Uploaded by ROHIL UTEKAR

Integrating Advertising and Publicity

advertisement
Journal of Advertising
ISSN: 0091-3367 (Print) 1557-7805 (Online) Journal homepage: https://www.tandfonline.com/loi/ujoa20
Integrating Advertising and Publicity
Jooyoung Kim , Hye Jin Yoon & Sun Young Lee
To cite this article: Jooyoung Kim , Hye Jin Yoon & Sun Young Lee (2010) Integrating Advertising
and Publicity, Journal of Advertising, 39:1, 97-114, DOI: 10.2753/JOA0091-3367390107
To link to this article: https://doi.org/10.2753/JOA0091-3367390107
Published online: 04 Mar 2013.
Submit your article to this journal
Article views: 811
View related articles
Citing articles: 8 View citing articles
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=ujoa20
Integrating Advertising and Publicity
A Theoretical Examination of the Effects of Exposure Sequence,
Publicity Valence, and Product Attribute Consistency
Jooyoung Kim, Hye Jin Yoon, and Sun Young Lee
ABSTRACT: Using several theories, including Information Integration Theory (IIT), Integrated Information Response
Model (IIRM), confirmation effect, and contrast theory, this study investigates the combined effects of advertising and
publicity under the varied conditions of exposure sequence, publicity valence, and product attribute consistency. Results
obtained through a 2 (sequence: ad then publicity versus publicity then ad) × 2 (attribute variation: same attribute versus
different attribute) × 2 (valence of publicity: positive versus negative) within-subjects factorial design show that advertising coupled with positive publicity induces confirmation effects regardless of sequence and attribute consistency, whereas
negative publicity combined with advertising mostly produces contrast effects, with mixed results depending on the
sequence and attribute consistency. Findings and implications are discussed.
Wide attention has been given to integrated marketing communication (IMC) in the past decade, by both academia and
industry, and the concept of IMC has found its way into the
mainstream of marketing practice and literature (Duncan and
Caywood 1996; Durkin and Lawlor 2001). Even though more
and more companies are taking an IMC approach to corporate
communication planning, relatively little empirical investigation has addressed its application (Eagle, Kitchen, and Bulmer
2007). Responding to this need, this study investigates the
integration effect of advertising and publicity, two of the
most widely accessible communication forms to consumers
(Balasubramanian 1994). For researchers and practitioners,
some guidelines are needed regarding how those two communication tools can be combined effectively. For example,
how would consumers, after being exposed to an advertisement
about the printing speed of a printer brand PRINTWELL
react to the brand if they read a negative review of the brand’s
printing speed? Would consumers react differently if the
exposure sequence were reversed? The number of potential
cases of advertising and publicity combinations extends far
beyond these examples. Simply put, given an advertisement
for a product focusing on attribute X, there are at least four
Jooyoung Kim (Ph.D., University of Florida) is an assistant professor of advertising, Department of Advertising and Public Relations,
Grady College of Journalism and Mass Communication, University
of Georgia.
Hye Jin Yoon (M.A., University of Georgia) is a doctoral student,
Department of Advertising and Public Relations, Grady College of
Journalism and Mass Communication, University of Georgia.
Sun Young Lee (M.A., University of Georgia) is a doctoral student,
School of Journalism and Mass Communication, University of North
Carolina.
possible combinations: ad and a positive publicity about the
same attribute, ad and a negative publicity about the same
attribute, ad and a positive publicity about another attribute,
and ad and a negative publicity about another attribute. Because advertising and publicity can be exposed to consumers
in many different ways, it is crucial to investigate the various
combinations.
Though rare, some studies provide research findings about
the integration effect of advertising and publicity. Loda and
Coleman (2005) examined whether there is a difference between
dependent variables such as message acceptance and message
response depending on the medium (advertising or publicity)
and on the sequencing of message delivery (advertising then
publicity or publicity then advertising). However, they only
considered the situations when the messages in advertising
and publicity are consistent and positive. Stammerjohan et al.
(2005) investigated the interactions between different types
of advertising media (radio and print) and the interaction
between brand familiarity and message valence in publicity
(positive or negative), but their study does not provide insight
into how the combinations of advertising and various types of
publicity (in terms of valence and attribute consistency) might
work to form particular attitudes for each case. In addition,
a few studies have focused on the joint effects of advertising
and consumer experiences other than publicity (e.g., ad and
trial [Smith 1993]; ad and negative word-of-mouth [Smith
and Vogt 1995]; ad and product sampling [Marks and Kamins
1988]).
Although there are some valuable findings in the literature, extant literature does not provide adequate empirical
guidance or acceptable theoretical frameworks for marketers
and researchers. Our study, therefore, attempts to investigate,
both empirically and theoretically, the combination effects of
Journal of Advertising, vol. 39, no. 1 (Spring 2010), pp. 97–113.
© 2010 American Academy of Advertising. All rights reserved.
ISSN 0091-3367 / 2010 $9.50 + 0.00.
DOI 10.2753/JOA0091-3367390107
98
The Journal of Advertising
advertising and publicity, focusing on three important conditional variables that are believed to serve as a simple but
comprehensive set of conditions for advertising–publicity
combinations: (1) exposure sequence, (2) publicity valence,
and (3) product attribute consistency.
Theoretical Framework
and Hypothesis
The literature shows that the virtue of integrating advertising
and publicity is the synergistic effect generated between them.
The synergy effect, which is the representative value of IMC,
refers to the benefits of multiple communication tools (i.e.,
both promotional tools and media tools) delivering a unified
message (Moriarty 1994; Naik and Raman 2003; Schultz
1996; Schultz 1996; Schultz and Kitchen 1997). In the same
vein, there are times when consumers are exposed to multiple
communication tools delivering ununified messages, or even
worse, antagonistic messages, generating a more negative
impact than being exposed to a single negative message. This
type of negative effect is called the counter-synergy effect (e.g.,
Baker 2002).
With this synergy effect of integration in mind, we developed our theoretical framework and hypotheses. The
theoretical framework for understanding the collective effect
of advertising and publicity can be developed by synthesizing some competing or supplementing theories found in the
literature. The theories we used in the present study include
Information Integration Theory (IIT) (Anderson 1971), Integrated Information Response Model (IIRM) (Smith and
Swinyard 1982), contrast theory (Hovland, Harvey, and Sherif
1957), and confirmation effect (LaBella and Koehler 2004).
Following a short review, each theory will be applied to several advertising publicity combinations. Hypotheses will be
formulated based on the predictive directions in favor of the
synergy concept of IMC or the counter-synergy concept against
which IMC exists.
Information Integration Theory
Anderson (1971) developed the Information Integration Theory
to explain how a person’s attitude is formed from integrating
different pieces of information. The theory assumes that attitudes are formed and modified as people interpret new information and integrate it with their prior attitudes. It involves two
basic operations: valuation and integration. Valuation refers to
the determination of the meaning, importance (i.e., weight),
and evaluation of the information. Integration is a process of
combining those valued pieces of information. The integration
process of advertising and publicity can be described in terms of
a simple algebraic model, such as R = wad*sad + wpub*spub , where
R = product evaluation after exposure to advertising and pub-
licity, wad = weight given to the ad information, sad = evaluative
position of message in the ad, wpub = weight given to the publicity information, and spub = evaluative position of information
in the publicity. There are two competing models for the rules
of integration: adding model and averaging model. The adding model assumes that people add each piece of information,
which results in the increase of R when the information pieces
are of the same sign. Therefore, each weight can range from
0 to 1. On the other hand, the averaging model assumes that
people integrate each piece of information by averaging it with
the prior set of integrated information. Therefore, R does not
necessarily become larger, and each new weight is dependent
on the weight of the information already integrated in the
model. In the averaging model, the weights of the algebraic
model above are interdependent, and the weights must sum
to one (e.g., wad = 1 − wpub) (Anderson 1971). In addition, the
weight (w) of each piece of information depends on its credibility and reliability, leading to the “weighted averaging”
(Eagly and Chaiken 1993, p. 247) procedure. Accordingly, a
consumer’s attitude toward the brand (Ab) after being exposed
to both advertising and publicity would be the average of “Ab
based on ad” and “Ab based on publicity,” where the weights
of ad and publicity can be adjusted based on their credibility
and reliability. The integration theory also provides a way of
understanding the sequential effect of informational stimuli
using the process of “inconsistency discounting” (Anderson
1971, p. 173) through which a new piece of information that
is inconsistent with the existing set of information is given
a decreased weight. Therefore, it is expected that consumers
will discount the weight of advertising or publicity if it is inconsistent (in terms of valence) with what they were formerly
exposed to.
Integrated Information Response Model
The Integrated Information Response Model is similar to
the averaging model of IIT, but it specifically focuses on the
combination of advertising and direct product experience.
This model, developed by Smith and Swinyard (1982), using Fishbein and Ajzen’s (1975) expectancy-value theory
(EV = Σ[bici ]ei , where EV = expectancy value, bi = belief
confidence for attribute i, ci = belief strength for attribute i,
ei = evaluation of attribute i), incorporates the effects of consumer cognition, affect, and behavior. The model suggests that
consumers often have a low degree of information acceptance
for advertising, for it is generally known to be a vested-interest
source. Therefore, consumers often form weakly (belief confidence: ci ) held attitudes (EV) based on the lower-order beliefs
(belief strength: bi ) created by the advertising exposure. Thus,
consumer cognition will be based on lower-order beliefs that
will result in a lower-order affect. Conversely, consumers who
have direct product experience will put more trust in their
Spring 2010 99
experience and form stronger beliefs that are higher-order
beliefs (i.e., expectancy is closer to 1), which result in higherorder affect (Smith and Swinyard 1983, 1988).
Because publicity is generally perceived as more credible
than advertising (e.g., Loda and Coleman 2005), publicity
might also result in a higher-order belief compared to advertising. When present, publicity, due to its higher-order
belief, will be a greater influence on overall evaluation than
advertising. Therefore, IIRM’s overall prediction would be that
the combined effect of advertising and publicity (regardless
of sequence) will be similar (or equal) to the effect of publicity alone.
Confirmation Effect
Although IIT’s weighted averaging process predicts that less
credible information is given little to no weight and IIRM
theorizes that a lower-order belief combined with a higherorder belief will have nearly zero weight (bi and ci ), some
empirical studies have shown that probability judgments can
still be influenced by nondiagnostic evidence (e.g., Nisbett,
Zukier, and Lemley 1981; Zukier and Jennings 1983).
For example, LaBella and Koehler (2004) found the so-called
confirmation effect, where the combination of nondiagnostic
evidence with diagnostic evidence increased the overall evaluation. That is, initial judgments can become more extreme
with the introduction of either diagnostic or nondiagnostic
evidence, when the evidence matches the person’s expectation,
because “people are more sensitive to, or give greater weight
to, evidence that confirms their currently held beliefs” (LaBella
and Koehler 2004, p. 1083). For example, when a consumer
tries a product and finds that it works as promised in an ad
he or she has watched before, IIT and the IIRM will predict
that the previous advertising effect will mostly disappear.
Thus, the product evaluation based on an ad-trial condition
and a trial-only condition should show no difference based
on the IIT or the IIRM. Conversely, the confirmation effect
theory will predict that an ad-trial condition (when ad and
trial confirm each other) will show a higher evaluation than a
trial-only condition because of the confirmation effect. As such,
when ad is combined with publicity, IIT and IIRM predict
that the advertising effect will mostly disappear, showing no
differences from the publicity-only condition. Conversely,
the confirmation theory predicts that the combined effect of
ad and publicity will be stronger than either the ad-only or
publicity-only condition due to confirmation effects.
Contrast, Assimilation, and Assimilation-Contrast
Effect
Compared to confirmation effect theory, which focuses on the
positive effect of an expected outcome between two experi-
ences, contrast effect theory explains the surprise effect of an
unexpected outcome (Oliver 1981). When expectations are not
matched (either positively or negatively) by an actual product
experience, contrast theory predicts that the contrast between
expectations and actual experience will cause the consumer to
inflate the discrepancy (Anderson 1973). Although there are
other related theories, such as assimilation theory or a hybrid
model called assimilation-contrast theory (see Eagly and Chaiken [1993] for a more detailed review), when advertising and
publicity information (in terms of valence) is inconsistent, and
thus unexpected, only the contrast effect theory is applicable.
When consumers are exposed to either advertising or publicity
that delivers inconsistent information (i.e., “negative publicity
after ad” or “ad after negative publicity”), the contrast theory
predicts that the combined effect will be much more negative
than the publicity-only condition. In the same situation, IIT
predicts that the combined condition and the publicity-only
condition will result in a similar effect because, for the combined condition, the weight of advertising will be nearly zero
(weighted averaging) or because any later inconsistent communication will be discounted (inconsistency discounting).
Hypotheses for Advertising and Publicity Integration
Theoretically, how might advertising and publicity work together? The traditional two-step model (e.g., Deighton 1984)
of advertising explains that in the first step, an exposure to
advertising induces the consumer to form a hypothesis about
the product. Because advertising is viewed as a vested-interest
source, the hypothesis should be tentative and beliefs might
be weakly held. In the second step, if the consumer encounters
any evidence, he or she will test the ad-based belief in order
to confirm it and increase confidence. The model suggests
that the effects of each step can interact so that the initial
expectation can change in any direction after the second step
(Deighton 1984). The strength of influence that the evidence
has in the second step depends on the credibility of the content and source (e.g., Birnbaum and Stegner 1979; Eagly and
Chaiken 1993).
Generally, publicity is thought to be more credible and
more influential than company-controlled forms of communications (Bond and Kirshenbaum 1998). Therefore, the power
of publicity to influence purchase decisions over other forms of
promotion would not be surprising (Chaiken and Maheswaran
1994; Sternthal, Dholakia, and Leavitt 1978). To lay the
foundation for our theoretical framework and subsequent
hypotheses, the following hypothesis was set forth first:
H1: Believability of publicity will be significantly higher than
that of advertising.
How do consumers integrate the information from advertising and publicity? To answer this question, we characterize
100
The Journal of Advertising
potential cases of ad (Ad) and publicity (Pub) combinations.
Given an ad for a product focusing on an attribute X (i.e.,
Adx ), there are at least four possible cases of publicity: a
positive publicity statement about the same attribute X (i.e.,
+Pubx ), a negative publicity statement about the same attribute X (i.e., –Pubx ), a positive publicity statement about
another important attribute Y (i.e., +Puby ), and a negative
publicity statement about another important attribute Y
(i.e., –Puby ). These four different cases of publicity give
us four Ad-Pub combinations: (1) Adx & +Pubx, (2) Adx &
–Pubx, (3) Adx & +Puby, and (4) Adx & –Puby. In addition,
the sequence of exposure needs consideration: Ad then Pub
or Pub then Ad. Crossing all possible Ad-Pub cases with
two possible sequences yield eight (4 × 2) combinations. In
each case, the integration procedure that combines the pieces
of information from both advertising and publicity will be
different. For each case, we discuss and compare the relevant
theories reviewed previously and formulate a hypothesis that
is in favor of the synergy or counter-synergy effect, for or
against which IMC exists. Hypothesized predictions and their
competing or alternative predictions with respective theories
are shown in Table 1.
Adx & +Pubx
This case combines an advertisement and a positive publicity
statement that focus on the same attribute. IIRM predicts
that publicity, due to higher credibility, will be more heavily
weighted than the ad. The weighted averaging principle of IIT
also predicts that the weight of advertising will be adjusted
to accommodate the greater publicity weight. Furthermore,
there should be no difference in total expectancy between the
Ad/Pub or Pub/Ad sequence, for IIT does not predict the discounting effect for the presentation of consistent information.
Conversely, alternative results can be predicted by confirmation effect theory when the information in advertising and
publicity is consistent.
All in all, IIT and IIRM theory predict that publicity will
dominate the advertising effect and make it negligible, but
the confirmation effect theory proposes an increase in evaluation that reflects a synergistic effect. In agreement with the
confirmation effect theory predicting increased evaluation, the
following hypotheses were proposed:
H2a: Attitude toward the brand under the combined (Adx
& +Pubx) communication condition in the Adx then +Pubx
sequence will be significantly higher than that of the ad-only
or publicity-only condition.
H2b: Attitude toward the brand under the combined (Adx
& +Pubx) communication condition in the +Pubx then Adx
sequence will be significantly higher than that of the ad-only
or publicity-only condition.
Adx & +Puby
When information between advertising and publicity is inconsistent in attribute (i.e., attribute X and Y) but consistent
in valence (i.e., positive), IIT’s weighted averaging model
suggests that +Puby might still gain higher weight but that
Adx should not lose too much weight due to its unique piece
of information (i.e., X). In this case, IIT’s adding model can
also explain the additive effect of two pieces of information
conveying different attributes. In a similar vein, within the
IIRM principle, the consumer will simultaneously possess two
separate stages of beliefs about the brand based on two pieces of
information (advertising about attribute X and publicity about
attribute Y), instead of exclusively choosing one information
source that provides a higher-order belief. In addition, there
should be no inconsistency discounting (because the attributes
are not opposite in evaluation). Will the confirmation effect
still be present even when the attributes are different? Previous
studies suggest that the confirmation effect might still occur
based on the affirmed valence, despite the lack of attribute
confirmation. For example, in their interpersonal performance
appraisal study, Robbins and DeNisi (1998) found that affect
consistency influenced the weighting of recalled information
and subsequent ratings.
For this case of Adx & +Puby, accordingly, IIT and IIRM’s
prediction would be that publicity will gain more weight but
the advertising will also retain its own weight. In addition,
the confirmation effect appears to be present because of the
congruent positive valence. IIT’s and IIRM’s predictions are
much closer to additive, while confirmation effect theory’s prediction is more synergistic. However, all three theories predict
that the attitude from the Adx & +Puby situation will be more
favorable than either the Adx-only or +Puby-only situation.
H3a: Attitude toward the brand under the combined (Adx
& +Puby ) communication condition in the Adx then +Puby
sequence will be significantly higher than that of the ad-only
or publicity-only condition.
H3b: Attitude toward the brand under the combined (Adx
& +Puby ) communication condition in the +Puby then Adx
sequence will be significantly higher than that of the ad-only
or publicity-only condition.
Adx & –Pubx
This case is widely prevalent in the marketplace and generally works against the advertiser (Ahluwalia, Burnkrant, and
Unnava 2000). IIRM does not predict much in this case, for
it does not have a theoretical property to explain how two
conflicting pieces of information can be integrated. But IIT’s
weighted averaging principle proposes that the weight of advertising will be adjusted to accommodate the greater weight
Spring 2010 101
Table 1
Hypothesized and Competing Predictions
Hypotheses and competing predictions
Theories
Prediction bases
H2a
A(Adx then +Pubx ) > A(Adx ), > A(+Pubx )
Confirmation effect
Confirmation (a, v)
Competing predictions:
> A(Adx ), ≤ A(+Pubx ) IIT Averaging
> A(Adx ), = A(+Pubx ) IIRM Expectancy weight of ad = 0
Confirmation effect
Confirmation (a, v)
H2b
A(+Pubx then Adx ) > A(Adx ), > A(+Pubx )
Competing predictions:
> A(Adx ), ≤ A(+Pubx ) IIT Averaging
> A(Adx ) = A(+Pubx ) IIRM Expectancy weight of ad = 0
Confirmation effect
Confirmation (v)
H3a
A(Adx then +Puby ) > A(Adx ), > A(+Puby )
IIT Adding
IIRM Expectancy weight of ad > 0
H3b
A(+Puby then Adx ) > A(Adx ), > A(+Puby )
Confirmation effect
Confirmation (v)
IIT Adding
IIRM Expectancy weight of ad > 0
H4a
A(Adx then –Pubx ) < A(–Pubx )
Contrast effect
Contrast (a, v)
Competing predictions:
≥ A(–Pubx ) IIT Weighted averaging
> A(–Pubx ) IIT Inconsistency discounting
Contrast effect
Contrast (a, v)
H4b
A(–Pubx then Adx) < A(–Pubx )
Competing predictions:
≥ A(–Pubx ) IIT Weighted averaging
≥ A(–Pubx) IIT Inconsistency discounting
Contrast effect
Contrast (v)
H5a
A(Adx then –Puby ) < A(–Puby )
Competing predictions:
> A(–Puby ) IIT Weighted averaging
> A(–Puby ) IIT Inconsistency discounting
Contrast effect
Contrast (v)
H5b
A(–Puby then Adx ) < A(–Puby )
Competing predictions:
> A(–Puby ) IIT Weighted averaging
> A(–Puby ) IIT Inconsistency discounting
Notes: A = attitude; a = attribute; v = valence; IIT = Information Integration Theory; IIRM = Integrated Information Response Model.
Items in bold represent the hypotheses and their corresponding theories.
of publicity, resulting in a nearly zero weight for advertising.
In addition, IIT’s inconsistency discounting process predicts
that the negative publicity will get a discounted weight
for the Adx then –Pubx sequence because the information
is inconsistent with the attitude formed by the original ad.
Inconsistency discounting can also happen in the –Pubx then
Adx sequence, for the ad information will be inconsistent with
the evaluative belief formed from the negative publicity. In
addition, this inconsistency discounting process suggests that
the attitude from the Adx then –Pubx sequence will be higher
than that from the –Pubx then Adx sequence. For a simple
example, if attitudes based on Adx only and –Pubx only are
5 and 2 on a 1–7 scale, respectively, and the discount rate is
50%, the attitude from the Adx then –Pubx sequence will be
(5 + 1)/2 = 3 and that from the –Pubx then Adx sequence will
be (2 + 2.5)/2 = 2.25.
Contrary to the predictions of IIT, different results are
predicted by the contrast theory. For the Adx then –Pubx sequence, because expectation based on Adx is not matched by
a more credible piece of information (–Pubx ), contrast theory
predicts that the surprising contrast between two pieces of
information will cause the consumer to magnify the discrepancy (Anderson 1973), resulting in a counter-synergy effect.
Though smaller in magnitude, the contrast effect can also be
present for the –Pubx then Adx sequence as the expectation
based on –Pubx is not matched by a new piece of information
(Adx ). The following hypotheses were formulated based on the
contrast effect theory to predict the counter-synergy effects for
a misintegrated combination of Adx & –Pubx. Since the ad-only
condition provides only positive information, we would also
expect this combined condition to be lower, but we are interested in a combined effect compared to the negative publicity
alone, not compared to positive advertising alone.
H4a: Attitude toward the brand under the combined (Adx
& –Pubx ) communication condition in the Adx then –Pubx
sequence will be significantly lower than that of the publicityonly condition.
H4b: Attitude toward the brand under the combined (Adx
& –Pubx ) communication condition in the –Pubx then Adx
sequence will be significantly lower than that of the publicityonly condition.
102
The Journal of Advertising
Adx & –Puby
This case is an unfortunate one for marketers, but it might
be less severe than Adx & –Pubx because of attribute inconsistency. When the information of advertising and publicity
is inconsistent in both attribute and valence, IIT’s weighted
averaging model suggests that –Puby might still gain more
weight (due to credibility) than advertising, but that advertising is also likely to sustain its weight because of its unique
piece of information. In addition, inconsistency discounting
might occur because of the valence inconsistency. That is,
the evaluations based on either –Puby in the Adx then –Puby
sequence or Adx in the –Puby then Adx sequence will be
discounted because they are opposite in valence. The contrast effect might also be observed. For the Adx then –Puby
sequence, because the valence expectation based on Adx is not
matched by a more credible piece of information (–Puby ),
contrast theory might suggest that the surprising contrast
between two pieces of information will cause the consumer to
inflate the valence discrepancy. Though smaller in magnitude
than it was in the case of the –Pubx then Adx sequence, the
contrast effect might also be present for the –Puby then Adx
sequence as the valence expectation based on –Puby is not
matched by the new piece of information (Adx ). Accordingly,
IIT’s prediction is much closer to subtractive than countersynergistic, while contrast effect theory’s prediction is more
counter-synergistic. The following hypotheses reflect this
counter-synergistic effect:
H5a: Attitude toward the brand under the combined (Adx
& –Puby ) communication condition in the Adx then –Puby
sequence will be significantly lower than that of the publicityonly condition.
H5b: Attitude toward the brand under the combined (Adx
& –Puby ) communication condition in the –Puby then Adx
sequence will be significantly lower than that of the publicityonly condition.
Absolute Attitude Changes
Lastly, using the confirmation effect and the contrast effect
theory, we hypothesize that there will be different magnitudes
of effects of exposure sequence on brand evaluations because
the quality and strength of confirmation or contrast will be
greater for the Ad then Pub sequence than for the Pub then
Ad sequence. The confirmation effect difference was speculated
by Deighton: “evaluative messages may be more prone to
confirmation than factual messages, and so may interact more
with evidence than do factual assertions” (1984, p. 764). For
both Adx & +Pubx and Adx & +Puby conditions, the size of
the confirmation effect would be different depending on the
sequence of exposure because, as Deighton (1984) asserted,
evaluative messages (Adx ) will be more prone to confirmation than factual messages (+Pubx or +Puby ). Therefore, the
Ad then +Pub sequence condition will show a much stronger
confirmation effect than the reversed sequence (+Pub then
Ad). Regarding the contrast effect for negative publicity cases,
Cohen and Goldberg’s (1970) finding shows that participants
reversed choices when strong or weak preceding expectations
were strongly disconfirmed but not when strong priors were
weakly disconfirmed, implying that the initial attitude will
negatively change in the Adx then –Pubx or the Adx then –Puby
sequence, but it will not positively change in the –Pubx then
Adx or the –Puby then Adx sequence. Our last hypothesis addresses this sequential effect difference; the absolute value of
the attitude changes is relevant because it shows the degree
of change independent of directionality (Marks and Kamins
1988).
H6: Absolute attitude change will be significantly greater
for the Ad then Pub sequence group than for the Pub then Ad
sequence group.
Method
A 2 (sequence: Ad then Pub versus Pub then Ad) × 2 (attribute variation: same attribute versus different attribute) × 2
(valence of publicity: positive versus negative) within-subjects
factorial design was used. Through pretests prior to the
main experiment, four suitable product categories (printer,
car insurance, highlighter, and disposable battery) and their
important attributes to consumers were chosen. Although
adding four product categories to a 2 × 2 × 2 design brings
32 combinations of conditions, 16 experiment cells were used
by including two products in each cell. A total of 537 undergraduate students from a large U.S. university participated
in the experiment for extra credit and were free at any time
to opt out of participating. Because each participant evaluated stimuli of two product types, their responses produced
1,074 cases.
The main experiment used a repeated-measures design,
measuring the dependent variable (i.e., attitude toward
brand) twice, once after exposure to advertising and once
after exposure to publicity. A repeated-measures design was
preferred because it allows for the examination of changes in
participant attitudes during a sequence of stimuli exposures
and can furnish more efficient use of subject resources, greater
comparability of the conditions, and reduced error variance
(Keppel and Wickens 2004). Although incidental effects
(i.e., systematic differences incidental to the actual treatment
manipulation, such as carryover effect) might be a limitation
of the within-subject design, if the point of the study is the
sequence of observations, the order of the stimuli is not an incidental problem to the study (Keppel and Wickens 2004).
Spring 2010 103
Pretests
Three pretests were conducted to determine (1) suitable
product categories of high- and low-involvement products,
(2) attributes that people might consider important for a
given product category, and (3) the fictitious brand names for
the products.
First, we chose utilitarian products over hedonic products
because the process of consumer attitude formation for utilitarian products is relatively more straightforward ( Johar and
Sirgy 1991). Furthermore, product categories with different
levels of involvement (i.e., high versus low) were considered
because communication effects and effectiveness can vary as a
function of product involvement (Marks and Kamins 1988;
Smith and Swinyard 1982). Highlighter and battery were
chosen as low-involvement products and laser printer and
auto insurance were chosen as high-involvement products
after pretesting their purchase decision involvement (Mittal
1989). The attributes were determined by free-association
tests: long-lasting experience and grip comfort for highlighter,
battery life and stable power output for battery, printing speed
and toner longevity for printer, customer service and accident
forgiveness for auto insurance.
Fictitious brand names were selected to eliminate any possible confounds. The names generated were low in familiarity,
low in implying product category benefit, neutral in quality,
and neutral in liking: “Thompson Auto Insurance,” “Pendley
Highlighter,” “Rizer Battery,” and “Kasler Laser Printer.”
Main Experiment
Stimuli (Advertisements and Publicity Articles)
Stimulus materials were designed to resemble print ads and
articles in a newspaper. The format of the ads was adapted
from previous studies (e.g., Cline, Altsech, and Kellaris 2003).
The ad consisted of the brand name, headline copy, and an image showing a close-up of the product. The size, layout, and
background of the ads were identical across product types. The
publicity stimuli were designed to resemble actual product
review articles in newspapers. The publicity articles featured
either positive or negative information and contained information either on the same attribute or a different attribute from
the one addressed in the ad. A set of sample stimuli is shown
in the Appendix.
Experimental Procedures
The experiment was conducted online. Links to the online
survey were provided, and participants underwent a selfadministered survey. A random link generator randomly
assigned participants to one of the 16 cells. Participants
viewed a total of two ads and two publicity articles in random
order (see Table 2). Stimuli were counterbalanced to control
any undesired incidental (or nuisance) effects systematically
(e.g., interactions between unrelated products, confirmation
bias effects [Deighton 1984] between unrelated products, or
context effects [Keppel and Wickens 2004]). For example, a
negative review for Kasler printer was placed after a positive
review for Pendly highlighter in cell no. 1, but it was placed
before a positive review about Rizer battery in cell no. 6.
Each participant viewed a mix of low- or high-involvement
products, positive or negative reviews on either the same or
different product attribute. Also, participants were specifically
instructed to disregard previous brands and focus only on the
stimulus brand at hand. After each stimulus was shown, attitude toward the message, message credibility, and attitude
toward the brand were measured.
Manipulation and Confounding Checks
Manipulation checks were implemented for the following
criteria: (1) attribute given in ad, (2) attribute given in publicity article, (3) valence of the publicity article, and (4) product
involvement. Cases were deleted where incorrect answers were
given on any one of the above manipulation criteria, resulting in 70 deleted cases from a total of 1,074 (=537 × 2). This
adjustment reduced the total number of cases to 1,004.
Product involvement was measured at the beginning of
the experiment using the same purchase decision involvement (PDI) scale to confirm manipulation success in the
experiment (Mittal 1995). Results showed that the product
categories of highlighter (M = 2.99, SD = 1.49) and battery
(M = 4.04, SD = 1.30) had lower mean scores of PDI than the
high-PDI products, printer (M = 5.47, SD = 1.09) and auto
insurance (M = 6.15, SD = .92). The difference between the
means of high- and low-PDI products were significant, F(2,
1002) = 757.08, p < .001. Subsequent analyses for hypothesis
tests, therefore, used pooled data of printer and auto insurance
as “high-involvement product” (HI), and highlighter and battery as “low-involvement” product (LI).
As a confounding check, participants’ general attitude toward advertising (see Pollay and Mittal 1993 for details about
the scale) and general attitude toward publicity (in a reduced
form of the scale from Yang and Oliver 2004) were measured
at the beginning. Results showed no difference in each measure
for Aad in general, F(7, 995) = . 893, p = .511 and for Anews
(attitude toward the news) in general; F(7, 995) = 1.393,
p = .205, across all cells and also no difference for any pair of
post hoc comparisons (all ps > .193) between cells.
In addition, attitudes toward the stimulus (i.e., advertising
or publicity) were measured to ensure that all participants
had equivalent attitudes toward the same stimulus, though
in different cells. For example, attitude toward the Rizer
104
The Journal of Advertising
Table 2
Order of the Stimuli Given and Mean of Ab in Each Condition
Cell no.
n (Cases)
C1
74
C2
74
C3
70
C4
77
C5
66
C6
70
C7
66
C8
63
C9
48
C10
51
C11
57
C12
40
C13
57
C14
57
C15
59
C16
75
First stimulus
Second stimulus
Third stimulus
Fourth stimulus
P_AD
(M = 3.88, SD = .72)
P_AD
(M = 3.70, SD = 1.03)
P_AD
(M = 3.79, SD = 1.02)
P_AD
(M = 3.71, SD = .98)
P_+Y
(M = 4.99, SD = .83)
P_–Y
(M = 2.95, SD = .72)
P_+X
(M = 4.64, SD = .74)
P_–X
(M = 2.79, SD = .64)
I_AD
(M = 3.30, SD = .80)
I_AD
(M = 3.54, SD = .92)
I_AD
(M = 3.50, SD = .76)
I_AD
(M = 3.23, SD = .70)
I_+Y
(M = 4.92, SD = .74)
I_–Y
(M = 2.91, SD = .63)
I_+X
(M = 4.90, SD = .75)
I_–X
(M = 2.54, SD = .72)
H_+X
(M = 4.88, SD = .76)
H_–X
(M = 2.63, SD = .78)
H_+Y
(M = 4.57, SD = .89)
H_–Y
(M = 3.13, SD = .83)
B_–X
(M = 2.90, SD = .71)
B_+X
(M = 4.99, SD = .70)
B_–Y
(M = 2.69, SD = .71)
B_+Y
(M = 5.03, SD = .69)
H_AD
(M = 5.01, SD = .78)
H_AD
(M = 4.44, SD = .70)
H_AD
(M = 4.99, SD = .84)
H_AD
(M = 4.79, SD = .76)
B_AD
(M = 3.62, SD = .59)
B_AD
(M = 3.90, SD = .71)
B_AD
(M = 3.84, SD = .95)
B_AD
(M = 3.91, SD = .84)
P_–Y
(M = 2.71, SD = .91)
P_+Y
(M = 5.07, SD = .94)
P_–X
(M = 2.21, SD = .86)
P_+X
(M = 5.11, SD = 1.07)
P_AD
(M = 5.10, SD = 1.02)
P_AD
(M = 3.21, SD = .82)
P_AD
(M = 4.91, SD = 1.14)
P_AD
(M = 2.71, SD = 1.11)
I_+Y
(M = 5.41, SD = 1.11)
I_–Y
(M = 2.26, SD = .83)
I_+X
(M = 5.25, SD = .94)
I_–X
(M = 2.02, SD = .79)
I_AD
(M = 5.31, SD = 1.02)
I_AD
(M = 2.64, SD = .78)
I_AD
(M = 5.59, SD = .89)
I_AD
(M = 2.39, SD = .83)
H_AD
(M = 5.57, SD = .92)
H_AD
(M = 2.44, SD = .84)
H_AD
(M = 5.21, SD = .96)
H_AD
(M = 3.42, SD = .90)
B_A
(M = 2.34, SD = .90)
B_AD
(M = 5.05, SD = 1.10)
B_AD
(M = 2.61, SD = 1.01)
B_AD
(M = 5.44, SD = .90)
H_–X
(M = 2.04, SD = .58)
H_+X
(M = 5.58, SD = .90)
H_–Y
(M = 3.27, SD = 1.16)
H_+Y
(M = 5.58, SD = 1.04)
B_–X
(M = 1.88, SD = .69)
B_+X
(M = 5.57, SD = 1.02)
B_–Y
(M = 2.04, SD = .83)
B_+Y
(M = 5.54, SD = .88)
Notes: Ab = attitude toward the brand; P = printer; I = insurance; H = highlighter; B = battery; AD = advertising; + = positive publicity; – = negative
publicity; X = publicity about product attribute consistent with ad; Y = publicity about product attribute inconsistent with ad; M = mean;
SD = standard deviation.
battery ad (attitude toward the ad itself, not toward the brand)
should not be different between any pair of C13, C14, C15,
and C16. It is important to note that in order to compare
pure attitudes, only attitudes measured on the first stimuli
were used in the analyses because the second stimuli could
be contaminated by the previous stimuli for the same brand.
For example, attitudes toward the Rizer advertisement in
C5, C6, C7, and C8 cannot be considered as pure as those in
C13, C14, C15, and C16. A four-item, seven-point semantic
differential scale (i.e., bad/good, dislike/like, uninteresting/
interesting, irritating/not irritating) was used to measure the
attitudes toward the ad and publicity article (Mitchell and
Olson 1981). Post hoc pairwise comparison tests of analysis
of variance (ANOVA) showed that there was no difference
in the Aad measure (all ps > .05) for the same advertisements
across a set of cells where the ad exposure came first (before
seeing the same-brand publicity). For the comparison tests
of Anews, we assumed that if the only difference between two
publicity articles is the product attribute (X versus Y) and the
attributes are regarded as equally important, Anews should not
be different between Anews about X and Anews about Y . For example,
the attitude toward the positive publicity article about attribute X of Rizer (in C6) should be the same as the attitude
toward the positive publicity article about attribute Y of
Rizer (in C8). Post hoc pairwise comparison tests of ANOVA
showed that there was no difference in the Anews measure (all
ps = 1.00) across the set of same-valence cells where the article
exposure came first. All scales used were internally reliable.
(See Table 3 for specific information on measures and scale
reliabilities.)
Spring 2010 105
Table 3
Measures and Scale Reliabilities
Name of scale
Items (all seven-point scales)
Cronbach’s a
Purchase decision
Not care at all/care a great deal as to which brand I buy;
involvement (PDI)
Various types and brands in the market are all very alike/all very different ;
(Mittal 1995)
Not at all important/extremely important to make a right choice when
buying product ;
Not at all concerned/very much concerned about the outcome of choice
.94
Aad in general
(Pollay and Mittal 1993)
Overall, I like advertising;
My general opinion about advertising is unfavorable (RC);
I consider advertising a good thing
.90
Anews in general
(Yang and Oliver 2004)
Overall, I consider news to be valuable;
Generally, I think news is trustworthy;
I find news to be interesting
.65
Message credibility
(Hallahan 1999a)
Not informative/informative;
Untrustworthy/trustworthy;
Inaccurate/accurate;
Unconvincing/convincing;
Not believable/believable
Ad = .87;
publicity = .90
Attitude toward brand
(Simons and Carey 1998)
Brand looks bad to me/brand looks good to me;
I don’t like the brand/I like the brand;
The brand is undesirable/the brand is desirable;
I feel negatively about the brand/I feel positively about the brand
Ad = .93;
publicity = .98
Note: Aad = attitude toward the ad; Anews = attitude toward the news; RC = reverse coded.
Results
To test the hypotheses, comparison tests were performed on
the relevant measures, but the measures from the exposure
to first and second stimuli (whether advertising or publicity)
were treated differently. As explained previously, when ad exposure comes prior to publicity, the ad measures assess the pure
effects of the ad; the measures for publicity subsequent to the
ad assess the effects of exposure to both the ad and publicity.
The same logic applies to the opposite sequence. For example,
in the Pub then Ad sequence, publicity will be the measure of
the stimuli given first and will be treated as a control. Marks
and Kamins (1988) successfully used this approach in their
advertising-product sampling sequence study.
Hypothesis 1 suggests that the credibility of publicity will
be significantly higher than the credibility of advertising in
all conditions. Credibility of advertising and publicity was
compared in all eight combination conditions. As we had two
sequences and two products, the total of 32 data blocks (16
cells × 2 products) were divided by 4 (2 × 2), resulting in eight
conditions to be analyzed: HI+X, HI−X, HI+Y, HI−Y, LI+X,
LI−X, LI+Y, and LI−Y, where HI, LI, +, −, X, and Y refer
to high-involvement products, low-involvement products,
positive publicity, negative publicity, consistent attribute,
and inconsistent attribute, respectively. The results from every condition supported H1 by showing that the credibility
of publicity is significantly higher than that of advertising
(HI+X: t[137] = −12.31, p < .001; HI−X: t[104] = −11.93,
p < .001; HI+Y: t[142] = −15.49, p < .001; HI−Y:
t[105] = −13.02, p < .001; LI+X: t[145] = −10.44, p < .001;
LI−X: t[101] = −9.81, p < .001; LI + Y: t[142] = −9.84,
p < .001; LI−Y: t[120] = −6.63, p < .001). Because the data
strongly supported H1, we could carry out the subsequent
hypothesis tests.
Hypothesis 2a proposed that the attitude toward the brand
under the combined communication condition in the Adx then
+Pubx sequence would be significantly higher than that of the
ad-only or publicity-only condition. For this hypothesis and
other subsequent hypotheses, a mean of combined condition
(i.e., attitude measured after second stimuli) was compared
with (1) the mean of the first condition within the same group,
and (2) the mean of the first condition from the reversed sequence group, using a paired-samples t-test. We also compared
the combined effect measure with an uncontaminated measure
of +Pubx, which can only be found in the reversed sequence
condition. A one-sample t-test, which examines whether the
sample mean is significantly different from expected (i.e.,
null hypothesis mean), was used to compare the means. For
H2a, the null hypothesis mean was the mean of “+Pubx only”
from the +Pubx then Adx sequence. All variables used in the
paired-samples t-test and the one-sample t-test conformed
to the normality assumption (−1.96 < all skewness < 1.96,
106
The Journal of Advertising
−1.96 < all kurtosis < 1.96). t-test results shown in Table 4
(HI/+X and LI/+X group) indicate that the attitude toward
the brand after exposure to both ad and publicity in the Ad
then Pub sequence was significantly higher than the means of
both ad-only and publicity-only conditions, supporting H2a
(all p < .002 for HI and LI conditions). Hypothesis 2b was also
supported as the measures of attitude toward the brand under
the combined (Adx & +Pubx) communication conditions in the
Pub then Ad sequence were significantly higher than ad-only
or publicity-only conditions (for both HI and LI conditions,
all ps < .001), thus demonstrating a confirmation effect.
Hypothesis 3a suggested that the measures of attitude
toward the brand under the combined communication conditions in the Adx then +Puby sequence would be significantly
higher than the ad-only or publicity-only conditions. Results
in Table 4 show that the hypothesis was supported (all ps < .05
for HI and LI). Again, results for the different sequence were
the same. Hypothesis 3b was also supported because the
measures of attitude toward the brand under the combined
communication conditions in the +Puby then Adx sequence
were significantly higher than the ad-only or publicity-only
condition (all ps < .05 for HI and LI).
Hypothesis 4a hypothesized that the measures of attitude
toward the brand under the combined communication conditions in the Adx then –Pubx sequence would be significantly
lower than that of the publicity-only conditions. Results
in Table 4 indicate that all conditions supported H4a (all
ps < .001). Hypothesis 4b suggested that the measures of
attitude toward the brand under the combined communication conditions in the –Pubx then Adx sequence would be
significantly lower than the publicity-only conditions. Results
in Table 4 show that for the Pub then Ad sequence, attitude
toward the brand after exposure to negative publicity (–X)
about the high-involvement product was not significantly different from the attitude based on the combined communication
condition ( p = .40). On the other hand, the low-involvement
condition supported H4b ( p = .005).
Hypothesis 5a proposed that the measures of attitude
toward the brand under the combined communication conditions in the Adx then –Puby sequence would be significantly
lower than the publicity-only conditions. Results show that
the hypothesis was supported for high-involvement products
( p < .003) but not for low-involvement products. As shown
in Table 4, for the Ad then Pub sequence, attitude toward
the brand after exposure to negative publicity (–Y) about the
low-involvement product was not significantly different from
the attitude based on the combined communication condition ( p = .13). The next hypothesis, H5b, suggested that the
measures of attitude toward the brand under the combined
communication conditions in the –Pubx then Ady sequence
would be significantly lower than the publicity-only condition. This hypothesis was not supported. As indicated in Table
4, for the Pub then Ad sequence, attitude toward the brand
after exposure to negative publicity (–Y) about both high- and
low-involvement products was not significantly different from
their respective attitude based on the combined communication condition ( p = .81 for HI; p = .28 for LI).
Finally, our last hypothesis, H6, proposed that the absolute
attitude changes would be significantly greater for the Ad
then Pub sequence groups than for the Pub then Ad sequence
groups. One-way ANOVA (for homogeneity of variance conditions) and Welch’s test of equality of means (for heterogeneity
of variance conditions) were used for the analyses. Welch’s test
was used in cooperation with ANOVA due to unequal sample
size between groups and the heterogeneity of variance, which
can be problematic for the t-test (Lewis-Beck, Bryman, and
Liao 2003). Welch’s test (Day and Quinn 1989; Welch 1947;
Wilcox 1993) uses adjusted degrees of freedom to protect
against increased type 1 errors under variance heterogeneity
(Quinn and Keough 2002). It is a powerful alternative to the
t-test and F-test when sample sizes are unequal and heterogeneity of variance is present (Lewis-Beck et al. 2003; Maxwell and
Delaney 2004). Results in Table 5 show that the hypothesis
was supported in all conditions (all ps < .001).
In addition, two post hoc analyses were performed to find
any differences in the attitudes of the combined condition
(1) between the Ad then Pub sequence versus reverse sequence
for each group, and (2) between the attribute-consistent versus
attribute-inconsistent condition under the same valence and
same sequence. For the first analysis, the combined communication effects were compared between the Ad then Pub and
the Pub then Ad sequences (i.e., between-sequence comparisons) for each group. This additional analysis was meant to
verify and extend the recent findings of Loda and Coleman
(2005), which showed that the Pub then Ad sequence produces
higher attitude than the reversed sequence. One-way ANOVA
and Welch’s tests produced interesting results. As shown in
Table 6, regardless of the involvement level of product and
product attribute consistency, there was no significant difference between the Ad then Pub sequence and the reverse
sequence for positive publicity conditions, but the attitudes in
the Ad then Pub sequence were significantly lower than those
in the reverse sequence for negative publicity conditions.
Another post hoc analysis was implemented to determine
whether there was any difference in the attitudes of the combined condition between attribute consistent and attribute
inconsistent under the same valence and same sequence (i.e.,
within-sequence comparison). Welch’s test was also jointly used
with ANOVA for this analysis. Some of the results in Table 7
demonstrate that positive valence groups show no difference in
attitude between attribute-consistent and attribute-inconsistent groups (e.g., HI/+X versus HI/+Y) regardless of sequence
condition. However, other results in Table 7 show that when
the attribute advertised was negatively featured in a publicity
Not supported
Not supported
Not supported
Supported
Supported
Not supported
Supported
Supported
Supported
Supported
Supported
Supported
Supported
Supported
Supported
Supported
Hypothesis
supported
Notes: HI = high involvement; LI = low involvement; Ab_ad = attitude toward the brand based on ad only; Ab_pub = attitude toward the brand based on publicity only; Ab_ad+pub = attitude
toward the brand based on the combination of ad and publicity.
p < .001
p < .05
p < .001
p < .001
p < .001
p < .001
p < .001
p < .001
p < .001
p < .05
p < .001
p < .001
p < .001
p < .05
p < .001
p < .001
p < .001
p < .001
p < .001
p < .001
p < .001
p = .40
p < .001
p < .01
p < .001
p < .003
p < .001
p = .13
p < .001
p = .81
p < .001
p = .28
3.61 (.88)
5.18 (1.00)
4.77
4.16 (.75)
5.57 (.96)
4.93
3.61
5.25 (1.07)
4.77 (.75)
4.16
5.34 (1.03)
4.93 (.73)
3.53 (.95)
5.21 (1.02)
4.96
4.26 (.91)
5.56 (.94)
4.78
3.53
5.20 (1.02)
4.96 (.79)
4.26
5.32 (.93)
4.78 (.83)
3.62 (.96)
2.15 (.83)
2.65
4.26 (.97)
1.96 (.64)
2.75
3.62
2.53 (.97)
2.65 (.69)
4.26
2.39 (.86)
2.75 (.76)
3.73 (.82)
2.52 (.89)
2.94
4.44 (1.06)
2.68 (1.16)
2.93
3.73
2.97 (.84)
2.94 (.68)
4.44
3.07 (1.02)
2.93 (.81)
H2a
HI/+X
Ad/Pub
LI/+X
Ad/Pub
H2b
HI/+X
Pub/Ad
LI/+X
Pub/Ad
H3a
HI/+Y
Ad/Pub
LI/+Y
Ad/Pub
H3b
HI/+Y
Pub/Ad
LI/+Y
Pub/Ad
H4a
HI/–X
Ad/Pub
LI/–X
Ad/Pub
H4b
HI/–X
Pub/Ad
LI/–X
Pub/Ad
H5a
HI/–Y
Ad/Pub
LI/–Y
Ad/Pub
H5b
HI/–Y
Pub/Ad
LI/–Y
Pub/Ad
–10.09 (67)
3.32 (67)
–9.73 (62)
5.53 (62)
12.75 (69)
–4.75 (69)
10.45 (82)
–4.04 (82)
–12.20 (69)
2.08 (69)
–9.49 (67)
6.79 (67)
14.05 (72)
–2.00 (72)
9.85 (74)
–5.00 (74)
8.36 (42)
–3.93 (42)
13.58 (38)
–7.75 (38)
–8.80 (61)
.86 (61)
–17.24 (62)
2.94 (62)
8.77 (47)
–3.25 (47)
11.04 (49)
–1.53 (49)
–6.91 (57)
–.25 (57)
–11.29 (70)
–1.09 (70)
Ab_ad
Ab_pub
Ab_ad+pub
t (df )
Significance
Relevant hypothesis
Group
Sequence
Table 4
Results by Analysis Group and Hypotheses
Spring 2010 107
108
The Journal of Advertising
Table 5
Absolute Attitude Changes
Absolute attitude change
Relevant Ad/Pub
hypothesis
Group
sequence
H6
HI/+X
HI/+Y
LI/+X
LI/+Y
HI/–X
HI/–Y
LI/–X
LI/–Y
Pub/Ad
sequence
Statistic
Significance
1.65 (1.16)
1.82 (.92)
1.48 (1.07)
1.41 (.98)
1.54 (1.05)
1.29 (.86)
2.33 (1.05)
1.77 (1.11)
.72 (.64)
.74 (.76)
.72 (.71)
.81 (.69)
.75 (.77)
.78 (.56)
.75 (.69)
.77 (.69)
W(1, 104.24) = 33.64
F(1, 141) = 58.28
W(1, 101.70) = 23.88
W(1, 118.21) = 17.26
F(1, 103) = 19.49
W(1, 77.78) = 14.62
W(1, 58.24) = 66.10
W(1, 75.13) = 32.07
p < .001
p < .001
p < .001
p < .001
p < .001
p < .001
p < .001
p < .001
Hypothesis
supported
Supported
Supported
Supported
Supported
Supported
Supported
Supported
Supported
Notes: HI = high involvement; LI = low involvement.
Table 6
Comparisons Between Sequences by Analysis Group
Ab_ad+pub
Ad/Pub
Group
sequence
Pub/Ad
sequence
HI/+X
LI/+X
HI/+Y
LI/+Y
HI/–X
LI/–X
HI/–Y
LI/–Y
5.25 (1.07)
5.34 (1.03)
5.20 (1.02)
5.32 (.93)
2.54 (.97)
2.39 (.86)
2.97 (.84)
3.07 (1.02)
5.18 (1.00)
5.57 (.96)
5.21 (1.02)
5.56 (.94)
2.15 (.83)
1.96 (.64)
2.52 (.89)
2.68 (1.16)
Statistic
F(1, 136) = .17
F(1, 144) = 1.98
F(1, 141) = .01
F(1, 141) = 2.32
F(1, 103) = 4.48
W(1, 96.46) = 8.61
F(1, 104) = 6.93
F(1, 119) = 3.75
p-value
Significance
p = .68
p = .16
p = .94
p = .13
p < .05
p < .01
p < .05
p = .05
n.s.
n.s.
n.s.
n.s.
significant
significant
significant
significant
Notes: Ab_ad+pub = attitude toward the brand based on the combination of ad and publicity; HI = high involvement; LI = low involvement; n.s. = not
significant.
article, regardless of the sequence, participants evaluated the
brand significantly less favorably than the different-attribute,
negative publicity condition. Together with the first post hoc
study result, this result implies that the strongest contrast
effect might occur for the Adx & −Pubx condition.
Discussion
The purpose of this study was to examine the effects of varying combinations of advertising and publicity, in terms of the
attribute consistency of publicity, valence of publicity, and
the sequence of exposure, on attitude toward the brand. Our
hypotheses were based on the basic assumption that publicity is
generally thought to be more credible and more influential than
other forms of company-controlled communication (Bond and
Kirshenbaum 1998). Hypothesis 1 tested this assumption and
was unanimously supported by all conditions. Theories of Information Integration (Anderson 1971), Integrated Information
Response (Smith and Swinyard 1982), confirmation effect (LaBella and Koehler 2004), and contrast effect (Hovland, Harvey,
and Sherif 1957) were used to propose and test the hypotheses.
Overall, compared to IIT and IIRM, the confirmation effect
theory was more useful and accurate in predicting the combined
effects of positive publicity and advertising. In predicting the
combined effects of negative publicity and advertising, however, the proposed usability of the contrast effect theory was
only noteworthy for the Adx then −Pubx sequence (H4a), the
low-involvement condition of the −Pubx then Adx sequence
(H4b), and the high-involvement condition of the Adx then
−Puby sequence (H5a). Taken as a whole, however, the results
of this study demonstrate the overall ability of the confirmation
effect (for synergy) and the contrast effect (for counter-synergy)
to explain integrated communication conditions.
The results show that for all positive publicity conditions, the combined communications condition produced a
significantly higher attitude toward the brand than either
LI/+Y
5.57
F(1, 129) = .02
p = .90
(.96)
HI/–Y
2.52
F(1, 89) = 4.14
p < .05
(.89)
LI/–Y
2.68
W(1, 79.23) = 14.04
p < .001
(1.16)
LI/+X
5.57
(.96)
HI/–X
2.15
(.83)
LI/–X
1.94
(.64)
LI/–X
2.39
(.86)
HI/–X
2.54
(.97)
LI/+X
5.34
(1.03)
HI/+X
5.25
(1.07)
LI/–Y
3.07
(1.02)
HI/–Y
2.97
(.84)
LI/+Y
5.32
(.93)
HI/+Y
5.20
(1.02)
F(1, 132) = 16.78
F(1, 118) = 6.66
F(1, 156) = .02
F(1, 141) = .63
Statistic
Pub-Ad sequence
Notes: Ab_ad+pub = attitude toward the brand based on the combination of ad and publicity; HI = high involvement; LI = low involvement.
HI/+Y
5.21
F(1, 136) = .05
p = .83
(1.02)
HI/+X
5.18
(1.00)
Ab_ad+pubAb_ad+pub
Means
Statistic
Significance
Means
Ad-Pub sequence
Table 7
Comparisons Between Attribute-Consistent Versus Attribute-Inconsistent Condition
p < .001
p < .02
p = .88
p = .79
Significance
Spring 2010 109
110
The Journal of Advertising
the advertising-only or publicity-only condition, regardless
of product attribute consistency and exposure sequence, thus
upholding the confirmation effect.
For negative publicity and consistent product attribute (i.e.,
H4), the proposed contrast effects were supported in most conditions of the Ad then Pub sequence. In addition to the overall
contrast effect, we further predicted that the magnitude of the
contrast effect would be greater for the Ad then Pub sequence
than for the reverse sequence. As predicted, for both high- and
low-involvement conditions under the Ad then Pub sequence,
the combination of negative publicity with advertising produced a significantly more negative attitude compared to the
publicity-only condition, supporting the contrast effect of H4a.
Under the Pub then Ad sequence, the low-involvement condition only demonstrated the contrast effect with a significantly
more negative attitude in the combined condition, while no
significant contrast effect was found for the high-involvement
products in the −Pubx then Adx sequence.
For negative publicity conditions and inconsistent product
attribute, the proposed contrast effects were observed only for
the high-involvement condition in the Ad then Pub sequence
(partial support of H5a).
These four unsupported cases provide useful insights. IIT’s
inconsistency discounting process seems more typically and
significantly evident for the −Pub then Ad sequence (except
for the low involvement–consistent attribute condition) than
the Ad then −Pub sequence, probably because the negative
publicity was credible enough to discount the advertising effect.
In addition, those unexpected results of H4b, H5a, and H5b
might possibly result from primacy-recency effects (e.g., Lana
1963; Springbett 1958). Based on Lana’s (1963) finding that
people with high interest in a topic exhibit the primacy effect,
the unforeseen result (A[−Pubx then Adx] = A[−Pubx]) of the
high involvement product situation of H4b might have been
caused by the primacy effect, as the persuasion consequence
of the initial communication (−Pubx) dominated the effect of
Adx. In the same vein, the unpredicted result (A[−Puby then
Adx] = A[−Puby]) of the high-involvement situation of H5b
might also be explained by the primacy effect. On the other
hand, lack of support for H5a (A[−Puby then Adx] = A[−Puby])
might be due to the recency effect, which is theorized to be
present under low-involvement conditions (Brunel and Nelson
2003).
We also found, from H6, that the absolute attitude change
was significantly greater for the Ad then Pub sequence group
than for the Pub then Ad sequence group, indicating a strong
support for greater contrast and confirmation effect in the Ad
then Pub sequence. As Marks and Kamins (1988) showed, the
absolute value of the attitude changes is relevant because it
shows the degree of change independent of directionality.
Our two post hoc analyses added more findings. The first
finding was that there was no significant difference in final at-
titude between the Ad then Pub sequence and reverse sequence
for high-involvement products, but the attitudes in the Pub
then Ad sequence were significantly greater than those in reverse
sequence, regardless of involvement level and product attribute
consistency when negative publicity was combined. Contrary to
the prediction of IIT’s inconsistency discounting process (that
the later stimulus in sequence should get a discounted weight
for both the Adx then −Pubx and −Pubx then Adx sequence),
our result showed that the −Pub then Ad sequence generated
significantly lower attitudes than the Ad then −Pub sequence.
As we discussed briefly for H4 and H5, the contrast effect could
be powerful enough to reverse the inconsistency discounting
effect. In addition, the inconsistency discounting effect might
happen retroactively if the former attitude stimulus has less
credibility. This idea needs further theoretical investigation in
future studies. The second post hoc analysis showed that the
contrast effect was stronger and that participants evaluated the
brand significantly less favorably for the combined condition
of Adx & −Pubx than the Adx & −Puby condition, regardless of
sequence. Together with the sequence effect finding from the
first post hoc study result, this finding implies that the Adx &
−Pubx condition will produce the greatest contrast effect.
There are some limitations within which our results should
be interpreted, such as the laboratory study conditions under
which the data were collected, the limited generalizability to
a larger population beyond college students, and the less controlled environment of an online experiment. In addition, some
of our results may not successfully differentiate additive versus
synergistic or subtractive versus counter-synergistic effects. For
example, though the result of H2 showed that the combination
of advertising and publicity produced more positive attitudes
compared to either the advertising- or publicity-only condition,
the result doesn’t sufficiently indicate whether a synergistic or a
combined effect for advertising and publicity was the cause. Future research could extend our study and compare the combined
effect to multiple exposures of the same ad or publicity. If, for
example, the Adx & +Pubx condition generated a more favorable
attitude compared to the double exposure condition of Adx or
+Pubx, that result might be considered much stronger evidence
for the confirmation effect. The same empirical test could be
done for negative publicity. If, for example, the Adx & –Pubx
condition generated a less favorable attitude compared to the
double exposure condition of –Pubx, that result might be considered much stronger evidence for the contrast effect. Within
our theoretical framework, however, our results at least upheld
the confirmation and contrast effect over IIT and IIRM.
Nonetheless, our study offers many insights into the current knowledge of advertising practice and theory regarding
the combination effect of advertising and publicity in various
conditions, which were relatively simple (compared to the
real world) but reasonably comprehensive. Our overall finding
is consistent with prior studies: Combined effects of various
Spring 2010 111
communication tools are more significant than the effects of a
single tool (Stammerjohan et al. 2005). Marketers might find
our findings useful when they design a marketing plan with
multiple communication tactics or when they react to a marketing or public relations crisis caused by negative publicity.
When hit with negative publicity of a brand, which might
convey information about a specific attribute, marketers and
advertisers could decide to counter it by launching a campaign
that addresses the positive qualities of that attribute or reinforces some other piece of information to decentralize the focus
given to the negative publicity, ultimately balancing out the
negative effects it might have on the brand. Our results suggest otherwise, however. Trying to balance out the negative
publicity effect by showing ads that have attributes that are
consistent or inconsistent with the publicity might cause even
more damage to brand attitude. Instead of using advertising
to counter negative publicity, it might be more beneficial to
release publicity for damage control. Future studies should
explore this strategy. On the other hand, when the brand receives positive publicity, our findings suggest that advertising
will further boost consumer attitudes, regardless of timing
of execution (before or after publicity). Future studies could
investigate various other forms of communication integration. For example, the cross-promotion possibility between
advertising and publicity (i.e., publicity featured in advertising and advertising in publicity) is a relatively unexplored yet
worthwhile topic that merits future investigation. This type
of integration might create a whole different level of synergy
where the confirmation effect might occur on multiple levels.
Future studies could also investigate how consumers might
react to the combination of advertising and other types of communication (such as sales promotion and direct marketing) that
might have similar levels of perceived credibility. Moreover,
one could examine combinations of advertising and other types
of more credible experiences, such as guided product trials in
a store. Furthermore, the combination of more than two types
of communication, such as ad, publicity, and sales promotion
or direct marketing, in various sequences, types of valence, and
levels of attribute consistency, could be studied. Investigation of
more than two communication methods might reveal whether
there are diminishing patterns of marginal confirmation or
contrast effects as the number of communication methods
increases. Such studies could provide useful insights into the
optimal number of different communication tools that can be
integrated for the best return on investment.
References
Ahluwalia, Rohini, Robert E. Burnkrant, and Rao H. Unnava
(2000), “Consumer Response to Negative Publicity: The
Moderating Role of Commitment,” Journal of Marketing
Research, 37 (May), 203–214.
Anderson, Norman H. (1971), “Integration Theory and Attitude
Change,” Psychological Review, 78 (May), 171–206.
Anderson, Rolph E. (1973), “Consumer Dissatisfaction: The
Effect of Disconfirmed Expectancy on Perceived Product
Performance,” Journal of Marketing Research, 10 (February),
38–44.
Baker, Michael J. (2002), The Marketing Book, Oxford:
Butterworth-Heinemann.
Balasubramanian, Siva K. (1994), “Beyond Advertising and Publicity: Hybrid Messages and Public Policy Issues,” Journal
of Advertising, 23 (December), 29–58.
Birnbaum, Michael H., and Steven E. Stegner (1979), “Source
Credibility in Social Judgment: Bias, Expertise, and the
Judge’s Point of View,” Journal of Personality and Social Psychology, 37 ( January), 48–74.
Bond, Jonathan, and Richard Kirshenbaum (1998), Under the Radar: Talking to Today’s Cynical Consumer, New York: Wiley.
Brunel, Frederic F., and Michelle R. Nelson (2003), “Message
Order Effects and Gender Differences in Advertising Persuasion,” Journal of Advertising Research, 43 (3), 330–341.
Chaiken, Shelly, and Durairaj Maheswaran (1994), “Heuristic
Processing Can Bias Systematic Processing: Effects of Source
Credibility, Argument Ambiguity and Task Importance on
Attitude Judgment,” Journal of Personality and Social Psychology, 66 (March), 460–473.
Cline, Thomas W., Moses B. Altsech, and James J. Kellaris
(2003), “When Does Humor Enhance or Inhibit Ad Responses?” Journal of Advertising, 32 (Fall), 31–45.
Cohen, Joel B., and Marvin E. Goldberg (1970), “The Dissonance
Model in Post-Decision Product Evaluation,” Journal of
Marketing Research, 7 (August), 315–321.
Day, R. W., and G. P. Quinn (1989), “Comparison of Treatments
After an Analysis of Variance in Ecology,” Ecological Monographs, 59 (4), 433–463.
Deighton, John (1984), “The Interaction of Advertising and
Evidence,” Journal of Consumer Research, 11 (December),
763–770.
Duncan, Thomas R., and Clarke Caywood (1996), “The Concept,
Process, and Evolution of Integrated Marketing Communication,” in Integrated Communication: Synergy of Persuasive
Voices, Esther Thorson and Jeri Moore, eds., Mahwah, NJ:
Lawrence Erlbaum, 13–34.
Durkin, Mark, and Margaret-Anne Lawlor (2001), “The Implications of the Internet on the Advertising Agency–Client Relationship,” Service Industries Journal, 21 (April), 175–190.
Eagle, Lynne, Philip J. Kitchen, and Sandy Bulmer (2007), “Insights into Interpreting Integrated Marketing Communications: A Two-Nation Qualitative Comparison,” European
Journal of Marketing, 41 (November), 956–970.
Eagly, Alice H., and Shelly Chaiken (1993), The Psychology of
Attitudes, Fort Worth, TX: Harcourt Brace Jovanovich
College.
Fishbein, Martin, and Icek Ajzen (1975), Belief, Attitude, Intention,
and Behavior: An Introduction to Theory and Research, Reading,
MA: Addison-Wesley.
Hovland, Carl I., O. J. Harvey, and Muzafer Sherif (1957),
“Assimilation and Contrast Effects in Reactions to
112
The Journal of Advertising
Communication and Attitude Change,” Journal of Abnormal
and Social Psychology, 55 ( July), 244–252.
Johar, J. S., and M. Joseph Sirgy (1991), “Value-Expressive Versus
Utilitarian Advertising Appeals: When and Why to Use
Which Appeal,” Journal of Advertising, 20 (September),
23–34.
Keppel, Geoffrey, and Thomas D. Wickens (2004), Design and
Analysis: A Researcher’s Handbook, Englewood Cliffs, NJ:
Pearson Prentice Hall.
LaBella, Carla, and Derek J. Koehler (2004), “Dilution and
Confirmation of Probability Judgments Based on Nondiagnostic Evidence,” Memory and Cognition, 32 (October),
1076–1089.
Lana, Robert E. (1963), “Interest, Media and Order Effects in Persuasive Communications,” Journal of Psychology, 56, 9–13.
Lewis-Beck, Michael, Alan Bryman, and Tim Futing Liao, eds.
(2003), The Sage Encyclopaedia of Social Science Research Methods, London: Sage.
Loda, Marsha D., and Barbara Carrick Coleman (2005), “Sequence
Matters: A More Effective Way to Use Advertising and
Publicity,” Journal of Advertising Research, 45 (December),
362–372.
Marks, Lawrence J., and Michael A. Kamins (1988), “The Use
of Product Sampling and Advertising: Effects of Sequence
of Exposure and Degree of Advertising Claim Exaggeration on Consumers’ Belief Strength, Belief Confidence, and
Attitudes,” Journal of Marketing Research, 25 (August),
266–281.
Maxwell, Scott E., and Harold D. Delaney (2004), Designing Experiments and Analyzing Data: A Model Comparison Perspective,
Mahwah, NJ: Lawrence Erlbaum.
Mitchell, Andrew A., and Jerry C. Olson (1981), “Are Product
Attribute Beliefs the Only Mediator of Advertising Effects
on Brand Attitude?” Journal of Marketing Research, 18 (August), 318–332.
Mittal, Banwari (1989), “Measuring Purchase-Decision Involvement,” Psychology and Marketing, 6 (Summer), 147–162.
——— (1995), “A Comparative Analysis of Four Scales of
Consumer Involvement,” Psychology & Marketing, 12 (7),
663–682.
Moriarty, Sandra E. (1994), “PR and IMC: The Benefits of Integration,” Public Relations Quarterly, 39 (3), 38–45.
Naik, Prasad A., and Kalyan Raman (2003), “Understanding the
Impact of Synergy in Multimedia Communications,” Journal
of Marketing Research, 40 (November), 375–388.
Nisbett, Richard E., Henry Zukier, and Ronald Lemley (1981),
“The Dilution Effect: Nondiagnostic Information Weakens
the Implications of Diagnostic Information,” Cognitive Psychology, 13 (April), 248–277.
Oliver, Richard L. (1981), “Measurement and Evaluation of Satisfaction Processes in Retail Settings,” Journal of Retailing,
57 (Fall), 25–48.
Pollay, Richard W., and Banwari Mittal (1993), “Here’s the Beef:
Factors, Determinants, and Segments in Consumer Criticism
of Advertising,” Journal of Advertising, 57 ( July), 99–114.
Quinn, Gerald P., and Michael J. Keough (2002), Experimental
Design and Data Analysis for Biologists, Cambridge: Cambridge
University Press.
Robbins, Tina L., and Angelo S. DeNisi (1998), “Mood Vs. Interpersonal Affect Identifying Process and Rating Distortions
in Performance Appraisal,” Journal of Business and Psychology,
12 (Spring), 313–325.
Schultz, Don E. (1996), “The Inevitability of Integrated Communications,” Journal of Business Research, 37 (November),
139–146.
———, and Philip J. Kitchen (1997), “Integrated Marketing
Communications in U.S. Advertising Agencies: An Exploratory Study,” Journal of Advertising Research, 37 (September/
October), 7–18.
Smith, Robert E. (1993), “Integrating Information from Advertising and Trial: Processes and Effects on Consumer Response
to Product Information,” Journal of Marketing Research, 30
(May), 204–219.
———, and William R. Swinyard (1982), “Information Response
Models: An Integrated Approach,” Journal of Marketing, 46
(Winter), 81–93.
———, and ——— (1983), “Attitude-Behavior Consistency:
The Impact of Product Trial Versus Advertising,” Journal of
Marketing Research, 20 (August), 257–267.
———, and ——— (1988), “Cognitive Response to Advertising
and Trial: Belief Strength, Belief Confidence and Product
Curiosity,” Journal of Advertising, 17 (3), 3–14.
———, and Christine A. Vogt (1995), “The Effects of Integrating
Advertising and Negative Word-of-Mouth Communications
on Message Processing and Response,” Journal of Consumer
Psychology, 4 (2), 133–151.
Springbett, B. M. (1958), “Factors Affecting the Final Decision in
the Employment Interview,” Canadian Journal of Psychology,
12 (1), 13–22.
Stammerjohan, Claire, Charles M. Wood, Yuhmiin Chang, and
Esther Thorson (2005), “An Empirical Investigation of the
Interaction Between Publicity, Advertising, and Previous
Brand Attitudes and Knowledge,” Journal of Advertising, 34
(Winter), 55–67.
Sternthal, Brian, Ruby Dholakia, and Clark Leavitt (1978), “The
Persuasive Effect of Source Credibility: Tests of Cognitive Response,” Journal of Consumer Research, 4 (March), 252–260.
Welch, Bernard L. (1947), “The Generalization of Student’s
Problem When Several Different Population Variances are
Involved,” Biometrika, 34 (1/2), 28–35.
Wilcox, Rand R. (1993), “Robustness in ANOVA,” in Applied
Analysis of Variance in Behavioral Science, Lynne K. Edwards,
ed., New York: Marcel Dekker, 345–374.
Yang, Hyeseung, and Mary Beth Oliver (2004), “Exploring the
Effect of Online Advertising on Readers’ Perceptions of
Online News,” Journalism and Mass Communication Quarterly,
81 (Winter), 733–749.
Zukier, Henry, and Dennis L. Jennings (1983/84), “Nondiagnosticity and Typicality Effects in Prediction,” Social Cognition,
2 (3), 187–198.
Spring 2010 113
Appendix
Example of Ad and Publicity Stimuli (Pendley Highlighter)
Ad
+Pubx
Publicity Examples
(+Pubx) PENDLEY Highlighter Leads the Pack in Long-Lasting Experience
ATLANTA (Feb. 6, 2007) – Consumer Reports announced today the results of a study that tested a variety of office supplies for
their performance and functionality. Among similarly priced non-toxic highlighters, PENDLEY ranked highest in long-lasting
experience. “Longevity has been an important feature of highlighters,” said Lance Jones, associate director of Consumer Reports.
“This study showed PENDLEY to be superior in longevity compared to its competitors.”
(–Pubx) PENDLEY Highlighter Lags Behind in Long-Lasting Experience
ATLANTA (Feb. 6, 2007) – Consumer Reports announced today the results of a study that tested a variety of office supplies
for their performance and functionality. Among similarly priced non-toxic highlighters, PENDLEY ranked the lowest in longlasting experience. “Longevity has been an important feature of highlighters,” said Lance Jones, associate director of Consumer
Reports. “This study showed PENDLEY to be inferior in longevity compared to its competitors.”
(+Puby) PENDLEY Highlighter Leads the Pack in Grip Comfort
ATLANTA (Feb. 6, 2007) – Consumer Reports announced today the results of a study that tested a variety of office supplies
for their performance and functionality. For highlighters, among products cleared as non-toxic and in the similar price range,
PENDLEY ranked the highest in grip comfort. “Comfort in grip has been an important feature of highlighters. If the body
of the highlighter is too thick or too thin, customers experience discomfort,” said Lance Jones, associate director of Consumer
Reports. “This study showed PENDLEY to have just the right grip size, providing more comfort than its competitors.”
(–Puby) PENDLEY Highlighter Lags Behind in Grip Comfort
ATLANTA (Feb. 6, 2007) – Consumer Reports announced today the results of a study that tested a variety of office supplies
for their performance and functionality. For highlighters, among products cleared as non-toxic and in the similar price range,
PENDLEY ranked the lowest in grip comfort. “Comfort in grip has been an important feature of highlighters. If the body of the
highlighter is too thick or too thin, customers experience discomfort,” said Lance Jones, associate director of Consumer Reports.
“This study showed PENDLEY to have an inadequate grip size that causes grip discomfort compared to its competitors.”
Download