The Quintessential Snack Food: Measurement of Product Prototypes

James Ward, University of Minnesota
Barbara Loken, University of Minnesota
ABSTRACT - A laboratory study was conducted to assess the applicability of family resemblance and prototypicality measures developed by Rosch and Mervis (1975) to product categories. Free recall category member listings, family resemblance scores, and mean prototypicality ratings were rank-ordered and correlated for 16 snack foods and 16 brands of shampoo. Snack food, but not shampoo, data were consistent with earlier findings in categorization research. Possible explanations for these differences, including the level of category abstraction and the use of a visual stimulus, are discussed.
[ to cite ]:
James Ward and Barbara Loken (1986) ,"The Quintessential Snack Food: Measurement of Product Prototypes", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 126-131.

Advances in Consumer Research Volume 13, 1986      Pages 126-131

THE QUINTESSENTIAL SNACK FOOD: MEASUREMENT OF PRODUCT PROTOTYPES

James Ward, University of Minnesota

Barbara Loken, University of Minnesota

[The authors wish to thank Lisa Snyder, Jon Strobel, and Jan Shimanski for their comments and assistance in the study design.]

ABSTRACT -

A laboratory study was conducted to assess the applicability of family resemblance and prototypicality measures developed by Rosch and Mervis (1975) to product categories. Free recall category member listings, family resemblance scores, and mean prototypicality ratings were rank-ordered and correlated for 16 snack foods and 16 brands of shampoo. Snack food, but not shampoo, data were consistent with earlier findings in categorization research. Possible explanations for these differences, including the level of category abstraction and the use of a visual stimulus, are discussed.

INTRODUCTION

The growing prominence in the marketplace of generic products, private labels, and "me-too" products may be one factor which has prompted recent attention to the effects of product typicality and similarity on consumer behavior. Nedungadi and Hutchinson (1984), Troye (1984), Howard (1983), Cohen (1982), and Hirschman (1981) have discussed the effects of "typicality" on issues ranging from the composition of the awareness set to strategic planning. But an important issue that has not been addressed in the consumer behavior literature is what underlying factors determine whether a product is judged more or less typical of a product category.

According to Rosch and Mervis (1975), objects within natural categories vary in their typicality because they vary in their degree of family resemblance to one another. Rosch defines the degree of a category member's family resemblance to other category members as the degree to which it has attributes in common with the other members. More typical category members have a greater degree of family resemblance, i.e. have more attributes characteristic of other members of the category. Around this core of prototypical members radiate less typical items that share progressively fewer characteristics of other category members as the fringes of the category are approached. So typicality effects are a function of the relation of a category member to the attribute structure of a category. Rosch and Mervis (1975) demonstrate that family resemblance scores and independent measures of typicality that do not rely on attribute listings are correlated in the .8 to .9 range for several natural categories.

Purposes

This paper has five purposes: (1) to discuss the implications of the family resemblance structure of categories for several theoretical and applied issues, (2) to address the issue of what factors underlie typicality judgements by demonstrating the graded structure of product categories, (3) to assess the relation of family resemblance scores to such variables as brand awareness and subjects' subjective perceptions of typicality, measured independently of attribute listings, (4) to lay out a practical, efficient methodology for computing family resemblance scores and (5) to discuss some of the problems that may be encountered in attempted applications of the techniQue.

Further insight into how family resemblance is measured and how it relates to the attribute structure of a category can be gained from an example. Suppose a simple category is made up of four products, each of which consumers perceive to have three discrete attributes. The products are ABC, BCD, ADE, and AFG. Note that no single attribute is common to all the products but that each product shares at least one attribute with others. The family resemblance of these products could be computed by setting up the product by attribute matrix shown in Figure 1, checking the attributes each product possesses, and then weighting each shared attribute by the number of products that possess it. Finally, each product's weighted sum of scores is totaled to yield an overall family resemblance score. In this category ABC has the highest family resemblance score and would tend to be rated as most typical of the category. ABC's attributes overlap the more frequently occurring attributes of other category members to a greater degree than any other member's attributes.

FIGURE 1

EXAMPLE OF FAMILY RESEMBLANCE SCORE COMPUTATION

The measurement of family resemblance has a number Or very practical marketing implications. A few are discussed below.

Cognitive Effects Or Typicality

The measurement of family resemblance might help marketers to understand the effects of typicality on a product's likelihood of entering the consumer's awareness set and from there his or her evoked set. Several cognitive advantages accrue to prototypical members of a category. They come into a consumer's "awareness set" first as category members, and are classified more quickly and accurately than less prototypical stimuli as category members (Rips, Shoben, and Smith, 1973; Mervis and Rosch, 1981). Measuring family resemblance shows which attributes most influence the degree of a product's prototypicality within a category. Given this understanding, products could be designed or repositioned so they would "come to mind" quickly and easily as members of a particular category. Perhaps with this intent, trade associations of both dairy producers and orange growers have attempted to promote milk and orange juice as snack foods good anytime and anywhere, two attributes that our empirical data will show are very

typical of snack foods.

Consumers may also use typical products as reference points for the evaluation of other products. Rosch (1975) has shown that in statements of the form ______" is essentially _____ ", subjects consistently place more prototypical category members in the latter position. This finding suggests that consumers may use prototypical products as references for evaluating other less typical category members.

Mervis and Pani (1980) have shown that people learn about a category more quickly and accurately if they are initially exposed to more prototypical examples of the category rather than nonrepresentative examples. Furthermore, typical members of a category tend to be learned before atypical members by children (Mervis and Rosch, 1981; Mervis, 1980). These findings suggest that trade associations or marketers who wish to educate consumers about a new product category could introduce consumers to more representative examples of the category first if their goal is to maximize the speed and efficiency Or learning.

Strategic Rationales for Typicality

The strategic circumstances that favor designing a more typical product are several. The marketer may wish to facilitate the acceptance Of a new type Of product in an established category. Family resemblance scores might show that "Taro root" would be more readily categorized as a snack food if the product were designed to have prototypical snack food attributes (e.g. convenient packaging, easy to eat finger rood).

If a marketer chooses not to pursue a strategy of product differentiation, the design of a more prototypical product might have certain advantages. A prototypical product might encourage consumers to infer that the attributes Or competitors' products also apply to the prototypical product (Mervis and Pani, 1980). Therefore, marketers who wish to compete with larger, better known rivals whose products have a reputation for quality may benefit by designing a prototypical product that maximizes generalization from competitors' products (cf. Peterson 1985).

Awareness Set and Evoked Set

Past marketing studies have explored the effect of typicality on cognitive variables but no past study has assessed what features determine the judged typicality of a product. Nedungadi and Hutchinson (1984) demonstrated the effect of typicality on the salience of a brand in a consumer's awareness set. They found that the prototypicality Of a brand of beverages or magazines is significantly related to a brand's order of mention in unaided recall Of brands in a designated product category and, furthermore, that prototypicality is related to brand attitudes. Troye (1984) also demonstrated the effect of typicality on the consumer's awareness. However, neither Nedungadi and Hutchinson (1984) nor Troye (1984) empirically assess the perceived attributes of products that underlie and produce their subjects' variations in typicality or brand similarity ratings.

Product Positioning

Howard (1983) has suggested use Or Rosch's theory in product positioning. Howard suggests that the concept of a "product hierarchy", a consumer's cognitive picture Or a product market, is important for marketers to understand because consumers cognitively position a product by its family resemblance to other products in the hierarchy. The product hierarchy has several uses: (1) It suggests a customer responds to a new and unknown product by matching it to the product it most closely resembles in the hierarchy. This match determines who existing competitors to the product will be. (2) The hierarchy is a means Of mapping and identifying market segments. If a customer's ideal product is distant from the prototypical product in a category, a market segment is suggested. (3) The product hierarchy helps determine whether a product will be perceived as an innovation or not. Howard characterizes the product hierarchy as a very practical concept, but argues that it is not well-measured and that the measurement techniques used by Rosch are not easily adapted to marketing practice. The present study is a step toward assessing the practicality of Rosch's concepts and measures in an applied context

Study Overview

The purposes of the first empirical study were to (1) test whether product categories have a graded structure that can be measured by family resemblance scores, (2) test the relation of such family resemblance scores to prototypicality ratings and free recall production ranks, and (3) lay out a procedure for computing family resemblances. These purposes are intended to explain and validate an alternative, richer measure of product typicality. Two categories, types of snack foods and brands Of shampoo, were used as stimulus categories. These were selected because they differed in their level of abstraction. Snack foods represented a superordinate category Or product types, very similar to the natural categories used by Rosch and Mervis (1975). In contrast, brands of shampoo represented a more subordinate category, consisting of a relatively homogeneous set of stimuli. Consumers may focus on product types or brands depending upon the decision they need to make, how their attention is directed by promotions and situational factors. Brands of shampoo were selected to include a representative sample, including both national and private-label brands, both frequently and infrequently used shampoos. Protoptypicality ratings, free recall of products or brands and attribute listings used to compute family resemblance scores were collected for each product category. The data collection procedure is reported below.

METHODOLOGY

Overview and Procedure

Snack Foods. Procedures developed previously by Rosch and Mervis (1975) were used, in a laboratory setting, to obtain family resemblance scores and two other measures of prototypicality. The study consisted of three parts. In part 1, subjects were asked to list al' the snack foods they could think of, in a free recall format, in the order that they thought Of them. Subjects wrote each snack food on a separate line, and up to 20 different types or snack foods could be listed. Subjects were given 1.5 minutes to complete this task. No one gave all 20 responses. In part 2 of the study, subjects completed prototype ratings (described more fully later) for 16 types of snack foods. Subjects were given 20 seconds to complete each rating. Finally, subjects listed the attributes Of each snack food to provide the raw data for computation Of family resemblance scores.

Shampoos. The same procedures discussed above were used on an independent group of subjects to obtain family resemblance scores, free recall of brand names, and prototypicality ratings for 16 brands Of shampoo. Differences between procedures used for shampoo and snack food stimuli were slight, but where appropriate are noted below.

Subjects

Subjects were students in two introductory marketing classes who participated in the study as part Or the course. Sample sizes were 66 in the first class, where shampoo stimuli were tested, and 33 in the second class, where snack foods were tested.

Free Recall Category Member Listings

At the start Or the experiment, prior to being exposed to individual brands of shampoo or individual snack food, subjects were asked to list, on a page, "as many brands Or shampoo [snack foods] that you can think Of, in the order that you think of them. n Subsequently for each of the sixteen shampoo stimuli and each of the sixteen snack food stimuli an average rank order of mention was computed from the free recall lists provided by subjects. Production norms for the stimuli were then created by ranking the snack food and shampoo sets from lowest to highest average rank.

Prototypicality Measures

Following Rosch and Mervis (1975), subjects were instructed to rate their perception of the prototypicality of products as shown below. In the case of snack foods (shown in brackets), the instructions varied slightly in that the type Of snack food was written on the questionnaire form and a slide Of the snack food was not shown.

. . . In this study we would like you to judge how good an example of a product category various brands in [members of] the category are. The brands in [members of] the category will be shown in a picture; you will be told the name Of the product category and then shown pictures of some items in category.

. . . Then rate how good an example Of the category each brand [food] is on a 0-10 point scale. A 10 means that you feel the picture [food] is a very good example Of your idea or image of what the category is; a O means you feel the picture [food] fits very poorly with your idea or image Of the category (or is not a member at all). A 5 means you feel the picture [food] fits moderately well. Use the other numbers of the 0-10 point scale to indicate intermediate judgments

A brief example was provided for color ("red" vs. less "red") and dog ("German Shepherd" vs. "Pekinese") categories to explain this technique.

Subjects completed prototypicality ratings on 0-10 scales with endpoints "extremely poor example" and "extremely good example". Sixteen brands Of shampoo were rated in class 1 and sixteen snack foods were rated in class 2

Family Resemblance Measures

Each Or the 16 snack foods and 16 shampoos were printed at the top of a page. Subjects in class 1 each gave family resemblance attribute listings for one of 4 different sets cf 4 shampoos. Each subject in class 2 provided attribute listings for one Of two different sets Of 8 snack foods. Within each class, all sets Or stimuli were counter-balanced to control for order Or presentation effects. Instructions (again following Rosch and Mervis, 1975) included the following:

. . . The next part Of the project is a simple experiment to find out the characteristics and attributes that people feel various brands (snack foods ) possess .

For example, if I asked you to list the characteristics of Tylenol pain reliever, you might list such characteristics as red bottle cap, extra-strength, round bottle, safety-sealed, red and white capsules, acts fast, is expensive, etc.

. . . At the top of the page . . . you will see the first brand [snack food] you should write about, then the next on the second page and so on .

At the top Of each Or the pages intended for listing attributes was the name Or the product (snack food, shampoo) and the clause "Attributes or characteristics you think describe this product:"

To compute a measure Of family resemblance--the degree to which an object has attributes that overlap those of other category members--for snack foods and shampoos, procedures developed by Rosch and Mervis (1975) were employed. First, all attributes mentioned by one or more subjects for any snack rood (or shampoo) were written down on a master list. Second, each snack food (or shampoo) for which that attribute had been listed was credited with having that attribute. Following Rosch and Mervis' instructions, the two researchers deleted cases for which the brand-attribute match was clearly false, and included cases for which a brand-attribute match was clearly true but not mentioned by subjects. No new attributes were added to the master list for snack foods or shampoos (i.e. no attributes not listed by subjects were added). In cases where the two judges disagreed, no brand-attribute match was included. Heavily value-laden attributes were omitted from analyses (e.g. "high quality") since a judgment about which brands should be credited with these attributes would be highly subjective.

Next, each attribute in the master list received a score from 1 to 16, representing the number Of types Or snack foods (or shampoos) that contained that attribute. Therefore, attributes were weighted by the number Of category members that contained them. Finally, to compute the total family resemblance score for each snack food and each shampoo, the weighted scores of each or the attributes that had been listed for a particular product or brand were summed.

RESULTS

Results, reported in Table 1, demonstrate that family resemblance is related in expected ways to the two typicality measures in the case of the superordinate product category "snack foods." Spearman rank order correlations between median free recall production ranks, mean prototypicality ratings, and family resemblance scores for snack foods are all highly significant, ranging from .82 to .92. Rank order data are shown in Table 2. As shown, the rank orders yielded by the different measures are markedly similar. Furthermore, the rankings have face validity. Quintessential snack foods like potato chips, popcorn and peanuts tended to rank highly (be very typical) on all measures. Inspection Or Table 2 supports Rosch and Mervis' (1975) argument that family resemblance underlies and determines typicality. Snack foods such as popcorn which have the highest family resemblance scores--i.e. have attributes which overlap those Or other snack foods to the greatest degree--have the highest typicality scores.

TABLE 1

SPEARMAN RANK/ORDER CORRELATIONS

TABLE 2

RANK ORDER DATA FOR SNACK FOODS

Further analysis Or our data provides insight into the underlying graded structure of product categories. According to Rosch and Mervis (1975), WA family resemblance relationship consists of a set of items of the form AB, BC, CD, DE. That is, each item has at least one, and probably several, elements in common with one or more other items, but no, or few, elements in common to all items" (p. 575). Figure 2 shows the number of attributes that respondents applied to only one product type, two types. three and so on to all sixteen types.

FIGURE 2

NUMBER OF ATTRIBUTES THAT APPLY TO EACH NUMBER OF CATEGORY MEMBERS

The figure supports Rosch and Mervis' (1975) argument that members of a category do not possess necessary and defining features that are common to all members of that category. No attribute was shared by all sixteen category members. Instead, snack foods have features that overlap some but not all other members of the category. The figure reveals that the different product types in the category were perceived as possessing relatively few (an average of 4-5) completely unique product characteristics

The correlations for the shampoo data between mean prototypicality ratings, free recall production ranKs and family resemblance scores are also reported in Table 1. The relationship between the rank order of mean prototypicality ratings and median free recall production rank is strong and significant although significantly lower (p C .05) than the correlation for snack foods. However, the correlation of family resemblance scores with the other measures is markedly low and nonsignificant, and significantly lower than the corresponding correlations for snack foods (p < .05). The possible reasons for the difference in results between the two types of categories--a superordinate category of product types and a more subordinate category of brands within a type--suggest the circumstances under which family resemblance scores can and cannot be validly computed.

By inspection of the attribute listings, we found that respondents reacted to the shampoo stimuli very differently than the snack food stimuli. Differences in family resemblance scores for shampoos were determined to a great extent by small, idiosyncratic differences in package design (e.g. "gold trim on label"). We speculate this is because shampoo attributes (e.g. "cleans hairs') did not differentiate the brands to which subjects were exposed, and therefore were not mentioned. In contrast, the family resemblance scores for snack foods were better related to the degree to which the foods shared salient attributes of the category. The subjects' tendency to focus on minor details of package design in the shampoo attribute listings but not the snack food listings may have been accentuated by three factors: (1) subjects viewed slides of shampoos but not snack foods while listing attributes; (2) the instructions may have focused subjects on details of package designs which were salient for shampoos but not snack foods; (3) subjects listed attributes of different brands of shampoo and different types of snack food. The use of slides for shampoos but not snack foods likely increased subjects' mention of package details in describing the shampoos simply by making such details more available. In the absence of the slides, subjects might have focused more on attributed related to the brands' functions or images. Subjects also might have focused on package details simply because brands of many personal care and household products tend to be differentiated more by the details of package design than by substantive differences. Furthermore, the correlations between the measures for brands of shampoo may have been weaker than for types of snack food because of the influence Or other variables like the relative advertising budgets of brands, their length of time in the marketplace, and familiarity. Unfortunately, the influence of these latter factors is difficult to assess in an experiment using natural stimuli. A similar study using artificial brands could control for such variables as frequency of exposure to advertising.

DISCUSSION

Measurement of Family Resemblance

To increase the comparability of our data with that of past studies, we closely followed Rosch and Mervis' (1975) procedure for computing family resemblance scores. In doing so, we noted that some aspects of the procedure were problematic and might be improved upon for our stimuli. The instructions to subjects, as reported in Rosch and Mervis (1975. D. 578) and adapted here, did little to focus subjects on the more salient common attributes of objects so subjects tended to mention many trivial attributes which made coding and computation complicated and time consuming. Perhaps instructions to focus on salient common attributes, or a time limitation placed on responses would simplify the procedure.

Another difficulty pertained to the way family resemblance scores were computed. If one or more subjects mentioned that an object had an unique attribute, not shared by any other item in the category, the unique attribute still increased the object's family resemblance score by "1". Thus, an object with a plethora of unique characteristics could conceivably gain a higher "family resemblance" score than another very typical object. This scoring procedure seems to be a serious conceptual flaw in the measure.

Finally, the Rosch and Mervis procedure allows researchers to credit an object with attributes not mentioned by any subject but "clearly and obviously" (p. 570) true of the object. Allowing such a degree of judgement may compromise the objectivity of the measure.

Considering the problems discussed above, an opportunity exists to improve upon the current means of measuring family resemblance. Further research is needed on the reliability of the current measurement procedure across different stimulus sets at different levels Of abstraction from subordinate to superordinate categories. The robustness of Rosch and Mervis' technique to changes in task instruction and to the means of presentation of the stimulus items also requires further investigation.

Structure and Manipulation of the Evoked Set

The measurement of family resemblance could be useful in the study of a number of applied issues. Members of a category having more family resemblance to other members are more likely to be part or a consumer's awareness set for a product category and thus have a greater chance of inclusion in a consumers' evoked set since the consumer can only select products from this set that "come to mind" as a member Of the awareness set. In measuring prototypicality, family resemblance scores have an advantage over simple ratings because they do more than merely predict inclusion in the awareness set. The data entering into their computation shows which Of a product's attributes have the most weight in determining a product's membership in and order Of mention in the awareness set.

Attributes shared by the greatest number Or snack foods have the most weight in making a product more typical of snack foods and thus more likely to be mentioned as a snack food. Table 3 shows the attributes shared by eight or more snack foods. The Table shows that a food is more likelY to be perceived as a snack item if it has different flavors or varieties, it is good with other foods, it is easy to prepare and easy to eat, it tastes good and it can be eaten any time of the day in a variety of situations. The most typical snack foods-popcorn, potato ships, peanuts--share almost all these attributes. Information on the determinants Of typicality could be useful to a marketer in a number of ways. For example, a marketer wishing to promote cauliflower as a snack item could advertise its versatility, convenience, and comPatibility with many other types Of foods.

TABLE 3

COMMONLY SHARED ATTRIBUTES OF SNACK FOOD

Besides their utility for new product positioning and design, the process Of computing family resemblance scores revealed some insights into why consumers perceived existing products as more or less prototypical of a category. For example, subjects ranked apples as a rather typical snack food--more prototypical than other fruits and vegetables such as carrots or salad. Inspection of the brand by attribute matrix used to compute the family resemblance scores revealed some possible reasons for this outcome. Apples appear to share attributes with traditional snack foods like popcorn, potato chips and peanuts. Like these snacks, apples are crunchy, easy to prepare, appropriate for many occasions, liked by most people, and easy-to-eat "finger foods". Subjects also noted that apples are "roundish" like traditional snack foods, divisible into small pieces, have many varieties and flavors, and go well with many other foods. Apples were perceived as like popcorn in certain ways including low in calories, and a good source Or roughage. Overall, then, apples were very prototypical Or the snack food category.

For the marketer who wishes to persuade consumers that a product is a member of another category or appropriate for some use or occasion, the concept of family resemblance offers interesting insights. Efforts to convince consumers the product is similar to members of the target category on commonly occurring characteristics could be effective. For example, a marketer who wished to encourage consumers to think of apples as snack foods could create promotions emphasizing the characteristics that apples share with very prototypical snack foods like popcorn, potato chips and peanuts.

Application to Consumer Inferential Reasoning and Advertising

The measurement Or family resemblance may be useful in the study Of consumers' inferential reasoning about the members Of a product category. Rips (1975) found that people tend to generalize new information about a product category asymmetrically. In reasoning about products, consumers may be more likely to generalize new information about more prototypical products to less prototypical products than vice versa. For example, if consumers perceive McDonald's hamburgers as prototypical and hear the rumor that McDonald's hamburger contains red worm meat, they may be more likely to generalize the suspicion to other fast food hamburgers than if they heard the same rumor about a less prototypical burger.

These insights into the effect Of family resemblance on inferential processes have implications for advertisers, particularly comparative advertisers. A more prototypical product may benefit less from point-by-point comparative advertising or advertising that focuses on specific features Of the product than its competitors because consumers may tend to infer less prototypical products have the attributes Of more prototypical products. In actual practice, many dominant, well-known brands like McDonald's avoid the use of point-by-point comparative advertising. Additional research should be useful to clarify these issues and their marketing applications.

Correlated Attributes

Natural categories appear to reflect the correlational structure cf the environment. That is, items in a natural category have attributes that tend to occur in correlated clusters. Thus if a consumer classifies a food as a snack food, he or she is likely to infer the food possesses a set of correlated attributes such as "different varieties," "good with other foods," "easy to prepare," "easy to eat finger food," and "appeals to kids" (Table 3). The consumer may be most confident, in a probabilistic sense, that the food has more typical attributes Of snack foods since these occur among snack foods with the greatest frequency.

If a consumer classifies a product as a member Of a category with correlated structure, then notes the product has a particular attribute like "different flavors" (perhaps emblazoned on the package) then he or she is likely to infer the product also has attributes highly correlated with the attribute initially known. For example, if a product is promoted as having many varieties, consumers might infer it has high calories if these two attributes tend to co-occur in a particular category. This tendency could lead consumers to make unexpected inferences in response to a promotion stressing a particular feature.

The measurement of family resemblance affords insight into the correlational structure of a category. Such insight may provide marketers with clues to the inferences consumers make as a result of categorizing a product or crediting it with a particular attribute.

REFERENCES

Cohen, J. B. (1982), "The Role of Affect in Categorization: Toward a Reconsideration of the Concept of Attitude," Advances in Consumer Research, Andrew A. Mitchell, (ed.), 9, 94-100.

Hirschman, Elizabeth C. (1981), "The Roles of Perception and Preference in Consumer Decision Making," Proceedings 1981 Educators' Conference, Series 47, American Marketing Association, 185-188.

Howard, John A. (1983), "Marketing Theory of the Firm," Journal of Marketing, 47 (Fall), 90-100.

Mervis, C. B. (1980), "Category Structure and the Development of Categorization," Theoretical Issues in Reading Comprehension, R. Spiro, B. C. Bruce, W. F. Brewer, (eds.), Hillsdale. NJ: Erlbaum.

Mervis, C. B. and J. R. Pani (1980), "Acquisition of Basic Object Categories," Cognitive Psychology, 12, 496-522.

Mervis, C. B. and E. Rosch (1981), "Categorization of Natural Objects," Annual Review of Psychology, M. R. Rosenzweig and L. W. Porter (eds.), 32, 89-115.

Nedungadi, Prakash and J. Wesley Hutchinson (1984), "The Prototypicality of Brands: Relationships with Brand Awareness, Preference and Usage," working paper, Marketing Department, University of Florida, Gainesville, FL.

Peterson, Robin T. (1985), "Product Differentiation is Not for Everyone," Marketing News, Vol. 19, 3, AMA.

Rips, L. J. (1975), "Inductive Judgments about Natural Categories," Journal of Verbal Learning and Verbal Behavior, 14, 665-681.

Rips, L. J., E. J. Shoben, and E. E. Smith (1973), "Semantic Distance and the Verification of Semantic Relations," Journal of Verbal Learning and Verbal Behavior, 12, 1-20.

Rosch, E. (1975), "Cognitive Reference Points," Cognitive Psychology, 7, 532-547.

Rosch, E. and C. B. Mervis (1975), "Family Resemblance Studies in the Internal Structure of Categories," Cognitive Psychology, 7, 537-605.

Rosch, E., Carol Simpson and R. Scott Miller (1976), "Structural Bases of Typicality Effects," Journal of Experimental Psychology: Human Perception and Performance, 2, 4, 491-502.

Troye, S. V. (1984), "Evoked Set Formation As a Categorization Process," Advances in Consumer Research, Thomas C. Kinnear (ed.), 11, 598-603.

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