Product Categorization and Inference Making-1987

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Journal of Consumer Research, Inc.

Product Categorization and Inference Making: Some Implications for Comparative Advertising
Author(s): Mita Sujan and Christine Dekleva
Source: Journal of Consumer Research, Vol. 14, No. 3 (Dec., 1987), pp. 372-378
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/2489498 .
Accessed: 18/08/2013 22:33

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR to
digitize, preserve and extend access to Journal of Consumer Research.

http://www.jstor.org

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions
Product Categorization and Inference
Making: Some Implications for
Comparative Advertising

MITASUJAN
CHRISTINEDEKLEVA*

A categorization approach to inference making was used to determine when the


effects of comparative advertising would differ from those of noncomparative ad-
vertising. Comparative advertising led to different responses for experts relative to
either product class or product type level noncomparative advertising, but for novices
only when the noncomparative ad referredto a broad product class. These differences
were for measures of similarity,distinctiveness, and ad informativeness. Mean values
on brand attitudes were not differentially influenced by comparative versus non-
comparative advertising, though comparative advertising resulted in more polarized
attitudes. These results were interpreted within the categorization framework and
point to the general usefulness of this conceptual framework.

Studies that have examined the effects of compara- noncomparative advertising as a result of differences in
tive advertising have found that, in general, com- inferences generated.
parative advertising is no different from noncompara-
tive advertising (e.g., Belch 1981; Goodwin and Etgar A CATEGORIZATION APPROACH TO
1980). Indeed, the evidence has not indicated when or INFERENCE MAKING
on what measures comparative ads will differ from
noncomparative ads. One problem for earlier work in One fundamental premise of the categorization ap-
comparative advertising has been the lack of a useful proach is that objects can be grouped at varying levels
conceptual framework. Though cognitive response the- of specificity. For example, products could be grouped
ory (Greenwald 1968) has been applied, it has not been into superordinate or product class categories such as
particularly useful in predicting differences between "cars," into less inclusive product type categories such
comparative and noncomparative advertising (cf. Gorn as "sports cars," or into specific brand level categories
and Weinberg 1984). We employ categorization theory such as "Nissan 300ZX cars." Categorization theory
(e.g., Mervis and Rosch 1981) to investigate this issue. also specifies that there is one basic level of inclusiveness
The approach is useful in predicting the inferences that at which individuals naturally categorize and sponta-
consumers make, i.e., the attributes that consumers as- neously name objects. The basic level maximizes par-
cribe to a product even when these are not explicitly simony-a few "rich" and "distinct" categories are
mentioned in the advertising. We derive propositions formed. Richness refers to the number of attributes de-
about the likely differences between comparative and scribing the category, and distinctiveness refers to how
differentiated the category is from other categories at
the same level. For products, the product type level is
likely to constitute the basic level of categorization. For
*Mita Sujanis AssistantProfessorof Marketing,Collegeof Busi-
ness Administration,The PennsylvaniaStateUniversity,University example, for the category of cars, sports cars and family
Park, PA 16802. ChristineDekleva was a graduatestudent in the cars are likely to be perceived as distinct subcategories,
MarketingDepartment,The PennsylvaniaState University, at the but various brands of sports cars are likely to be seen
time this researchwas conducted,and her Master'sthesis was part as having many shared attributes and few distinct ones.
of this researcheffort.Jim Bettman'scomments on this projectare Categorization ideas further suggest that few infer-
gratefullyacknowledged.
ences can be made about a product by categorizing it

372
? JOURNAL OF CONSUMERRESEARCHe Vol. 14 C December 1987

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions
CATEGORIZATIONAND INFERENCE MAKING 373

at the product class level (e.g., a camera can best be to a noncomparative ad if the product is presented
described in terms of "takes pictures," "uses film," etc.). within the broad product class (e.g., as a car). However,
Categorization at the product type level allows many a comparative ad may not be very different from a non-
more inferences to be drawn, and these inferences are comparative ad that positions the product at the basic
also likely to be evaluative in character (e.g., a 35mm product type level (e.g., as a sports car).
camera can be described as "takes good pictures," "uses
a large film format," "versatile," etc.). Further, the in-
ferences that can be made about contrasting product Moderating Effects of Category Expertise
types are likely to be distinct (e.g., the attributes ascribed The categorization approach suggests that increas-
to a 110 camera-"handy to carry," "cartridge film"- ingly specific categories (e.g., brand level categories) are
are likely to be different from those ascribed to a 35mm utilized by individuals to process information with in-
camera). Finally, categorizing at the brand level rather creasing knowledge about the domain (Mervis and
than at the product type level is likely to result in some Rosch 1981). Thus experts, but not novices, might ben-
increase in the attributes that can be used to describe efit from brand level categories in the number and
the product (e.g., a Canon AE-1, in addition to being quality of inferences that can be made. Hence, for ex-
described as "takes good pictures," and "versatile," can perts, but not novices, comparative brand level ads may
also be described as "Japanese made," and "reliable"), also differ from noncomparative product type ads.
but this increase is likely to be small (e.g., many of the The general finding that comparative advertising is
inferences that can be made about the Canon AE- 1 can no different from noncomparative advertising (e.g.,
be made based on its membership in the 35mm camera Belch 1981; Golden 1979) is not surprising when viewed
category). within the categorization framework. These studies
To summarize, the categorization research suggests generally use product type cues and categories in which
that more attributes can be inferred about a product by there is little scope for expertise (e.g., fluoride tooth-
categorizing it at more specific levels and that the in- pastes). However, as argued earlier, differences can be
crease is greater in moving from the product class to expected between comparative and noncomparative
the product type level than from the product type to advertising when the product is presented within a broad
the brand level. product class or the advertising is aimed at expert con-
sumers. The purpose of this article is to test these prop-
ositions.
Implications for Comparativeversus
NoncomparativeAdvertising STUDY 1
In considering comparative advertising, we examine Before the effects of inference making on comparative
comparative ads for unknown or new brands that ex- advertising can be addressed, it is important to verify
plicitly name a competing brand and contain little in- the predictions made by categorization theory in the
formation beyond the general comparison made. This product context. Thus, the purpose of Study 1 was to
distinction between comparative and noncomparative examine the nature of inferences at the product class,
ads can be viewed in terms of category level. Noncom- product type, and brand levels.
parative advertising may position the new brand as a The procedure used to search for basic level categories
member of a broad product class (e.g., Renault 181 . in the object domain was adapted from Sujan (1985b).
a truly fine automobile), or a more specific product type Seven product classes were selected (cars, cameras, tape
(e.g., Renault 181 . . . a remarkable sports car). Com- recorders, video cassette recorders, electronic games,
parative advertising can present cues at the most specific tennis racquets, and foods). Three product types were
level by comparing the advocated brand to a well-known identified for each product class, and two familiar
brand already present in the consumer's memory struc- brands were identified for each product type. Thus, 10
ture (e.g., Renault 181 . . . comparable to the Nissan categories were generated for each product.
300ZX). Given that the advertised brand is relatively One hundred twenty students participated in the
unknown, any inferences formed about attributes not study. Each subject listed characteristic attributes for
mentioned will depend on the cues in the ad, and one one category from each of the seven products included
important cue is the category level at which the product in the study. The procedure produced 12 attribute lists
is positioned. Further, since categorization theory sug- for each product-category combination, with each list
gest that inferences will increase from the product class generated by a different subject. Attributes generated
to the product type level and level off from the product by two or more subjects were counted as descriptive of
type to the brand level, it can be predicted that the effects the category. These lists of inferences served as the units
of providing a cue at the brand level (e.g., comparing of analyses. Since the number of observations was dif-
it to a Nissan 300ZX) will be more pronounced relative ferent at the product class (7), product type (21), and

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions
374 THE JOURNAL OF CONSUMER RESEARCH

brand (42) levels, the observations at the product type STUDY 2


and brand levels were averaged so that a comparison
between the product class, the "average" product type Overview
in the product class, and the "average" brand in the
product class, or product type (as appropriate), could The purpose of Study 2 was to examine the effects of
be made. varying product cue level in the context of comparative
advertising. An understanding of the differences in in-
Results ferences at various levels of a product hierarchy was
used to predict not only contingencies under which dif-
Richness of Inferences. The attributes at the more ferences in comparative and noncomparative ads can
specific level that were not mentioned at the higher be expected, but also outcome measures on which these
level(s) under which they were nested were counted for differences are likely to occur.
each product. For example, for cars, "rapid accelera-
tion" was counted as an additional attribute at the Ad Informativeness. Study 1 demonstrated that the
product type level whereas "has wheels" was not, as it increase in inferences was large between the product
was also listed at the product class level. Seven obser- class and product type level and leveled off between the
vations were obtained at each level, one for each prod- product type and brand level. Though not investigated
uct. For the sample of products used, the gain in attri- in Study 1, categorization research also suggests that
butes was greater in moving from the product class to the increase in inferences between the product type and
the product type level than in moving from the product brand level will be exaggerated for experts. This pattern
type to the brand level (6.5 vs. 2.6, t(12) = 2.9, p < 0.01). for inferences should be mirrored by perceptions of
The Evaluative Nature of Inferences. An inference perceived ad informativeness. Thus, based on inferences
was counted as being evaluative if it was judged as car- made, comparative ads will be perceived as more in-
rying some positive or negative evaluation. Usually an formative than noncomparative product class ads for
evaluative inference was an attribute qualified by a de- all consumers. For experts, but not for novices, com-
scriptor (e.g., "rapid acceleration"). Two judges, blind parative ads are also likely to be perceived as more in-
to the hypotheses, rated the inferences (agreement formative than noncomparative product type ads.
= 0.93). Seven observations were obtained at each level,
Brand Attitudes. More specific inferences may result
one for each product. The proportion of evaluative in-
in a more extreme positive or negative evaluation. Since
ferences was lower at the product class level and not
inferences may be positive or negative depending upon
different between the product type and brand levels ( 14
other factors (e.g., idiosyncratic prior experience), the
percent vs. 78 percent vs. 83 percent; F(2, 18) = 22.78,
richer, more evaluative inferences generated by more
p < 0.01). There was also a larger gain in evaluative
specific cues may elicit more polarized evaluations from
inferences in moving from the product class to the
each consumer, but not necessarily different evaluations
product type level than in moving from the product
(due to the mix of positive and negative evaluations)
type to the brand level (6.2 vs. 2.6, t( 12) = 2.2, p < 0.05).
when averaged across consumers. Since arguments
Distinctiveness of Inferences. Similarity across in- similar to those made earlier for ad informativeness can
ference lists was defined as the ratio of shared to non- be made here, the expected pattern of results for these
shared attributes. Twenty-one observations were ob- measures will be the same as that for ad informativeness.
tained at each level, three for each of the seven product
classes. Comparison of the similarity between product Similiarity/Distinctivenessof the AdvertisedBrand.
types within a product category and brands within the Moving to more specific levels in the product hierarchy
same product type indicated that inferences at the also increases the distinctiveness of inferences. This is
product type level are more distinct from each other likely to positively influence the extent to which the
than inferences at the brand level (0.26 vs. 0.45, t(40) advertised product is perceived as similar to like prod-
= 2.0, p < 0.05). Comparison of the similarity between ucts by increasing knowledge of shared attributes and
product types within a product category and brands be- perceived as different from contrast products by in-
longing to different product types indicated that brand creasing knowledge of distinct attributes. Thus, com-
level inferences across product types are marginally, parative advertising has the potential of placing the
though not significantly, more distinct (0.26 vs. 0.17, advertised brand in the "same league" as the named
t(40) = 1.7, p < 0.08). referent (Gorn and Weinberg 1984) and hence making
In sum, the results indicate that the "basic" level of it distinct from contrast brands. Again, since inferences
categorization occurs at the product type level. The in- are expected to mediate, the expected pattern of results
ferences at this level are rich, evaluative, and quite dis- for these measures is the same as that for the other mea-
tinctive. sures.

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions
CATEGORIZATIONAND INFERENCE MAKING 375

Hypotheses uct type ad as Tamron's newest 35mm SLR camera,


and in the comparative ad as compared to the Canon
Based on these arguments, the following hypotheses AE- 1. Other information in the ads was identical. The
are forwarded. text referred to picture-taking capabilities, ease of use,
HI: For experts, comparative ads will differ from and convenience of use. In a pretest, these attributes
noncomparative ads that position or label a were rated as being fairly important and characteristic
brand as a member of a broad product class of any type of camera.
category resulting in: Expertise. An objective knowledge scale on photog-
(a) greater perceived ad informativeness, raphy and cameras (Sujan 1985a) was applied after all
of the dependent measures were collected. Based on a
(b) more extreme brand evaluations, median split, subjects were classified as experts or nov-
(c) greater similarity of the advertised brand ices.
to like brands and to the referent, and
(d) greater distinctiveness of the advertised
Dependent Variables
brand from contrast brands. Inferences. Subjects listed any features that they
thought the advertised brand might possess, even if the
H2: For novices, comparative ads will differ from
feature was not explicitly mentioned in the ad. The
noncomparative ads that position or label a
number of inferences listed was counted for each subject
brand as a member of a broad product class
and coded by two judges who were knowledgeable about
category resulting in:
cameras and blind to the hypotheses. The interjudge
(a) greater perceived ad informativeness, reliability was 77 percent. Disagreements were resolved
by discussion, so that all inferences were coded. Attri-
(b) more extreme brand evaluations, butes coded as being characteristic of all cameras (e.g.,
(c) greater similarity of the advertised brand can use a flash) were given a score of 1, 35mm camera
to like brands and to the referent, and attributes (e.g., interchangeable lenses) were given a
score of 2, and attributes specific to a subset or particular
(d) greater distinctiveness of the advertised brand of 35mm camera (e.g., programmable) were given
brand from contrast brands. a score of 3. The average score indicated the average
H3: For experts, comparative ads will differ from level of specificity of the inferences generated. The sum
noncomparative ads that position or label a of the scores of all the inferences generated produced
brand as a member of a product type category an inference index that reflected both the number and
on measures (a) to (d). specificity of inferences.
Attribute lists that included two general camera class
H4: For novices, comparative ads will not differ attributes, two 35mm attributes, and two attributes
from noncomparative ads that position or la- specific to the Canon AE- 1 were also used to measure
bel a brand as a member of a product type inferences. The attributes used were drawn from the
category on measures (a) to (d). attribute lists from Study 1. Subjects rated how likely
H5: The effects specified in H 1 to H4 will be me- it was that the advertised product possessed each of the
diated by inferences. features. Subjects' responses on the two scales within
each of these three types were averaged to form three
attribute likelihood indices (coefficient alpha: product
Method class index = 0.71, product type index = 0.84, brand
index = 0.79).
Ninety-four students participated. Subjects saw a
simulated print ad. Their task was to form an impres- Ad Informativeness. This was measured on two
sion of the advertised product. Subjects then filled in scales (informative/uninformative, precise/vague) and
the dependent measures and expertise scales and were averaged (alpha = 0.76).
finally debriefed.
Product Evaluation. Three scales (good/bad, favor-
able/unfavorable, positive/negative) were averaged to
IndependentVariables form one index of product evaluation (alpha = 0.89).
Type of Ad. Three print ads were created for a fic- Extremity of evaluation was computed as the absolute
titious brand of camera (Tamron A-25). The brand was deviation of the index from the midpoint of the scale.
positioned in the noncomparative product class ad as Similarity/Distinctiveness. 110 Instamatic cameras
Tamron's newest camera, in the noncomparative prod- were defined as the "contrast" category. Subjects were

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions
376 THE JOURNAL OF CONSUMER RESEARCH

asked to think about a particular type of camera (i.e., (H5) Mediational Effects of Inferences. Analyses on
a 35mm, a 110 camera, or the Canon AE-1) and then the inference index combined both number and spec-
rate how similar/different the advertised brand was to/ ificity of the open-ended inferences. As expected, the
from this camera. analyses indicated that for experts the inference index
was significantly greater when a comparative ad was
used than when either a noncomparative product class
Results ad (1 5.9 vs. 3.6, t = 7.0, p < 0.01) or a noncomparative
The hypotheses were tested using directional t-tests product type ad (15.9 vs. 8.9, t = 4.0, p < 0.01) was
(with 88 degrees of freedom). used. For novices, this index was significantly greater
when a comparative ad was used than when a noncom-
(H]) Comparative vs. Noncomparative Product Class parative product class ad was used (7.7 vs. 3.6, t = 2.3,
Ads: Expert Consumers. The comparative ad led to p < 0.05), but not when a noncomparative product type
higher ratings on measures of perceived informativeness ad was used (7.7 vs. 6.6, t < 1, n.s.). Thus, the pattern
(3.1 vs. 1.8, where 7 is most informative, t = 3.3, p of results for inferences mirrored the pattern of results
< 0.01), extremity of evaluation (1.2 vs. 0.8 where 3 is for the outcome measures.
most extreme, t = 1.9, p < 0.05; though not mean values
The closed-ended attribute likelihood indices mea-
on brand evaluations, 3.9 vs. 3.9), similarity to like
sured the extent to which subjects ascribed product class
brands (2.7 vs. 4.2 where 1 is most similar, t = 2.7, p
< 0.05), similarity to the Canon AE-1 (3.2 vs. 5.3, t (camera), product type (35mm camera), and brand
= 4.2, p < 0.01), and distinctiveness from contrast (Canon AE- 1) attributes to the advertised product. Note
that the most general inferences (e.g., uses a flash) can
brands (6.1 vs. 3.9 where 7 is most distinctive, t = 5.5,
be made across the product class, product type, and
p < 0.01). Thus Hla through Hld were supported.
brand ads. Thus, no effects were expected for the prod-
(H2) Comparative vs. Noncomparative Product Class uct class attribute likelihood index, and the results up-
Ads: Novice Consumers. Novices, like experts, found held this hypothesis.
the comparative ad to be more informative (2.4 vs. 1.8, For both experts and novices, as expected, there was
t = 1.8, p < 0.05), to result in greater similarity to like a greater likelihood of ascribing product type and brand
brands (2.9 vs. 4.3, t = 2.6, p < 0.05) and to the referent level attributes to the product when a comparative ad
(3.2 vs. 4.8, t = 3.2, p < 0.01), and to lead to greater was used than when a noncomparative product class
distinctiveness from contrast brands (6.1 vs. 4.5, t = 5.6, ad was used (experts: product type index = 5.6 vs. 3.3
p < 0.01) than the noncomparative product class ad, where 7 is the most likely, t = 4.8, p < 0.01, brand index
thus supporting H2a, H2c, and H2d. Evaluations, = 4.5 vs. 3.0, t = 3.3, p < 0.01; novices: product type
however, were not more polarized (0.7 vs. 0.8) reject- index = 4.8 vs. 3.5, t = 2.7, p < 0.05, brand index = 3.9
ing H2b. vs. 2.8, t = 2.4, p < 0.05). Since product type inferences
can be made for both product type and brand ads, no
(H3) Comparative vs. Noncomparative Product Type significant differences were expected for the product
Ads: Expert Consumers. Experts found the compara-
type index between the comparative and noncompar-
tive ad to be marginally more informative than the ative product type ads. The results upheld this hypoth-
noncomparative product type ad (3.1 vs. 2.6, t = 1.3, esis (experts: 5.6 vs. 5.2, t < 1, n.s.; novices: 4.7 vs. 5.0,
p < 0. 10). Experts were also more extreme in their eval- t < 1, n.s.).
uations (1.2 vs. 0.6, t = 2.8, p < 0.05). The comparative For experts, differences were expected for the brand
ad did not increase perceptions of similarity to like index between the comparative and noncomparative
brands (2.7 vs. 2.8), but it did lead to greater similarity product type ads, and the results directionally supported
to the referent (3.2 vs. 4.1, t = 1.8, p < 0.05) and greater
this (4.5 vs. 4.0, t = 1, n.s.). For novices, there were no
distinctiveness from contrast brands (6.1 vs. 4.9, t = 3.0, differences on this measure between the comparative
p < 0.01). Thus, H3a-H3d were at least weakly sup- and noncomparative product type ads (3.9 vs. 3.8, t
ported. < 1, n.s.). Thus, these results were consistent with the
(H4) Comparative vs. Noncomparative Product Type notion that more specific attributes are perceived to be
Ads: Novice Consumers. As expected, there were no characteristic of the product when comparative ads are
significant differences (t < 1, n.s., in all cases) on per- used than when noncomparative ads are used, especially
ceived informativeness (2.4 vs. 2.5), extremity of eval- by experts.
uations (0.7 vs. 0.9), similarity to like brands (2.9 vs. To directly test the mediational effects of inferences,
2.7), or distinctiveness from contrast brands (5.3 vs. regression models were run using Chow F-tests to com-
5.0). Thus H4a, H4b, and H4d were supported. How- pare nested models. In the full model, the outcome
ever, the comparative ad did produce greater percep- measure was predicted using the independent variables
tions of similarity to the referent (2.8 vs. 3.9, t = 2.2, and the inference index. In the restricted or nested
p < 0.05), rejecting H4c. model, the outcome measure was predicted using the

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions
CATEGORIZATIONAND INFERENCE MAKING 377

inference index alone. A comparison of the two models 1984). This study demonstrates that inferences gener-
would indicate whether the independent variables had ated, i.e., the attributes ascribed to the product even
any direct effect on the outcome measure beyond that when not explicitly stated, are important mediators of
due to the effect on inference. both ad perceptions and brand perceptions. Thus, the
The results of the restricted models indicated that study points to one fruitful avenue for future research-
the inference index was a significant predictor of all of examining the effects of inferences as mediators of ad
the outcome measures (F values significant at p < 0.01 and product evaluations (cf. Huber and McCann
in all cases). The Chow tests comparing the restricted 1982)-and further suggests that the categorization ap-
and full models indicated that the independent variables proach is a heuristic framework for generating hy-
had no direct effect on two of these measures: ad in- potheses about the number and nature of inferences.
formativeness and extremity of evaluation (F(3,89) < 1, Future research might examine both cognitive re-
n.s. for both). However, the independent variables did sponses and inferences in a single study to determine
directly influence similarity to like brands (F(3,89) the conditions under which each type of cognition me-
= 3.3, p < 0.05), similarity to the referent (F(3,89) diates outcome measures. Subjects can also be in-
= 9.9, p < 0.01), and distinctiveness from contrast structed to rate the valence of their inferences so that
brands (F(3,89) = 5.9, p < 0.01 ) beyond the effects due the coding of inferences more closely parallels the cod-
to inferences. ing of cognitive responses as counter and support ar-
A stronger test of the role of inferences was conducted guments. This might also help predict the direction of
in which it was determined that the independent vari- evaluations rather than just their extremity.
ables had a significant effect on inferences beyond those The advertising stimuli used in this research con-
obtained by using perceived ad informativeness (F(3,89) tained cues at different levels of specificity, and there
= 17.5, p < 0.01) and extremity of evaluation (F(3,89) was no mixing of levels of informational cues within
= 22.0, p < 0.01) as "predictors" for inferences. This each ad. However, more commonly, ads might contain
suggests that in a causal sequence, inferences precede a mixture of cues. Based on the categorization literature,
(or mediate) the effects on the dependent variables. one might predict that the inferences generated would
Thus, H5, relating to the mediational effects of infer- depend upon the most specific cue present in the stim-
ences, was strongly supported for some measures but ulus ad. Future research could test this prediction. Fur-
not for others. ther, since a single product and single ad execution was
used at each level, the key findings of this study need
Discussion to be verified by research using more realistic ads and
multiple products.
The results of Study 2 support the findings of Study The study did not support the hypothesized media-
1: product inferences vary systematically depending on tional effects of inferences on perceptions of similarity
the specificity of available cues. The findings also point to like brands and to the referent and distinctiveness
to the critical role of expertise in inference generation from contrast brands. These measures were directly
especially at the subordinate or brand levels. Compar- predicted by the conditions in the experiment. One rea-
ative advertising containing brand cues led to different son for this might be that judgments of similarity/dis-
responses for experts relative to either product class or tinctiveness could be made based upon the cues explic-
product type level noncomparative advertising but for itly mentioned in the ad (i.e., the ad labelled the product
novices only in relation to product class level noncom- as a camera, a 35mm camera, or similar to the Canon
parative advertising. Further, Study 2 determined that AE- 1). Thus, consumers may rely on inferences to make
inferences were important mediators of some measures judgments only when information is not directly avail-
of effectiveness (perceived ad informativeness and ex- able (e.g., to make judgments of product liking when
tremity of evaluations). Finally, the study confirmed all product information is not provided). However,
previous findings (Gorn and Weinberg 1984) that com- these findings suggest the need for future research that
parative advertising results in greater perceptions of directly assesses the "boundaries" of the impact of in-
similarity to the referent regardless of the contingencies ferences on information processing.
considered.
[Received May 1986. Revised April 198 7.]
GENERAL DISCUSSION
Previous research that has used cognitive responses REFERENCES
as mediators of the effects of comparative advertising
has found that differences in cognitive responses are Belch, George E. (1981), "An Examination of Comparative
not associated with corresponding differences in re- and Noncomparative Television Commercials: The Ef-
sponses to the ad or brand (cf. Gorn and Weinberg fects of Claim Variation and Repetition on Cognitive

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions
378 THE JOURNAL OF CONSUMER RESEARCH

Responses and Message Acceptance," Journal of Mar- Greenwald, Timothy C. Brock, and Thomas M. Ostram,
keting Research, 18 (August), 333-349. New York: Academic Press, 147-170.
Golden, Linda L. (1979), "Consumer Reactions to Explicit Huber, Joel and John McCann (1982), "The Impact of In-
Brand Comparisons in Advertisements," Journal of ferential Beliefs on Product Evaluations," Journal of
Marketing Research, 16 (November), 517-532. Marketing Research, 19 (August), 324-333.
Goodwin, Stephen and Michael Etgar (1980), "An Experi- Mervis, Carolyn B. and Eleanor Rosch (1981), "Categoriza-
mental Investigation of Comparative Advertising: Impact tion of Natural Objects," Annual Review of Psychology,
of Message Appeal, Information Load, and Utility of Vol. 12, 89-115.
Product Class," Journal of Marketing Research, 17 Sujan, Mita (1985a), "Consumer Knowledge: Effects on
(May), 187-202. Evaluation Processes Mediating Consumer Judgments,"
Gorn, Gerald J. and Charles B. Weinberg (1984), "The Impact Journal of Consumer Research, 12 (June), 31-46.
of Comparative Advertising on Perception and Attitude: (1985b), "The Nature and Structure of Consumer
Some Positive Findings," Journal of Consumer Research, Product Categories," Working Paper No. 156, Marketing
11 (September), 719-727. Department, College of Business Administration, The
Greenwald, Anthony A. (1968), "Cognitive Learning, Cog- Pennsylvania State University, University Park, PA
nitive Response to Persuasion and Attitude Change," in 16802.
Psychological Foundations ofAttitudes, eds. Anthony G.

This content downloaded from 128.119.168.112 on Sun, 18 Aug 2013 22:33:35 PM


All use subject to JSTOR Terms and Conditions

You might also like