Advances in Consumer Research Volume 19, 1992 Pages 260-267
COGNITIVE ASSOCIATIONS AND PRODUCT CATEGORY COMPARISONS: THE ROLE OF KNOWLEDGE STRUCTURE AND CONTEXT
Deborah J. MacInnis, University of Arizona
Kent Nakamoto, University of Arizona
Gayathri Mani, University of Arizona
Research in such domains as choice among noncomparable alternatives and brand extension strategies is rooted in the cognitive associations consumers make among product categories. This paper describes an exploratory study aimed at identifying the domains of consumer knowledge regarding product categories, and the contextual variables that influence the salience of these knowledge domains. The results have both theoretical and methodological implications for research on product category comparisons.
Consumer researchers are increasingly interested in understanding the cognitive processes involved when consumers make comparisons among two or more product categories. Two relevant research domains in which knowledge in this area is central are (l) consumer choice and decision making and (2) brand extension strategies.
Choice among product categories (i.e., noncomparable alternatives) is thought to occur by comparisons of abstract attributes which the two product categories share (Johnson 1984). For example, even though fur coats and stereos are not comparable on any physical attributes, they may be compared (i.e., are similar) in terms of the level of status each communicates. Johnson (1984) proposes that this abstraction process occurs because the two product categories belong to a similar superordinate level category, while Bettman and Sujan (1987) propose that this abstraction process occurs because the product categories achieve similar goals, and are hence members of the same goal-derived category (Barsalou 1985). Because relatively little is known about the structure of consumer knowledge regarding product categories, and the various conditions that make certain kinds of knowledge more or less salient, the nature of the abstraction process operating in a given consumption context is unclear.
The second domain in which knowledge of product category comparisons is relevant is that of category extensions. A category extension occurs when a brand in an established product category (Woolite) is extended to a novel category (rug cleaners). Although it has been suggested that consumers' evaluations of an extended brand depend on attribute similarity between the established and extended product categories (Aaker and Keller 1990), little is known about the knowledge domains consumers use to judge similarity among product categories.
The emergent findings in these two research areas raise several questions about the structure and use of consumer knowledge. First, what are the knowledge domains consumers use to judge comparability or similarity across products? Although Johnson (1984) uses the term "attribute" as a knowledge domain, the term is vague and obscures the potentially rich domains that may comprise knowledge of product categories. Products may share not only physical attributes, but also benefits, usage occasions, usage locations, users, etc.
Second, are certain knowledge domains consistently more salient (i.e., more strongly tied to the product category) than others? For example, across comparison contexts, are the physical attributes which two product categories share more likely to come to mind than similarities in other domains? Attribute similarities may in fact be more salient due to their physical observability (Paivio 1971) and their role in defining category membership. On the other hand, Barsalou (1982) shows that beyond a core of defining features, attribute salience is a function of the context--notably the composition of the comparison set and the associated judgment task.
Third, the salience of specific bases may be a function not only of knowledge structure and the task (as indicated above), but also of a spreading activation effect achieved through elaboration (Collins and Loftus 1975). To what extent, then, will the activation of similarities between product categories in a certain knowledge domain also affect perceived differences among product categories? Will, for example, consideration of the similarities between two product categories in a knowledge domain also cue consumers to differences between the product categories in that domain? Or will the goal-directed salience of consistent attributes inhibit activation of these differences (Fiske and Taylor 1984)?
Finally, what factors moderate the salience or activation of knowledge domains? Context effects created by the specific product categories being compared, the task (similarity, choice, evaluation), and other information inherent in the task context may influence the extent to which certain bases are salient (Chakravarti and Lynch 1983). Understanding the moderators of salience is important from a marketing standpoint since marketers are often in a position to communicate attribute similarities or differences between two product categories that may not be immediately obvious (salient) to consumers.
We report an exploratory study designed to address these issues. Using a thought-listing task, we examine what aspects of a consumer's knowledge structure for a product category become salient in comparing categories. We also investigate the role that articulated similarities have on articulated differences among product categories. We consider two primary methodological questions as well--first, the extent to which articulated similarities and differences predict overall similarity between product categories, and second, the impact of the thought-listing task itself on global similarity judgments.
REFERENT AND COMPARISON PRODUCT CATEGORIES
Most importantly, we examine the influence of three contextual factors on the salience of various domains of knowledge. One context factor is the "base" or "referent" product category. More specifically, product categories are likely to differ in the extent and domains of knowledge available to be used in a comparison judgment. A second context factor is the comparison product category. By manipulating the comparison category we are likely to vary the amount and types of knowledge regarding the referent category that will be relevant to the comparison. A final context variable concerns the more general knowledge bases that may be cued from the situation. Via experimental instructions, we attempt to cue particular functional bases for relating the categories being compared. We show that these contextual variables are all critical to understanding the category comparison process and its informational antecedents.
Subjects, Design, Procedures
One hundred thirty-eight undergraduate and graduate students participated in this study. Each subject completed a two-part questionnaire. Instructions at the beginning of the questionnaire told students that their task was to examine product category pairs and to judge the extent to which they were similar. The instructions also contained a cueing manipulation. The cue was designed to prime specific knowledge structures relating the two categories. One third of the students were simply told to assume that the products were sold by the same company. The second group of students was told to assume that a company that produces the first named product decides to manufacture and market the second product (a brand extension). The third group of students was given the brand extension instructions, and in addition was told that in many cases factors such as manufacturing and marketing synergies are important in extending brands. The three cues were designed to provide progressively stronger priming of potential marketing and manufacturing-related similarities and differences between the two products.
In part 1 of the questionnaire, subjects evaluated the overall similarity between a given (referent) product category and five different comparison product categories. [The five variants were selected as part of a pretest designed to identify products that represented a slight variant on the existing product (control), or had high vs. low similarity on naturally salient features and high vs. low similarity on nonsalient features. Subsequent analyses indicated that this manipulation was not successful. Only two broad classes tended to emerge--those comparison categories that had obvious high levels of feature overlap with the referent category and those for which the overlap, if it did exist, went largely unnoticed.] The referent and comparison categories are shown in Table 1. For example, one group of subjects evaluated the similarity of tomato sauce to tomato paste, tomato puree, apple juice, dill pickles, and white wine. Subjects were asked to rate each of the five product category pairs on a 1 to 9 similarity scale (1= not at all similar; 9= very similar).
SALIENCE AND IMPORTANCE OF BASES USED TO JUDGE SIMILARITIES AMONG PRODUCT CATEGORIES
Part 2 of the questionnaire followed immediately. As in part 1, each subject was shown a referent category and five comparison categories drawn from those shown in Table 1. The pairings of base categories in parts 1 and 2 were counterbalanced across subjects. Thus, different product categories always served as referents in the two parts of the questionnaire. In addition, presentation order of the comparison products was randomized within groups in both parts 1 and 2.
In part 2, for each referent-comparison product combination, subjects were asked to articulate the ways in which the product categories were similar and different. Following this thought-listing task, consumers were asked to rate the overall similarity of the two product categories, again on a 1-9 point similarity scale. They were then asked to rate on a 1-9 point importance scale the extent to which each articulated similarity and difference influenced their overall similarity judgment.
A content analysis of the thought-listing data was conducted to identify the knowledge domains tapped when comparing product categories. This analysis identified 11 general categories which are presented in Table 2. The number of similarities and differences represented for each knowledge domain, and the average weight of the items from each domain were used as dependent variables in the analysis. In addition, the similarities and differences were used to predict the overall similarity rating, by multiplying the average weight of the items listed for a domain by the number of items listed from that domain. These weighted scores for similarities and differences were then used in a regression analysis to predict the overall similarity rating.
Product Knowledge Salience in Category Comparison
The number of items identified as similarities and differences between product categories for each of the identified knowledge domains was tabulated, with the average importance value attached to the items in each domain (Table 2). These data allow us to analyze the structure of product category knowledge and the salience of knowledge domains. The mean number of similarities and differences articulated for each of the eleven domains indicated that some domains were consistently more salient than others. That is, some domains were consistently used to discuss ways in which the products were similar and different. The most common domain for judging both similarities and differences concerned physical product attributes. A mean of 1.39 statements reflected physical similarities between the product categories while 1.33 statements reflected differences between the attributes of the product categories. Thus, physical similarities and differences among product categories appear to be salient in the context of category comparisons.
A second general domain of consumer knowledge for product categories concerned consumers' self-orientation (e.g., usage) with the product. This general self vs. product domain contained several sub-domains. Consumers' considered similarities and differences in how products are used ( = .41 and .46 similarities and differences, respectively), when the products are used ( = .28 and .19), why the products are used (their benefits) ( = .16 and .15), and who uses the products ( = .10 and .13, respectively).
Less salient to consumers, though still articulated, were thoughts regarding the similarities and differences between products in terms of their marketing or production aspects. Noted similarities and differences between products in their packaging ( = .16 and .13, respectively), their distribution ( = .16 and .04), their production ( = .07 and .05), and their marketing tactics ( = .03 and .02) were not very common. The finding that consumers rarely consider marketing and manufacturing synergies between two products when judging overall similarity has obvious implications for brand extension research. Although product category extensions are often based on marketing or manufacturing synergies that the established and extended product categories share, consumers may not consider such synergies when they judge the value of an extended product.
It is likely that many of the thoughts expressed arise from an elaborative spreading-activation process. The question was raised earlier as to whether domains accessed as sources of similarity would also cue differences from the same domain. The number of items in each domain expressed as similarities and differences are notably symmetric, i.e., the more similarities noted in a domain, the more differences were likely to be expressed as well. While this symmetry may be related to the structure of the thought-listing task, it is suggestive that, at least in the context of category comparison, domain-specific priming rather than item-type (similarity vs. difference) priming appears to be operating.
The other measure taken with respect to these items was their subjective importance to overall similarity judgment (see Table 2). While there are meaningful differences among these ratings which will be considered later, it is notable at this point that subjects generally perceived the self-orientation domains to be most important. Thus, even though product attribute associations were more numerous, they were not perceived to be the most central in judging overall similarity. However, the variation in weights is not large; in other words, given that an item in a domain was identified, it was perceived to be at least moderately important to overall judgment.
These results indicate that the subjective importance of a knowledge domain and the extent of salient knowledge in that domain are separate constructs. The number of items expressed likely reflects a broad range of salient product knowledge, while the importance weights are likely to provide a link to the specific judgment task. Thus, it is likely that both aspects are required in examining the relationship between salient product knowledge and global judgments.
Articulated Features and Global Category Similarity
In order to examine the mapping of expressed knowledge and global judgments, the articulated similarities and differences were used to predict overall similarity of the referent and comparison categories. The number of items from each similarity and difference domain was weighted by the average importance rating of those items. These indices were used as independent variables in a regression analysis of the overall rated similarity of the two categories. The overall model was highly significant (p<.001), with the variables in the equation accounting for approximately 43% of the variance in overall similarity scores. A similar regression using only the number of items (unweighted by subjective importance) accounted for only 32% of the variance in overall similarity ratings, again indicating the need to consider the importance weights.
As would be expected, the parameter estimates for each knowledge domain indicate that all similarities between the referent and comparison product category are positively related to overall similarity judgments (except marketing, which has a nonsignificant negative coefficient), while all differences between the referent and the comparison product category are negatively related to overall similarity judgments. Moreover, the coefficient estimates suggest that bases which are most important (i.e., self-related bases) have the strongest influence on overall similarity assessment. Thus, similarities in product usage and users have strong positive effects on similarity assessment (b=.22, b=.20, respectively), while differences in product usage and users have strong negative effects on similarity assessments (b=-.13 and -.17, respectively). These findings suggest that the number of items identified and the importance weights attached to articulated items are relatively good predictors of perceived similarity. [The regression results also suggest that the respondents' subjective importance weights underweight important knowledge domains. In other words, the subjective weights are the respondent's estimates of the coefficients in this model. Were they accurate, the model coefficients would all have been equal. In fact, they add even more weight to the domains perceived by respondents to be important.]
Referent Product Category Knowledge and Associated Context Effects
ANOVA's on the total number of similarities and differences and the number of similarities and differences regarding each of the eleven bases and their respective importance weights (see Table 3) revealed several interesting findings. First, the total number of similarities and differences revealed significant referent product category effects (p<.001, p<.001 for total similarities and differences respectively). The mean number of similarities and differences for each product category reveal that consumers articulate significantly more similarities and differences for some products (i.e., cereals, ice-cream, pens) than for others (tomato sauce, dogfood). These results suggest that the knowledge structure of some product categories is more elaborate than the knowledge structure of others, leading to the identification of more similarities and differences with the comparison product category.
NUMBER OF SIMILARITIES AND DIFFERENCES
Some product categories, therefore, may be more easily compared to seemingly unrelated product categories than others. Consumers' knowledge structures for some categories may be sufficiently elaborate to allow the use of a broader range of both similarities and differences as bases for category comparisons (cf. Beattie 1981). As a result, some product categories may be more or less extendable than others. How these perceptions map onto consumer acceptance of an extended brand remains to be examined.
Second, significant referent product category effects were found for many knowledge domains. For example, consumers articulated fewer attribute similarities and differences among product categories when the referent product category was dogfood ( = .928 and .741 for similarities and differences, respectively) than when the referent product category was ice cream ( = 1.83 and 1.42). Thoughts regarding similarities and differences in product usage were most common with pens ( =.66 and .76) and least common for dogfood ( = .18 and .20). Similarities and differences regarding product benefits were most evident when the referent product category was cereal ( = .27 and .27) and least evident when the referent product category was pens ( = .06 and .08). Knowledge of the multiple uses to which cereal can be put and knowledge of its multiple benefits (nutrition, taste, satisfying hunger) suggest that this domain of knowledge for cereals is highly elaborated, while for pens it is not. These findings suggest that product specific knowledge in effect frames or provides a context for the nature of the comparison by identifying the bases on which the comparison can be made.
Context Effects Associated with the Comparison Product
Significant main effects for comparison product category and interactions between referent and comparison product were found across many of the knowledge domains (see Table 3). Although the complexity of the data preclude complete reporting of the nature of these interactions, their presence illustrates a different context effect from the referent category knowledge effect noted above. The knowledge domains tapped in judging similarity of product categories is influenced by the context in which the referent product category is evaluated (i.e., the comparison product)(see Tversky 1977). Comparing products to different product categories means that different bases will be salient. In effect, this finding confirms Barsalou's (1982) notion that associations that are likely to be salient for a product depend on the context in which the product is embedded.
A central problem in the study of multiple distinctive categories is the identification of some a priori basis for characterizing the relationship among the categories. The effects of referent and comparison category found here suggest that the explicit elicitation of shared and unique features may provide a fruitful approach to this problem.
Context Effects Associated with the Cueing Manipulation
The study also revealed that the number of similarities and differences noted within each domain was influenced by the presence of an explicit cue. Specifically, the cueing manipulation appeared to influence which aspects of product categories were selected for processing (see Chakravarti and Lynch 1983). Although the cueing manipulation was generally weak, and did not affect the total number of similarities and differences consumers noted between two product categories, it did have the effect of reducing the number of similarities consumers saw between two products in terms of their attributes (p<.01). It also increased the number of similarities consumers saw between the purchase locations of products (i.e., distribution; p<.01) and the number of similarities regarding the purchase situation (p<.01) (see Table 3).
Thus, it is possible to alter via cueing the nature of associations consumers make between two product categories. Moreover, marketing managers may have control over the type of associations that are made. Through marketing communications, marketers can cue for consumers similarities and differences on attributes that may not be immediately salient to them. This effect is important both for enhancing consumers' perceptions of the viability of an extended brand and in altering the nature of their comparisons between seemingly noncomparable products.
Limitations of the Thought-Listing Method
An obvious issue regarding the methodology applied in the present study is that the thought-listing task may lead to self-fulfilling overall judgments. In order to assess the magnitude of this problem, we compared the overall ratings taken in part 2 of the study with those provided in part 1. In part 1, no thought-listing task was imposed on judgment. However, the cueing manipulation was present for both parts.
The impact of thought-listing was examined in an ANOVA on the global similarity judgments using referent category and cue as between-subject factors, and comparison product and presence or absence of thought-listing as within-subject factors. As expected, thought-listing had a significant effect on judgments (F1,89=106.91, p<.001), an effect that was moderated by a significant interaction with cueing condition (F2,89=3.96, p<.05). In addition, the type of comparison product affected the impact of thought-listing (F4,356=8.42, p<.001).
The mean global similarity ratings are shown in Table 4. In general terms, thought listing increased the global similarity rating between categories. However, the impact of this elaboration diminished as the comparison product category was perceived to be less similar to the base category. Thus, in elaborating, subjects appeared to search for more bases for similarity, an exercise that was more successful when there was obvious consistent feature overlap between categories (comparison categories 1 and 2) than when there was less obvious feature overlap between the categories (comparison categories 3, 4, and 5). In addition, the cue alerting the subjects to the possible importance of marketing and manufacturing synergies also decreased the impact of elaboration. The types of thoughts elicited shed further insight on this finding; this cue increased the number of purchase location similarities but decreased the number of physical feature similarities. This is again consistent with the notion that the increase in global similarity from elaboration was based on elaboration of obvious consistent feature overlap.
SUMMARY AND IMPLICATIONS
In this study, we have examined featural bases underlying judgments of the similarity of categories. Our first results provide insight on the domains of knowledge applied to this comparison task. These results indicate that consumer knowledge about products spans several knowledge domains: knowledge about physical product attributes, benefits, usage, users, and marketing-related attributes. In making judgments about category relationships, salience and importance of a knowledge domain are different: concrete elements such as physical attributes tend to be highly salient, but self-related elements associated with product consumption tend to be most important. In general, while marketing-related elements lack salience, they are moderately important once elicited. In addition, elaboration does not seem to distinguish between similarities and differences; rather, salience appears to derive from the knowledge domain.
COMPARISONS OF OVERALL SIMILARITY ASSESSMENTS GIVEN THE PRESENCE OR ABSENCE OF THOUGHT-LISTING
From a methodological perspective, articulated knowledge bases and importance weights strongly predict overall similarity assessments, suggesting that these information elements are linked to the similarity judgment task. Our initial concern was the thought-listing task itself would severely distort the judgment process. However, our results suggest that while thought listing magnifies the knowledge effects (most likely because of the extended effort at memory elaboration), it does not appear to change the basic patterns of judgment. It should be noted, however, that our results must be interpreted in light of the specific similarity judgment tasks assigned.
We view the referent category, comparison category, and instructional cues as defining a judgment context. With respect to referent category, we found that product category knowledge varies in its level of elaboration, thus influencing the number of similarities and differences consumers can perceive between the referent and comparison brand. Moreover, the elaboration is domain-specific, and the domains elaborated vary across referent categories. In combination, these findings suggest that the referent category, to a significant extent, defines the "rules" of comparison. The comparison category appears to moderate this referent category effect, emphasizing as critical bases for comparison overlapping and unique features within the relevant domains. The impact of the cue further suggests that external stimuli can affect the salience and importance of specific knowledge domains, and thus the featural bases used in the judgment process.
While speculative, these results are suggestive of the types of abstractions likely to emerge from the processes advanced by Bettman and Sujan (1987) and Johnson (1984). From the standpoint of the design of new products, the present results suggest factors critical to understanding consumer evaluations of brand extensions. Consideration of physical attribute similarities and economic efficiencies alone may well be insufficient to predict consumer response. More importantly, the present study begins to examine the informational determinants of category comparisons. While this task is clearly different from choice or evaluation tasks, to the extent that comparison processes are integral to consumer evaluation of brand extensions, the present results begin to develop a basis for predicting the likely consumer response to such new products.
Aaker, David A. and Kevin Lane Keller (1988), "Consumer Evaluations of Brand Extensions," Journal of Marketing, 54 (Jan), 27-41.
Barsalou, Lawrence W. (1982), "Context-independent and Context-dependent Information in Concepts," Memory and Cognition, 10 (January), 82-93.
Barsalou, Lawrence W. (1985), "Ideals, Central Tendency, and Frequency of Instantiation as Determinants of Graded Structure in Categories," Journal of Experimental Psychology: Learning, Memory, and Cognition, 11 (October), 629-654.
Beattie, Ann E. (1981), "Effects of Product Knowledge on Comparison, Memory,Evaluation, and Choice: A Model of Expertise in Consumer Decision-Making," in A.A. Mitchell (ed.), Advances in Consumer Research, Vol. 9,Ann Arbor, MI: Association for Consumer Research, 336-341.
Bettman, James R. and Mita Sujan (1987), "Effects of Framing on Evaluation of Comparable and Noncomparable Alternatives by Expert and Novice Consumers," Journal of Consumer Research, 14 (September), 141-154.
Chakravarti, Dipankar and John G. Lynch, Jr. (1983), "A Framework for Exploring Context Effects in Consumer Judgment and Choice," in Richard P. Bagozzi and Alice M. Tybout (eds.), Advances in Consumer Research, Vol. 10, Ann Arbor, MI: Association for Consumer Research, 289-297.
Collins, A.M. and E.F. Loftus (1975), "A Spreading-Activation Theory of Semantic Processing," Psychological Review, 82, 407-428.
Fiske, Susan T. and Shelley E. Taylor (1984), Social Cognition, Reading, MA: Addison-Wesley.
Johnson, Michael D. (1984), "Consumer Choice Strategies for Comparing Noncomparable Alternatives," Journal of Consumer Research, 11 (December), 741-753.
Paivio, A. (1971), Imagery and Verbal Processes, New York: Holt, Rinehart, and Winston.