
inform the consumer about the veracity of promotional messages after the fact; thus the
seller’s recognition that buyers may ultimately discover the “truth” about a product might
inhibit the producer or marketer from exercising seller control through misleading or
false marketing communications.
Nelson’s proposition is that experience is an alternative form of search that
consumers use in order to obtain valid information when that information is more
difficult or expensive to obtain. Coincident with the idea that some information is more
difficult to obtain was recognition that sellers can distort information and willfully create
uncertainty (Akerlof, 1970). Buyers may use market statistics (or reviews or brand
information) to judge quality and, as Akerlof suggests, this reliance on reputation (i.e.
indirect knowledge of the product’s quality) is an incentive for sellers to market poorer

16

quality goods in order to increase profits. This is consistent with Bok’s (1999) moral
asymmetry in which lies are more likely to occur when the truth is seen as ineffective.
Darby and Karni (1973) conceptualized this as “optimal fraud.” But to explain the forces
contributing to optimal fraud, they found it necessary to add a third class of properties,
which they labeled credence qualities. Darby and Karni provide the following
comparison:
We distinguish then three types of qualities associated with a particular purchase:
search qualities which are known before purchase, experience qualities which are
known costlessly only after purchase, and credence qualities which are expensive
to judge even after purchase. (p. 69)
By expensive, Darby and Karni mean that objective information is not readily
available even after purchase and use of the product or service. The buyer is left to rely
on the credibility or honesty of the seller in order to accept that he or she has received fair
value for the price paid. In some cases, it may be possible to determine whether or not the
product is as specified or the service was actually performed, but the cost to do so would
not be warranted. In other cases, where information about the good or service is solely
controlled by the seller, the buyer may never know with certainty the quality of the
product.
Marketers, particularly those whose function is to create marketing messages,
often trade in uncertainty. Advertising, public relations, sales promotions, and other
forms of marketing communications are designed to present product qualities in such a
way that they are valued favorably by the potential buyer in comparison to similar
products from competing sellers. In order to do this marketers focus on qualities for

17

which their products have a competitive advantage and they ignore or diminish
information about properties for which the seller’s product lacks advantage. For the buyer
attempting to make a fair comparison of product alternatives, the information available is
often a patchwork of competing claims about different properties that do not line up
neatly for evaluation. As Stigler (1961) indicated, it is already difficult for the buyer to
know what the price is for various alternatives at a given point in time. When the buyer
must also assess both knowable and unknowable qualities, the potential for uncertainty
and the need to rely on incomplete or imperfect information is increased. Zeithaml (1981)
points out additional difficulties in her discussion of how consumers evaluate goods and
services differently. She notes that most goods have searchable qualities while services
tend to be judged on the basis of experience and credence. The characteristics which help
to define a service – intangibility, nonstandardization, and the inseparability of
production and consumption – make services much more difficult to evaluate and,
therefore, much more dependent upon experience (or credence) to supply evaluative
information.
Both buyers and sellers, in general, realize that consumers often have to make
choices without having complete information. Thus it may be the responsibility of the
buyer, caveat emptor, to understand that marketing claims are created with the intent to
persuade. But it is also incumbent upon sellers to promote their products ethically.
However as Akerlof (1970) and Darby and Karni (1973) theorize, the greater the extent
of uncertainty, the more likely the seller will have an incentive to cheat. Stigler (1961)
focuses on price, which is the obvious example of the kind of factual information for
which buyers can search; but there are other facts, attributes, and qualities that buyers

18

could know before purchasing. Nelson (1970) uses the example of canned tuna to
distinguish experience qualities from search qualities. He suggests buyers can easily buy
multiple brands of tuna in order to determine preference, and at the low cost of
experience there is insufficient demand for establishments to sell tastes of various brands
of tuna fish. Darby and Karni cite repair service as having credence qualities, noting that
in many cases consumers will be unaware of the ability of the service to satisfy a given
want, or lack the expertise to evaluate the service performance, and may only be able to
evaluate the outcome, not the procedure. As these examples illustrate, search, experience,
and credence product properties represent successively greater opportunities for
uncertainty and increasingly greater potential for producer/marketer deception. Thus the
consumer’s perceived ability to search for information is likely to be inversely related to
the perception of risk (Zeithaml, 1981) and manifest itself as greater uncertainty and
increased suspicion (lack of trust).
In summary, the likelihood of marketing deception increases with the degree of
uncertainty about the quality of the product. From an economic standpoint search is a
cost of product acquisition (Stigler, 1961) and the consumer engages in certain trade-offs
(e.g. time, shopping effort, reading or hearing advertising). The extent of this effort is
related to the value of the purchase. Asymmetry between the information held by the
seller and information obtained by the buyer could alter the outcome of the transaction
(Akerlof, 1970), opening the door for cheating and mistrust. One important reason for
this asymmetry is that not all product information can be known through search prior to
purchase (Nelson, 1970, 1974). In some cases it is necessary to experience the product
before one can know if the quality and characteristics match expectations, with those

19

expectations often based on the advertising and promotional messages associated with the
product and a specific brand. Darby and Karni (1973) added a third condition, suggested
that in some circumstances, even after purchase and use, the consumer is forced to rely
on information from the provider regarding the quality or characteristics of the product;
they referred to these as credence goods and services. The categorization of goods and
services, or their attributes, as search, experience, or credence is an efficient way of
ordering the degree of information asymmetry. The three SEC levels imply that prior to
purchase the consumer is likely to be able to reduce uncertainty, but differentially so,
depending on the type of product or attribute. It also implies an increasing opportunity for
exposure to deceptive marketing communications as search gives way to reliance on
experience and/or the truthfulness of others (credence).
The dissertation research addresses the topic of consumers’ abilities to detect
deception in marketing communication by applying truth-bias theory to marketing
messages. This research provides a conceptual replication of the Levine, Kim, Park, &
Hughes (2006) test of Park and Levine’s (2001) probability model for veracity base rates.
More critically, it generalizes their work on deception detection from interpersonal
communication interactions to mediated marketing communication. This generalization is
important for two reasons: (1) the majority of deception detection theory is limited by
conditions associated with interpersonal communication. Because TBT is not limited by
those conditions, this study provides a first test separating the message from the sender’s
behavior. (2) Marketing communication often involves persuasive arguments using
incomplete information (emphasizing strengths and ignoring weaknesses), which is
inherently deceptive (though the level of intent to deceive may vary). Consumers may

20

project persuasive motives to the sources of marketing messages (Campbell & Karmani,
2008; Levine et al., 2010) and their expectations for whether or not deception is occurring
will be influenced by message factors such as branding and by the degree of asymmetry
for the information found in marketing messages.
Hypotheses:
The first set of hypotheses is derived from the Levine et al. (2006) test of the
Park-Levine probability model. This provides the conceptual replication and will be used
to generalize the model to non-interpersonal marketing messages:
H1: Consumers are truth-biased. When presented with equal numbers of truths and
lies, consumers will judge a greater proportion of the messages as truthful
(base on Levine et al. H1).
H2: Truth accuracy will be greater than lie accuracy (Levine et al. H2).
H3: The accuracy of judging marketing truths and lies is a function of the message
veracity base rate; as the ratio of truthful messages to total messages
increases, accuracy will increase. (based on Levine et al. H4)
H4: The relationship between the ratio of truthful to total messages and detection
accuracy is linear and positive (Levine et al. H5).
Levine et al. (2006) included a hypothesis that truth accuracy would be higher
than 50% (equal chance of judging correctly or incorrectly) and lie accuracy would be
lower than 50%. Specific levels of truth-bias are an implication of the veracity effect
without factoring in the potential effects of projected motive or other contextual
information that might alter the slope and intercept of the graph expressing the
relationship between the ratio of truthful messages and detection accuracy; this will be

21

addressed by subsequent hypotheses. Failure to reject H1 through H4 should provide
strong evidence that the Park-Levine model generalizes to marketing messages. While
this study is concerned with establishing that truth accuracy is greater than lie accuracy
(H2), the shape of the distribution is expected to be affected by brand and SEC
moderators; therefore, a second set of hypotheses replaces the Levine et al. hypothesis for
variation around the 50% accuracy level. Levine et al. included a total of nine conditions
(including control) using large samples for each condition in order to increase the rigor of
the test. This rigor was expressed as a series of hypotheses about the precision of the
results. In order to examine moderators related to marketing deception, this research
requires a substantial number of conditions and, by necessity, has smaller sample sizes.
Therefore the hypotheses related to precision are not included.
Two moderators are of interest in this study: (1) the presence or absence of brand
is believed to affect consumer trust, and (2) the three SEC levels that can be associated
with a marketing claim may alter consumer expectations regarding the veracity of the
message. The effects of both factors should be reflected by the degree of truth (or lie)
bias and by the slope and intercept of the line describing the relationship between
veracity base rates and detection accuracy. This leads to the following truth-bias
hypotheses:
H5: Branding increases truth-bias; consumers will judge branded messages as
truthful more often than unbranded messages.
H6: The ability to verify the marketing claim increases truth-bias; consumers will
judge messages about search characteristics as truthful more often than
messages about experience characteristics (H6a); consumers will judge

22

messages about experience characteristics as truthful more often than
messages about credence characteristics (H6b), and by the logic of ordering
consumers will judge messages about search characteristics as truthful more
often than messages about credence characteristics (H6c).
H7: There will be an interaction between branding and SEC levels such that
branded messages about search characteristics will be judged truthful at a rate
significantly higher than unbranded messages about credence characteristics.
While truth-bias is positively associated with expectations about the
trustworthiness of the message and source, accuracy is expected to be undermined by the
effect of truth-bias. By varying the base rate of message veracity, we should observe a
change in the slope and intercept of the line describing the relationship between veracity
base rates and detection accuracy. When expectations of honesty are high, such as when
consumers are presented with branded messages about search characteristics, a steep
slope with a low intercept would reflect the tendency to judge truths accurately and lies
inaccurately. In the branded search conditions dominated by truthful messages (which
match well with expectations), accuracy should be high. In the branded search conditions
dominated by dishonest messages (which violate expectations), accuracy should be low.
At the other extreme, when consumers judge unbranded credence messages, they should
have lower expectations for honesty. These expectations should result in judging more
honest messages as lies and more dishonest messages as truthful. The effect across
conditions will be a flattening of the line such that the slope is closer to zero than in the
branded search conditions, or even negative. It is worth noting that a real difference in
judgment accuracy would result in a translation of the line (same slope with a higher or

23

lower intercept); however, TBT would not predict this and this study does not offer any
hypotheses about this effect. The hypotheses regarding accuracy are the following.
H8: Branding increases the expectation of honesty; as the number of truthful
claims increases as a proportion of total claims, judgment accuracy will
increase, and the increase for branded claims will have a steeper slope and
lower intercept than for unbranded claims.
H9: The ability to verify the marketing claim increases the expectation of honesty;
as the number of truthful claims increases as a proportion of total claims,
judgment accuracy will increase. The increase for search claims will have a
steeper slope and lower intercept than for experience claims (H9a); the
increase for experience claims will have a steeper slope and lower intercept
than for credence claims (H9b); and by the logic of ordering the increase for
search claims will have a steeper slope and lower intercept than for credence
claims (H9c).
H10: There will be a significant and positive interaction between branding and SEC
levels such that the relationship between the ratio of truthful to total messages
and detection accuracy will have a steeper slope and lower intercept for
branded search messages than for unbranded credence messages.
Even though we expect consumers to be more skeptical or weary when receiving
marketing messages than when they are engaging in interpersonal interaction, on average
consumers should expect marketing messages to be truthful and will judge these
messages to be truthful more often than they judge them to be false. This research does
not include a test of marketing messages versus interpersonal messages. Nonetheless, it

24

will be interesting to see if the levels of truth-bias vary significantly from the Levine et
al. (2006) results. This leads to the following research questions:
RQ1: Do consumers have less truth-bias toward marketing messages than they have
toward interpersonal messages?
RQ2: Does the inherent deception (or awareness of persuasive intent) increase
consumer skepticism to the level that consumers are lie-biased when judging
the veracity of marketing messages?

25

CHAPTER 2: METHOD
This study tests the propositions that (1) the receivers of marketing messages will
judge those messages according to their degree of truth-bias and (2) the accuracy of their
veracity judgments is a function of the distribution of true and false claims, moderated by
branding and by the search condition of the claim – that is, whether categorically the
claim addresses a search, experience, or credence (SEC) characteristic. An experiment
was conducted replicating the Levine et al. (2006) procedure. This experiment replaced
the interpersonal truths and lies stimuli with a set of honest and dishonest marketing
claims. Furthermore, the experiment was repeated for six groups; each group represented
one combination of the branding and categorical SEC moderating conditions.
Main Experiment
Subjects. The sample for this marketing deception detection experiment consisted
of 1,507 American adults (18 years of age or older) obtained using the Synovate eNation
omnibus panel. The omnibus panel is a commercial survey research tool used for daily,
multi-client studies. The studies are primarily consumer-oriented and the panel
approximates a nationally representative sample (cf. Serota, Levine, & Boster, 2010 for a
discussion of the strengths and limitations of this approach). The unweighted sample was
52.4% female and the mean age was 43.2 years, (SD = 15.29). Results are typically poststratification weighted (Kish, 1965) in order to improve external validity. Key panel
demographics (age, gender, income, and region) are matched to the U.S. Census Bureau’s
monthly Current Population Survey (CPS). Synovate also uses weighting to adjust
partially for the underrepresentation of Hispanics and other ethnic minorities in the
sample. In addition to approximating the distribution of the U.S. adult population, the

26

panel approach uses random assignment of subjects to treatment groups in order to
achieve comparable samples for each of the conditions in the research design. Subjects
were included in a prize drawing as the incentive to participate.
Design. The design of this experiment was adapted from the Levine et al. (2006)
test of the Park-Levine probability model of deception detection accuracy (Park &
Levine, 2001). Levine et al. used a one-way design to show that a linear relationship
exists between the number of truthful messages included in deception detection
experiments and the accuracy of veracity judgments; each treatment was a systematic
variation in the base rate of truths and lies. This study expanded upon that design and
consisted of a three-factor independent groups design, A x B x C, and full random
assignment of subjects for each of the treatment conditions.
The primary measurement in this study was judgment of eight marketing claims
as truthful or not truthful. By comparing these judgments with the actual veracity of the
statements, the dependent measures of accuracy (number of correct judgments as a
proportion of total judgments) and truth-bias (number of honest judgments as a
proportion of total judgments) were derived.
The three between-subjects factors were (A) the independent measures of brand
identification (real brand versus no brand identification), (B) the SEC category (search
versus experience versus credence) for the claim, and (C) the message veracity base rate.
Each subject completed the judgment task for eight messages; the messages were either
branded or unbranded claims from one of the three SEC categories. The study included
nine base rate conditions within each of the six branding x SEC categories and each base
rate condition ranged from 0 to 8 honest marketing claims. Levine et al. (2006) used the

27

50% honest/50% dishonest condition as an offset control group. In this study, the 50/50
mix was not offset; however, the equal mix condition serves as a de facto control group
for the base rate conditions within each of the six branding x SEC categories. Figure 1
provides a schematic view of the research design.

A

Unbranded
Branded
Search

Experience

B
Credence
0

1

2

3

4

5

6

7

8

C
Figure 1. Three factor design with 54 independent groups; (A) branding by (B) SEC
characteristics by (C) number of truthful claims (among 8 claims evaluated).
Procedures and Measures. Data for this study were collected using the Synovate
e-Nation online omnibus data collection system. Subjects received an e-mail invitation
asking them to participate in an omnibus survey on December 22 or December 27, 2010;
the total sample of 1,507 was obtained between December 22 and 29, 2010. The
invitation was directed to a specific member of the household identified by age and
gender. The invitation instructed the specified individual to click on a link to the survey
website; 1,238 subjects (82.1%) were the specified respondent while 269 surveys (17.9%)
were completed by a different member of the household. On the website, subjects were
provided with instructions, asked questions confirming participant identification, asked
28

omnibus survey questions for several unrelated topics including the deception detection
questions, and asked a series of demographic questions. Subjects recruited for the
December 22 survey answered questions (in order) about whiskey, truthfulness of claims,
and software logos. For the December 27 survey subjects were asked about cameras,
truthfulness of claims, software logos, deodorant, optical purchases, and hardware stores.
Since nearly all members of the population were expected to be consumers and have
experience with the consumer behaviors of information seeking and purchase decision
making, no attempt was made to screen the subjects for consumer activity. The study
procedures were IRB-approved.
Each subject evaluated eight messages. All messages were selected from one of
the search, experience, and credence categories by random assignment. Each subject was
randomly assigned to either the real brand group or the unbranded group. Subjects were
also randomly assigned to a veracity base rate (number of truthful claims among the eight
messages); the set of eight messages were presented in random order. After each message
was presented the subject was asked whether the message was a truthful claim or not
truthful. Appendix A provides the text of a typical questionnaire and Appendix B is the
complete list of 96 messages used in the study.
Creation of the Marketing Claims Induction
This stimulus generation required two distinct steps. First, it was necessary to
create a pool of brand and product characteristics that were reliably classified into the
three SEC categories. Second, for each claim to be used as part of an experimental
condition, four messages needed to be created. Initially two messages about the brand or
product characteristic had to be created, one that was known to the message creator to be

29

truthful and one that was known to the message creator to be false. These messages were
to include an established and potentially recognizable brand name. Then, a second set of
two messages was created by taking the two branded messages and removing the brand
name from each; the absence of a brand name assured that message would lack real,
accrued brand equity.
Search/experience/credence classification. In order to obtain lists of products and
product characteristics that fall distinctly into each of the three categorical conditions,
student coders participated in a two-stage stimulus generation process in order to assign
SEC categories to the characteristics of various goods and services. Although numerous
lists have been generated as an outcome of the evolution of Stigler’s (1961) economics of
information theory, these lists have tended to reflect objectives associated with specific
theoretical perspectives. Categories tend to be influenced by the researchers’ academic
orientation. Some of the studies and articles were economic in nature and dealt with the
elasticity of demand given varying assumptions about information availability (Nelson,
1970, 1974; Darby & Karni, 1973; Laband, 1986). Others were marketing promotional in
nature, which more often focuses on the inseparability of production and consumption
(Shostack, 1977; Zeithaml, 1981; Ekelund, Mixon, & Ressler, 1995). In practical terms,
this means that the economics-based lists most often focus on the products themselves
while the marketing lists usually take into account the characteristics of the products and
how they are consumed. The latter better recognizes that for a given product or brand, the
product may have characteristics in all three categories. For example, a restaurant might
advertise, “Eat at Joe’s. For just $8.99, our diners get a delicious meal made with only the
best ingredients!” In such a case, the price of the restaurant meal may be verified before

30

purchasing and is therefore searchable; the tastiness of the meal is subjective and requires
the consumer’s dining experience before the veracity of the statement can be confirmed;
and the consumer is left to trust that the ingredients are the best or even what they are
purported to be when the menu or server is consulted. While existing lists are constrained
by the theoretical and empirical needs for which they were created, time and technology
are also critical factors in the process of stimulus generation. Many new products and
services did not exist 40-50 years ago when economic theories of information were first
tested and, more importantly, the tools for information search have been greatly refined
in the last two decades. The breadth of product availability and the ubiquitousness of the
Internet reduce the cost of search to the point that the quality of many characteristics once
thought of as experience or credence attributes can be known prior to purchase.
The ideation process took place in two stages. Initially, four student assistants
participated in a group discussion of SEC products and characteristics moderated by the
researcher. The students were enrolled in a strategic management course at a large
Midwestern university and received extra credit as undergraduate research assistants. To
start the discussion, they were presented with a short written explanation of the concepts
of search, experience, and credence based on the theoretical material in the introduction
to this study (Appendix C). The description defined each of the terms, explained that they
might apply to entire products (goods, services, and ideas) or only to certain
characteristics of products. A discussion followed in which the research assistants were
allowed to ask questions and clarify their understanding of the terms. They were then
asked to recommend entire products or characteristics associated with specific products
for each category. Three separate lists were maintained during the brainstorming session;

31

the participants were encouraged to discuss, challenge, debate, and modify the
recommended categorical assignments. The entire group discussion lasted approximately
two hours. The result was an extensive list of products with their key characteristics or
attributes assigned to the three search categories.
In the second phase of the ideation process, the lists from the group discussion
were organized into a 100-item battery that was presented as a self-administered coding
exercise (Appendix D). The coding exercise was completed by 18 volunteer coders
recruited from students enrolled in a management strategy course at a large Midwestern
university. The students received extra credit for assisting with the item coding. The list
of products and product characteristics was coded online so that the items could be
presented to the coders in random order. As in the initial phase of the process, the three
SEC concepts were briefly explained. For each item on the list, coders were asked to
indicate whether their knowledge of the product and their ability to evaluate the product
could be enhanced by (1) searching, or if it required (2) first-hand experience with the
product, or if it required them to (3) trust the information provided by the producer or
seller because the product quality or characteristic could not be objectively verified. Each
product or product characteristic was scored for consistency of response. In general, the
response consistency is highest for search characteristics. Coders were somewhat less
likely to agree on the assignment of experience characteristics, and much less likely to
agree on the assignment of credence characteristics. The final list of 24 products and
characteristics (eight each for the three SEC categories) was determined by removing
inconsistent coders, re-calculating the consistency for each item, and selecting those
items with the highest consistency within each of the three categories. Coder reliability

32

was assessed at several points in this process.
Initially, 57 items were most frequently coded as search, 35 items were most
frequently coded as experience, and 8 items were most frequently coded as credence. The
scores were assessed for inter-rater reliability using Fleiss’ kappa for multiple raters of
categorical data (Fleiss, 1971); the initial k = .260 indicated a need to review individual
coder results. Across the 100 items, pairwise coder agreement for the 18 coders ranged
from a high of 70% to a low of 29%. The overall mean agreement was 52.8% (SD =
7.09%); one coder had an average agreement of more than 2 SD below the mean and was
removed. Three additional coders exhibited a pattern of bias against coding items as
credence and were removed. Based on judgments by the remaining 14 coders, 41 items
were classified as search, 19 were classified as experience, and 12 were classified as
credence; 18 ties were removed from the list. This improved the Fleiss k to .411 with 14
coders and 72 items. In order to further refine the item list, all items for which there was a
clear second choice assignment were eliminated; at this step Fleiss k = .564 with 14
coders and 42 items. The final step was to select the 24 most consistently judged items
(eight for each SEC category); this increased the Fleiss k to an acceptable .594. The final
eight search items had a coder agreement range of 93% to 100%. For experience items
the range was 86% to 100%. Coders had some difficulty classifying items into the
credence category; coder agreement ranged from 54% to 79%, with the lowest agreement
item also judged to be a search item by 36%.
Final claim generation. In order for the induction messages to be of known
veracity, it was necessary for the claims to be created with the explicit intent of being
either truthful or not truthful. The process of creating and sending messages for

33

advertising, public relations, or other elements of the promotional mix typically involves
multiple individuals. In order for a false or dishonest message to enter the marketing
environment some or all of these individuals must be complicit in the deception. The
dishonesty may reside in the facts of the strategic marketing or communication plan that
is typically created within the firm that produces the goods or services being marketed.
Or, dishonesty may be introduced in the creative brief, a document usually produced by
the marketing agency’s account management in order to translate the marketing plan into
specific communication objectives that guide the creative process of message
development. The facts of the creative brief may be altered by copywriters and art
directors in order to create false but favorable brand or product impressions in the
finished message. Though less likely to influence the content of the message, the final
communication passes through a production and editorial process in which arbitrary
alterations or errors, some of which may be misleading or falsifying, can be introduced.
Rather than relying on the many stages of the message creation process to
produce a mix of true and false claims, a standard procedure was established. For each of
the 24 messages, an initial abbreviated creative brief was developed using information
believed by the researcher to be correct and accurate about a branded product. Actual
brands were matched to the questions selected from the second phase of the ideation
process. Brand and product websites, promotional messages, media reviews, rating
services (e.g. Consumer Reports, Better Business Bureau), and government agency
reports (e.g. US FDA filings) were culled for information that could be used to create the
truthful and not truthful claims. Information in the brief was fact-checked using at least
two sources each for the background and description.

34

The brief identified the product or service and the brand name; it included the
question used to establish the product and characteristic SEC classification, and it
included a short statement of facts relevant to the claim. Key sources were also noted.
Four versions of the claim were added to the brief upon completion: the branded and
unbranded truthful claims and the branded and unbranded false claims (see Appendix E
for an example).Other elements that are typically part of a creative brief, such as
communication objectives, a target audience profile, and guidance on the tonality of
messages were excluded.
A brief was developed for each of the 24 products or product characteristics
selected from the ideation phase of the stimulus generation process. In several instances
the product category was slightly modified or a different product category with equal
classification agreement was substituted; this was done to obtain sufficient information to
assure the veracity of the truthful claim. For false claims, the messages incorporated a lie
about a key detail or fact in order to create more favorable positioning vis-à-vis the
competition and/or contradicted the factual information of the truthful claim so that the
product or service would appear to be more favorable to the audience. The final step was
to replace the brand name with the generic “we” or “our [product]” to create unbranded
versions of both the truthful and not truthful claims. This provided a battery of 96 claims,
four versions for each of the 24 products (eight in each of the three SEC categories). All
claims were written by the researcher; some were original and some were adaptations of
existing claims for the specific brand and product.

35

CHAPTER 3: RESULTS
The 1,507 subjects judged a total of 12,056 claims with each subject judging eight
marketing claims. The number of truthful claims ranged from zero to eight. The total
number of truthful claims judged was 6,083 or slightly more than half. Due to slight
variation in sample sizes across the 54 cells in the study, if all judgments were made
randomly, the expected percentage judged truthful would be 50.46%. Overall, more than
half of the claims were judged to be truthful (M = 4.49, SD = 2.09) yielding an overall
truth-bias of 56.4% with 95% CI = 55.1 – 57.7%. A one-sample t-test showed that the
average number of truth judgments exceeded the expected value of 50.46%, t (1506) =
8.82, p < .001, d = .228.
Total accuracy across the 54 cells ranged from 24.5% to 72.5% with an overall
average accuracy of 58.9% (95% CI = 57.8 – 60.0%). The one-sample t-test showed that
the total judgment accuracy exceeded 50%, t (1506) = 15.77, p < .001, d =.408 and did
not support RQ2, which asked if consumer skepticism leads to lie-bias. Truthful claims
were judged correctly 65.1% of the time while 53.9% of the not truthful claims were
judged correctly; non-overlapping confidence intervals indicate that the difference is
significant (truthful 95% CI = 63.4 – 66.9%; not truthful 95% CI = 52.1 – 55.7%). A
paired samples t-test (excluding those only truthful claims and those judging only not
truthful claims) confirmed this difference, t(1170) = 6.82, p < .001, r = -.307, dz =.199.
Correctly judged truthful and not truthful claims are both significantly greater than 50%.
Hypotheses 1 and 2: Assuming Equal Truthful and Not Truthful Claims
The Park-Levine probability model (2001) posits that deception detection
accuracy is a function of message veracity base rate rather than people’s ability to

36

distinguish truths from lies. Levine, Kim, Park, and Hughes (2006) support this
prediction, showing that the slightly better than average accuracy observed in many
deception detection studies (Bond & DePaulo, 2006) was partially due to the 50/50 mix
of truths and lies used to control for stimulus effects in most experiments. Levine et al.
(2006) obtained veracity judgments from an offset control group using the 50/50 stimulus
mix. In order to provide a conceptual replication from outside the interpersonal realm, the
first two hypotheses of this study address the data for cells with equal numbers of truthful
and not truthful claims aggregated across moderator conditions. Hypotheses 3 and 4 are
addressed by data obtained from the total sample. The remaining hypotheses address the
brand and search moderators both separately and in combination.
A total of 168 subjects judged 1,344 marketing claims across the six equal
truthful/not truthful conditions. Each subject judged four truthful and four not truthful
claims (the subjects did not know how many claims were truthful). The task for this
group was conceptually equivalent to the Levine et al. (2006) control group. Hypothesis 1
predicted that consumers are truth-biased. When presented with equal numbers of truths
and lies, consumers will judge a greater proportion of the messages as truthful. Overall
the data were consistent with this prediction. Subjects judged 54.2% of the claims truthful
(M = 4.32 claims, SD = 2.04). A one-sample t-test showed the observed proportion of
truthful judgments to be greater than chance, t(167) = 2.12, p < .05, d = .163. However,
consistent with RQ1, an independent groups t-test showed that truth-bias is significantly
lower when evaluating truthful and untruthful marketing claims than the truth-bias
Levine et al. observed when subjects judged interpersonal truths and lies (M = 66.1%, SD
= 14.4%, N = 50), t(216) = 3.16, p < .005, d = .576.

37

Hypothesis 2 predicted that truth accuracy will be greater than lie accuracy. The
168 subjects in the six equal truthful/not truthful conditions had accuracy ranging from
44.5% to 71.9% with an average accuracy of 60.0% (M = 4.80 claims judged correctly,
SD = 1.50) with a 95% CI = 57.2 – 62.9%. A one sample t-test indicates that this is
significantly better than chance, t(167) = 6.94, p < .001, d = .535. Correct judgments
were obtained for 64.0% of the truthful claims (95% CI = 59.2 – 68.7%) while correct
judgments of not truthful claims were obtained 56.1% of the time (95% CI = 51.3 –
60.9%). Hypothesis 2 is supported by a paired sample test, t(167) = 2.00, p < .05, dz =
.155. However, the overlapping confidence intervals indicate only that truth accuracy is
directionally higher than lie accuracy; unequivocal evidence that truthful claims are
judged accurately more often than not truthful claims was not obtained. Levine et al.
(2006) predicted that interpersonal lie accuracy would be below 50%. While their
hypothesis of below average accuracy was confirmed, consumer judgments of untruthful
marketing claims were significantly better than chance, t(167) = 2.47, p <.05, d = .191.
Hypotheses 3 and 4: Overall Accuracy Predictions
Hypothesis 3 predicted that the accuracy of judging marketing truths and lies is a
function of the veracity base rate (the number of truthful claims divided by total claims);
as the ratio of truthful to total messages increases, accuracy will increase. This hypothesis
was initially tested using the total sample data (N = 1,507), combining results across the
moderator conditions. The 2 x 3 moderators used to define six conditions were measured
with approximate equal proportion. There is no evidence that these exist in equal
proportion in the population of marketing messages. In fact, it is reasonable to assume
that most claims either are branded or are unbranded statements in the context of an

38

advertisement or other promotional message for which the brand is known. Based on the
ideation phases of the stimulus generation process coders identified more than half of all
claims as having search characteristics rather than experience or credence attributes.
The hypothesis that accuracy increases as the base rate for actual veracity
increases was tested using a one-way ANOVA. The influence of proportion of truthful
messages to total messages on accuracy was significant, F(8, 1498) = 9.84, p < .001;
2

however, the effects size was relatively small, η = .05, η = .22. The overall results are
depicted in Figure 2.
0.70
0.65

Accuracy

0.60
0.55
Upper CI
0.50

Observed
Lower CI
Linear Fit

0.45

Polynomial Fit
0.00
0

1

2

3

4

5

6

7

8

Actual Veracity (Number of Truthful Claims)

Figure 2. Observed accuracy as a function of actual veracity (number of truthful claims
2
out of eight presented). The linear fit is y = 0.0167x + 0.5213, R = 0.78; the curvilinear
2

2

(polynomial) fit is y = -0.0033x + 0.0043x + 0.4903, R = 0.94.
Drawing upon the interpersonal results from Levine et al., Hypothesis 4 predicted
the relationship between the ratio of truthful to total messages and detection accuracy is
39

linear and positive. As Figure 2 shows, accuracy means are linearly increasing from the 0
to 5 truthful claims; but from 5 to 8 truthful claims, there is little gain in accuracy (the
somewhat lower mean for the 7 truthful claims condition is primarily the result of an
inconsistently low score in the branded x search moderator set; with 54 unique cells in
the experimental design and an average sample size of n < 28 it should be reasonable to
find one or two cells with discrepant results). Both linear and curvilinear equations were
2

fit to the data; the linear fit resulted in R = .78 but the curvilinear equation yielded a
2

substantially better R = .94. Data in Table 1 show that Hypotheses 3 and 4 are only
partial supported; above 62.5% (five truthful claims out of eight total claims), accuracy
ceases to increase. Two of the values deviated significantly from a linear fit but, when the
nonlinear fit was tested, none of the t-values was significant at p < .05.
Table 1
Observed and Expected Accuracy Rates by Number of Truthful Claims (Total Sample)
Truthful
Claims Actual

Fit

Linear
Dev

Nonlinear
Fit
Dev

t

t

95% CI

0

0.500

0.521

0.02

-0.99

0.490

-0.01

0.45

0.457 – 0.542

1

0.515

0.538

0.02

-1.33

0.530

0.02

-0.88

0.480 – 0.549

2

0.562

0.555

-0.01

0.50

0.564

0.00

-0.09

0.532 – 0.591

3

0.598

0.571

-0.03

1.95 *

0.590

-0.01

0.57

0.571 – 0.624

4

0.600

0.588

-0.01

0.85

0.610

0.01

-0.68

0.571 – 0.629

5

0.640

0.605

-0.04

2.95 **

0.624

-0.02

1.37

0.616 – 0.664

6

0.635

0.622

-0.01

0.97

0.631

0.00

0.28

0.608 – 0.660

7

0.611

0.638

0.03

-1.47

0.631

0.02

-1.08

0.574 – 0.647

8

0.632

0.655

0.02

-1.17

0.625

-0.01

0.35

0.592 – 0.671

Note: * p < .10, ** p < .05
40

Hypotheses 5-7: Moderator Influences on Truth-bias
Bok’s veracity principle (1999) and Levine’s projected motive model (Levine,
Kim, & Blair, 2010) argue that lies occur for a reason and, in the absence of a reason or
motive, those judging messages expect the messages to be truthful. With regard to claims
made in marketing communications, two moderators have been hypothesized to influence
the extent of truth-bias by consumers receiving marketing messages. First, the presence
of a brand serves as a signal that the message can be trusted (Erdem & Swait, 1998), thus
increasing the likelihood that the message will be judged truthful. Second, the perceived
ability to verify information in the claim can lead to an inference of honesty. Derived
from economic theories of information, search claims are those that can be verified
before purchasing, experience claims can be verified through purchase and use, and
credence claims require the consumer to trust the seller even after the product has been
purchased and used. Hypotheses 5-7 predict that truth-bias increases when messages are
branded and perceived to be verifiable before purchase.
Hypothesis 5 states that branding increases truth-bias and consumers will judge
branded messages as truthful more often than unbranded messages. To test the validity of
this hypothesis, half of the messages presented included a brand name while the other
half were the same messages presented generically. Results for branded messages were
aggregated across SEC conditions in order to compare them with results for unbranded
messages, also aggregated across the SEC conditions. Truth-bias for branded claims was
59.7% (95% CI = 57.8 – 61.6%) while truth-bias for unbranded claims was 53.0% (95%
CI = 51.3 – 54.8). In addition to non-overlapping confidence intervals, Hypothesis 5 was
supported by an independent samples t-test, t(1505) = -5.01, p < .001, d = .258.

41

Hypothesis 6, which addresses SEC characteristics and truth-bias, states that the
ability to verify marketing claim increases truth-bias. Consumers will judge messages
about search characteristics as truthful more often than messages about experience
characteristics (Hypothesis 6a). Consumers will judge messages about experience
characteristics as truthful more often than messages about credence characteristics
(Hypothesis 6b). By the transitive ordering logic, consumers will judge messages about
search characteristics as truthful more often than messages about credence characteristics
(Hypothesis 6c). In order to test the components of Hypothesis 6, data for each of the
three SEC characteristics was aggregated across branded and unbranded conditions and
compared pairwise using independent groups t-tests. Truth-bias for search claims was
63.8% (95 CI = 61.6 – 66.0%); truth-bias for experience claims was 51.7% (95% CI =
49.5 – 53.8%); and truth-bias for credence claims was 53.7% (95% CI =51.4 – 56.0%).
Hypothesis 6a was confirmed, t(1004) = 7.70, P < .001, d = .486; truth-bias is
significantly greater when judging search claims than when judging experience claims.
Hypothesis 6b could not be confirmed, t(998) = -1.26, p = .209 (based on Levene’s Test
variances could not be assumed to be equal), d = .079; the mean truth-bias for experience
claims was lower than the mean truth-bias for credence claims, with substantially
overlapping confidence intervals. Despite the unexpected order transposition, Hypothesis
6c was confirmed, t(1002) = 6.195, p < .001, d = .391; truth-bias is significantly greater
when judging search claims than when judging credence claims. Overall, there is
substantially more truth-bias associated with the ability to verify information before
purchasing than exists for claims that cannot be verified in advance, whether those claims
involve experience or credence. Because Hypothesis 6b could not be supported by the

42

data, truth-bias results for experience and credence claims were combined, M = 52.7%
(SD = 25.8%; 95% CI = 51.1 – 54.3%). An independent groups t-test of search versus
other claims shows that truth-bias for search claims is significantly different from the
truth-bias observed when the inference of prior search is not part of the claim, t(1505) =
7.97, p < .001, d = .438.
Hypothesis 7 states there will be an interaction between branding and SEC levels
such that branded messages with search characteristics will be judged as truthful at a rate
significantly higher than unbranded messages about products with credence
characteristics. In order to test this proposition, two analyses were performed, a simple
significance test of the two extreme conditions in the experiment and a two-way ANOVA
to test for main effects and interaction between the branding and SEC moderators. Table
2 shows the mean truth-bias, standard deviations, and confidence intervals for the six
moderator conditions. In addition to having non-overlapping confidence intervals, an
independent groups t-test shows that truth-bias for branded search claims is significantly
higher than truth-bias for unbranded credence claims, t(500) = 8.12, p < .001, d = .725.
Table 2.
Average Truth-bias for Six Moderator Conditions, Branding by SEC Characteristic
Type of Claims
Search
Experience
Credence

Total

Mean

SD

95% CI

n

Branded

68.4

25.6

65.3 - 71.5%

256

Unbranded

59.0

23.5

56.1 - 61.9%

247

Branded

53.1

26.0

49.9 - 56.4%

247

Unbranded

50.2

24.0

47.3 - 53.2%

256

Branded

57.3

27.3

54.0 - 60.7%

255

Unbranded

49.9

25.4

46.7 - 53.1%

246

56.4

26.1

55.1 - 57.7%

1507

43

This hypothesis was also evaluated using a two-way ANOVA to test for main
effects of branding and SEC characteristics and interaction between the two moderator
variables. The branding and SEC inductions significantly influenced truth-bias in the
2

overall model, F(5, 1501) = 19.18, p <.001, with an effect size of partial η = .060. Main
effects on truth-bias were also significant; for branding, F(1, 1506) = 25.29, p < .001,
2

2

partial η = .017, and for SEC categories, F(2, 1505) = 32.67, p < .001, partial η =
.042. However, the interaction between branding and SEC categories did not significantly
impact truth-bias.
Table 3.
Two-way ANOVA for Truth-bias with a Comparison of Branding versus SEC Categories
as Measured and Branding versus SEC Categories Combined (S versus EC)
2

df

MS

F-test

A (Branding)

1

1.62

25.29

< 0.001

0.017

B (SEC)

2

2.10

32.67

< 0.001

0.042

2

0.14

2.19

0.112

0.003

p

Partial η

Branding x SEC categories
Main Effects

Interaction
AB

Branding x Search vs combined Experience and Credence
Main Effects
A (Branding)

1

0.34

7.34

0.007

0.005

B (S vs. EC)

2

0.95

20.20

< 0.001

0.013

2

0.23

4.86

0.028

0.003

Interaction
AB

Since the tests of Hypothesis 6 showed no significant difference for truth-bias

44

across experience and credence categories, these categories were combined and the twoway ANOVA was repeated. Table 3 compares the results of the initial and revised
models. Although collapsing experience and credence into a single category had the
positive result of creating a significant interaction between branding and the reduced SEC
2

factor, F(3, 1504) = 4.86, p < .05, partial η = .003, the overall model and its individual
components were substantially weakened.
Hypotheses 8-10: Moderator Influences on Judgment Accuracy
The final set of hypotheses addresses the moderating effects of branding and the
perceived ability to verify claims (the SEC category induction) on the accuracy of
judging claim veracity. Whereas truth-bias reflects a basic tendency toward truthfulness
and the receiver expectation that messages will be truthful, judgment accuracy reflects
the extent to which the individual’s truth-bias conforms to the actual veracity of messages
sent. Analyses of Hypotheses 5-7 examined the extent to which branding and SEC
category influence, or moderate, truth-bias. Hypotheses 8-10 are propositions regarding
how accuracy changes when truth-bias is affected by the moderating conditions.
Hypothesis 8 states that branding increases the expectation of honesty; as the
number of truthful claims increases as a proportion of total claims, judgment accuracy
will increase, and the increase for branded claims will have a steeper slope and lower
intercept than for unbranded claims. This hypothesis was tested using a two-way
ANOVA to examine the relationship between branding, base rate, and accuracy. Results
2

for the overall model, F(17, 1489) = 6.04, p < .001, partial η = .065, show that branding
and number of truthful messages in the set of eight messages evaluated significantly
influenced the accuracy of respondent judgments. Main effects of branding on accuracy
45

and number of truthful claims on accuracy were both significant as was the effect of the
interaction between branding and number of truthful claims on accuracy. Table 3
summarizes the results of this analysis.
Table 4.
Two-way ANOVA for Accuracy with Branding versus Number of Truthful Claims
2

MS

F-test

17

0.27

6.04

< 0.001

0.065

A (Branding)

1

0.23

5.08

0.024

0.003

C (N-Truthful)

8

0.43

9.63

< 0.001

0.049

8

0.10

2.32

0.018

0.012

df

p

Partial η

Branding x Number of Truthful Claims
Model
Main Effects

Interaction
AC

A t-test of simple slopes (Aiken & West, 1991) for branded versus unbranded
accuracy (with number of truths centered) yields t(1503) = 3.29, p < .001, supporting
Hypothesis 8 but a t-test of the intercept (number of truthful claims = 0) did not support
the hypothesis, t(164) = -1.27, p = .21. Visual comparison of the accuracy trends for
branded and unbranded claims provides further support for Hypothesis 8 but it also raises
questions about the strength of this hypothesis. Figure 3 depicts the results for all branded
2

claims. The linear trend line has a good fit with the observed data resulting in R = .94.
Figure 4 depicts the results for all unbranded claims. The linear trend line for unbranded
claims, with one exception, falls within or at the confidence interval for the observed
2

data; however, the fit results in a relatively weak R = .31. Using a polynomial curve fit
2

improves the fit of predicted to observed results, R = .81.

46

0.75
Branded Claims
0.70

Accuracy

0.65
0.60
0.55
Upper CI
Observed
Lower CI
Linear Fit

0.50
0.45
0.00
0

1

2

3

4

5

6

7

8

Actual Veracity (Number of Truthful Claims)
Figure 3. Observed accuracy as a function of actual veracity (number of truthful claims
2
out of eight presented). The linear fit is y = .0234x + .4827, R = 0.94.

0.75
Unbranded Claims
0.70

Accuracy

0.65
0.60
0.55

Upper CI
Observed
Lower CI
Linear Fit
Polynomial Fit

0.50
0.45
0.00
0

1

2

3

4

5

6

7

Actual Veracity (Number of Truthful Claims)
Figure 4. Observed accuracy as a function of actual veracity (number of truthful claims
2
out of eight presented). The linear fit is y = .0093x + .5636, R = 0.31; the curvilinear
2

2

(polynomial) fit is y = -.0052x + .0513x + .5147, R = 0.81.

47

8

Hypothesis 9 states that the ability to verify the marketing claim increases the
expectation of honesty and as the number of truthful claims increases as a proportion of
total claims, judgment accuracy will increase. The increase for search claims will have a
steeper slope and lower intercept than for experience claims (H9a); the increase for
experience claims will have a steeper slope and lower intercept than for credence claims
(H9b); and by the logic of ordering the increase for search claims will have a steeper
slope and lower intercept than for credence claims (H9c). This hypothesis was first tested
using a two-way ANOVA to examine the overall relationship between SEC category,
base rate, and accuracy. Results for the overall model, F(26, 1480) = 10.34, p < .001,
partial η

2

= .154, show that SEC category and number of truthful messages in the set of

eight messages evaluated significantly influenced the accuracy of respondents’ veracity
judgments. As Table 5 shows, the main effects as well as the interaction between the
number of truthful claims and SEC categories were significant.
Table 5.
Two-way ANOVA for Accuracy with SEC Category versus Number of Truthful Claims
df

MS

F-test

p

2

Partial η

SEC Category x Number of Truthful Claims
26

0.42

10.34

< 0.001

0.154

B (SEC Category)

2

1.55

37.83

< 0.001

0.049

C (N-Truthful)

8

0.48

11.73

< 0.001

0.060

16

0.28

6.88

< 0.001

0.069

Model
Main Effects

Interaction
BC

Pair-wise t-tests of the simple slopes were used to compare SEC results. The test
of search versus experience slopes supported Hypothesis 9a, showing that search claims
48

yield a steeper slope than experience claims, t(1002) = 7.62, p < .001. A t-test of the
intercepts (number of truths = 0) also supported this hypothesis, showing that the
observed search intercept was significantly lower than the observed experience intercept,
t(113) = -5.77, p <.001. The test of experience versus credence slopes did not supported
Hypothesis 9b, showing that credence claims yield a steeper slope than that of experience
claims, t(1000) = -2.21, p < .05; this result was significant but in the opposite direction of
the prediction. In addition, a t-test of the intercepts, t(112) = -.44, p = .661, was not
significant. Hypothesis 9c which predicted that search claims would produce a steeper
slope than credence claims was supported, t(1000) = 5.62, p < .001. The test of intercepts
showed the search intercept was significantly lower than the credence intercept, with
t(101) = -6.74, p < .001. Figures 5, 6, and 7 depict the observed results.

0.80
Search Claims

Accuracy

0.70
0.60
0.50
Upper CI
Observed
Lower CI
Linear Fit
Polynomial Fit

0.40
0.30
0.00
0

1

2

3

4

5

6

7

Actual Veracity (Number of Truthful Claims)
Figure 5. Observed accuracy as a function of actual veracity (number of truthful claims
2
out of eight presented). The linear fit is y = .405x + .389, R = 0.74; the curvilinear
2

2

(polynomial) fit is y = -.00972x + .1177x + .298, R = 0.95.

49

8

0.80
Experience Claims

Accuracy

0.70
0.60
0.50
0.40

Upper CI
Observed
Lower CI
Linear Fit

0.30
0.00
0

1

2

3

4

5

6

7

8

Actual Veracity (Number of Truthful Claims)
Figure 6. Observed accuracy as a function of actual veracity (number of truthful claims
2
out of eight presented). The linear fit is y = .00002x + .562, R = 0.00.

0.80
Credence Claims

Accuracy

0.70
0.60
0.50
0.40

Upper CI
Observed
Lower CI
Linear Fit

0.30
0.00
0

1

2

3

4

5

6

7

Actual Veracity (Number of Truthful Claims)
Figure 7. Observed accuracy as a function of actual veracity (number of truthful claims
2
out of eight presented). The linear fit is y = .011x + .6059, R = 0.66.

50

8

Hypothesis 10 states there will be a significant and positive interaction between
branding and SEC levels such that the relationship between the ratio of truthful to total
messages and detection accuracy will have a steeper slope and lower intercept for
branded search messages than for unbranded credence messages. This hypothesis was
first tested using a three-way ANOVA to examine the overall relationship between
accuracy, number of truthful claims (out of eight claims evaluated), and the moderators
of branding and search category. Results of the three-way ANOVA indicate that the
overall model has a significant affect on judgment accuracy, F(53, 1454) = 6.74, p <
2

.001, partial η = .197. As Table 6 shows, the main effects for branding, SEC category,
and number of truthful claims as well as all two-way interaction effects and the three-way
interaction effect significant impacted accuracy at or beyond p < .01.
Table 6.
Three-way ANOVA for Accuracy with Branding versus SEC Category versus Number of
Truthful Claims
df

MS

F-test

p

2

Partial η

SEC Category x Number of Truthful Claims
53

0.27

6.74

< 0.001

0.197

A (Branding)

1

0.30

7.67

0.006

0.005

B (SEC Category)

2

1.47

37.06

< 0.001

0.049

C (N-Truthful)

8

0.46

11.48

< 0.001

0.059

AB

2

0.28

7.04

0.001

0.010

AC

8

0.11

2.86

0.004

0.016

BC

16

0.29

7.42

< 0.001

0.076

ABC

16

0.09

2.15

0.005

0.023

Model
Main Effects

Interactions

51

In addition, the simple slopes and intercepts of the six moderator conditions were
compared (with particular attention to the extreme conditions of branded search claims
and unbranded credence claims). Table 7 shows the six linear equations predicting
accuracy from number of truthful claims for the six combinations of moderators and the
predicted and actual intercepts (mean accuracy) when 0 truthful claims are evaluated. In
order to support Hypothesis 10, the slopes of the linear equations should decrease from
branded to unbranded conditions within SEC category and from search to credence
claims within branded and unbranded categories. Similarly, the observed intercepts (at
0/8 truthful claims) should be increasing. While the intercepts match the expected pattern,
the slopes do so from branded to unbranded conditions but not across SEC categories.
The slopes for branded and unbranded search are steepest but the slopes for branded and
unbranded credence are higher than those for branded and unbranded experience.
Table 7.
Regression Equations and Observed Intercepts for the Relationship between Accuracy
and Number of Truthful Claims by the Six Combinations of Moderator Variables

Regression
Equation *

Accuracy at Truthful Claims = 0
Observed
+/Predicted Observed
Predicted
0.299
0.245
-0.054

Branded Search

Ŷ = .054x + .515

Unbranded Search

Ŷ = .024x + .575

0.479

0.369

-0.110

Branded Experience

Ŷ = .001x + .542

0.538

0.563

0.025

Unbranded Experience

Ŷ = -.001x + .583

0.587

0.602

0.015

Branded Credence

Ŷ = .017x + .663

0.595

0.571

-0.024

Unbranded Credence

Ŷ = .004x + .636

0.620

0.643

0.023

Note: * Number of truthful claims centered; predicted accuracy at truthful claims = 0
corresponds to the intercept for uncentered truthful claims.

52

Table 8.
Pair-wise Comparison of Simple Slopes and Intercepts for all Moderator Combinations

t

Simple Slope
df
p

Observed Intercept *
t
df
p

Branded Search vs.
Unbranded Search

4.27

499

<.001

-2.23

50

< .05

Branded Experience

7.49

499

<.001

-5.02

60

<.001

Unbranded Experience

7.63

508

<.001

-4.97

51

<.001

Branded Credence

5.28

507

<.001

-5.35

54

<.001

Unbranded Credence

7.16

498

<.001

-6.44

45

<.001

Branded Experience

3.26

490

< .05

-3.14

60

< .05

Unbranded Experience

3.54

499

<.001

-3.32

51

< .05

Branded Credence

1.02

498

ns

-3.41

54

< .05

Unbranded Credence

2.93

493

< .05

-4.55

45

<.001

0.28

499

ns

-0.55

61

ns

Branded Credence

-2.41

498

< .05

-0.13

64

ns

Unbranded Credence

-0.57

493

ns

-1.20

55

ns

Branded Credence

-2.55

507

< .05

0.44

55

ns

Unbranded Credence

-0.71

498

ns

-0.54

46

ns

1.91

497

ns

-1.14

49

ns

Unbranded Search vs.

Branded Experience vs.
Unbranded Experience

Unbranded Experience vs.

Branded Credence vs.
Unbranded Credence

Note: * Mean accuracy at 0 truthful claims (out of each claims judged).
To more rigorously test these observations, Table 8 shows the 15 pair-wise t-tests
of the simple slopes and intercepts for the six moderator combinations. For this
comparison, the number of truthful statements was centered (with the equal number of
truthful and untruthful claims = 0; range = -4 to +4). Positive t values for the slopes
53

indicate that the slope in the first moderator condition is steeper than that of the second
moderator condition. Similarly, higher t values for the observed intercepts indicate a
higher intercept for the first condition when compared to the second condition.
Hypothesis 10 is partially supported by these data. For the simple slope and observed
intercept of accuracy across number of truthful claims the t values are significantly higher
for branded search claims than for claims in the other five conditions, including
unbranded search claims. When the messages evaluated consist of unbranded search
claims, t values are significantly higher in all cases, with the exception of the noted
comparison with branded search claims, and with the slope of branded credence claims.
The remaining t-tests for slopes and intercepts were not significant, with two exceptions;
the slope for branded experience claims was significantly less steep than that of branded
credence claims (in the opposition direction of the prediction) and the slope for
unbranded experience claims was significantly less steep than that of branded credence
claims. Given the interaction between branding and SEC category, the slope of branded
credence claims was expected to exceed only that of unbranded credence claims.
The most extreme comparison is for branded search claims and unbranded
credence claims. The simple regression for branded search is Ŷ = .054x + .515 and has a
significantly steeper slope than the simple regression for unbranded credence, Ŷ = .004x +
.636, with t(498) = 7.16, p < .001. The intercepts for the simple regression are centered;
with the number of truthful claims = 0 the predicted accuracy for branded search claims
is .299 and the observed value is .245 while unbranded credence claims had predicted
accuracy of .620 and the observed value is .643. For the observed intercepts t(45) = -6.44,
p < .001. The results for this extreme comparison support the hypothesis that marketing

54

deception detection accuracy is influenced by the interaction of branding, perceived
ability to search and verify claims prior to purchase, and the number of truthful claims (as
a proportion of total claims) being evaluated.

55

CHAPTER 4: DISCUSSION
The goals of this study were to provide a conceptual replication of the Levine et
al. (2006) test of the Park-Levine probability model (2001) and to extend the model to
show that moderators of truth-bias will alter the relationship between veracity base rates
and deception detection accuracy. The research was conducted using written product
claims such as those found in advertising and other promotional messages. The objectives
of this message choice were twofold. First, a successful replication would show that a
theory of deception detection based on the receiver’s truth-bias can be generalized
beyond interpersonal communication; without substantive adaptation, prior theories of
deception detection cannot be generalized to non-interpersonal situations. Second, a
successful test would demonstrate that truth-bias plays an important role in consumer
decision making. Products messages with search, experience, and credence characteristics
were created. Thus, each message implied one of the three levels of consumer ability to
verify the message content. Each of the messages was also varied by being presented
with or without a brand, and for each message within each of the six SEC by branding
conditions a truthful and not truthful version was generated. For each of the six
moderator conditions, the base rate, or number of truthful messages was experimentally
varied. The truthfulness of eight messages was judged, truth-bias and accuracy were
calculated, and the results were compared to predicted patterns.
Overall, the study shows that the interpersonal model of detection accuracy
replicates when the test is applied to non-interpersonal marketing messages and that the
probability model is generally supported when truth-bias is high. Furthermore, the study
shows that under different circumstances truth-bias will vary; that is, truth-bias toward

56

marketing messages is influenced by the presence or absence of a specific brand name
and whether or not the receiver is presented with claims about product qualities that can
be verified prior to purchase. The study also confirms the expectation, based on the ParkLevine model, that when truth-bias has been altered, detection accuracy across base rates
will change.
Conceptual Replication
The Park-Levine model rests on several critical assumptions. Levine at al. (1999)
showed that receivers of interpersonal messages are most often truth-biased and, because
people make truth judgments more often than lie judgments, they are more likely to make
correct judgments when judging truths than lies (the veracity effect). A correct judgment
occurs when a truthful statement is judged true or an untruthful statement is judged not
true; in other terms, accuracy is the hit rate of correct judgments. By implication, the
overall hit rate for detection accuracy is contingent upon the ratio of lies to the total
number of statements being evaluated and slightly better than chance. Accuracy will be
higher when judging a set of statements with more truths than lies and, because of the
veracity effect, it will be lower when judging a set that contains more lies that truths.
The current study included several hypotheses derived from the Levine et al.
(2006) test of the Park-Levine model. Results of this experiment supported the
propositions that consumers are truth-biased and provided further evidence for the
veracity effect, finding truth accuracy to be greater than lie accuracy. The veracity effect
occurs because new information is by default deemed truthful (Gilbert, 1991; Gilbert,
Krull, & Malone, 1990) and because truth-bias reflects the typically low frequency of lies
and lying in everyday life (Anderson, Ansfield, & DePaulo, 1997; Serota, et al., 2010).

57

Replication of the Park-Levine test without interpersonal messages adds support for the
theoretical model across a broader range of communication phenomena and provides a
strong argument for generalizing the Park-Levine model.
The original test of the Park-Levine model showed that detection accuracy is a
linear and positive function of base rate. The current study provides evidence that
accuracy is a function of base rate and that in the most common, everyday circumstances
the function is linear and positive. However, the data do raise questions about whether or
not this proposition holds in all situations. Despite the ANOVA results, visual inspection
(see Figure 2) and curve fitting suggest that there may be a threshold for accuracy as the
proportion of truthful messages increases. While the data did not provide any evidence
that consumer skepticism leads to lie-bias, it does suggest that consumers are less truthbiased when it comes to marketing communication than they are when evaluating
messages that occur interpersonally. The Park-Levine test included a hypothesis that lie
accuracy would be less than 50%. Due to the nature of marketing messages as well as the
possibility that variation across moderator conditions might confound the results (branded
and unbranded messages did produce distinctly different patterns), this hypothesis was
not part of the assessment of marketing truths and lies. Had it been, it would have been
rejected; claims that were not truthful were still judged correctly more than half the time.
All together this suggests that when people take on the role of consumer they may not
become lie-biased but they are more skeptical of marketers than of people they interact
with interpersonally in their day-to-day experience.
Moderation of Marketing Claims
The study replicated and validated key elements of the test of the Park-Levine

58

model by demonstrating that consumers are truth-biased and that this affects accuracy
when judging the veracity of claims about products and their attributes. In addition, the
study examined whether or not truth-bias could be moderated so that the accuracy of
consumer judgments would be altered. Park and Levine (2001) noted that many
researchers have tried to find factors that can affect detection accuracy, such as training,
familiarity, and receiver suspicion. In most instances, the studies cited (cf. Park &
Levine, 2001) were designed to examine direct influences on judgment accuracy.
However, it seems more likely, in light of direct testing of the probability model and
subsequent research on the antecedents of detection accuracy (e.g. Levine et al., in press)
that most factors moderate truth-bias directly but influence the outcome indirectly as the
resulting degree of truth-bias interacts with the base rate in order to determine the
veracity judgment accuracy. The results of the current study support the proposition that
truth-bias can be moderated, and that when base rate effects are combined with the
moderated levels of truth-bias accuracy will increase or decrease.
Marketing messages have many characteristics, some that are part of the
communication process and others that are associated with the product or the marketing
exchange. Factors such as message strength, elemental aspects of the message (verbal and
visual components, length, type face, arrangement), form of a persuasive argument,
source credibility (when a presenter is used or an authority is cited), channel features
(print versus radio or television or Internet and whether or not the message is
synchronous or asynchronous), noise, and receiver comprehension might influence how
the message is received and interpreted. Other elements such as product category, brand
equity, consumer knowledge and/or experience, and the consumer’s likelihood of acting

59

upon the message due to motivation, resources, or ability also constitute interpretive
influences. Among these, the study focused on branding (in its simplest form, presence or
absence) and the economic cost/benefit trade-off associated with information search prior
to purchase. Brand is both a signal (Erdem & Swait, 1998) and, in this research, a proxy
for many kinds of information a consumer might obtain about a product from reading ads
or watching commercials or receiving other kinds of promotional messages. The SEC
categorization captures a critical aspect of the consumer decision-making process by
representing levels of difficulty in the information search phase; although marketing
messages may be targeted to any phase of the decision process (for example, to build
awareness or enhance customer post-purchase satisfaction), during information search the
consumer may be most proactive with regard to processing marketing communications.
Thus, branding and SEC categories were selected to test for moderating effects.
Moderation of truth-bias. The study provides clear evidence that truth-bias is
significantly higher when evaluating branded messages than when evaluating the same
claims without the brand. In many ways, branding serves the same kind of trust-building
function that McCornack and Parks (1986) noted in their review of familiarity as an
influence in relational deception detection. Brands convey information about the product
and the company. Those brands that are familiar, strong, favorably evaluated, have
unique associations, and are accessible in memory have high brand equity (Keller, 1993);
brand equity is a predictor of consumer trust. Even at the most basic level, the mere
presence of a brand affords the opportunity to convey brand equity while the absence of a
brand does not. Though some brands may convey negative equity, most consumers treat
branding as a heuristic inferring that the presence of the brand is at a minimum an effort

60

to obtain the consumer’s trust. Even though different brands are likely to convey different
degrees of trust (e.g. Fidelity Investments versus Madoff Investment Securities), the
presence of the brand, even when not well known, may be taken as a sign that the product
can be trusted (e.g. Madoff Investment Securities prior to the scandal). The study results
suggest that branding engenders a willingness to trust, and that truth-lie judgments will be
more truth-biased in the presence of a brand.
Results of the study show that the ability to search before purchasing leads to
more truth- bias than when the consumer must rely on post-purchase experience or
credence. However, the subordinate proposition that claims with experience qualities lead
to more truth-bias than claims with credence qualities was not supported. There are
several ways that the three types of claims are differentiated including association with
product types and differing degrees of difficulty in product evaluation. Whether or not a
consumer actually engages in search behaviors, the knowledge that one has the ability to
verify the information presented in a marketing claim appears to be sufficient to induce
trust as demonstrated by higher levels of truth-bias. However, the failure to find different
levels of truth-bias when judging experience and credence claims suggests that the prepurchase versus post-purchase distinction is a more critical feature of SEC claims than
either the difficulty of evaluation, which is central to Zeithaml’s (1981) view or the basic
cost-benefit proposition of Darby and Karni’s original (1973) classification.
Similarly, the study found mixed support for the interaction between branding and
SEC. A simple comparison of the extreme conditions, claims that are branded and can be
searched prior to purchase versus claims that are unbranded and rely on the credence or
perceived veracity of the seller, yields significantly different levels of truth-bias. But an

61

ANOVA of the 2 x 3 moderator conditions supports only the main effects. When
experience and credence claims are combined in order to test the pre-purchase versus
post-purchase conceptualization of search in conjunction with branding, both main
effects and the interaction are supported. When consumers are presented with a brand and
have the ability to verify the information given in the message they are significantly more
likely to manifest truth-bias than if they only know the brand or only know that the
information is verifiable prior to purchase.
Effect on accuracy. The study inductions demonstrated that the level of truth-bias
can be increased or decreased through the influence of branding and SEC categories
(albeit the later influence might better be described as the ability to engage in prepurchase versus post-purchase verification). The Park-Levine model allows for the
possibility of different levels of truth-bias but notes that accuracy is a function of the
correct identification of both truths and lies. As truth-bias increases, by definition the
judgment of lies will become less accurate. Total accuracy is an additive function of both
truth and lies judgments, P (T ∩ H) + P (~T ∩ ~H); total accuracy equals the probability
of a true statement being judged as honest plus the probability of a not true statement
being judged as not honest. Unless the two probabilities are equal, the total accuracy will
vary with the proportion of lies to total messages. As the probability that the message
being judged is a lie increases, accuracy will decline. To understand the overall effects of
the moderators on truth-bias it was necessary to examine accuracy as well as truth-bias
across the moderator conditions. This was done be systematically varying the base rate
and examining the slopes and intercepts for accuracy plotted against base rates.
The original test of the Park-Levine model required subjects to judge eight

62

videotaped interviews in which the respondent was either lying or telling the truth. In
order to test the propositions of the study, the number of lies were varied from 8, P (T) =
0.0 to 0, P (T) = 1.0. Critical measures of the outcome of the study were the slope and
intercept, which described the relationship of total judgment accuracy to the base rate. In
order to test the effects of moderators on the marketing messages, this design was
replicated for each of the six moderator conditions and the slopes and intercepts were
compared across conditions.
Branding and the perceived ability to verify the information presented influence
accuracy when the base rate is varied. As expected, two-way ANOVA results for
branding and base rate, two-way ANOVA for SEC categories and base rate, and the
three-way ANOVA for branding, SEC categories, and base rate all had significant main
effects and interactions with accuracy as the dependent measure. The nature of these
effects, when examined as slopes and intercepts, is more complicated. The slopes and
intercepts need to be examined separately from the analysis of variance.
Comparing branded conditions to unbranded conditions, as predicted the slope of
accuracy scores across base rates is much steeper when the messages are branded. The
higher the truth-bias, the steeper the slope of accuracy will be across base rates. Since
branding promotes an increase in truth-bias, accuracy for conditions that consist mostly
of untruthful statements will have low accuracy and conditions in which the statements
are mostly truthful claims will have high accuracy. In practical terms this means that
branding encourages consumers to expect the truth. To most consumers a brand is
supposed to stand for something positive and it is a violation of normative expectations
for the owners of a brand to lie in their promotional communications. Sellers also

63

understand that the consumer has the power to sanction the brand by not purchasing, by
not re-purchasing, by giving it low customer satisfaction ratings, or by attacking the
reputation of the brand through negative word-of-mouth. The data show that the absence
of a brand does not lead to lie-bias but the strength of the truth-bias is less than when the
brand is present. Therefore, the base rate conditions that are dominated by lies are severe,
non-normative trust violations and judgment accuracy is correspondingly low. It is worth
noting that this is also strong evidence that consumers cannot tell the truth from a lie.
Accuracy patterns for the three SEC categories are consistent with the results for
moderator influences on truth-bias. The slope of accuracy scores over base rates in the
search condition was significantly steeper than the slopes in the experience and credence
conditions. However, when the slopes for experience claims and credence claims were
compared, the results were not in the predicted direction. As with the truth-bias results,
accuracy results lead toward the conclusion that the pre-purchase ability to verify
information increases truth-bias and that conditions dominated by lies will have low
accuracy while conditions dominated by truthful statements will have high accuracy.
Akerlof (1970) argued that asymmetry of information leads to a greater likelihood of
cheating by sellers. The ability to search before buying should be considered by
consumers as an opportunity to achieve symmetry. Knowing that search reduces
uncertainty, consumer are likely to anticipate a higher level of seller truthfulness and will
be inclined toward truth-bias.
In addition to examining the slopes, this study looked at the intercepts; these
represent the accuracy for base rate conditions in which the subjects were presented with
only lies. It was presumed that the steepness of the slopes would result in lower intercepts

64

corresponding to high truth-bias and higher intercepts corresponding to low truth-bias.
Follow the logical development of the Park-Levine model, the hypotheses of this study
did not account for the possibility that truth-bias might vary across base rate conditions.
Support for the intercept component of the hypotheses was mixed.
The current study anticipated that truth-bias is independent of base rate. As
expected, the slopes of truth-bias plotted against base rates were nearly zero (flat) when
consumers were faced with the pre-purchase verification ability suggested by search
claims. But in the post-purchase verification situations associated with experience and
credence claims, truth-bias was shown to increase as the base rate changes from mostly
lies to mostly truths. In these conditions, accuracy remains nearly stable (flat) across the
base rates. This pattern suggests an unmeasured influence on truth-bias. The effect is that
consumers are more likely to distinguish truths from lies. One possibility is familiarity
with the brand. Kardes, Posavac, and Cronley (2004) suggest that a brand name heuristic
involving reputation can be a source of product inference, especially in low motivation
evaluation situations (Mahewswaran, Mackie, & Chaiken, 1992). The current study used
claims involving real brands that vary along geographic lines (local to national),
prominence (extent to which the product comes for a leader brand), and time in the
market (new brands versus established brands). Although Park and Levine (2001) note
that factors such as familiarity, suspicion, and probing are often not general across truth
and lie accuracy, these influences might account for variation in the difference between
truth accuracy and lie accuracy. Further, when consumers evaluate a mix of better-known
and lesser-known brands under conditions of higher uncertainty (inability to verify the
information prior to purchase) the results may cancel out, creating a pattern of constant

65

truth-bias across base rates. Given the limitations of the current study, this may be fruitful
ground for future investigation.
When both branding and SEC category moderators are measured in conjunction
with a base rate induction, the interaction is expected to be significant (it was), and the
slopes and intercepts for accuracy across base rates should indicate that truth-bias has
more influence on accuracy when the consumer is relying on branding and when the
benefit of verifying the claim is high relative to the search costs (they do). However, even
though the three-way ANOVA shows significance for all main effects, all two-way
interactions, and the three-way interaction there remains an important caveat. The data
show that SEC category is an influential moderator because of the pre-purchase versus
post-purchase evaluative quality of search claims versus both experience and credence
claims. But the differences between effects of experience claims and credence claims are
not consistent with the hypotheses. When consumers feel that they can only verify claims
after purchase (experience and credence claims), they are less likely, on average, to be
truth-biased and their accuracy is more likely to be uniform across base rates. A higher
degree of truth-bias, as noted in the branded and search claim conditions will be
associated with greater variability in accuracy. This variability will be consistent with the
Park-Levine mode.
Limitations
The current study has several limitations that may affect the observed results.
First, the claims used and the way in which they were presented may have both
methodological and conceptual implications for the results. The claims were created
using an ideation process to assign products and product qualities to the three SEC

66

categories. Each claim was paired with a specific brand and a truthful claim was written.
The claim was then adapted to create a not truthful version and finally, two generic
versions were created by removing the brand identification from both the truthful and not
truthful claims. Since the three categories are a continuum representing the difficulty of
information acquisition and evaluation, and this continuum is believed to be correlated
with factors that distinguish between goods and services, the categories included
distinctly different product sets (cf. Zeithaml, Parasuraman, & Berry, 1985 for a review
of the distinguishing factors). In the current study there was no attempt to assure that the
products or brands were a representative cross-section within each category. Instead,
products were selected because they were the highest scoring on coder agreement for
assignment across categories.
In addition, the possible lack of representativeness may be exacerbated by the
imbalance of products across categories. The ideation process generated many search
products; therefore the high scoring and most consistently items may not representative
the range of products that fall into this category. Conversely, the number of credence
products was small and all of the products that met a minimum agreement criterion were
selected. Tests of the data involving SEC categories were the most likely to be
unsupported, but this could be the result of a weak operationalization rather than weak
theory.
It should also be noted that currently there is no research providing an objective
measure of the distribution across SEC categories of products available in the
marketplace. This raises concerns about ecological validity. Results based on aggregated
moderator conditions may be skewed to the extent that the equal distributions used in the

67

study deviate from the real-world. A related question is the validity of interpreting results
for SEC categories and unbranded claims. Logically, the product category and qualities
that are part of each claim should imply the extent to which consumers might expect to
search and verify the claim. However, without an identified brand, consumers might not
factor searchability into their veracity judgment. If the analysis of SEC category effects
had been limited to only data from the branded conditions, the results of the study would
not have changed but the evidence might be consider stronger.
Another issue is the randomization of claims assigned to the base rate inductions.
The assignment of product claims to truthful and not truthful positions within each base
rate cell was random. However, all subjects evaluating a cell read the same set of eight
claims (the order of presentation was randomized across subjects). Ideally each subject
would have received a unique mix of truthful and not truthful claims (e.g. Subject 1 in the
3/8 not truthful base rate condition would have read a truthful claim for Arcadia bottled
water while Subject 2 would have read the not truthful claim for the same product).
Because of the product and brand assignment concern already noted, the fixed set of
claims may introduce an unmeasured bias. For reasons associated with cost and
feasibility (subject control within the online panel) the study did not include an induction
check. This is a serious shortcoming. Research to validate the finds of the present study
should include additional measures to confirm that subjects perceive the claims fit into
the categories to which they were assigned.
Concern about the item distribution also raises a conceptual issue. Studies of
brands and products obtain evaluations across many dimensions. Products are
differentiated along the goods and services continuum, with services distinguished by a

68

higher degree of intangibility, non-standardization, and the inseparability of production
and consumption. Goods may also be durable or consumable. Brands are differentiated
by familiarity, brand equity, and the components of brand equity such as brand strength
of associations, brand loyalty, perceived quality, and brand value. Brands may also be
distinguished by their owners’ brand architecture; the brand may be a master brand or a
sub-brand, it may represent a house of brands or a branded house. Any of these may be a
confounding factor that affects the consumer’s truth-bias. For example, consumers may
use different schema for Ford (the company) and Ford (the car division), and the brand’s
contribution to truth-bias may differ depending on which of these the consumer
associates with a specific marketing claim.
Implications
Validation of the Park-Levine model. This study validates the Levine et al. (2006)
test of the Park-Levine probability model using marketing messages instead of
interpersonal interviews. It shows that consumers cannot readily distinguish the truth or
falseness of marketing claims. And consistent with the Park-Levine probability model,
and theory developed in the context of interpersonal communication, truth-bias interacts
with the base rate of actual message veracity so that the accuracy of veracity judgments is
low when the messages are predominantly not truthful and high when the messages are
predominantly truthful. This is an important generalization for theories of deception
detection. It shows that truth-bias theory is supported beyond interpersonal messaging.
The psychophysical model used by Ekman, the trait model developed by Zuckerman,
DePaulo, and Rosenthal, and, especially, the IDT model proposed by Buller and Burgoon
are dependent upon the abilities of receivers to decode the cues and signals of a human

69

sender. As this study shows, Levine’s truth-bias theory, which is in part predicated on the
Park-Levine probability model, can be used to examine and explain messages that are
mediated, interposed, and asymmetrical; in short, the theory works with noninterpersonal as well as interpersonal messages.
Extending the Park-Levine model to include moderators. This study provides
ample evidence that truth-bias is moderated by external influences. This extends the
Park-Levine model in a complex but useful way. As Park and Levine (2001) note, there
have been many efforts to find factors that affect detection accuracy (cf. Park & Levine,
and Levine et al., 2006 for discussion). However, in nearly every instance the research
has attempted to show a direct effect of these factors on accuracy. Park and Levine show
that accuracy is a function of truth-bias, that truth and lie accuracy must be addressed
separately, and that truth-bias has an interaction with veracity base rate(s).
Philosophically (Bok, 1999) and empirically (DePaulo et al., 1996; Serota et al., 2010),
real-life interactions and messaging are inherently more honest than not; far more so than
is represented by the typical 50% truths and 50% lies present in most deception detection
studies (Bond & DePaulo, 2006). Thus, most efforts to identify the factors affecting
detection accuracy fail because those models try to predict overall accuracy without
accounting for the conditional nature of truth and lie accuracies. The current study shows
that certain factors have clear moderating effects on truth-bias, which in turn are reflected
in varying degrees of accuracy across base rates.
Future research. These results suggest many opportunities for additional research,
two of which have the most immediacy. First, the evidence to support branding and SEC
category effects suggests that other variables related to the veracity and believability of

70

marketing claims need to be investigated. These include the measures of brand equity that
capture dimensions of the brand such as strength and value. They might also include
measures of how information is processed when making these judgments. For example,
Kardes, Posavac, and Cronley (2004) categorized consumer inferences according to a 2 x
2 x 2 model consisting of induction versus deduction, stimulus-based versus memorybased, and singular versus comparative judgments. While Kardes et al. (2004) focus on
inferences about product qualities leading to a purchase decision, these approaches could
equally well be applied to assessing the veracity of the information on which that
decision is based. Second, success of the current study invites reconsideration of
influences on interpersonal veracity judgments. For example, familiarity and suspicion
have been shown to have some effect on accuracy, but if it can be demonstrated that they
have differential effects on the interaction between truth-bias and base rate, these
variables may turn out to be more powerful predictors than original believed.
Marketing applications. Much attention in the marketing literature on deception
has been given to the risks associated with false information. Some of this has focused on
fairness to the competition; more recently, concern for the welfare of consumers has been
conjoined to concerns for the enterprise. Boush et al. (2009) suggest that too much
attention is given to deception research in support of legal remedies and government
regulation to provide consumer protection. Akin to Ekman’s view that people can be
better lie detectors if they learn to read the nonverbal cues, Boush et al. argue that
consumer self-protection should be the goal of research on marketing deception. The
current study suggests this goal may be misdirected. The data make a case that consumers
need to be more self-aware, especially of the human tendency toward truth-bias. But that

71

does not mean the consumers can be better lie detectors. Simply being more skeptical
would shift the odds in favor of the consumer when marketing lies are pervasive. But it
would also make consumers less accurate in situations where marketing messages are
basically truthful. Government regulation and legal enforcement may be good solutions
if, as study results suggest, consumers are not good lie detectors. Akerlof’s economic
theory of asymmetrical information argues that sellers act in their own self-interest.
Sellers who understand truth-bias and consumer fallibility are more likely to engage in
deception as strategy when the market offers no effective sanctions. Given the results of
this study, regulation is likely to be a more effective safeguard than trying to promote
self-protection by trying to make consumers better deception detectors.
Park, Levine, McCornack, Morrison, & Ferrara (2002) provide evidence that
detection needs to focus more on contextual features, such as a priori knowledge, third
party information, and overt challenges to elicit truthful information in interpersonal
communication. Similarly, future research on marketing deception might include more
work on contextual influences, like the SEC characteristics. It would also be useful to
determine the actual prevalence of marketing deception. Finally, it would be valuable to
examine the relationship between base rates for marketing lies and the availability of
norm-enforcing influences such as industry self-regulation, government regulation, and
the reputational influences of third-party product evaluations, consumer blogs, and wordof-mouth.
Conclusion
This study provides a validation of the Park-Levine probability model. The results
were obtained by evaluating truthful and not-truthful marketing claims. These claims

72

were moderated by the presence or absence of a brand and the perceived ability to verify
the claim (operationalized by assignment of claims to three categories derived from the
economics of information – search, experience, and credence). Results show that the
moderators influence consumer truth-bias and, as a consequence, the pattern of accuracy
across veracity base rates was altered. This outcome is important because it demonstrates
the generalizability of the Park-Levine model beyond its application to interpersonal
interactions; competing theories are inherently limited to interpersonal communication or
the observation of human nonverbal behavior. The successful addition of moderator
variables adds explanatory power to this theoretical extension into marketing
communication, and it invites a reassessment of previously-considered influences on
interpersonal deception detection.

73

APPENDICES

74

Appendix A: Questionnaire
Appendix A illustrates the basic form of the study questionnaire. The survey was
presented as a single page on the Synovate e-Nation omnibus survey website (other pages
on the website included instructions, other omnibus client questions, demographics, and
the closing redirection to a sweepstakes site for survey participants. Subjects read the
introductory statement and then, for each of eight claims, indicated whether or not they
believed the marketing claim statement was truthful or not truthful. The number of actual
truthful and not truthful statements varied by treatment group; the order of statement
presentation was randomized for each subject. A sample of one questionnaire from the
branded search cell is shown:
The following list includes eight statements that you might find in an advertisement or
commercial or on a company or store website. Read each statement and tell us whether or
not you think the statement is truthful. Even if you are not sure, please tell us if you think
the statement is truthful or not. Check one box for each statement:
Only the Samsung Galaxy’s DLNA output lets you stream content
effortlessly from your phone to other devices like your TV.

[ ] Truthful
[ ] Not truthful

Gain expertise in energy management. Begin or finish your energy
management degree at Walsh College.

[ ] Truthful
[ ] Not truthful

See Morocco now. Delta flies you direct to Casablanca via Royal Air
Moroc.

[ ] Truthful
[ ] Not truthful

Welcome to Circle Cross Ranch. This inviting new neighborhood is
close to local shopping malls.

[ ] Truthful
[ ] Not truthful

Flatbush Gardens apartment complex is home to an eclectic
community of families, young professionals, and seniors.

[ ] Truthful
[ ] Not truthful

Jobe’s Organics All-Purpose Fertilizer Spikes give gardeners an
environmentally friendly option for healthy plants.

[ ] Truthful
[ ] Not truthful

Colony Bank’s Albany branch is open to service your banking needs
on Saturday.

[ ] Truthful
[ ] Not truthful

Carfax estimates a used car’s average annual miles driven based on the [ ] Truthful
last odometer reading reported to Carfax.
[ ] Not truthful

75

Appendix B: Claims
Appendix B lists the statements used in the study. Separate sets of statements
were used for each of the three search conditions and these were further divided by
creating branded and unbranded versions. A total of 96 statements were created, 16 for
each of the six moderator conditions (two branding conditions x three search conditions).
Each set of 16 statements consisted of truthful and not truthful versions of eight claims.
For each claim, a subject only saw one version, either the truthful or not truthful claim.
The veracity base rate was determined by the number of truthful claims presented among
the eight claims judged; the base rate ranged from 0% (0/8 claims) to 100% (8/8 claims).
The 96 claims, by moderator condition, are:
Branded Search – Not Truthful
110
120
130
140
150
160
170
180

Only the Samsung Galaxy’s DLNA output lets you stream content effortlessly
from your phone to other devices like your TV.
Gain expertise in energy management. Begin or finish your energy management
degree at Walsh College.
See Morocco now. Delta flies you direct to Casablanca.
Welcome to Circle Cross Ranch. This inviting new neighborhood is close to
local shopping malls.
Flatbush Gardens apartment complex is home to an eclectic community of
families, young professionals, and seniors.
Jobe’s Organics All-Purpose Fertilizer Spikes give gardeners an environmentally
friendly, safe-to-handle option for healthy plants.
Colony Bank’s Albany branch is open to service your banking needs on
Saturday.
Carfax obtains up-to-date mileage reports for every vehicle registered in the
United States and Canada.

Branded Search – Truthful
111

The Samsung Galaxy’s DLNA output lets you stream content effortlessly from
your phone to other devices like your TV.
121 Gain expertise in management. Begin or finish your management degree at
Walsh College.

76

131
141
151
161
171
181

See Morocco now. Delta flies you direct to Casablanca via Royal Air Moroc.
Welcome to Circle Cross Ranch. This inviting new neighborhood is only 12
miles from local shopping malls.
Flatbush Gardens apartment complex is home to a predominantly homogeneous
community of Black, West Indian, and working class families.
Jobe’s Organics All-Purpose Fertilizer Spikes give gardeners an environmentally
friendly option for healthy plants.
Colony Bank’s Albany branch is open for drive-thru and ATM service on
Saturday.
Carfax estimates a used car’s average annual miles driven based on the last
odometer reading reported to Carfax.

Unbranded Search – Not Truthful
210
220
230
240
250
260
270
280

Only our DLNA output lets you stream content effortlessly from your phone to
other devices like your TV.
Gain expertise in energy management. Begin or finish your energy management
degree at our college of business.
See Morocco now. We fly you direct to Casablanca.
Welcome. This inviting new neighborhood is close to local shopping malls.
This apartment complex is home to an eclectic community of families, young
professionals, and seniors.
Our fertilizer spikes give gardeners an environmentally friendly, safe-to-handle
option for healthy plants.
We are open to service your banking needs on Saturday.
We obtain up-to-date mileage reports for every vehicle registered in the United
States and Canada.

Unbranded Search - Truthful
211
221
231
241
251
261

Our DLNA output lets you stream content effortlessly from your phone to other
devices like your TV.
Gain expertise in management. Begin or finish your management degree at our
college of business.
See Morocco now. We fly you direct to Casablanca via our code-sharing partner.
Welcome. This inviting new neighborhood is only 12 miles from local shopping
malls.
This apartment complex is home to a predominantly homogeneous community of
Black, West Indian, and working class families.
Our fertilizer spikes give gardeners an environmentally friendly option for
healthy plants.

77

271
281

We are open for drive-thru and ATM service on Saturday.
We estimate a used car’s average annual miles driven based on the last odometer
reading reported to us.

Branded Experience – Not Truthful
310
320
330
340
350
360
370
380

Alice Huang’s massage therapy is guaranteed to successfully treat patients for
back pain.
Patagonia Capilene active briefs are the only choice for serious hiking and
trekking.
Enjoy new Pepsi Blue, the fusion of berry and cola that tastes like real
blueberries with something extra.
Notox Hangover Treatment is an all natural herbal supplement that helps combat
the effects of alcohol consumption.
Prego Traditional Italian Sauce brings you the taste of authentic Italian home
cooking.
Mina’s Salon offers unparalleled service.
You will love the fragrance of Chanel No. 5, the best-selling perfume of all time.
Bounty paper towels have the absorbency and cloth-like durability that works for
every cleaning job.

Branded Experience - Truthful
311
321
331
341
351
361
371
381

Alice Huang has used massage therapy to successfully treat patients for back
pain.
Patagonia Capilene active briefs provide quick-drying comfort for serious hiking
and trekking.
Enjoy new Pepsi Blue, the fusion of berry and cola that tastes unlike anything
else on the market.
Notox Hangover Treatment is an all natural herbal supplement that rarely
produces side effects.
Prego Traditional Italian Sauce brings you the taste of sweet tomatoes and savory
Italian herbs.
Mina’s Studio has been voted the best salon in the Hearld-Sun Reader’s Choice
Awards.
Be one of the women who love the fragrance of Chanel No. 5. Try the bestselling perfume of all time.
Bounty paper towels have the absorbency and cloth-like durability that you
would expect from a leading brand.

78

Unbranded Experience – Not Truthful
410
420
430
440
450
460
470
480

Our massage therapy is guaranteed to successfully treat patients for back pain.
Our synthetic active briefs are the only choice for serious hiking and trekking.
Enjoy the fusion of berry and cola that tastes like real blueberries with something
extra.
This hangover treatment is an all natural herbal supplement that helps combat the
effects of alcohol consumption.
Our traditional Italian sauce brings you the taste of authentic Italian home
cooking.
Our salon offers unparalleled service.
You will love the fragrance of this best-selling perfume of all time.
This paper toweling has the absorbency and cloth-like durability that works for
every cleaning job.

Unbranded Experience - Truthful
411
421
431
441
451
461
471
481

We have used massage therapy to successfully treat patients for back pain.
Our synthetic active briefs provide quick-drying comfort for serious hiking and
trekking.
Enjoy the fusion of berry and cola that tastes unlike anything else on the market.
This hangover treatment is an all natural herbal supplement that rarely produces
side effects.
Our traditional Italian sauce brings you the taste of sweet tomatoes and savory
Italian herbs..
Our salon has been voted the best salon in the local paper’s readers’ poll.
Be one of the women who love the fragrance. Try the best-selling perfume of all
time.
This paper toweling has the absorbency and cloth-like durability that you would
expect from a leading brand.

Branded Credence – Not Truthful
510
520

Try Arcadia bottled natural water – 100% pure.
Lafayette Auto Repair technicians will always provide the correct diagnosis for
your car repair.
530 Eating 1.5 ounces of Diamond walnuts each day can reduce the risk of heart
disease.
540 If your transmission needs to be replaced, an AAMCO expert technician rebuilds
it with better than original parts.
550 At Jackson Hewitt we know the ins and outs of tax preparation, so you can feel

79

confident we’ll find all the deductions and credits you’re entitled to.
560 La Prairie Cellular Luxe Lip Colour hydrates and makes the lips shine to prolong
a youthful appearance.
570 CofmanTownsley Injury Lawyers – when others leave you suffering, we have the
Power to win the compensation you deserve.
580 Enrolling in the engineering program at Kansas State University will open the
door to an assured career in engineering.
Branded Credence - Truthful
511
521
531
541
551
561
571
581

Try Arcadia bottled natural water - filtered and treated for purity.
Lafayette Auto Repair technicians can efficiently diagnose your car repair.
Diamond walnuts are high in unsaturated fat. Unsaturated fat has been shown to
contribute to the reduction of bad cholesterol.
If your transmission needs to be replaced, an AAMCO expert technician rebuilds
it with new OEM-quality or better parts.
At Jackson Hewitt, preparing tax returns is what we do. And that could mean
more money in your pocket.
La Prairie Cellular Luxe Lip Colour hydrates and makes the lips shine for a
beautiful appearance.
CofmanTownsley Injury Lawyers – when others leave you suffering, we stand up
for your rights to ensure you get the money you need to cover costs.
Enrolling in the engineering program at Kansas State University can open the
door to professional careers in engineering.

Unbranded Credence – Not Truthful
610
620
630
640
650
660
670
680

Try our bottled natural water – 100% pure.
Our auto repair technicians will always provide the correct diagnosis for your car
repair.
Eating 1.5 ounces of walnuts each day can reduce the risk of heart disease.
If your transmission needs to be replaced, our expert technicians completely
rebuild it with better than original parts.
We know the ins and outs of tax preparation, so you can feel confident we’ll find
all the deductions and credits you’re entitled to.
This lipstick hydrates and makes the lips shine to prolong a youthful appearance.
When others leave you suffering, our personal injury lawyers have the Power to
win the compensation you deserve.
Enrolling in the engineering program will open the door to an assured career in
engineering.

80

Unbranded Credence - Truthful
611
621
631
641
651
661
671
681

Try our bottled natural water - filtered and treated for purity.
Our auto repair technicians can efficiently diagnose your car repair.
Walnuts are high in unsaturated fat. Unsaturated fat has been shown to contribute
to the reduction of bad cholesterol.
If your transmission needs to be replaced, our expert technicians rebuild it with
new OEM-quality or better parts.
Preparing tax returns is what we do. And that could mean more money in your
pocket.
This lipstick hydrates and makes the lips shine for a beautiful appearance.
When others leave you suffering, our personal injury lawyers stand up for your
rights to ensure you get the money you need to cover costs.
Enrolling in the engineering program can open the door to professional careers in
engineering.

81

Appendix C: Ideation Stage 1 - Stimulus Paper
Appendix C is a short paper provided to the undergraduate research assistants
who participated in the first stage of the claim generation process. The research assistants
were asked to read the paper, which explains the economic concepts of search,
experience, and credence. The research assistants were given the opportunity to ask
questions and clarify their understanding of the concepts. This was followed by a
brainstorming session to generate examples of each of the three types of products and/or
product characteristics. The full text of the paper follows:

[Paper begins.]
Search, Experience, and Credence in Consumer Behavior: An Explanation
In 1961, George J. Stigler’s paper, The Economics of Information, was published
in The Journal of Political Economy. Stigler proposed that the search for information is
an important aspect of economic activity and that the nature and availability of
information is critical to the determination (ascertainment) of market price. In 2001
Stigler shared the Nobel Prize in Economics with A. Michael Spence and George A.
Akerlof for this work, which collectively became know as the analysis of markets with
asymmetrical information.
In a sentence, asymmetrical information means that buyers and sellers do not have
equal information when they enter into a transaction and the extent to which that
information is unequal will affect the outcomes of the transaction. In general, the seller
has more control over information about the product and the exercise of this control will
influence the economic power of both parties. Seller control means that the seller may

82

provide full disclosure of information, withhold some information, withhold all
information, or create false information.
Market price is the price the buyer ultimately pays for the product. Stigler
proposes that prices vary frequently and, with few exceptions, no one can know all the
prices that sellers of a given product are quoting at a given time. Therefore, to ascertain
the most favorable price, buyers must search to find various price quotes. Stigler goes on
to assess the cost of search against its value in reducing the price to the buyer. If the
prospect of significantly reducing the price is high, more searches are warranted; if the
prospect is low, the value of the search activity is also low and the buyer will engage in
limited search activity.
Stigler focuses primarily on the price of goods, noting that the dispersion of prices
is a measure of ignorance in the market. He also observes that this measure is biased
because the products are never homogeneous; the products can vary by the services the
seller performs, by the range of products stocked, and (though Stigler explicitly avoids it)
the quality of the products. Responding to Stigler’s concern, Nelson (1970, 1974)
examined the role of quality and the difficulty of obtaining information about the
performance characteristics of the product under various conditions of use. He argues that
the utility of quality would vary more than the utility of price because of the difficulty of
obtaining information about quality and performance. Nelson further argues that one
solution to this problem is that buyers rely on their post-purchase experience with the
product. Some products (perhaps many) do not afford the opportunity to search and
determine which has the best quality. In these cases, the buyer may choose to use the

83

product and assign a posterior utility to the purchase which in turn informs future
decision making.
Coincident with the development of Nelson’s proposition that experience offers
an alternative to search in order to obtain more difficult information was recognition that
sellers can distort information and willfully create uncertainty (Akerlof, 1970). Buyers
may use market statistics (or reviews or brand information) to judge quality and, as
Akerlof suggests, this reliance on reputation is an incentive for sellers to market poorer
quality goods in order to increase profits. Darby and Karni (1973) conceptualized this as
“optimal fraud” and sought to explain the contributing forces. In order to do so, they
found it necessary to add a third of class of properties which they labeled as credence
qualities. Darby and Karni provide the following comparison:
We distinguish then three types of qualities associated with a particular purchase:
search qualities which are known before purchase, experience qualities which are
known costlessly only after purchase, and credence qualities which are expensive
to judge even after purchase. (p. 69)
By expensive, Darby and Karni mean that objective information is not readily
available even after purchase and use of the product or service. The buyer is left to rely
on the credibility or honesty of the seller in order to accept that he or she has received fair
value for the price paid. In some cases, it may be possible to determine whether or not the
product is as specified or the service was actually performed, but the cost to do so would
not be warranted. In other cases, where information about the good or service is solely
controlled by the seller, the buyer may never know with certainty the quality of the
product.

84

Marketers, particularly those whose function is to create marketing messages,
often trade in uncertainty. Advertising, public relations, sales promotions, and other
forms of marketing communications are designed to present product qualities in such a
way that they are valued favorably by the potential buyer in comparison to similar
products from competing sellers. In order to do this marketers focus on qualities for
which their products have a competitive advantage and they ignore or diminish
information about properties for which the seller’s product lacks advantage. For the buyer
attempting to make a fair comparison of product alternatives, the information available is
often a patchwork of competing claims about different properties that do not line up
neatly for evaluation. As Stigler (1961) indicated, it is already difficult for the buyer to
know what the price is for various alternatives at a given point in time. When the buyer
must also assess both knowable and unknowable qualities, the potential for uncertainty
and the need to rely on incomplete information is increased.
Both buyers and sellers, in general, realize that this is the case. Thus it is the
responsibility of the buyer, caveat emptor, to understand that marketing claims are
created with the intent to persuade. It is also incumbent upon sellers to promote their
products ethically. However as Akerlof (1970) and Darby and Karni (1973) theorize, the
greater the extent of uncertainty, the more likely the seller will have an incentive to cheat.
Stigler (1961) focuses on price, providing the obvious example of the kind of factual
information for which buyers can search; but there are other facts that buyers can know
before purchasing. Nelson (1970) uses the example of canned tuna to illustrate
experience. He suggests buyers can easily buy multiple brands of tuna in order to
determine preference, and at the low cost of experience there is insufficient demand for

85

establishments to sell tastes of various brands of tuna fish. Darby and Karni cite repair
service as having the credence quality, noting that in many cases consumers will be
unaware of the ability of the service to satisfy a given want and may only be able to
evaluate the outcome, not the procedure. As these examples illustrate, search, experience,
and credence product properties represent successively greater opportunities for
uncertainty.

References
Akerlof, G. A. (1970). The market for lemons: Quality uncertainty and the market
mechanism. Quarterly Journal of Economics, 90(4), 599-617.
Darby, M. R. & Karni, E. (1973). Free competition and optimal amount of fraud. Journal
of Law and Economics, 16, 67-88.
Nelson, P. (1970). Information and consumer behaviour. Journal of Political Economy,
78(2), 311-329.
Nelson, P. (1974). Adverting as information. Journal of Political Economy, 82(4), 729754.
Stigler, G. J. (1961). The economics of information. Journal of Political Economy, 49(3),
213-225.
[Paper ends.]

86

Appendix D: Ideation Stage 2 - Coding
Appendix D shows the second phase of the claim generation process. This
includes the explanation of the task to the student coders, a brief description of search,
experience, and credence product characteristics, and the list of items that were
categorically judged. The item order was randomized for each coder.
Task Explanation
[Explanation begins.]
Thank you for assisting with this project. By completing the task and entering
your name at the end, you will receive 2 points of extra credit in MGT 435. Your name
will only be used for awarding extra credit. Your individual answers are strictly
confidential and will be used as part of a statistical analysis of the task.
The task involves reading statements (questions) and assigning them to one of
three categories. This task is designed to help develop stimulus material for a research
study. The purpose of the task is to identify those questions that can be clearly assigned
to one category versus those that are difficult to put into a single category. Therefore, we
are looking for consistent assignment by multiple coders.
It is important that you read the Introduction, then read each question and make
your best guess about the category in which it belongs. Note that randomly assigning
questions to categories will not help complete the task. In addition, the statistical analysis
used to establish the reliability of the coding will also indicate which coders are making
random assignments.
There are 100 items to be coded into the three categories. Do not spend a lot of
time deliberating about individual items; work quickly and record your first impression.

87

The entire task should take approximately 15 minutes to complete.
[Explanation ends.]
Introduction and Coding Instructions
[Introduction and instructions begin.]
The following is a list of questions someone might ask when purchasing, or
considering the purchase of, various goods and services. The goal of this coding task does
not require you to answer the question. Rather, you should read the question and
determine how a consumer might obtain the answer to the question.
There are three main ways that consumers answer these kinds of questions:


The first approach is search. This means that the question can be answered
before purchasing the goods or service by searching for information or
shopping around.



The second approach is experience. This means that the question can be
answered, but typically only after purchasing and using the goods or services.
This does not mean that the answer could not be found before purchasing;
however, the cost (in time, money, or effort) would be so great that it is less
costly to buy the product and try it than to search for the answer before
purchasing.



The third approach is credence. This means that the question cannot be
answered with certainty before purchasing, and even after purchasing the
consumer will usually trust the credibility or honesty of the seller with regard
to knowing the answer to the question.

Read each question and for that question, indicate how you think a typical

88

consumer would obtain the answer. Is the answer to the question most likely to be
obtained by search, experience, or credence? If the wording of the question refers to a
"specific brand" or "this make and model" assume that the consumer would be asking the
question about one brand such as "Budweiser," "Nikon CoolPix," "Chevrolet," or
"Beaumont Hospital." If you think the answer might be obtained in more than one way,
indicate the one way you feel describes best how a consumer would get the answer to this
question.
[Introduction and instructions end.]
Item List
The following list contains statements categorized by coders in the second phase
of the ideation process. The 24 items selected as best representing product characteristics
in the category are identified as (S) search, (E) experience, and (C) credence. During the
claims generation several of these items were modified to fit with an existing product and
known qualities or substituted with another item having equal coder agreement.

1
2
3
4
5
6
7
8
9
10

What is the % alcohol content of a specific brand of bottled beer?
Is this woman's makeup product a liquid or a solid?
Is the fragrance of this perfume one that you like?
E
Is the fragrance of this perfume one that your friends will like on you?
How many sheets are on the role of this brand of paper toweling?
Does cotton underwear fabric feel comfortable?
E
What is the texture of the fabric used to make a specific brand of sports shirt?
What are the specifications for this make and model of digital camera?
Which make and model car has the styling you most prefer?
How many miles are on the used car you are considering?
S

11
12

Does the car you want to buy come with the features you want?
What is the square footage of the model home in a new subdivision
development?
How far is the apartment you are considering renting from the place where
you work?

13

89

14
15
16
17
18
19
20

Is the apartment you are considering close to where you work?
Is the available unit in that apartment building a rental unit or a condo for
sale?
What are the demographics of the neighborhood you are planning to move
to?
Are the neighbors in the subdivision where you are considering buying
friendly?
Will the works in the gallery show satisfy my aesthetic needs?
Will the works on display in the art gallery fit with other artwork I own?
What majors are offered by the college?

21
22
23
24
25
26
27
28
29
30

Does this airline offer flights to my destination?
Does that soft drink taste good?
Will drinking this brand of beer give me a hangover?
Is this skim milk or 2% milk?
Does your dog prefer wet food or dry food?
What does this ethnic food mix taste like?
Is this cosmetic hypo-allergenic?
Will this lipstick make me look more attractive to other people?
Will this brand of paper toweling absorb more water than other brands?
Will sheets of this brand of paper toweling last long enough for my clean up
jobs?

31
32
33
34
35
36
37
38
39

Will this fertilizer formula help me grow bigger tomatoes?
Does this medical device really work for my condition?
Does that bar and restaurant have the best Friday-night entertainment?
I fell on a broken sidewalk; will this attorney get me a big settlement?
How easy is it to use this model of digital camera?
How long should I expect the batteries to last in this model of video camera?
Will this cell phone camera work with my computer?
If my cable television goes out will the cable company fix it quickly?
Does your brand of cellular phone service provide better sound quality than
the other brands?
How long should I expect this car to last?

40
41
42
43
44
45
46
47
48
49

How safe is this SUV when compared to other SUVs?
Does this car have good handling?
Was this home built with craftsmanship?
Is that the real price or just the advertised price of that new refrigerator?
Is this university a good choice for my needs?
The credit counselor says I should stop paying my credit card; should I really
do that?
Does this university really have one of the top 10 programs in my major?
Does this university provide the atmosphere and lifestyle I want?
How does the quality of professors at my university compare to the quality at
similar universities?

90

S
S

S
S
E
E
E
C
E

C
S

50

How is the quality of service at that hair salon?

51
52
53
54
55
56
57

Should I believe the stories I hear on this television station's local newscast?
Are the reporters for this television network unbiased?
Are the cows that produced this milk organically fed?
Is this brand of dog food nutritional for my dog?
Is bottle of water really pure?
Does the water in this bottle come from a natural spring?
How much performance enhancement should I expect from this energy
drink?
How well do these vitamins work?
Is the vitamin content exactly as stated on the label?
Will this sunblock provide effective protection for my skin?

58
59
60
61

68
69
70
71
72
73
74
75
76
77
78
79
80

Is the celebrity endorsing this charity being paid for producing this
commercial?
How much protection does SPF 50 sunblock provide for my skin?
How were those organic paper napkins produced?
Does this fertilizer contain only organic material?
How effective is this drug for treating my illness?
What side effects does this drug have?
Will these new running shoes with the rocker soles strengthen my leg
muscles?
Will the design of these tennis shoes help me improve my tennis game?
Do these sunglasses block out UV rays?
Has this used car ever been in an accident?
Is this make and model likely to have mechanical problems?
Will this car keep me safe from injury in the event of an accident?
Does the contractor have the proper training to do this job?
How skilled is the contractor I am considering for this job?
I just had my jewelry appraised; are these fair replacement values?
Is this art deco painting authentic?
If I major in this subject will I be able to get a job in my field?
Should I believe it when the mechanic says I need a valve job?
Was this transmission repair done the way it was explained to me?
Will going through psychotherapy make me a happier person?

81
82
83
84
85
86
87
88
89

Doctor, should I have a C-section or a natural delivery?
Can this mortgage broker get me the lowest rate available?
The attorney says you should always sue in this situation; is that true?
Can the accountant assure me that I will pay the least amount of income tax?
Is this a legitimate tax deduction?
Will this insurance policy protect my home in case of a natural disaster?
Will this massage therapy give me relief from the pain in my back?
Is this tour really all-inclusive?
Will buying these gold coins help protect me against inflation?

62
63
64
65
66
67

91

E

C

C

S

C
C
C

C
E

90

Is the jeweler paying me a good premium for my old jewelry?

91

Will I save more on this phone if I pay a higher price but get a mail-in rebate
coupon?
92 Does this cell phone have the very latest technology?
93 The price of this DVD set is really inexpensive; is this a bootlegged copy?
94 What is the hourly rate the gym's personal trainers charge?
95 Does the bank have Saturday hours?
96 Will I be able to get my money out of these investments when I need it?
97 If I use this diet, will I be able to maintain my weight loss?
98 Can my professor really catch me if I cheat on this assignment?
99 How quickly will I get attention if I go to this hospital's emergency room?
100 Does the new, more efficient cereal package give me better value for my
money?

92

S

Appendix E: Example Creative Brief
Appendix E provides an example of the creative brief for this project:
SEC Category and Stimulus Number: Credence 35
Question: Can the accountant assure me that I will pay the least amount of income tax?
Brand/Product: Jackson Hewitt Tax Service, tax preparation
Background/Description: Jackson Hewitt is a tax preparation firm; they employ many
preparers who are not accredited or certified public accounts. They compete with H&R
Block as well as independent CPA firms. Aside from some scattered comments by
customers on online bulletin boards, reviews about the Jackson Hewitt tax preparation
chain are almost uniformly negative. In 2007, the U.S. Dept. of Justice field lawsuits
claiming fraud at 125 Jackson Hewitt locations.
Branded Truthful Claim (3511): At Jackson Hewitt, preparing tax returns is what we
do. And that could mean more money in your pocket.
Branded False Claim (3501): At Jackson Hewitt we know the ins and outs of tax
preparation, so you can feel confident we'll find all the deductions and credits you're
entitled to. [from the Jackson Hewitt website; the claim is contrary to consumer opinion
and federal litigation]
Unbranded Truthful Claim (3510): Preparing tax returns is what we do. And that
could mean more money in your pocket.
Unbranded False Claim (3500): We know the ins and outs of tax preparation, so you
can feel confident we'll find all the deductions and credits you're entitled to.
Source(s) and/or Support:
http://www.bls.gov/oes/current/oes132082.htm
http://www.jacksonhewitt.com/Prepare-Your-Taxes/In-Office-Products-Services/InOffice-Products-and-Services/
http://www.consumersearch.com/tax-preparation-services/jackson-hewitt
http://www.consumeraffairs.com/finance/jackson_hewitt.html

93

REFERENCES

94

REFERENCES

Aaker D. A. (1991). Managing brand equity: Capitalizing on the value of a brand name.
New York: The Free Press.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting
interactions. Newbury Park, CA: Sage.
Akerlof, G. A. (1970). The market for lemons: Quality uncertainty and the market
mechanism. Quarterly Journal of Economics, 90(4), 599-617.
Argyle, M. (1972). Non-verbal communication in human social interaction. In R. Hinde
(Ed.), Non-verbal communication (pp. 243-269). Cambridge, England: Cambridge
University Press.
Anderson, D. E., Ansfield, M. E., & DePaulo, B. M. (1997). Love’s best habit: Deception
in the context of relationships. In P. Philippot, R. S. Feldman, & E. J. Coats
(Eds.), The social context of nonverbal behavior. Cambridge: Cambridge
University Press.
Berger, C. R., & Calabrese, R. J. (1975). Some explorations in the initial interaction and
beyond: Toward a developmental theory of interpersonal communication. Human
Communication Research, 1, 99-112.
Berger, C. R., & Gudykunst, W. B. (1991). Uncertainty and communication. In B. Dervin
& M. Voight (Eds.), Progress in communication sciences (pp. 21–66). Norwood,
NJ: Ablex.
Blair, J. P., Levine, T. R., & Shaw, A. J. (2010). Content in context improves deception
detection accuracy. Human Communication Research, 36, 420-439.
Blakeman, R. (2007). Integrated marketing communication: Creative strategy from idea
to implementation. Lanham, MD: Rowman & Littlefield Publishers.
Bok, S. (1999). Lying: Moral choice in public and private life. New York: Vintage.
Bond, C. F., Jr., & DePaulo, B. M. (2006). Accuracy of deception judgments. Review of
Personality and Social Psychology, 10, 214-234.
Boush, D. M., Friestad, M., & Wright, P. (2009). Deception in the marketplace: The
psychology of deceptive persuasion and consumer self-protection. New York:
Routledge.

95

Buller, D. B., & Burgoon, J. K. (1996). Interpersonal deception theory. Communication
Theory, 6, 203–242.
Burgoon, J. K., & Buller, D. B. (2008). Interpersonal deception theory. In L. A. Baxter &
D. O. Braithwaite (Eds.), Engaging theories in interpersonal communication:
Multiple perspectives (pp. 227–239). Los Angeles, CA: Sage.
Burgoon, J. K., & Levine, T. R. (2009). Advances in deception detection. In S. W. Smith
& S. R. Wilson, New directions in interpersonal communication research (pp.
223-242). Thousand Oaks, CA: Sage.
Burke, R. R., DeSarbo, W. S., Oliver, R. L., & Robertson, T. S. (1988). Deception by
implication: An experimental investigation. Journal of Consumer Behavior, 14,
483-494.
Campbell, M. C., & Kirmani, A. (2008). I know what you’re doing and why you’re doing
it: The use of the persuasion knowledge model in consumer research. In C.
Haugtvedt, P. M. Herr, & F. R. Kardes (Eds.), The handbook of consumer
psychology (549-573). Mahwah, NJ: Erlbaum.
Cooke, E. F. (1997). Greed leads to better marketing, healthier competition, economy.
Marketing News 31(23), 4.
Cooke, E. F. (2004). The crossroads between morality, greed, and competition. In L.E.
Pelton & S. L. True (Eds.), Business ethics: Perspectives on corporate
responsibility (6th ed., pp. 51-56). Boston: Houghton Mifflin.
Darby, M. R., & Karni, E. (1973). Free competition and optimal amount of fraud. Journal
of Law and Economics, 16, 67-88.
Day, G. S. (1984). Strategic market planning: The pursuit of competitive advantage. St.
Paul, MN: West Publishing.
DePaulo, B. M., Ansfield, M. E., & Bell, K. L. (2006). Theories about deception and
paradigms for studying it: A critical appraisal of Buller and Burgoon's
interpersonal deception theory and research. Communication Theory, 6, 297-310.
Duncan, T. (2005). Principals of advertising and IMC (2nd ed.). Boston: McGrawHill/Irwin.
Ekelund, R. B., Mixon, F. G., & Ressler, R. W. (1995). Advertising and information: An
empirical study of search, experience, and credence goods. Journal of Economic
Studies, 22(2), 33-43.
Ekman, P. & Friesen, W. V. (1969a). Detecting deception from the body or face. Journal
of Personality and Social Psychology, 29(3), 288-298.

96

Ekman, P., & Friesen, W. V. (1969b). Nonverbal leakage and clues to deception.
Psychiatry, 32, 88-106.
Erdem, T., & Swait, J. (1998). Brand equity as a signaling phenomenon. Journal of
Consumer Psychology, 7(2), 131-157.
Fein, S., & Hilton, J. L. (1994). Judging others in the shadow of suspicion. Motivation
and Emotion, 18(2), 167-198.
Fleiss, J. L. (1971). Measuring nominal scale agreement among many raters.
Psychological Bulletin, 76(5), 378-382.
Gardner, D. M. (1975). Deception in advertising: A conceptual approach. Journal of
Marketing, 39, 40-46.
Gilbert, D. T. (1991). How mental systems believe. American Psychologist, 46, 107–119.
Gilbert, D. T., Krull, D. S., & Malone, P. S. (1990). Unbelieving the unbelievable: Some
problems in the rejection of false information. Journal of Personality and Social
Psychology, 59, 601-613.
Goffman, E. (1963). Behavior in public places: Notes on the social organization of
gatherings. Glencoe, IL: The Free Press.
Healy, P. M. & Palepu, K. G. (2003). The fall of Enron. Journal of Economic
Perspectives, 17(2), 3-26.
Jacoby, J., & Hoyer, W. D. (1987). The comprehension and miscomprehension of print
communications: An investigation of mass media magazines. New York:
Advertising Educational Foundation.
Kardes, F. R., Posavac, S. S., & Cronley, M. L. (2004). Consumer inference: A review of
processes, bases, and judgment contexts. Journal of Consumer Psychology, 14(3),
230-256.
Keller, K. L. (1993). Conceptualizing, measuring, and managing customer-based brand
equity. Journal of Marketing, 57(1), 1-22.
Kirmani, A. & Wright, P. (1989). Money talks: Perceived advertising expense and
expected product quality. Journal of Consumer Research, 16, 344-353.
Kish, L. (1965). Survey Sampling. New York: John Wiley & Sons.
Knapp, M. L. (2008). Lying and deception in human interaction. Boston: Pearson Allyn
and Bacon.

97

Laband, D. N. (1986). Advertising as information: An empirical note. The Review of
Economics and Statistics, 68(3), 517-521.
Levine, T. R. (2009). Truth bias theory: A new theory of deception. Unpublished
manuscript, Michigan State University.
Levine, T. R., Kim, R. K., & Blair, J. P. (2010). (In)accuracy at detecting true and false
confessions and denials: An initial test of a projected motive model of veracity
judgments. Human Communication Research, 36, 82-110.
Levine, T. R., Kim, R. K., Park, H. S., & Hughes, M. (2006). Deception detection
accuracy is a predictable linear function of message veracity base-rate: A formal
test of Park and Levine’s probability model. Communication Monographs, 73,
243-260.
Levine, T. R., Park, H. S., & McCornack, S. A. (1999). Accuracy in detecting truths and
lies: Documenting the “veracity effect.” Communication Monographs, 66, 125144.
Levine, T. R., Serota, K. B., Shulman, H., Clare, D., Park, H. S., Shaw, A. S., Shim, J. C.,
& Lee, J. H. (in press). Sender demeanor: Individual differences in sender
believability have a powerful impact on deception detection judgments. Human
Communication Research.
Lewicki, R. J. (1983). Lying and deception: A behavioral model. In M. H. Bazerman &
R. J. Lewicki (Eds.), Negotiating in organizations (pp. 68–90). Beverly Hills, CA:
Sage.
Lewicki, R. J., & Robinson, R. (1998). Ethical and unethical bargaining tactics: An
empirical study. Journal of Business Ethics, 17, 665–682.
Mahewswaran, D., Mackie, D. M., & Chaiken, S. (1992). Brand name as a heuristic cue:
The effect of task importance and expectancy confirmation on consumer
judgments. Journal of Consumer Psychology, 1, 317-336.
McCornack, S. A., & Parks, M. R. (1986). Deception detection and relationship
development: The other side of trust. In M. L. McLaughlin (Ed.), Communication
yearbook 9 (pp. 377-389). Beverly Hills, CA: Sage.
Murphy, P. E. (1999). Character and virtue ethics in international marketing: An agenda
for managers, researchers and educators. Journal of Business Ethics, 18, 107-124.
Murphy, P. E. (2002). Marketing ethics at the millennium: Review, reflections, and
recommendations. In N. E. Bowie (Ed.), The Blackwell guide to business ethics
(pp. 165-185). Oxford, England: Blackwell.

98

Nelson, P. (1970). Information and consumer behaviour. Journal of Political Economy,
78(2), 311-329.
Nelson, P. (1974). Adverting as information. Journal of Political Economy, 82(4), 729754.
Olson, J. C. (1977). Price as an informational cue: Effects on product evaluations. In A.
G. Woodside, J. N. Sheth, & P. D. Bennett (Eds.), Consumer and Industrial
Buying Behavior (pp 267-286). New York: North Holland Publishing.
Park, H. S., & Levine, T. R. (2001). A probability model of accuracy in deception
detection experiments. Communication Monographs, 68, 201-210.
Park, H. S., Levine, T. R., McCornack, S. A., Morrison, K., & Ferrara, M. (2002). How
people really detect lies. Communication Monographs, 69, 144–157.
Parasuraman A. (1997). Reflections on gaining competitive advantage through customer
value. Journal of Academy of Marketing Science 25 (2): 154–161.
Porter, M. E. (1985). Competitive Advantage. New York: Free Press.
Preston, I. L. (1977). The FTC’s handling of puffery and other selling claims made “by
implication”. Journal of Business Research, 5,155-181.
Rao A. R. & Monroe, K. B. (1989). The effect of price, brand name, and store name on
buyers’ perceptions of product quality: An integrative review. Journal of
Marketing Research, 26, 351-357.
Robin, D. P. & Reidenbach, R. E. (1987). Social responsibility, ethics, and marketing
strategy: Closing the gap between concept and application. Journal of Marketing,
51, 44-58.
Russo, J. Edward (1976). When do advertisements mislead the consumer: An answer
from experimental psychology. Advances in Consumer Research, 273-275.
Russo, J. E., Metcalf, B. L., & Stevens, D. (1981). Indentifying misleading advertising.
Journal of Consumer Research, 8(2), 119-131.
Schweitzer, M. E., Ordóñez, L., & Douma, B. (2004). Goal setting as a motivator of
unethical behavior. Acadamy of Management Journal. 47, 422-432.
Serota, K.B., Levine, T. R., & Boster, F. J. (2010). The prevalence of lying in America:
Three studies of self-reported lies. Human Communication Research, 36, 2-25.

99

Serota, K. B., & Bhargava, M. (2010). Brand-usage networks: Understanding brand
architecture from a consumer perspective. Journal of Strategic Marketing 18(6),
435-455.
Sethi, S. P. (1994). Multinational corporations and the impact of public advocacy on
corporate strategy: Nestle and the infant formula controversy. Dordrecht,
Netherlands: Kluwer Academic.
Shostack, L. (1977). Breaking free from product marketing. Journal of Marketing, 41(2),
73-80.
Spence, M. (1973). Job marketing signaling. The Quarterly Journal of Economics, 87(3),
355-374.
Srivastava, R. K., Alpert, M. I., & Shocker, A. D. (1984). A customer-oriented approach
for determining market structures. Journal of Marketing, 48(Spring), 32-45.
Stigler, G. J. (1961). The economics of information. Journal of Political Economy, 49(3),
213-225.
Woodruff, R. B. (1997). Customer value: The next source for competitive advantage.
Journal of Academy of Marketing Science, 25(2), 139-153.
Zeithaml, V. A. (1981). How consumer evaluation processes differ between goods and
services. In J. H. Donnelly & W. R. George (Eds.), Proceedings of the AMA
Marketing of Services Conference, Chicago, 186-190.
Zeithaml, V. A., Parasuraman, A., & Berry, L. L. (1985). Problems and strategies in
services marketing. Journal of Marketing, 49, 33-46.
Zuckerman, M., DePaulo, B. M., & Rosenthal, R. (1981). Verbal and nonverbal
communication of deception. In L. Berkowitz (Ed.), Advances in experimental
social psychology (Vol. 14, pp. 1-59). New York: Academic Press.

100