Consumer Inferences and Family Branding Strategies: a Demonstration of Category-Based Induction

ABSTRACT - A useful approach to understanding how consumers evaluate the strength of category-based product attribute statements has been suggested by a recent paper by Osherson, Smith, Wilkie, Lopez, and Shafir (1990) which examined category-based induction and argument strength. The approach is potentially useful as a framework for understanding how consumers evaluate statements about specific properties of family brand categories and individual products within those categories. Our study attempted to demonstrate both the potential and the limitations of the application of these concepts in a marketing environment and provided some support for the usefulness of the concepts, but also raised concerns about the categorization issues underlying their application.

Citation:

Christopher Joiner and Barbara Loken (1994) ,"Consumer Inferences and Family Branding Strategies: a Demonstration of Category-Based Induction", in NA - Advances in Consumer Research Volume 21, eds. Chris T. Allen and Deborah Roedder John, Provo, UT : Association for Consumer Research, Pages: 188-194.

Advances in Consumer Research Volume 21, 1994      Pages 188-194

CONSUMER INFERENCES AND FAMILY BRANDING STRATEGIES: A DEMONSTRATION OF CATEGORY-BASED INDUCTION

Christopher Joiner, University of Minnesota

Barbara Loken, University of Minnesota

ABSTRACT -

A useful approach to understanding how consumers evaluate the strength of category-based product attribute statements has been suggested by a recent paper by Osherson, Smith, Wilkie, Lopez, and Shafir (1990) which examined category-based induction and argument strength. The approach is potentially useful as a framework for understanding how consumers evaluate statements about specific properties of family brand categories and individual products within those categories. Our study attempted to demonstrate both the potential and the limitations of the application of these concepts in a marketing environment and provided some support for the usefulness of the concepts, but also raised concerns about the categorization issues underlying their application.

Although categorization theories have recently been used to successfully understand brand extension and family brand management strategies (e.g., Boush and Loken 1991; Park, Milberg and Lawson 1991; Romeo 1991) future consumer research could benefit from more formal and unified sets of propositions. An understanding of how consumers evaluate statements about attributes of a family brand category and individual products within the category, and how these relate to consumers' perceptions of similarity between the family brand, the individual products in the category, and potential new products, would be valuable to marketing researchers and practitioners. Recently, Osherson, Smith, Wilkie, Lopez and Shafir (1990) have proposed a model that is potentially useful for understanding how consumers evaluate the strength of product attribute statements by assessing category-based induction and argument strength (cf. Kardes 1993). Our study is an initial attempt to demonstrate both the potential and the limitations of the application of these concepts in a marketing environment.

The Osherson et al. Framework

Consider an "argument" that contains one or more premises, P₁...P_n, and a conclusion, C. Suppose both premises and conclusion have the form "all members of X have property Y," where X is a category and Y is an attribute or property that remains fixed across premises and conclusion (Osherson et al. 1990). An example of a marketing-related categorical argument is the following:

Ivory shampoo is mild

Ivory soap is mild

Therefore, all Ivory health and beauty products are mild

Osherson et al. (1990) propose that the strength of a categorical argument increases as a function of two factors: (a) the degree to which the premise categories are similar to the conclusion category, and (b) the degree to which the premise categories are similar to members of the lowest level category that includes both the premise and conclusion categories (e.g., 'All Ivory health and beauty products' might be the lowest level category which includes the categories 'Ivory shampoo' and 'Ivory soap'). Based on these two factors, the authors developed and found empirical support for a series of propositions relating to categorical argument strength. [The framework proposed by Osherson et al. is different from a more recent model (Sloman 1993), not tested here, in that it focuses on the role of categories and perceptions of similarity between categories in assessments of argument strength, rather than on the role of properties or attributes of the categories. Although prior brand research has often classified "categories" such as "Ivory shampoo" as exemplars of the family brand category "Ivory products", our alternative interpretation of such products as categories is consistent with the conception of categories in the Osherson et al. model. For example, in the Osherson et al. (1990) paper, "grizzly bears" is a more specific category than "bears", even through, alternatively, grizzly bear could be viewed as an exemplar of the bear category.]

A key aspect of the Osherson et al. framework is that categorical arguments can be labeled as general, specific, or mixed. An argument is general if the categories associated with each of the premises [CAT(P₁)...CAT(P_n)] are all properly included in the category of the conclusion [CAT(C)]. For example the argument,

Grizzly Bears love onions

Polar Bears love onions

Therefore, all Bears love onions

is general, since "Grizzly Bears" and "Polar Bears" are included in the "all Bears" category. An argument is specific if any category that properly includes one of CAT(P₁)...CAT(P_n), CAT(C) also includes the others. The following is an example of a specific argument:

Mosquitoes have X

Ants have X

Therefore, Bees have X

Finally, an argument is mixed if it is neither general nor specific. For example the argument,

Flamingoes have X

Mice have X

Therefore, all Mammals have X

is not general because "Flamingoes" are not included in the "Mammal" category and it is not specific because the "Bird" category includes "Flamingoes" but not "Mice" or "Mammals" (Osherson et al., p.186).

In the present research, we selected seven of the original thirteen propositions (Osherson et al. 1990) that we felt would potentially be applicable to studies of existing and new products (e.g., brand and line extensions) that comprise a family brand category. A formal description of each proposition, and an example of each from the present study, are shown in Table 1.

Proposition 1, labeled premise typicality, states that the more representative or typical CAT(P₁)...CAT(P_n) are of CAT(C), the more P₁...P_n confirm C. For example, in the following pair of arguments, the first argument "a" is claimed to be stronger than the second argument "b" (assuming that Brand B includes a wide variety of snack foods), because potato chips are more typical of the category snack foods than are granola bars:

(a):Brand B potato chips are low in sodium

All Brand B snack foods are low in sodium

Is Stronger Than

(b):Brand B granola bars are low in sodium

All Brand B snack foods are low in sodium

Consistent with this proposition, a moderately typical product in a family brand category has been shown to dilute a general family brand belief (i.e., confirm a negative belief) to a greater extent than a less typical product (Loken and John 1993).

TABLE 1

SAMPLE STIMULI, EACH OF SEVEN PROPOSITIONS TESTED

Proposition 3, conclusion specificity, states that the more specific the category in the conclusion, the stronger the argument. As noted in Table 1, when consumers are given "facts" about Brand C windbreakers and trenchcoats, they should more easily draw conclusions about the more similar, specific subcategory "Brand C outerwear" than about the broader, less specific category "All Brand C products". Proposition 3 is illustrated by the role individual products have in confirming beliefs about brand subcategories (Boush 1993) as compared to the brand category as a whole.

Propositions 5 and 6 are useful in examining how beliefs about individual products (existing or new) in a family brand may confirm beliefs about other existing products or potential brand extensions. Proposition 5, premise-conclusion similarity states that the more similar the premise categories are to the conclusion category, the more the premises confirm the conclusion. This proposition can be used to understand the conditions under which an extension of a brand to a new product category will be successful (e.g., Aaker and Keller 1990; Boush and Loken 1991; Park et al. 1990).

Proposition 6, premise diversity, which states that the less similar premise categories are among themselves, the more the premises confirm the conclusion, could be used to compare broad versus narrow brand categories (cf. Boush and Loken 1991) and their ability to confirm beliefs about existing or new products. Research on sequential introductions of brand extensions (cf. Keller and Aaker 1992) could also be analyzed within this framework.

Propositions 9 and 10, nonmonotonicity-general and -specific, demonstrate how an argument is weakened when a new premise is added that introduces a different category to an argument. Learning that another brand's product has a similar attribute as a member of a target family brand should decrease the strength of the consumer's beliefs about the group of products in the target family brand. Thus, in the earlier Ivory products example, addition of the premise that Neutrogena soap is mild should reduce the strength of the argument that All Ivory health and beauty products are mild.

Finally, proposition 11, inclusion fallacy, states that a specific argument can sometimes be made stronger by increasing the generality of its conclusion. As shown in the example in Table 1, the conclusion that 'All Brand B products' have a particular attribute should be perceived as stronger than the conclusion that an "atypical" Brand B product has the attribute. If so, this "fallacy" would have implications for communication efforts for certain family brands and individual products.

Prior Knowledge

In the present study, as in Osherson et al. (1990), we attempted to control for effects of attribute knowledge (e.g. "machine washable") by using fictitious brands and category attributes whose implications would not affect argument strength (Osherson et al. used obscure biological properties to assure "blank" attributes or properties). Subjects were expected, however, to vary in their knowledge of categories used in arguments (e.g. electronics equipment). Therefore, we will examine an exploratory measure of subjects' product category knowledge as a potential mediator of assessments of categorical argument strength.

Prior research suggests that expertise or prior knowledge is associated with richer, more detailed representations of a category in memory (e.g., Alba and Hutchinson 1987). Experts should better understand relations (e.g. similarities) between categories and between category members, and should process similarity judgements (in argument strength choices) more swiftly and accurately than novices (cf. Muthukrishnan and Weitz 1991).

Furthermore, in the present context, experts may form more specific (sub)categories within a hierarchy, thereby influencing their perceptions of the "lowest level category" that includes both premise and conclusion categories; however, such greater specificity among experts (than novices) should not increase their accuracy in responses as long as experts and novices agree that the relevant lowest level category is more general in the "weaker" argument.

In sum, due to their detailed category knowledge structures, experts may be more adept at judging similarities between categories and may more accurately process categorical arguments. On the other hand, while experts are more likely to perceive specific subcategories as their lowest level categories, this skill may not translate into more accurate argument choices. A final possibility is that prior knowledge will actually interfere with experts' ability to logically process arguments; that is experts' richer knowledge representations may "overload" the available work space (cf. Wyer and Srull 1986) in memory and interfere with the execution of logical argument processing. Given the tentative nature of these competing hypothetical outcomes, we will treat product category knowledge, as a factor mediating subjects' perceptions of argument strength, as exploratory.

METHODOLOGY

Study Overview and Procedure

Fifty-three undergraduate marketing students completed a questionnaire designed to test the validity of each of the seven propositions in Table 1. Subjects were presented with two arguments and asked to select the argument ("a" or "b") whose conclusion was most believable based on the premises (Osherson et al. 1990).

The first page of the questionnaire contained instructions that were read aloud to the subjects. Subjects were told to assume that the statements above the line were facts and to choose the argument whose facts provided a better reason for believing its conclusion (Osherson et al. 1990). The products were described as fictitious, and an example of the task was presented.

Subjects completed judgements for each of the seven propositions, for each of three product classes: (1) electronics, (2) snack foods, and (3) clothing. The order in which the stimuli were presented was systematically varied: for one-third of the subjects, electronics stimuli appeared first (followed by snack foods and clothing), for another third the snack food stimuli were presented first (followed by clothing and electronics), and for the final third clothing stimuli were presented first (followed by electronics and snack foods). The order in which the propositions were presented was held constant across all three stimulus conditions. However, within each stimulus category set the order in which the seven propositions was presented was randomized. Thus, three different random orders of propositions were used for the first, the second, and the third product categories (regardless of the particular category that appeared in that order). The "expected" (i.e. "correct") answer for the contrasting pair of arguments varied, so that 62% of the time the "correct" answer was "a" and 38% of the time the "correct" answer was "b". Following judgements for the 21 stimuli pairs, subjects reported their perceived knowledge in each of the three product classes.

Stimuli

Three superordinate product categories (electronics, clothing, snack foods) were selected in order that exemplars were familiar to the subject population. For each category, relying on researchers' intuition, we selected specific exemplars that were representative or unrepresentative of the category, and that varied in similarity to one another, depending upon the proposition tested. Finally, product attributes were selected that were general or abstract enough to be applied credibly to a range of products within each category. See Table 1 for examples of questionnaire items.

In developing the stimuli, we used fictitious brands for two reasons. First, we were concerned that prior beliefs about a particular brand name would impair subjects' abilities to evaluate the validity of propositions independently of extraneous factors such as company expertise. Also, while the particular product attributes used should be superfluous to the study findings, we believed that they may, in fact, impact judgements of the validity of conclusions for known brands. The use of fictitious brands made it more likely that subjects would follow the directions to assume the premises were true. Second, using fictitious brands allowed us to use a vast array of products within the category (cf. Osherson et al. 1990). In presenting the stimuli, we stated that the particular brand (e.g. "Brand A") includes a wide range of products (e.g. for the electronics category: computers, fax machines, camcorders, TVs, calculators, clock radios, smoke detectors, stereo and other audio equipment, VCRs, and other electronics equipment).

Measures

For each set of arguments, subjects made two judgements. First, subjects were asked which of the two sets of arguments presented was more believable (Osherson et al. 1990), set "a" or set "b". Second, subjects were asked to rate their confidence in this judgement, on a scale from 0 ("not at all confident") to 10 ("extremely confident"). Since preliminary analyses suggested that confidence ratings for high and low knowledge subjects were similar, results of these ratings were dropped. Finally, subjects were asked to rate their overall (subjective) knowledge for each of the three product categories (electronics, snack foods, and clothing) on three single-item scales from 0 ("not at all knowledgeable") to 10 ("extremely knowledgeable").

RESULTS

Overall Ratings

Results of the choice data, across all subjects, presented in Table 2 (see "Overall" column), indicate support for five of the seven propositions. For propositions 1 (premise typicality), 3 (conclusion specificity), 5 (premise-conclusion similarity), 6 (premise-diversity), and 11 (inclusion fallacy), subjects were significantly (p< .005) more likely to select the "correct" set of arguments than the "incorrect" set of arguments, with one exception.

For example, for proposition 1 (premise typicality), subjects were more likely to say that the premises confirmed the conclusion when the premise exemplars were prototypical of the conclusion category. The one exception across these five propositions occurred in the snack foods category for proposition 6. According to proposition 6, the less similar the premise categories are among themselves, the more the premises should support the (specific) conclusion. However, in our study, if Brand B potato chips and Brand B corn chips possessed an attribute ("no cholesterol") then Brand B mixed nuts were also perceived to possess the attribute, more so than if Brand B potato chips and Brand B candy bars possessed the attribute. In retrospect, the particular stimuli selected for this item may have been problematic; the lowest level category that included the premises and the conclusion may have been different than our a priori expectations. While we expected the lowest level category to be "snack foods" in the two arguments, our subjects may have felt, for example, that the lowest level category in the first case was "salty snack foods" and "snack foods" in the second.

Results also indicate a lack of support for propositions 9 (nonmonotonicity-general) and 10 (nonmonotonicity-specific). With regard to proposition 9, for two of the three product categories, when a premise was added to a general argument converting it to a mixed argument, subjects were no more likely to judge a conclusion believable than when the premise was not added to the argument. Thus, findings yielded nonsignificant differences between the two sets of argument choices, for the electronics and clothing categories. For the snack foods data, findings were significant, and in the expected direction. With regard to proposition 10, when a premise was added to a specific argument converting it to a mixed argument, again subjects were no more likely to judge a conclusion believable than when the premise was not added. The lack of significant increase in choice for the specific (vs. mixed) argument occurred in all three product categories (electronics, snack foods, and clothing). In fact, for snack foods, findings were actually significant in the reverse direction, supporting the mixed over the specific argument. Possible explanations for the lack of support for propositions 9 and 10 may be that the propositions are not valid for (1) the types of stimuli used in this study (e.g. fictitious stimuli, the particular product categories selected), (2) consumer products, or (3) consumers who are less knowledgeable about the products. It seems unlikely that the artificiality of the stimulus situation would decrease the validity of certain arguments; in fact, by maintaining a high degree of control over the stimuli, we anticipated that the validity of the proposed arguments would be optimized. However, this possibility is raised again later in the discussion section. It also seems unlikely that the particular product categories selected were unusual, since we included three replicates that represented a broad range of product categories. The effects of knowledge on validity judgements are explored next.

Self-Reported Knowledge of the Product Category

Table 2 presents validity judgements for the seven propositions as a function of self-reported knowledge of the product category. Sample respondents were split into two groups, representing high and low knowledge subjects. A median split was used for each of the three product categories (the median self-reported knowledge rating was 6.0 for electronics, 7.0 for snack foods, and 7.0 for clothing).

As the data indicate, results varied depending upon the product category selected. For the electronics category, high knowledge individuals tended to yield stronger overall support for the propositions. As indicated by the number of individuals selecting the "correct" set of arguments, higher knowledge individuals showed significant support for six of the seven propositions (all but proposition 10), whereas lower knowledge individuals showed significant support for only three of the seven propositions.

TABLE 2

OVERALL, HIGH AND LOW KNOWLEDGE SUBJECTS PERCENT (%) "CORRECT" RESPONSE

For the snack foods category, high and low knowledge subjects both yielded support for five of the seven propositions, all but propositions 6 and 10. For the clothing category, high and low knowledge subjects again both yielded support for five of the seven propositions, in this case, all but propositions 9 and 10.

In sum, self-reported knowledge of the product category appeared to influence validity judgements for the electronics but not the snack foods or clothing categories. Thus, the earlier lack of support shown for propositions 9 and 10 cannot be explained solely by differences in subjects' knowledge. While the raw percentages in Table 2 for propositions 9 and 10, for all three product categories, show higher proportions of "correct" responses for high than for low knowledge subjects (with one exception), the overall pattern of significant chi-squares does not show impressive support for these differences.

An examination of the means and standard deviations for the knowledge measures may suggest at least a partial explanation for the different results across the product categories. A ceiling effect in knowledge scores appears to have existed in the snack foods (mean = 6.87, standard deviation = 1.66) and clothing (7.06 and 1.73) categories, but not in the electronics category (5.60 and 2.12). In fact, over 60% of the subjects in the snack foods and clothing categories rated themselves as 7 or above on the 10-point scale. The wider variability and lower overall mean in the electronics (as compared to the snack foods or clothing) category may have generated a "truer" indication of the differences in knowledge between the high and low groups.

DISCUSSION

Summary of Findings and Limitations

Our preliminary investigation indicated support for five of the seven (Osherson et al. 1990) propositions. Propositions 9 and 10, and proposition 6 in the snack food category, were not supported by the data. These mixed results raise a number of issues about the applicability and generalizability of the Osherson et al. propositions to a marketing setting and about the propositions themselves. A closer look at the propositions that were not supported highlights these issues.

Theoretical Implications

As discussed earlier, the characteristics of the specific stimuli used in the testing of the arguments' believability may have affected the proportion of "correct" responses chosen by subjects. In the case of proposition 6 in the snack foods category, subjects' judgements of the lowest level category including products in the conclusion (mixed nuts) and the premises (potato chips and corn chips; potato chips and candy bars) may have had a greater impact on their assessment of argument strength than the diversity of products presented in the premise. In the other two categories, however, this was not the case; the less similar the products in the premises were among themselves, the stronger the argument, as predicted. Future research should examine to what degree differences in consumers' category structures of consumer products influence the strength of categorical arguments. A limitation of the current study was that the stimuli were chosen on the basis of the authors' judgements of similarity and relevant category hierarchies. Future research should include pretests and/or manipulation checks to verify these judgements.

The findings from proposition 10 are interesting and informative as well. Although this proposition was strongly supported in the Osherson et al. study it was not supported in this experiment, and in fact, the findings for the snack foods category were significant in the reverse direction. One possible explanation for the current results is that, in our experiment, including a common "Brand X" label across products introduced a set-size confound. That is, when presented with information about a category consisting of a group of products produced by the same company (a fact that was made salient by the directions and the common Brand X label), subjects may have placed more credence on an argument containing a greater number of premises than one with fewer premises, regardless of the similarity of the products in the premises and conclusion. Further, the expected hierarchical category structure may have been affected by the Brand X label. Rather than think in terms of a category structure such as "All Brand B products, Brand B salty snacks, Brand B potato chips, etc.," subjects may simply have considered the lowest level category to be "All Brand B products". Thus, in proposition 10, the addition of a new premise to a single existing premise (to convert the argument from specific to mixed) may have made the argument stronger. Proposition 9 could be expected to demonstrate evidence of a similar confound, although these arguments involved the addition of a premise to two existing premises rather than just one. In fact, it seems likely that a set-size effect would be more evident under certain conditions, including cases with (1) a small initial number of premises, and (2) a common label across all premises and conclusions (e.g. "Brand B"). With respect to the latter, the stimuli used by Osherson et al. did not have this type of label framing the premises and conclusions of their arguments.

A final issue raised in this demonstration study concerns the validity of the specific-general distinction that is at the core of many of the hypothesized propositions. A fundamental assumption associated with the propositions is that subjects largely agree with each other about facts related to the hierarchical level of different stimulus objects. For example, for one of our marketing stimuli, it was necessary to assume that the majority of subjects conceptualized cassette tape players and microcassette recorders as belonging to the same level category, that "audio products" was at the next level of the hierarchy, and that the category "electronics" was one level above it. Research on the role of context effects in categorization (e.g., Barsalou 1982; Ratneshwar and Shocker 1988), and on goal-derived and ad-hoc categories (e.g., Barsalou 1982, 1983) raises questions about the conditions under which category groupings may change. Depending on what cues were salient to a given subject during the administration of the experiment, a cassette tape player and microcassette recorder may have been included in an intermediate category such as "products that run on batteries," a category that emphasizes different determinant attributes and contains a more diverse set of members than the "audio products" category.

Although an examination of the hierarchical nature of categories and the effects of context seems particularly relevant to the product categories investigated in marketing, similar concerns may arise for the natural object categories described in the original Osherson et al. research. In an example from their paper, subjects were assumed to be in universal agreement on the fact that bees, ants, and mosquitoes were all members of the next level category "insects" rather than another category such as "household pests". Since the propositions of interest are conceptually based on similarity judgements between the premise and conclusion categories, as well as between the premise categories and members of the lowest level category that includes both premise and conclusion categories, lack of consensus about the existence, and hierarchical levels, of the categories may influence expected results. Osherson et al. acknowledge this limitation by stating that greater predictive accuracy for their model of argument strength would require supplementary principles to account for "on-line" categories (p.200). At a minimum, these considerations would suggest that future research involving the application of any categorical argument strength propositions would require pretesting in order to establish how the majority of consumers think about, and represent, the hierarchical levels associated with different consumer categories. Alternative models (e.g. Sloman 1993), or modifications to the Osherson et al. model, which do not assume the existence of these types of stable category structures, should also be researched to see if they provide better explanations of argument strength in a consumer environment.

Managerial Implications

Although further research will be needed to explore the boundary conditions of these propositions and to assess their generalizability to a wider range of product categories, there are a number of potential managerial applications for this approach to analyzing consumers' judgements and decisions. Because the framework includes propositions for general, specific, and mixed arguments, it can be used to investigate a wide range of issues important to marketing management. The categorical argument conceptualization is broad enough to be applied to issues involving current individual products and family brand categories, product category and sub-category relationships, potential brand and/or line extensions, the fortification and dilution effects associated with extension information, and the merits/consequences of broad versus narrow family brands. Although some of the phenomena suggested by the propositions have already been demonstrated in consumer research (e.g., the typicality or "fit" of the extension category) other areas have yet to be investigated. Additionally, by providing a formal, relatively broad framework within which to explore these managerially relevant issues, the categorical argument propositions may provide a unifying approach to tie together some of the different findings in extension and product category research.

REFERENCES

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

Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (March), 411-454.

Barsalou, Lawrence W. (1982), "Context Independent Information in Concepts," Memory and Cognition, 10 (January), 82-93.

Barsalou, Lawrence W. (1983), "Ad Hoc Categories," Memory and Cognition, 11 (May), 211-227.

Boush, David M. (1993), "Brands as Categories," in Brand Equity & Advertising, ed. David A. Aaker and Alexander L. Biel, Hillsdale, NJ: Lawrence Erlbaum Associates, 299-312.

Boush, David M. and Barbara Loken (1991), "A Process Tracing Study of Brand Extension Evaluations," Journal of Marketing Research, 28 (February), 16-28.

Kardes, Frank (1993) "Category-Based Induction and Product-Line Coherence," paper presented at the American Marketing Association Winter Educators' Conference", Newport Beach, CA.

Keller, Kevin L. and David A. Aaker (1992), "The Effects of Sequential Introduction of Brand Extensions," Journal of Marketing Research, 24 (February), 35-50.

Loken, Barbara and Deborah Roedder John (1993), "Diluting Brand Beliefs: When Do Brand Extensions Have a Negative Impact?" Journal of Marketing, 57 (July), 71-84.

Muthukrishnan, A.V. and Barton A. Weitz (1991), "Role of Product Knowledge in Brand Extension," in Advances in Consumer Research, Vol. 18, ed. Rebecca H. Holman & Michael R. Solomon, Provo, UT: Association for Consumer Research, 407-413.

Osherson, Daniel N., Edward E. Smith, Osmond Wilkie, Alejandro Lopez, and Eldar Shafir (1990), "Category-Based Induction," Psychology Review, 97(2), 185-200.

Park, C. Whan, Sandra Milberg, and Robert Lawson (1991), "Evaluation of Brand Extensions: The Role of Product Feature Similarity and Brand Concept Consistency," Journal of Consumer Research, 18 (September), 185-193.

Ratneshwar, S. and Allan D. Shocker (1988), "The Application of Prototypes and Categorization Theory in Marketing: Some Problems and Alternative Perspectives," in Advances in Consumer Research, Vol. 15, ed. Michael J. Houston, Provo, UT: Association for Consumer Research, 280-285.

Romeo, Jean B. (1991), "The Effect of Negative Information on the Evaluations of Brand Extensions and the Family Brand," in Advances in Consumer Research, Vol. 18, ed. Rebecca H. Holman and Michael R. Solomon, Provo, UT: Association for Consumer Research, 399-406.

Sloman, Steven A. (1993), "Feature-Based Induction," Cognitive Psychology, 25, 231-280.

Wyer, Robert S. Jr. and Thomas K. Srull (1986), "Human Cognition in Its Social Context," Psychological Review, 93, 322-359.

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