Measures of the Attribute Structure Underlying Product Typicality

ABSTRACT - The attribute structure of product categories and its effects on product typicality were examined, for 14 brands of shampoo, in laboratory studies with college students. An alternative to Rosch and Mervis' (1975) family resemblance measure of the attribute structure underlying typicality is proposed, based on multi-attribute attitude theory. Results confirm a positive relationship between perceived product typicality, the proposed measure or attribute structure, and attitude toward the product. Implications for prior conceptual work in categorization, attitude theory, and consumer behavior are discussed.

Citation:

Barbara Loken and James Ward (1987) ,"Measures of the Attribute Structure Underlying Product Typicality", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 22-26.

Advances in Consumer Research Volume 14, 1987      Pages 22-26

MEASURES OF THE ATTRIBUTE STRUCTURE UNDERLYING PRODUCT TYPICALITY

Barbara Loken, University of Minnesota

James Ward, Arizona State University

ABSTRACT -

The attribute structure of product categories and its effects on product typicality were examined, for 14 brands of shampoo, in laboratory studies with college students. An alternative to Rosch and Mervis' (1975) family resemblance measure of the attribute structure underlying typicality is proposed, based on multi-attribute attitude theory. Results confirm a positive relationship between perceived product typicality, the proposed measure or attribute structure, and attitude toward the product. Implications for prior conceptual work in categorization, attitude theory, and consumer behavior are discussed.

INTRODUCTION

When consumers think of snack foods, they recall products like popcorn, applies, and yogurt. Most consumer recall popcorn sooner than yogurt, and regard popcorn as a better, more typical example of a snack food. Consumers also might think of "McDonald's" as a better, more typical example of "fast rood restaurants" than "Colonel Sanders" or "Pizza Hut." More generally, consumers appear to regard some products and brands as more typical than others. More typical products and brands are recalled faster as category members, and may be learned sooner and even used as "standards of comparison" for other products and brands. More typical brands may also be more preferred by consumers. But what is it that determines whether one product or brand is more "typical" of a category than another?

In the present research we explore the determinants of typicality in product categories, and in particular examine the attribute structure underlying typicality, and explore alternative ways of measuring it. Our study advances prior work by (1) developing an improved measure or the attribute structure underlying differences in typicality among products, (2) theoretically and empirically showing how the measure relates to not only the perceived typicality or brands but also affect toward brands, and (3) discussing the implications of these findings for consumer behavior and marketing practice.

Typicality Effects

Research in cognitive psychology shows that categories have graded structure (Lingle, Altom and Medin 1984). That is, people perceive members or natural object categories (like birds, trees, and vehicles) and goal-derived categories (like things to take on a picnic) to vary in their representativeness or typicality of the category.

As category members become more typical, they gain increasing priority in a number of cognitive tasks. Research (Mervis and Rosch 1981 ) shows that more typical instances of a category tend to be:

- named first in free recall or category instances

- classified raster than less typical instances

- classified with fewer errors

l- earned more rapidly as a category member

- used as cognitive reference points in comparisons (more typical members tend to be "standards of comparison" for less typical members)

Recent research also suggests that more typical members of goal-relevant and natural object categories tend to be more preferred than less typical members (Barsalou 1983, 1985), perhaps since they tend to possess attributes that are useful for goal fulfillment to a greater degree than less typical category members.

Marketing researchers have Just begun to explore the implications of prototypicality effects for consumer behavior (see, e.g. Cohen 1982). Initial research interest has focused on whether members of product categories vary in their typicality, and whether typicality is associated with the effects noted above. Recent studies have round that members of product categories vary in typicality, and more typical members tend to be recalled sooner than less typical members (Nedungadi and Hutchinson 1985, Ward and Loken 1986). Aside from Nedungadi and Hutchinson's (1985) finding of a relationship between rated liking and typicality in two product categories (magazines and beverages), the relationship between perceived typicality, affect, and beliefs about product attributes has been little explored.

Given the current interest of consumer researchers in the effects of typicality in product categories, further research is necessary on (1) what factors determine typicality, (2) how these characteristics relate to preference, and (3) how the product or brand characteristics underlying typicality should be measured.

Determinants of Typicality

Research by Barsalou (1985) suggests that the perceived typicality of a category member may be influenced by at least three factors: (1) the category member's family resemblance to other members or the category, (2) the extent to which it has "ideal" attributes, i.e. attributes useful to achieving the goal served by the category, and (3) its frequency of instantiation, i.e., the number or times the category member has been encountered as a member of the category. The relative influence of these factors appears to vary between natural object and goal-derived categories. In natural object categories, Barsalou (1985) round that family resemblance had the strongest correlation with perceived typicality. The other two factors had significant, though smaller, influences. In goal-derived categories, the relationship between family resemblance and perceived typicality was not significant. Only the degree to which an item had an attribute relevant to goal achievement and its frequency of instantiation appeared to be associated with typicality.

The attributes of category members can be viewed, and measured, from two different perspectives. Family resemblance measures the degree to which an item shares attributes with other members or a natural object or taxonomic category. From Barsalou's perspective, only attributes which facilitate achievement of a goal are relevant. These measures, and their potential limitations, are discussed below.

Measures of the Attribute Structure Underlying Typicality

The typicality of a category member, as described by Rosch and Mervis (1975), is a function of its family resemblance, that is, the degree to which the exemplar possesses attributes of other category members and the frequency of those attributes among the members of the category. The measure of family resemblance developed by Rosch and Mervis (1975) is illustrated by the example shown in Figure 1 (Ward and Loken 1986). The example supposes that a product category is made up of four products, and each product has three attributes, represented as ABC, BCD, ADE, and AFG. So compute a family resemblance score, each attribute of every product is weighted by the number of products in the category set that possess it. Next, an overall family resemblance score is computed for each product by summing the weights of each of its attributes.

This procedure assigns a product a higher family resemblance score to the extent that it possesses more attributes, and to the extent these attributes are shared by other category members. In our example, the Product ABC has the greatest family resemblance score.

FIGURE 1

EXAMPLE OF FAMILY RESEMBLANCE SCORE COMPUTATION

Rosch and Mervis (1975) have shown that family resemblance scores correlate highly (in the .8 to .9 range) with independent measures of typicality, such as a single-scale measure that asks for a rating of "how good an example of the category each brand is" (cf. Ward and Loken 1986), for several natural categories.

Recently, Ward and Loken (1986) corroborated these findings for the category "snack foods"--a product category of types of things like popcorn, apples, and yogurt similar to the natural categories Rosch and colleagues have explored in their research. Ward and Loken found a correlation or .87 between the family resemblance scores of snack foods and their rated typicality. However, they found a very low correlation of only .03 between the family resemblance scores of shampoo brands and their rated typicality.

Methodological Problems with Family Resemblance Measures

The above results suggest that Rosch and Mervis' (1975) procedure for computing family resemblance scores may be limited in its application to certain categories (e.g. superordinate categories like types of products) similar to stimuli Rosch has used in her research. Several procedural aspects may contribute to this limited application The measure is computed from subJects' responses to open-ended questions about the "attributes or characteristics you think describe this object," and all attributes and characteristics listed by subjects for any object in the category are ultimately used in the computational process. The list of attributes generated by subJects can be very long, and may include many attributes that are not "salient." In our previous study, some subJects mentioned many more small, idiosyncratic differences in package design (e.g. "gold trim on label") than salient functional and physical differences. Such remarks received heavy weights in determining family resemblance scores even though other data, to be discussed, revealed that features like "color of cap" are not salient purchase criteria for shampoo. The use of a visual stimulus (slides) for exposing subjects to brands of shampoos but not snack foods may have accentuated these tendencies. However, in a test of the effect of presenting category members to subJects via written cues or actual observation of the items, Rosch et al. (1976) round that the measure performed equally well for both stimulus formats.

Another methodological problem with the family resemblance measure is that unique attributes mentioned by one or more subjects about a category member increase the object's family resemblance score (by "1"), seemingly counter-intuitive to the procedure's objective. Furthermore, a good deal of subJectivity is involved in the procedure. For example, coders are allowed to credit a category member with attributes not mentioned by any subject but "clearly and obviously" true of the member (Rosch and Mervis, p. 570). Finally,the family resemblance score of a category member depends on the specific set or products measured. If the set changes, so does the score.

Conceptual Problems with Family Resemblance Measures

Besides these methodological problems, a more fundamental flaw may exist in the we of family resemblance as an explanation for typicality differences among brands. Consumers may Judge the typicality of a brand not by its family resemblance to other brands but instead by the degree to which the brand has attributes related to the goals or uses or the category (Barsalou 1983, 1985). This view is consistent with most models of consumer behavior (e.g. Engel and Blackwell 1982) that assume that consumers view products and brands-as means to satisfy their goals.

The proposed view becomes particularly plausible in view of the recent finding, corroborated by the present study, that a positive relationship exists between typicality and brand preference (Nedungadi and Hutchinson 1985). This seemingly curious relationship would follow logically from an assumption that more typical brands have more valued attributes.

Barsalou's Measure

In a recent study, Barsalou (1985) attempted to measure the degree to which category members possessed "ideal" attributes by a simple procedure. He intuitively chose a dimension along which he felt subJects would evaluate category members. Then he asked subjects to rate each item in a set of categories by the "amount" of the dimension. For example, he asked subJects to rate birthday presents by how happy people are to receive the presents; he rated vehicles by how efficient each is as transportation. The ratings were collected by asking ten subJects to rate a category member on a nine-point scale that measured whether the subJect perceived the item to have a very low amount or a very high amount of the chosen dimension.

Barsalou acknowledges that this measure is an incomplete way to assess the degree to which a category member is ideal. He notes that,

Because only one ideal was observed for each category, and because each ideal was picked intuitively, these experiments only show that ideals are related to typicality; they do not provide accurate estimates of the strength of this relation . . . (p. 641)

Alternative Measure of Attribute Structure

To address the above concerns, an alternative procedure to family resemblance for measuring the attribute structure underlying typicality was developed ln the present research. This procedure measures the degree to which the members of a product or brand category have attributes that are salient in a purchase decision. Therefore, this is an initial attempt to design a measure more objective and complete than Barsalou's, and to solve the problems associated with Rosch's technique. This alternative procedure is based heavily upon prior conceptual work of Fishbein and Ajzen (1975) in the development or a multiattribute attitude model. First, an adaptation of Fishbein and Ajzen's technique for eliciting a hierarchy of salient beliefs was used. In the present case, we asked subJects, ln an open-ended format, to list all the attributes of shampoo that would influence whether they would (1) buy the shampoo or (2) not buy the shampoo. The most frequently mentioned attributes were taken as the "modal salient" attributes of brands of shampoo. This procedure differs from Fishbein and Ajzen's in that the set or sailent beliefs was not measured uniquely for each brand of shampoo within the category, but does allow the development of a measure of graded structure across all items in a category.

In the next step subJects were asked to rate the likelihood that each member of a category had each salient attribute. The average or summed rating across all subjects, for each brand and each salient attribute, was then computed. The sum of these ratings across all salient beliefs represented the brand's "attribute structure" score.

Purpose of Study

In the present study we attempted to test this alternative formulation of attribute structure, as an indicant of the attributes underlying perceived typicality, by relating its results to former data on typicality, family resemblance, and physical similarity. The conceptual basis of our measure has several additional implications. First, since we propose that typicality should be related to a measure of attribute structure, based on the attitude model of Fishbein and Ajzen, we should also rind a link between typicality and a more global measure of attitude. This argument provides a conceptual rationale for previous findings linking typicality ratings to preference (Nedungadi and Hutchinson 1985), and to goal fulfillment (Barsalou 1985).

Second, while for certain product categories the physical features of the product are important to its overall evaluation (Ward, Loken, Ross and Hasapopoulos 1986) or to the formation of functional beliefs about the product (Loken, Ross and Hinkle 1986, Ross and Loken 1983), these features should represent only part of the consumers' underlying belief structure about the product. Since this belief structure is presumed to underlie attitude toward the brand, it should be more strongly related than mere physical similarities to attitude. Therefore, the relationship between family resemblance scores (which appear to be heavily influenced by physical features) and attitude toward brands should be weaker than the relationship between the proposed attribute structure scores and attitude. Another implication of this point it that the physical similarity of brands to all other brands should be more strongly correlated to their family resemblance scores than their attribute structure scores, and thus attitudes. Finally, if consumers primarily think of the brands in product categories in terms of their relevance to purchase goals, attribute structure scores should be correlated more strongly with typicality than physical similarities. Analysis of this relationship should provide insight into whether consumers structure product categories by the goal relevance of items or their physical similarities.

In sum, the present study explores the relationship between perceived typicality and several measures of potential determinants--family resemblance, an alternative measure of the attribute structure of members of product categories, and physical similarities among brands. The study then explores the relationship between attitudes toward brands, their perceived typicality, and the attribute structure that may influence both perceived typicality and brand attitudes.

METHODOLOGY

Overview

Data collected from a set of laboratory studies were used to test the above hypotheses. In the first study, a sample of 66 college students completed measures of typicality and family resemblance for 14 different brands of shampoo. In a second study, 82 students were asked to rate the same 14 brands of shampoo along belief and attitude (affect) dimensions. Finally, a third study had groups of 112 college students rate the physical similarities of all possible pairs of the 14 brands of shampoo. The students were not selected for prior knowledge or use of the brands or lack thereof. Although separating the measures of physical similarity, typicality, beliefs, and affect was not intentional (each study was designed with different purposes in mind, tests different hypotheses from the present study, and is reported in more detail elsewhere), this separation helped to ensure that each type of measure would not affect the others.

Selection or the 14 shampoos was based upon prior research which showed that (1) this set included popular brands as well as less popular private labels (four of the brands were private label imitators of four of the national brands included), and (2) the brands represented a diverse range in terms of physical similarity.

Measures and Procedures

Global Typicality Ratings. As part of Study 1, procedures developed previously by Rosch and Mervis (1975) were used to obtain a global typicality measure of the 14 shampoos. Subjects were asked to judge how good an example of the product category various brands in the category were. While viewing a slide showing each brand of shampoo In its usual package, subjects completed typicality ratings on 0-10 scales With endpoints "extremely poor example" and "extremely good example." So as to avoid confusion with this scale and a "good-bad" attitude scale, subjects were explicitly instructed to use the scale to rate the typicality of items, not their preference for the items. An example of this was provided to them. Barsalou (1985), using scales with "excellent" and "poor" endpoints, found that typicality was correlated with preference for only certain categories (goal-oriented and not "common" categories), providing some evidence that subjects do not confuse "goodness-of-example" ratings with attitude ratings.

Computation of Family Resemblance. To derive attitudes to use in computing family resemblance scores, each subJect in Study 1 gave attribute listings for four different shampoos. At the top or each of the pages intended for listing attributes was the name or the brand or shampoo (e.g. "PERT") and the clause "Attributes or characteristics you think describe this product:".

To compute family resemblance--the degree to which an object has attributes that overlap those of category members--a multi-stage procedure developed by Rosch and Mervis (1975) was followed. First, all attributes mentioned by one or more subjects for any shampoo were written down on a master list. Second, each shampoo for which that attribute had been listed was credited with having that attribute. Following Rosch and Mervis' instructions, 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 shampoos (i.e. no attributes not listed by subJects were added). In cases where the two judges disagreed, no brand-attribute match was included. We round that certain heavily value-laden and ambiguous attributes (e.g. "high quality") had to be omitted from analyses in order to follow the Rosch and Mervis technique, since a judgement about which brands should be credited with these attributes would be highly subjective. These omissions, while infrequent, may have contributed modestly to the low correlation between family resemblance and typicality found in Ward and Loken (1986).

Next, each attribute in the master list received a score that represented the number of brands of shampoo that contained the attribute. Therefore, attributes were weighted by the number of category members that contained them. Finally, to compute the total family resemblance score for each shampoo, the weighted scores of each of the attributes that had been listed for a particular product or brand were summed.

Attitude and Belief Ratings. SubJects in the second study were shown slides of the 14 brands or shampoo, one at a time. The order in which the 14 brands were shown was randomized, then held contant across all subJects. While viewing each brand, subjects completed rating scales about their perceptions of the product's attributes, their attitudes toward the product, and self-reported knowledge of the product. Each of the 14 brands was rated along 9 belief dimensions (see Ward, Loken, Ross, and Hasapopoulos (1986) for a list of the dimensions). These dimensions were developed through open-ended pretests with a sample characteristic or our subject population. In the pretest, subjects were asked to list "the positive and negative attributes, qualities or characteristics of a shampoo that would increase (or decrease) your chances or purchasing it." This procedure is an adaptation of Fishbein and Ajzen's modal salient belief elicitation procedure (1975). The eight most frequently mentioned attributes were used to form belief statements in the main experiment. An additional attribute ("controls dandruff") was added, since two shampoos were dandruff shampoos.

The nine beliefs were measured on likelihood scales from 0 (extremely unlikely) to 10 (extremely likely). Since all attributes were positive, an attribute structure score was computed by taking the average belief rating across all subJects, and summing these ratings across all 9 attributes for each brand.

Attitude toward the brand was measured by asking subjects to rate each shampoo ("Overall, how would you rate this shampoo?") on three 0-10 evaluative scales (high-low quality; good-bad; satisfactory-unsatisfactory).

Physical Similarity. In a third and final study, subjects provided data on the physical similarities between the same 14 different brands of shampoo. Four groups of subjects rated all possible pairs of shampoos, and were asked to make similarity ratings on a scale from 0 (extremely dissimilar) to 10 (extremely similar). Subjects were carefully instructed to judge whether the pairs were similar in overall appearance, not in their functions or quality attributes: "We want you to tell us whether or not the two products look alike in appearance . . . [we] do not want you to tell us whether the two products are similar in what they do for your hair. n An illustration was provided to clarify this difference, and subjects did not appear to have any difficulty in understanding the nature or the task.

Using the data on pairwise physical similarity between brands, we computed a measure of each brand's relative similarity to all the other brands by summing the similarity scores of the brand to all other brands. Brands that were perceived as very similar to other brands received higher summed scores than brands perceived as dissimilar to others.

RESULTS

Pearson's correlations between mean typicality ratings, family resemblance scores, mean attitudes, physical similarities, and attribute structure scores were computed. Correlations are reported in Table 1, and means and ranks are in Table 2 for each brand of shampoo. The correlations show that typicality ratings are strongly related to the proposed measure or attribute structure and to attitudes toward the individual brands. As reported previously (Ward and Loken 1986) and listed in Table 1, typicality was not related to family resemblance for these stimuli. Furthermore, the proposed attribute structure measure is more strongly related than family resemblance to attitudes.

The analyses revealed a strong and significant correlation between family resemblance and physical similarity. This result seems to confirm our belief that in previous research the family resemblance measure captured too much detail about the physical appearance of the objects rated.

Physical similarity was not strongly related to either the proposed attribute structure measure or to typicality.

TABLE 1

PEARSON'S CORRELATIONS

The most notable exception to the overall pattern of correlations was Head & Shoulders shampoo. This brand was ranked high in typicality and attitude, but low in a:tribute structure. These results may have occurred because dandruff control should have received a disproportionate weight ln computing the attribute structure score for Head & Shoulders. In future research the attribute structure measure might be improved by the addition or importance or evaluation weights.

Another explanation for the anomalous performance of Head & Shoulders is the possibility that attributes influencing the typicality of items in the sub-category or "dandruff shampoos" differed from the superordinate "shampoo" category. The relationship between category level, attributes salient for superordinate and subordinate category members, and typicality needs further exploration.

DISCUSSION

The study's results, and the proposed measure or the attribute structure underlying product typicality, have the potential to help integrate several lines or inquiry including research on typicality, goal-derived categorization, and attitude theory. The findings also offer a number of implications for marketing practitioners.

Attribute Structure

The present results suggest that typicality is strongly related to consumers' salient beliefs associated with the brand. The more a brand of shampoo has each or these attributes. the greater that brand's rated typicality.

These results provide evidence that the proposed measure of attribute structure is better related to typicality, and may provide more insight into the underlying determinants of typicality, than Rosch and Mervis' measure of family resemblance. This outcome is encouraging because our procedure has several advantages over traditional family resemblance measures. First, the proposed measure is less subJective and more quantitative than family resemblance measures. Second, it is easier to administer and requires less time and effort ln coding and computation. Finally, it is based upon prior conceptual and measurement work in attitude theory,and provides an alternative theoretical rationale for various findings in the typicality literature (such as the relationship between typicality and preference, noted earlier). The procedure and our results are a step towards integrating prior work in categorization with not only attitude theory but also Barsalou's (1983) view of the structure or goal-oriented categories, a view that is currently attracting the interest of consumer researchers (Nedungadi and Hutchinson 1985).

Evoked Set

Our data, and Nedungadi and Hutchinson's (1985) results, show that more typical brands are recalled as members of a category faster than less typical brands. If typical brands come to mind sooner, and it the consumer screens brands for entry into his or her evoked set by a heuristic such as "I will further consider the first X brands that I recall which pass a minimum performance criterion," then more typical brands should be especially likely to enter the evoked set. Without further explanation, the possible relationship between typicality and propensity to enter the evoked set does not appear functional for the consumer. But our model suggests that more typical brands tend to have higher values on attributes useful for the solution of a consumption problem and therefore should be more likely to enter the evoked set. Consistent with our model, Laroche, Rosenblatt, Brisoux, and Shimotakahara (1982) found that products in the evoked set are more favorably evaluated than other products the consumer is aware Or.

TABLE 2

RANK ORDER OF RAW DATA, BRANDS, EACH MEASURE

Future Research

The theory and results discussed in the present study suggest some possibilities for further exploration. First, future studies might continue to explore the application of attitude theory to understanding the relationship between preference and typicality. In particular, the attribute structure measure might be further improved by the addition of weights for the salient beliefs. Further study or whether the source attributes are salient in determining typicality at superordinate and subordinate category levels would also be useful.

The positive relationship between typicality and preference raises an interesting question: Why do apparently atypical products sell? For example. most people may not rate a Porsche as a typical car. However, as noted before, a Porsche might be regarded, at least by a limited segment or consumers, as a typical "expensive sports car." Future research might also begin to explore how product category prototypes are formed. Research questions might include the influence or early entrance to the category, the influence or market share, and the influence of the consumer's particular sequence of learning about the category (first brand encountered, etc.) on prototype formation.

REFERENCES

Barsalou, L. V. (1983), "Ad-Hoc Categories," Memory and Cognition, 11 (3), 211-227.

Barsalou, L. W. (1985), "Ideals, Central Tendency, and Frequency or Instantiation as Determinants of Graded Structure. n Journal of Experimental Psychology: Learning Memory, and Cognition, 11 (October), 629-654.

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

Engel, J. F. and Blackwell, R. D. (1982), Consumer Behavior, 4th edition, Hinsdale, IL: The Dryden Press.

Fishbein, M. and I. Ajzen (1975), Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research. Reading, MA: Addison-Wesley, 1975.

Laroche, M, J. Rosenblatt, J. Brisoux, and R. Shimotakahara (1982), "Brand Categorization Strategies In RRB Situations: Some Empirical Results," in Richard Bagozzi and Alice Tybout (eds.), Advances in Consumer Research, Vol. 10, 549-554.

Lingle, J. H., M. Altom, and D. Medin (1984), "of Cabbages and Kings in Assessing the Extendibility of Natural ObJect Concept Model to Social Things," in R. Wyer and T. Srull, Handbook of Social Cognition, NJ: Erlbaum.

Loken, B., I. Ross, and R. L. Hinkle ( 1986), "Consumer 'Confusion' of Origin and Brand Similarity Perceptions," Journal of Public Policy and Marketing, Vol. 5.

Mervis, C. and E. Rosch (1981), "Categorization of Natural ObJects," Annual Review or Psychology, M. R. Rozenweig and L. W. Port .E ( eds . ), 32, 89-115.

Nedungadi, P. and J. Hutchinson ( 1985) , "The Prototypicality of Brands: Relationships with Brand Awareness, Preference and Usage," in Elizabeth Hirschman and Morris Holbrook (eds.), Advances in Consumer Research, Vol. 12, 498-503.

Rosch, E. and C. Mervis (1975), "Family Resemblances: Studies in the Internal Structure or Categories," Cognitive Psychology, 7, 573-605.

Ross, I. and B. Loken (1983), "Consumer Psychological Theory Bearing on Trademark Infringement Issues," presented at the American Psychological Association Convention, Division 23, Anaheim, CA.

Ward, J. and B. Loken (1986), "The Quintessential Snack Food: Measurement of Product Prototypes," in Richard Lutz (ed.), Advances in Consumer Research, Vol. 13, in press.

Ward, J., B. Loken, I. Ross, and T. Hasapopoulos (1986), "The Influence of Physical Similarities on Generalization of Affect and Attribute Perceptions from Private Label to National Brands," American Marketing Association Educators' Conference.

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