Available online at www.sciencedirect.

com
Journal of
CONSUMER
PSYCHOLOGY
Journal of Consumer Psychology 21 (2011) 324 – 337

The contrasting effects of negative word of mouth

in the post-consumption stage
Haksin Chan a,⁎, Selina Cui b
a
Department of Marketing, The Chinese University of Hong Kong, Shatin, Hong Kong
b
Arkadin (HK) Ltd., 2402, Citicorp Centre, Causeway Bay, Hong Kong

Received 12 January 2010; revised 16 November 2010; accepted 23 November 2010

Available online 30 December 2010

Abstract

We explore the effects of negative word of mouth (NWOM) from worse-off or similar others in the post-consumption stage. In four
experiments, we show that the ramifications of NWOM are more complex than portrayed in the literature. Specifically, we demonstrate that
attribute-based NWOM has a negative (i.e., aggravating) effect on dissatisfied consumers, whereas experience-based NWOM has a positive (i.e.,
alleviating) effect. Thought-listing data reveal distinct processes underlying the contrasting effects. On one hand, these results are consistent with
the predictions of attitude polarization and downward comparison research. On the other hand, they are explainable in terms of the disconfirmation
model.
© 2010 Society for Consumer Psychology. Published by Elsevier Inc. All rights reserved.

Keywords: Word of mouth; Post-consumption; Dissatisfaction; Attitude polarization; Social comparison; Self-enhancement

Consumers are increasingly sharing product information and
consumption experiences with each other. With the prolifera-
tion of online communities and social media, consumers may
conveniently share their experiences or opinions and, at the
same time, be easily exposed to word-of-mouth (WOM)
influence (Wheat, 2010). This exposure, whether intentional
or incidental, has enormous implications for all stages of
decision making (Schindler & Bickart, 2005). In particular, the
growing opportunities for WOM influence have the potential to
transform the post-consumption processes.
Consider the following excerpts from two Internet consumer
forums.
Consumer A “The hotel was obviously old but it could've been
cleaner. The room had a musty smell; the furniture
was broken & should not have been in the room
since it obviously needed to be replaced. The carpet
was stained & sticky; it also needed to be replaced
or could've been cleaned…There was a microwave,
refrigerator & TV which were all plugged into one
extension cord which…we're sure the Fire Marshal
would not appreciate seeing.”1
Consumer B “I [used] a Clinique sample from one of those free
gift things...I used a very small amount and it
actually made my skin feel great…and [last night]
I used it again. On this application, almost
instantaneously I noticed red blotches forming
on my skin which soon turned to rash like
bumps…When I woke up this morning my face is
bright red over my whole cheeks and most of my
forehead feels like I have been badly sunburnt.”2
How would such postings influence other disgruntled guests
of the hotel and consumers allergic to Clinique products?
Would these consumers be more displeased because they are
more convinced that the hotel/Clinique products are bad (per

⁎ Corresponding author. Fax: +852 2603 5473. 1

http://travel.yahoo.com/p-hotel-376105 (accessed July 15, 2010).
2
E-mail addresses: hchan@cuhk.edu.hk (H. Chan), s.cui@arkadin.com http://forums.vogue.com.au/showthread.php?p=4280911 (accessed July 15,
(S. Cui). 2010).

1057-7408/$ - see front matter © 2010 Society for Consumer Psychology. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.jcps.2010.11.005
 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
attitude polarization research), or would they feel better
knowing that they are not the only victims of a service/product
failure (per downward comparison research)?
The WOM literature is silent on these intriguing questions.
Prior research on WOM influence has focused on the pre-
purchase stage, documenting strong WOM effects on product
involvement (Giese, Spangenberg, & Crowley, 1996), brand
attitude (Herr, Kardes, & Kim, 1991; Laczniak, DeCarlo, &
Ramaswami, 2001), purchase intention (Gilly, Graham,
Wolfinbarger, & Yale, 1998), and purchase behavior (Brown
& Reingen, 1987; Liu, 2006). Nonetheless, the impacts of
WOM on other consumers in the post-consumption stage
remain largely unexplored.
Interestingly, two bodies of social psychological literature—
one on attitude polarization and one on downward comparison—
point to opposite effects of negative WOM (NWOM) on other
dissatisfied consumers. Given the distinct processes that underlie
attitude polarization and downward comparison, we propose and
demonstrate that NWOM (from worse-off or similar others) may
aggravate or alleviate dissatisfaction, contingent on message
content—which may trigger one process versus another. As we
will discuss, both the aggravating effect (per attitude polarization
research) and the alleviating effect (per downward comparison
research) are conceptually linked to the disconfirmation model of
consumer (dis)satisfaction.
A general framework of message content is lacking in the
WOM literature, although recent studies have recognized the
complexity (Laczniak et al., 2001) and adaptability (Schellekens,
Verlegh, & Smidts, 2010; Sengupta, Dahl, & Gorn, 2002) of
WOM content. As conceptualized in this article, attribute-based
WOM highlights the product—what it is and how it performs,
whereas experience-based WOM highlights the consumer—what
happens as a result of purchase and consumption. This distinction
draws on Gutman's (1982) conceptualization that products
possess attributes but consumers experience the consequences
of product attributes. As anecdotal examples, consumer A's
posting is attribute-based and consumer B's posting is experience-
based.
In short, our research investigates the effects of attribute-
based and experience-based NWOM (from worse-off or similar
others). The results of four experiments reveal an aggravating
effect on dissatisfaction when the message is attribute-based and
an alleviating effect when the message is experience-based. The
evidence also suggests that the contrasting effects are driven,
respectively, by a product-evaluative and a social-comparative
process. Moreover, these effects seem to reach beyond
dissatisfaction and contribute to corresponding changes in
repurchase intent. Note, however, that message context and
experience comparability create boundary conditions for these
NWOM effects. Before presenting the empirical work, we lay
out the theoretical foundations for the opposite effects of
NWOM on consumer dissatisfaction.
Attitude polarization
Attitude polarization refers to a social psychological
phenomenon in which an attitude becomes more extreme after
325
exposure to, deliberation on, and/or communication about
attitude-congruent information (Abelson, 1995; Eagly &
Chaiken, 1993). For example, a voter who opposes a public
policy typically becomes more opposed to it after reading an
article against the policy; a consumer who has a favorable
opinion of a product typically becomes more favorable after
hearing positive comments about it.
The attitude polarization effect is most evident when group
discussion is involved. Not only does group discussion facilitate
the exchange of persuasive arguments among individuals, but it
also motivates individuals to adjust their positions in order to
meet or exceed the perceived norm (Eagly & Chaiken, 1993).
These two types of influence—informational and normative—
often occur in combination to produce the polarization effect
(Isenberg, 1986).
A number of other conditions also contribute to attitude
polarization. Evaluating new information, even that which is
mixed or inconclusive, can cause an attitude to become more
extreme (Lord, Ross, & Pepper, 1979), and mere exposure to
the attitude positions of like-minded others is sufficient for
normative forces to exert their polarizing influence (Myers,
1978). In fact, attitude polarization can be induced by sheer
intrapersonal processes such as thinking about an attitude object
(Tesser, 1978) and repeated expression of an attitude (Downing,
Judd, & Brauer, 1992). In a broader framework, these
intrapersonal processes may be conceptualized “as involving
imagined or anticipated interaction with like-minded indivi-
duals, thus implicating the same mediating mechanisms as in
group polarization” (Abelson, 1995, p. 29). Regardless of the
varied theoretical perspectives, attitude polarization is a robust
phenomenon that manifests itself under a variety of conditions,
interpersonal or otherwise.
Attitude polarization among consumers
Previous research has confirmed consumer susceptibility to
informational and normative influences (Bearden, Netemeyer, &
Teel, 1989; Lascu, Bearden, & Rose, 1995). In particular, the
polarization effect as a result of group interactions has been shown
to be pertinent to different aspects of consumer behavior. In their
study of sorority women in the naturalistic environment, Ward
and Reingen (1990) found interesting patterns of polarization for
consumer beliefs and choices. They further observed that group
characteristics may promote or inhibit polarization. On the basis
of experimental data, Rao and Steckel (1991) verified that a
decision model that incorporates the polarization effect outper-
forms models that do not. Recently, Bohlmann, Rosa, Bolton, and
Qualls (2006) demonstrated that satisfaction judgments of both
individual consumers and organizational buyers are susceptible to
the polarizing influence of group interactions—a phenomenon
they labeled as “escalation.”
The aggravating effect of NWOM
A large amount of WOM occurs in the context of group
discussions with family and friends (Ward & Reingen, 1990),
colleagues (Bohlmann et al., 2006), and strangers (Schlosser,
326 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
2005). When dissatisfied consumers exchange NWOM infor-
mation in a group setting, dissatisfaction should be escalated.
Apart from group discussion, other forms of NWOM commu-
nication may also be conducive to dissatisfaction escalation. The
polarization literature suggests that a NWOM message would
have an aggravating effect on the attitudes and judgments of
other dissatisfied consumers, as long as the message conveys
negative opinions and/or arguments in support of such opinions.
Depending on the format of communication (e.g., face-to-face,
telephone, email, Internet forum, etc.), different factors may be
responsible for the aggravating effect. These variations aside, the
polarization literature generally points to an aggravating effect
of NWOM on consumer dissatisfaction.
Downward comparison
In a review and synthesis of social psychology literature,
Wills (1981) convincingly argues that stress and negative affect
evoke a self-enhancement motivation to compare downward
with worse-off targets. In contrast to Festinger's (1954) classic
theory of social comparison, which concerns the evaluation of
abilities and opinions, downward comparison theory highlights
the enhancement of subjective well-being. Specifically, the
favorable comparison between the self and less fortunate others
enables a person to feel better about an unpleasant situation.
In essence, the use of negative information about others to
enhance subjective well-being is a learned “cognitive” strategy
for coping with stress and negative affect (Wills, 1981). It may
involve a passive process (i.e., taking advantage of available
downward targets) or an active process (i.e., finding or creating
downward targets).
Downward comparison theory has received support across
qualitative, experimental, and correlational studies. In an
influential field study, Taylor, Wood, and Lichtman (1983)
found widespread evidence of downward comparison among
breast cancer patients, some of whom even compared
themselves with hypothetical downward targets. The self-
enhancing effects of downward comparison were subsequently
confirmed in lab experiments (e.g., Buunk, Oldersma, & de
Dreu, 2001; Gibbons & Boney McCoy, 1991). Moreover,
surveys of individuals with physical or emotional problems
have generally reported a positive relationship between
subjective well-being and the extent to which one's situation
is perceived as better than those of others (e.g., Dias & Lobel,
1997; Helgeson & Taylor, 1993). Overall speaking, the
downward comparison phenomenon and the resulting allevia-
tion of negative affect are well-documented (Buunk & Gibbons,
2007).
Downward comparison among consumers
Stress and negative affect are “endemic” to consumption
(Duhachek, 2005). Therefore, the downward comparison
phenomenon is likely pervasive among consumers. The
motivation to engage in downward comparison is particularly
high in situations where “frustration or misfortune has occurred
that is difficult to remedy through instrumental action” (Wills,
1981, p. 245). For example, consumers stuck in a slow-moving
queue may derive some comfort from the long line behind them,
even though they cannot do anything to improve the situation.
In support of this view, Zhou and Soman (2003) demonstrated
in a series of experiments that the number of people behind has
a positive influence on the experience of waiting and the
likelihood of staying in the queue. Their research provides solid
evidence that knowing about the existence of less fortunate
others enhances consumer affect and increases consumer
tolerance. By the same token, NWOM from less fortunate others
may have an alleviating effect on consumer dissatisfaction.
The alleviating effect of NWOM
It is common for dissatisfied consumers to experience
negative emotions ranging from disappointment to distress
(Westbrook, 1987; Westbrook & Oliver, 1991). According to
downward comparison theory, these negative emotions are
relieved to some extent by the realization that “things could be
worse.” Therefore, dissatisfied consumers may feel better about
their own situation when they learn that others have worse
experience. In other words, NWOM may have a positive effect
on dissatisfied consumers; namely, the alleviation of dissatis-
faction. Buunk et al.'s (2001) research on relationship
satisfaction lends support to this argument. For individuals in
stable relationships, Buunk et al. found a satisfaction-enhancing
effect of downward comparison. In their analysis, engaging in
downward comparison helps “buffer” against relational discon-
tent by creating a lower reference point.
The role of message content
Our literature review suggests opposite effects of NWOM
(from worse-off others) on consumer dissatisfaction: (1) an
aggravating effect based on attitude polarization research, and
(2) an alleviating effect based on downward comparison
research. This apparent contradiction can be resolved by
recognizing the different thrusts of the two streams of research.
Specifically, the former concerns how others' viewpoints may
influence the evaluation of an issue or an object, whereas the
latter concerns how others' experience may influence the
standard against which a personal situation is compared. We
argue that a NWOM message would produce an aggravating
effect if it triggers a product-evaluative process, and would
produce an alleviating effect if it triggers a social-comparative
process. In terms of (dis)satisfaction theory, these effects and
the underlying processes are in line with a popular view of (dis)
satisfaction formation—the disconfirmation model.
Attitude polarization research emphasizes the deliberation
process—the evaluation of arguments and evidence offered by
others—and the beliefs and opinions thus formed (Abelson,
1995; Eagly & Chaiken, 1993; Sunstein, 2002). It is no
coincidence that the polarization effect is often investigated in
the context of social and political issues (e.g., Downing et al.,
1992; Lord et al., 1979; Myers, 1978). The deliberative/
evaluative focus is also evident in pertinent consumer research,
which has verified the polarizing influence of “evaluative
 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
beliefs” (Ward & Reingen, 1990), “preference evaluations”
(Rao & Steckel, 1991), and “product performance evaluations”
(Bohlmann et al., 2006). In a post-consumption NWOM
context, the exposure to negative evaluations of salient product
attributes likely prompts dissatisfied consumers to deliberate on
the performance of these attributes. Given informational and
normative influences, the result would be a negative shift in
performance perceptions and increased dissatisfaction.
In contrast to attitude polarization research, downward
comparison research involves a process that is subjective/
instinctive rather than deliberative/evaluative. The latter stream
of research documents various psychological benefits (e.g.,
stress reduction, mood improvement, and increase in satisfac-
tion) derived from a favorable comparison between the self and
worse-off others (Buunk & Gibbons, 2007; Wills, 1981). In a
post-consumption NWOM context, negative experience shared
by worse-off others likely prompts dissatisfied consumers to
draw such a comparison. Given a self-enhancement motivation,
the exposure to a salient downward target would contribute to
decreased dissatisfaction.
To recapitulate, the contrasting effects of NWOM on
dissatisfied consumers are driven, respectively, by a product-
evaluative process (that focuses on product attributes) and a
social-comparative process (that focuses on consumer experi-
ence). It follows that the relative strengths of these effects are a
function of the salience of attribute information versus
consumer experience in the NWOM message. When attribute
information is salient (e.g., when a consumer enumerates the
clumsy features of a software product), attitude polarization
research suggests an aggravating effect. When consumer
experience is salient (e.g., when a consumer recounts the
damage done by a defective software product), downward
comparison research suggests an alleviating effect.
We refer to the first type of NWOM as attribute-based and
the second type as experience-based. Specifically, attribute-
based NWOM highlights the product (i.e., what the product is
and how it performs) whereas experience-based NWOM
highlights the consumer (i.e., what happens to the consumer
as a result of purchase and consumption). This conceptualiza-
tion parallels Gutman's (1982) classic distinction between
attributes and consequences—that products possess attributes
but consumers experience the consequences of product
attributes. Both types of NWOM have been featured in
previous research. Schlosser's (2005, Experiment 1) NWOM
manipulation exemplifies attribute-based NWOM: “While a
well done claymation, it seems to lack the ability to tell its
viewer what it is about until the end. The preceding parts of the
show are barren and oblique, as the setting, and in my opinion,
unnecessary.” In contrast, Herr et al.'s (1991, Experiment 2)
NWOM condition is mostly experience-based: “It's the worst
car he's ever had. It seems like it's always in the shop being
repaired. I think he's spent more to keep it running than it
originally cost him.”
The contrasting effects of attribute-based and experience-
based NWOM are also explainable in terms of a major theory
of (dis)satisfaction. According to the disconfirmation model,
(dis)satisfaction reflects the discrepancy between (1) the
327
consumer's perception of product performance and (2) the
standard or reference point adopted by the consumer (Oliver,
1997). Attribute-based NWOM exerts an influence on the first
variable by heightening dissatisfied consumers' negative
perceptions of product performance (thus increasing dissatis-
faction). This influence of NWOM is consistent with prior
research that views consumer perceptions of product perfor-
mance as malleable (Deighton, 1992) and sensitive to social
factors (Fournier & Mick, 1999). On the other hand,
experience-based NWOM exerts an influence on the second
variable by creating a lower reference level for dissatisfied
consumers (thus decreasing dissatisfaction). As proposed by
Woodruff, Cadotte, and Jenkins (1983), WOM communica-
tions constitute one important type of “experience-based
norms” that consumers may use as a frame of reference for
disconfirmation judgments.
In summary, a synthesis of attitude polarization and
downward comparison literatures suggests contrasting effects
of attribute-based and experience-based NWOM on dissatis-
faction. The contrasting effects are also grounded in a major
framework of post-consumption behavior. Formally, we
advance the following hypothesis.
Hypothesis 1. (a) Attribute-based NWOM from a worse-off
other will increase dissatisfaction. (b) Experience-based
NWOM from a worse-off other will decrease dissatisfaction.
Besides these contrasting effects, the literatures on attitude
polarization and downward comparison also point to distinct
processes. As discussed above, a product-evaluative process is
largely responsible for the aggravating effect of NWOM,
whereas a social-comparative process is at the core of the
alleviating effect of NWOM. We expect to find corresponding
process evidence to accompany the hypothesized effects
associated with attribute-based and experience-based NWOM.
Hence, we propose a second hypothesis.
Hypothesis 2. Attribute-based (vs. experience-based) NWOM
from a worse-off other will elicit (a) more product-evaluative
thoughts and (b) fewer social-comparative thoughts.
Pilot experiment
The pilot experiment provided a preliminary test of
Hypotheses 1 and 2 and led to some refinements in
methodology. In this experiment, the 60 participants (45%
male) first rated their experimentally induced, pre-NWOM
dissatisfaction (T1-D) with a bank before receiving either
attribute-based or experience-based NWOM from a worse-off
other. After a brief delay, they reported their post-NWOM
dissatisfaction (T2-D) and performed a thought-listing task.
Consistent with Hypothesis 1, dissatisfaction increased in
the attribute-based condition (MT1-D = 7.49, MT2-D = 8.24,
t = 5.82, p b .001) and decreased in the experience-based
condition (MT1-D = 7.32, MT2-D = 6.86, t = −2.79, p b .05).
However, there was only partial support for Hypothesis 2.
As predicted by Hypothesis 2(a), participants in the attribute-
based condition reported more product-evaluative thoughts
328 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
than did those in the experience-based condition (MAttr = 1.77,
MExp = 1.27, F(1, 58) = 4.61, p b .05). But there was no
support for Hypothesis 2(b), despite the 7 to 13 ratio of
social-comparative thoughts in the attribute-based condition
versus the experience-based condition (MAttr = .23, MExp = .43,
F(1, 58) = 2.08, p N .10).
The lack of statistical support for Hypothesis 2(b) was, for
the most part, due to the small number of social-comparative
thoughts reported (20 out of 135 overall). In retrospect, the
thought-listing task might have elicited only a small portion of
these thoughts. As Wills (1981, p. 265) suggests, there is a
general tendency “to keep [downward] comparison private and
implicit” because it is considered less than admirable. To
circumvent this problem, we introduce a new variable—
message context—in Experiment 1. Arguably, contextual cues
that draw attention to others' experiences would facilitate the
reporting of social-comparative thoughts.
The role of message context
Previous research evinces powerful effects of contextual
cues on consumers' encoding and retrieval of information
(Meyers-Levy & Tybout, 1997). The influence of a message is
generally context-dependent, with message processing biased
by different types of contextual information (Tormala &
Clarkson, 2007).
Typically, NWOM is received in the context of other
messages. If these messages are focused on a particular topic,
that could create a facilitative or inhibitive context for the
processing of the NWOM message. As far as the thought-listing
task is concerned, a people-focused context (i.e., a context that
focuses on individuals and their experiences) makes social
information salient, thus facilitating social comparison and the
reporting of social-comparative thoughts. In this context, we
expect to obtain clear evidence for Hypothesis 2(b). On the
other hand, Hypothesis 2(a) should hold up well in a product-
focused context (i.e., a context that focuses on products and
their evaluations).
Moreover, crossing message context (product-focused vs.
people-focused) with message content (attribute-based vs.
experience-based) provides a stronger and more nuanced test
of the contrasting effects of NWOM on dissatisfaction. Since a
product-focused context draws attention to product attributes, it
should strengthen the impact of attribute-based NWOM (and
possibly weaken the impact of experience-based NWOM).
Conversely, since a people-focused context directs attention to
others' experiences, it should strengthen the impact of
experience-based NWOM (and possibly weaken the impact of
attribute-based NWOM). Hypothesis 3 summarizes the joint
effects of message context and message content on consumer
dissatisfaction.
Hypothesis 3. (a) In a product-focused (vs. people-focused)
context, attribute-based NWOM from a worse-off other will
increase dissatisfaction to a greater extent. (b) In a people-focused
(vs. product-focused) context, experience-based NWOM from a
worse-off other will decrease dissatisfaction to a greater extent.
Experiment 1
Method
One hundred and twenty one undergraduates (56% male)
participated in a “consumer study” in exchange for a small gift.
They were randomly assigned to a 2 (message context: product-
focused, people-focused) × 2 (message content: attribute-based,
experience-based) × 2 (time: T1, T2) repeated-measures design.
At the onset of the experiment, participants were instructed to
imagine themselves waiting inside a bank before they read a
baseline scenario depicting understaffed and slow service.
Participants then rated their pre-NWOM dissatisfaction (T1-D)
toward the bank. Following a brief distractor task, participants
were instructed to imagine themselves conversing with friends
at a dinner gathering before they read a manipulation-embedded
scenario containing a NWOM message about the same bank.
Message context was manipulated with a boldfaced sentence
describing the conversation as either focusing on “some low
quality products and services on the market” (product-focused)
or “some unpleasant experiences that you and your friends had
encountered” (people-focused). Message content was manipu-
lated by presenting either an attribute-based or an experience-
based NWOM message (refer to Appendix A for the baseline
and NWOM scenarios). A pretest confirmed that the manipula-
tions were successful (see Manipulation checks).
After the manipulations, participants reported their post-
NWOM dissatisfaction (T2-D) and performed a thought-listing
task (which asked them to “write down whatever feelings or
thoughts” they had upon receipt of the NWOM message). The
dissatisfaction items were shuffled to reduce the impact of
short-term memory. Participants also rated the scenarios on
realism and comprehensibility. All the above items were
measured with 10-point Likert scales (1 = strongly disagree,
10 = strongly agree). Finally, participants provided demographic
data and tried to guess the purpose of the study (at which none
succeeded). Appendix B contains all the measures and their
reliability statistics (where applicable).
Results
Manipulation checks
A pretest (n = 80) provided the context, content, and
negativity checks. The context manipulation was assessed
with a two-item context measure capturing the extent to which
the NWOM context was people-focused versus product-focused
(such that a higher score would indicate a stronger people focus
and a weaker product focus). A 2 (message context) × 2
(message content) ANOVA on the context measure yielded
only a significant main effect of message context (MProd = 5.16,
MPeop = 7.39, F(1, 76) = 75.06, p b .001), thus validating the
context manipulation. The content manipulation was assessed
with a two-item content check measuring the salience of
consumer experience relative to product attributes in a NWOM
message (such that a higher score would indicate a stronger
emphasis on consumer experience relative to product attri-
butes). A 2 (message context) × 2 (message content) ANOVA
 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
on the content measure yielded only a significant main effect
of message content (MAttr = 4.06, MExp = 7.55, F(1, 76) = 59.59,
p b .001), which indicated that the content manipulation was
successful. Moreover, both NWOM scenarios were perceived
as more negative than the baseline scenario (ps b .001), and at
the same time equivalent in terms of negativity (F b 1). Realism
and comprehensibility ratings taken during Experiment 1 were
high (Ms N 7.30) and equivalent (Fs b 1).
Hypotheses tests
Dissatisfaction data were analyzed with a 2 (message
context) × 2 (message content) × 2 (time) repeated-measures
ANOVA. The ANOVA revealed a context × time inter-
action (F(1, 117) = 8.67, p b .01) and a content × time interaction
(F(1, 117) = 16.62, p b .001). No other effects were significant.
Since the baseline scenario induced equivalent levels of T1-D
(F b 1), these interactions were indicative of the relevance of
both message variables to NWOM influence.
Hypothesis 1 received support when the data were collapsed
across message context. As predicted by Hypothesis 1(a), there
was an increase in dissatisfaction from T1 to T2 in the attribute-
based condition (MT1-D = 7.37, MT2-D = 7.92, t = 2.19, p b .05).
Conversely, dissatisfaction decreased in the experience-based
condition (MT1-D = 7.42, MT2-D = 7.12, t = −3.38, p b .01), thus
supporting Hypothesis 1(b).
A more nuanced pattern emerged when the message-
congruent and message-incongruent contexts were analyzed
separately. Specifically, the aggravating effect of attribute-
based NWOM and the alleviating effect of experience-based
NWOM were significant in the message-congruent contexts but
not in the message-incongruent contexts. The presence of an
aggravating effect of attribute-based NWOM in a product-
focused context (MT1-D = 7.27, MT2-D = 8.15, t = 5.84, p b .001),
coupled with its absence in a people-focused context (MT1-D =
7.47, MT2-D = 7.68, NS), supported Hypothesis 3(a). Likewise,
Hypothesis 3(b) was supported by an alleviating effect of
experience-based NWOM in a people-focused context (MT1-D =
7.50, MT2-D = 6.93, t = −4.96, p b .001) and by the absence of
this effect in a product-focused context (MT1-D = 7.33, MT2-D =
7.31, NS).
Thought-listing data provided a glimpse into the underlying
processes. Two independent judges coded the verbatim data
into three categories: product-evaluative thoughts (e.g., “The
bank needs to improve its service ASAP”), social-comparative
thoughts (e.g., “I was more lucky than my friend”), and other
thoughts (e.g., “I would never wait in a long queue”). Interjudge
agreement was 94%, and discrepancies were resolved through
discussion.
In support of Hypothesis 2, attribute-based NWOM in a
product-focused context elicited more product-evaluative
thoughts (MAttr-Prod = 1.84, MExp-Peop = 1.23, F(1, 59) = 7.57,
p b .01) and fewer social-comparative thoughts (MAttr-Prod = .19,
MExp-Peop = .87, F(1, 59) = 13.45, p b .001) than experience-based
NWOM in a people-focused context. By contrast, there was no
support for Hypothesis 2 in the message-incongruent contexts
(product-evaluative thoughts: MAttr-Peop = 1.90, MExp-Prod = 1.57,
329
Table 1
Summary of mediation analysis (Experiment 1).
NWOM effect Step Dependent Independent Estimate t-statistic
variable variable
Dissatisfaction (D) 1 ΔD Message content 1.45 7.60 a
2 Thought Message content −.61 −3.91 a
type
3 ΔD Message content 1.20 5.89 a
Thought type −.42 −2.75 b
a
p b .001.
b
p b .01.
NS; social-comparative thoughts: MAttr-Peop = .50, MExp-Prod = .33,
NS).
Further analysis illuminated the mediating role of thought
type in the message-congruent contexts. To perform mediation
analysis, we constructed a thought type variable which captured
the salience of social-comparative thoughts relative to that of
product-evaluative thoughts.3 Following Baron and Kenny's
(1986) recommendations, we performed a series of regressions
involving message content (attribute-based = 1; experience-
based = 0), thought type, and change in dissatisfaction (ΔD).
We first demonstrated significant effects of message content on
ΔD and also on thought type. Then we included both message
content and thought type in the regression model for ΔD. A
significant Sobel test (Sobel's z = 2.25, p b .05) verified the
mediating role of thought type. Table 1 presents the step-by-step
mediation analysis.
Discussion
Experiment 1 establishes that NWOM from a worse-off other
may heighten or dampen dissatisfaction. The findings support
both the aggravation hypothesis for attribute-based NWOM and
the alleviation hypothesis for experience-based NWOM. In
addition, the evidence suggests that a message-congruent (vs.
message-incongruent) context strengthens the impact of
NWOM and a message-incongruent context can render
NWOM ineffective.
Notably, the augmented effects of attribute-based NWOM in
a product-focused context and experience-based NWOM in a
people-focused context provide indirect evidence for the
process hypotheses—that the aggravating effect is driven by a
product-evaluative process and the alleviating effect is driven
by a social-comparative process. The main purpose of a
message-congruent context is, however, methodological; name-
ly, to facilitate the reporting of thoughts in connection with a
NWOM message. With the help of message-congruent
contexts, the thought-listing data confirm that the two types
of NWOM trigger distinct psychological processes in line with,
respectively, attitude polarization and downward comparison
research. Mediation analysis further confirms that these
3
For each participant, thought type was 1 if the thoughts were social-
comparative, 0 if the thoughts were mixed, and −1 if the thoughts were
product-evaluative.
330 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
processes are, in turn, responsible for the contrasting NWOM
effects.
Notwithstanding the promising results, Experiment 1 is
limited in several respects. Theoretically, downward compar-
ison is a “cognitive” strategy to enhance subjective well-being
(Wills, 1981). But the experience-based NWOM scenario that
was found to elicit this cognitive strategy contains affective
details (e.g., “I was really frustrated”) as well as objective
information (e.g., “I was kept waiting for over 40 minutes”).
Although emotions are a natural part of consumer experience,
the emotional content in the scenario might have triggered
extraneous psychological processes such as emotional conta-
gion (Hatfield, Cacioppo, & Rapson, 1994).4 Besides this
potential confound, a narrow scope constitutes another
limitation of the experiment, which deals with a rather
restrictive case of NWOM influence—the influence of
NWOM from a worse-off other. The use of facilitative contexts
also weakens the generalizability of the results.
These limitations are addressed in Experiment 2, a
longitudinal study disguised as a pilot test of an investment
service. As compared with a NWOM message on waiting time
during a single bank visit, a NWOM message on investment
return over a period of time should be more conducive to the
reporting of social-comparative thoughts. This is especially the
case when the comparison involves a similar (rather than worse-
off) other. To present a more comprehensive picture of NWOM
influence, Experiment 2 also examines the opposite effects of
NWOM beyond dissatisfaction. The lateral comparison litera-
ture provides the theoretical basis for the experiment.
Lateral comparison
According to Wills (1981), enhancement of subjective well-
being can be achieved through downward comparison with
worse-off others or lateral comparison with similar others. Like
downward comparison, lateral comparison allows individuals to
gain comfort through the realization that “I'm not the only
victim” or “someone else also suffers.” Evidence for the self-
enhancing effects of lateral comparison abounds in counseling
and social work. For example, the therapeutic effects of self-
help groups (i.e., groups of individuals helping each other work
on common issues or concerns) are partly attributable to
comparison with similar others (Schiff & Bargal, 2000;
Weinberg, Uken, Schmale, & Adamek, 1995). Not surprisingly,
the finding that distressed individuals report higher subjective
well-being after discovering that others are in similar situations
4
A reviewer suggested another potential confound. Namely, since the
information in the attribute-based NWOM condition (e.g., closed counters and
slow clerks) is consistent with what is portrayed in the baseline condition, it
might be conducive to stable attributions which could lead to an aggravating
effect. But the experience-based NWOM condition also contains information
(e.g., a long wait) that is consistent with the baseline condition. So any impact
of this potential confound would likely be small. Moreover, the stability
dimension of attribution theory typically concerns whether or not the same
consumer has “stable negative experiences with the brand across times and
situations” (Laczniak et al., 2001; p. 59). This is different from the focus of our
research, which concerns how one consumer is influenced by another
consumer's NWOM.
is known as the “all in the same boat phenomenon” (Shulman,
2006).
In light of this evidence, we predict an alleviating effect for
experience-based NWOM from a similar other. This prediction
is consistent with recent findings that dissatisfied consumers do,
as a coping strategy, turn to those with similar experiences
(Duhachek, 2005). In terms of the disconfirmation model, the
rationale for this prediction lies in the lowering of the
comparison standard. By definition, dissatisfaction implies the
use of a relatively high standard. It follows that experience-
based NWOM from a similar other conveys a substandard
experience. That helps create a lower reference point, which
contributes to dissatisfaction alleviation by virtue of more
favorable disconfirmation outcomes.
In the case of attribute-based NWOM from a similar other,
we predict an aggravating effect on dissatisfaction. As
polarization research consistently demonstrates, an initial
tendency is strengthened by exposure to judgments and
evaluations similar to one's own (Abelson, 1995; Eagly &
Chaiken, 1993). In terms of the disconfirmation model,
attribute-based NWOM exerts its polarizing influence by
heightening dissatisfied consumers' negative perceptions of
product performance. A negative shift in performance percep-
tions results in dissatisfaction aggravation by virtue of less
favorable disconfirmation outcomes.
Hypothesis 4. (a) Attribute-based NWOM from a similar other
will increase dissatisfaction. (b) Experience-based NWOM
from a similar other will decrease dissatisfaction.
Repurchase intent is a natural consequence of dissatisfaction,
and an inverse relationship exists between the two variables
(Oliver, 1997). Given the hypothesized effects of NWOM on
dissatisfaction, a corresponding pair of predictions regarding
repurchase intent is in order.
Hypothesis 5. (a) Attribute-based NWOM from a similar other
will decrease repurchase intent. (b) Experience-based NWOM
from a similar other will increase repurchase intent.
As with Hypothesis 2 (which concerns NWOM from a
worse-off other), we expect to find evidence of the processes
underlying the hypothesized effects. The following is an
extension of Hypothesis 2 to NWOM from a similar other.
Hypothesis 6. Attribute-based (vs. experience-based) NWOM
from a similar other will elicit (a) more product-evaluative
thoughts and (b) fewer social-comparative thoughts.
Experiment 2
Method
Eighty one part-time MBA students (39% male) participated
in Experiment 2, which involved a 2 (message content:
attribute-based, experience-based) × 2 (time: T1, T2) repeated-
measures design. Experiment 2 was disguised as a pilot test of
 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337 331

an innovative financial service named AiAA. Participants were negative (MAttr = 7.46, MExp = 7.20, F b 1). Furthermore, emo-
recruited at the end of a lecture on market research methods. As tionality was low and equivalent (MAttr = 3.41, MExp = 3.08,
one of several “hands-on learning opportunities,” participation F b 1).
in the ostensible market research was voluntary.
The main feature of the fictitious financial service was a
state-of-the-art asset allocation technology that would generate Hypotheses tests
customized portfolios for “small investors” according to their Both dissatisfaction and purchase intent data were analyzed
personal situations and investment goals. Participants played with a 2 × 2 repeated-measures ANOVA. For dissatisfaction, a
the role of an investor who had just opened a US$15,000 content × time interaction emerged (F(1, 67) = 15.48, p b .001).
account with the company. They attended a briefing session and No other effects were significant. Given equivalent levels of
provided the information (e.g., target return, investment T1-D (F b 1), the interaction indicated that T2-D differed across
horizon, risk tolerance, etc.) required for a customized portfolio. the message conditions. As predicted by Hypothesis 4(a), there
In the four weeks that followed, participants received weekly was an increase in dissatisfaction after exposure to attribute-
email updates on portfolio valuation and return. They were based NWOM (MT1-D = 7.07, MT2-D = 7.68, t = 3.30, p b .01).
instructed to send a reply after reading each update and to abide Hypothesis 4(b) also received support, as dissatisfaction
by the no-communication rule (i.e., refrain from disclosing the decreased in the experience-based condition (MT1-D = 6.95,
“proprietary information” in the email updates). MT2-D = 6.53, t = −2.27, p b .05).
All participants received identical email updates. Weekly For purchase intent, only a content × time interaction was
returns of the purportedly “customized” portfolio were mixed: significant (F(1, 67) = 16.57, p b .001). Given equivalent levels
−2.06%, 1.45%, −2.78%, and 2.22%—mirroring but under- of T1-PI (F b 1), this indicated that T2-PI differed across the
performing the MSCI AC World Indices. There was a net loss message conditions. Supporting Hypotheses 5(a) and 5(b),
of US$187.80 (−1.25%) for the four-week period. purchase intent decreased in the attribute-based condition
The fourth and last email update included a feedback survey, (MT1-PI = 3.88, MT2-PI = 3.48, t = −2.08, p b .05) and increased in
which established both T1-D and pre-NWOM purchase intent the experience-based condition (MT1-PI = 3.75, MT2-PI = 4.31,
(T1-PI). A second survey was email-administered five days t = 3.95, p b .001).
after the first survey was returned. The second survey was As in Experiment 1, thought-listing data were coded into
ostensibly to compensate for the limitations caused by the no- three categories (interjudge agreement = 91%). There were more
communication rule. Participants were provided with the verbal product-evaluative thoughts in the attribute-based condition
feedback of “a fellow participant” whose response was than in the experience-based condition (MAttr = 1.25, MExp = .66,
“representative of quite a few people.” This constituted the F(1, 67) = 7.35, p b .01). In contrast, the number of social-
manipulation of message content: the verbal feedback was comparative thoughts was higher in the experience-based
either attribute-based or experience-based (see Appendix A). A condition than in the attribute-based condition (MAttr = .18,
thought-listing task immediately followed the content manip- MExp = .64, F(1, 67) = 10.38, p b .01). The pattern was consistent
ulation. Thereafter, measures of T2-D and post-NWOM with Hypothesis 6. As in Experiment 1, we performed further
purchase intent (T2-PI)—the same measures used for T1-D analysis on the mediating role of thought type. Mediation test
and T1-PI—were taken. The survey closed with an open-ended results showed that thought type mediated the effects of
question on the purpose of the research (to which none of the NWOM on dissatisfaction (Sobel's z = 2.57, p b .01) and
answers came close). purchase intent (Sobel's z = −2.19, p b .05). Table 2 contains
Manipulation checks were conducted in a pretest. In addition details of the mediation analysis.
to the revised content and negativity measures, there was a new
check (adapted from Moore & Homer, 2000) on the
emotionality of the two NWOM conditions. The pretest as
Table 2
well as experimental items were anchored by 1 (strongly Summary of mediation analysis (Experiment 2).
disagree) and 10 (strongly agree). Appendix B contains the
NWOM effect Step Dependent Independent Estimate t-statistic
measures and their reliability statistics (where applicable).
variable variable
Twelve participants failed to complete one or more of the
Dissatisfaction (D) 1 ΔD Message content 1.02 3.93 a
aforementioned tasks and their data were not analyzed. Of the
2 Thought Message content −.66 −3.73 a
69 participants in the final sample, 33 were in the attribute- type
based condition and 36 were in the experience-based condition. 3 ΔD Message content .64 2.41 c
Thought type −.58 −3.51 a
Results Purchase intent (PI) 1 ΔPI Message content −.95 −4.07 a
2 Thought Message content −.66 −3.73 a
type
Manipulation checks 3 ΔPI Message content −.67 −2.75 b
A pretest (n = 57) confirmed that the experience-based Thought type .41 2.71 b
message scored higher on the content measure than did the a
p b .001.
attribute-based message (MAttr = 3.18, MExp = 8.34, F(1, 55) = b
p b .01.
245.57, p b .001). Both messages were perceived as equally c
p b .05.
332 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
Discussion
In one of the few experimental studies on lateral comparison,
we show that experience-based NWOM from a similar other,
like that from a worse-off other, has an alleviating effect on
dissatisfaction. This, along with the aggravating effect of
attribute-based NWOM from a similar other, upholds the
moderating role of message content. Note that the opposite
effects of NWOM extend beyond dissatisfaction. We document
a pattern of repatronage intentions that dovetails with the pattern
of dissatisfaction ratings. The behavioral intention data confirm
the far-reaching impacts of the two distinct types of NWOM.
Moreover, we obtain clear evidence—without the help of
facilitative contexts—that attribute-based NWOM triggers a
product-evaluative process and experience-based NWOM
triggers a social-comparative process. Mediation analysis
lends further support to our theoretical explanation of NWOM
effects.
Experiment 2 also addresses a methodological issue in
Experiment 1: the lack of control for emotional content in the
NWOM conditions. With this potential confound removed, it is
gratifying that the results are consistent with all the hypothe-
sized effects and processes. Two other methodological
improvements—the longitudinal design and the real-life
facade—add credence to the robustness of our results.
Experience comparability
The provocative finding that NWOM can alleviate dissatis-
faction hinges on a social-comparative process triggered by the
salient negative experience of a worse-off or similar other.
However, consumer experiences are, by definition, driven by
product attributes (Gutman, 1982). It is natural to associate
negative consumer experiences with inferior product attributes.
Indeed, thought-listing data of the foregoing experiments
indicate that product-evaluative thoughts are prevalent even in
the experience-based NWOM condition. This raises an
interesting issue that merits further investigation: Would some
experience-based NWOM messages induce sufficient product-
evaluative thoughts to attenuate or eliminate the alleviating
effect?
So far, we have designed the experiments such that the same
attributes are responsible for the negative experience of the
dissatisfied consumer and that of the NWOM communicator.
With consumer experiences directly comparable, the social-
comparative process predominates. Nevertheless, negative
experiences such as loss of time (Experiment 1) and loss of
money (Experiment 2) can be driven by a variety of attributes.
For example, a long wait in a restaurant may be due to the wait
staff or kitchen staff, and being overcharged may be the result of
human or machine error. The comparison of consumer
experiences becomes more difficult when different attributes
are involved. In addition, the product-evaluative process may
intensify because the NWOM message now points to new
inferior attributes.
Hence, we identify experience comparability as a moderator of
the alleviating effect of experience-based NWOM. If experience
comparability is high (i.e., the same attributes drive the experience
of the dissatisfied consumer and that of the NWOM communi-
cator), we expect the social-comparative process to prevail and a
significant alleviating effect to emerge—as in the foregoing
experiments. If experience comparability is low (i.e., different
attributes drive the experience of the dissatisfied consumer and
that of the NWOM communicator), we expect the product-
evaluative process to intensify relative to the social-comparative
process. The result would be an attenuation of the alleviating
effect. The above rationale applies to downward as well as lateral
comparison. However, since Experiment 3 only concerns lateral
comparison, we state the following hypotheses in terms of
experience-based NWOM from a similar other.
Hypothesis 7. (a) When experience comparability is high,
experience-based NWOM from a similar other will decrease
dissatisfaction. (b) When experience comparability is low, the
effect of (a) will be attenuated.
Hypothesis 8. (a) When experience comparability is high,
experience-based NWOM from a similar other will increase
repurchase intent. (b) When experience comparability is low,
the effect of (a) will be attenuated.
Hypothesis 9. When experience comparability is low (vs.
high), experience-based NWOM from a similar other will elicit
(a) more product-evaluative thoughts and (b) fewer social-
comparative thoughts.
Experiment 3
Method
Fifty four undergraduates (44% male) participated in
Experiment 3 for extra credit in an introductory marketing
course. They were randomly assigned to a 2 (experience
comparability: high, low) × 2 (time: T1, T2) repeated-measures
design. The experiment was similar to Experiment 1, except for
a new set of scenarios, a longer distractor task (which lasted
approximately 20 min), and an additional dependent variable
(purchase intent). The baseline scenario depicted a 15-minute
service delay at a restaurant. Prior to the comparability
manipulation, participants were instructed to imagine them-
selves visiting an Internet forum “designed for consumers to
share their negative experiences.” The manipulation involved a
“posting about the same restaurant” from someone who
experienced a 15-minute service delay in either a high-
comparability or a low-comparability condition. Appendix A
contains all the scenarios and Appendix B contains all the
measures and their reliability statistics (where applicable).
Results
Manipulation checks
A pretest (n = 47) confirmed that the baseline scenario was
perceived as more comparable to the NWOM scenario in the
high-comparability condition than in the low-comparability
 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337 333

condition (MHigh = 8.04, MLow = 4.79, F(1, 45) = 56.50, p b .001). Table 3
Hence, the comparability manipulation was successful. More- Summary of mediation analysis (Experiment 3).
over, both NWOM conditions scored high on the content NWOM effect Step Dependent Independent Estimate t-statistic
measure (MHigh = 7.70, MLow = 7.52, F b 1) and low on the variable variable
emotionality measure (MHigh = 3.74, MLow = 3.49, F b 1). In Dissatisfaction (D) 1 ΔD Message content .68 2.68 a
terms of negativity, both were rated as similar to the baseline 2 Thought Message content −.63 −3.30 a
type
scenario (Fs b 1). Realism and comprehensibility ratings taken
3 ΔD Message content .40 1.52
during Experiment 3 were high (Ms N 7.60) and equivalent Thought type −.44 −2.49 b
(Fs b 1). Purchase intent (PI) 1 ΔPI Message content −.67 −2.43 b
2 Thought Message content −.63 −3.30 a
Hypothesis tests type
3 ΔPI Message content −.43 −1.46
A 2 × 2 repeated-measures ANOVA on dissatisfaction
Thought type .38 1.93
revealed a comparability × time interaction (F(1, 52) = 8.53, a
p b .01.
p b .01). No other effects were significant. Since the baseline b
p b .05.
scenario induced equivalent levels of T1-D (F b 1), the
interaction indicated a significant T2-D difference between
the comparability conditions. In support of Hypothesis 7(a), also experienced a 15-minute delay) does not necessarily have
dissatisfaction decreased from T1 to T2 in the high- an alleviating effect. If different attributes (e.g., reservation
comparability condition (M T1-D = 6.85, M T2-D = 6.38, t = system and order system) are responsible for the respective
−2.35, p b .05). This effect was almost reversed in the low- delays, the alleviating effect could be eliminated altogether.
comparability condition (MT1-D = 6.98, MT2-D = 7.23, t = 1.74, In the downward/lateral comparison research tradition, the
p b .10), which clearly supported the attenuation prediction of focus is on the severity of one's negative experience relative to
Hypothesis 7(b). that of another person. The theory predicts an alleviating effect
Purchase intent data were analyzed in the same way. The as a result of a favorable comparison with a worse-off or similar
2 × 2 ANOVA yielded only a comparability × time interaction other. Experiment 3 contributes to this theory by illuminating
(F(1, 52) = 5.91, p b .05). Given equivalent levels of T1-PI the role of experience comparability while holding experience
(F b 1), the interaction indicated that T2-PI differed across the severity constant. Note that we define experience comparability
comparability conditions. As predicted by Hypothesis 8(a), in terms of whether the experiences of different consumers are
purchase intent increased from T1 to T2 in the high- driven by the same attributes. Future research may explore other
comparability condition (MT1-PI = 4.04, MT2-PI = 4.52, t = 2.23, forms of comparability such as experience quantifiability (e.g.,
p b .05). Hypothesis 8(b) was also supported, with purchase loss of time or money vs. loss of comfort or esteem).
intent virtually unchanged in the low-comparability condition
(MT1-PI = 3.89, MT2-PI = 3.70, t = −1.10, NS). General discussion
As before, thought-listing data were coded into three categories
(interjudge agreement = 95%). Consistent with Hypothesis 9, the Previous research is unequivocal about the negative
low-comparability condition elicited more product-evaluative influence of NWOM (for a notable exception, see Laczniak
thoughts (MHigh = 1.44, MLow = .89, F(1, 52) = 5.51, p b .05) et al. [2001]). In four experiments, we demonstrate that the
and fewer social-comparative thoughts (MHigh = .37, MLow = .82, ramifications of NWOM are more complex than portrayed in
F(1, 52) = 6.20, p b .05) than did the high-comparability the literature. Specifically, NWOM from a worse-off or similar
condition. Mediation tests verified that thought type mediated other can have a negative (i.e., aggravating) or positive (i.e.,
the effects of NWOM on dissatisfaction (Sobel's z = 1.98, p b .05) alleviating) effect on dissatisfied consumers. Consistent with
and, to a less extent, on purchase intent (Sobel's z = −1.66, p b .10). attitude polarization research, an attribute-based message
Table 3 details the mediation analysis. triggers a product-evaluative process and results in dissatisfac-
tion aggravation. In contrast, as downward comparison research
Discussion suggests, an experience-based message triggers a social-
comparative process and results in dissatisfaction alleviation.
Experiment 3 identifies an important boundary condition for These findings are corroborated by behavioral intention data,
the alleviating effect of experience-based NWOM. The results and more important, explainable in terms of the disconfirmation
suggest that this effect is significant only under high experience model of (dis)satisfaction. On the other hand, message context
comparability and it disappears when experience comparability is and experience comparability serve as boundary conditions to
low. Thought-listing data and mediation analysis largely support these findings.
our theoretical explanation—that the product-evaluative process Our investigation calls attention to a neglected dimension of
intensifies relative to the social-comparative process under the WOM research; namely, the influence of WOM in the post-
latter condition (when the NWOM message points to new inferior consumption stage. The phenomenal growth of consumer-
attributes). generated media (e.g., discussion forums, blogs, social
To a dissatisfied restaurant customer (who experienced a 15- networking sites, etc.) has allowed consumers to easily compare
minute delay), a NWOM message from a similar other (who experiences. A cursory review of consumer discussion forums
334 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
indicates that dissatisfied consumers often go to great lengths to
describe their miserable experiences. According to ethnograph-
ic research, experiences as well as opinions are widely shared in
these forums (de Valck, van Bruggen, & Wierenga, 2009).
Experience-based stimuli are commonly used as NWOM
manipulations. Consider these NWOM fragments from prior
research: “I think he's spent more to keep [his car] running than
it originally cost him” (Herr et al., 1991), “The regular Ben and
Jerry's was so rich it made me sick” (Giese et al., 1996), and “It
just seemed like every time I boot [my computer] up I have
some kind of a problem” (Laczniak et al., 2001). Sharing these
experiences would probably deter consumers in the pre-
purchase stage. As we demonstrate, however, it could also
alleviate the dissatisfaction and increase the repurchase intent of
those in the post-consumption stage.
Arguably, the alleviating effect of experience-based NWOM is
pronounced when solid reference points or standards do not exist,
as with new technologies. Fournier and Mick's (1999) study on
technology consumption reveals that standards do emerge and
evolve over time. Our research highlights the influence of
NWOM in this evolution process. The downward comparison
explanation also resonates with Fournier and Mick's view that
coping is an integral part of technology consumption. As a simple
yet effective coping strategy, downward comparison may be
pervasive among technology consumers.
More generally, our research points to a new, psychosocial
perspective on NWOM influence. Previous research typically
conceptualizes NWOM influence in terms of product evaluations
and choice in the pre-purchase stage. In this article, we show that a
social-comparative, rather than product-evaluative, process may
predominate in the post-consumption stage, as dissatisfied
consumers find relief in the realization that someone else has a
similar or worse experience. This finding is consistent with classic
experiments that demonstrate a stress-affiliation tendency
(Schachter, 1959) and also recent studies that identify social
support seeking as a major category of consumer coping strategies
(Duhachek, 2005). In fact, Duhachek's research indicates that
consumers often make a conscious effort to manage their negative
experience. This being the case, some dissatisfied consumers may
actually seek out NWOM in order to relieve stress and anxiety.
Self-enhancing social comparison may involve a passive or an
active process (Buunk et al., 2001; Wills, 1981). Dissatisfied
consumers may take the passive approach by attending to horror
stories from other consumers. They may also create self-
enhancing opportunities by seeking out similar or worse-off
others. This is a viable strategy for self-enhancement because
“everyone is better off than someone as long as one picks the right
dimension” (Taylor et al., 1983, p. 30). In this cyber era, the
myriad consumer discussion forums and complaint sites provide
ample opportunity and an unobtrusive way for self-enhancement.
A recent ethnographic study confirms that much of online forum
behavior is indeed social in nature, driven by a variety of
motivations other than information search (de Valck et al., 2009).
Despite the self-enhancement motivation, NWOM from a
similar or worse-off other does not always result in dissatisfaction
alleviation. The alleviating effect occurs only when consumer
experience is salient in the NWOM message. An attribute-based
message, which emphasizes the product rather than the consumer,
triggers a product-evaluative process that leads to the aggravation
of dissatisfaction. In general, this finding is in line with the view
that performance perceptions are susceptible to influence
(Deighton, 1992). In particular, it meshes with research on the
biasing effects of post-consumption advertising (Braun, 1999). In
addition to Braun's memory-based account, our research suggests
that interpersonal influence is another reason why performance
perceptions can be changed retrospectively.
Overall, our research highlights the pivotal role of message
content in triggering distinct processes that lead to the
contrasting NWOM effects. To date, there has been very little
effort toward conceptualizing the content of WOM. The void is
especially evident in experimental research, which largely
manipulates WOM content in terms of valence (e.g., Giese
et al., 1996; Herr et al., 1991) and captures WOM behavior as a
probability (e.g., Frenzen & Nakamoto, 1993; Ryu & Feick,
2007). In this regard, the differentiation between attribute-based
and experience-based messages is a step toward a more nuanced
understanding of WOM communication.
It is worth mentioning that Dichter's (1966) research on
WOM motivations suggests a third dimension of message
content beyond the dichotomy proposed in this article. Apart
from “product-involvement” (which likely prompts the sharing
of attribute information) and “self-involvement” (which likely
prompts the sharing of personal experience), Dichter also
identifies “other-involvement” as an important WOM motiva-
tion. This motivation directs attention to the WOM recipient(s)
and likely prompts the sharing of purchase and consumption
advice. Such advice may be highly influential in the
marketplace, as some consumers are known to act on advice
without acquiring attribute information (Olshavsky & Granbois,
1979).
There is a growing consensus that WOM can be leveraged as
a marketing tool. Recent research, for example, offers
suggestions on how to manage online forums (Godes &
Mayzlin, 2004) and referral programs (Ryu & Feick, 2007). A
richer, multidimensional conceptualization of the WOM
construct may inform this line of research, which has abstracted
away the content of WOM (by focusing on variables such as
volume, valence, and likelihood). In particular, the main thesis
of this article—that experience-based NWOM can have an
alleviating effect on dissatisfied consumers—offers new
insights for conceptualizing WOM management. The finding
that contextual factors can strengthen or weaken the influence of
NWOM also sheds useful light on the topic.
Some limitations of our experiments point to avenues for
further research. First, the focal issues of dissatisfaction (i.e.,
waiting time and return on investment) are quantitative. This
makes for easy comparison with others. Conceivably, qualitative
issues (e.g., comfort and hygiene) could hinder the comparative
process and diminish the alleviating effect of NWOM. Second,
we contrast the pure forms of attribute-based and experience-
based NWOM. More often than not, NWOM is a mix of product
information and personal experience. Further research is needed
to determine the impacts of mixed messages. Finally, we
ascertain immediate effects of NWOM, but it is not clear how
 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
long these effects would last. Would one effect dissipate more
quickly than the other? What variables might influence the
robustness of these effects? Answers to these questions await
further research.
Acknowledgments
The authors thank Alan Wong for his assistance in literature
research and data collection, and Lisa Wan for her assistance in
data analysis. They also appreciate the guidance of the editor
and associate editor, as well as the helpful comments of two
anonymous reviewers.
Appendix A. Manipulations
A.1. Experiment 1
Baseline scenario:
You wait in line at a bank to make an important payment.
You expect to spend only a few minutes. There are a number of
people ahead of you. The service is very slow. You notice that
only 4 of the 7 counters are open. You end up waiting 20 min
before you are served.
Attribute-based NWOM condition:
You had dinner with a group of friends later that day.
[Message context manipulation] After you told them what
happened at the bank, one of your friends had this to say about
the same bank:
“They did it again. Last time many people were waiting but
the bank only had 2 of the 7 counters open. The clerks were
extremely slow. The service was really inefficient. They should
definitely hire more people and train them better!”
Experience-based NWOM condition:
You had dinner with a group of friends later that day.
[Message context manipulation] After you told them what
happened at the bank, one of your friends had this to say about
the same bank:
“You are not alone. My experience was worse. Last time
many people were waiting and I was kept waiting for over
40 minutes. I was late for an appointment because of that. I was
really frustrated. I feel terrible about this experience!”
A.2. Experiment 2
Attribute-based NWOM condition:
“The performance of AiAA is not very good. They are actually
losing money. Their technology has generated negative returns in
two of the past four weeks. The market has been a little volatile,
but their technology is supposed to produce good results.”
Experience-based NWOM condition:
“My experience with AiAA is not very good. I am actually
losing money. My investment has given me negative returns in
two of the past four weeks. The market has been a little volatile,
but I would like to see my investment increase in value.”
335
A.3. Experiment 3
Baseline scenario:
You make a dinner reservation with a new restaurant. You
arrive at the restaurant on time. Your table is not ready. You end
up waiting 15 min before you are seated.
High-comparability NWOM condition:
Later that day, you visited an Internet forum designed for
consumers to share their negative experiences. You noticed this
posting about the same restaurant:
“I had a bad experience at this new restaurant. I booked a
table for dinner one day in advance. But I had to wait like
15 minutes after I got there on time!”
Low-comparability NWOM condition:
Later that day, you visited an Internet forum designed for
consumers to share their negative experiences. You noticed this
posting about the same restaurant:
“I had a bad experience at this new restaurant. I ordered extra
food to take home when I placed my dinner order. But I had to
wait like 15 minutes after I paid the check!”
Appendix B. Measures
B.1. Experiment 1
Context check:
1. The conversation at the dinner was mainly about unpleasant
personal experiences.
2. The conversation at the dinner was mainly about low quality
products and services. (R)
Note. (R) = reverse coded; r = .75
Content check:
1. Your friend's message focused more on the bank's
performance than on personal experience. (R)
2. Your friend's message focused more on personal feelings
than on evaluations of the bank.
Note. (R) = reverse coded; r = .84
Negativity check:
1. This scenario/message describes a serious problem of the
bank.
2. This scenario/message portrays a very unpleasant consump-
tion experience.
Note. r = .79
Dissatisfaction measure:
1. I feel unhappy about my experience with the bank.
2. I am dissatisfied with the bank's performance.
3. I am not pleased with the bank's performance.
Note. α = .89
Realism check:
1. The scenarios described above are realistic.
336 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
Comprehensibility check:
1. The scenarios described above are easy to understand.
B.2. Experiment 2
Content check:
1. The focus of this message is on the person rather than on
AiAA.
2. This message emphasizes AiAA's performance over the
person's experience. (R)
Note. (R) = reverse coded; r = .92
Negativity check:
1. The tone of this message is negative.
2. This message is unfavorable to AiAA.
Note. r = .80
Emotionality check:
1. This message is filled with emotion.
2. This message contains references to specific feelings.
Note. r = .86
Dissatisfaction measure:
1. I feel happy about my experience with AiAA. (R)
2. I am satisfied with AiAA's performance. (R)
3. I am not pleased with AiAA's performance.
Note. (R) = reverse coded; α = .87
Repurchase intent measure:
1. I will consider making future investments through AiAA.
B.3. Experiment 3
Comparability check:
1. The experience described in the above scenario and the
experience described in this message are directly comparable.
Content check:
1. The focus of this message is on the person rather than on the
restaurant.
2. This message emphasizes the restaurant's performance over
the person's experience. (R)
Note. (R) = reverse coded; r = .89
Negativity check:
1. The tone of this scenario/message is negative.
2. This scenario/message is unfavorable to the restaurant.
Note. r = .85
Emotionality check:
1. This message is filled with emotion.
2. This message contains references to specific feelings.
Note. r = .82
Dissatisfaction measure:
1. I feel happy about my experience with the restaurant. (R)
2. I am satisfied with the restaurant's performance. (R)
3. I am not pleased with restaurant's performance.
Note. (R) = reverse coded; α = .91
Repurchase intent measure:
1. I will consider visiting the restaurant again.
Realism check:
1. The scenarios described above are realistic.
Comprehensibility check:
1. The scenarios described above are easy to understand.
References
Abelson, R. P. (1995). Attitude extremity. In R. E. Petty, & J. A. Krosnick
(Eds.), Attitude strength: Antecedents and consequences (pp. 25−42).
Mahwah, NJ: Lawrence Erlbaum.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychology, 51,
1173−1182.
Bearden, W. O., Netemeyer, R. G., & Teel, J. E. (1989). Measurement of
consumer susceptibility to interpersonal influence. Journal of Consumer
Research, 15, 473−481.
Bohlmann, J. D., Rosa, J. A., Bolton, R. N., & Qualls, W. J. (2006). The effect of
group interactions on satisfaction judgments: Satisfaction escalation.
Marketing Science, 25, 301−321.
Braun, K. A. (1999). Postexperience advertising effects on consumer memory.
Journal of Consumer Research, 25, 319−334.
Brown, J. J., & Reingen, P. H. (1987). Social ties and word-of-mouth referral
behavior. Journal of Consumer Research, 14, 350−362.
Buunk, A. P., & Gibbons, F. X. (2007). Social comparison: The end of a theory
and the emergence of a field. Organizational Behavior and Human Decision
Processes, 102, 3−21.
Buunk, B. P., Oldersma, F. L., & de Dreu, C. K. W. (2001). Enhancing
satisfaction through downward comparison: The role of relational discontent
and individual differences in social comparison orientation. Journal of
Experimental Social Psychology, 37, 452−467.
de Valck, K., van Bruggen, G. H., & Wierenga, B. (2009). Virtual communities:
A marketing perspective. Decision Support Systems, 47, 185−203.
Deighton, J. (1992). The consumption of performance. Journal of Consumer
Research, 19, 362−372.
Dias, L., & Lobel, M. (1997). Social comparison in medically high-risk pregnant
women. Journal of Applied Social Psychology, 27, 1629−1649.
Dichter, E. (1966). How word-of-mouth advertising works. Harvard Business
Review, 44, 147−166.
Downing, J. W., Judd, C. M., & Brauer, M. (1992). Effects of repeated
expressions on attitude extremity. Journal of Personality and Social
Psychology, 63, 17−29.
 H. Chan, S. Cui / Journal of Consumer Psychology 21 (2011) 324–337
Duhachek, A. (2005). Coping: A multidimensional, hierarchical framework of
responses to stressful consumption episodes. Journal of Consumer
Research, 32, 41−53.
Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth,
TX: Harcourt Brace College Publishers.
Festinger, L. (1954). A theory of social comparison processes. Human
Relations, 7, 117−140.
Fournier, S., & Mick, D. G. (1999). Rediscovering satisfaction. Journal of
Marketing, 63, 5−23.
Frenzen, J. K., & Nakamoto, K. (1993). Structure, cooperation, and the flow of
market information. Journal of Consumer Research, 20, 360−375.
Gibbons, F. X., & Boney McCoy, S. (1991). Self-esteem, similarity, and
reactions to active versus passive downward comparison. Journal of
Personality and Social Psychology, 60, 414−424.
Giese, J. L., Spangenberg, E. R., & Crowley, A. E. (1996). Effects of product-
specific word-of-mouth communication on product category involvement.
Marketing Letters, 7, 187−199.
Gilly, M. C., Graham, J. L., Wolfinbarger, M. F., & Yale, L. J. (1998). A dyadic
study of interpersonal information search. Journal of the Academy of
Marketing Science, 26, 83−100.
Godes, D., & Mayzlin, D. (2004). Using online conversations to study word-of-
mouth communication. Marketing Science, 23, 545−560.
Gutman, J. (1982). A means-end chain model based on consumer categorization
processes. Journal of Marketing, 46, 60−72.
Hatfield, E., Cacioppo, J. T., & Rapson, R. L. (1994). Emotional contagion.
New York: Cambridge University Press.
Helgeson, V. S., & Taylor, S. E. (1993). Social comparisons and adjustment among
cardiac patients. Journal of Applied Social Psychology, 23, 1171−1195.
Herr, P. M., Kardes, F. R., & Kim, J. (1991). Effects of word-of-mouth and
product-attribute information on persuasion: An accessibility-diagnosticity
perspective. Journal of Consumer Research, 17, 454−462.
Isenberg, D. J. (1986). Group polarization: A critical review and meta-analysis.
Journal of Personality and Social Psychology, 50, 1141−1151.
Laczniak, R. N., DeCarlo, T. E., & Ramaswami, S. N. (2001). Consumers'
responses to negative word-of-mouth communication: An attribution theory
perspective. Journal of Consumer Psychology, 11, 57−73.
Lascu, D. -N., Bearden, W. O., & Rose, R. L. (1995). Norm extremity and
interpersonal influences on consumer conformity. Journal of Business
Research, 32, 201−212.
Liu, Y. (2006). Word of mouth for movies: Its dynamics and impact on box
office revenue. Journal of Marketing, 70, 74−89.
Lord, C. G., Ross, L., & Pepper, M. R. (1979). Biased assimilation and attitude
polarization: The effects of prior theories on subsequently considered
evidence. Journal of Personality and Social Psychology, 37, 2098−2109.
Meyers-Levy, J., & Tybout, A. M. (1997). Context effects at encoding and
judgment in consumption settings: The role of cognitive resources. Journal
of Consumer Research, 24, 1−14.
Moore, D. J., & Homer, P. M. (2000). Dimensions of temperament: Affect
intensity and consumer lifestyles. Journal of Consumer Psychology, 9,
231−242.
Myers, D. G. (1978). Polarizing effects of social comparison. Journal of
Experimental Social Psychology, 14, 554−563.
Oliver, R. L. (1997). Satisfaction: A behavioral perspective on the consumer.
New York: McGraw-Hill.
337
Olshavsky, R. W., & Granbois, D. H. (1979). Consumer decision making—Fact
or fiction? Journal of Consumer Research, 6, 93−100.
Rao, V. R., & Steckel, J. H. (1991). A polarization model for describing group
preferences. Journal of Consumer Research, 18, 108−118.
Ryu, G., & Feick, L. (2007). A penny for your thoughts: Referral reward
programs and referral likelihood. Journal of Marketing, 71, 84−94.
Schachter, S. (1959). The psychology of affiliation. Palo Alto, CA: Stanford
University Press.
Schellekens, G. A. C., Verlegh, P. W. J., & Smidts, A. (2010). Language
abstraction in word of mouth. Journal of Consumer Research, 37, 224−237.
Schiff, M., & Bargal, D. (2000). Helping characteristics of self-help and support
groups: Their contribution to participants' subjective well-Being. Small
Group Research, 31, 275−304.
Schindler, R. M., & Bickart, B. (2005). Published word of mouth: Referable,
consumer-generated information on the Internet. In C. P. Haugtvedt, K. A.
Machleit, & R. F. Yalch (Eds.), Online consumer psychology: Understand-
ing and influencing consumer behavior in the virtual world (pp. 35−61).
Mahwah, NJ: Lawrence Erlbaum.
Schlosser, A. E. (2005). Posting versus lurking: Communicating in a multiple
audience context. Journal of Consumer Research, 32, 260−265.
Sengupta, J., Dahl, D. W., & Gorn, G. G. (2002). Misrepresentation in the
consumer context. Journal of Consumer Psychology, 12, 69−79.
Shulman, L. (2006). The skills of helping individuals, families, groups, and
communities (5th ed). Belmont, CA: Thompson.
Sunstein, C. R. (2002). The law of group polarization. Journal of Political
Philosophy, 10, 175−195.
Taylor, S. E., Wood, J. V., & Lichtman, R. R. (1983). It could be worse: Selective
evaluation as a response to victimization. Journal of Social Issues, 39, 19−40.
Tesser, A. (1978). Self-generated attitude change. In L. Berkowitz (Ed.),
Advances in experimental social psychology, Vol. 11. (pp. 289−338)
New York: Academic Press.
Tormala, Z. L., & Clarkson, J. J. (2007). Assimilation and contrast in
persuasion: The effects of source credibility in multiple message situations.
Personality and Social Psychology Bulletin, 33, 559−571.
Ward, J. C., & Reingen, P. H. (1990). Sociocognitive analysis of group decision
making among consumers. Journal of Consumer Research, 17, 245−262.
Weinberg, N., Uken, J. S., Schmale, J., & Adamek, M. (1995). Therapeutic
factors: Their presence in a computer-mediated support group. Social Work
with Groups, 18, 57−69.
Westbrook, R. A. (1987). Product/consumption-based affective responses and
post-purchase processes. Journal of Marketing Research, 24, 258−270.
Westbrook, R. A., & Oliver, R. L. (1991). The dimensionality of consumption
emotion patterns and consumer satisfaction. Journal of Consumer Research,
18, 84−91.
Wheat, R. (2010). The empowered shopper.May/June: The Hub Magazine
42–46.
Wills, T. A. (1981). Downward comparison principles in social psychology.
Psychological Bulletin, 90, 245−271.
Woodruff, R. B., Cadotte, E. R., & Jenkins, R. L. (1983). Modeling consumer
satisfaction processes using experience-based norms. Journal of Marketing
Research, 20, 296−304.
Zhou, R., & Soman, D. (2003). Looking back: Exploring the psychology of
queuing and the effect of the number of people behind. Journal of Consumer
Research, 29, 517−530.