The Prototypicality of Brands
The Prototypicality of Brands
The Prototypicality of Brands
Prakash Nedungadi and J. Wesley Hutchinson (1985) ,"The Prototypicality of Brands: Relationships With Brand
Awareness, Preference and Usage", in NA - Advances in Consumer Research Volume 12, eds. Elizabeth C.
Hirschman and Moris B. Holbrook, Provo, UT : Association for Consumer Research, Pages: 498-503.
[ direct url ]:
http://acrwebsite.org/volumes/6441/volumes/v12/NA-12
Pages 498-503
Awareness of a product or the ability to bring it to mind has often been considered an important determinant of
choice. Practitioners and consumer researchers alike have been interested in indicators of memorability, since brand
recall may play an important role in determining whether a product is considered for purchase at all. The notion of the
evoked set (Howard & Sheth, 1969; Campbell, 1969; Narayana & Markin, 1975) posits that a set of brands or
products are retrieved by consumers and considered for purchase. Various standard measures, such as aided and
unaided brand name recall and top-of-mind awareness, rest on the assumption that the ability of the consumer to
remember a brand or product will strongly affect the probability of its being considered for purchase.
In an attempt to understand what drives product awareness researchers have examined the relationship between
awareness and a wide variety of marketing and consumer behavior variables (e.g., Axelrod, 1968; Haley and Case,
1979). Important among these are variables such as brand attitude, preference and usage, and advertising recall.
Each of these have been found to be related to brand recall in systematic ways.
Research in psychology and consumer behavior, within the information processing paradigm, has examined the
"strength of association" between concepts and has demonstrated the importance of variables such as frequency and
recency (of exposure) in determining strength of association and thereby in affecting retrievability of a certain concept
given relevant cues. Recent work on categories and categorization processes has established the importance of
"prototypicality" as a determinant of the associative strength between a category concept and members of the
category. Prototypicality (or simply typicality), is a measure of how representative an object is of a category.
Operationally, the prototypicality (of an object) is measured by individuals' ratings of how "good an example" they
consider the object, of a category (Rosch, 1973).
Internal Structure
In classical concept formation experiments, categories were thought of as composed of undifferentiated, functionally
equivalent instances. Any stimulus which fit the definition of the concept (i.e., possessed the necessary and sufficient
criteria for membership) was considered as good an example of the category as any other. However, it has been
subsequently established in a number of naturally occurring domains that all members are not equally representative
of the categories to which they belong, and necessary and sufficient conditions for category membership seldom
exist. More specifically, it has been established that:
(1) Some instances are better examples of a category than others (e.g., Rosch, 1973, 1975; Rips, Shoben & Smith,
1973).
(2) Items exist whose category membership is uncertain (e.g., Labov, 1973; McCloskey & Glucksberg, 1978).
(3) Non-members of a category vary in how similar they are to the category concept (e.g., Barsalou, 1983).
These findings have been considered indicative of the fact that categories possess internal structure. To quote Rosch
(1973), "Categories are composed of a 'core meaning' which consists of the clearest cases (best examples) of a
category, surrounded by other members of decreasing similarity to that core meaning." Thus, objects in a category
vary in their typicality leading to graded, internal structure within the category.
Determinants of Prototypicality
A number of explanations have been proposed for how people assign or arrive at the clearest cases within
categories. In general, two distinct factors have been identified. First, objects that are frequently encountered as
instances of the category are perceived to be representative of the category. Second, objects that possess attributes
that occur frequently within the category are also perceived to be representative. Thus, it is possible for an item to
have high prototypicality even though it is rarely encountered if its attributes are typical of the category.
In research using artificial, statistically generated categories, such as random dot patterns, the 'prototype' could be
identified a priori as the central, on average, member of the category. However for most domains, prototypes do not
precede the category but appear to be formed through the processing of relationships between items within the
categories. Rosch (1975) has argued that the majority of objects encountered in nature possess attributes that are
highly correlated with one another. The attributes "has feathers" and "lay eggs" are not statistically independent
properties; likewise, the "curb weight" and "luxuriousness" of cars tend to be correlated. This correlational attribute
structure leads to natural clusters of objects which get named in various ways so as to form categories. As a result,
objects within the same category tend to have many attributes in common. Moreover, these attributes occur
infrequently in contrasting categories. This high intra-category similarity is referred to as family resemblance (Rosch &
Mervis, 1975)
Cohen (1982) suggests that we should consider affect as residing in the category itself and that we may well
categorize our world to reflect our evaluations of it. This would be especially true for product categories, where
evaluation clearly serves an important purpose. If consumer categories are to be functional, they may well group
products into categories that incorporate affect (e.g., "affordable cars" or "movies I would take the family to"). This
suggests that the prototypicality of a brand could be affected by such marketing variables as the importance and
determinance of attributes, brand evaluations, preferences and levels of usage.
Effects of Prototypicality
Prototypicality has been found to affect a number of major dependent variables used in psychological research which
carry clear implications for consumer behavior. Speed of processing and response times for verification of category
membership were found to be faster for representative members of a category. Importantly the frequency and order
of production of members of a category (in response to free elicitation tasks) was found to be significantly correlated
with degree of typicality (Mervis, Catlin & Rosch, 1976). All of these measures are natural indices of brand awareness
and should be related to the ease with which a brand is recalled when the need for a particular product class arises.
Category membership is learned first for the most representative exemplars and last for the least representative
ones. In addition, categories are learned more easily if initial exposure is to representative exemplars only (Mervis &
Pani, 1980). There is no reason to believe that learning would be any different for product categories. Thus, the
representativeness of a specific brand could have a major impact on the acquisition of brand-related knowledge,
particularly for consumers with low product familiaritY.
Extensions of Prototypicality to Other Domains
The notion of representativeness in categorization has recently spread to quite a few domains besides object
categories. While findings of typicality have been quite robust, the explanations for the development of internal
structure have not always been the same.
Barsalou (1983) studied prototypicality in ad hoc or goal-derived categories -- such as "things to take on a camping
trip", or "things not to eat on a diet". His findings differ from those for natural object categories in certain respects and
are quite relevant to consumer product categorization. Barsalou finds that although the members of such categories
are quite heterogeneous they still give rise to typicality effects. However, typicality was not determined by family
resemblance (i.e., its general similarity to other members of the category) but by
(a) the value of the item on the goal-relevant dimension and
(b) the frequency with which the item had been used as an instance of the concept in the past.
For example, "things not to eat on a diet" would have high calories and edibility as goal-relevant dimensions.
Chocolate and butter, although they may not normally belong to the same product category, are thought of as highly
typical of this ad hoc category.
The determinants of typicality hypothesized by Barsalou may be quite related to usage, frequency of exposure and
liking, and further underscore the need to conduct domain-specific research to establish the locus of typicality effects.
Other interesting research has demonstrated how representativeness could have an effect on processing of
information about categories. For instance. in the area of inductive reasoning Rips (1975) shows how novel
information about a category is often generalized asymmetrically with respect to typicality. New information about
typical members is more likely to be generalized to atypical members than vice-versa. This type of generalization
would be especially important for consumer expectations about pricing. In most circumstances, consumers do not
have access to price information for all brands and all retail outlets. Since prices change on a continuing basis,
inferences about price are a virtual necessity.
Finally, Kahnemann and Tversky (1973) and Tversky and Kahnemann (1973) demonstrate how individuals use
various decision heuristics based purely on the availability and representative less of certain events/items. This often
leads to significant biases in judgment.
AN EXPLORATORY INVESTIGATION OF BRAND PROTOTYPICALITY
Objectives of the Study
Prototypicality appears to be an important product related variable that deserves further study. It could have effects
on diverse consumer behaviors, such as retrieval of product information, generalization within product categories,
making of inferences and ultimately brand choice. It is quite surprising that given the robustness of these findings in
other domains, no attempts have been made to verify its importance in consumer behavior.
The objectives of this study were thus to:
(1) Verify the existence of "prototypical products" and graded, internal structure in consumer product categories.
(2) Examine the importance of typicality as a determinant of product/brand awareness.
(3) Explore the relationships between prototypicality and other important variables such as brand preference and
usage.
(4) Examine the effects of the level of generality at which the product class is defined (i.e., the breadth of market
boundaries) on the above relationships.
Methodology
The principal manipulations in this study pertained to the level of abstraction at which the product classes were
defined. Two superordinate level product categories, magazines and beverages, and four subordinate level
categories, news magazines, business magazines, soft drinks and beers, were used in the study (see Figure 1). A
target set of brands was defined for each subordinate category based on estimated market penetration (Simmons,
1982). The principal dependent measures were unaided brand name recall, rated prototypicality, rated liking and selfreported rate of usage.
FIGURE 1
LEVELS OF ABSTRACT - ION FOR PRODUCT CATEGORIES
Each subject completed three questionnaires during the experimental session. The order in which the questionnaires
were given was the same for each subject. In the first questionnaire, unaided brand name recall was measured.
Subjects were given five minutes to recall as many brands as possible belonging to a particular product class. This
task was performed twice, once for each product class (i.e., magazines or beverages). The three levels of generality
for each product class (i.e., the superordinate or one of the two subordinates) were varied randomly across subjects.
Thus, each subject was assigned to one of nine (3 x 3) possible conditions. - The order in which product classes
occurred (i.e., magazines first vs. beverages first) was also varied randomly across subjects. Unaided brand recall
yielded three empirical measures. Brand Recall was defined as one if the brand was listed, zero otherwise. Top-ofMind Awareness was defined as one if the brand was the first listed, zero otherwise. Finally, Conditional Awareness
was the rank of a brand in the list, given that it was recalled. That is, the first brand listed had rank 1, the second
listed had rank 2, etc. All measures were computed only for those brands that were in the target set. If a target brand
was not recalled, Conditional Awareness was treated as a missing observation.
The second questionnaire asked subjects to rate the prototypicality of each brand on a nine point scale.
Prototypicality was defined and given examples in a manner very similar to that used by Rosch (1973). The
distinction between prototypicality and preference were specifically noted. Two randomly constructed orderings of the
brands, and their corresponding reverse orderings were used, one ordering per subject. These four orderings were
counter-balanced across subjects. The category level to be used in making the prototypicality judgment (i.e.,
superordinate vs. subordinate) was varied randomly within subjects on a brand-by-brand basis and counterbalanced
across subjects.
The third questionnaire asked subjects to rate each brand on a nine point scale indicating their liking for the brand
and to indicate the approximate number of times that they had used the brand in the last month. Note that no
particular level of generality was specified in this task. The order of presentation of brands was randomized anew and
counterbalanced as in the second questionnaire. Seventy-two undergraduates at the University of Florida participated
in this study in return for course credit.
Results and Discussion
Typicality effects. Subjects found the task of rating prototypicality readily understandable and appeared to have no
problem in rating representativeness of brands within a product class. Table 1 shows the mean ratings on typicality
obtained by each brand for both subordinate and superordinate levels. Mere examination of the results indicate quite
a few points of interest.
The phenomenon of graded structure is readily apparent. Subjects consistently rated some brands as being more
typical than others. An analysis of variance found significant main effects of Brand and Level (F[37,1796] = 14.94, p <
.00001, and F[1,1796] = 25.92, p < .00001, respectively). The interaction was also significant F[37,1796] = 3.83, p < .
00001). Subjects was included as a random factor and was significant (F[71,17963 = 6.31, p < .00001). Mean
prototypicality ratings are presented in Table 1. The obtained values appear to be consistent with intuitive
expectations about brand prototypicality. For instance, Time and Coca Cola received high prototypicality ratings at
both the subordinate and the superordinate levels, while New Republic and R.C. Cola received low ratings for both.
Quite in line with research in other domains, the phenomenon of typicality appears relevant to product categorization.
TABLE 1
MEAN PROTOTYPICALITY RATINGS BY BRAND AND LEVEL OF ABSTRACT - ION
An examination of the prototypicality ratings for different category levels reveals expected changes in typicality.
Brands thought of as poor examples of their subordinate category are thought of as poor if not worse examples of
their superordinate category. An interesting reversal is seen for many brands listed as News Magazines. In fact, only
U.S. News-& World Report and New Republic were rated higher at the subordinate level. This illustrates one effect of
misclassifying brands in "ill-fitting" product classes. It also provides evidence that the two measures of prototypicality
are distinguishable and do not simply measure the same underlying construct.
This finding is relevant to the establishment of product-market boundaries in line with consumer perceptions of
product-usage (Day, Shocker, and Srivastava, 1979). If products are classified into categories where they do not fit
"naturally" they will have more in common with contrasting categories and will be far less typical, resulting in poor
associations with the category concept and poor recall. Some of the results discussed in the next section suggest
further implications of product misclassification.
Relationships Between Variables. Prior work on typicality effects using natural object categories, has attempted to
separate the effects of typicality on dependent measures (such as speed of processing or production) from the effects
of other variables such as liking, frequency and familiarity. Prototypicality is posited as a pure 'cognitive' construct
unaffected by more attitudinal or evaluative measures. Rosch (1973) detected low correlations between the typicality
of members in most object categories and liking for these members. Notable exceptions were "crime" and "disease"
where degree of typicality was seen as being tied to negative affect for the category. McCloskey (1980) and Malt &
Smith (1982) show that while typicality is affected by familiarity, variations in typicality do exist independent of
variations in familiarity. Barsalou (1983) shows how typicality in ad hoc categories could be affected by frequency of
prior exposure and goal-relevant evaluation.
As Cohen (1982) points out, product categories could function quite differently from the more "passive" object
categories that have been the traditional domain of research. Consumers are constantly exposed to advertising,
point-of-purchase displays and messages that proclaim benefits of specific brands, often forcing them to make
evaluations prior to purchase. In such a scenario one would expect to see quite different and systematic effects of
affect and frequency. While we do not take a position as strong as Cohen's -- that "affect resides at the category
level" -- it does seem plausible that affect would play a more influential role in the semantics of consumer products
than in other natural categories. For instance, attitudes toward a given brand and personal experience with that brand
might become part of the meaning of the brand concePt.
In order to examine the relationship between prototypicality, brand awareness and brand preference, the mean
values of each variable were computed across subjects for each brand. Prototypicality was measured at two levels of
generality (i.e., Superordinate and Subordinate). Similarly, Brand Recall, Top-of-Mind Awareness and Conditional
Awareness were measured at two levels of generality. Note that means for each level of Prototypicality were
computed across both levels of generality for the memory variables, and vice versa. This aspect of the design
counterbalances certain task artifacts that could generate spurious correlations between the prototypicality and
memory measures. The reported usage rates were used to generate individual usage shares defined at the
superordinate and subordinate levels. Specifically, superordinate usage shares were computed as the brand usage
rate divided by the total usage rate for the superordinate product category. Subordinate usage shares were computed
in the same way, except the total-usage rate was defined at the subordinate level. Since the task did not specify a
level of generality for liking, and ratings are not generally thought to be a ratio scale measure, there was no natural
way in which two levels of liking could be specified. The intercorrelations between all of these variables are given for
Magazines in Table 2 and Beverages in Table
TABLE 2
MAGAZINES (N = 18)
TABLE 3
BEVERAGES (N = 20)
Prototypicality
The most obvious result is that most of the correlations of other variables with Prototypicality are in the expected
directions and are significant. In particular, both levels of Prototypicality were found to be highly correlated with Liking
for both Magazines and Beverages. This is consistent with our earlier discussion of the role of affect in determining
the internal structure of product categories. Specifically, this result suggests that the semantic importance and the
evaluative importance of attributes are similar for these product categories. It is likely that these correlations have
been inflated by method variance attributed to the fact that Prototypicality and Liking are the only rating scale
measures. If this were the sole determinant of the correlations, however, Liking and the two levels of Prototypicality
would be expected to exhibit the same patterns of correlation with the other variables. It will become evident from the
discussion that follows that this is not the case.
A second interesting aspect of the relationship between the two levels of Prototypicality and Liking is that it interacts
with product class. For Beverages, Liking is equally correlated with both levels of Prototypicality. For Magazines,
however, the correlation is higher with Superordinate Prototypicality. This pattern is also present for the correlations
between the two levels of Prototypicality and Superordinate Usage share. Given our earlier discussion of the
misclassification of News Magazines, this result suggests that Liking will be strongly related to Prototypicality only
when "natural" product categories are used. An interest g area for future research will be to examine the relationship
between cognitive and evaluative determinants of naturally occurring market boundaries.
Brand Awareness Measures
The various brand awareness measures exhibited differential sensitivity to the other variables. In general, Top-ofMind Awareness was most highly correlated with Usage Share. Conditional Awareness, on the other hand, was most
correlated with Prototypicality and Liking. Brand Recall did not consistently differentiate between Prototypicality,
Liking, and Usage Share. This suggests that consumers access a product category first through brands they use and
then retrieve brands based on structural aspects of memory such as product attributes. Interestingly, Usage Share is
the variable that is most sensitive to the frequency with which a brand is encountered in the external environment.
Conversely, Prototypicality and Liking are more related to internal, cognitive representations.
Finally, the internal validity of the distinction between superordinate and subordinate levels of generality was
supported by a consistent interactive pattern of correlations. Variables measured at the same level of generality were
consistently correlated more highly than the variables measured at different levels of generality. This was true for both
product classes
CONCLUSIONS
In this paper we have outlined several ways in which prototypicality is likely to be of interest to consumer researchers.
Although the reported results must be considered exploratory, there are several clear implications. First, there appear
to be significant differences between brands with respect to judged prototypicality. Second, contrary to the trend for
ordinary object categories (Rosch, 1973), the prototypicality of brands appears to be significantly related to personal
preference. Finally, different memory-based measures of brand awareness appear to be differentially sensitive to
internal and environmental aspects of product familiarity. In particular, Conditional Awareness was found to be
strongly related to Prototypicality and Liking, but not to Usage Share. Top-of-Mind Awareness exhibited the opposite
trend. Issues regarding the reliability and validity of the brand prototypicality construct and its casual relationships
with other theoretical constructs are important areas of future research.
REFERENCES
Axelrod, J. N. (1968), "Advertising Measures that Predict Purchase," Journal of Advertising Research, 8, 3-17.
Barsalou, L. W. (1983), "Ad-Hoc Categories," Memory & Cognition, 11(3), 211-227.
Campbell, B. H. (1969), "The Existence of Evoked Set and Determinants of its Magnitude in Brand Choice Behavior,"
Unpublished Doctoral Dissertation, Columbia University.
Cohen, J. B. (1982), "The Role of Affect in Categorization: Toward a Reconsideration of the Concept of Attitude," in
Andrew A. Mitchell (ed), Advances in Consumer Research, Vol 9, 94-100.
Day, G. S., A. O. Shocker, and R. K. Srivastava (1979), "Customer-Oriented Approaches to Identifying ProductMarkets," Journal of Marketing, 43, (Fall), 8-19.
Haley, R. I., & P. 8. Case (1979), "Testing Thirteen Attitude Scales for Agreement and Brand Discrimination," Journal
of Marketing, 43, 20-32.
Heider. E. R. (1972), "Universals in Color Naming and Memory," Journal of Experimental Psychology, 93, 10-20.
Howard, J. A., and J. Sheth (1969), The Theory of Buyer Behavior, New York, Wiley.
Kahnemann, D., and A. Tversky (1973), "Subjective Probability: A Judgement of Representativeness," Cognitive
Psychology, 3, 430-454.
Labov, W. (1973), "The Boundaries of Words and Their Meanings." in C. J. N. Bailey & R. W. Shuy (eds.), New Ways
of Analyzing Variation in English. Vol. 1, Washington, D. C.: Georgetown University Press.
Malt, B. C., and E. E. Smith (1982), "The Role of Familiarity in Determining Typicality," Memory and Cognition, 10(1),
69-75.
McCloskey, M. E. (1980), "The Stimulus Familiarity Problem in Semantic Memory Research," Journal of Verbal
Learning and Verbal Behavior, 19, 485-502.
McCloskey, M. E., & S. Glucksberg (1978), "Natural Categories. Well Defined or Fuzzy Sets?," Memory & Cognition,
6, 462-72.
Mervis, C. B., J. Catlin, and E. Rosch (1976), "Relationship Among Goodness-of-Example, Category Norms, and
Word Frequency," Bulletin of The Psychonomic Society, 7, 283-84.
Mervis, C. B., & J. R. Pani (1980), "Acquisition of Basic Object Categories," Cognitive Psychology, 12.
Narayana, L. L., and R. T. Markin (1975), "Consumer Behavior and Product Performance: An Alternative
Conceptualization," Journal of Marketing, 39, 1-6.
Rips, L. J. (1975), "Inductive Judgements 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. (1973), "On the internal Structure of Perceptual and Semantic Categories," in T. E. Moore (ed) Cognitive
Development and the Acquisition of Language, 111-44, New York, Academic Press.
Rosch, R. (1975), "Cognitive Representations of Semantic Categories," Journal of Experimental Psychology, 104,
192-233.
Rosch, R. (1978), "Principles of Categorization," in E. Rosch and B. B. Lloyd (ed) Cognition and Categorization, 2748.
Rosch, E. & C. B. Mervis (1975), "Family Resemblance Studies in the Internal Structure of Categories," Cognitive
Psychology, 7, 573-605.
Simmons Market Research Bureau, Inc. (1982), Simmons 1982 Study of Media and Markets, Vols. 1, 17, 18.
Tversky, A., and D. Kahnemann (1973), "Availability: A Heuristic for Judging Frequency and Probability," Cognitive
Psychology, 5, 207-32