Using Network Analysis to Understand Brands

Geraldine R. Henderson, Howard University
Dawn Iacobucci, University of Arizona
Bobby J. Calder, Northwestern University
ABSTRACT - In this paper, we model brand associations as associative networks. Although the idea of brand associative networks is well accepted, rarely are these networks elicited and modeled explicitly. Here, we do so in the context of a branding experiment by using a combination of qualitative (repertory grid) and quantitative (network analysis) techniques. This elicitation and analysis technique was used prior to and after subjects saw an advertisement for one of three types of brand extensions for a popular core brand. Results confirmed the usefulness of network analysis for identifying brand association structure and assessing changes in brand associations as a result of brand extension activity.
[ to cite ]:
Geraldine R. Henderson, Dawn Iacobucci, and Bobby J. Calder (2002) ,"Using Network Analysis to Understand Brands", in NA - Advances in Consumer Research Volume 29, eds. Susan M. Broniarczyk and Kent Nakamoto, Valdosta, GA : Association for Consumer Research, Pages: 397-405.

Advances in Consumer Research Volume 29, 2002     Pages 397-405

USING NETWORK ANALYSIS TO UNDERSTAND BRANDS

Geraldine R. Henderson, Howard University

Dawn Iacobucci, University of Arizona

Bobby J. Calder, Northwestern University

ABSTRACT -

In this paper, we model brand associations as associative networks. Although the idea of brand associative networks is well accepted, rarely are these networks elicited and modeled explicitly. Here, we do so in the context of a branding experiment by using a combination of qualitative (repertory grid) and quantitative (network analysis) techniques. This elicitation and analysis technique was used prior to and after subjects saw an advertisement for one of three types of brand extensions for a popular core brand. Results confirmed the usefulness of network analysis for identifying brand association structure and assessing changes in brand associations as a result of brand extension activity.

Recently, a new approach to understanding brand equity has been introduced to the literature (Henderson, Iacobucci, and Calder 1998). In their paper, the authors present a conceptual framework to uncover ten branding effects based on brand associations (Keller 1993). In this paper, we put the Henderson, et al. (1998) framework to an empirical test. We then map these 10 branding effects into three branding constructs: positioning, complementarity, and substitutability.

Below, we present a brief review of the consumer associative network approach. We then pursue these conjectures with an empirical test: Consumers were exposed to information about a purported forthcoming brand extension. Consumers made judgments of associative links and their perceptions are represented as associative networks. Post-manipulation perceptions are compared to those held prior to the intervention, and we examine the resultant structures for brand constructs.

CONSUMER ASSOCIATIVE NETWORKS

Representation

It is commonly held that consumers store information in memory in the form of associative networks (Anderson and Bower 1973; Ellis and Hunt 1992). In general, researchers contend that knowledge is represented as links of associations among concept nodes (cf. Sirsi, Ward, and Reingen 1996). The nodes are units of information such as brands, attributes, advertisements, etc., and the links contain the relational tie between the concepts (e.g., a brand "possesses" much of an attribute, a brand image is "like" the spokesperson). The links make various associations by connecting nodes together to form a network of ideas, or a knowledge structure. [Associations in network representations of mental models can also possess strengths, e.g., for an association based on many experiences or exposures to communications. In a graph, strength is indicated by the thickness of the line, the number of links between two nodes, or by a numerical indicator near the link. Asymmetrical relations may also be represented if one node evokes another but the reverse is not true. We are presenting symmetric, binary ties for the purpose of simplicity, though we note that all that we present is easily extended to ties with strength and direction using standard network methods (e.g., Knoke and Kuklinski 1982).]

Researchers in marketing have recently discussed brand associations, primarily as they relate to issues of brand equity, brand image, and brand knowledge (e.g., Keller 1993; Sirsi, et al. 1996). Many cognitive theories of consumer behavior posit associative network structures (e.g., Bettman 1971; Calder and Gruder 1989), and yet rarely are they elicited or modeled empirically. It is desirable to represent brand associations as networks in order to allow these structural data to be modeled in a manner most consistent with existing theoretical views of consumer memory structure.

Elicitation

Consumer brand associations can be elicited by a variety of data collection methods including free association and response (Green, Wind, and Jain 1973; Krishnan 1996), laddering (Reynolds and Gutman 1988), and pairwise similarity judgments (Hauser and Koppelman 1979). [Due to the relative newness of associative networks to marketing, we were tempted to make this paper be more tutorial in nature. Although we did our best to make it read well for those not as familiar with network approaches, we resisted making it a tutorial on network methods in marketing in general, or consumer associative networks in particular, due to the existence of fairly comprehensive reviews on these methods. Interested readers are directed to Iacobucci and Hopkins (1992) for the former and Henderson, Iacobucci, and Calder (1998) or Krishnan (1996) for more depth on these methods.]

In this research, we use a more qualitative method called the repertory grid (Henderson, et al., 1998; Kelly 1955; Zaltman and Coulter 1995). With the repertory grid technique, respondents choose their own stimulus brands and the attributes that they personally believe to be relevant for the comparisons. [A more detailed outline of the repertory grid technique and consumer associative networks is presented in Henderson, Calder, and Iacobucci 19998).] We use the repertory grid primarily because of its ability to bridge the gap between qualitative data collection media and quantitative analysis techniques (Green, et al., 1973). Although the repertory grid technique has been mentioned previously in the marketing literature (Sampson 1972; Olson and Muderrisoglu 1977), it has not received much attention in terms of a primary means of data collection until recently (Henderson, et al. 1998; Zaltman and Coulter 1995).

BRAND CONSTRUCTS

Keller (1993) suggests that brand association are the "building blocks" of a brand’s equity. However, we posit more abstract entities, or brand constructs, that comprise a brand’s equity: positioning, complementarity, and substitutability. These brand constructs are comprised of branding effects (brand dilution, branded features, co-branding, brand parity, brand confusion, and cannibalization) that are in turn comprised of brand associations. The first brand construct, positioning, focuses on a particular brand (relative to others) whereas the latter two, complementarity and substitutability, focus on relationships between brands. We describe each construct and the methods by which each could be studied. Figure 1 presents the relationships between branding constructs, observable branding effects, and associative networks.

FIGURE 1

BRANDING CONSTRUCTS, BRANDING EFFECTS, AND ASSOCIATIVE NETWORK METHODS

Brand Positioning: Brand Dilution/Branded Features

A brand is positioned as having more good attributes and fewer poor attributes relative to the other competitive market offerings. If a brand has a particularly strong and favorable heritage, the manager may consider leveraging those positive associations by introducing a brand extension. The positive associations consumers hold for the parent brand are thought to transfer to the new introduction by the activation of the original brand name. Two brand positioning phenomena enjoying current popularity are "brand dilution" and the "branding of features."

Brand dilution is the extent to which capitalizing on brand associations (e.g., by introducing brand extensions) may harm the original brand (Loken and John 1993). Brand dilution would be demonstrated most pointedly by a decrease in positive associations, or an increase in negative associations with the focal brand. However, the category itself may be diluted (e.g., as when the centralities for positive attribute nodes decrease, or those for negative attributes increase). Branded features are attributes closely associated with the host brand and are candidates to be further branded and differentiated themselves. Candidates for branded features would be those attributes high on centralityCthose that are important to the brand and category. With network methods, we could detect changes in cognitive associations regarding the parent brand. In particular, centrality indices may be used to verify that the links to the focal brand are still positive and active.

Complementary Brands: Co-Branding

Complementarity capitalizes on the associations between brands, seeking, for example, opportunities for co-branding. Co-branding effects include the use of ingredient brands or composite brands. Ingredient brands can be used as a portion of some product (e.g., an Intel Pentium Chip inside an IBM ThinkPad Notebook Computer, Starbucks coffee served aboard United Airlines flights), whereas composite brands are the "bundling of two brands to provide an enhanced consumer benefit or reduced cost" (e.g., Microsoft’s and General Electric’s MSNBC Cable/Internet offering). Cohesion associative network methods offer empirical possibilities for co-branding. Groups of attributes and brands defined on their mutually connected associations would be fitting candidates for opportunities to incorporate multiple brands. These groups reflect structures of a natural complementarity of the products that already exists in the consumer mind.

TABLE 1

DEGREE CENTRALITIES PRE- AND POST- FOR ALL THREE BRAND-EXTENSIONS

Substitutable Brands: Brand Parity, Brand Confusion, and Cannibalization

Whereas complements are brands managers seek to package together, substitutes are brands that compete with one another (Aaker and Keller 1990). Substitutability includes brand parity, brand confusion, and cannibalization. Brand parity is the perception of sameness among brands. Brand parity taken to extreme yields brand confusion: brands are indistinguishable or commodity-like, and the stimulation of one brand may evoke thoughts of another (Kapferer 1995). Cannibalization occurs when one of a firm’s brands steals share away from another.

METHOD

Below, we describe an experiment in which we explore the effect of the three Porsche brand extensions and the resultant brand associations in a pre-/post- design. Participants were graduate students in marketing. A 3 (Brand Extension Type) x 2 (pre-/post-extension response) mixed design was used. The Brand Extension Type condition represented a between-subjects manipulation whereas the pre-/post- condition was manipulated on a repeated measures basis. Participants saw an advertisement for one of three types of brand extensions: regular, family, or economy. They completed the data task before and after being exposed to the brand extension intervention.

We advertised Porsche as introducing one of three types of brand extensions: For 17 participants, "Model X" was as an economy car, featuring low price and good mileage. For 15 participants, "Model X" was a family car with features relating to the safety and comfort of a family of five (e.g., enlarged trunk space, built-in child seats, airbags, four-wheel drive, luggage racks). Finally, 13 participants served as a control group and were told the "Model X" was "everything you ever thought a Porsche to be" with no new attributes featured. [Ironically, shortly after we had begun our research, Porsche announced plans to develop and market a sports utility vehicle in a co-venture arrangement with Mercedes. This announcement added a dimension of truth to our cover story; we debriefed participants by providing them with the Wall Street Journal article. However, we informed them that the brand names used within the context of the experiment, including Porsche, were used just for the sake of example and not because of any relationship with the actual parent companies.] We chose two different extensions (economy and family) to be consistent with extant literature on brand extensions (e.g., Loken and John 1993). In the latter, for instance, the authors introduce brand extensions that are 'close,’ 'medium,’ and 'far’ from their parent brand. Thus, in our exercise, the regular condition would be 'close,’ the family condition would be 'medium’ and the 'economy’ condition would be 'far.’

The 45 participants named (up to) seven sports car brands and compared them via triadic elicitation. [Normally, a sample size of 45 would not be worthy of consideration for experimental purposed. However, since this research actually represents the quantification of otherwise qualitative data. A sample of 45 is quite healthy.] [For a detailed description of triadic elicitation see Henderson, et al. (1998).] Respondents were asked: "When you think about sports cars that you would like to own, which ones come to mind? Please list seven [below] and assume that money is not a major constraint." They were then provided with seven blank lines to write in their brand names. As for the second factor, prticipants completed their data collection task prior to and following the brand extension intervention. A multimedia campaign was used: The new Porsche was announced through word-of-mouth (a conversation excerpt), concept board (a print ad), expert reviews (Consumer Reports), and a news report (mock article from the Wall Street Journal).

RESULTS

Each respondent generated a list of cars and proceeded through the triadic comparisons. A repertory grid matrix was generated for each participant. Each of the participant’s matrices consisted of between 1 and 8 attributes and between three to seven cars. A content analysis determined a common set of 29 attributes across all respondents. Similarly, a list of 28 vehicles was elicited. For the purpose of illustration, these lists were parsed down further to 12 cars and 16 attributes based on the agreement of at least 4 of the 45 respondents (10 percent of the sample). [Researchers desiring fewer (or more) exploratory and idiosyncratic results would move this criterion up (or down). We chose to err for idiosyncrasy and richness of results.] We would normally present visual depictions of these associative networks, however, for the sake of space, we have included only the associative network summary indices: centrality, cliques, and structural equivalents.

Centralities

Table 1 has the degree centralities before and after the interventions for each of the three brand extensions. Few significant changes resulted from the advertising communications intervention. The increases (+) and decreases (-) are those over 1.98 times the average standard deviation of the pre- and post- conditions for each brand extension condition. A * indicates those that exceeded 1.98 times the average standard deviation of the pre- and post- conditions within each brand extension condition. For "economy" and "regular," the difference had to exceed 7 to be significant. For "family," a difference of 5 was significant (seconomy=3.64, sfamily=2.49, sregular=3.31).

There were no significant changes in the economy condition. In the family condition, there were significant changes to Lexus (-6), "non-British" (+6), "less-classy" (-8), and "low-quality" (+6). In the regular condition, there were significant changes to Jaguar (-7) and Slower (-7).

Cliques

Cliques are the primary measures of cohesion between nodes in a network and are subgroups based on complete mutuality (Knoke and Kuklinski 1982). Table 2 has the results on the cliques that formed in the consumer association networks before and after the manipulations.

There were 28 cliques in the pre-economy condition: 4 four-member cliques and 24 three-member cliques. There were 11 cliques in the post-economy condition: 3 five-member cliques, 1 four-member clique, and 7 three-member cliques. There were 6 three-member cliques formed in the pre-family condition and none in the post-family condition. There were 13 cliques formed in the pre-regular condition: 3 four-member cliques and 10 three-member cliques. There were 16 cliques formed in the post-regular condition: 3 four-member cliques and 13 three-member cliques.

Equivalence Substitutes

Two nodes are structurally equivalent if they have identical linkages to other network nodes (cf.., Hopkins, Henderson, and Iacobucci 1995). Table 3 has the equivalence groupings for these six conditions. In the pre-economy condition, 3 groups were based on structural equivalence. In all other conditions (post-economy, pre-/post-family, and pre-/post-regular), 2 groups were formed.

DISCUSSION

Brand Positioning

Brand Dilution. In the economy condition, there was no evidence of brand dilution. On the contrary, brand enhancement occurred for Chrysler. Perhaps individuals exposed to the economy version of the Model X by Porsche began to consider Chrysler to be a "sportier" brand if the most prototypical sports car manufacturer (Porsche) was indeed entering the low-end market.

In the family condition, there was significant brand dilution: fewer associations (4) were made to Lexus post- versus pre- family Model X by Porsche. Since Lexus is known more for its sedans than its coupes, it could be that those individuals who elicited it prior to the introduction of Model X, were less inclined to do so after the intervention. That is, there was clearly a segment of consumers (participants) for whom "sports car" represented sporty sedans (e.g., Lexus, Jaguar) or even SUV’s (e.g., Jeep). However, once they were made to believe that Porsche was entering in to this more family-oriented market, they generated fewer associations to its Japanese rival, Lexus.

In the regular condition, Jaguar suffered significant brand dilution along with insignificant losses to Mercedes Benz and Corvette. Perhaps the introduction of the regular Model X, which was touted as "everything you ever expected a Porsche to be," reminded consumers of what a "sports car" was really supposed to be. Most subjects provided Porsche as one of their more top-of-mind (or prototypical) "sports cars," and placed Mercedes, Jaguar, and Corvette farther down the list (if at all). However, surprising was a marked increase, yet not significant, in the centrality of Jeep from the pre-regular to the post-regular condition. Perhaps since there were no atypical features of sports cars provided, and consumers’ minds wee allowed to freely float to their perceived notions of "sports cars," Jeep was allowed into their thinking upon repeated reflection.

TABLE 2

COHESION CLIQUES OF ASSOCIATIONS

TABLE 3

EQUIVALENCE OF GROUPS OF COMPETITIVE SUBSTITUTES

Branded Features. Candidates for branded features would be those attributes high on centralityCthose that are important to the brand and category. There were several changes in degree centralities in the economy condition. Most notably, although not significant, was the decrease in low price associations. At first glance, this might seem counter-intuitive for the economy condition. However, it actually makes sense since our manipulation was for an $11,000 sports car by Porsche. Prior to the intervention, participants were probably anchoring on much higher reference prices as being "low." However, after being exposed to $11,000, they updated their priors to reflect a new understanding of what "low" might really mean. In addition, four attributes increased, again not significantly so, from the pre- to post- economy conditions: "Not Japanese," "Variety not Sport," Two Wheel Drive Not All Terrain," and "Sporty not Sedan." It is not clear whether any of these would be candidates for branded features as previously suggested.

Several changes to the perception of the sports car category occurred after the introduction of the family Model X. There were increases in perceptions of "classy" (less classy decreased from 8 to 0), "non-British", and "low quality." Oddly enough, but not significant, was the increased centrality of "less variety," perhaps due in part to our strong intervention that listed so many features (e.g., enlarged trunk space, built-in child seats, airbags, four-wheel drive, luggage racks) that the consumer may have thought that there were no additional options available, hence "less variety" for this family car. Since neither the economy nor regular conditions offered as many features, the consumer was more likely to assume that other options were available.

Since the regular condition functioned like a control group, it might appear odd to see any changes between pre- and post-intervention networks. However, there was still a new Porsche introduced and this introduction helped to create new associations ("non-Japanese," "Variety," "Two Wheel Drive," and "sporty") that did not exist prior to exposure to the intervention. In addition, some attributes that were highly central prior to the new car introduction, lost favor in light of the new entrant to the category: "unique" and "slower." "Unique" probably decreased in centrality due to the insistence by our intervention that the Model X was "everything you have come to expect from a Porsche." On the other hand, "slower" significantly decreased because most consumers have come to associate Porsche with speed. By adding another typical Porsche to the set of brand nodes, there was an update to the set of attributes associated with those sets of brands.

Complementarity: Co-Branding

In this section, we present the network findings from the pre- / post- experimental design for all three brand extensions. We begin by describing the brands that are cohesive (associated) and then proceed to examine those that are equivalent (similar). The cliques in Table 2 are based on a large dataset and thus, the results are fairly complex, so they are worth explaining here.

A first observation is that for the true brand extensions (economy and family), the number of cliques decreased from pre- to post manipulation. This finding suggests that the information in the communication became the focus of the consumer perceptionCthe marketing intevention clarified and simplified the associative network (at least in the short-term). The number of cliques remained the same in the regular (control) condition, and if anything, increased slightly, as if participants were elaborating on their own individual thoughts when given no truly new information.

In the pre- economy condition’s cliques 1B4, there were four associations between Benz and Jaguar. Mazda joins this group twice, and Corvette and Porsche are each activated once, indicating peripheral brands to the core Benz-Jaguar associative link. In the second clique are less traditional cars, while the fourth clique suggests the perception that Benz and Jaguar are manufacturers of a fuller line, a variety of cars, not just the sports cars being considered and compared to Porsche.

We summarize these cliques, looking for the qualities they have in common (i.e., factoring the cliques for their common and unique associations). Cliques 1-4 identify a core set of {Benz, Jaguar, Mazda} associations. Cliques 5-7 contain Benz, Jeep, classic, and faster. Cliques 8-12 describe the Benz as fast and sporty while cliques 13-18 associate Ferrari with fast. Cliques 20-28 describe both Jeeps and Mazdas as fast sports cars with common shapes.

After describing the economy Porsche, the clique structure changed quite a bit. The Benz-Jaguar was activated less frequently. Instead, cliques 1B4 show a connection between the brands: Benz, Ferrari, Jeep, and the regular Porsche and its new Model X. Cliques 5 and 6 classify Benz and non-Japanese togetherCa focus on European makes. Similarly cliques 8 and 9 contrast European/German with American, together with the attribute of common shape. Finally, cliques 10 and 11 describe that common shape is associated with perceptions of high quality.

In the pre-family condition, cliques 1-4 indicate a connection between Jaguar and Mazda and the quality of common shape. Clique 5 connects Benz and Jeep with classic, and clique 6 shows activations among faster, high price, and less variety (a high-end focused product line).

No cliques were found after the intervention. The family Porsche was described with a large number of qualities that were indeed difficult to reconcile with the status of a sports car, including leg room, full seating capacity, trunks, and bike racks etc., so perhaps the communication so sufficiently confused the consumer that they had not yet had time to reconvene sensible associations. Recall the centrality results that many changed associations were occurring in the minds of these consumers. Evidently dyadic associations were being formed and destroyed, but as yet, cliques (groups of three or more nodes) had not yet firmed up. At this point, when "sports car" was activated, it now brought along many weird new associations, none of which had yet been strongly assimilated into the sports car category, hence yielding no cliqued associations.

For the regular Porsche condition, cliques 1B5 showed similarity to the associations in the pre- conditions of the other cells (i.e., Benz and Mazda and common shape were associated). Cliques 6 and 7 emphasize the uniqueness of the Alfa Romeo. Cliques 8B10 characterized the Mazda as a Japanese make, and cliques 11 and 12 associate Lexus and non-American.

After consumers thought about Porsches, the cliques looked only somewhat different. Cliques 1B6 connected Benz to Porsche, and to a lesser extent, Lamborghini and sedans. Cliques 7B13 described the Mazda as a car with common shape made by a manufacturer that produces a small variety of cars. Lastly the cliques 16 and 17 connect the Lamborghini and high price nodes.

Substitutability

Brand Parity/Confusion/Cannibalization. As previously mentioned, Table 3 contains the equivalence groupings for these six conditions. Whereas the number of cliques usually got smaller post-intervention, the number of structural equivalents increased pre- to post- intervention. The brands and attributes that do not specifically appear in these charts are those that form one larger group, similar in structure by default by not being highly interconnected to other network nodes.

In the pre-economy condition, Mazda, Jaguar, and Jeep were the most similar interchangeable brands. The attributes of non-Japanese and common shape were seen as similar. Less-variety and faster were also attributes that factored together as similar, which is also sensible given that several auto manufacturers were included in the network with specialties in sports cars.

After introducing the inexpensive Porsche, the brands and attributes perceived to be similar change somewhat. Mazda and Jeep are now grouped with Ferrari. Jaguar is grouped with Chrysler. The attributes of non-Japanese and common shape still correlate, and now less-variety joins the cluster. Finally, sedan bodies are seen to be similar to the quality of high price.

In the family equivalence groups, there are again similarities among Jaguar, Benz, Mazda, and Jeep. There are also similarities between the attributes of less-variety and common shape, and non-Japanese and fast. After the introduction of a big, bulky family Porsche, respondents cluster together {non-Japanese, less-variety, sedan}, {Corvette, Mazda}, and {Chrysler, Jaguar, Ferrari}.

Finally, in the pre-regular condition, again, we see some concordance with the pre- perceptions in the other conditions: Mazda and Jeep are similar, as are fast, non-Japanese, and less-variety, and finally common shape is grouped with classic. After simply thinking more about Porsches, Benz is grouped with relatively high prices, Mazda and Jeep retain their similarity, and Lexus is grouped with fast and high status. Less-variety and sedan are grouped, as well as the group of non-Japanese and British (presumably in contrast to American and European makes).

The findings on the cliques and equivalence groups were not as simple and clear, as we would have liked. However, they demonstrated the richness and variety of the connections that consumers have in their minds when considering and comparing brands.

SUMMARY

Our goal was to employ consumer associative networks for the purpose of uncovering the brand constructs of positioning, complementarity, and substitutability. We began by discussing the nature of consumer brand associationsCresponses that are evoked when consumers think about brands. Existing cognitive theories of associative structures were connected to existing literature on structural networks for the purposes of representing consumer brand associations.

Associative networks were compared in a pre- / post- design during which respondents were exposed to mock advertising literature describing one of three hypothetical brand extensions. The associative networks draw from individual’s free association responses (i.e., listing their own relevant stimulus consideration sets) and from triadic comparisons that generate multiple attributes that distinguish among the brands. Given the qualitative and idiosyncratic nature of the data collection procedure, it is not surprising that the resulting associative networks were at timesfairly complex. Of greater importance is the ability of networks to distinguish brands that are associated, and therefore candidates for complementary brand action such as co-branding, from brands that are similar, and therefore substitutable competitors in the minds of the consumers.

Future Directions. Future research might include replication across non-automotive categories and elicitation and representation using theories of other knowledge structures. In terms of the former, we believe that automobiles were a product category ripe with instances of possibilities for the study of branding constructs. With that being said, future research could perhaps pursue other categories to see if they are as conducive to the study of brand constructs.

There are many different forms of knowledge structures that could have been employed for the purpose of uncovering the various branding constructs. More traditional methods bases on spatial representation such as MDS would be one option. It is our belief that consumer associative networks would prove superior to these approaches because of their tolerance of idiosyncratic responses. One of the primary differences between MDS and Consumer Associative Networks is that MDS operates using researcher-driven stimuli (brands) whereas the stimuli (brands) used in Consumer Associative Networks are consumer-driven. Thus, in the MDS example, it would have been impossible to uncover the association that consumers made between sports cars and what we consider to be sports utility vehicles unless the researcher had included an SUV brand in the pairwise similarities task given to the participants. What makes the associative network approach so rich is its ability to uncover associations such as these that would be counter-intuitive to researchers. It would be of interest to test this empirically. For the present, at the least, we wished to begin to demonstrate the utility of the consumer associative networks’ utility as an approach to begin to start addressing the many questions that might be asked regarding branding constructs. We believe that we have begun to make progress on this venture.

REFERENCES

Aaker, David. A. and Kevin Lane Keller (1990), "Consumer Evaluations of Brand Extensions," Journal of Marketing, 54 (January), 27-41.

Anderson, John R. and Gordon H. Bower, eds., (1973), Human Associative Memory: A Brief Edition, Hillsdale, NJ: Erlbaum.

Bettman, James R. (1971), "A Graph Theory Approach to Comparing Consumer Information Processing Models," Management Science, 18 (4), December, 114-128.

Calder, Bobby J. and Charles L. Gruder (1989), "Emotional Advertising Appeals," in A. Tybout and P. Cafferatta (eds.), Advertising and Consumer Psychology, Hillsdale, NJ: Erlbaum

Ellis, Henry C. and R. Reed Hunt (1992), Fundamentals of Cognitive Psychology, Fifth Edition, Dubuque, Iowa: W.C. Brown Co. Publishers, 169-193.

Green, Paul, Yoram Wind, and Arun K. Jain (1973), "Analyzing Free-Response Data in Marketing Research," Journal of Marketing Research, 10 (February), 45-52.

Hauser, John R. and Frank S. Koppelman (1979), "Alternative Perceptual Mapping Techniques: Relative Accuracy and Usefulness," Journal of Marketing Research, 16 (November), 495-506.

Henderson, Geraldine R., Dawn Iacobucci, and Bobby J. Calder (1998), "Brand Diagnostics: Mapping Branding Effect Using Consumer Associative Networks," European Journal of Operational Research, 111, 306-327.

Hopkins, Nigel, Geraldine R Henderson and Dawn Iacobucci (1995), "Actor Equivalence in Networks: The Business Ties That Bind," Journal of Business-to-Business Marketing, 2, 3-31.

Iacobucci, Dawn and Nigel Hopkins (1992), "Modeling Dyadic Interactions and Networks in Marketing," Journal of Marketing Research, 29, February, 5-17.

Kapferer, Jean-Nodl (1995), "Brand Confusion: Empirical Study of a Legal Concept," Psychology and Marketing, 12 (6), 551-568.

Keller, Kevin Lane (1993), "Conceptualizing, Measuring, and Managing Customer-Based Brand Equity," Journal of Marketing, January, 1-22.

Kelly, George A. (1955), The Psychology of Personal Constructs, Volumes 1 and 2, New York: Academic Press.

Knoke, David and James H. Kuklinski (1982), Network Analysis, Newbury Park, CA: Sage Publication, Inc.

Krishnan, H.S. (1996), "Characteristics of Memory Associations: A Consumer-Based Brand Equity Perspective," International Journal of Research in Marketing, 13 (October), 389-405.

Loken, Barbara and Deborah R. John (1993), "Diluting Brand Beliefs: When Do Brand Extensions Have a Negative Impact?" Journal of Marketing, July, 71-84.

Olson, Jerry C. and A. Muderrisoglu (1977), "The Stability of Responses Obtained by Free Elicitation: Implications for Measuring Attribute Salience and Memory Structure," in Advances in Consumer Research, 6, ed. W. L. Wilkie, 269-275.

Reynolds, Thomas J. and Jonathan Gutman (1988), "Laddering Theory, Method, Analysis, and Interpretation," Journal of Advertising Research, February/March, 11-31.

Sampson, Peter (1972), "Using the Repertory Grid Test," Journal of Marketing Research, 9, 78-81.

Schmitt, Bernd H., Nader T. Tavassoli, and Robert T. Millard (1993), "Memory for Print Ads: Understanding Relations Among Brand Name, Copy, and Picture," Journal of Consumer Psychology, 2 (1), 55-81.

Sirsi, Ajay K., James C. Ward, and Peter H. Reingen (1996), "Microcultural Analysis of Variation in Sharing of Causal Reasoning about Behavior," Journal of Consumer Research, 22, 345-372.

Zaltman, Gerald and Robin H. Coulter (1995), "Seeing the Voice of the Customer: Metaphor-Based Advertising Research," Journal of Advertising Research, July/August, 35-51.

----------------------------------------