Sie sind auf Seite 1von 14

A Process-Tracing Study of Brand Extension Evaluation

Author(s): David M. Boush and Barbara Loken


Source: Journal of Marketing Research, Vol. 28, No. 1 (Feb., 1991), pp. 16-28
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/3172723
Accessed: 17/12/2010 07:05

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ama.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

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.

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Marketing Research.

http://www.jstor.org
DAVIDM. BOUSHand BARBARA
LOKEN*

The authorsexplore the implicationsof consideringa brandas representinga


category consistingof its products.Theyreportresultsof a laboratoryexperiment
in whichresponsetimesand verbalprotocolswere usedto examineprocessesrelated
to the evaluationof brandextensions.Evaluationsof brandextensionswere influ-
enced both by the extension'ssimilarityto the brand'scurrentproducts(brandex-
tensiontypicality)and by the variationamong a brand'scurrentproducts(brand
breadth).An invertedU describesthe relationshipbetweenbrandextensiontypi-
cality and evaluationprocess measures.Moderatelytypicalextensionswere eval-
uated in a more piecemealand less global way thanwere eitherextremelytypical
or extremelyatypicalextensions.Subjects'attitudestowardbrandextensionswere
correlatedhighlywith theirratingsof brandextensiontypicality.

A Process-Tracing
Study of Brand Extension
Evaluation

Brandextensionhas been called "theguidingstrategy many strategicconsiderations,includingthe appropri-


of productplannersin the 1980's"(Tauber1988, p. 26). atenessof a company'scorporatestructure,applicability
With such productsas Zenithcomputers,Atarifax ma- of capitalresources,and abilityof personnelin the new
chines, and Ivory shampoo,businessesattemptto use market.It also requiresthat a favorableprior attitude
the equity built up in establishedbrandnames to help towardcurrentbrandedproductstransferto a new prod-
launchnew products.1Capitalizing on an establishedbrand uct. For example,the makersof Zenithcomputersmust
nameis a growthstrategythatseemsdestinedto increase believethatsome of the positiveattitudesassociatedwith
in popularityas the cost of launchinga new brandname Zenith television sets and the categorylabel "Zenith"
skyrockets.Recentcost estimatesfor developinga new will transferto computers.Understandingwhetherand
brandrangefrom$80 millionto $150 million(Aakerand how this transferoccursrequiresan understanding of the
Keller 1989;Tauber1988). way attitudesare linked to particularcategory labels.
Successful brand extension depends importantlyon Thoughconsumerattitudestowardbrandshave long been
recognizedas crucialin marketing,consumerresearch-
'The specificityof the term"product"affectswhethera new prod-
ers have not considereda brandand its productsas con-
uct offeringrepresentsa realextensionof the brand.A 19-inchport- stitutinga category.
able colortelevisionset also can be describedas an "electronicprod- The purposeof our articleis to explore the implica-
uct" or simply as a "televisionset." Here we generallyrefer to tions of viewing a brandand its productsas a "brand
"products"and "brandextensions"at the latterintermediatelevel of category"and, in particular,the effects of this categor-
specificity(e.g., "televisionset"). izationperspectivefor evaluatingpotentialbrandexten-
sions. Viewing a brandas a categoryraises, and pro-
*David M. Boush is AssistantProfessorof Marketing,University vides guidance for addressing, several interesting
of Oregon.BarbaraLokenis AssociateProfessorof Marketing,Uni-
versityof Minnesota.
questions.Do all family-branded productssharethe at-
The articleis based on the first author'sdissertationresearchand titudetowardthe brandcategoryequally?Do consumers
was partiallysupportedby grantsfromthe CurtisL. CarlsonSchool evaluate extensionsthat are similar to currentfamily-
of Managementand from Universityof MinnesotaAcademicCom- brandedproductsdifferentlyfromthose thatare dissim-
puterServicesandSystems.The authorsthankDel Hawkins,Debbie ilar to currentproducts?Do consumersevaluateexten-
John,JimWard,PeterWright,andthreeanonymousJMRreviewers sions of brandcategoriesthatrepresenta broadrangeof
for theircommentson draftsof the manuscript.
producttypesthe sameway thattheyevaluateextensions
16

Journal of Marketing Research


Vol. XXVIII(February1991), 16-28
BRANDEXTENSION
EVALUATION 17

of brandcategoriesrepresentinga narrowrangeof prod- Brand breadth. Anotheraspect of brandknowledge


uct types?To answerthese questions,we first examine structurethatmay influencejudgmentsabouta brandex-
the structureof brandcategories.Specifically, we dis- tension is brandbreadth.Brand breadthrefers to the
cuss how a brandcategorycan be describedby its breadth variabilityamongproducttypes representedby a brand
(i.e., by the variationamongits products)andposit that name.Forexample,the category"Heinzproducts"would
any currentor potential brandedproductcan be de- be extremelynarrowif the only producttype were ket-
scribedby how typicalit is of the brandcategory.Next chup;it wouldbe extremelybroadif it also includedlawn
we describethe processesby which a brandextension mowersanddishwashers.The currentHeinzproductof-
can be evaluated.Hypothesestest the way brandbreadth feringthatincludesbabyfood, sauces, soups, andother
and brandextensiontypicalityinfluencethe evaluation processedfoods is somewherebetween those two ex-
of brandextensions. Our intent is to demonstratehow tremesin breadth.
theoriesof categorizationcan help marketersto antici- Brandbreadthappearsto be a resultof the typicality
pate the effects of brandextensions,andthusplan more of brandextensions.If brandmanagersconsistentlyex-
coherentbrandextensionstrategies. tend the brandby offering new productsthat are very
muchlike (i.e., typicalof) currentones, a narrowbrand
BACKGROUND results.If brandextensionsarevery differentfrom(i.e.,
Category Structure atypicalof) currentproducts,a broadbrandresults. It
seems clear, furthermore,that as new productsbecome
MervisandRosch (1981, p. 89) statethat "a category
established,people's beliefs aboutwhat is "typical"of
exists whenevertwo or moredistinguishableobjectsare the categorywill be revised.
treatedequivalently."However,it is the nonequivalence
of cateogrymembers,or "gradedstructure,"that sets The Effect of Brand Breadth on Perceived Typicality
categoriesapartfrom unorderedsets (Rosch, Simpson, Now let us suppose that one of two conditionspre-
andMiller1976;Smith,Shoben,andRips 1974).Graded vails: a brandname is associated(1) exclusively with
structureand a second characteristicof categorystruc-
soups (a narrowbrand)or (2) with soups, condiments,
ture, brandbreadth,are relevantfor understanding how and frozenvegetables(a broadbrand).Considera new
judgments of new brandsare formed. kind of soup andcannedvegetablesas two potentialex-
Gradedstructure.People perceivemembersof most tensionsof eitherthe narrowbrandor of the broadbrand.
naturallyoccurringcategoriesas varyingin theirdegree For the narrowsoupbrand,anothersoupis an extremely
of those categories.
of typicality,or representativeness,
For example,a robinis perceivedas moretypicalof the typicalmemberof the family-brandcategory.The new
soup probablysharesnearlyall the salientfeaturesof the
category "bird"than is an ostrich. Furthermore,even currentfamily-branded products(e.g., a healthfulfood,
nonmembersdifferin how typicala nonmemberthey are a stand-alonemeal, wateryconsistency).The potential
of a category.For example, an unrelatedobject such as brandextensioncannedvegetables,in contrast,is not a
a chairis a betternonmemberof the category"bird"than
is a butterfly.This rangein categoryrepresentativeness very typicalmemberbecauseit sharesonly a few char-
acteristicswith soups (e.g., it is a healthfulfood butnot
from the most representativemembersof a categoryto a stand-alonemeal). Further,soup cannotbe as typical
the nonmembersthat are least similarto the categoryis a brandextension for the broadbrandas it is for the
called "gradedstructure"(see, e.g., Barsalou 1985; narrowbrandbecause it sharesfewer of the character-
Mervisand Rosch 1981) andhas been demonstratedfor istics of currentproducts(e.g., it shares most charac-
a varietyof consumercategories(LokenandWard1987; teristicswith soups but few with condimentsor frozen
Wardand Loken 1986). In the contextof brandexten-
vegetables).Cannedvegetables,however,are moretyp-
sion, gradedstructureimpliesthatsomeproductsaremore ical of the broadbrandthan of the narrowsoup brand
representativeof a brandcategorythan are others, and becausecannedvegetablessharemanyof the featuresof
preliminarydata supportthis assumption.2 frozenvegetables,one of the currentfamily-brandmem-
bers.
As the precedingexample illustrates,brandbreadth
2Toverifythatgroupsof productsrepresentedby brandnamescan shouldinteractwith brandextensiontypicality;percep-
have categorystructure,30 men and46 women 18 to 75 yearsof age
were interviewedby telephonefroma sampleof 190 namesdrawnat tions of typicality should be more extreme for narrow
randomfromthe telephonedirectoriesof two large midwesternmet- brands than for broad brands. Brand extensions that are
ropolitanareas. Respondentswere asked to rate variousproductsas essentially the same as the brand'scurrentproductsshould
examples(0 = very poor example, 10 = very good example)of the
brandsSony, Jello, Kraft,andGucci. For each brand,threeproducts be perceived as more typical if the brand is narrow than
were very good examples(e.g., Sony television set, Gucci shoes), if the brand is broad. However, brand extensions that
threeweremoderatelygood examples(e.g., Sony camera,Guccicos- are very different from (any of) the brand's currentprod-
metics), andthreewere verypoorexamples(e.g., Sony shoes, Gucci ucts should be perceived as less typical if the brand is
televisionset). Intersubjectagreementaboutthe relativegoodnessof narrow than if the brand is broad.
examplefor these productswas highly significant(p < .000) across
all four brandsas measuredby Kendall'scoefficientof concordance It is importantto note that the preceding propositions
(Barsalou1983). would not hold if categorization of novel objects oc-
18 OF MARKETING
JOURNAL RESEARCH, 1991
FEBRUARY

curredby matchingthe new instancewith any current and affect is computedthrougha weightedcombination
category exemplar(e.g., its "nearestneighbor").Re- of attributes.
views of recentevidence(cf. Barsalou1990;Malt 1989) Smith, Shoben, and Rips (1974) proposea two-step
supporta model in which either exemplaror abstract modelforjudgingwhethera conceptor objectis a mem-
prototypescan be used. Therefore, ber of a category.In the first step, which is rapid,one
of narrow tries to matchthe featuresof the categorywith the fea-
H1:Typicalityratingsforextensions brandsare
turesof the object.Whenthereis a clearmatchor a clear
moreextremethantypicalityratingsfor extensions
of broadbrands. mismatch,the processis complete.Whensome features
matchand some do not match, a second, slower stage
Piecemealand CategoricalEvaluationProcesses of processingis necessary.Duringthe second stage, a
Relevantto the studyof brandextensionis the way in morecarefulcomparisonof definingfeaturesof the cat-
whichthe affectassociatedwiththe categorygeneralizes egory is madeto determinewhetherthe objectis a mem-
to the brandextension. Following Fishbeinand Ajzen ber of the category.
(1975), we view an attitude(e.g., towarda brandex- Both the Fiske and Pavelchakmodel and the Smith,
tension) as the location of an object or concept on an Shoben, and Rips model involve a two-stage process.
affectivedimension.Attitudestowardbrandextensions The first stage is rapidand global whereasthe second
can be formedin at least two ways. In one, an attitude stage is slower and more deliberate.However,the for-
is "computed"from specific attributesof the extension. mermodeldescribesaffectiveresponsesto a new object;
Such a process has been termed "analytical"(Cohen the lattermodel addressescategorymembershipbut not
1982), "piecemeal"(Fiske 1982), or "computational" affect. In the case of brandextensions,we are interested
(Brooks1978). A varietyof computational models(e.g., primarilyin the way in which people evaluatethe new
Bettman,Capon,andLutz 1975)haveprovedusefulboth productaffectively,which in turndependson the new
for predictingconsumerattitudesandfor diagnosingthe product's"fit"to the originatingbrand.Oncepeopleare
bases of attitudes.However, the models do not purport informedthata new extensionhas been madeby a par-
to describeconsciousevaluationprocesses. ticularcompany,they may evaluatethe new productas
A differentgroupof models (cf. Brewer 1988; Fiske desirableor undesirableon the basis of (1) whetherthey
1982;SrullandWyer 1989)relieson categorizationpro- like the originatingbrandand (2) whetherthe new prod-
cesses to describeattitudeformation.If a new instance uct is representativeof or similarto the brand'scurrent
(e.g., a brandextension)is identifiedas belongingto a products.
previouslydefinedcategory(e.g., a brand),the attitude Thoughnot explicitly addressingthis issue of evalu-
associatedwith that categorycan be transferredto the ation, the Smith, Shoben, andRips modelraisesthe in-
new instance.Fiske, for example, arguesfrom dataon terestingpossibilitythat when an extensionis very dis-
personperceptionthatan objectwill receivethe attitude crepantfrom the original category, a rapid evaluation
associatedwith a categoryor schemato the extent that may occur.The Fiske andPavelchakmodelpredictsthat
it is perceivedto fit the category. Similarly,Srull and mismatches,whetherclear or ambiguous,evoke slow,
Wyer propose that people attempt to form general piecemeal processes. Following Smith and his coau-
impressionsof other people and will use such impres- thors, we believe thereis an alternativeto the Fiske and
sions both to assess new informationand to make sub- Pavelchakmodel wherebyboth clear matchesand clear
sequentjudgments.Otherresearcherswho describeat- mismatchesof the originalbrandcategoryare evaluated
titudesas being associatedwith a categoryratherthan morerapidlythanmoderatemismatches.This prediction
generatedby inferencesaboutindividualattributeshave is also consistentwith the literature(e.g., Mervis and
examinedaffectreferral(Wright1976)andcategory-based Rosch 1981), whichshowsthatboth(1) the moresimilar
processing(Sujan 1985). an exemplaris to a prototypeof the category,the more
Retrievalof prioraffect and computationof affect are quicklyit will be judged as a memberof thatcategory,
two processesthatshouldnot be assumedto be mutually and (2) the moredissimilara nonmemberof a category
exclusive in any given affectivereaction.For example, is to a prototype,the more quicklyit will be judged to
retrievedaffect could influencebeliefs abouta product, be a nonmemberof (or "poorfit" to) the category.Eval-
which in turn could influence the attitudetoward the uatinga new brandextensionpresentspeoplewith a cat-
product.Some models,two of whicharediscussedhere, egory verification task (e.g., they verify that a new con-
explicitly incorporateboth retrievaland computational diment is an acceptable member of the brand category
processes.In one suchmodel,FiskeandPavelchak(1986) "Heinz products"). Very atypical extensions are quickly
proposea two-stepprocessfor evaluation.The first step judged to be a "poor fit" with the current brand image
involves an attemptto matchthe new objectwith a cur- and elicit fewer piecemeal processes than do moderately
rentcategory.If categorizationis successful(i.e., if there typical brand extensions. Very typical brand extensions
is a match),the affect associatedwith the categorylabel also should elicit only the first stage of evaluation, with-
is appliedto the new object and the evaluationprocess out need for slow, piecemeal processes.
is complete.If thereis a poor matchbetweenthe object In summary, the relationship between brandextension
and categoryknowledge,piecemealprocessesare evoked typicality and evaluation response time is hypothesized
BRANDEXTENSION
EVALUATION 19

to be an inverted U. That is, extremely atypical potential inal brand has a negative affective label, the more dis-
members of a brand (e.g., Heinz floor wax) should be crepant the brand extension is, the less positively it will
evaluatedrapidlybecause a clear mismatchbetween brand be evaluated as an instance of the original brand. Intu-
and brand extension elicits only the first stage of eval- itively, from a brand management perspective, this no-
uation. Moderatelyplausiblebrandextensions (e.g., Heinz tion seems logical. Atypical brand extensions are less
instant coffee) should elicit the second, slower stage of likely to share the advantages that the brand name has
evaluation because neither category membership nor cat- accrued, and the resulting evaluation will be less positive
egory nonmembershipis obvious. The consumer is likely than the evaluation of a brand extension that is typical
to judge, in a piecemeal way, whether the attributesthat of the family brandcategory (Boush et al. 1987). Hence,
make up the abstract representation associated with the
current brand name can be used to develop the potential Ha: Forpositivelyevaluatedbrands,a brandextension's
extension. Finally, extremely typical brand extensions degree of typicalityof the family brandis related
positivelyto subjects'evaluationsof the extension.
(e.g., a new Heinz sauce), like extremely atypical brand Thatis, brandextensionsare liked betterif they are
extensions, should require only the rapid first stage of typicalof the originalfamily-brandcategorythanif
evaluation. Because piecemeal processes are indicated they are not.
by slow response times and more attribute-focused or
inferential cognitive responses (cf. Sujan 1985), we pre- Finally, if the breadth of a brand's current products
dict: influences the perceived typicality of potential brandex-
tensions, then, by the same arguments proposed previ-
H2: Extremelytypicalbrandextensionsareevaluatedmore
ously, brand breadth should likewise influence attitude
rapidlythanmoderatelytypicalbrandextensions. toward the brand extensions (cf. Judd and Lusk 1984;
H3: Extremelyatypical brandextensions are evaluated
more rapidlythan moderatelytypical brandexten- Linville 1982; Parducci and Wedell 1986). Therefore,
sions. Hg: Subjects'reportedaffect towardan extension of a
H4: Extremelytypical brandextensionselicit fewer at- narrowbrandis more extremethan that towardan
tribute-related
(piecemeal)cognitive responsesthan extensionof a broadbrand.
do brandextensionsthat are moderatelytypical of
the brand. METHOD
H5: Extremelyatypicalbrandextensionselicit fewer at-
tribute-related The objective of our study was to determine the effect
(piecemeal)cognitive responsesthan
do brandextensionsthat are moderatelytypical of of brand category structure, in particular the effect of
the brand. brand breadth and brand extension typicality, on the
evaluation of potential brand extensions. Though the sit-
Additionally, because narrow brands are expected to uation described here for brand extensions differs from
promote extreme perceptions of brand extension typi- that described in the person perception literature-the
cality (HI), it follows that extensions of narrow brands objects are productsratherthanpeople-the models should
should be less likely than extensions of broad brands to nevertheless be relevant to testing hypotheses about the
require the second stage of evaluation. Therefore: processes and outcomes of categorizations of potential
H6: The evaluationof an extensionof a narrowbrandis new extensions.
more rapidthanthe evaluationof an extensionof a Brand information was manipulatedand subjects eval-
broadbrand. uated a series of potential brand extensions that varied
H7: The evaluationof an extension of a narrowbrand systematically from an established brand concept. The
elicits fewerpiecemealcognitiveresponsesthandoes evaluation process was traced by using response times
the evaluationof an extensionof a broadbrand. and verbal protocols. Attitudes toward the potential ex-
tensions were assessed by means of rating scales. Rather
Effects of Brand Extension Typicality on Evaluation than using actual brands, for which brand breadth could
So far we have examined the nature of the evaluation not be varied systematically without introducing brand
process rather than the outcomes of categorization pro- history confounds into the study design, we developed
cesses. Suppose people have information about products fictitious brandsthat systematicallyvaried in brandbreadth
representing a family brand and then report their attitude and the typicality of brand extensions.
toward a potential brand extension. A spreading acti-
vation model of memory (Collins and Loftus 1975) de- Design
picts this information as nodes connected by links that The study was a 2 x 5 mixed design with two levels
indicate relative proximity in a semantic network. Such of brand breadth (narrow or broad) and five levels of
a model suggests that typical category members (i.e., brand extension typicality. The latter was a repeated
typical brand extensions) are linked more closely to the measure. Two replicates of this experiment were con-
affect of the category (the brand). The concept of sche- ducted for each subject, one involving brands associated
matic fit (Fiske 1982) also suggests that similar objects with grocery products and the other involving brands as-
elicit similar affect. Therefore, a judgment of poor fit is sociated with electronic products. All brand extensions
likely to have affective consequences. Unless the orig- in the grocery products replicate were extensions of a
20 JOURNAL
OF MARKETING 1991
FEBRUARY
RESEARCH,

branddescribedin a cover story as "brandB" and all throughthe following procedurefor each of two repli-
extensionsin the electronicproductsreplicatewere de- cates. The subjectwas askedto reada bookletcontaining
scribedas extensionsof a brandcalled "brandG." threekindsof informationabouta fictitiousbrand(e.g.,
brandB) andits competitors(e.g., brandsA, C, andD).
Independent Variables The firstkindof brandinformationconveyedsimilarities
Brand extension typicality. Brand extension typicality in size and age of the parentcompanies.This informa-
was manipulatedas a within-subjectsvariable.Subjects tion was providedas an overviewof the companiesand
evaluatedpotentialbrandextensionsthatrangedfromvery was includedto minimizeunwantedinferencesby sub-
similarto the brand'scurrentproductsto very different jects when they were told lateraboutdifferencesin the
fromcurrentproducts.Ratingscales havebeenusedpre- productssold undereach brand.The secondkind of in-
viouslyas an appropriate measureof perceivedsimilarity formationwas intendedto manipulatebrandbreadth.It
(Sujanand Dekleva 1987; Wardand Loken 1986), and includeda descriptionof each brand'sproductmix. The
typicalitymeasuresthat employ similarityscales have thirdkind of informationprovidedmore specific infor-
been used frequently(e.g. Barsalou 1985; Rosch and mationabouteach brand'sproducts.It was designedto
Mervis 1975;Tversky 1977). inducea positiveattitudetowardthe targetbrand.Each
A pretestwas conductedto select potentialbrandex- brandwas describedby listingproductsalongwith eval-
tensionsthatwouldprovidea rangein similarityto soup, uationsof thoseproductsin a ConsumerReportsformat.
condiments,and frozenvegetables(the "current"brand At the end of the bookletwere instructionsfor the sub-
B products)and to televisionsets, calculators,and dig- ject to look over all the informationon the brandsuntil
ital watches (the "current"brandG products).Pretest "you feel you have formedan impressionof the four
subjectswere also askedto thinkof extensionsthatwere brands."The rationalefor includinginformationabout
extremelyunlikely. On the basis of the graded mean the competitivebrandswas to providea contextfor the
similarityto currentproducts,the potentialextensionsof target information.When the subject had formed an
brandB (the groceryproducts)were selectedto be fro- impressionof each brand,he or she was instructedto
zen vegetables,chunkystew, steaksauce, breakfastcer- turn the page and rate each brand'sproductson three
eal, cannedfruit, toothpaste,floor wax, and pet food. semantic differentialattitudescales (describedsubse-
The potentialextensions of brand G (the electronics quently).
products)were selectedto be televisionsets, calculators, Afterevaluatingthe four brands,the subjectreturned
digital watches, cameras,refrigerators,ballpointpens, the first bookletand receiveda second booklet. He or
bicycles, and garbagecans. Potentialbrandextensions she was askedto readalongas the experimenter readthe
in each replicateincludedproductsthatwere "thesame" instructionsaloud. Subjectswere told thatthe next part
as those of the currentbrand, productsthat were ex- of the studypertainedto the way they would rate some
tremelydifferentfromthose of the currentbrand(almost new productsthatmightbe made by brandB (or brand
nonsensical),and productsthat were moderatelydiffer- G) andthatthey wouldbe askedto rateseveralpotential
ent fromthose of the currentbrand. brandB (or brandG) productsone at a time. They were
Brand breadth. Brandbreadthwas manipulatedfor asked to thinkout loud while evaluatingeach product.
fictitiousbrandsas a between-subjectsfactor with two The subjectthenturnedthe page andreadalouda ques-
levels. Brandswere constructedartificiallyto be com- tion such as, "How would you rate brandB breakfast
posed either of very similarproductsor of a more di- cereal?",and completedtwo semanticdifferentialatti-
verse groupsof products.A narrowbrandin the grocery tude ratingscales (describedsubsequently).The subject
productsreplicatewas composedof soups only, condi- evaluatedall of the potentialbrandextensionsin thisway.
ments only, or frozen vegetables only; a broadbrand (The first was for practiceonly.) The time requiredfor
was composedof soups, condiments,and frozen vege- the subjectto evaluateeach brandextensionandthe ver-
tables. A narrowbrandin the electronicproductsrepli- balizationsmade duringeach evaluationwere recorded
cate was composedof television sets only, calculators by meansof a videocamera(PentaxPV-C55A)equipped
only, or digitalwatchesonly;a broadone was composed with a stopwatchfunction(RCA CGA020). The order
of television sets, calculators,and digital watches. By in whichsubjectsevaluatedthe brandextensionswas de-
includingthreenarrow-brand conditionsfor each broad- terminedrandomlyandthenreversedfor half of the sub-
brand condition, we attempted to hold the type of prod- jects. The order of the replicates was reversed for half
uct constant between the two breadth levels and so elim- of the subjects and the design was balanced for combi-
inate confounds due to type of product. Also note that, nations of order effects between replicates.
for the broad brands, three different potential extensions After subjects had evaluated brand extensions for both
in each replicate could be "the same" as current prod- replicates, they completed backgroundquestions and gave
ucts. personaldemographicinformation.Finally, similarityand
recall measures were taken to ensure that brand breadth
Procedure and brand extension typicality were manipulated as in-
The subject sat at a desk that was facing a video cam- tended. The entire procedure took from 20 to 30 minutes
era approximatelyseven feet away. The subject then went per subject.
BRANDEXTENSION
EVALUATION 21

Dependent Variables accurateestimatesof type I errorswithoutan assumption


of any particularform of the variancesand covariances
Typicalitywas measuredby askingsubjectsto ratethe
similarityof each potentialbrandextensionto "products amongtherepeatedmeasures(LaTourandMiniard1983).
the brandcurrentlymakes"on a 7-pointbipolarscale (1 Manipulation Checks
= dissimilar, 7 = similar). (Remember that typicality of
the brandextensionis includedin the studyas not only Affect toward the brands. The assumption was that in
an independentvariablebut also a dependentvariable orderto transferattitudefroma brand'scurrentproducts
becauseperceived typicalityis predictedto dependon to new products,subjectshad to constructan attitude
whetherthe brandis narrowor broad.) towardthe brand'scurrentproducts.Providingproduct
Attitudestowardthe brandextension(evaluativeout- ratingsin a ConsumerReportsformat(as describedpre-
comes) were measuredon two 7-point semanticdiffer- viously)was the meansby whichwe triedto createthese
entialscales anchoredby favorable/unfavorable andde- positive attitudes.Evaluationson three 7-point bipolar
(cf. BurkeandEdell 1989).Evaluative
sirable/undesirable adjectivescales anchoredby desirable/undesirable,fa-
processes were measuredby the lengthof time takenby vorable/unfavorable,and high quality/low quality, re-
a subjectto makean evaluationandby concurrentverbal spectively, yielded means rangingfrom 5.3 to 6.0, all
protocols.The responsetime used in these analyseswas acceptablypositive. There were no significantdiffer-
measuredfromwhen subjectsfinishedreadingthe ques- ences betweenbrandson these measures.
tion on the brandextensionuntil they markedthe first Brand extension typicality. Subjects were asked to rate
the similarityof eachpotentialbrandextensionto "prod-
ratingscale. This intervalomits both the readingtime uctsthe brandcurrentlymakes"on a 7-pointbipolarscale
and the time for subjectsto read and considermultiple
scales, which can vary across subjects. (1 = dissimilar, 7 = similar). The results indicated that
the potentialbrandextensions representeda range in
Subjects similarityto the brand'scurrentproductsin both repli-
cates (see Table 1). However, among electronicprod-
Subjectswere 144 universitystudents,72 in each of ucts, the ballpointpen and bicycle were equallydissim-
the two breadthconditions.The narrowbrandconditions lar to currentproducts. Because the bicycle did not
were balancedfurtherby product.That is, in the elec- constitutean additionallevel of the typicalitymanipu-
tronics replicate, 24 subjects reacted to brandsrepre- lation, it was droppedfrom the analyses.3
sentedby television sets only, 24 to brandsrepresented Brand breadth. Subjects were asked to name the
by calculatorsonly, and24 to brandsrepresentedby dig- productsthey recalledthat brandB and brandG "cur-
ital watchesonly. In the narrowbrandconditionsof the rentlymake."Forthe manipulationto have worked(and
groceryproductsreplicate,24 subjectsreactedto brands be recalled)perfectly,all subjectsin the narrowbrand
representedonly by frozenvegetables,condiments,and conditionsshould have rememberedonly one kind of
soups, respectively.Subjectseach were paid $5.00 and, product(e.g., frozenvegetablesonly, condimentsonly,
in some cases, given extracreditfor participation.Sub- or soup only) and all subjectsin the broadbrandcon-
jects were assignedrandomlyto experimentalconditions ditionsshouldhave namedthreekindsof products(e.g.,
within each of the two replicates.Those who were as- frozen vegetables,condiments,and soup). The manip-
signed to the narrowbrandexperimentalconditionsin ulationworkedwell, if not perfectly.Nearlyall subjects
the first replicatewere assignedto the broadbrandcon- (98%)in the narrowbrandconditionsfor both brandB
dition in the second, and vice versa. and brandG recalledone producttype. Most subjects
RESULTS (78%)in the broadbrandconditionsaccuratelyrecalled
threeproducttypes.
The responseswere analyzed,for the most part, in a
2 (narrowbrand/broadbrand)x 5 (potentialbrandex- Brand Extension Typicality as a Dependent Measure
tensions rangingfrom the same as currentproductsto H, predictsthat typicalitywould interactwith brand
very differentfrom currentproducts)repeatedmeasures breadth;specifically, that typicalityratings for exten-
analysisof variancedesign. sions of narrowbrandswouldbe moreextremethantyp-
Hypothesizedeffects were tested by performingboth icality ratingsfor broadbrands.As predicted,thereis a
univariate and multivariateanalyses of variance. The ho- significant interaction between brand breadth and the
mogeneity of variance assumption was not met in any of product within-subjects factor in multivariateanalyses of
the analyses. That is, the observations were significantly variance for both the grocery products replicate (F4,139 =
correlated within subjects (Greenhouse-Geisser epsilon 4.27; p < .003) and the electronicproductsreplicate(F3,138
ranged from approximately .80 to .90). Conventional
univariateanalyses therefore would produce inflated type
I errors, implying an overly liberal test of the hypotheses
3Analyses conducted without the bicycle and without the ballpoint
(Geisser and Greenhouse 1958). The preferred solution pen were substantivelythe same. The bicycle was chosen to be dropped
to this problem is to perform multivariate analyses of because of indications that it produced the only significant order effect
variance on the within-subjects effects, which produce in the analyses.
22 JOURNAL
OF MARKETING FEBRUARY
RESEARCH, 1991

Table 1
CELLMEANSFORDEPENDENT
MEASURES

Percent
Typicality rating Attitude rating Response time piecemeal responses
Narrow Broad Narrow Broad Narrow Broad Narrow Broad
brand brand Overall brand brand Overall brand brand Overall brand brand Overall
Grocery products
"Same" product 6.72 6.32 6.52 12.22 11.44 11.83 8.41 9.66 9.02 6.9 15.3 11.1
Canned fruit 4.67 5.07 4.87 9.76 10.53 10.14 8.93 11.14 10.03 18.1 18.1 18.1
Breakfast cereal 3.34 3.76 3.56 8.66 8.94 8.80 9.45 12.52 10.98 11.1 11.1 11.1
Toothpaste 1.99 1.92 1.95 6.45 6.87 6.66 8.58 10.58 9.57 6.9 9.7 8.2
Floor wax 1.38 1.32 1.35 5.59 5.79 5.69 9.50 8.41 8.97 6.9 5.6 6.3

Electronic products
"Same" product 6.83 6.56 6.69 11.66 12.22 11.94 8.04 8.26 8.15 4.2 21.4 12.7
Camera 4.69 5.53 5.11 10.00 10.83 10.44 10.65 10.23 10.44 22.2 20.0 21.1
Refrigerator 3.08 3.76 3.43 8.14 8.94 8.53 10.25 12.55 11.40 12.5 17.1 14.5
Ballpoint pen 2.26 2.11 2.18 7.87 7.66 7.79 11.56 10.27 10.91 9.7 7.1 8.4

= 9.74; p < .000). In both replicates, results indicate analyses of variance reportedhere employed naturallog-
that brand breadth influenced the perceived typicality of arithmic transformationsof the response times.4
products. For potential extensions that were essentially Effects of typicality on response time. H2 predicts that
the same as current products, the perceived typicality extremely typical brand extensions would be evaluated
ratings were higher when the brand originally repre- more rapidly than moderately typical brand extensions;
sented a narrowrange of product types than when it rep- H3 predictsthat extremelyatypical brandextensions would
resented a broad range of product types (mean = 6.72 be evaluated more rapidly than moderately typical ones.
vs. 6.32 in the grocery replicate; 6.83 vs. 6.56 in the A main effect for the product within-subjects factor is
electronics replicate). significant in the electronic products replicate (F3,133 =
The opposite pattern of results occurred for potential 14.20; p < .000). Results vary in the predicted direction
extensions that were moderatelydifferentfrom the brand's but are not significant in the grocery products replicate
current products. Canned fruit, for example, was rated (F4,127 = 1.92; p < .111). Brand extensions that were
as more typical of the current grocery products brand "the same" as current products were rated more rapidly
when the brand represented a broad range of products than any of the other brand extensions in the eletronics
than when it represented a narrow range (mean = 5.07 replicate, supportingH2 (see Table 1). Further,the brand
vs. 4.67). Similarly, the camera was rated as more typ- extensions that were most atypical were evaluated faster
ical of current electronic products when the brand was than more moderatelyatypical extensions, supportingH3.
broad than when it was narrow (mean = 5.53 vs. 4.69). Because typicality also was included in the study as a
Hence, narrow brands increased the perceived typi- dependent variable, and hypotheses pertain to perceived
cality of brand extensions that were about the same as typicality of extensions, we believe it is also useful to
current products and decreased the perceived typicality examine the changes in response times as a function of
of products that were moderately different from current this continuous-level variable. Therefore, for each rep-
products, supporting HI. However, as shown in Table licate, response times are also shown plotted against de-
1, when the potential extensions were very different from gree of reportedtypicality in Figure 1, A and B. Again,
the brand's current products, differences between nar- overall, the predicted curvilinear relationship between
row and broad brand conditions were small. perceived typicalityand speed of judgment(responsetime)
is indicated.
Effects on Response Time Effect of brand breadth on response time. H6 predicts
that extensions of narrowbrandswould be evaluatedmore
Several hypotheses were made about the speed with
which subjects make judgments. Faster judgments are
assumed to be associated with categorical processing and 4Response times could be recorded fairly precisely by using slow-
slower judgments with piecemeal processes. Preliminary motion and freeze-frame controls. A subset of 30 randomly chosen
responses were timed twice. The mean difference between the times
analyses of response times revealed substantial positive was just under .15 second and the maximum difference was .30 sec-
skewness, indicating that a transformationto normalize ond. Errors of this magnitude should not significantly affect the anal-
the data was appropriate (Kirk 1982). Therefore the yses.
EVALUATION
BRANDEXTENSION 23

Figure 1
OF BRANDEXTENSION
EFFECT TYPICALITY
AND BRANDBREADTHON RESPONSE
TIMES,PIECEMEAL
COGNITIVE
AND ATTITUDE
RESPONSES, RATINGS

A. ResponseTim ByTypcaltyRating B. ResponseTmeByTypicalty


Rating
OroceryProducts lectron Products
Response e (snds)
ReposTime(Reoneim (seconds)

12 . . . . . .' . . .
......... *............. ... .....
12

3 41 2 3 4 6 6 1

TykFt PyTyp mFtnP

C. Piecemeal Responses By TypicalityRating D. Piecemeal Responses By TypicalityRating


roceryProducts Electronic
Products
PRtPiecemesl
Responses PctPiecemeal
Responses
20 25

18- ..
18

....
......... .. ...
15...o.. ....
8
?5

I I 0 I I I I
41iI
1 2 3 4 5 7 1 2 3 4 5 86 7
6
Typicaty
Pating TypicWny
Rang

E. AttitudeRatingByTypicality
Rating F. AttitudeRatingByTypicality
Rating
OroceryProducts Electronic
Products
AttitudeRating AttitudeRating
14 13

11
10
10

6I -

SI I I 7 I I I I
1 2 3 4 5 7 1 2 3 4 5 6 8
Typicality
Rating Typicality
Rating

Narrow Broad Overall


i--] 0
.... .. ......• ....
24 OF MARKETING
JOURNAL RESEARCH, 1991
FEBRUARY

rapidlythanextensionsof broadbrands.Thoughon av- extensions.Goodness-of-fittests confirmthis resultboth


erage potentialextensionsof narrowbrandswere eval- for groceryproducts(x2 = 9.58; p < .05) and for elec-
uatedmorerapidlythanthose of broadbrands(mean = tronicproducts(x2 = 14.01;p < .003). The overallper-
8.97 vs. 10.46 for the groceryreplicate;mean = 10.12 centage of subjects renderingpiecemeal responses is
vs. 10.32 for the electronicsreplicate),the main effect greaterfor the two moderatelytypicalbrandextensions
of brandbreadthis not significantin eitherreplicate(F,1o30 in each replicate(18.1%for cannedfruit;21.1% for the
= 1.98;p < .161 in the groceryreplicate;F1,135 < 1 in camera)than for eitherthe "same"(11.1% in the gro-
the electronicproductsreplicate).However, an interac- cery replicate;12.7%in the electronicsreplicate)or the
tion betweenbrandbreadthandproducttypicalityis sig- extremelyatypicalbrandextensions(6.3%for floorwax;
nificantin multivariatetests involvingthe groceryprod- 8.4% for ballpintpen). H4 and H5 are supported.Fur-
ucts replicate(F4,127 = 2.41; p < .05) and is close to thermore,in contrastto responsetimeresults,whencog-
conventionalsignificancein the electronicproductsrep- nitive responsesare plottedby perceivedtypicality(see
licate (F3,133 = 2.14; p < .098). Figure1, C andD), the predictedinverted-U-shaped re-
Because of its marginalsignificance,this interaction lationshipappears to be more evident for narrow than
must be interpretedcautiously;however, it can be in- for broadbrands.
terpretedin two ways. First, as shown in Table 1, ex- Effect of brand breadth on cognitive responses. H7
tensionsof narrowbrandswere evaluatedmore rapidly predictsthat extensionsof a narrowbrandwould evoke
thanthose of broadbrands,but only when the potential fewer piecemeal thoughtsthan extensions of a broad
extensionwas not verydifferentfromthe brand'scurrent brand.This hypothesisis supportedonly for "thesame"
product(the one exceptionwas the camerain the case productin the electronicproductsreplicate(X2= 8.1; p
of electronicproducts).For example, as predicted,the < .004). Brandextensionsthat were the same as the
productthat was "the same" as one that the brandal- brand'scurrentproductselicitedpiecemealprocessesmore
readymadewas evaluatedmorerapidlyfor narrowbrands often for broad(21.4%)thanfor narrow(4.2%)brands.
thanfor broadbrands(mean= 8.41 vs. 9.66 secondsin Resultsfor "the same"productin the groceryproducts
the groceryreplicate;mean = 8.04 vs. 8.26 secondsin replicate(15.3%vs. 6.9% for broadandnarrowbrands,
the electronicsreplicate).However, the potentialbrand respectively)vary in the expecteddirectionbut are not
extensionthat was most atypicalof the brand'scurrent significant.A new productthatis just like thosethe brand
productswas evaluatedfasterfor broadbrandsthanfor alreadymakes seems more likely to evoke thoughtful-
narrowones (for floor wax, mean = 8.41 vs. 9.50 sec- ness if the brandconsists of a varietyof productsthan
onds; for ballpointpen, mean = 10.27 vs. 11.56 sec- if it consists of only one kind of product.Thoughthis
onds). A secondinterpretation of these data sheds some is an interestingresult, H7 is not supportedoverall.
on
light H2 and H3. As can be seen in Figure 1, A and
B, the predictedinverted-U-shaped relationshipbetween Effects on Attitude
typicalityand responsetime is more evident for broad The correlationsbetween the "desirability"and "fa-
thanfor narrowbrands.For narrowbrands,not only is vorability"scales rangefrom .81 to .92. Consequently,
the shapeof the curveflatter,butalso the responsetimes attitudewas measuredby summingthe two scalesto pro-
for the most atypicalextensionincrease. duce a single 14-pointscale (with a ratingof 8 as the
neutralpoint).
Effects on Cognitive Responses Effect of product typicalityon attitude. H8 predictsthat
Protocolswere coded into categoriesfor thoughtsre- the more typical a potentialextension is of the brand,
latingto specificproductattributesandthoughtsdrawing the more positively it would be evaluated.The results
inferencesabout the products.Thoughtsaboutproduct supportthat hypothesis.Main effects for producttypi-
attributesor inferencesdrawnaboutthe suitabilityof the cality are significantin both the groceryproductsrep-
potentialbrandextensionsexpressed "piecemeal"pro- licate (F4,139 = 117.71; p < .000) and the electronic
cesses. Subjects'thoughtswere coded by threejudges. productsreplicate(F3,140 = 64.98; p < .000). Atypical
One judge coded all the responsesand two judges each extensionswereperceivedvery negativelyin the grocery
coded responsesfrom a differentgroupof 20 subjects. productsreplicateand slightlybelow the neutralpointin
Disagreementswere resolvedby the first authorand all the electronicproductsreplicate(see Table 1). The cor-
responses were coded. Interjudge agreement averaged relation between attitude and perceived typicality ratings
83%. Because the majority of subjects made no piece- averages .50 in the grocery replicate and .48 in the elec-
meal responses, cognitive responses were treated as a tornics replicate. This relationshipbetween typicality and
categorical variable and the proportion of subjects ren- attitude ratings is demonstrated graphically in Figure 1,
dering piecemeal responses when evaluating a particular E and F.
product was compared across factors. Effect of brand breadth on attitude. H9 predicts that
Effects of typicality on cognitive responses. H4 and H5 extensions of narrow brands would elicit more extreme
propose respectively that extremely typical and ex- attitudes than extensions of broad brands. The results
tremely atypical brand extensions would evoke fewer generally support that hypothesis. The main effect of
piecemeal processes than would moderatelytypical brand breadth is significant in the electronic products replicate
BRANDEXTENSION
EVALUATION 25

(F1,142 = 5.07; p < .026); the interactionbetweenbrand is that different classificationshave different conse-
breadthand producttypicalityis significantin the gro- quences. You shouldbe less botheredwhen a librarian
cery productsreplicate(F4,139 = 2.92; p < .023) and is does not fit your "librarian"
stereotypethanwhen a doc-
close to conventional significance for the electronic torwho is aboutto operateon you does not fit your"sur-
productsreplicate(F3,14 = 2.40; p < .070). geon" stereotype.Similarly,thoughto a lesser degree,
In the groceryproductsreplicate,as predicted,therange negativeconsequencesare impliedby atypicalbrandex-
of attitudeis greaterfor narrowthan for broadbrands. tensions.For example,an atypicalbrandextensionmay
Forexample,brandextensionsthatwere the sameas the be regardednegativelyif it is perceivedto requireex-
brand'scurrentproductswere perceivedas more desir- pertiseoutsidethe originatingcompany.
able when the brandwas narrow(mean = 12.22) than
when the brandwas broad(mean = 11.44). In contrast, Brand Breadth
brandextensionsthatweremostdissimilarto the brand's Resultsindicatethatbrandbreadthinteracted withbrand
currentproductswere perceivedas less desirablewhen extensiontypicality.Specifically,when the brandmade
the brandwas narrow(mean = 5.59) thanwhen it was a varietyof products,an extensionthat was essentially
broad(mean = 5.79). In bothreplicatesthe attituderat- the same as a currentproductwas perceivedas not as
ings werehigherfor moderatelytypicalbrandextensions typicalas when the brandmade only one type of prod-
whenthe brandwas broadthanwhenit was narrow.(For uct. However, greaterbreadthincreasedthe perceived
the moderatelytypicalextensioncannedfruit, the mean typicalityof moderatelydiscrepantextensions.The pre-
is 10.53 for the broadbrandand 9.76 for the narrow ceding discussionmay imply that a narrowbrandsuch
brand.For the camera,the meanis 10.83 for the broad as Campbell'shas an advantageover a broaderbrand
brandand 10.00 for the narrowbrand.)Apparently,ex- such as Heinz in offeringa new soup, but Heinz has an
tensionsthatare somewhatdifferentfromthe currentof- advantageover Campbell'sin offeringa moderatelydif-
ferings are consideredmore acceptableif the company ferentextensionsuchas a new line of frozenvegetables.
has alreadyextendedto differentproducts. Finally, brandbreadthhad little effect on the perceived
Finally, attitudeswere plottedagainstperceivedtyp- typicalityof extremelydiscrepantextensions,suggesting
icalityratings.However,note thatbecausebrandbreadth that breadthand typicalityare relative, ratherthan ab-
had the same effect on typicalityratingsas on attitude solute, magnitudes.For any particularlevel of brand
ratings,the plots in Figure1, E andF, disguisethe effect breadth,some productsare so discrepantas to makethe
of brandbreadth. brandcategoryseem relativelynarrow.Therefore,nei-
ther a narrowbrandsuch as Campbell'snor a broader
DISCUSSION brandsuchas Heinzwouldhavean advantagein offering
Consistentwith conceptualframeworksdevelopedin an extremelydiscrepantextensionsuch as toothpaste.
cognitive and social psychology, brandextensiontypi- A narrowbrandmay not always have an advantage
cality and brandbreadthhad significanteffects on the whenofferingnew productsthatareessentiallythe same
evaluationprocessesandevaluationoutcomesassociated as currentones. Results in the electronicproductsrep-
with a potentialbrandextension.Thesefindingsandim- licateindicatethatthe extensionsof the broadbrandgen-
plicationsfor categorizationmodels that examineeval- erallywereperceivedmorefavorablythanthe extensions
uation processesare addressedin the sections that fol- of the narrowbrand.In the replicatein whichbrandnar-
low. rowness seemed disadvantageous,the brand's current
BrandExtensionTypicality productswere calculators,television sets, and digital
watches.Perhapsa brandthatmadeonly digitalwatches
Both replicatesof our study show a direct linearre- or calculatorswas viewed as overspecialized.Recent
lationshipbetweentypicalityandattituderatingsfor po- findings (Aaker and Keller 1989) substantiatethe im-
tentialbrandextensions. As predicted,attitudetransfer portanceof perceivedexpertisein positive evaluations
from the currentbrandto potentialextensionsdid not of brandextensions.Possiblygroupsof relatedproducts
occurin an all-or-nothingway. The negativeattitudeto- fortifya brandimage,particularly one in whichtechnical
wardextremelyunlikelyor atypicalproductsthata brand expertisehas an importantrole.
mightmakewentbeyondthe failureof a positiveattitude Evaluation Processes
to "ruboff" on the new product.Particularlyin the gro-
cery productsreplicate,subjectsseemedto find atypical The evidence reported here is consistent with an eval-
brandextensionsfundamentally wrongfor the brand.This uation model that includes both categorical and piece-
result is differentfrom what has been predictedin the meal processes. As hypothesized, categorical processes
person perceptionliteraturefor schematicfit. Specifi- predominated when consumers evaluated products that
cally, Fiske (1982) positedthatwhen a persondoes not were "the same" as the brand's current products and
fit a particularschema, the affectiveresponseis neutral piecemeal processes predominated when consumers
or slightly positive pendingidentificationof the person evaluated products that were moderately different from
as a good fit with anotherschema. the brand's current products. However, there are also
One possible reasonfor this differencein predictions several deviations from the predicted U-shaped relation-
26 JOURNAL
OF MARKETING 1991
FEBRUARY
RESEARCH,

ship betweenbrandextensiontypicalityand the process the experimentaltask representedthe real issues inves-
measures. tigated. The primarydifferencesbetween the experi-
Responsetimes for extensionsof narrowbrandsde- mentaltask and an evaluationof a truebrandextension
viate from the predictedcurvilinearpatternin that ex- are that experimentalevaluations(1) did not occur in a
tremelyatypicalextensionsof narrowbrandstook longer store as they often would in consumerdecision situa-
to evaluatethandid moderatelytypicalextensions,a re- tions, (2) werebasedon vicariousexperienceratherthan
sult consistentwith the Fiske andPavelchakmodel. Per- personalexperiencewith productsin use, and (3) were
haps subjectswere somewhatconfusedby the prospect basedon information presentedin exclusivelyverbalform
of a companyextendingits brandfrom, say, condiments ratherthanas a reactionto othertactile and visual cues
to toothpaste.Atypicalextensionsof broadbrandsmay suchas wouldoccurin mostproductevaluationsettings.
have evoked less confusionbecause the brandat least In addition,thereis a strongpossibilitythatthe experi-
had alreadydemonstratedexpertisein multipleproduct mentaltaskprimedthe typicalitygradientrepresentedby
lines. potentialbrandextensions (it was probablyevident to
The relationshipbetweentypicalityandresponsetime subjectsthat potentialbrandextensionsvariedin simi-
is muchflatterfor narrowbrandsin the groceryproducts larityto currentbrandedproducts).
replicatethan in the electronicproductsreplicate.One A secondlimitationof the studyis thatwhenresponse
explanationfor this differenceis the types of products timesandverbalprotocolsare collectedconcurrently they
tested.Subjectsmay haveevaluatedgroceryproductsand may becomeconfounded.Thatis, the act of verbalizing
electronicproductsvery differently.Because electronic probablylengthensresponsetime. Thoughtverbalization
productsare more technicallycomplex, the evaluative may either accentuateor attenuatethe extent of piece-
criteriafor a brandextensionmay be based on the per- meal processingas measuredby responsetime because
ceivedexpertiseof the manufacturer. However,the eval- all verbalresponsestake time and only some of them
uationof a groceryproductbrandextensionmaybe based indicatepiecemealprocessing.
moreon subjectivecriteriafor the appropriatenessof us-
ing the samebrandon productswith extremelydifferent Concluding Remarks
attributes.Perhapsbrandsthatmaketechnicallycomplex Despite the limitations,the resultsof our experiment
products(calculators,television sets) encourage con- reveal some interestingaspects of the relationshipbe-
sumersto scan their memoriesmore thoroughlyfor a tween categorystructureand the process of evaluating
way to matchthe brandextensionwith a featureof the potentialnew membersof the category.The strengthsof
brand'scurrentproducts.Futureresearchmightexplore the researchmethodinclude the experimentalmanipu-
this issue further. lation of brandbreadthand the comparisonof response
A notableaspectof the responsetimes and cognitive timesandverbalprotocolmeasuresacrosstwo replicates
responsesin the electronicproductsreplicateis thatthe of the experiment.The results suggest that category
extremelyatypicalproduct(ballpointpen) took as long structurehelpsto shapeevaluationprocessesandthatat-
to evaluateas the moderatelytypical products(camera titudeand typicalityare closely linked.
and refrigerator)but did not elicit piecemealcognitive To the extent that the experimentaltask mirroredas-
responsesin nearlyas many subjects.This findingsug- pects of an actual productevaluation,the results also
gests thatthe two processindicatorsmay have measured provideinsightinto the way brandextensionevaluation
differentaspectsof piecemealprocessing.The response is influencedby brandextension typicalityand brand
times seem to have measuredthe difficultyof finding categorybreadth.Specifically,brandextensionsappear
pointsof similarityor differencebetweenthe brandex- to be evaluatedmore highly to the extent that they are
tensionand currentproducts,whereasthe cognitivere- perceivedas similarto (or typicalof) the brand'scurrent
sponsesmeasuredthefrequencywithwhichrelevantpoints productoffering.The perceivedtypicalityof a potential
of similarityor differencewere found. Apparently,sub- extensionis influenced,in turn, by the breadthof the
jects scannedtheirmemoriesfor points of similarityor brand's currentproductoffering. Narrowbrand cate-
differencebetweencurrentproductsandthepotentialbrand gories appearto have advantagesfor extensionsthatare
extension,and verbalizedany thatseemedrelevant.Re- highlysimilar,andbroadbrandcategoriesappearto have
sponse time is arguablythe best measureof the extent advantagesfor extensionsthataremoderatelysimilar,to
of the evaluative processing because it includes the silent the brand's current products. In our study, broad brand
search of memory. In this context, verbal protocols may categories included a greater variety of products than did
be more useful to characterize the nature of processing narrow brand categories. This variety may be specified
than to measure its extent. better as having both range (operationally defined as the
difference between the two most different products) and
Limitations
variability (operationally defined as a summary measure
The strategy in our study was to manipulate structural of interproductdifferences, such as pairwise differences
variables that seemed likely to influence evaluation. As squared and summed). Future research might investigate
with all laboratoryexperiments, interpretationof the re- range and variability as separate factors of brand cate-
sults must be tempered by an understandingof how well gory breadth.
EVALUATION
BRANDEXTENSION 27

It is significant that some potential brand extensions tions to Social Perception," in Affect and Cognition: The
were perceived as moderately typical rather than either 17th Annual Carnegie Symposium on Cognition, Margaret
extremely typical or extremely atypical of the brand cat- S. Clarkand SusanT. Fiske, eds. Hillsdale,NJ: Lawrence
ErlbaumAssociates, 171-90.
egory. These moderately typical extensions appeared to
- andMarkA. Pavelchak(1986), "Category-Based Ver-
generate greater thought and processing time than either
sus Piecemeal-BasedAffectiveResponses:Developmentsin
extremely typical or extremely atypical extensions. Per-
Schema-Triggered Affect," in The Handbook of Motivation
haps they would be aided by promotions, and other mar- and Cognition: Foundation of Social Behavior, Richard W.
keting devices, that take into account this greater pro- Sorrentinoand E. Tory Higgins, eds. New York:Guilford
cessing. Press, 167-203.
Our findings suggest that categorization theory pro- Geisser, Seymourand Samuel W. Greenhouse(1958), "An
vides a promisingframeworkfor furtherresearchon brand Extensionof Box's Resultson the Use of F Distributionin
extension. By better understandingthe structureof brand Multivariate Analysis," Annals of Mathematical Statistics,
categories and the way differences (or changes) in that 29 (April), 885-91.
structureaffect brand attitudes, marketerscan better an- Judd,CharlesM. and CynthiaM. Lusk (1984), "Knowledge
Structuresand EvaluativeJudgments:Effects of Structural
ticipate the effects of brand extension.
Variableson JudgmentalExtremity,"Journalof Personality
REFERENCES and Social Psychology, 46 (6), 1193-207.
Kirk, Roger E. (1982), Experimental Design: Procedures for
Aaker, David A. and Kevin Lane Keller (1989), "Consumer the Behavioral Sciences. Monterey, CA: Brooks/Cole Pub-
Evaluations of Brand Extensions," Journal ofMarketing, 54 lishing Company.
(January),27-41. LaTour,StevenA. andPaulW. Miniard(1983), "TheMisuse
Barsalou,LawrenceW. (1983), "AdHoc Categories,"Mem- of RepeatedMeasuresAnalysis in MarketingResearch,"
ory and Cognition, 11 (3), 211-27. Journal of Marketing Research, 20 (February), 45-57.
- (1985), "Ideals,CentralTendency,and Frequencyof Linville, PatriciaW. (1982), "TheComplexity-Extremity
Ef-
Instantiationas Determinantsof GradedStructurein Cate- fect and Age-BasedStereotyping,"Journal of Personality
gories," Journal of Experimental Psychology: Learning, and Social Psychology, 42 (2), 193-211.
Memory and Cognition, 11 (4), 629-48. Loken, Barbaraand JamesWard(1987), "Measuresof Attri-
(1990), "On the Indistinguishabilityof Exemplar buteStructureUnderlyingProductTypicality,"in Advances
Memoryand Abstractionin CategoryRepresentation,"in in Consumer Research, Vol. 14, Melanie Wallendorf and
Advances in Social Cognition, Vol. 3, T. K. Srull and R. Paul F. Anderson,eds. Provo, UT: Associationfor Con-
S. Wyer, Jr., eds. Hillsdale, NJ: LawrenceErlbaumAs- sumerResearch,22-8.
sociates. Malt, BarbaraC. (1989), "An On-LineInvestigationof Pro-
Bettman,JamesR., Noel Capon,and RichardJ. Lutz(1975), totype and ExemplarStrategiesin Classification,"Journal
"Cognitive Algebra in MultiattributeAttitude Models," of Experimental Psychology: Learning, Memory and Cog-
Journal of Marketing Research, 12 (May), 151-64. nition, 15 (4), 539-55.
Boush, David, ShannonShipp, BarbaraLoken, Ezra Genc- Mervis, CarolynB. and EleanorRosch (1981), "Categoriza-
turk,SusanCrockett,Ellen Kennedy,BettyMinshall,Den- tion of Natural Objects," Annual Review of Psychology, 32,
nis Misurell,LindaRochford,andJon Strobel(1987), "Af- 89-115.
fect Generalization to Similar and Dissimilar Brand Parducci,Allen and Douglas H. Wedell (1986), "The Cate-
Extensions," Psychology and Marketing, 4 (3), 225-37. gory Effect With Rating Scales: Number of Categories,
Brewer, MarilynnB. (1988), "A Dual Process Model of Numberof Stimuli, and Methodof Presentation,"Journal
Impression Formation," in Advances in Social Cognition, of Experimental Psychology: Human Perception and Per-
Vol. 1, T. K. SrullandR. S. Wyer, Jr., eds. HillsdaleNJ: formance, 12 (4), 496-516.
LawrenceErlbaumAssociates, 1-36. Rosch, Eleanorand CarolynB. Mervis (1975), "FamilyRe-
Brooks, Lee (1978), "NonanalyticConcept Formationand semblances:Studiesin the InternalStructureof Categories,"
Memory for Instances," in Cognition and Categorization, Cognitive Psychology, 7 (4), 573-605.
E. Rosch and B. B. Lloyd, eds. Hillsdale, NJ: Lawrence - , CarolSimpson,and R. Scott Miller (1976), "Struc-
ErlbaumAssociates, 169-215. tureBases of TypicalityEffects,"Journalof Experimental
Burke,MarianChapmanandJulieA. Edell (1989), "TheIm- Psychology: HumanPerception and Performance, 2 (4), 491-
pactof Feelingson Ad-BasedAffect andCognition,"Jour- 502.
nal of Marketing Research, 26 (February), 69-83. Smith,EdwardE., EdwardJ. Shoben,andLanceJ. Rips(1974),
Cohen,Joel B. (1982), "TheRole of Affect in Categorization: "Structureand Process in SemanticMemory:A Featural
Towardsa Reconsideration of the Conceptof Attitude,"in Model for SemanticDecisions,"PsychologicalReview, 81
Advances in Consumer Research, Vol. 9, Andrew Mitchell, (3), 214-41.
ed. Ann Arbor, MI: Associationfor ConsumerResearch, Srull, ThomasK. andRobertS. Wyer(1989), "PersonMem-
94-100. ory and Judgment,"PsychologicalReview, 96 (1), 58-83.
Collins, AllanM. andElizabethF. Loftus(1975), "A Spread- Sujan,Mita (1985), "ConsumerKnowledge:Effects on Eval-
ing ActivationTheoryof SemanticProcessing,"Psycholog- uationStrategiesMediatingConsumerJudgments,"Journal
ical Review, 82 (6), 407-28. of Consumer Research, 12 (June), 31-46.
Fishbein,Martinand Icek Ajzen (1975), Belief, Attitude,In- - and ChristineDekleva (1987), "ProductCategoriza-
tention and Behavior. Reading, MA: Addison-Wesley Pub- tion andInference-Making:
Some Implicationsfor Compar-
lishing Company. ative Advertising," Journal of Consumer Research, 14 (De-
Fiske, SusanT. (1982), "Schema-Triggered
Affect: Applica- cember, 372-8.
28 OF MARKETING
JOURNAL 1991
FEBRUARY
RESEARCH,

Tauber,EdwardM. (1988), "BrandLeverage:Strategyfor Wright,PeterL. (1976), "An AdaptiveConsumer'sView of


Growthin a Cost ControlWorld,"Journalof Advertising AttitudesandChoiceMechanismsas Viewedby an Equally
Research,31 (August/September),26-30. Adaptive Advertiser," in Attitude Research at Bay, William
Tversky,Amos (1977), "Featuresof Similarity,"Psycholog- D. Wells, ed. Chicago:AmericanMarketingAssociation,
ical Review, 84 (4), 327-52. 113-311.
Ward,Jamesand BarbaraLoken(1986), "TheQuintessential
SnackFood: Measurementof ProductPrototypes,"in Ad-
vances in Consumer Research, Vol. 13, Richard Lutz, ed.
Provo, UT: Associationfor ConsumerResearch,126-31. ReprintNo. JMR281101

fIY1
~L~I0 0

1 bl

publication are now avai a e ro

Yes!I wouldliketo knowmoreaboutUMIArticle Name


Clearinghouse.I aminterestedin electronicordering
Title
throughthe followingsystem(s):
Institution/Company
D DIALOG/Dialorder D ITTDialcom
O OnTyme O OCLC ILLSubsystem Department
EI Other (pleasespecify) Address
El I am interestedin sendingmyorderbymail.
City State Zip
l Pleasesendmeyourcurrentcataloganduser
instructionsforthe system(s)I checkedabove. Phone
Mailto: University International
Microfilms
300 NorthZeebRoad,Box91 AnnArbor,MI48106

Das könnte Ihnen auch gefallen