Content uploaded by Eric R. Spangenberg
Author content
All content in this area was uploaded by Eric R. Spangenberg on Dec 15, 2014
Content may be subject to copyright.
Marketing Letters 7:2 (1996): 187-199
© 1996 Kluwer Academic Publishers, Manu&ctured in the Netherlands
Effects of Product-Specific Word-of-Mouth
Communication on Product Category Involvement
JOAN L. GIESE
Department (^Marketing, «bshmgton Slate University, Pullman,
WA
99164-4730
ERIC R. SPANGENBERG
Department of Marketing,
Wiishington
State University, Pullman,
WA
99164-4730
AYN E. CROWLEY
Department of Marketing, Drake University, Des Moines, IA 50311
Key words: word-of-mouth, involvement, product category
Abstract
Drawing primarily from categorization theoiy, this paper presents justification for the effects of word-of-mouth
(WOM) communication on product category involvement. Results of an empirical test of this relationship are
presented showing an enduring effect of positive WOM communication on product category involvement; this
effect was not found for negative WOM. These results suggest that positive WOM about one firm's brand may
help competitors t^ increasing involvement, thus generating more sales (not necessarily of one's own brand) in
an entire product category. Our findings, coupled with categorization theory, provide support for a series of
prop-
ositions presented concerning the effects of changes in product category involvement on nondiscussed brand atti-
tudes and purchase intentions.
As perhaps the most influential source of consumer information, the study of
word-of-
mouth (hereinafter WOM) communication holds much potential for consumer researchers
and practitioners. Although not extensively researched, studies examining the importance
of WOM suggest it is a key &ctor in consumer decision making (e.g., Leonard-Barton,
1985;
Price and Feick, 1984; Richins, 1983), and it has substantial influence on product
evaluations and purchase decisions (Brown and Reingen, 1987; Price and Feick, 1984).
WOM communication, for example, has been shown to have greater impact on product
evaluations than printed information (Boigida and Nisbett, 1977; Herr, Kardes, and Kim,
1991).
WOM may indeed play a key role in consumer decision making, but existing research
leaves many unanswered questions regarding this role.
In marketing practice, proactive marketing strategies recognizing and taking into con-
sideration the effects of WOM communication are commonly encouraged. WOM com-
munications are perceived as so critical to success (or gdlure) that service firms are en-
couraged to proactively manage WOM communications (e.g., Zeithaml, I^rasuraman, and
Berry, 1985). WiUde (1994) outlined four strategies for WOM: (1) discourage unfavorable
WOM (e.g., dealing with rumors), (2) create favorable WOM (e.g., influentials' stimula-
tion, (3) stimulate additional direct sales through WOM (referrals), and (4) simulate WOM
in advertising (e.g., slice-of-life ad technique). While frequently accepted and employed
188
J.L.
GIESE,
E.R.
SPANGENBERG AND
A.E.
CROWLEY
by many marketers, these strategies are not well justified in the literature. The effects of
any of the above outlined strategies have not been fully identified, nor have they been em-
pirically verified through research. Thus, the wisdom of incorporating WOM in marketing
strategies, and how to develop this kind of strategy, is questionable until scholarly investiga-
tion provides clarification.
An important issue focuses on how WOM infonnation ultimately affects product evalua-
tions and purchase decisions. Exploratory findings suggest that WOM communication m^
indirectly affect sales by influencing consumers' level of involvement with the product
category. Indeed, with source credibility as a covariate, involvement with resume writing
software has been found to be significantly different as a result of the WOM condition
to which subjects were exposed (Spangenberg, Giese, and Crowley, 1994). This signifi-
cant interaction between valence of WOM information and involvement suggested that in-
volvement may be something other than a covariate or moderating variable as it is com-
monly modeled. Specifically, Spangenbeig, Giese, and Crowl^ (1994) found that positive
(negative) WOM infonnation about a specific product increased (decreased) consumers'
involvement with an entire product category. If this is the case, negative information mj^
not alw^ result in only negative affective responses toward the mentioned brand (cf. Mizer-
sld, 1982) but would result in weaker affective responses toward aU brands in a product
category. If this e^ct does generalize to a product category, the implications for managers
could be significant in strategy development.
The above situation would most likely occur when consumers are not highly involved-
conditions when WOM information would be most effective because consumers' product
belief are not strongly held and consumers view brands as being less differentiated
(Z^chkowsky, 1986). Conversely, this situation may not apply to low-involvement con-
sumers "because they may not believe the communication would have personally relevant
effects" (Zaichkowsky, 1986, p. 6). Thus, moderate involvement categories represent the
most likely condition where product-specific WOM information would generalize across
the product category. Moderate involvement categories, however, have been all but ignored
in involvement research (see Andrews, Durvasula, and Akhter, 1990). This is ultimately
important because of the potential impact on the many product categories for which con-
sumers do not exhibit extreme (low or high) levels of involvement.
This paper presents an experimental test of theoretically driven hypotheses regarding
the effects of positive and negative WOM information on a factor seldom considered as
a dependent variable—product category involvement. Because involvement as a dependent
variable is unaddressed in the literature, and word-of-mouth communication is relatively
unexplored and difficult to manipulate, we designed a simple yet innovative study to test
these relationships. Our somewhat surprising results raise questions concerning conven-
tional beliefs regarding the effects of WOM communication; therefore, propositions for
future research are developed.
1.
Conceptual background
Zaichkowsky (1985, p. 342) defines product involvement as "a person's perceived relevance
of the object based on inherent needs, values, and interests." Personal relevance is based
EFFECTS OF PRODUCT-SPECmC WORD-OF-MOUTH COMMUNICATION
189
on the antecedents of involvement (Andrews, Durvasula, and Akhter, 1990) such as in-
herent needs, values, and interests (Zaichkowksy, 1985) evidenced by a person's knowledge,
experience,
and
cognitive structure (Celsi
and
Olson, 1988) regarding the product category.
Thus,
by definition, high-involvement consumers have strongly held needs, values, and
interests; conversely, low-involvement consumers have weakly held (or nonexistent) needs,
values, and interests. As a consequence, involvement
plays
a role in determining consumers'
attention and comprehension processes (Celsi and Olson, 1988). High-involvement con-
sumers are more motivated to process messages, while low-involvement consumers lack
the motivation to process messages and, instead, consider peripheral cues (such as source
likeability) in
forming
attitudes. Furthermore, if consumers process messages more exten-
sively, attitude change tends to be more enduring (Petty, Cacioppo, and Schumann, 1983).
When the personal relevance of
a
message is moderate or ambiguous, as we would ex-
pect under moderate levels of product category involvement, the source may influence the
extent of message processing (Petty and Cacioppo, 1986). Subsequently, when consumers
do not have well-defined brand beliefe, a personal information source is more persuasive
in product evaluations than a nonpersonal source like advertising (Herr, Kardes, and Kim,
1991).
Thus, when consumers are neither high nor low with regard to their level of involve-
ment, the message source serves as a stimulus to encourage more extensive message pro-
cessing (similar to high-involvement consumers) rather than, or in addition to, serving as
a peripheral cue. Unlike consumers with high product category involvement, however, these
moderately involved consumers are likely to be relatively less knowledgeable (Higie and
Feick, 1989) and have more basic cognitive structures (Sujan and E>ekleva, 1987).
Consideration of a person's cognitive structure
regarding
a product category (that is, how
consumers categorize information), may shed insight on the nature of involvement. Bartlett
(1932) contended
that
individuals organize their worlds according to
"chunks"
of knowledge.
Rosch (1975) found that people group these chunks of knowledge into categories based
on perceived similarities. Thus, an object is placed in a particular category if the object
shares certain features with other objects in the category (that is, the object is typical of
objects in the category). Furthermore, objects are categorized at the level that is most cog-
nitively efficient
and
at which the information value of the categories is maximized (Rosch
et al., 1976)—the most basic level. What constitutes the basic level varies by individual,
depending (at least partially) on the level of expertise held by the individual. As individuals
develop increased knowledge about a particular topic, categories implemented to process
information become less basic and more specific (Mervis and Rosch, 1981).
Consistent with the literature on categorization, Sujan and Dekleva (1987) indicate that
product type is most likely to be the basic level of categorization for most product offer-
ings because of the perception of many shared attributes. This suggests that consumers
view various brands of a particular product type as having sinular attributes and thus
categorize them together. This would be the case particularly for individuals who are rela-
tively less knowledgeable about a particular domain (cf., Fiske, Kinder, and Larter, 1983).
Fbr example, consumers with limited knowledge about ice cream would most likely use
"premium ice cream" as a level of categorization rather than "Haagen Dazs."
Because of the dynamic nature of memory, schemas or scripts (knowledge structures)
are constantly modified to incorporate new knowledge, thus enhancing learning (Schank,
1982).
This categorization system in memory
facilitates
generalization of knowledge learned
190
J.L.
GIESE,
E.R.
SPANGENBERG AND A.E. CROWLEY
in one context to a separate, connected context as they are activated. For example, Schank
(1982) described the memory structure of the more general "fest ft)od" schema by indexing
Burger King so that when entering McDonald's for the first time, the customer is reminded
of Burger King. This occurs because the
t>vo
fest food outlets share many attributes. There-
fore, one would expect that if a consumer had limited knowledge about a product category,
he or she would organize brands by product type, subsequently generalizing new informa-
tion about a particular brand to other brands in the product category.
A lack of distinction within product categories is commonly associated with lower prod-
uct involvement compared to high-involvement situations in which consumers clearly dif-
ferentiate between alternative brands. Under the low-involvement scenario, brands in a prod-
uct category would be perceived as nondifferentiated, acceptable substitutes (Zaichkowsky,
1986).
Kbigaonkar and Moschis (1982) use this notion of differentiation between alternative
brands to distinguish between high-involvement and low-involvement product categories,
finding that subjects were less suscq>tible to negative infonnation if products were classified
as high involvement. This rationale suggests that consumers under high involvement hold
firm beliefs about product attributes and are influenced only by strong quality arguments,
whereas under lower involvement, beliefe are not strongly held and hence are more easily
influenced. This conclusion suggests that product-specific WOM information (or any prod-
uct information for that matter) may affect perceptions of an entire product category where
consumers have organized brands according to a product type schema. This is most likely to
occur when perceived differences among brand alternatives are smaD or nonexistent, perhaps
due to relatively low levels of knowledge or involvement concerning the product category.
As mentioned above, Higie and Feick (1989) indicate that consumers with high involve-
ment are likely to be more knowledgeable about a respective product category. Much earlier,
Howard and Sheth (1969) considered involvement with products to lead to greater percep-
tion of difference between attributes—that is, the more knowledgeable or involved con-
sumers are, the more able or motivated they are to detect differences between attributes
of brands within product categories. While knowledge is not interchangeable with involve-
ment, it is reasonable to expect that the implications for categorization structures would
be comparable for the two constructs. This could explain how brand-specific information
would generalize to the entire product category under some conditions. Thus, as a conse-
quence of more extensive message processing within a basic cognitive structure, an in-
crease or decrease in interest of a mentioned brand could result in increased or decreased
consumer interest in the entire product category. By definition, this increase (decrease)
in interest would positively (negatively) influence consumer involvement with the product
category. Furthermore, the change in involvement should be relatively enduring because
of the central processing approach (Petty, Cacioppo, and Schumann, 1983).
2.
Hypotheses
Given the empirical and conceptual support described above, we expect levels of involve-
ment for a product category, not merely cognitive and affective perceptions of a specific
brand within the category, to be affected by product-specific WOM communication. Thus,
under initial conditions of moderate involvement, we hypothesize the following:
EFFECTS OF PRODUCT-SPECinC WORD-OF-MOUTH COMMUNICATION 191
Hypothesis 1: Brand-specific negative
WOM
injbrnuaion will significantly decrease pwd-
uct category involvement.
Hypothesis 2: Brand-specific positive WOM information will significantly increase
prod-
uct category involvement.
Hypothesis 3: These changes in product category will be enduring.
3.
Method
3.1.
Design and subjects
A between-subjects design with a control group was used. The treatment factor consisted
of negative WOM infonnation, positive WOM information, and no infonnation. Eighty-
two subjects from an undergraduate course at a large northwest university participated in
the experiment as part of their regular coursework and for extra credit. An initial screen-
ing survey and the subsequent experiment coincided with discussions of marketing research
in the course; the experiment was therefore perceived as part of their course and not a
research project.
3.2. Materials and measures
3.2.L Videotape. Each of the three treatment groups watched a fifteen-minute videotaped
focus group. The focus group was videotaped earlier using typical college students discussing
their activities, interests, and opinions so that a hypothetical company could make a deter-
mination about how to segment their markets for premium ice cream products. No brand-
specific information was discussed in the videotaped session.
3.2.2.
Product category. A pretest using seventy undergraduate students was conducted
to identify an appropriate product category. Of the several product categories tested, premium
ice cream was considered most appropriate because pretest subjects indicated generally
moderate to high levels of femiliarity (86 percent indicated that they were either "somewhat
femiliar" or "quite femiliar") and showed large relative variance in product category in-
volvement. Familiarity was used to choose the product category as a sufficient level of
the construct was necessary to insure that the manipulation would be processed (Andrews,
Durvasula, and Akhter, 1990). We also sought a category with variance in involvement
in our student population; if a majority of subjects were highly involved with the product
category, our hypotheses could not have been tested.
3.2.3.
Independent variable. Two levels of valence (positive or negative) of WOM infor-
mation were manipulated in the experimental sessions. The confederate student made one
of two statements in respective sessions. The control group received no WOM informa-
tion. Treatment manipulation statements were as follows:
192
J.L.
GIESE,
E.R.
SPANGENBERG AND
A.E.
CROWLEY
• Cell I (positive): "The other day a friend and I tried both the regular and the low-fet
versions of Ben and Jerry's ice cream. We couldn't believe how good they both tasted.
They were so creamy. We were really impressed."
• Cell 2 (negative): "The other day a friend and I tried both the regular and the low-fiit
versions of Ben and Jerry's ice cream. The regular Ben and Jerry's was so rich it made
me sick, but the low fet version had no taste. We were both disappointed."
• Cell 3 (control): No statement was made.
3.2.4. Dependent variable. Product category involvement was used as the main criterion
variable in this study. Pre- and post-manipulation, and two-week delay product category
involvement measures were taken using the ten-item revised version of Zaichkowsky's (1985)
Personal Involvement Inventory (PII) (Zaichkowsky, 1994).
3.2.5.
Potential covariates. WOM information transfer is an interpersonal form of com-
munication. As such, a measure of susceptibility to interpersonal influence (Bearden,
Netemeyer, and Teel, 1989) was included as a potential covariate; this construct did not
account for a significant amount of variance in any of our tests, and therefore we do not
discuss the construct hereafter. Pretest results suggested that level of femiliarity varies with
level of involvement; whether this is a result of, or results in, greater involvement is not
known. This theorization led to inclusion of femiliarity as a potential covariate which is
discussed further below.
3.3. Procedure
During the first week of class sessions, subjects completed a survey to determine baseUne
involvement scores for several product categories including the focal product category of
premium ice cream; five subjects were dropped at this point because they had no femiliar-
ity and extremely low pretest involvement with the target product category. Subjects were
asked to volunteer for one of three sessions as role-playing participants in a focus group.
In each condition, four round tables seating twelve to thirteen people were used. In oider
to involve subjects in the experiment and give the "feel" of a focus group as much as possible,
each subject received a can of soda and a name card. Because subjects were assuming
the role of focus group participants, they were asked not to talk with other classmates be-
tween sessions so that otfier participants could attend the session as fiee firom bias as possible.
To further minimize potential demand artifacts, all sessions were run in the same after-
noon of one day; sessions were also spaced one hour apart to limit interaction between
arriving and departing subjects.
Each of the three treatment groups watched the same fifteen-minute videotaped focus
group, ostensibly providing students with insights concerning focus group participation
and fecilitation. No evaluative statements were made concerning product categories or brands
within product categories on the videotape or in the treatment sessions other than the
manipulation.
Immediately following each viewing of the videotape, the moderator/instructor asked
for comments. In the two treatment groups, a paid student confederate made a single
EFFECTS OF PRODUCT-SPECIFIC WORD-OF-MOUTH COMMUNICATION 193
comment (negative or positive, respectively) concerning a specific brmui of premium ice
cream. The conditions were assigned randomly to time slots prior to the experiment and
were blind to the moderator until the comment was made. After the spoken manipulation,
but before any other comments could be made, the moderator politely thanked the con-
federate (by the name "Jennifer"; her real name was not shared with anyone in the par-
ticipating classes) for the comment, explaining, however, that the purpose of the special
session was to understand the techniques of focus group interviewing not
to
make evaluative
judgments. A ten-minute structured lecture concerning focus group interviewing and a short
quiz followed. The quiz reinforced the connection of the special session with the regular
course and also provided a manipulation check to determine the level of attentiveness to
the session. Fbllowing the experiment, another experimenter entered the room (and the
moderator left) to collect data for an "unrelated research study." The dependent measures
were imbedded in this survey. The same procedure was used for the control group; the
confederate was in attendance but did not speak. The confederate was seated in the same
position in the room in all three conditions.
As a measure of the lasting effect of our manipulation, approximately two weeks later
an experimenter posing as a
representative
from a local grocery store came into each class-
room and asked for subjects' participation. He indicated that marketing researchers at the
university had assisted in the design of
this
survey to help a local grocer better understand
college students' buying behavior. This survey included involvement scales for several prod-
uct categories including the target product category, ongoing search measures, attitudes,
and purchase behavior measures. As a manipulation check, after completing this survey,
students were asked by a separate researcher whether they
remembered
"Jennifer," 53 per-
cent of exposed subjects recalled her presence in the sessions. An earlier manipulation
check would have more appropriately demonstrated the effectiveness of
the
manipulation;
however, we could not risk hypothesis guessing prior to collecting our two-week delay
measure of involvement.
The experimental procedure described above was pretested prior to the actual experi-
ment using a different but demographically similar group of college students. This pretest
group was debriefed in an attempt to detect any demand effects that could result in bias;
none of the pretest subjects detected the presence of the confederate or guessed our
hypotheses.
In every case, the
relevant
measurements
regarding
the target product category were im-
bedded in several measures
regarding
distractor product categories (such as credit cards,
discount stores, and laundry detergent). This was to disguise the product category of in-
terest thus minimizing potential demand artifacts or hypothesis guessing.
4.
Results
The main tests of significance were conducted as analyses of covariance in
the
three-condition
(WOM: positive, negative, control) experiment using the immediate posttest and two-week
delay measures of involvement as criterion variables and the pretest measure of involve-
ment as covariate (cf., Cronbach and Furby, 1970; Peter, Churehill, and Brown, 1993).
Familiarity was included as a significant covariate in the analysis. Because the data collection
194 J.L. GIESE, E.R. SPANGENBERG AND A.E. CROWLEY
sessions were separated by time, experimenter, and location, there is a difference in number
of subjects at each stage (that is, some subjects did not show up for every session).
Figure
1
shows adjusted mean values for pretest, immediate posttest, and two-week delay
product category involvement. There was no statistically significant difference between
the negative WOM and control conditions for both the immediate and delayed measures
(both F's < 1). Thus, Hjrpothesis 1 was not supported: there was no effect of negative
WOM communication as com])ared to the control condition. H)fpothesis 2 was supported:
there was a statistically significant immediate increase in product category involvement
for subjects exposed to positive WOM communication compared to subjects in the control
and negative WOM conditions (F(l, 81) = 4.72, p = .033). The immediate effects on
involvement also held over time; for the two-week delayed product category involvement
measure, the positive WOM condition was significantly greater than the negative WOM
and control conditions (F(l, 66) = 4.10, p = .047) which did not differ statistically at
the delay. Thus, Hypothesis 3 was supported for the positive WOM condition.
5.
Discussion
This study demonstrates that product category involvement is increased by positive product-
specific WOM information, but not decreased by negative product-specific WOM infor-
mation. A major implication of this finding is that positive WOM about your brand may
47-r
46 -
45 -
44 -
43-
42 -
41 -- 41.1{+)
40-
39
46.0445.5
-Positive (N>i34)
-Negative (N=28)
-Control (N>20)
39.85
H h
Pretest lmme<fiate Delayed
Posttest Posttest
Figure 1. Product category involvement by WOM communication condition.
EFFECTS OF PRODUCT-SPECmC WORD-OF-MOUTH COMMUNICATION 195
help your competition by increasing involvement level, thus generating more sales (not
necessarily of your brand) in the entire product category. That is, if product category in-
volvement is increased by positive, product-specific, WOM communication, the ultimate
purehase made by consumers m^ not necessarily be the brand about which the positive
WOM infonnation was advanced. In other words, a nondiscussed brand may be piuchased.
From a different perspective, if involvement increases within a product category to the
extent that consumers differentiate more carefully between brands, the finding that prod-
uct category involvement is affected by WOM information is of further import to practi-
tioners, particularly in a market with many customers locked in a routine problem-solving
(RPS) framework. A WOM campaign may generate greater consumer involvement within
a product category, resulting in more infonnation-seeking and thus better-informed con-
sumers. If we have a better product than that of the competition, we may capture the more
involved consumers from competitors' previously uninvolved loyal RPS purchasers; however,
this also suggests the counterintuitive notion that positive WOM could be ultimately damaging
to market share for some firms. Further research should be conducted to determine how
WOM can influence variables lite brand-specific purchase intention and purchase frequency.
Our unexpected result regarding negative WOM informadon may not be overly surpris-
ing. This result coincides with that of Ray and Wilkie (1970) regarding advertising fear
appeals: if a given ai)peal or message is too strong or negative, people m^ avoid the message
or mentally "write off' the recommendation being made in the message. Thus, the kind
of information presented may partially account for our results. Furthermore, incongruity
between the very positive overall context and the very negative infonnation could have elicited
these results. Mizerski (1982) suggests that an attribution threshold operates in situations
where "the expectancy of favorable information is particularly strong, [thus,] un&vorable
information would not be viewed as credible" (p. 308). It is interesting to note that, in
a study replicating Petty, Cacioppo, and Schumann's (1983) test of their elaboration like-
lihood model, negative nonverbal behaviors disrupted argument processing (Huddleston,
1985).
This implies that negative communication effects are possibly more complex than
positive communication effects, and it is not uncommon to obtain unanticipated negative
commimication results. Future research is warranted to address negative communication
processing.
Our results are consistent with the findings of Sujan and Dekleva (1987) indicating that
product type is most likely to be the basic level of categorization for product offerings because
of the perception of many shared attributes (product category is likely more cognitively
efficient than brand-level categorization). Although our data do not directly address the
issue, and what constitutes a basic level of categorization varies by individual, our findings
are not counter to the idea that some product categories are more or less differentiated
than others. In general, this study supports the expectation that consumers with limited
knowledge about a product cat^ory organize brands by product type, subsequently generaliz-
ing new information about a specific brand to the product category.
As with any study, this one has limitations. Most important, the study focuses on only
one brand and one product category. Thus, the generalizability of our conclusions across
product categories and settings cannot yet be determined. Given this limitation, further
empirical evidence is necessary before we can generally conclude that positive and negative
WOM communication will have the effects we have found in other contexts.
196
J.L.
GIESE,
E.R.
SPANGENBERG AND
A.E.
CROWLEY
6, Propositions for future research
Research on WOM communication, its definition, measurement, manipulation, and subse-
quent effects on various dependent variables has been sporadic and generally unfocused
as a stream of research. This is due, in part, to special challenges regarding effective
manipulation of WOM in the laboratory, while maintaining an acceptable level of external
validity. Little generalizable research has emerged and methods of study are diverse and
unstandardized. More specifically, further research should be conducted to determine how
WOM can influence variables like brand-specific purchase intention and frequency. Based
on our findings and relevant theoretical relationships, we have developed a set of proposi-
tions to guide future research.
6.1.
Context propositions
Motivating consumers to process messages is an essential first step in persuasion (Petty
and Cacioppo, 1986). Contextual fectors may infiuence the extent to which consumers per-
form this first step. Two contextual issues are presented based on relevant research. First,
although little research has directly addressed the issue. Petty and Cacioppo (1986) suggest
that when consumers are moderately involved, the source mjry affect their motivation to
process the message. Research has shown that WOM communication has a greater impact
on product evaluations than printed infonnation (Borgida and Nisbett, 1977; Herr, Kaides,
and Kim, 1991). Thus, we propose that when a consumer is moderately involved with
a product category, a personal source providing WOM information would be more infiuen-
tial on consumers' motivation to process messages than nonpersonal sources, such as adver-
tising. In cases where consumers experience an "attribution threshold" (Mizerski, 1982),
consumers would be less motivated to process the WOM message:
Proposition 1: Under moderate involvement, consumers are more motivated to process
messages when the source is personal rather than nonpersonal.
Proposition 2: Consumers are less motivated to process WOM messages when the valence
of the context is incongruent with the valence of the message.
6.2. Brand attitudes and purchase intention propositions
Categorization theory (discussed in Section 1) would suggest that under conditions where
message recipients do not have a detailed cognitive structure, which would be the case
under moderate levels of involvement (cf. Celsi and Olson, 1988), consumers would tend
to generalize brand-specific information to the entire product category because that is most
cognitively efficient and infonnation value is maximized (Rosch et al., 1976).
Although not empirically tested, anecdotal evidence supports this contention. In 1991,
bottled water sales were growing at less than 1 percent, well below the expected double-
digit growth trend. Some industry analysts blamed the overall economic recession, while
EFFECTS OF PRODUCT-SPECIHC WORD-OF-MOUTH COMMUNICATION 197
"other bottled water analysts attribute the cat^ory's woes in part to industry leader Perrier's
1990 recall. Negative publicity and broken habits may have spurred some bottled water
loyalists to fall away from the fold" (Beverage
World,
1992-1993, p. 16). This is consis-
tent with our contention that information about a single brand can affect an endre product
category. Thus, we propose examining brand-specific WOM information's impact on all
brands in a product category:
Proposition 3: If consumers do not perceive attribute differences between brands,
brand-
specific word-of-mouth information affects consumer's attitudes toward all brands in the
prodtict category.
Proposition 4: If consumers do not perceive attribute differences between brands,
brand-
specific word-of-mouth information affects purchase intentions of all brands in the product
category.
Proposition 5: If cotisumers perceive attribute differences between brands, brand-specific
word-of-mouth information affects only consumers' attitude toward the disctissed
brand.
Proposition 6: If consumers perceive attribute differences between brands, bnmd-specific
word-qf-mouth information affects only consumers' purchase intention of the discussed
brand.
7. Conclusion
The main contribution of this article is a balanced theoretical justification for the effects
of WOM communication on product category involvement and a set of testable proposi-
tions focusing on two heretofore relatively unexamined topics—WOM communication and
moderate levels of product category involvement. WOM communication research is im-
portant and relevant to marketing researchers and managers. The results of our study show
an effect of positive WOM communication on product category involvement. Furthermore,
our results combined with categorization theory provide the support to make propositions
concerning the effects of changes in product category involvement on nondiscussed brand
attitudes and purchase intentions. Given the potential impact on product sales, this paper
calls for future research to address these important issues. It is our hope that our work
will motivate and &cilitate more clearly directed, programmatic research addressing both
theoretical and normative questions regarding WOM communication.
References
Andrews, J. Craig, Srinivas Durvasula. and Syed H. Akhter. (1990). "A Framework fbr Conceptualizing and
Measuring the Involvement Constnict in Advertising Research." Joumai of Advertising 19(4), 27-40.
Bartlett, F.C. (1932). Remembering: A Study in Experimental and Social Psychology. Cambridge: Cambridge
University Press.
198 J.L. GIESE, E.R. SPANGENBERG AND A.E. CROWLEY
Beaiden, miliam O., Richard G. Netemeyer, and Jesse E. Teel. (1989). "Measurement of Consumer Suscept-
ibility to Interpeisonai Influence." Journal of Consumer Research IS (March), 473-481.
Bevemge Wbrld Databank. (1992-1993). "The Beverage Market Index fbr 1992: BotUed \V&ter." Great Neck,
NY: Keller International Publishing Corporation.
Borgida, Eugene and Richard E. Nisbett. (1977). "The Differential Impact of Abstract vs. Concrete Information
on Decisions." Journal of Applied Social Psychology 7 (July-August), 258-271.
Brown, Jacqueline Johnson and Peter H. Reingen. (1987). "Social lies and Word-of-Mouth Referral Behavior."
Journal of Consumer Research 14 (December), 350-362.
Celsi, Richard L. and Jerry C. Olson. (1988). "The Role of Involvement in Attention and Comprehension Pro-
cesses." Journal of Consumer Research 15 (September), 210-224.
Cionbach, Lee J. and Lita Furby. (1970). "How We Should Measure 'Change' —or Should We?" Psychological
Bulletin 74 (July), 68-80.
Fiske, Susan T, Donald R. Kinder, and W.M. Larter. (1983). "The Novice and the Expert: Knowledge-Based
Strategies in B>litical Cognition." Journal of Experimental Social Psychology 19, 381-400.
Herr, Paul M., Frank R. Kardes, and John Kim. (1991). "Effects of W)rd-of-Mouth and Product-Attribute Infor-
mation on Persuasion: An Accessibility-Diagnosticity Perepective."
Journal
of Consumer Research 17 (March),
454-462.
Higie, Robin A. and Lawrence F. Feick. (1989). "Enduring Involvement: Conceptual and Measurement Issues."
Advances in Consumer Research 16, 690-696.
Howard, John A. and Jagdish N. Sheth. (1969). The Theory cf Buyer Behavior. New York: John Wiley.
Huddleston, B.M. (1985). "An Experimental Investigation of the Influence Deceptive Nonverbal Cues Exert on
Persuasive Processes." Doctoral dissertation, Utiiversity of Missouri.
Korgaonkar, P.K. and Goerge P. Moschis. (1982). "An Experimental Study of Cognitive Dissonance, Product
Involvement, Expectations, Performance and Consumer Judgment of Product Performance." Journal of Adver-
tising 11(3), 32-44.
Leonard-Baiton, Dorothy. (1985). "Experts as Negative Opinion Leaders in the Diffusion of a Ibchnological
Innovation." Journal of Consumer Research 11 (March), 914-926.
Mervis, Carolyn B. and Eleanor Rosch. (1981). "Categorization of Natural Objects."
Annual
Review of Psychology
32,
89-115.
Mizerski, Richard W. (1982). "An Attribution Explanation of the Disproportionate Influence of Unfavorable In-
formation." Journal of Consumer Research 9 (December), 301-310.
Peter, J. Paul, Gilbert A. Churchill, Jr., and Tbm J. Brown. (1993). "Caution in the Use of Difference Scores
in Consumer Research." Journal of Consumer Research 19 (March), 655-662.
Petty, Richard E. and John T. Cacioppo. (1986). Communication and Persuasion: Central and Peripheral Routes
to Attitude Change. New \brk: Springer-Verlag.
Petty, Richard E., John T. Cacioppo, and David Schumann. (1983). "Central and Peripheral Routes to Advertis-
ing Effectiveness: The Moderating Role of Involvement." Journal of Consumer Research 10, 134-148.
Price, Linda L. and Lawrence F. Feick. (1984). "The Role of Interpeisonal Sources in External Search: An
Informational Perspective." Advances in Consumer Research 11. 250-255.
Ray, Michael L. and William L. Wilkie. (1970). "Fear: The Potential of an Appeal Neglected by Marketing."
Journal of Marketing 34 (January), 54-62.
Richins, Marshal L. (1983). "Negative ^brd-of-Mouth by Dissatisfied Consumers: A Pilot Study." Journal of
Marketing 47 (Winter), 68-78.
Rosch, Eleanor. (1975). "Cognitive Representations of Semantic Categories." Journal of Experimental
Psychology:
General 104, 192-223.
Rosch, Eleanor, et al. (1976). "Basic Objects in Natural Categories." Cognitive Psychology 8, 382-439.
Schank, Roger C. (1982). Dynamic Memory: A Theory of Reminding and Learning in Computers and People.
New York: Cambridge University Press.
Spangenbeig, Eric R., Joan L. Giese, and Ayn E. Crowley. (1994). "WDrd-of-Mouth Communication: Some
New Issues." Working paper, ^tehington State University.
Sujan, Mita, and Christine Dekleva. (1987). "Product Categorization and Inference Making: Some Implications
for Comparative Advertising." Journal of Consumer Research 14 (December), 372-378.
Wilkie, William L. (1994). Consumer Behavior. New York: John WUey.
EFFECTS OF PRODUCTSPECIFIC WORD-OF-MOUTH COMMUNICATION 199
Zaichkowsky, Judith Lynne. (1985). "Measuring the Involvement Construct." Joumal
<tf
Consumer Research
12 (December), 341-352.
Zaichowsky, Judith Lynne. (1986). "Conceptualizing Involvement." Journal of/Advertising 15(2), 4-14.
Zaichowsky, Judith Lynne. (1994). "The Peisonal Involvement Inventory: Reduction, Revision and Application
to Advertising." Joumal of Advertising 23(4), (December), 58-70.
Zeithaml, \Uarie A., A. Paiasutaman, and Leonard L. Berry. (1985). "Problems and Strategies in Services
Marketing." Joumal cf Marketing 49 (Spring), 33-46.
