Role of brand-specific associations in brand extension

MISSING IMAGE

Material Information

Title:
Role of brand-specific associations in brand extension
Physical Description:
vii, 214 leaves : ill. ; 29 cm.
Language:
English
Creator:
Broniarczyk, Susan M., 1965-
Publication Date:

Subjects

Subjects / Keywords:
Branding (Marketing)   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1992.
Bibliography:
Includes bibliographical references (leaves 210-213).
Statement of Responsibility:
by Susan M. Broniarczyk.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001751615
notis - AJG4551
oclc - 26529458
System ID:
AA00003689:00001

Full Text










ROLE OF BRAND-SPECIFIC ASSOCIATIONS IN BRAND EXTENSION


By

SUSAN M. BRONIARCZYK


















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

1992


U~:R.R~ITY














ACKNOWLEDGEMENTS


I feel a tremendous sense of relief as I reach closure on

the dissertation. My energies were so directed at attaining

this single goal that it became an abstraction that never

quite seemed possible. I am deeply indebted to many people who

through their encouragement and support helped me overcome my

mental roadblocks.

One constant source of support was Joe Alba to whom I can

never express my gratitude. This dissertation and my career

are a testament to his concern for my scholarly and personal

growth. In essence, he took me by the hand and taught me to be

a researcher. That simple statement was a huge undertaking of

his time and energy that always seemed in endless supply. He

was a source of strength when I was stressed out and believed

in me when I often did not believe in myself. I could not have

asked for a better friend, colleague, or mentor.

I am grateful to Professor John Lynch for his

encouragement and guidance on statistical analyses, and

moreover, for the enthusiasm towards research that he shares

with all those who are fortunate to come in contact with him.

Thanks are also offered to Professors Wes Hutchinson and Bart

Weitz for their guidance in refining my ideas and positioning

this work.






I would also like to express my sincere thanks to my

family and friends in Chicago who loved me enough to encourage

me to pursue my dreams. They surrounded me with their love and

support and were always there when I needed them.

Finally, I would like to thank my fellow doctoral

students, especially Cindy Copp and Michelle DeMoss, who

helped make Gainesville home.


iii













TABLE OF CONTENTS



ACKNOWLEDGEMENTS .... . .ii

ABSTRACT . ... . vii


CHAPTERS

I INTRODUCTION . 1

II CONCEPTUAL BACKGROUND ON THE ROLE OF BRAND-
SPECIFIC ASSOCIATIONS IN THE EVALUATION OF
BRAND EXTENSIONS . 4

Brand-Specific Associations . 5
Attitude Toward the Brand . 8
Brand-Specific Associations versus Category
Similarity . 10
Brand Knowledge . 12

III STIMULUS DEVELOPMENT . ... .. 13

Overview . . .. 13
Pretest 1 .. . 14
Pretest 2 . . 15
Pretest 3 ... . .. 17
Pretest 4 . . 24
Manipulation Check 1 . 30
Manipulation Check 2 . .. 31

IV EXPERIMENT 1: AN EMPIRICAL INVESTIGATION OF
THE ROLE OF BRAND-SPECIFIC ASSOCIATIONS AND
BRAND AFFECT ON THE EVALUATION OF BRAND
EXTENSIONS . . 39

Overview . . 39
Experimental Design . .. 40
Experimental Procedure . .. 41
Results . . 43






V EXPERIMENT 2: AN EMPIRICAL INVESTIGATION OF
THE ROLE OF BRAND-SPECIFIC ASSOCIATIONS AND
PRODUCT CATEGORY SIMILARITY ON THE
EVALUATION OF BRAND EXTENSIONS .. 59

Overview . . 59
Stimulus Materials . .. 62
Experimental Design . .. 68
Experimental Procedure . .. 69
Results . . 71

VI EXPERIMENT 3: AN EMPIRICAL INVESTIGATION OF
THE MODERATING ROLE OF EXPERTISE ON BRAND-
SPECIFIC ASSOCIATIONS AND BRAND AFFECT IN
THE EVALUATION OF BRAND EXTENSIONS 86

Stimulus Materials. . ... 86
Experimental Design. . 88
Experimental Procedure . .. 88
Results . . 89

VII EXPERIMENT 4: AN EMPIRICAL INVESTIGATION OF
THE MODERATING ROLE OF EXPERTISE ON BRAND-
SPECIFIC ASSOCIATIONS AND PRODUCT CATEGORY
SIMILARITY IN THE EVALUATION OF BRAND
EXTENSIONS . . 95

Overview . . .95
Experimental Design. . 98
Experimental Procedure . .. 99
Results . . .. 100

VIII GENERAL DISCUSSION . ... 107

APPENDICES

I DETAILED LISTING OF TASKS FOR PROTESTS 114

II DETAILED LISTING OF TASKS FOR EXPERIMENT 1 152

III DETAILED LISTING OF TASKS FOR EXPERIMENT 2 163

IV DETAILED LISTING OF TASKS FOR EXPERIMENT 3 .194

V DETAILED LISTING OF TASKS FOR EXPERIMENT 4 .201

REFERENCE LIST ........ . 210

BIOGRAPHICAL SKETCH ........ .. 214












Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

ROLE OF BRAND-SPECIFIC ASSOCIATIONS IN BRAND EXTENSION

By

Susan M. Broniarczyk

May 1992

Chairperson: Joseph W. Alba
Major Department: Marketing

The dissertation investigates the role of brand-specific

associations in brand extensions. Existing research has

examined the determinants of successful brand extensions at

the product category level. Empirical results have found that

brand affect and product category similarity influence

consumer perceptions of brand extensions.

This dissertation postulates that brand-specific

associations moderate the effect of brand affect and product

category similarity on brand extension judgments. Brand-

specific associations are defined as associations that

differentiate a brand from its product level and other brands

in its category. It is suggested that brands have value

because of these specific associations and that consumer

judgments of extensions are influenced by whether a brand's

association is relevant in an extension category.






Four experiments are conducted to examine the impact of

brand-specific associations on brand extensions. The first

study examines the moderating role of brand-specific

associations on brand affect and provides evidence that brand-

specific associations may lead to preference reversals from

the original to the extended category. The second experiment

examines the moderating role of brand-specific associations on

product category similarity and shows that brands may extend

to physically dissimilar product categories.

The last two experiments investigate how knowledge of the

brand mediates the role of brand-specific associations in

brand extensions. These studies find that brand-specific

associations moderate the effect of brand affect and product

category similarity on brand extension judgments only for

consumers high in brand knowledge.

Taken together, the results indicate very strong

interactive effects among brand-specific associations and

previously identified determinants of extension. In doing so,

the results highlight the importance of recognizing the

specific associations of brands when evaluating opportunities

for extension.


vii












CHAPTER I
INTRODUCTION


Launching a new product is often a risky endeavor due to

its high costs and low probability of success (Tauber, 1988).

One strategy for improving the odds involves extending an

existing brand name to the new product. The familiarity of the

name may reduce the amount of advertising needed to generate

awareness, and the reputation of the brand may enhance initial

evaluation by consumers. Such intangibles constitute brand

equity (Farquhar, 1989) and reinforce the notion that a brand

name represents more than the functional attributes of a

product (Jones, 1986).

Tauber (1981) identifies two possible brand name

extensions--a line extension and a franchise extension. A line

extension occurs when the brand name is extended by adding new

flavors, sizes, or varieties within the existing product

category. A franchise extension occurs when the brand name is

extended to a different product. Whereas applying an existing

name to a line extension is ubiquitous and intuitively

appealing, franchise extensions are less obvious and produce

less predictable results. Specifically, the greater upside

potential of a brand extension must be weighed against a

higher likelihood of failure and potential negative impact on






2

the brand name itself (Jones, 1986; Tauber, 1988). This

dissertation will examine franchise extensions.

Despite the benefits and risks of brand extension and its

long tradition, research on the topic is in its infancy.

Initial studies have followed the natural tendency to

determine at a general level which variables are associated

with successful brand extension. The preliminary findings of

this research serve as partial motivation for this

dissertation. Specifically, two factors have emerged that

appear to account for extendibility of a brand name:

similarity of the original and extended product classes and

consumer attitude toward the brand in question (Aaker and

Keller, 1990; Boush and Loken, 1991; University of Minnesota

Consumer Behavior Seminar, 1987).

At one level these conclusions seem plausible and

straightforward, and they undoubtably affect consumer reaction

to brand extensions. However, research and intuition suggest

a more complicated picture. Prior research has mainly

addressed product category effects and ignored brand effects.

This dissertation attempts to demonstrate that brand-specific

associations moderate the impact of these product category

effects on consumer judgments of brand extensions.

Chapter II begins with a survey of the literature on

brand extensions and brand-specific associations. The role of

brand-specific associations in brand extensions is

conceptualized. Hypotheses are developed that involve the







3

interaction of brand-specific associations and brand affect

and product category similarity. The influence of consumer

brand knowledge is also discussed.

Chapter III presents a series of protests for stimulus

development. Chapter IV presents the method and results of an

experiment designed to test the interaction of brand-specific

associations and brand affect. Chapter V presents the method

and results of an experiment designed to test the interaction

of brand-specific associations and product category

similarity.

The next two chapters examine the moderating role of

brand knowledge on the influence of brand-specific

associations in brand extensions. Chapter VI presents the

method and results of an experiment designed to examine the

effect of brand knowledge on the influence of brand-specific

associations and brand affect in extension judgments. Chapter

VII presents the method and results of an experiment designed

to test the effect of brand knowledge on the influence of

brand-specific associations and product category similarity in

extension judgments.

Chapter VIII provides a general discussion of the

findings and the implications of these results for research on

brand extensions. Managerial implications of these findings

and directions for future research are also discussed.












CHAPTER II
CONCEPTUAL BACKGROUND ON THE ROLE OF BRAND-SPECIFIC
ASSOCIATIONS IN THE EVALUATION OF BRAND EXTENSIONS



Overview


The value of a brand name may be defined not only in

terms of the advantages it provides in its current competitive

arena but also in terms of the potential advantages it offers

in untapped markets (Srivastava and Shocker, 1991). This

realization has led to increasing reliance on brand extension

as a strategy for achieving corporate growth (Springen and

Miller, 1990). The rationale underlying this new product

strategy is that extended brands offer enhanced profit

potential as a result of lower introductory costs and higher

consumer acceptance (Tauber, 1988).

The increase in brand extensions has led to a recent

upsurge in research, particularly with respect to consumer

reactions to brand extensions. Specifically, there is now much

interest in identifying those factors that moderate consumer

evaluations of the extended brand. Initial research has been

exploratory and attempted to isolate at a general level the

variables that influence perceptions of an extension (Aaker

and Keller, 1990; MacInnis and Nakamoto, 1990). Initial

results suggest that attitude toward the brand and product

4






5

category similarity are factors that influence the

extendibility of a brand.

Other research has been more concerned with isolating the

processes that underlie consumer reactions and consequently

has used a more experimental approach. Ironically, however,

these studies have either explicitly framed brand effects out

of the equation by using fictitious brand names (Boush and

Loken, 1991; Keller and Aaker, 1992; University of Minnesota

Consumer Behavior Seminar, 1987) or have been designed in such

a way that brand effects cannot be isolated from product class

effects (Aaker and Keller, 1990; Chakravarti, MacInnis and

Nakamoto, 1990).

Yet, brand-specific associations are an integral part of

a brand's strength for leveraging. This paper focuses on the

role of brand effects in brand extensions.


Brand-Specific Associations


A brand-specific association is defined as an association

that differentiates a brand from its product level and other

brands in its category (Chakravarti, MacInnis, and Nakamoto,

1990; Srinivasan 1979). Brands have value because of their

associations to specific attributes or benefits. Marketing

efforts to differentiate a brand may generate an association

not included in the product category representation or make

the association stronger for a brand than its category level

or other brands in its category (Sujan and Bettman, 1989).







6
Brand-specific associations may include physical attributes

(e.g., Pringles potato chips come in a can), benefits (e.g.,

Pringles' packaging is resealable), usage situations (e.g.,

Pringles are easy to pack for trips), or users (e.g., Pringles

are for light snack eaters). Although a brand may solely

possess an association, a subset of brands in a category often

have similar positioning as evidenced by market segmentation.

Because firms invest considerable resources to position their

brands in consumer minds (Ries and Trout, 1981), it seems

reasonable that they would want to leverage this asset in

potential extensions.

Furthermore, the fact that consumers think in terms of

benefits is consistent with theory in marketing and cognitive

psychology. Benefit segmentation is based on the principle

that people seek the benefits that products provide rather

than the physical product itself. Day, Shocker, and Srivastava

(1979) posit that to identify market structures from a

consumer perspective, product markets need to be broadly

defined as the set of product substitutes for a specific usage

situation in which similar benefits are desired.

Additionally, research on categorization suggests that

these positioning may cut across product categories. At some

level of abstraction, brand-specific associations help

consumers achieve goals (Park and Smith, 1989), and seemingly

unrelated products may be categorized together if they share

a common goal (Barsalou, 1983, 1985). As a need arises,






7

individuals are quite adept at forming goal-derived

categories, and recent evidence suggests that such categories

may be established in memory (Barsalou, 1991). In fact, in a

marketing context, Ratneshwar and Shocker (1991) find that

consumers have established memory structures for one type of

brand-specific association, usage situations. Thus, if a brand

association is relevant in another product category, a brand

extension into that category is likely to be viewed favorably

because the product benefits desired are already associated

with the brand name.

As noted, existing research has curiously ignored this

brand effect by examining brand extensions from a product

class level (Boush and Loken, 1991; University of Minnesota

Seminar, 1987), using fictitious brand names (Keller and

Aaker, 1992), or confounding brand and product category

effects by only extending one brand from each product category

(Aaker and Keller, 1990; Chakravarti et al., 1990). Some

research has recognized brand effects but has found little

positive effect (MacInnis and Nakamoto, 1990; Rangaswamy,

Burke, and Oliva, 1990). In these cases, however, extension

categories were not selected to be congruent with the brand-

specific association.

Fortunately, one study has examined brand-specific

associations and their importance in determining extension

categories (Park, Milberg, and Lawson, 1991). They examined

two broad types of brand associations and found that if the







8
brand-specific association was relevant in the extension

category, evaluations of the brand extensions were more

favorable. Consistent with Park et al. (1991), this paper

takes the strong position that category effects should not be

isolated from brand effects. That is, the effect of brand-

specific associations may be so pervasive and overwhelming

that main effects of brand attitude and product category

similarity may be misleading. In the present research,

multiple brand-specific associations are examined, and

moreover, are directly compared with the other factors

previously reported to be important moderators of

extendibility.


Brand-Specific Associations Versus Attitude Toward the Brand


One possible mechanism by which the extended brand is

thought to be evaluated is a transfer of the brand's affect to

the extended product. This affect transfer is thought to occur

only if the original and extended product categories are

similar. In support of this, Aaker and Keller (1990) find that

a product's quality primarily transfers to the extended

product if the new product is deemed a substitute or

complement of the original product and Boush and Loken (1990)

find a direct relationship between product category similarity

and affect transfer.

Although this process may occur, the affect transfer

model (Fiske and Pavelchak, 1986) involves a process different






9

than that proposed in brand extensions. In the Fiske and

Pavelchak model, an object is compared to objects in another

category, and if object similarity is high, attitude toward

the category is transferred to the object. In brand extension,

two categories are supposedly compared, and if similarity is

high, attitude from the object is transferred. This

misconception does not rule out affect transfer, albeit by a

different route, but leaves open the possibility that the

extended product is evaluated based on its attributes

regardless of product category similarity. This would seem to

be especially true if, unlike the aforementioned studies,

brands had salient associations that were relevant in the

extension category. Park et al. (1991) did not investigate the

process of extended brand evaluation and purposively

controlled for brand affect.

The preceding logic suggests that when purchasing a

product, the consumer is trying to satisfy a need and this

drives their selection among alternative brands. If an

extended brand's associations satisfy this need, the brand

should be favorably evaluated. Thus, an extended brand may be

evaluated based on whether its association is relevant in the

extended category and its brand affect. Affect transfer

implies that the only reason brands from the same original

product category would have differential prospects for success

in the extended category is that some brands engender greater

positive affect. However, based on goal relevance, it is







10
predicted that there will be a brand-specific association X

brand affect interaction. The strongest form of this

interaction suggests that the preference ordering fr9m the

original to the extended product category may be reversed when

the brand-specific association of a less preferred brand is

relevant in the extension category. Experiment 1 examines

brand-specific associations and brand affect in extensions.


Brand-Specific Associations versus Category Similarity


Product category similarity has face validity as a factor

in brand extensions. However, Murphy and Medin (1985) argue

that the standard definition of similarity as the degree of

feature overlap (Tversky, 1977) is inadequate because it

ignores the relevance of needs, goals, and theories. They

propose that the presence of a single relational match may

produce a high similarity judgment if the relation is

important in a product category. Here, similarity is not a

blind mapping of features but rather the search for

connections between otherwise dissimilar objects, as with

noncomparable alternatives (Johnson, 1984). We argue that

there is no necessary reason to expect consumers to make

judgments of overall similarity when evaluating the extended

product. When connections are present, the individual must

comprehend them, and from a managerial perspective, be

persuaded by them. When connections are nonobvious, inference

may be required.







11

The results on product category similarity in brand

extensions have been equivocal. Some have found that product

category similarity and brand extension evaluation are

positively related (Aaker and Keller, 1990; Boush and Loken,

1991; MacInnis and Nakamoto, 1990; University of Minnesota

Consumer Behavior Seminar, 1987), and others have found little

or no effect (Keller and Aaker, 1992; Park et al., 1991). The

most pertinent finding is from Park et al. (1991) who found

that product category similarity seemed to have less effect on

a prestige-oriented brand than a functional brand, perhaps

because the abstract prestige image connection was easier to

make. However, they did not directly compare brand-specific

associations with product category similarity to see which a

manager should be more concerned about when extending a brand.

Moreover, their results imply that nonimage brand effects may

be limited.

If the brand-specific association is salient and relevant

in the extension category, consumers should respond to the

congruity of two products sharing the same name, regardless of

the type of association. Therefore, it is predicted that not

only will there be a brand-specific association X product

category similarity interaction, but that brand-specific

associations will dominate product category similarity if the

association is important in the extended category. Experiment

2 directly compares the roles of brand-specific associations

and product category similarity in extension judgments.









Moderating Effect of Brand Knowledge


Knowledge of the brand is obviously necessary for brand

effects to influence extension judgments. Although familiarity

with the brand name may produce some brand effects, a

knowledge of the brand-specific association is required for

consumers to make the connection that the brand benefit is

relevant for the extension category.

The likelihood that consumers will make these inferences

increases with knowledge (Alba and Hutchinson, 1987). More

important, knowledgeable consumers are more likely to be

analytic when judging the congruity between an original and

extended brand (cf. Alba and Hutchinson, 1987; Brewer, 1988;

Muthukrishnan and Weitz, 1991).

If the foregoing logic is true, then brand knowledge

should moderate the effects of brand-specific associations in

brand extensions. Thus, if brand knowledge is low, one would

expect main effects of brand affect and product category

similarity to predominate in brand extensions, whereas if

brand knowledge is high, one would expect brand effects to

interact with these determinants. Experiments 3 and 4 examine

the moderating effect of brand knowledge on the use of brand-

specific associations in extension evaluations.













CHAPTER III
STIMULUS DEVELOPMENT


Overview


The reasoning developed in the last chapter concerned how

brands may have specific associations that differentiate them

from their product category level and other brands in their

product category. These brand-specific associations are

expected to interact with brand affect in extension judgments.

Furthermore, brand-specific associations may allow a focal

brand to be more preferred in an extended product category

than other brands, even though the other brands are more

preferred in the original product category.

This chapter discusses four protests required to test the

hypotheses and two posttests that served as manipulation

checks. The protests were designed to identify a comprehensive

set of brands that

(a) were familiar to all subjects,

(b) had specific associations that were highly salient

but not image-based,

(c) had associations that differentiated them from the

their product categories and other brands in those

categories, and

(d) had not been extended previously.

13






14

The protests sequentially narrow the set of possible

brands for experimental testing. The first and second protests

involve identifying brands with specific associations. The

third pretest examines brand affect and image ratings. The

fourth pretest generated potential extensions for the brands.

Two tests that served as manipulation checks were conducted

after the first experiment and are reported at the end of this

chapter.

Pretest 1


The purpose of this pretest was to identify all brands

that would meet the aforementioned requirements. The first

required identification of a set of product categories and

examination of the brands within each category to determine if

they met the constraints.

A comprehensive set of 119 product categories along with

the major brands within each was gathered from examining a

retail catalog, Simmons market research data, a supermarket

inventory list, and other catalogs. The categories include

durables, clothing, food products, healthcare products, and

services. An entire listing of categories and brands is

presented in Appendix IA-1. This set was examined to determine

product categories that contained brands with specific

associations.

Specifically, three judges examined the list of product

categories. The judges were instructed to examine each product

category to determine if it contained at least two familiar






15

brand names with specific, nonoverlapping associations that

had not been multiply extended. The exact instructions to the

judges are detailed in Appendix IA-2.

The three judges were in complete agreement on 76% of the

product categories. In cases in which there was disagreement,

the product categories were retained for further protesting if

two judges agreed that the criteria had been met. Most

categories of durable goods were eliminated because their

brands had been extended previously. For instance, the Black

& Decker name is applied to product categories that range from

irons to power tools. Surprisingly, most services were also

eliminated because the judges felt that there was little

differentiation among brands. In total, 31 product categories

were liberally identified as meeting the criteria and were

retained for additional protesting. These categories are

denoted by an asterisk in Appendix IA-1.


Pretest 2


The purpose of this pretest was to identify the specific

associations of the remaining brands. A free association task

was employed to determine each brand's specific association

and subsequently to compare the associations among brands

within a product category for uniqueness.

For each product category, associations were taken for

the product category level and at least two high profile

brands within that category. Seventy-two subjects wrote their






16

thoughts for brand names, while thirty subjects wrote their

thoughts for product categories. Subjects were run in groups

of up to twenty participants. Each subject provided their

associations for ten to twelve brands (product categories),

with each brand coming from a different product category. Each

brand (category) was printed atop its own page in the stimulus

booklet. The task instructions informed subjects that 30

seconds would be allotted to generate associations for each

item, and that these associations would include physical

product features, usage situations, or any other thoughts (see

Appendix IB-1).

The associations for each brand and product category that

were mentioned by at least 20 percent of its respondents are

listed in Appendix IB-2. Associations to a given brand were

compared with associations to other brands in its category and

to the category itself. Brand associations that appear in bold

in Appendix IB-2 are those considered to be brand-specific

associations. Brands were considered to have a brand-specific

association if they had an association that was (a) unique

from other brands and the category level or (b) twice as

salient for the brand than other brands and the category.

The results show that most brands had specific

associations. These specific associations included physical

features of the product such as size (e.g., Honda was viewed

as small), color (e.g., Froot Loops is a multi-colored

cereal), texture (e.g., Wonder is a soft bread), and packaging






17

(Pringles potato chips come in a can). Brand-specific

associations also included benefits (e.g., Bactine is a

soothing first aid remedy), usage situations (e.g., Pepsi soda

can be a morning drink), and users (e.g., Apple computers are

used by kids).

Nine product categories were dropped from further

protesting because they contained brand-specific associations

that were derived solely from their commercials or because

further examination revealed that the brands already had

multiple extensions. The luggage, automobile tires, and paper

towel product categories were eliminated because their brand-

specific associations related to their ad executions or

spokesperson (e.g., Rosie in Bounty commercials) which were

deemed unacceptable as a basis for extension. The pen,

cleansing cream, fruit juice, toys, and battery, and picnic

cooler product categories were eliminated because brand-

specific associations only belonged to brands that had been

multiply extended.


Pretest 3


To test the hypothesis that inferences about brand-

specific associations can lead to brand preference reversals

between the original and the extended product categories, it

was necessary to measure brand preference in the original

product category. In addition to having a brand-specific

association, the focal brand had to be less preferred in its






18
original product category than a comparison brand. The purpose

of this pretest was identify brands that possessed specific

associations and were less preferred than another brand in

their product category. Additionally, because the focus of

this research is on brand-specific associations unrelated to

image, brands with prestige images needed to be identified and

eliminated from consideration.

At least four brands from each of the remaining 22

product categories were rated for their affect and image.

Affect was measured by having subjects indicate their brand

preference on a nine point scale from dislike to like. Image

was defined as conveying prestige or high status to other

people and measured by having subjects indicate the extent to

which brands have a self-expressive image to other people on

a nine point scale from not at all to very much. Instructions

to subjects included an example on rating brands for their

image (see Appendix IC-1).

Sixty undergraduate subjects participated in this

pretest. To avoid carryover effects from one scale to the

other, the categories were divided into two sets. One group of

30 subjects gave preference ratings for Set 1 and then image

ratings for Set 2. Another group of 29 subjects first gave

preference ratings for Set 2 and then image ratings for Set 1.

The ratings for affect and image are shown in Table III-1.

The differences in mean ratings were analyzed by

performing contrasts on a repeated measures analysis of








TABLE III-1



BRAND MEANS FOR PRETEST 3


1)LAUNDRY DETERGENT
A)Surf
B)Wisk
C)Cheer
D)Tide

2)SUNTAN LOTION
A)Coppertone
B)Hawaiian Tropic
C)Sea & Ski
D)Bain de Soleil

3)WINE
A)Gallo
B)Inglenook
C)Paul Masson
D)Taylor

4)STOMACH REMEDY
A)Maalox
B)Rolaids
C)Pepto-Bismol
D)Alka-Seltzer

5)BICYCLE
A)Cannondale
B)Schwinn
C)Raleigh
D)Huffy

6)BREAD
A)Wonder
B)Bakery Goodness
C)Pepperidge Farm
D)Nature's Own


Preference

5.400
5.32D
6.10D
7.52ABC


5.90 D
6.68c
4.3 3ABD
7. 0Ac


5.06c
4.27
4.39A
4.61


4.57B
5.60
5.42
5.71


5.87
6.74
6.35
4.65Asc


4. 90C
5.28"C
6.58ABD
7.67ABC


Image

2.86CD
3.240
4.03AD
4.45ABC


5.038C
6.00AC
3.97ABD
5.97c


4.72c
4.52C
5.48AB
4.93


2.86BcD
3.61A
4.11A
3.93A


6.28D
6.00D
5.48
4.03AB


3.83CD
3.07"D
5.93"AB
5.45 AB








TABLE III-1


7)TOPICAL FIRST AID
A)Campho-Phenique
B)Bactine
C)Lanacane
D)Neosporin

8)PEANUT BUTTER
A)Roddenbury
B)Peter Pan
C)Skippy
D)Jif

9)BEER
A)Budweiser
B)Coors
C)Michelob
D)Lowenbrau

10)POTATO CHIPS
A)Lay's
B)Ruffles
C)Pringles
D)Wise

11)WATCH
A)Seiko
B)Swatch
C) Rolex
D) Timex

12)DEODORANT
A)Arrid
B)Secret
C)Sure
D)Right Guard

13)COMPUTER
A) IBM
B)Compaq
C)Apple
D)Hewlett-Packard


-- continued


Preference

5.90BCD
4.77AD
4.67A
6.77ABC


4.50 BCD
6.74
6.26A
7.10A


6.84 8
5.58ACD
7.45BD
4.22ABC


6.77
6.90CD
5.90B
4.97AB


6.72BD
4.83Ac
7.72 8
5. 00AC


3.97c
4.55
4.82A
4.79


8.14BCD
5.32A
6.52A
5.93A


Image

3.29
3.68
3.36D
4.25c


4.69
5.14
5.28
4.72


6.14c
5.66CD
7.37AB
6.93B


4.31Bc
5.52AD
5.27AD
3.76BC


6.13CD
5.19c
8.65ABD
4.51AC


2.65BD
4. 10AC
3.138
3.55A


7.45BCD
5.42ACD
6.60AB
6.27AB








TABLE III-1 -- continued


Preference Image
14)AUTOMOBILE
A) BMW 7.72CD 8.8BCD
B)Honda 7.07c 6.68AC
C) Ford 4.79ABD 4.97ABD
D)Acura 6.55AC 7.29AC

15)TOOTHPASTE
A)Crest 7.07B"D 4.87B
B)Close-Up 5.97AD 5.68ACDE
C)Aqua Fresh 5.31AD 4.388
D)Ultra Brite 4.21"ACE 4.17B
E)Colgate 6.310 4.688

16)SOAP
A)Irish Spring 5.148 4.39
B)Ivory 6.52AC 4.35
C)Camay 4.24"8 3.80
D)Dial 6.10c 3.65

17)SHAMPOO
A)Head & Shoulders 4.14 4.23c"
B)Suave 4.34 3.190
C)Prell 4.14 3.06AD
D)Vidal Sassoon 5.03 5.45AB

18)PAIN RELIEVER
A) Bayer 5.10Bc 3.778C
B)Tylenol 7.41AD 4.77AD
C)Advil 7.38AD 5.19AD
D)Bufferin 4.628c 3.35BC

19)SOFT DRINK
A)Coca-Cola 7.41" 6.178
B)7-Up 6.17A 4.77A
C)Dr. Pepper 5.38A 5.30
D)Pepsi 6.69 5.58

20)GUM
A)Trident 5.48 4.268
B)Carefree 5.07 3.63A
C)Bubble Yum 4.45D 3.81
D)Big Red 5.93c 4.10








TABLE III-1 -- continued


Preference Image

21)GYM SHOES
A)Reebok 6.64s 7.04c
B)L.A. Gear 3.76ACD 7.92co
C)New Balance 5.65 0 4.09AB
D)Nike 7.60BC 5.67AB

22)CEREAL*
A)Lucky Charms 5.34
B)Cheerios 5.10C
C)Froot Loops 6.480
D)Wheaties 4.83c

Means were on a 9 point Scale (N=29 or N=31) and a letter
indicates a significant difference of p < 0.05.

* The cereal product category inadvertently left out
certain brands and these data were collected for 29
subjects at a later time.

variance for each product category. Differences that were

significant at a 0.05 level are superscripted by a letter

indicating the contrast brand. Based on these results,

fourteen product categories were eliminated.

The nine product categories of deodorant, shampoo, peanut

butter, pain reliever, bicycle, topical first aid remedy, soft

drink, stomach remedy, and suntan lotion were eliminated

because preference differences did not exist on brands

identified as having brand-specific associations.

The product category of soap only showed a marginal

preference for Irish Spring over Camay, F(1,29) = 2.96, E <

0.09. The overall difference in soap preference was due

exclusively to a preference by men for Irish Spring, [(M =

5.27 vs. M = 3.47), F(1,14) = 6.78, p < 0.02]; females showed






23

no preference difference, [(M = 5.00 vs. M = 5.07), F(1,13) =

0.08, R > 0.78]. It was thought that this example would

provide a strong test of the influence of affect in brand

extension evaluations. That is, if both men and women equally

like the potential brand extensions of Camay, then further

evidence of the relative role of brand-specific associations

over affect is provided. For this reason, Camay was retained

as a potential stimulus brand.

The categories of wine and gym shoes were eliminated

because their preference and image ratings were in opposition.

That is, the less preferred brand with a specific association

had a higher image than the more preferred brand which would

confound a preference reversal result.

Unfortunately, many image differences existed between

brands in the remaining thirteen product categories. To test

the hypotheses on as large a set of stimuli as possible, it

was deemed acceptable to covary out these image differences in

the experiment as long as preference and image differences

among the brands were in the same direction and a large

preference differential did not exist. Three product

categories were dropped because the image differential between

brands was deemed to be excessive: watch (Timex vs. Rolex),

bread (Nature's Own vs. Wonder), and auto (Ford vs. Honda).

Thus, eight product categories remained for further

protesting. The focal and comparison brands for each product

category are shown in Appendix IC-2. The focal brand possessed






24
a brand-specific association and was less preferred than its

comparison brand. While the comparison brand often had a

brand-specific association, this was not a requirement. Its

main purpose was to serve as a representative for its product

category. This point will be discussed further in Posttest 1.


Pretest 4


The purpose of the fourth pretest was to have subjects

generate potential extensions for brands that would be used as

product extension stimuli for experiment 1. By having subjects

generate potential extensions, the test of preference

reversals would be on a set of extensions that subjects

believed were plausible.

Thirty-three subjects generated potential brand

extensions for the eight product categories. Eight subjects

generated potential extensions to a product category level

cue. Twenty-five other subjects generated potential extensions

to the brand name cue of either the focal or comparison brand

within a given category. The order of product categories was

randomized. Subjects were given one minute to generate

potential extensions to each cue. They were asked to help a

firm identify possible areas of expansion for a product (see

Appendix ID-1).

Subjects generated a wide range of potential extensions

and appeared to consider a brand's specific association when

deciding where to extend. For instance, possible Cheerios'






25

extensions such as oatmeal and trail mix incorporate its

brand-specific association of plain taste. Table III-2

contains the extensions generated by subjects for each cue in

the product categories. Extensions that were common responses

across the three cues are listed in the common category. The

remaining extensions generated are grouped after their

respective cue.

This pretest also pointed out that three additional

product categories should be deleted. The lack of

differentiation between cues in the laundry detergent product

category pointed out that stain removing had erroneously been

deemed a unique association of Wisk in Pretest 2. Because

stain-fighting was equally salient at the product category

level and the brand level of Wisk, it is not a brand-specific

association, and the laundry detergent category was

eliminated. The gum category was eliminated because it

overlapped with potential extensions of toothpaste and cereal.

Because the focal brand of Bubble Yum had the same brand-

specific association as Froot Loops cereal and the comparison

brand of Big Red had the same association as Close-Up

toothpaste, the presence of gum in the set might have

contaminated the results of the main experiment in as much as

product category was a within-subjects factor. The potato









TABLE III-2

RESULTS OF PRETEST 4

SUBJECT-GENERATED EXTENSIONS


LAUNDRY DETERGENT
Common Extensions
Across Cues:


Wisk:


Tide:


Category:



POTATO CHIPS
Common Extensic
Across Cues:


Pringles:





Ruffles:



Category:


,ns


Fabric Softener, Bleach, Dishwashing
Liquid, Washer, Dryer, Household
Cleaner, Stain Remover, Soap, Laundry
Basket

Air Freshener, Fabric Protector, Starch,
Dishwasher, Shampoo, Lint Remover,
Sponges

Car Wax, Air Freshener, Toothpaste,
Towels, Shampoo, Carpet Cleaner,
Upholstery Cleaner

Clothes, Clothes Hangers, Clothes Line,
Ironing Board, Iron, Starch, Machine
Degreaser


Chip Dip, Soft Drinks, French Fries,
Mashed Potatoes, Corn Chips, Tortilla
Chips,Popcorn

Instant Potatoes, Microwave Side Dishes,
Pretzels, Potato Skins, Crackers,
Cookies, Bread, Snack Cakes, Cheese
Puffs, Frozen Pizza, Candy, Vegetable
Oil, Cheese


Potato Skins, Chip Container, Flavors,
Bread, Candy, Cereal, Crackers, Tater
Tots

Pretzels, Beer, Chip Warmer, Chip Clip,
Pork Rinds, Tater Tots, Instant
Potatoes, Peanuts, Salt, Oil, Flavors,
Cheese Spread, Preservatives









TABLE III-2 -- continued


CEREAL
Common Extensions
Across Cues:

Cheerios:




Froot Loops:





Category:




BEER
Common Extensions
Across Cues:


Coors:

Budweiser:


Category:



COMPUTER
Common Extensions
Across Cues:


Apple:


IBM:


Poptarts, Bread, Milk, Juice, Pastries,
Breakfast Bar, Granola

Oatmeal, Vitamins, Breakfast Drink,
Trail Mix, Sweet Kid Variety, Bird Feed,
Candy, Snacks, Cereal Bowl, Flavor
Variety, Cookies

Cookies, Candy, Toys, Fruit-Roll Up,
Ice Cream, Gum, Popsicles, Video Game,
Flavors, Crayons, Chips, Low Sugar
Variety, Cereal Bowls, Oatmeal, Games


Pancake Mix, Pancake Syrup, Oatmeal,
Muffins, Animal Feed, Alcohol Products,
Dried Fruit, Boxes, Toys, Party Mix,
Grits, Snacks, Sugar


Mugs, Glasses, Clothes, Beach Towel,
Potato Chips, Soft Drinks, Wine Coolers,
Water, Snack Foods

Ice Chest, Party Supplies

Hard Alcohol, Flag, Posters, Sports,
Kegs, Hat, Watch, Pizza, Bottle Opener

Hard Alcohol, Can Insulator, Coasters,
Flavors, Tea, Ice Chest, Signs,
Cigarettes


Software, Video Games, Stereo, Printer,
Monitor, Television, Modem, Instruction
Manual, Calculator, Telephone

VCR, Copy Machine, Desk, CB, Radio, Disk
Container

Typewriter, Ribbon, Modem, Camcorder,









TABLE III-2 -- continued


COMPUTER continued
Category:





BAR SOAP
Common Extensions
Across Cues:


Camay:


Irish Spring:





Category:


Desk, Chairs, Disks, Disk Container,
Paper, Keyboard, Covers, Typewriter,
Sunglasses, Mouse, Joystick, Desk Lamp,
Ribbon, Course



Shampoo, Conditioner, Liquid Soap,
Body Lotion, Towels, Toilet Paper, Dish
Detergent, Laundry Detergent, Soap Dish

Facial Moisturizer, Facial Cleanser,
Bath Beads, Sunscreen, Make-up, Scented
Variety, Hand Lotion, Talcum Powder,
Deodorant,Fresh Wipes, Bath Rug

Deodorant, Cologne, Aftershave,
Detergent, Air Freshener, Bubble Bath,
Household Cleaners, Facial Soap, Make-
up, Air Conditioner, Shower Gel, Talcum
Powder, Perms, Combs, Bath Mat

Bath Mat, Bath Rug, Shower Curtain,
Facial Soap, Household Cleaners, Facial
Cleanser, Washclothes, Combs, Brushes,
Hand Lotion, Shaving Cream, Make-up
Remover


TOOTHPASTE
Common Extensions
Across Cues:


Toothbrush, Dental Floss,
Dental Drill, Fluoride


Mouthwash,


Close-Up:


Breath Mints, Gum, Fluoride Variety,
Tooth Whitener, Acne Medicine, Fluoride
Rinse, Denture Cleaner, Soap

Dental Kit, Plaque Remover, Water Pik,
Gum Medication, Electric Toothbrush,
Gum, Sugarfree Candy, Throat Lozenges,
Breath Mints, Shampoo, Soap, Towels


Crest:








TABLE III-2 -- continued


TOOTHPASTE continued
Category: Toothbrush Holder, Teething Medicine,
Dental Polish, Tooth Whitener, Fluoride
Rinse, Denture Cleaner, Dental Adhesive,
Toothpicks, Deodorant, Candy

GUM
Common Extensions Candy Bars, Baseball Cards, Blow Pops,
Across Cues: Breath Mint, Sugarfree Variety, Hard
Candy

Bubble Yum: Soft Drink, Toys, Comic Book, Gum Balls,
Jello, Mint Variety, Cereal, Clothes,
Skateboards, Bikes, Cereal, Shoes

Big Red: Cinnamon Red Hots, Mouthwash,
Toothpaste, Soft Drink, Ice Cream,
Flavor Variety, Gum Wrapper

Category: Tic Tac, Chewing Tobacco, Gum Wrapper,
Medicated Gum, Toothpaste, Gumball
Machine, Taffy

chips category was eliminated because subjects had trouble

generating extensions to the attributes of ridges, can, and

processed. Thus, five product categories remained to test the

hypotheses.

These five product categories met the criteria outlined

at the beginning of this chapter. The product categories

contained two familiar brands, one of which had a specific

association, had not been multiply extended, and was less

preferred than its counterpart. Although it is not possible to

eliminate the effects of image completely, the brands that

survived all four protests are notably low on the image

dimension, particularly in comparison to brands used by Park

et al. (1991).








Manipulation Check 1


From the pool of responses obtained from Pretest 4, the

experimenter selected extension categories that manipulated

the match between a brand's specific association and the

relevance of that association in the extension category. For

each product category, two extension categories were selected.

One extension matched the focal brand's association and the

other extension matched the comparison brand's or the product

category's association.

To corroborate the experimenter's choice of extensions,

a manipulation check was performed. Recall that the hypothesis

concerned whether preference for the focal brand would reverse

from the original to the extended product category. Thus, the

emphasis in stimulus selection was on the focal brand. The

focal extension category was chosen such that there was a

match between the focal brand's specific association and the

relevant attributes in that product category. The other

extension category served as a control where the comparison

brand was expected to retain its preference advantage. To

facilitate the selection of stimuli, this comparison extension

category either contained an association that was relevant

specifically for the comparison brand or for the product

category.

Fourteen subjects participated in this test. Each subject

rated how relevant a specific association was for four product

categories on a nine point scale from not at all relevant to






31

very relevant. For each association, two focal extension

categories and two comparison categories were tested. The

instructions to subjects and a sample of the stimulus test are

presented in Appendix IE.

The mean relevance ratings are exhibited in Table III-3.

The means were analyzed by a within-subject analysis of

variance. Contrasts among the means for a given association

were performed. These results demonstrate that the focal

brand's association was relevant in the selected focal

extension category and the comparison brand or product

category association was relevant in the comparison extension

category.


Manipulation Check 2


Different methodologies for extracting brand associations

may produce different results (Barnard and Ehrenberg, 1990).

Therefore, this test examines brand associations using a

rating scale method. Given the short generation period, the

free association task used in the preceding test was a measure

of top of mind awareness. Thus, the associations listed were

likely to be highly salient. The rating scale measure may be

more sensitive to subtle differences in association strength.

An association may be an integral component of a brand

concept, but may not be listed in a free association task









because it seems too mundane. For example, a subject may not

list "cleans" as an attribute for a soap brand, because he/she

TABLE III-3



MANIPULATION CHECK 1

MEANS FOR ASSOCIATION RELEVANCE IN EXTENSION CATEGORIES


1)SKIN SOFTENING
A) Deodorant 1.86s0
B)Moisturizer 8.71ACD
C)Cologne 2.29sD
D)Cleansing Cream 7.50ABC

2)SWEETNESS
A)Lollipops 8.578D
B)Oatmeal 3.00ACD
C)Toasted Pastry 7.86"0
D)Waffles 4.14AC

3)DENTAL PROTECTION
A)Toothbrush 8.93c
B)Dental Floss 8.50"
C)Mouthwash 6.07ABD
D)Breath Mints 4.00ABC

4)TECHNOLOGY
A)Cellular Phone 8.29B
B)Instructional Tapes 3.71ACD
C)Stereo 7.868
D)Video Games 7.57B

5)LIGHT TASTE
A)Scotch 2.57"
B)Wine Cooler 5.79A
C)Bottled Water 6.21A
D)Beer Mug 3.93

6)SCENT
A)Deodorant 7.7 18C
B)Moisturizer 6.29AC
C)Cologne 8.86ABD
D)Cleansing Cream 5.93AC









TABLE III-3 -- continued



7)BLANDNESS
A) Lollipops 2.14BCD
B)Oatmeal 5.50AC
C)Toasted Pastry 2.93AB
D)Waffles 4.21A

8)BREATH FRESHENING
A)Toothbrush 2.29cD
B)Dental Floss 2.21"
C)Mouthwash 8.86AB
D)Breath Mints 9.00AB

9)USER FRIENDLINESS
A)Cellular Phone 7.14
B)Instructional Tapes 6.79
C)Stereo 6.29
D)Video Games 5.71

10)LOGO
A)Scotch 9.00
B)Wine Cooler 5.79c
C)Bottled Water 2.50"
D)Beer Mug 8.00

11)COLOR
A) Lollipops 8.07BCD
B)Oatmeal 2.50AC
C)Toasted Pastry 5.57AB"
D)Waffles 3.07AC

12)MOUNTAIN SPRING
A)Scotch 2.00C
B)Wine Cooler 3.43c
C)Bottled Water 8.86AB"
D)Beer Mug 2.57c

13) ELECTRONICS
A)Cellular Phone 8.43B
B)Instructional Tapes 3.71ACD
C)Stereo 8.868
D)Video Games 7.868









TABLE III-3 -- continued


14)HIGH ALCOHOL CONTENT
A)Scotch 8.438co
B)Wine Cooler 7.21ACD
C)Bottled Water 1.00ABD
D)Beer Mug 3.71ABC

15)HEALTHY GRAINS
A)Lollipops 1.07BCD
B)Oatmeal 8.07"
C)Toasted Pastry 3.43ABD
D)Waffles 5.57Asc

16)KIDS
A)Video Games 8.79BCEFGH
B) Waffles 6.64ADEFGH
C)Stereo 5.71ADEF
D) Lollipops 8.71BCEFGH
E)Toasted Pastry 8.00ABcDFGH
F) Instructional Tapes 2.29ABCDEG
G) Oatmeal 4.36ABDEFH
H)Cellular Phone 1.Q00ACDEG

Means on 9 Point Scale (n = 14). Superscripts indicate ratings
of contrast categories that are significant at the 0.05 level.



feels that this is obvious. On the other hand, the rating

scale measure may introduce demand by cuing associations that

were not salient in the brand concept.

To control for potential demand effects, subjects rated

how strongly they associated an attribute with a brand in a

comparative manner. Thus, whereas brand was a between-subjects

factor for the free association task, brand was a within-

subject factor for this rating task.

Twenty-four subjects participated in this test. For each

association, subjects rated how strongly it was associated

with four brands in a product category: the focal brand, the







35

comparison brand, and two filler brands. The instructions to

subjects and a sample of the stimulus material are described

in Appendix IF.

The mean ratings for brand association strength are given

in Table III-4. The difference in association strength for the

focal and comparative brand was analyzed by a t test. All the

focal brand associations used in the experiment were confirmed

by the rating task methodology except for the beer category.

Coors and Budweiser were rated equally for the association of

light taste. (Subjects seemed confused on how to interpret

rating questions in the beer category, as three subjects put

question marks and skipped these questions). In general,

however, the results of the rating scale measure of brand

association strength strongly converged with the findings of

the free association task in Pretest 2.


TABLE III-4

MANIPULATION CHECK 2
MEAN BRAND RATINGS FOR ASSOCIATION STRENGTH


1)SKIN SOFTENING
A)Irish Spring 1.58
B)Ivory 6.79
C)Dial 2.67
D)Camay 7.96

Difference between A and D, t(1,23) = 12.96, E < 0.0001

2)TECHNOLOGICAL
A)IBM 8.58
B)Compaq 5.13
C)Apple 7.13
D)Hewlett-Packard 6.79

Difference between A and C, t(1,23) = 1.94, p < .06









TABLE III-4 -- continued


3)BLAND
A)Froot Loops 1.54
B)Wheaties 6.50
C)Cheerios 5.25
D)Grape Nuts 6.75

Difference between A and C, t(1,23) =6.37, E < 0.001

4)LIGHT TASTE
A)Michelob 5.81
B)Lowenbrau 2.76
C)Coors 5.95
D)Budweiser 5.24

Difference between C and D, t(1,20) =1.15, p > 0.2641

5)BREATH FRESHENING
A)Crest 6.00
B)Close-Up 8.50
C)Ultra Brite 3.79
D)Colgate 5.75

Difference between A and B, t(1,23) = 5.23, E < 0.0001

6)FLAVORS
A)Froot Loops 8.75
B)Wheaties 2.08
C)Cheerios 3.38
D)Grape Nuts 2.25

Difference between A and C, t(1,23) = 9.12, p < 0.001

7)MOUNTAIN SPRINGS
A)Michelob 3.45
B)Lowenbrau 1.90
C)Coors 7.05
D)Budweiser 3.80

Difference between C and D, t(1,19) = 3.41, R < 0.004

8)KIDS
A)Froot Loops 8.83
B)Wheaties 2.13
C)Cheerios 5.79
D)Grape Nuts 1.42

Difference between A and C, t(1,23) = 4.99, P < 0.001









TABLE III-4 -- continued


9)SCENT
A)Irish Spring 8.79
B)Ivory 3.04
C)Dial 5.63
D)Camay 3.88

Difference between A and D, t(1,23) = 8.17, p < 0.001

10)USER FRIENDLY
A)IBM 6.43
B)Compaq 4.17
C)Apple 8.00
D)Hewlett-Packard 3.35

Difference between A and C, t(1,22) = 2.42, E > 0.0242

11)DENTAL PROTECTION
A)Crest 8.79
B)Close-Up 4.33
C)Ultra Brite 3.58
D)Colgate 7.92

Difference between A and B, t(1,23) = 9.26, E < 0.001

12)LOGO
A)Michelob 5.77
B)Lowenbrau 3.36
C)Coors 5.50
D)Budweiser 7.55

Difference between C and D, t(1,21) = 3.20, R > 0.0043

13)HEALTHY
A)Froot Loops 1.92
B)Wheaties 8.13
C)Cheerios 5.83
D)Grape Nuts 8.38

Difference between A and C, t(1,23) = 7.47, p < 0.001

14)KIDS
A)IBM 3.25
B)Compaq 2.20
C)Apple 7.00
D)Hewlett-Packard 1.92

Difference between A and C, t(1,23) = 5.99, p<0.001









TABLE III-4 -- continued


15)HIGH ALCOHOL CONTENT
A)Michelob 4.95
B)Lowenbrau 6.40
C)Coors 4.30
D)Budweiser 5.00

Difference between C and D, t(1,19) = 1.45, E > 0.1625

16)SWEET
A)Froot Loops 8.75
B)Wheaties 2.13
C)Cheerios 3.17
D)Grape Nuts 1.96

Difference between A and C, t(1,23) = 10.79, p < 0.0001


Means on a 9 point scale (N=24)












CHAPTER IV
EXPERIMENT 1: AN EMPIRICAL INVESTIGATION OF THE ROLE OF
BRAND-SPECIFIC ASSOCIATIONS AND BRAND AFFECT ON THE
EVALUATION OF BRAND EXTENSIONS



Overview


The reasoning developed in Chapter II concerned the role

of brand-specific associations and brand affect in the

evaluation of brand extensions. It was argued that brand

associations were an important component of brand equity. Two

different types of brand associations are the liking for that

brand and the specific beliefs that differentiate the brand in

its product category. Because both types of brand associations

are likely to be important when a brand is extended to a new

product category, it was hypothesized that there would be a

brand-specific association X brand affect interaction.

Although not predicted, it was tentatively proposed that if a

brand's specific associations are judged by the consumer to be

beneficial for the extended product category, the brand may

have an advantage over its original competitors when extending

to that new product category, even if the competitors have

higher brand affect. This chapter details the methodology

employed in Experiment 1 and the analyses and results of

Experiment 1.









Experimental Design


The overall design is a 2 (Brand) X 2 (Extension

Relevance) X 5 (Product Category) X 2 (Extension Category Set)

mixed design. Brand name was a between-subjects factor of

focal or comparison brand (e.g. Close-Up vs. Crest). The

second factor, extension relevance, varied whether the focal

or comparison brand's association was relevant in the

extension category (e.g. The breath mint extension category

valued breath-freshening, whereas the toothbrush extension

category valued dental protection). It was treated as a

within-subject factor. The last two factors of parent product

category and set were replicates included to increase the

generalizability of the results. Pretesting showed that five

parent product categories were eligible to test the hypothesis

(i.e., toothpaste, soap, cereal, computer, and beer). All five

were included as a within-subject factor. Extension category

set was a between-subjects factor that replicated the

extension product categories designated as matching the

association of either the focal or comparison brand (e.g. The

association of breath freshening was relevant in the extension

categories of breath mint and mouthwash). Thus, this factor

provided an opportunity to demonstrate that the predicted

interaction was due to a match between the focal brand's

associations and the relevance of that association in the

extension product category, rather than some other aspect of

the extension product category.






41

The entire stimulus set for this experiment is shown in

Appendix IIA-1. Each subject evaluated two brand extensions

for each of the five product categories. For one brand in each

product category, subjects judged two extensions, one that was

consistent with the focal brand's association and one that was

consistent with the comparison brand's association. The order

of association relevance in the extension category was

counterbalanced as was the order of product categories. One

additional product category in which both extensions were

appropriate for the brand was included as filler to reduce

potential hypothesis guessing. The order of product category

sequences is described in Appendix IIA-2. In total, subjects

evaluated twelve potential extensions.


Experimental Procedure


Seventy-six volunteer subjects enrolled in the

Introductory Marketing course at the University of Florida

participated in this experiment. Subjects received extra

credit for their participation. The subjects were run in

groups of 10-12 participants. The stimulus materials were

contained in a booklet with brands and categories

counterbalanced across subjects and within experimental

sessions. The introductory page defined brand extensions and

gave examples of existing brand extensions. Subjects were told

that the experimenter was interested in reactions to some

potential brand extensions (see Appendix IIB-1). To







42

familiarize subjects with the procedure, they completed a

practice set of questions about a potential extension for

Ford. Subjects were then paced through the remainder of the

study by the experimenter.

For each extension, subjects were allotted one minute to

answer three items (see Appendix IIB-2). The first two items

on the questionnaire were nine point scales that measured

extension evaluations. The first scale assessed overall

evaluation of the potential brand extension relative to

existing brands in the extension product category and ranged

from one of the worst (1) to one of the best (9). The second

scale assessed preference for the potential brand extension

and ranged from dislike (1) to like (9). The third item

concerned cognitive responses to the potential brand

extension. Subjects were asked to list their thoughts about

the potential brand extension.

For each product category, subjects answered the three

items for two potential extensions for the same brand.

Afterward, subjects answered three additional questions about

the brand name. Specifically, they were asked to rate their

preference, familiarity, and image perception regarding the

brand name. The first item served as a manipulation check of

the affect ratings from the pretest. The last two items served

as covariates in the analyses. The three items measuring

perceptions of the brand name are presented in Appendix IIB-3.






43

Finally, subjects completed a preference ranking task.

Subjects were asked to rank order their preference for three

brands in an extension product category. One brand was the

existing leader of the product category, and the other two

were potential extensions by the focal and comparison brand.

Recall that each focal and comparison brand had two extensions

that valued it association but not its counterpart for a total

of 24 possible extensions across the six product categories.

Whereas brand name was a between-subjects factor in the main

design, the ranking test replicated the design with brand name

as a within-subject factor forcing subjects to choose whether

they preferred the focal or comparison brand for each

extension. Subjects ordered their preferences for a set of

twelve extensions (one designated focal and comparison

extension from each product category), but not the same set

that they evaluated earlier. The instructions for rank

ordering preferences are described Appendix IIB-4. The entire

procedure took approximately forty-five minutes to complete.

An outline of the procedure for each subject is presented in

Appendix IIA-3.

Results


The manipulation check of brand affect yielded results

that were consistent with those found in the pretest. Across

the five product categories, the brands that were initially

identified as focal brands were rated lower in brand

preference (M = 6.16) than the brands that were identified as






44

comparison brands (M = 6.86), and this difference was

significant F(1,375) = 7.42, p < .001. Further support for the

affect manipulation was provided by the fact that for each

product category, the focal brand was directionally less

preferred than the comparison brand. However, this result was

only significant for two of the five product categories. This

inconsistency with Pretest 3 is likely to have been caused by

methodological differences. Recall that subjects here gave

affect ratings for one brand in each category after evaluating

two potential extensions for that brand. In contrast, subjects

in Pretest 3 gave affect ratings for four brands from each

product category. Apparently, rating brand preference without

the context of other brands in its category increased the

standard error of the test relative to the pretest in which

several brands within a product category were rated. These

results are presented in Table IV-1.


TABLE IV-1

MEAN PREFERENCE RATINGS FOR BRANDS



PRODUCT CATEGORY FOCAL BRAND COMPARISON BRAND

TOOTHPASTE* Close-Up 6.42 Crest 7.89
CEREAL Cheerios 5.74 Froot Loops 6.32
SOAP Camay 5.25 Irish Spring 5.46
COMPUTER* Apple 7.18 IBM 8.43
BEER Coors 6.14 Budweiser 6.30

*Indicates difference significant at p < .05.








Evaluation Judgments of Brand Extensions

The critical test is to examine the Brand X Extension

Relevance interaction. This corresponds to a test of the brand

affect X brand-specific association interaction. A secondary

test involves the relative preference ordering of the focal

brand and the comparison brand in an extension category in

which the focal brand's specific association is relevant.

Dependent Measures. Two measures of brand extension

preference were collected: a measure of comparative

preference with existing brands in that category and a measure

of general liking. These two evaluative ratings of potential

brand extensions were highly correlated across all conditions

(Pearson r = 0.84) and were averaged to produce the single

dependent measure used in the analyses.

Analysis. For purposes of analysis, the design was

treated as a hierarchical nested design. A factor is nested

when the levels of that factor are uniquely defined at the

levels of another factor (Keppel, 1982). Therefore, the Brand

factor is nested in Product Category because each set of focal

and comparison brands is specific to its original product

category. Likewise, Extension Relevance and Set are also

treated as nested in Product Category. The data were analyzed

in a mixed ANOVA design with between-subjects factors of Brand

and Set and within-subject factors of Extension Relevance and

Product Category. The results are first presented treating all

factors as fixed. Therefore, the statistical generalizations







46
from these results are limited to effects observed with the

specific stimuli used in this experiment. The results are

reported in Table IV-2 to Table IV-4.

Hypothesis testing. The hypothesis predicted a Brand X

Extension Relevance interaction. The analysis of variance in

Table VI-5 clearly supports this prediction. The Brand X

Extension Relevance interaction was significant, F(4,218) =

4.48, R < .002, and produced a cross-over pattern. In

extension product categories where the focal brand-specific

association was relevant, the focal brand was more preferred

as an extension (M = 6.15) than was the comparison brand (M =

5.04). On the other hand, in extension product categories

where the focal brand-specific association was not relevant,

the comparison brand (M =5.73) retained its preference

advantage over the focal brand (M = 4.94). The strongest test

of the interaction is the former effect of brand at the

designated focal extension category. Although the means

indicate the preference ordering of focal and comparison

brands reversing in extension categories that valued the focal

brand's association, complications from the nested design did

not permit a statistical test across categories. However, this
0
result is confirmed by the individual category analyses

reported later. Thus, the brand-specific association X brand

affect interaction was significant and indicated preference

reversals. Even though the focal brand was less preferred in

its original product category than the comparison brand, its








brand-specific association provided the advantage for it to be

more preferred in an extension category where its association

was valued.

TABLE IV-2


ANALYSIS OF VARIANCE
TESTS OF HYPOTHESES FOR BETWEEN SUBJECTS EFFECTS



SOURCE TYPE IV SS DF F VALUE P > F

BRAND(CAT) 13.1181 4 0.67 0.6120
SET(CAT) 6.2112 4 0.32 0.8656
BRAND*SET(CAT) 21.5388 4 1.10 0.3558
SUB(BRAND SET) 1068.5666 219


TABLE IV-3

ANALYSIS OF VARIANCE
TESTS OF HYPOTHESES FOR WITHIN SUBJECTS EFFECTS


SOURCE TYPE IV SS DF F VALUE P > F

EXTE::fCAT) 65.6031 4 7.22 0.0001
EXTEN*BRAND(CAT) 40.7273 4 4.48 0.0017
EXTEN*SET(CAT) 91.6723 4 10.09 0.0001
EXTEN*BRAND*SET(CAT) 17.1385 4 1.89 0.1139
EXTEN*SUB(BRAND SET) 495.1689 218


CAT 40.8232 4 2.83 0.0270
CAT*SUB(BRAND SET) 497.6768 138









TABLE IV-4


LEAST SQUARE MEANS USING COMBINED DEPENDENT MEASURE



BRAND 1 BRAND 2
TOOTHPASTE (Less Prefer) (More Prefer)


Mouthwash
Dental Floss
Breath Mint
Toothbrush




Oatmeal
Toasted Pastry
Waffles
Lollipops


Moisturizer
Deodorant
Cleansing Cream
Cologne


Close-Up
6.80
6.20
7.58
5.95


Crest
7.36
6.69
6.17
6.81


Cheerios FrootLoops
6.23 3.48
5.40 4.95
5.36 3.96
4.08 6.54


Camay
6.53
5.03
6.18
3.77


COMPUTER


IrishSprinq
4.53
5.42
5.29
3.54


Video Games
Stereo
Instructional Tape
Cellular Phone



Wine Coolers
Beer Mug
Bottled Water
Scotch


Apple
6.45
4.82
6.85
6.21


Coors
5.67
6.83
5.91
3.68


IBM
6.08
4.66
7.34
7.22


Budweiser
3.29
6.43
3.03
4.71


* Extension 1 was chosen so that Brand 1's association was
relevant, whereas Extension 2 was chosen so that Brand
2's association was relevant. Within each parent product
category, Set was a replicate of the match of brand
associations to extension categories.


SET 1

SET 2


CEREAL

SET 1

SET 2


SOAP

SET 1

SET 2


EXT 1
EXT 2
EXT 1
EXT 2


EXT 1
EXT 2
EXT 1
EXT 2


EXT 1
EXT 2
EXT 1
EXT 2


SET 1

SET 2


BEER

SET 1

SET 2


EXT
EXT
EXT
EXT


EXT
EXT
EXT
EXT






49

The results reported in Table IV-2 and Table IV-3

utilized Type IV Sum of Squares because this reduced the CPU

power needed to fully test this nested design using a general

linear model. The model accounted for 90% of the variance in

the evaluation of brand extensions, but this analysis did not

provide tests of the covariates. In more limited model

specifications, tests of the covariates repeatedly resulted in

the same outcome, the familiarity covariate was marginally

significant, F(1,219) = 3.70, p < .06, and the image covariate

was not significant, F(1,219) = 2.17, p > .14.

One consequence of a nested design is that tests for

interaction effects between a nested factor and the factor

within which it is nested are not permissible. Thus, a test

for the three-way interaction of Brand X Extension Relevance

X Product Category was not possible. The significant Extension

Relevance X Set interaction, F(4,218), p < .001 suggests that

the replicates of focal and comparison extension categories

were not identical in their manipulation strength of

association relevance. This is not surprising and does not

diminish the implications of the results. Moreover, the

critical result of interest did not interact with Set; Brand

X Extension Relevance X Set interaction was not significant,

F(4,218) = 1.87, E > .06.

To further examine the effect of brand-specific

associations and brand affect, analysis of variance tests were

performed for each product category. The results show that the







50

Brand X Extension Relevance interaction was significant in

four of the five product categories and eight of the ten focal

extension categories exhibited preference reversals.

In the toothpaste product category, the Brand X Extension

Relevance was significant, F(1,65) = 5.99, E <.02, but so was

its higher order interaction, Brand X Extension Relevance X

Set, F(1,65) = 8.07, E < .001. Therefore, simple main effects

tests for brand in the focal extension category in each set

were performed. In the first set, Close-Up was not preferred

to Crest in the mouthwash product category and although the

means were in the wrong direction they were not significant

(M = 6.80 vs. M = 7.36; F(1,35 = 0.88, p > .35). Cognitive

responses revealed that this result was due to different

perceptions of the mouthwash category by subjects in each

brand condition. Subjects who evaluated Close-Up mouthwash

perceived it as a minty breath freshener, whereas subjects who

rated Crest mouthwash perceived it as an anti-plaque dental

rinse. In the second set, Close-Up (M = 7.58) was

significantly preferred over Crest (M = 6.17) as an extension

in the breath mint product category, F(1,38) = 4.63, E < .04.

In the cereal product category, the Brand X Extension

Relevance interaction was also significant, F(1,66) = 28.24,

E < .001. Cheerios was preferred to Froot Loops in both the

focal extension categories of oatmeal (F(1,37) = 23.11, p <

.001) and waffles (F(1,37) = 2.99, p < .10).






51

In the soap product category, the Brand X Extension

Relevance interaction was significant, F(1,64) = 16.95, p <

.001. Although there was a significant three-way interaction

of Brand X Extension Relevance X Set, F(1,64) = 4.45, p < .04,

it was due to an unexpected reaction to the comparison brand

in the comparison extension category of Set 2. That is, the

comparison brand of Irish Spring (M = 3.54) was equally

preferred to the focal brand of Camay (M = 3.77) in the

comparison extension of cologne, F(1,36) < 1. Cognitive

responses revealed that although Irish Spring possessed the

important attribute of scent, it was evaluated negatively on

this dimension as being overpowering. The focal brand, Camay,

was more preferred in both of the focal extension categories

of moisturizer, F(1,37) = 8.65, p < .01, and cleansing cream,

F(1,35) = 3.20, p < .08.

Recall that the soap category was specifically chosen to

provide further insights into the role of affect in brand

extensions. Although Irish Spring was preferred to Camay in

its original category at the aggregate level, this preference

dominance was found only with males; females were ambivalent.

If the evaluations of the potential extension were moderated

by brand affect, one would expect different evaluations of the

extensions by gender. The results show that both genders

reacted almost identically to potential extensions. The Brand

X Extension Relevance X Gender interaction was not

significant, F (2,63) = 0.56, p < .58. For both males and







52

females, the Brand X Extension Relevance interaction was

significant and in the same cross-over pattern, males (F(1,27)

= 12.24, p < .01) and females (F(1,27) = 5.32, E < .03). Thus,

regardless of their original preferences, both genders equally

recognized the relevance of the focal brand's association in

the focal extension category and made their extension

evaluations accordingly.

In the computer product category, the Brand X Extension

Relevance interaction was not significant, F(1,65) = 0.34, p

> .56. The primary reason appeared to be the focal extension

of instructional tapes in which the comparison brand, IBM (M

= 7.34) was unexpectedly more preferred than the focal brand,

Apple (M = 6.85). Nonetheless, this difference was not

significant, F(1,34) = 0.15, E > .70. Cognitive responses

showed that subjects believed IBM lacked ease of use and

therefore an instructional tape extension would be desirable.

Apparently, subjects evaluated extensions based on two

different aspects of relevance, one in which the brand's

association was relevant in the extension category and one in

which the extension category would be a relevant addition to

the core product. This suggests that a brand may benefit from

an extension by adding new value to the core brand through the

reverse transfer of a new association. In the other focal

extension of video games, Apple (M = 6.45) was more preferred

than IBM (M = 6.08), although this was not significant,

F(1,34) = 0.11, p > .74.






53

Finally, in the beer product category, the Brand X

Extension interaction was significant, F(1,66) = 25.39, p <

.001. The focal brand of Coors was more preferred than
o
Budweiser in both the focal extension categories of wine

cooler (F(1,37) = 17.62, p < .001) and bottled water (F(1,37)

=18.74, p < .001). Thus, the category replicates provide

strong evidence of a brand-specific association X brand affect

interaction. They also indicate that brand-specific

associations seem to dominate brand affect in extension

evaluations.

ANOVA effects treating brands, extensions, and product

categories as random factors. To this point, the analyses

have treated Brand, Extension Relevance, and Product

Categories as fixed factors. Thus, statistical generalizations

are limited to treatment effects observed with the specific

conditions of this experiment. To increase generalizability,

these factors were analyzed as random factors.

The usual error terms cannot be used to estimate random

effects. The appropriate test is a F ratio in which the

denominator is a combination of terms that contribute to the

mean squares of the effect (Keppel, 1982). If an effect in the

model matches the variation in the treatment effect, it serves

as the error term and is distributed as a normal F ratio. If

no single effect matches, quasi F ratios are used in which the

denominator is a combination of effects that together provide

an appropriate test of the treatment effect. Quasi F ratios







54

have degrees of freedom that are adjusted according to

Satterthwiate's (1946) method (see in Appendix IIC).

The results are presented in Table IV-5 and Table IV-6.

The Brand X Extension Relevance interaction is not significant

with random factors, F(4,4) = 2.38, p > .21, although the

three-way Brand X Extension Relevance X Set approaches

significance, F(4,218) = 1.89, E < .12. Thus, the hypothesis

is statistically supported with the treatment brands and

product categories used in this experiment, but it is not

generalizable to all brands and product categories.


Table IV-5


ANALYSIS OF VARIANCE FOR RANDOM EFFECTS
NORMAL F RATIOS


SOURCE TYPE IV SS DF DF F VALUE P > F

EXTEN*BRAND(CAT) 40.7273 4 4 2.38 0.2112
EXTEN*SET(CAT) 91.6723 4 4 5.35 0.0666
EXTEN*BRAND*SET 4.2846 4 218 1.89 0.1139


Table IV-6

ANALYSIS OF VARIANCE FOR RANDOM EFFECTS
QUASI F RATIOS



SOURCE TYPE IV SS DF DF F VALUE P > F

EXTEN(CAT) 65.6031 4 5 0.57 0.2500
BRAND(CAT) 13.1181 4 3 0.29 0.2500

CAT 40.8232 4 -1 N/A N/A
SET(CAT) 6.2112 4 4 0.06 0.2500









Preference Ranking

This task specifically examined the possibility of brand

preference reversal between the original and extended product

category using a rank order dependent measure in which Brand

was a within-subject factor. A forced choice measure was

constructed from these ranks based on which brand received the

higher ranking. Implicitly, subjects chose whether they

preferred the focal or comparison brand in extension

categories. These forced choice results are presented in

Tables IV-7 and IV-8. Chi square analyses were performed for

each extension category. These results show that in six of the

ten focal extension categories the focal brand was selected

over the comparison brand as predicted at the 0.02 level. In

the other four focal extension categories, there was no

difference in brand choice, X2 < 0.50. These instances of

equal preference in the focal extension category still provide

evidence of the effect of brand-specific associations because

the original preference dominance of the comparison brand is

erased. A less preferred brand was able to be equally or more

preferred than its comparison brand in an extension category

that valued its association.

These forced choice results parallel the simple effect

tests of the evaluative rating task reported earlier, and the

latter most closely resembles the decision task for an

extension judgment. A consumer evaluates a brand extension in

isolation or in comparison to existing members of the









TABLE IV-7


FORCED CHOICE RESULTS
NUMBER OF SUBJECTS CHOOSING FOCAL OR COMPARISON BRAND
IN EXTENSION CATEGORY


TOOTHPASTE


Mouthwash
Dental Floss
Breath Mint
Toothbrush




Oatmeal
Toasted Pastry
Waffles
Lollipops


Moisturizer
Deodorant
Cleansing Cream
Cologne


BRAND 1

Close-Up
17
8
32
7


BRAND 2


Crest
18
28
8
33


Cheerios Froot Loops
25 10
14 22
36 4
1 21


Camay
28
9
29
11


Irish Spring
8
26
11
29


COMPUTER


Video Games
Stereo
Instructional Tape
Cellular Phone




Wine Cooler
Beer Mug
Bottled Water
Scotch


Apple
16
5
19
3


Coors
19
14
29
15


IBM
19
30
21
37


Budweiser
15
25
10
21


SET 1

SET 2


CEREAL

SET 1

SET 2


SOAP

SET 1

SET 2


EXT 1
EXT 2
EXT 1
EXT 2


EXT 1
EXT 2
EXT 1
EXT 2


EXT
EXT
EXT
EXT


SET 1

SET 2


BEER

SET 1

SET 2


EXT
EXT
EXT
EXT


EXT 1
EXT 2
EXT 1
EXT 2









TABLE IV-8


CHI SQUARE ANALYSIS OF FORCED CHOICE RESULTS


EXTENSION
CATEGORY N CHI SQUARE P > X2

TOOTHPASTE
Mouthwash 34 0.03 0.25
Dental Floss 35 11.11 0.01
Breath Mint 39 14.40 0.01
Toothbrush 39 16.90 0.01

CEREAL
Oatmeal 34 6.43 0.02
Toasted Pastry 35 1.78 0.25
Waffles 39 25.60 0.01
Lollipops 22 18.18 0.01

SOAP
Moisturizer 35 11.11 0.01
Deodorant 34 8.26 0.01
Cleanse Cream 39 8.10 0.01
Cologne 39 8.10 0.01

COMPUTER
Video Games 34 0.26 0.25
Stereo 34 17.86 0.01
Instruct Tape 39 0.10 0.25
Cellular Phone 39 28.90 0.01

BEER
Wine Cooler 35 0.47 0.25
Beer Mug 38 3.10 0.10
Bottled Water 38 9.26 0.01
Scotch 35 1.00 0.25


extension category, not in comparison to hypothetical

extensions from the brand's current competitors.

In summary, the evaluative rating results strongly

support the hypothesis of a Brand X Extension Relevance

interaction. The overall interaction was significant and the

interaction was significant in four of the five replicate







58

categories. Moreover, in the ten extension categories that

valued the less preferred brand's association, this brand was

more preferred than a comparison brand in seven extensions and

equally preferred in three extensions. These three apparent

exceptions to the general hypothesis were easily explained by

cognitive responses that indicated the construct of extension

relevance might be expanded to include satisfying a benefit

desired in the core product. The preference ranking task

displayed a similar pattern of results. Overall, the results

consistently show strong brand effects such that a brand

possessing an association valued in an extension category is

more preferred than its original competitors when extending to

that new category, even if its competitors have higher brand

affect.













CHAPTER V
EXPERIMENT 2: AN EMPIRICAL INVESTIGATION OF THE ROLE OF
BRAND-SPECIFIC ASSOCIATIONS AND PRODUCT CATEGORY SIMILARITY ON
THE EVALUATION OF BRAND EXTENSIONS


Overview




Prior research has supported the commonsense notion that

product category similarity plays a role in brand extension.

That is, a brand extends more easily when the new category is

similar to the original category than when it is dissimilar--

all things equal (Aaker and Keller, 1990; Boush and Loken,

1991). Recently, however, one paper has argued that this

relationship may be moderated by the identity of the brand.

Specifically, Park et al. (1991) compared the extendibility of

brands originating from the same product category but

differing along a prestige-functional continuum. Their results

suggest that, although the basic similarity relationship

holds, prestige brands are influenced less by product category

similarity than are functional brands.

This study represents an important step because it

recognizes the influence of the brand in brand extension.

Unlike previous research, it employed real brands with unique

identities and did not confound brand identity with product

class. However, it may be limited in its ability to capture

59







60

the unique value of a brand name for two reasons. First, the

distinction between functionality and prestige is too broad,

inasmuch as each concept may refer to a diverse set of

meanings. For example, among automobiles, functionality may be

reflected in reliability (e.g., Honda), energy efficiency

(e.g., Geo), safety (e.g., Chrysler), and durability (e.g.,

Ford). Similarly, a brand may be prestigious for a variety of

reasons including its expensive price, superior product

quality, or exclusivity of consumers (Park, Jaworski, and

MacInnis, 1986). Indeed, it could be argued that prestige and

function orientations are subcategories within a product class

--more specific than the category but less specific than

individual brands. For example, experiment 1 showed that

consumers reacted very differently to extensions from two

functional brands of cereal such as Froot Loops and Cheerios.

This problem is exacerbated in the Park et al. (1991) study by

the use of only a single prestige brand (Rolex) and a single

functional brand (Timex), resulting in a confound of sorts

between brand and subcategory.

A second limitation of the study involves its focus and

the consequent managerial implications of its results. Park et

al. (1991) were interested primarily in examining the joint

effects of product class similarity and brand concept

consistency on extendibility. Their results indicate that the

best scenario for successful extension involves extending a

brand into a similar and consistent category (e.g., functional






61

brand into a similar functional category). We suspect that

most managers would concur with this finding. Functional

brands may lack the glamour needed to succeed in prestige

product classes and, all things equal, similar product

categories will produce better results than dissimilar

categories -- as suggested by most research in this area. Once

the incongruity between prestige and functionality is

recognized, however, a critical question for managers involves

the extent to which their opportunities are limited by

category similarity. The central theme of the present research

is that associations specific to the brand are a primary

determinant of extendibility.

In the present context, this suggests that an appropriate

brand-specific association may compensate for lack of product

class similarity and, moreover, extensions that are not

consistent with the brand's identity may be ill-conceived,

regardless of category similarity. In experiment 2 we directly

examine the differential leverage provided by product category

similarity and brand-specific associations when extending to

a brand. Following the logic of experiment 1, it is predicted

that there will be a brand-specific association X product

category similarity interaction. Moreover, it is hypothesized

that an extended brand will be more favorably evaluated in a

dissimilar category in which its association is valued than in

a similar category in which its association is not relevant.

That is, brand effects are predicted to dominate the effects






62

of category similarity when the brand association is strong

and valued in the extension category.

We focus on functional brands because the findings of

Park et al. (1991) imply that brand effects and extension

opportunities may be limited for functional brands.

Additionally, this work extends the previous research by

showing that brand associations are not limited to the generic

categories of prestige-oriented or functional-oriented, but

can be any association that is strongly linked to the brand.


Stimulus Materials


One additional pretest was necessary to develop stimulus

materials for this experiment. The purpose of the pretest was

to identify extension categories that met two requirements.

First, extension categories had to vary in their similarity to

the original product categories. Second, the specific

association of the focal brands needed to be important in

their respective dissimilar extension categories. The focal

brand for each product category and its corresponding brand-

specific association are presented in Appendix IIIA-1.

The product categories that were identified as having

brand-specific associations in Pretest 2 were reexamined to

see if they were eligible. An initial screening for a range of

extension similarity was performed by examining subject-

generated extensions. In addition to those products tested in

Experiment 1, extensions were generated for the other product







63

categories by ten subjects. Seven product categories were

selected for further protesting: toothpaste, cereal, gum,

soap, gym shoes, watch, and automobile tires. To provide a

strong test of brand effects, an effort was made to include

durable categories. Manufacturing capabilities are likely to

be prominent in the evaluation of durable goods (Aaker and

Keller, 1990) and consequently product level effects should be

most likely to occur for these extensions. Although several of

the durable brands had previously been extended, they were

strongly associated with one product category.

Thirty four subjects participated in this pretest.

Subjects first indicated their overall similarity ratings for

92 product category pairs on a nine point scale ranging from

not at all similar to very similar. The seven original product

categories were paired with seven or eight potential extension

categories and the forty remaining product pairs related to

other research. (see Appendix IIIA-2). To further examine the

similarity construct, subjects then rated the 52 original and

extension category pairs on the same nine point scale for

three more specific types of category similarity: similarity

of physical features, similarity of usage situations, and

similarity of benefits provided (see Appendix IIIA-3).

Finally, subjects rated these product categories for the

importance of brand-specific associations on a nine point

scale ranging from not at all important (1) to very important

(9). The instrument is shown in Appendix IIIA-4.







64

The results of this pretest are presented in Appendix

IIIB. The mean similarity ratings for the four types of

similarity are given in Appendix IIIB-1 and the mean

association relevance ratings are given in Appendix IIIB-2.

The overall similarity rating was highly correlated with the

three more specific measures of similarity across product

categories (Pearson r > 0.50) and was used to determine the

selection of stimuli for the similarity manipulation. Selected

product categories had extension categories that varied over

3 similarity levels, and the focal brand's association was

important in the dissimilar extension category.

Three levels of similarity were identified from

protesting. A line extension is an item from the same product

class as the original product. These were highly similar to

the originals with mean ratings ranging from 7 to 9. A similar

extension is from a different product category that is

moderately similar to the original product class with a mean

similarity rating ranging from 4 to 6. A dissimilar extension

is from a different product category that is highly dissimilar

to the original product category with a mean similarity rating

ranging from 1 to 3.

Four original product categories had extension categories

at all 3 similarity levels. To provide a strong test of the

hypothesis, two dissimilar extension categories were desired

for each original product category. Because similarity and

extension relevance ratings are on different scales, it is not






65

possible to equate their manipulation strength. Thus, if brand

associations dominate similarity in brand extension judgments,

it may be due to a stronger manipulation of extension

relevance than similarity, not because brand associations have

an inherently stronger effect than similarity. Therefore, only

highly dissimilar extension categories were chosen and more

than one dissimilar extension was tested from each product

category. These extension categories and their mean similarity

ratings are shown in Table V-l. The results show that for each

original product category, the line extension, similar

extension, and dissimilar extensions differed in their

similarity ratings with p values less than 0.05. Furthermore,

the two dissimilar extensions for each category did not differ

in their similarity ratings.

The second requirement was that the focal brand's

specific association be more important in the dissimilar

extension categories than in the similar and line extension

categories. The importance ratings of associations in these

extension categories are presented in Table V-2. For each

product category, the two dissimilar extensions had higher

association relevance ratings than their respective similar

extensions, t(1,28), p < .05. However, the line extension only

differed from the two dissimilar extensions in the cereal

(t(1,28) = 7.04, R < .0001) and soap product category (t(1,28)

= 7.71, p < .0001). Because very similar product categories

are likely to overlap on important attributes, it was







66

recognized that the focal brand's specific association would

often be valued in line extensions. Therefore, we decided to

focus the hypothesis testing on similar versus dissimilar

extensions and use the line versus dissimilar extensions as a

strong test of similarity in brand extensions.


TABLE V-1

MEAN SIMILARITY RATINGS BY TYPE OF EXTENSION CATEGORY


ORIGINAL
CATEGORY

CEREAL


SOAP



GYM SHOE



WATCH



Letter


TYPE OF EXTENSION CATEGORY

LINE EXTEN SIMILAR DISSIM 1 DISSIM 2
(A) (B) (C) (D)

Hot Cereal Waffles Lollipop Popsicle
7.97BCD 4.59ACD 1.44AB 1.59AB

Liquid Hand BubbleBath Shoe Room
Soap Deodorize Freshener
8.21BCD 5.65ACD 1.45AB 2.15AB

Ladies Canvas Wingtips Pain Rub Thirst
Gyms Quencher
7.978cD 3.64ACO 1.82AB 2.68A

PocketWatch Bracelet Alarm Outdoor
System Thermomet
7.29BCD 4.62ACD 2.21AB 1.88AB

denotes a t test t(1,28) difference of E < .05.


In Experiment 1, the comparison brands either were

representative of the focal-brand category or had their own

brand-specific associations in that category. For exploratory

purposes, two types of comparison brands were selected here to

examine this distinction. For the cereal, soap, and gym shoe

product categories, the comparison brands were chosen so that









Table V-2


MEAN ASSOCIATION RELEVANCE RATINGS BY
TYPE OF EXTENSION CATEGORY


TYPE OF EXTENSION CATEGORY


FOCAL BRAND


FROOT LOOPS
Flavor
Sweet
Kids


IRISH SPRING

Scent


NIKE

Athlete
Sports


TIMEX

Durability
Reliability


LINE EXT
(A)

Hot Cereal
6.57CD
4.33CD
5.07CD


Liquid Hand
Soap
6.72BCD


Ladies Canvas
Gyms
6.56
7.418


PocketWatch

6.90c
8.48BD


SIMILAR
(B)

Waffles
6.67CD
3.83CD
4.43CD


BubbleBath

7.10ACD


Wingtips

2.00ACD
2.14ACD


Bracelet

6.43c
5.28ACD


DISSIM 1
(C)

Lollipop
8.19AB
8.92AB
8.86AB


Shoe
Deodorize
8.24ABD


Pain Rub

6.938
7.03BD


Alarm
System
8.00AD
8.72BD


DISSIM 2
(D)

Popsicle
8.29AB
8.58AB
8.57AB


Room
Freshener
8.96ABC


Thirst
Quencher
7.598
7.59BC


Outdoor
Thermomet
7.10c
6.86ABC


Letter denotes a t test t(1,28) difference of p < .05.


they had brand-specific associations that were different from

the focal brand. For the watch product category, the

comparison brand was representative of its product category

and positioned similarly to the focal brand. However, the

important association was at least twice as salient for the

focal brand as for the comparison brand in the initial free

association pretest. In all cases, extension categories were







68

selected based on their similarity and extension relevance to

the focal brand, not the extension relevance of the comparison

brand. The focal and comparison brands for each original

product category are presented in Appendix IIIB-3.


Experimental Design


The design of this experiment was a 2 (Brand) X 3

(Similarity of Extension Product Category) X 4 (Product

Category) X 2 (Dependent Measure) mixed design. Brand name was

a within-subject factor of focal vs. comparison brand.

Similarity was a within-subjects factor consisting of four

levels: line extension, similar extension, and two dissimilar

extensions. Product category is a within-subject replicate of

extensions from the four original product categories.

Finally, to address an important issue in brand extension

research, the dependent measure was a between-subjects factor

of either an extension preference judgment or an extension fit

judgment. Although fit has been argued to be the better

measure of brand extension judgments (Park, Milberg, and

Lawson, 1990), the construct of fit has not been clearly

defined. This factor is included for exploratory purposes.

Each subject evaluated four brand extensions, one for

each of the four original product categories. Furthermore, one

extension was a line extension, one a similar extension, and

two were dissimilar extensions. The selection of extensions

and the order in which subjects evaluated them was determined







69

by a 4 X 4 diagram-balanced Latin square design (see Appendix

IIIC-1). The stimuli were then constructed so that an

orthogonal 4 X 4 diagram-balanced Latin square for Brand was

imposed for each row of the Similarity X Product Category

Latin square. Thus, there were four sequences of brand

presentation for each row of the Similarity X Product Category

Latin Square, or 16 conditions. This stimulus design is shown

in Appendix IIIC-2. For each of these conditions, subjects

either gave evaluative or fit judgments, resulting in a total

of 32 conditions.


Experimental Procedure


One hundred fifty-nine subjects enrolled in the

Introductory Marketing course at the University of Florida

participated in this experiment. Subjects were run in groups

of 10-12 and received one extra credit point for their

participation. Subjects first received the same introduction

page used in Experiment 1. To familiarize subjects with the

task, they completed a practice set of questions about one

similar and one dissimilar extension. Subjects were then paced

through the study by the experimenter.

For each extension, subjects had one minute to answer

three items. For subjects in the Evaluative Measure condition,

the first two items on the questionnaire were nine point scale

items pertaining to evaluation of the extensions. The first

item asked subjects to rate how desirable the extension would







70
be from undesirable (1) to desirable (9). The second item

asked subjects to rate their preference for the potential

brand extension from dislike (1) to like (9). The third item

solicited cognitive responses and asked subjects for their

thoughts about the potential brand extension. These three

items are presented in Appendix IIID-1.

For subjects in the Fit Measure condition, the nine point

scales addressed the fit of extensions (see Appendix IIID-2).

The first item asked subjects to rate the extent to which the

extension fit with their knowledge of the brand name and

ranged from not at all (1) to very much (9). The second item

asked subjects to rate how appropriate the extension was for

the brand name and ranged from not appropriate (1) to very

appropriate (9). The third item was the cognitive response

question described previously.

Subjects answered the three items for four potential

extensions from different product categories. They then

completed a filler task before answering several questions

that served as covariates. These questions measured

preference, familiarity, perception of existing extension

breadth, image, and prototypicality of various brand names and

usage of the extension product category. The instructions and

rating scales for the covariates are presented in Appendix

IIIE-3. The entire procedure took approximately thirty-five

minutes.









Results


The test of the first hypothesis concerned whether there

would be a brand-specific association X product category

similarity interaction between a dissimilar extension category

that valued its association and a similar extension category

that did not value its association. Furthermore, it was

expected that the focal brand would be more preferred in the

dissimilar than the similar extension category.

Dependent Measures

Two measures of extension evaluation and two measures of

extension fit were collected. The two evaluative ratings of

potential brand extensions were highly correlated across all

conditions (Pearson r = 0.84) and were averaged to produce a

single evaluative dependent measure in the analyses. However,

the two fit ratings of brand extensions were only moderately

correlated across all conditions (Pearson r = 0.59). The first

item asked whether the extension fit with the brand name and

the second question asked whether the extension was

appropriate for the brand name. The appropriateness question

usually had a lower rating than the fit question, especially

for comparison brands. However, for the overall analyses,

these two questions will be averaged to allow a comparison of

evaluative and fit judgments in brand extensions.

Analysis

The design was treated as a hierarchical nested design.

The Brand factor is nested in Product Category because each







72

set of focal and comparison brands is specific to its original

product category. The Similarity factor was also treated as

nested in Product Category because the similarity manipulation

of extension categories was relative to each original product

category. The data were analyzed in a mixed ANOVA design with

between-subjects factors of (a) Dependent Measure, (b)

Sequence of brand presentation (Seq X), and (c) Sequence of

similarity-product category presentation (Seq Y), within-

subject factors of (a) Brand, (b) Similarity of Extension

Category, and (c) Original Product Category and the six

covariates. The results are presented treating all factors as

fixed. Therefore, the statistical generalizations from these

results are limited to effects observed with the specific

stimuli used in this experiment. Because the two dissimilar

extensions from each original product category showed no

differential effects of Similarity (Category), F(4,125) =

1.42, P > .24 or Similarity X Brand (Category) interaction,

F(4,125) = 1.24, p > .29, they were collapsed to form the

dissimilar extension condition. The results are reported in

Tables V-3 to V-5.

The test of between-subjects factors in Table V-3 showed

that the Dependent Measure factor was significant, F(1,151) =

11.91, p < .001. This factor will be examined further in a

later section. The sequence of similarity-product category

presentation was marginally significant, F(3,151) = 2.26, p <

.10 suggesting that perhaps subjects' reactions to later









extensions were affected by the similarity of previous brand

extensions to their original category.



TABLE V-3

ANALYSIS OF VARIANCE
TEST OF HYPOTHESES FOR BETWEEN SUBJECTS EFFECTS



SOURCE TYPE III SS DF F VALUE P > F

DEP 61.4496 1 11.91 0.0010
SEQ X 4.6424 3 0.30 0.2500

SEQ Y 34.6643 3 2.26 0.1000
SUB(DEP SEQX SEQY) 773.3003 151




TABLE V-4

ANALYSIS OF VARIANCE
TESTS OF HYPOTHESES FOR WITHIN SUBJECTS EFFECTS


SOURCE TYPE III SS DF F VALUE P > F

SIM(CAT) 479.6824 8 14.00 0.0001
CAT 136.4232 3 10.61 0.0001
BRAND(CAT) 11.3213 4 0.66 0.6196
SIM*BRAND(CAT) 134.9102 8 3.94 0.0002
SIM*DEP(CAT) 76.0558 8 2.22 0.0252
CAT*DEP 10.1013 3 0.79 0.5022
BRAND*DEP(CAT) 18.8053 4 1.10 0.3574
SIM*BRAND*DEP(CAT) 40.5584 8 1.18 0.3574
RESIDUAL 1807.9096 422










TABLE V-5

LEAST SQUARE MEANS COLLAPSING


DEPENDENT MEASURES


ORIGINAL CATEGORY

CEREAL
Focal: Fr
Comparison: Ch

SOAP
Focal: Ir
Comparison: Ca

GYM SHOE
Focal: Ni
Comparison: L.

WATCH
Focal: Ti
Comparison: Se


TYPE OF

LINE


oot Loops
eerios


ish Spring
may


ke
A.Gear


mex
iko


4.48
5.92


6.75
7.39


6.55
6.97


5.92
5.48


EXTENSION

SIMILAR


4.64
5.14


6.34
6.35


2.03
3.65


3.43
3.47


CATEGORY

DISSIM


5.01
2.62


5.85
4.06


4.76
4.88


6.15
6.52


The appropriate within-subject error terms for the main

effect tests of Similarity, Category, and Brand are confounded

by the Latin square design. In the analysis in Table V-4, the

main effects tests used the overall variability within

subjects as the error term. A more sensitive test would use

the variability within subjects for that factor as the error

term. Separate analyses were run for each main effect test

using the more specific error term. These analyses are shown

in Table V-6 to Table V-8. These results are consistent with

the former analysis and show a main effect of Similarity,

F(2,464) = 24.82, P < .0001, a main effect of Product

Category, F(3,444) = 11.67, p < .0001, and no main effect of

Brand, F(1,157) =2.52, p > .12.









The tests of covariates in Table V-9 showed that brand

affect was a major influence in brand extension evaluations,

F(1,151) = 37.61, p < .0001, although a brand's prestige

image, F(1,151) = 2.74, p < .10, and extension category usage,

F(1,151) = 2.68, p < .10 also affected extension judgments.



TABLE V-6

ANALYSIS OF VARIANCE
TEST OF MAIN EFFECT OF SIMILARITY


SOURCE TYPE III SS DF F VALUE P > F

SIM 262.7926 2 24.82 0.0001
SIM*SUB(DEP SQX SQY) 2456.0499 464



TABLE V-7

ANALYSIS OF VARIANCE
TEST OF MAIN EFFECT OF PRODUCT CATEGORY


SOURCE TYPE III SS DF F VALUE P > F

CAT 163.5637 3 11.67 0.0001
CAT*SUB(DEP SQX SQY) 2074.1376 444




TABLE V-8

ANALYSIS OF VARIANCE
TEST OF MAIN EFFECT OF BRAND


SOURCE TYPE III SS DF F VALUE P > F

BRAND 15.28637 1 2.52 0.1143
BRD*SUB(DEP SQX SQY) 951.7284 157









TABLE V-9


ANALYSIS OF VARIANCE
TESTS OF COVARIATES


SOURCE TYPE III SS DF F VALUE P > F

AFFECT 192.6085 1 37.61 0.0001
FAMILIARITY 4.8044 1 0.94 0.2500
BREADTH OF EXTENSION 3.7710 1 0.74 0.2500
IMAGE 14.0496 1 2.74 0.1000
PROTOTYPICAL 0.2069 1 0.04 0.2500
USAGE 13.7384 1 2.68 0.1000
SUB(DEP SEQX SEQY) 773.3003 151


The covariates of brand familiarity, breadth of previous

extension, and prototypicality had little influence on

extension judgments, F's < 1.00.

Hypothesis testing

The hypothesis predicted a Brand X Similarity of

Extension Category interaction. The analysis of variance in

Table V-4 clearly supports this prediction. The Brand X

Similarity interaction was significant, F(8, 422) = 3.94, p <

.0001. However, the critical comparisons are similar versus

dissimilar extensions and line versus dissimilar extensions.

Similar versus dissimilar extensions. Interaction

comparisons revealed that the Brand X Similarity interaction

between similar and dissimilar extensions was significant,

F(4,281) = 3.89, p < .005. Moreover, the second hypothesis of







77

brand-specific associations dominating product category

similarity is supported. The focal brand was more preferred in

the dissimilar than the similar condition, F(4,151) = 4.16, p

< .004, whereas the contrast brand was less preferred in the

dissimilar condition F(4,150) = 3.51, p < .009.

Given the nested design, it was not possible to examine

the three way interaction contrast of Similarity (Similar vs.

Dissimilar) X Brand X Product Category. As previously

discussed, the set of focal and comparison brands comprising

the Brand manipulation differed by product category. For the

cereal, soap, and gym shoe product categories, the focal and

comparison brands had different associations, and only the

focal association was valued in the dissimilar extension

category. For the watch category, the focal and comparison

brands had the same association valued in the dissimilar

product category, but their association strength differed. The

results of analysis of variance tests for each product

category are examined comparing similar and dissimilar

extensions below.

For two product categories in which the focal and

comparison brand had different brand-specific associations,

significant Similarity X Brand interaction contrasts were

obtained, ((Cereal, F(1,103) = 9.12, P < .004), and (Soap,

F(1,99) = 8.58, p < .005)). Simple main effects of Brand

showed that for the cereal product category, the focal brand

(Froot Loops) was directionally preferred in the dissimilar







78

(Lollipops and Popsicles) than the similar extension

(Waffles), (M =5.61 vs. M=4.76, F(1,48) = 1.50, E >.22),

whereas the opposite was true of the comparison brand

(Cheerios), (M = 2.51 vs. M = 4.26, F(1,48) = 10.91, p <

.002). For the soap product category, the focal brand (Irish

Spring) was equally preferred in the dissimilar (Shoe

Deodorizer and Room Freshener) and similar extensions (Bubble

Bath), (M = 6.47 vs. M = 6.60, F(1,44) = 0.54, p > .46),

whereas the comparison brand (Camay) was more preferred in the

similar extension, (M = 3.95 vs. M = 6.16, F(1,47) = 16.03, p

< .001). Although the focal brand was not more preferred in

the dissimilar than the similar extension category, it was at

least equally preferred in both extensions. On the other hand,

the comparison brand was more preferred in the similar than

dissimilar extension category. Therefore, these categories

show brand-specific associations moderating product category

similarity, but the results are weak with respect to brand-

specific associations dominating similarity.

However, there was no Similarity (Similar vs. Dissimilar)

X Brand interaction contrast for the gym shoe and watch

product categories [(Gym Shoe, F(1,101) = 1.52, p >.22), and

(Watch, F(1,101) = 0.06, R > .80)]. Examining the similar and

dissimilar comparison, one sees that this result is due to

both brands being more preferred in the dissimilar than the

similar extension. In the gym shoe product category, the focal

brand (Nike) was more preferred in the dissimilar (Pain Rub








79

and Thirst Quencher) than the similar extension (Wingtip), (M

= 4.76 vs.M = 2.03, F(1,48) = 11.29 R < .002) extension as

was the comparison brand (L. A. Gear), (M = 4.88 vs. M = 3.65,

F(1,46) = 4.04, R < .05). In the watch product category, the

focal brand (Timex) also was more preferred in the dissimilar

(Alarm System and Outdoor Thermometer) than the similar

extension (Bracelet), (M = 6.15 vs.M = 3.43, F(1,47) = 7.61,

E < .01) as was the comparison brand (Seiko), (M = 6.52 vs.

M = 3.47, F(1,47) = 11.84, p < .002). Thus, these category

replicates display extension judgments where product category

similarity appeared to have little effect as both brands are

more preferred in the dissimilar than the similar extension.

In the watch product category, when both brands possessed

associations that were valued in the dissimilar extension

category, both brands were more preferred in the dissimilar

than similar extension category. Surprisingly, differential

strengths of association did not seem to affect the brand

judgments in the dissimilar extension category. However,

posttest findings reveal that it was a manipulation problem.

Timex was actually less associated with reliability than

Seiko, although this difference was not significant (M= 7.29

vs. M = 7.71), t(1,23) = 0.85, R > .40) in ratings of brand

associations. This test was not conducted a priori because the

results of Experiment 1 Manipulation Check 2 had shown a high

correspondence between the free association task and the

association rating task. Consequently, no definitive







80

statements can be made about the differential effects of

strength of brand association on brand extension judgments.

Future research is needed to examine the role of brand

association strength in brand extensions (Farquhar, Herr, and

Fazio, 1990).

The results of the gym shoe category revealed that both

the brands of Nike and L.A. Gear were viewed more favorably in

sport-related extension categories than in a dress shoe

extension. This was expected for Nike because of its athletic

association, but not for L.A. Gear which was associated with

fashion. Cognitive responses suggest two possible

explanations. One is that some subjects may not have known

what wingtips were and the other is that L.A. Gear is

generally viewed as a women's brand and wingtips are men's

dress shoes. Interestingly, though these results suggest that

sport-related extensions would be the most favorably received,

Nike has recently extended into ladies pumps. One rationale

for this extension is the hope that a gym shoe would transfer

the association of comfort to the category of ladies pumps.

In summary, the category replicates generally showed the

focal brand being more preferred in a dissimilar extension

category that valued its association than a similar extension

category that did not value its association. Although the

results show a main effect of product category similarity in

extension judgments, this effect is qualified by its

interaction with the relevance of the brand-specific







81

association in the extension category. The results also

suggest that not only may brands have benefits that are valued

in dissimilar categories, but a product category association

may also be valued in a dissimilar extension category.

Line versus dissimilar extensions. Recall that it was

difficult to select line extensions that were low on extension

relevance for the focal brand. This was only possible in the

cereal and soap product categories, but even in these

instances the focal brand's association was rated above the

mean for extension relevance. In contrast, line extensions

are, by definition, a powerful manipulation of similarity and

this result was confirmed by pretest similarity ratings. The

line and dissimilar extensions for each category were rated at

opposite ends of the nine point similarity scale. Thus, this

contrast provides a very conservative test of the hypothesis.

The Brand X Similarity interaction contrast between line

and dissimilar extensions was significant, (4,279) = 5.75, p

< .002, however the nature of the interaction differed from

the similar vs. dissimilar comparison. Here, line extensions

are preferred to dissimilar extensions, but this difference is

not significant for focal brands (F(4,149) = 1.13 ,p > .34)

whereas it is for comparison brands (F(4,149) = 4.82, E <

.002). Therefore, even with a conservative test, the focal

brand is not more preferred in line than dissimilar

extensions.







82

Because the extension relevance manipulation only held

for two of the four product categories, this contrast was

examined for each product category. For both product

categories in which the line extension differed from the

dissimilar extension on extension relevance ratings,

significant Similarity (Line Extension vs.Dissimilar) X Brand

interactions were obtained, [(Cereal, F(1,102) = 21.00, p <

.0001) and (Soap, F(1,98) = 11.06, p < .002)]. In the cereal

product category, the focal brand (Froot Loops) was marginally

preferred in the dissimilar (Lollipops and Popsicles), (M =

5.01) than the line extension (Hot Cereal), (M = 4.48),

F(1,47) = 3.76, E < .06, whereas the comparison brand

(Cheerios) was more preferred in the line (M = 5.92) than the

dissimilar extension (M = 2.62), F(1,48) = 22.13, p < .0001.

However, in the soap product category, both brands were more

preferred in the line (Liquid Hand Soap) than the dissimilar

extension (Shoe Deodorizer and Room Freshener), although this

difference was smaller for the focal brand (Irish Spring), (M

= 6.75 vs. M = 5.85, F(1,46) = 5.06, E < .03) than the

comparison brand (Camay), (M = 7.39 vs. M = 4.06, F(1,45) =

46.07, R < .0001). As one would expect, the interaction

contrast of Similarity (Line Extension vs. Dissimilar) X Brand

interaction for product categories in which extension

relevance was not manipulated [(Gym Shoe, F(1,102) = 2.23, p

> .13 and (Watch, F(1,100) = 0.08, p > .77)]. Thus, in this

comparison of line extension vs. dissimilar extension that







83

favored similarity, brand-specific associations still

dominated for the cereal category and significantly reduced

similarity's influence in the soap category.

Fit versus evaluative dependent measures. This factor was

exploratory and no hypotheses were proposed. Park et al.

(1990) suggested that a fit measure encompasses more than

similarity, and MacInnis and Nakamoto (1990) found that fit

explained a large portion of the variance in extension

evaluations. The three way interaction of Similarity X Brand

X Dependent Measure was not significant, F( 8,422) = 1.18, p

>.30, but the interaction of Similarity X Dependent Measure

was significant, F(8,422) = 2.22, E < .03. Interaction

comparisons revealed that there was a significant Similarity

X Dependent Measure interaction between the line and

dissimilar extensions F(4,279) = 3.09, p < .02, but not in the

similar versus dissimilar (F(4,279) = 1.38, p > .23 or line

versus similar (F(4,126) = 1.04, p > .25) extensions. As shown

by the least square means for each dependent measure in Table

V-10, this appeared to be due to higher fit ratings in the

line extension condition, whereas the fit and evaluative

ratings were similar for dissimilar extensions. This result

suggests that the fit measure was actually more sensitive to

product category similarity than the evaluative measure.

However, the important comparisons are its sensitivity to

the Similarity X Brand interaction in similar vs. dissimilar

and line vs. dissimilar extension conditions. For both










TABLE VI-10

LEAST SQUARE MEANS FOR EACH DEPENDENT MEASURE

FIT DEPENDENT MEASURE

ORIGINAL CATEGORY LINE SIMILAR DISSIM

CEREAL
Focal Brand 4.78 4.76 5.61
Comparison Brand 5.89 4.26 2.51

SOAP
Focal Brand 7.67 6.60 6.47
Comparison Brand 8.08 6.16 3.95

GYM SHOE
Focal Brand 7.65 2.09 5.83
Comparison Brand 8.34 4.51 5.07

WATCH
Focal Brand 7.19 3.29 5.28
Comparison Brand 5.96 4.88 6.58


LIKE DEPENDENT MEASURE

ORIGINAL CATEGORY LINE SIMILAR DISSIM

CEREAL
Focal Brand 4.17 4.47 4.38
Comparison Brand 5.95 6.17 2.89

SOAP
Focal Brand 6.03 5.68 5.25
Comparison Brand 6.68 6.14 4.44

GYM SHOE
Focal Brand 5.33 1.99 3.63
Comparison Brand 5.54 2.48 4.40

WATCH
Focal Brand 4.71 3.41 7.06
Comparison Brand 5.33 2.30 6.63


comparisons, the three-way interaction of Similarity X Brand

X Dependent Measure were not significant, (F(4,281) = 0.78, p

> .53) and (F(4,279) = 1.52, p > .19) respectively. Thus,







85

these results provide little insight into what differentiates

the fit construct from evaluative judgments in brand

extensions. Future researchers may want to examine if ease of

manufacturing the extension category (Aaker and Keller 1990)

or other factors moderate this relationship.

In summary, this experiment found a brand-specific

association X product category similarity interaction for

functional products. Also, brand extensions to dissimilar

product categories were judged more favorably than extensions

to similar product categories if the brand's association was

relevant in the dissimilar category. Even when comparing

dissimilar to line extensions, brand-specific associations

were shown to dominate or offset category similarity. Thus, a

brand should leverage its strength of association and not

limit extension opportunities to similar product categories.

Although this experiment demonstrated how brand-specific

associations may dominate similarity, this does not imply that

similarity is uninfluential. Insofar as similarity plays a

role, its effects should be greatest when the influence of

brand-specific associations is at its lowest, when consumers

are incapable of incorporating brand-specific associations

into their extension judgments. One obvious factor that may

affect the influence of brand-specific associations is

expertise. The next two experiments examine the moderating

effect of knowledge of the brand's associations in extension

evaluations.













CHAPTER VI
EXPERIMENT 3: AN EMPIRICAL INVESTIGATION OF THE MODERATING
ROLE OF EXPERTISE ON BRAND-SPECIFIC ASSOCIATIONS AND
BRAND AFFECT IN THE EVALUATION OF BRAND EXTENSIONS


Overview


In experiment 1, brand-specific associations were

shown to moderate brand affect in extension evaluations such

that preference reversals occurred between the original and

extended category. This effect is predicted to occur only when

consumers have knowledge of the brand-specific associations

and these associations are relevant in the extension category.

This experiment examines the hypothesis that brand-specific

associations will be more influential than brand affect for

consumers high in brand knowledge, whereas brand affect will

be more influential for consumers low in brand knowledge.


Stimulus Materials


A product category was needed in which experts and

novices had the same brand preferences, but experts had a

higher level of brand knowledge. Additionally, both experts

and novices needed to be familiar with the product category.

The product category of personal computers was selected

because it is a well-known category, but one in which

technical knowledge varies across people.

86







87

Market share was used as a surrogate for brand preference

in the selection of brands. Based on market share, Apple was

selected as the more preferred personal computer brand (9.4%),

and Compaq was selected as the less preferred brand (4.0%),

(Standard & Poor's Industry Surveys, 1990). Discussions with

several computer personnel at the University of Florida

revealed that Apple was associated with good software-run

features such as graphics and sound, but poor hardware

attributes. On the other hand, Compaq was associated with good

hardware construction, but had no software associations.

Although undergraduates also associated Apple with graphics

and user friendliness, they lacked the technical knowledge of

hardware design, and associated Compaq as an IBM clone with

laptop computers.

Extension categories in which these brand associations

were valued were generated from discussions with computer

personnel. A machine reader was selected as an appropriate

extension for Apple because its associations of graphics and

sound were relevant. A machine reader is a hand-held device

used primarily by sight-impaired individuals that scans

written pages and then speaks the text aloud. A mainframe

computer was selected as an appropriate extension for Compaq

because its hardware design is several steps more advanced

than a personal computer. Mainframe computers are large,

stand-alone machines with substantial memory capacity and can

accommodate multiple person simultaneous use.









Experimental Design


A 2 (Brand) X 2 (Extension Relevance) X 2 (Brand

Knowledge) mixed design was used. Brand is a between-subjects

factor that pitted the more preferred Apple brand versus the

less preferred Compaq brand. Extension Relevance is a within-

subject factor involving whether the associations of Apple

(machine reader) or Compaq (mainframe) were relevant in the

extension category. Brand Knowledge is a between-subjects

factor (high vs. low knowledge of the brands). Computer

engineers with either masters or doctorate degrees served as

brand experts. They were employed at a specialized computer

firm that did not compete in the personal computer market.

Undergraduates at the University of Florida served as novices.

Subjects evaluated two extensions for the same brand

name. One extension was relevant to the brand association and

the other was not. The order in which subjects evaluated the

extensions was counterbalanced. Subjects were randomly

assigned to the brand conditions.


Experimental Procedure


Forty-six subjects participated in this experiment.

Fifteen expert and thirty-one novice subjects volunteered to

participate. One novice subject was dropped because he

identified himself as working part-time for a computer firm.

Subjects received an introductory page that explained brand

extensions and a second page that explained the focus of this







89

experiment was on personal computer extension (see Appendix

IV-1).

For each extension, the product category was defined at

the top of the page. Subjects then responded to the same two

evaluative items used in experiment 1. The first item asked

subjects to rate their overall evaluation of the extension

relative to existing brands in that category from one of the

worst (1) to one of the best (9). The second item asked

subjects to rate their preference for the potential brand

product from dislike (1) to like (9). The third item concerned

cognitive responses to the potential brand extension. These

items are presented in Appendix IV-2.

Subjects answered the three items for the two potential

brand extensions. They then answered several questions about

brand names that served as a manipulation check of brand

preference and as covariates. They rated their preference,

familiarity, perception of extension breadth, usage of

personal computer brands and their perception of the

difficulty making the extension products (see Appendix IV-3).


Results


It was predicted that the experts' judgments would be

dominated by relevance of brand-specific associations in the

extension categories and novices' extension judgments would be

dominated by brand affect. Thus, one would expect a Brand X

Extension Relevance X Brand Knowledge interaction.







90

Furthermore, one would expect preference reversals from the

original to the focal extension category for experts, but not

for novices.

Manipulation Check

In stimulus selection, market share served as a surrogate

for brand preference with Apple being more preferred than

Compaq. The results of subjects' brand preferences confirmed

this manipulation. For both experts (M = 7.07 vs. M = 4.80,

F(1,14) = 15.21, p <.01) and novices (M = 7.42 vs. M = 5.90,

F(1,29) = 12.51, E < .01), Apple was more preferred than

Compaq as a personal computer brand. As one would expect,

experts were more familiar than novices with Compaq (M = 7.47

vs. M = 4.39) and with Apple (M =8.33 and M = 7.90).

Dependent Measure

The two measures of extension evaluation were highly

correlated for each extension [(Mainframe, Pearson r = .72);

(Machine Reader, Pearson r = .76)] and were averaged to

produce a single dependent measure in the analyses.

Hypothesis Testing

The design was analyzed by a mixed ANOVA with between-

subjects factors of brand and expertise, within-subject factor

of extension relevance, and five covariates. The results are

presented in Tables VI-1 to VI-3. The statistical test of the

hypothesis is the same treating Brand as a fixed or random

factor.








91

The least square means are presented in Table VI-1.

Subjects appeared to see both extensions as plausible because

the means were generally above the midpoint of the scale.



Table VIII-1

LSMEANS OF EXTENSION JUDGMENTS


EXTENSION
CATEGORIES

Machine Reader

Mainframe


EXPERTS

Apple Compaq
(n = 8) (n = 7)


7.97

2.64


5.18

4.36


NOVICES

Apple Compaq
(n = 17) (n =13)


7.01

6.65


6.51

6.04


Table VIII-2

ANALYSIS OF VARIANCE FOR BETWEEN-SUBJECTS EFFECTS


SOURCE TYPE III SS DF F VALUE P > F

BRAND 1.6257 1 1.07 0.3083

EXPERT 23.5183 1 15.44 0.0004

BRAND*EXPERT 0.0266 1 0.02 0.8955

BREADTH 0.1856 1 0.12 0.7291

IMAGE 18.2012 1 11.95 0.0014

USAGE 1.4105 1 0.93 0.3422

DIFFICULTY 9.4725 1 6.06 0.0007

SUB(BRAND EXPERT) 56.3691 37









Table VIII-3

ANALYSIS OF VARIANCE FOR WITHIN-SUBJECT EFFECTS


SOURCE TYPE III SS DF F VALUE P > F

EXTEN RELEVANCE 22.5400 1 12.04 0.0013
EXTEN*BRAND 16.7062 1 8.92 0.0048
EXTEN*EXPERT 3.9228 1 2.10 0.1555
EXTEN*BRAND*EXPERT 21.5543 1 11.51 0.0016
DIFFICULTY 18.5470 1 9.91 0.0031
EXTEN*SUB(BRND EXPT) 74.8756 40


The between-subjects ANOVA in Table VI-2 shows a main

effect of Expertise, F(1,37) = 15.44, R < .001 and a

significant image covariate, F(1,37) = 11.95, p < .002. All

other effects were not significant, F's < 1, except the

difficulty of making the extension category, which is more

appropriately discussed as a within-subject factor.

The within-subject ANOVA in Table VIII-3 shows that the

predicted three-way interaction of Extension Relevance X Brand

X Expertise was significant, F(1,40) = 11.51, p < .002. Simple

effect tests for expertise show that the Extension Relevance

X Brand interaction was significant for experts, F(1,12) =

7.71, R < .02, but not for novices, F(1,27) = 0.08, p > .77.

Thus, consumers high in brand knowledge evaluated extensions

based on whether the brand's association was relevant in the

extension category, but novices were not influenced by brand-

specific associations. Although the means indicate that the







93

more preferred brand, Apple, received higher evaluations from

novices in both extension categories, the simple main effect

for brand affect was not significant, F < 1.

The difficulty of making a product in the extension

category covariate was significant, F(1,40) = 9.91, p < .004.

Surprisingly, the machine reader was perceived as a more

difficult product to make, perhaps because there are few in

existence today.

The test of the second hypothesis, a preference reversal

by experts between the original product category and the

extension category that valued the less preferred brand's

association, involves whether there was a Brand X Expertise

interaction for the extension category in which the focal

brand's association was valued. In the mainframe extension

category, the less preferred brand, Compaq, had a relevant

association and therefore was expected to be more preferred as

an extension than Apple by subjects high in brand knowledge.

A significant Brand X Expertise interaction was obtained for

the mainframe extension, F(1, 37) = 5.44, p < .03. Simple

effects show that this was the result of experts preferring

Compaq (M = 4.36) to Apple (M = 2.64), F(1,14) = 6.53, p <

.04, whereas novices preferred Apple (M = 6.65) to Compaq (M

= 6.04), although this effect was not significant, F < 1.

In summary, brand-specific associations were shown to

moderate the effect of brand affect on extension judgments

only for consumers high in brand knowledge. For brand experts,