Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
454
© 1991 by JOURNAL OF CONSUMER RESEARCH. Inc. • Vol. 17»March 1991
All rights reserved. 0O9J-53OI/9l/l7O4.O0O2$O2.0O
EFFECTS OF WORD OF MOUTH 455
TABLE 1
Results
OVERALL EVALUATIONS AS A FUNCTION OF PROCESSING
OBJECTIVES. ATTRIBUTE VALENCE, ANO ANECDOTE Effects of Prior Impressions on Judgmental Vivid-
VALENCE (EXPERIMENT 2) ness Effects. Overall product evaluations as a func-
tion of processing goals, attribute valence, and WOM
Memory set Impression set valence are presented in Table 1. To test the hypothesis
that vivid WOM communications should have a re-
Attribute Positive Negative Positive Negative
valence WOM WOM WOM WOM
duced effect on product judgments when a prior
impression ofthe product is available (vs. not avail-
Positive 9.10, 8.20^ 8.33.6 7.38c able) from memory (Hypothesis 2), a series of planned
Neutral 6.44. 5.49b 5.41b 4.71b contrasts was performed on overall product judg-
Negative 1.45. 1.85. 1.89. 2.21. ments.^ Because prior impressions are less likely to be
available in memory set than in impression set con-
NOTE. Higher scores indicate more favorable overall evaluations. Means in
the same row not sharing a common subscript differ at p < .05. n = 10 per cell ditions (Hastie and Park 1986; Kardes 1986; Lichten-
(except in the lower right entry, for which n = 9). stein and Srull 1985, 1987), WOM effects on judgment
should be weaker in impression set than in memory
set conditions. Consistent with this prediction, in pos-
itive-attribute conditions, a significant WOM effect was
(memory set). The description contained a brief sum- obtained in the memory set condition (X = 9.10 vs.
mary of standard features, a ranking (the brand was 8.20, p < .05). However, unexpectedly, a significant
ranked sixth, twentieth, or thirty-fourth best ofthe 40 WOM effect was also found in the impression set con-
brands tested), and information about nine important dition {X = 8.33 vs. 7.38, p < .05). Thus, partial sup-
attributes (engine performance, transmission, han- port for Hypothesis 2 was found in positive-attribute
dling, braking, ride, noise, displays, seating, and reli- conditions.
ability). The attributes were described favorably (e.g.,
braking was smooth), neutrally (e.g., braking was usu- A different pattern of results was found when neutral
ally smooth), or unfavorably (e.g., braking was abrupt). attributes were presented. In the memory set condition,
more favorable judgments were formed in positive-
Subjects participated in small groups of four to six.
than in negative-WOM conditions (X = 6.44 vs. 5.49,
After subjects had read the description, the confederate p < .05). Although a robust WOM effect on judgment
said that the car seemed similar to a car her father was obtained in the memory set condition, no WOM
owns. In positive WOM conditions she added, "It's effect was found in the impression set condition {X
the best car he's ever had. He hasn't spent a dime on = 5.41 vs. 4.71, NS). Thus, consistent support for Hy-
repairs since he bought it. He says if it ever wears out pothesis 2 was found in neutral attribute conditions.
he'll get another just like it." In negative WOM con-
ditions she said, "It's the worst car he's ever had. It EffTects of Negative-Attribute Information on Judg-
seems like it's always in the shop being repaired. I think mental Vividness Effects. To test the hypothesis that
he's spent more to keep it running than it originally the presence of extremely negative attribute infor-
cost him." The experimenter asked her to write the mation reduces the effect of anecdotal WOM com-
brand name of the car on a slip of paper, said it was munications on judgment (Hypothesis 3), planned
the described brand, and thanked and dismissed her. comparisons were performed separately in memory set
and in impression set conditions. Although robust
WOM effects on judgment were found when positive
Judgment and Recall Measures
'Planned comparisons should be performed instead of overall F-
After dismissal of the confederate, subjects were tests when interactions involving differences between specific cells
asked to recall as much attribute information as pos- are predicted (Hays 1981; Keppel 1982; Kirk 1982). Nevertheless,
sible within a three-minute period, rate each attribute the reader may be interested in the results of omnibus F-tests. A 2
(impression or memory set) X 3 (positive, neutral, or negative at-
on scales ranging from 0 (extremely low quality) to 10 tributes) X 2 (positive or negative WOM) X 2 (counterbalancing
(extremely high quality), and judge the overall quality order) between-subjects analysis of variance was performed on
of the target product. Finally, subjects were thanked overall evaluations. Counterbalancing order produced no significant
and debriefed. main effects or interactions. Main effects were found for processing
goal (F(l,95) = 6.97,/>< .01), attribute valence (F(2,95) = 474.44.
Two judges coded the recall protocols using a gist p < .001), and WOM valence (F(l,95) = 7.90, p < .01). More fa-
criterion, and high interrater reliability was attained vorable evaluations were formed in memory set than in impression
(97 percent). Attribute ratings were averaged to form set conditions, and more favorable evaluations were formed as at-
a single brand attitude index (Cronbach's alpha = .98, tribute and WOM information increased in favorability. Significant
processing goal by attribute valence (f(2,95) = 5.92, p < .01) and
p < .001). This index correlated highly with the overall attribute valence by WOM valence (/=^(2,95) = 5 . 8 l , p < .01) inter-
quality measure (r = .94, p < .001). actions were also found.
EFFECTS OF WORD OF MOUTH 459
or neutral attributes were presented to memory set was obtained (r = -.70, p < .05), whereas, in the neg-
subjects, the WOM effect was eliminated when ex- ative-WOM cell, the judgment-recall correlation was
tremely negative attributes were presented to memory not significant (p > .20). Thus, in the positive-WOM
set subjects (X = 1.45 vs. 1.85, NS). Similarly, in cell, as the number of neutral attributes that could be
impression set conditions, no WOM effect was found recalled increased, less favorable overall judgments
when extremely negative attributes were presented were formed. In the negative-WOM cell, relatively
(X = 1.89 vs. 2.21, NS). Thus, consistent support unfavorable overall judgments were formed regardless
was found for Hypothesis 3 across processing-goal of the number of neutral attributes available from
(impression vs. memory set) conditions. memory.
Judgment-Recall Correlations/Positive-Attribute Judgment-Recall Correlations/Negative-Attribute
Conditions. Judgment-recall correlations were ex- Conditions. Finally, in negative-attribute conditions,
amined to facilitate interpretation of the unexpected nonsignificant judgment-recall correlations were found
WOM effect on judgment obtained in the positive at- across impression/memory set conditions {p > .20)
tribute-impression set cell. Prior research has shown and across WOM-valence conditions {p > .20). These
that an impression set induces subjects to engage in results imply that on-line judgments were formed in
on-line processing and to form separate evaluation- both impression and memory set conditions. This
based and attribute-based representations in memory finding is consistent with previous research demon-
(Hastie and Park 1986; Kardes 1986; Lichtenstein and strating that extremely negative information can
Srull 1985, 1987). Because independent representa-
tions are formed, a lack of correspondence between prompt spontaneous judgment formation (Bettman
judgment and recall is likely to be observed. In con- and Weitz 1983; Weiner 1985).
trast, a memory set prompts subjects to form an attri-
bute-based representation in memory (no evaluation- Discussion
based representation is formed). When memory set Although WOM communications were found to
subjects are subsequently asked to evaluate a product, have a strong impact on product judgments in exper-
a judgment is .computed on the basis ofinformation iment 1, this effect was reduced when a well-defined
that is retrieved directly from the attribute-based rep- prior impression was available from memory or when
resentation. Consequently, a strong judgment-recall extremely negative attribute information was encoun-
correlation is likely to be observed. tered. Unexpectedly, a WOM effect on judgment was
found when positive attributes were presented to
Consistent with this model, significant judgment- impression set subjects. Analyses of the relationship
recall correlations were found in memory set condi- between judgment and recall, however, suggest that
tions (r = .51 and .59 for positive- and negative-WOM these subjects failed to form a clear, well-defined
conditions, respectively, p < .05). Moreover, in impression ofthe target product in the negative-WOM
impression set conditions, a nonsignificant relation- condition, presumably because the uncertainty created
ship between judgment and recall was found in the by the attribute valence-WOM valence inconsistency
positive-WOM cell (r = .26, NS). However, in the prompted a judgmental shift toward moderation
impression set-negative WOM cell, a significant cor- (Cialdini et al. 1973; Jaccard and Wood 1988).
relation between judgment and recall was obtained (r
= .64,> < .05). This unexpected finding implies that When extremely negative attributes were presented,
judgments formed in this cell were influenced signif- vivid WOM communications had no effect on product
icantly by memory for attributes. However, the un- judgments. Because extremely negative information is
certainty created by the inconsistency between the at- perceived as highly diagnostic (experiment 1) and be-
tribute versus the WOM information may have cause diagnostic information can reduce the judg-
prompted subjects to adjust their judgments toward a mental impact of less diagnostic but more accessible
more moderate position. This type ofjudgmental shift inputs, judgmental vividness effects were eliminated
is observed frequently under conditions of uncertainty completely when extremely negative attribute infor-
because a moderate stance is justified easily and is al- mation was provided.
tered readily as additional information becomes avail-
able (Cialdini et al. 1973; Jaccard and Wood 1988). GENERAL DISCUSSION
Judgment-Recall Correlations/Neutral-Attribute Consistent with the accessibility-diagnosticity model
Conditions. As expected, nonsignificant judgment- (Feldman and Lynch 1988; Lynch et al. 1988), the re-
recall correlations were found in impression set con- sults suggest that information accessibility mediates
ditions {p > .20). Unexpectedly, in memory set con- the effects of WOM information on product judgments
ditions, judgment-recall correlations differed as a (experiment 1). However, information-accessibility ef-
function of WOM valence. In the positive-WOM cell, fects on judgment are reduced, when more diagnostic
an inverse relationship between judgment and recall information, such as prior impressions or extremely
460 JOURNAL OF CONSUMER RESEARCH
negative attribute information, is available (experi- processes induce consumers to overestimate the valid-
ment 2). Thus, information accessibility mediates the ity of their prior impressions.
effects of vividly presented information on judgment, In consumer settings, negative information tends to
and the perceived diagnosticity of other available in- be more diagnostic or informative than positive or
puts moderates these effects. neutral information. Negative attributes strongly im-
The results also imply that judgmental vividness ef- ply membership in one category (i.e., low quality) to
fects are not as elusive as implied by previous research the exclusion of others, whereas positive or neutral
(Taylor and Thompson 1982). Robust vividness effects attributes are more ambiguous with respect to category
on judgment were obtained in two experiments em- membership. Positive and neutral features are asso-
ploying two different product classes. Consistent with ciated with many high-, medium-, and low-quality
the accessibility-diagnosticity model, the results sug- products. Negative features, on the other hand, have
gest that judgmental vividness effects are not elusive stronger implications for categorization. Even when
when (1) the judgment context provides the oppor- many positive features are exhibited (e.g., the soup has
tunity for differential encoding of vivid versus pallid fresh meat, fresh Grade-A potatoes, and fresh vita-
information, (2) prior impressions are unavailable min-rich vegetables), a single extremely negative fea-
from memory, and (3) information having extremely ture (e.g., the broth is rancid) can be highly informa-
negative judgmental implications is unavailable for tive. Consequently, negative-attribute information is
processing. Many ofthe studies included in Taylor and weighed heavily in judgment.
Thompson's (1982) highly cited review failed to meet In conclusion, WOM communications often have a
these criteria, and, consequently, judgmental vividness strong impact on product judgments because infor-
effects were unlikely to be observed. mation received in a face-to-face manner is more ac-
In addition to shedding light on the vividness con- cessible than information presented in a less vivid
troversy, the accessibility-diagnosticity model has the manner. Although WOM information is highly acces-
capacity to synthesize and integrate a wide variety of sible from memory, its itnpact on judgment is reduced
seemingly disparate judgmental phenomena. For ex- when more diagnostic information is available. Prior
ample, the results of our experiments suggest that the impressions are easy to interpret, and they influence
vividness effect (Kisielius and Sternthal 1984, 1986), encoding and retrieval processes in a manner that re-
the perseverance effect (Lord et al. 1979, 1984), and duces the correspondence between objective and per-
the negativity effect (Skowronski and Carlston 1987, ceived cue diagnosticity. Consequently, prior beliefs
1989) can be explained through the accessibility-di- (as opposed to other possible beliefs) are held with a
agnosticity model. high degree of confidence. Similarly, negative infor-
Experiment 1 demonstrated that WOM communi- mation tends to suggest one categorization over others.
cations have a greater impact on product judgments Information that strongly implies one hypothesis, in-
than less vivid printed infbrmation even when infor- terpretation, or categorization over other possibilities
mation content is held constant. Although the manner is weighed heavily in judgment, regardless of whether
in which information is presented does not affect its alternative possibilities are improbable or are simply
probative value, vividly presented information has a overlooked.
disproportionate effect on judgment. Furthermore,
consumers are likely to overestimate the diagnosticity
of accessible information because alternative inputs [Received June 1990. Revised August 1990.]
are likely to be overlooked.
Using the same methodology but a different product REFERENCES
class, experiment 2 showed that the judgmental effects Arndt, Johan (1967), "Role of Product-related Conversa-
of accessible WOM information are reduced when tions in the Diffusion of a New Product," Journal of
other more diagnostic pieces ofinformation are avail- Marketing Research, 4 (August), 291-295.
able. Consumers trust their own opinions more than Baron, Reuben M. and David A. Kenny (1986), "The Mod-
they trust the opinions of others (Hoch and Deighton erator-Mediator Variable Distinction in Social Psycho-
1989). Other individuals are likely to exhibit knowl- logical Research: Conceptual, Strategic, and Statistical
edge or reporting biases (see Wood and Eagly 1981), Considerations," Journal of Personality and Social
and, consequently, information provided by other Psychology, 51 (December), 1173-1182.
sources is often ambiguous (i.e., multiple interpreta- Bearden, William O. and Michael J. Etzel (1982), "Reference
Group Influence on Product and Brand Purchase De-
tions are possible). In contrast, self-generated infor- cisions," Journal of Consumer Research, 9 (September),
mation is less likely to be contaminated by knowledge 183-194.
or reporting biases (although, under certain circum- , Richard G. Netemeyer, and Jesse E. Teel (1989),
stances, self-deception has been observed; see, e.g., "Measurement of Consumer Susceptibility to Interper-
Berglas and Jones 1978; Gur and Sackeim 1979). sonal Influence," Journal of Consumer Research, 15
Moreover, selective encoding and selective retrieval (March), 473-481.
EFFECTS OF WORD OF MOUTH 461
Berglas, Steven and Edward E. Jones (1978), "Drug Choice toward Behavioral Alternatives," Journal of Personality
as a Self-handicapping Strategy in Response to Non- and Social Psychology. 54 (April), 580-591.
contingent Success," Journal of Personality and Social Kardes, Frank R. (1986), "Effects of Initial Product Judg-
Psychology, 36 (April). 405-417. ments on Subsequent Memory-based Judgments,"
Bettman, James R. and Barton A. Weitz (1983), "Attribu- Journal of Consumer Research, 13 (June), I - l l .
tions in the Boardroom: Causal Reasoning in Corporate Keppel, Geoffry (1982), Design and Analysis: A Researcher's
Annual Reports," Administrative Science Quarterly, 28 Handbook, Englewood Cliffs, NJ: Prentice-Hall.
(June), 165-183. Kirk, Roger E. (1982), Experimental Design: Procedures
Biehal, Gabriel and Dipanker Chakravarti (1986), "Con- for the Behavioral Sciences, Monterey, CA: Brooks/
sumers' Use of Memory and External Information in Cole.
Choice: Macro and Micro Perspectives," Journal of Kisielius, Jolita and Brian Sternthal (1984), "Detecting and
Consumer Research, 12 (March), 382-405. Explaining Vividness Effects in Attitudinal Judgments,"
Borgida, Eugene and Richard E. Nisbett (1977), "The Dif- Journal of Marketing Research, 21 (February), 54-64.
ferential Impact of Abstract vs. Concrete Information and Brian Sternthal (1986), "Examining the Vividness
on Decisions," Journal of Applied Social Psychology, 7 Controversy: An Availability-Valence Interpretation,"
(July-August), 258-271. Journal of Consumer Research, 12 (March), 418-431.
Brown, Jacqueline Johnson and Peter H. Reingen (1987), Lichtenstein, Meryl and Thomas K. Srull (1985), "Conceptual
"Social Ties and Word-of-Mouth Referral Behavior," and Methodological Issues in Examining the Relationship
Journal of Consumer Research, 14 (December), 350- between Consumer Memory and Judgment," in Psycho-
362. logical Processes and Advertising Effects: Theory, Research,
Cialdini, Robert B., Alan Levy, C. Peter Herman, and Scott and Application, ed. Linda F. Alwitt and Andrew A. Mitch-
Evenbeck (1973), "Attitudinal Politics: The Strategy of ell, Hillsdale, NJ: Erlbaum, 113-128.
Moderation," Journal of Personality and Social Psy- and Thomas K. Srull (1987), "Processing Objectives as
chology, 25 (January), 100-108. a Determinant of the Relationship between Recall and
Engel, James E., Roger D. Blackwell, and Robert J. Kegerreis Judgment," Journal of Experimental Social Psychology, 23
(1969), "How Information Is Used to Adopt an Inno- (March), 93-118.
vation," Journal ofAdvertising Research, 9 (December), Lord, Charles G.. Mark R. Lepper, and Elizabeth Preston
3-8. (1984), "Considering the Opposite: A Corrective Strategy
Feldman, Jack M. and John G. Lynch (1988), "Self-gener- for Social Judgment," Journal of Personality and Social
ated Validity and Other Effects of Measurement on Be- Psychology. 47 (December), 1231-1243.
lief, Attitude, Intention, and Behavior," Journal ofAp- , Lee Ross, and Mark R. Lepper (1979), "Biased Assim-
plied Psychology, 73 (August), 421-435. ilation and Attitude Polarization: The Effects of Prior The-
Feldman, Sidney P. and Merlin C. Spencer (1965), "The ories on Subsequently Considered Evidence," Journal of
Effect of Personal Influence in the Selection of Con- Personality and Social Psychology, 37 (November), 2098-
sumer Services," in Proceedings ofthe Fall Conference 2109.
of the American Marketing Association, ed. Peter D. Lynch, John G., Howard Marmorstein, and Michael F. Weigold
Bennett, Chicago: American Marketing Association, (1988), "Choices from Sets Including Remembered Brands:
440-452. Use of Recalled Attributes and Prior Overall Evaluations,"
Fischhoff, Baruch and Ruth Beyth-Marom (1983), "Hy- Journal of Consumer Research, 15 (September), 169-184.
pothesis Evaluation from a Bayesian Perspective," Psy- McGill, Ann L. and Punam Anand (1989), "The Effect of Vivid
chological Review, 90 (July), 239-260. Attributes on the Evaluation of Alternatives: The Role of
Gur, Ruben C. and Harold A. Sackeim (1979), "Self-De- Differential Attention and Cognitive Elaboration," Journal
ception: A Concept in Search ofa Phenomenon," Jour- of Consumer Research, 16 (September), 188-196.
nal of Personality and Social Psychology. 37 (February),
Mizerski, Richard W. (1982), "An Attribution Explanation of
147-169.
the Disproportionate Influence of Unfavorable Informa-
Hastie, Reid and Bernadette Park (1986), "The Relationship
tion," yoi/wfl/o/Co/jra/Mer/{eiearc/i, 9 (December), 301-
between Memory and Judgment Depends on Whether
310.
the Judgment Task Is Memory-based or On-Line," Psy-
chological Review, 93 (July), 258-268. Nisbett, Richard and Lee Ross (1980), Human Inference: Strat-
Hays, William L. (1981), Statistics, New York: Holt, Rine- egies and Shortcomings of Social Judgment, Englewood
hart & Winston. Cliffs, NJ: Prentice-Hall.
Higgins, E. Tory and Gillian King(1981), "Accessibility of Price, Linda L., Lawrence F. Feick, and Robin A. Higie (1989),
Social Constructs: Information-processing Conse- "Preference Heterogeneity and Coorientation as Determi-
quences of Individual and Contextual Variability," in nants of Perceived Informational Inttuence" Journal of
Personality, Cognition, and Social Interaction, ed. Business Research, 19 (November), 227-242.
Nancy Cantor and John F. Kihistrom, Hillsdale, NJ: Reingen, Peter H. and Jerome B. Kernan (1986), "Analysis
Erlbaum, 69-121. of Referral Networks in Marketing: Methods and Illustra-
Hoch, Stephen J. and John Deighton (1989), "Managing tion," Journal of Marketing Research, 23 (November),
What Consumers Learn from Experience," Journal of 370-378.
Marketing, 53 (April), 1-20. Richins, Marsha L. (1983), "Negative Word-of-Mouth by Dis-
Jaccard, James and Gregory Wood (1988), "The Effects of satisfied Consumers: A Pilot Study," Journal ofMarketing.
Incomplete Information on the Formation of Attitudes 47 (Winter), 68-78.
462 JOURNAL OF CONSUMER RESEARCH
Skowronski, John J. and Donal E. Carlston (1987), "Social Weiner, Bernard (1985), "'Spontaneous' Causal Thinking,"
Judgment and Social Memory: The Role of Cue Diagnos- Psychological Bulletin, 97 (January), 74-84.
ticity in Negativity, Positivity, and Extremity Biases," Jour- Wood, Wendy and Alice H. Eagly (1981), "Stages in the Analysis
nal of Personality and Social Psychology, 52 (April), 689- of Persuasive Messages: The Role of Causal Attributions
699. and Message Comprehension," Journal of Personality and
and Donal E. Carlston (1989), "Negativity and Ex- Social Psychology, 40 (February), 246-259.
tremity Biases in Impression Formation: A Review of Wright, Peter (1974), "The Harassed Decision Maker: Time
Explanations," Psychological Bulletin, 105 (January), Pressures, Distractions, and the Use of Evidence," Journal
131-142. of Applied Psychology, 59 (October), 555-561.
Taylor, Shelley E. and Suzanne C. Thompson (1982), "Stalking Wyer, Robert S. and Thomas K. Srull (1986), "Human Cog-
the Elusive 'Vividness" Effect," Psychological Review, 89 nition in Its Social Context," Psychological Review, 93
(March), 155-181. (July), 322-359.