Familiarity and the Structure of Product Knowledge

Jerry N. Conover, University of Arizona
ABSTRACT - The relationship between product familiarity and cognitive structure was assessed for two common consumer products. Responses to a modified repertory grid test indicated that more familiar consumers maintain relatively more complex knowledge structures, in contrast to theoretical predictions. Variable effects of familiarity on knowledge abstractness were observed . Implications of these findings for theories of product knowledge are discussed.
[ to cite ]:
Jerry N. Conover (1982) ,"Familiarity and the Structure of Product Knowledge", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 494-498.

Advances in Consumer Research Volume 9, 1982      Pages 494-498


Jerry N. Conover, University of Arizona


The relationship between product familiarity and cognitive structure was assessed for two common consumer products. Responses to a modified repertory grid test indicated that more familiar consumers maintain relatively more complex knowledge structures, in contrast to theoretical predictions. Variable effects of familiarity on knowledge abstractness were observed . Implications of these findings for theories of product knowledge are discussed.


The notion that consumers approach their decisions in different ways as they gain familiarity through experience with a product has long been popular with consumer researchers (e.g., Bettman, 1979; Howard & Sheth, 1969). The processes by which information is acquired, interpreted, and utilized in consumer decisions should reflect the quantity and nature of relevant information already gained through previous experience. Recent empirical evidence confirms the influential role of product familiarity in information processing (Bettman & Park, 1980; Johnson & Russo, 1981; Raju & Reilly, 1980).


Noting the need for a theoretical framework within which to understand the mechanism whereby familiarity exerts its influence, Marks and Olson (1981) suggested that researchers adopt a cognitive structure perspective. From this point of view, the information that a consumer gains from using or learning about a product is stored in a permanent memory which maintains that knowledge for future use. A basic characteristic of the consumer's permanent memory is that the information contained therein is highly organized for efficient retrieval. The structure of this organized product knowledge is subject to change as the consumer acquires new information to be integrated with the old. The effect of familiarity, then, is to modify the structure of the consumer's product knowledge.

The nature of the changes that familiarity causes in product knowledge structures warrants further investigation. In exploring this question, Marks and Olson (1981) reported evidence consistent with a "unitization" theory of knowledge development proposed by Hayes-Roth (1977). According to this theory, knowledge structures begin with the development and strengthening of lower-order knowledge units (concepts), each of which is activated in an all-or-none fashion. For example, at the earliest stages of familiarity with a product, a consumer learns isolated pieces of information about it -- perhaps such things as physical features, price, etc. With further product experience, however, these isolated concepts become associated with each other, as the consumer learns which pieces of product information "go together". With sufficient learning, related pieces of information may come to be activated as a unit ("unitized", in Hayes-Roth's terms), so that an appropriate stimulus elicits recall of an entire configuration of product information. Thus, after a period of increasing complexity of product knowledge (in the sense of comprising multiple separate concepts), further increments in familiarity lead to knowledge structures that are actually simpler (i.e., they contain relatively few, higher-order units, each of which comprises numerous related pieces of information). A corollary to the process of unitization, or decreasing complexity, of product knowledge at high levels of familiarity is the notion that well-learned knowledge is maintained at higher levels of abstraction. That is, each higher-order memory unit represents a relatively general notion subsuming the more concrete, detailed concepts of which it is comprised. Thus, knowledge complexity is hypothesized to be an inverted-U shaped function of familiarity. Abstractness of knowledge should increase at higher levels of familiarity; but from very low to moderate levels of familiarity, the addition of new, concrete, concepts should offset increases in the abstractness of concepts already acquired.

By way of example, consider the situation of a consumer learning about automobiles. When he first starts learning about the product, he may form isolated concepts corresponding to such concrete features as "rack and pinion steering", "McPherson struts", and "anti-sway bars". As the consumer's learning progresses, additional, isolated concepts are accumulated in his knowledge structure. At first, these concepts are relatively low in meaning to the consumer, but with experience, he comes to realize that these features are related to each other. They "go together" in that they all contribute to the "handling" characteristics of a car. That is, they form a higher order memory unit ("handling"). Although they may continue to be accessed as separate features, the more familiar consumer will likely consider them as a unitized, abstract characteristic. Such utilization of isolated bits of knowledge, of course, serves the useful function of simplifying the processing of vast amounts of automobile-related information.

Measuring Cognitive Structures

A challenging problem in research on product knowledge structures is how best to measure them. One approach relies on verbal protocols generated by subjects in a free elicitation procedure (Kanwar, Olson, & Sims, 1981; Olson & Muderrisoglu, 1979). Subjects' protocols are scored to index the "dimensionality" of their knowledge (i.e., the number of unique concepts in the protocol) and the abstractness or concreteness of the elicited concepts; this approach thus yields both quantitative and qualitative descriptions of product knowledge.

Marks and Olson (1981), employing this procedure, reported that business students (relatively low familiarity) had more highly dimensional knowledge about office chairs than secretaries (relatively high familiarity), consistent with the unitization theory. Contrary to theoretical predictions, however, the two groups did not differ significantly in the abstractness of their product knowledge (as rated by judgments of concepts in the protocol).

A different approach to assessing the structure of consumers knowledge derives from the psychological literature on cognitive complexity (Bieri et al., 1966; Goldstein & Blackman, 1978; Kelly, 1955). Consumer researchers have studied the complexity of product knowledge with a modification of the role construct repertory test (also known as the repertory grit, or "rep" test; e.g., Mazis, 1973; Tan & Dolich, 1981). This approach basically requires subjects to make judgments about certain stimuli (e.g., various brands) on each of several self-generated dimensions (product characteristics). The judgments are then mathematically analyzed for indices of the degree to which the product characteristics are used in a differentiated manner. The generated product dimensions may also be assessed in terms of relative abstractness. Unfortunately, very little research using this standard measure of cognitive complexity has been reported with respect to product familiarity. Tan and Dolich (1980) found no correlation between product familiarity and cognitive complexity as measured by tied ratings in a rep test. However, as explained in a subsequent section of the present paper, this approach to scoring the rep test has significant limitations; the Tan and Dolich findings, therefore. should be weighed accordingly.

Purpose of the Present Study

The present research was designed to help elucidate the role of knowledge structures as mediators of product familiarity effects. The only previous study directly addressing this issue (Marks & Olson, 1981) was rather limited in scope and sensitivity to the theoretical relationships outlined above. A major shortcoming of that study was its use of only two levels of product familiarity. Since the unitization theory predicts an initial increase in dimensionality (along with little or no increase in abstraction), followed by a drop in dimensionality (and an increase in abstraction) at higher levels of familiarity, at least three levels of product familiarity are required to test these predictions. The present research investigated knowledge structures of consumers varying widely in familiarity. Moreover, to explore the generality of any familiarity/structure effects, research was conducted with two different consumer products. Thus, this study represents a more extensive exploration of the relationship between familiarity and product knowledge.


Subjects and Products

Product knowledge and familiarity data were collected from 128 upper-class students enrolled in a course in Consumer Behavior at the University of Missouri-Columbia, who were given extra credit for their participation. About six weeks prior to the collection of the primary data of the study, these subjects completed questionnaires concerning their familiarity and experiences with 21 different classes of products. Based on responses to this preliminary questionnaire, two products (automobiles and athletic shoes) were selected for use in the present study. The preliminary data indicated that these two products were known well enough that many brand names would be familiar to the subjects, and also that subjects varied considerably in their overall familiarity with the products.


Instruments and Procedure

Each subject provided information about only one of the two products (automobiles- 57 subjects; athletic shoes; 63 subjects) [Eight subjects failed to properly complete the task; their data were excluded from all analyses.]. Product familiarity was assessed with a seven-point scale of the subject's knowledge about how to select the best brand within the product class. A rating of "7" indicated the subject thought she/he knew everything that was pertinent; a rating of "4" indicated as much knowledge as the average person; a rating of "1" meant the subject had no relevant product knowledge. [While this operationalization of product familiarity reflects the subject's self perception, the reader should note that other, more objective, measures might be defined. The author is currently conducting research to assess the correspondence of subjective familiarity with indices of product usage and knowledge.]

The structure of subjects' knowledge about the products was assessed from their responses to a modified rep test (Bieri et al., 1966); a completed example of the test is shown in Figure 1. This test consisted of a 10x10 matrix, whose columns were labeled with the names of 10 brands within the product class; these brands were among the most commonly identified brands in each product class on the earlier product familiarity questionnaire. To the right of each row of the matrix, two blank lines appeared. Within each row, three cells were marked with circles. Subjects were instructed to complete one row at a time. The subject was to think of a characteristic in terms of which any two of the three circled brands differed from the third brand. This characteristic describing the two similar brands was then written into the left-most blank at the end of the row, while the way in which the third brand differed from the other two was described in the right-hand blank. Thus, the subject filled in the blanks with words or phrases describing a dimension along which the brands were perceived to vary. Thus, after defining a product dimension for the row, the subject assessed each of the ten brands in terms of that dimension. A rating scale with values from 1 to 6 was employed for this purpose, with smaller values (1 through 3) representing the left-hand (similarity) end of the dimension, and larger values (4 through 6) representing the right-hand (contrast) end of the dimension. The subject first rated the three circlet brands, and then the remaining brands in the row.



Upon completing each row, the subject proceeded to the next row, which had three different brands circled. The same procedure was followed until all ten rows were completed. Subjects were not allowed to repeat a product characteristic that they had already identified on an earlier row.


Familiarity Ratings

Ratings of familiarity varied from 2 to 7 for automobiles, with a mean rating of 3.75, and from l to 7 for athletic shoes, with a mean of 3.68. Thus, a considerable range of familiarity with each product was present, permitting subsequent assessment of the relationship between familiarity and product knowledge structure variables.

Reliability assessments of the familiarity measure were obtained by correlating the ratings provided during the rep test session with identical measures taken six weeks earlier in the preliminary questionnaire. The test-retest reliability coefficients were 0.69 for automobiles, and 0.65 for athletic shoes (p < .0001), which indicate a moderately reliable subjective index of product familiaritY

Dimensionality Measures

Tie-based Scores. As originally developed, the rep test provides a measure of cognitive complexity based on the number of tied scores within each column of the grid (Bieri et al., 1966). This measure is grounded on the assumption that a subject who rates the brands on one characteristic in an identical fashion to the way he rates the brands on another characteristic is actually using two different names or descriptions for a single underlying product dimension. Thus, the more ties that occur within the ten ratings for each brand, the lower the dimensionality of the subject's knowledge structure underlying those ratings. That is, high tie-scores reflect low cognitive complexity, and vice versa.

Using this measure, the complexity of knowledge reflected in each subject's rep test responses was computed . With a 10x10 grid, this score may range from 40 to 450. The mean complexity score for athletic shoes was 130.7, and for automobiles, 98.7. Thus, automobiles were construed in a more complex fashion than athletic shoes. t(119)=4.89, p < .0001.

The more interesting question, however, is whether the complexity of product knowledge varies systematically with familiarity. Plots of tie scores versus familiarity ratings (from the rep test session) are shown for the two products in Figure 2. These plots reveal no systematic linear relationship between complexity scores and familiarity for athletic shoes (r=.059, p=.65)5 and only a marginal inverse linear relationship between complexity (i.e., low tie scores) and familiarity for automobiles (r=.263, p < .06). Nor is a curvilinear relationship (as predicted by the unitization hypothesis) evident in either plot. Thus, on the basis of the traditional tie-score measure of knowledge complexity, the relationship between product familiarity and the dimensionality of product knowledge appears very slight.



Factor Analysis Measure. Before concluding that no relationship exists between familiarity and knowledge complexity, one may legitimately question the validity of the tie-score approach to assessing rep test responses. A significant problem with this approach is that much information potentially available in non-tied scores is ignored by such a measure. The degree of covariation between ratings on two product characteristics will be influenced to some extent by the number of tied scores (if the relationship is positive). But two highly correlated characteristics need not have any pairs of scores that are exactly tied, in which case the traditional complexity index would consider the characteristics totally differentiated. Moreover, the tie-score approach would be biased against recognizing as similar two product characteristics which were negatively correlated (e.g., characteristics G and H in Figure 1). Few ties would be counted in comparisons of such characteristics yet they clearly reflect different versions of a common underlying product dimension. Thus, a better measure is needed for scoring the rep test responses.

Factor analysis is ideally suited to this problem. The rep test responses of each subject were subjected to principal components factor analysis, with the various brands representing ten different sets of responses to the ten identified product characteristics. For each subject, the number of "true" factors underlying his or her 10 product characteristics was estimated by counting all factors with eigenvalues greater than 1.0. The number of factors thus determined represents the dependent variable indicating the dimensionality of the subject's product ratings.

With this index of dimensionality, automobiles were again perceived in a more complex fashion (mean factors = 2.70) than athletic shoes (2.21 factors), t(118) = 3.96, p<.0001. Figure 3 provides evidence of a positive relationship between familiarity and dimensionality for both products (r=.302, p<.029 for athletic shoes; r=.368, p<.01, for automobiles). No curvilinear relationship was evident for either product. Hence, this more sensitive measure indicates that more familiar subjects had more dimensional knowledge structures.


Each product characteristic was classified as either concrete or abstract by two independent judges (Geistfeld, Sproles. & Badenhop, 1977). Characteristics that described specific, physical features of the product were considered concrete, as were non-product features (e.g., such marketing variables as price) which could be measured easily and precisely. More general product characteristics, especially those that were functions of more specific features, were rated as abstract. A rating of 0 was given to concrete characteristics, and 1 to abstract characteristics. The two judges agreed at first rating in about three-fourths o the cases; discrepancies were resolved through discussion of specific characteristics.



Characteristics describing automobiles were judged slightly less abstract (mean 0.70) than those describing athletic shoes (0.76), t(118)-2.10, p<.05). As indicated in Figure 4, more familiar subjects tended to use more concrete automobile descriptions than did less familiar subjects, r=.302, p<.03; this relationship was not found with athletic shoes, r=-.087, p=.50).



The plot for athletic shoes, though generally flat, suggests the possibility of an inverted U-shaped relationship between familiarity and abstractness (which would be opposite the hypothesis). To assess this possibility, an analysis was made of the quadratic component of the variance among familiarity levels, following an initial analysis of variance on abstractness scores. This trend analysis revealed the quadratic component to be nonsignificant, F(1,55)=0.290; familiarity did not influence the abstractness of knowledge about athletic shoes.


The relationships predicted from the unitization theory of knowledge development were not observed in the present data. The theory predicted an inverted U-shaped relationship between the dimensionality of product knowledge and the consumer's familiarity with the product; instead, dimensionaLitY was found to increase monotonically with familiarity. Moreover, the predicted U-shaped relationship between abstractness and familiarity did not obtain; instead, subjects more familiar with automobiles tended to use more concrete characteristics to describe them, while familiarity with athletic shoes did not influence the degree of abstractness or concreteness of subjects' responses.

These findings are rather surprising, in light of theoretical expectations and Marks and Olson's (1981) findings reported above. Several potential explanations for the present observations can be considered. One possibility is that the subjects actually ranged only from low to moderate in familiarity, without any truly "high familiarity" subjects in the sample. If this were the case, then the data would be consistent with the unitization theory, as they would represent points on the first half of the U-shaped functions. This explanation seems rather unlikely, however, given the descriptions of the familiarity scale endpoints, and the fact that most subjects rated their familiarity at moderate levels, with relatively fewer ratings of 1, 6 or 7. It is also possible that the single measure of product familiarity, which was not perfectly reliable (and thus not perfectly valid), may not have fully assessed familiarity. Further research is needed into this possibility.

It is conceivable that the measures of dimensionality and abstractness employed did not provide valid indices of the underlying constructs. However, the logic for preferring a factor analysis measure of rep test dimensionality over one based on tied scores, explained above, seems particularly compelling, given the quantity of data that the latter measure ignores. Tan and Dolich's (1980) results are consistent with those of the present study when tie scores are used, but they differ from the results counting number of factors. Further research is needed to clarify the best means for measuring knowledge dimensionality with the rep test. Other approaches to assessing knowledge dimensionality should be explored, too, in light of the admittedly crude and limited nature of rep test scores. Moreover, the present measure of abstractness is also relatively crude, since there were several responses that were difficult to classify readily as abstract or concrete. Further work is needed to refine measurement of the abstractness of product knowledge.

Aside from concluding that the unitization theory is incorrect, at least one additional explanation for the present results may be offered. According to Hayes-Roth (1977), the more familiar person, who has learned to chunk together related pieces of knowledge into higher-order units, does not lose the ability to decompose those relatively abstract chunks into their more concrete constituent parts as needed. If the present task required subjects to produce relatively concrete, detailed information about the products, then the most familiar subjects would not necessarily reveal their abilities to generalize from simpler knowledge structures. It is possible that this task did not encourage the use of high-level abstractions that the more familiar subjects were capable of, in which case the unitization theory would not be disconfirmed. As this possibility can not be ruled out, the present data may fall within the explanatory power of this theory.

Future Research

This study is clearly exploratory in nature, and additional research is needed to understand the role of cognitive structure as a mediator of product familiarity effects. As suggested in the preceding discussion, future research should attempt to explore the best means for measuring product knowledge structures. Replication of this basic design with multiple measures, using both the same and additional products, will help refine measurement techniques, as well as establish the generality of any familiarity - structure relationship.

Once such research has established the scope of the familiarity - structure relationship, additional research will be useful to assess the impact of product knowledge structures on information processing. attitudes, intentions, and actual behavior. The most direct tests of these relationships would involve actual manipulation of product familiarity (through controlled exposure to product use and information), with longitudinal assessment of various indices of processing, attitudes, and behaviors.

To fully understand the role of product familiarity, an extensive program of research along these lines will be required. It is hoped that the present study will serve to stimulate further work on the challenging question of what customers know about products.


Bettman, James R. (1979), An Information Processing Theory of Consumer Choice, Prading, MA: Addison-Wesley Publishing Co.

Bettman, James R. and Park, C. Whan (1980), "Effects of Prior Knowledge and Phase of Choice Process on Consumer Decision Processes: A Protocol Analysis," Journal of Consumer Research, 3, 233-40.

Bieri, J., Atkins, A. L., Briar, S., Leaman, R. L., Miller, H., and Tripoldi, T. (1966), Clinical and Social Judgment: The Discrimination of Behavioral Information, New York: John Wiley & Sons, Inc.

Geistfeld, Loren V., Sproles, George B., Badenhop, Suzanne B. (1977), "The Concept and Measurement of a Hierarchy of Product Characteristics," in Advances in Consumer Research Vol. 4, (ed.) William D. Perreault Jr., AtLanta: Association for Consumer Research, 302-307.

Goldstein, Kenneth M., and Blackman, Sheldon (1978), Cognitive Style: Five Approaches and Relevant Research, new York: John Wiley & Sons, Inc.

Hayes-Roth, Barbara (1977), "Evolution of Cognitive Structures and Processes," Psychological Review, 84, 260-78.

Howard, John A., and Sheth, Jagdish N. (1969), The Theory of Buyer Behavior, New York: John Wiley & Sons, Inc.

Johnson, Eric J., and Russo, Edward J. (1981), "Product Familiarity and Learning New Information" in Advances in Consumer Research. Vol. 8 (ed.) Kent B. Monroe, Ann Arbor, MI: Association for Consumer Research 151-155.

Kanwar, Rajesh, Olson, Jerry C., and Sims, Laura S. (1981), "Toward Conceptualizing and Measuring Cognitive Structures" in Advances in Consumer Research. Vol. 8, (ed.) Kent B. Monroe, Ann Arbor, MI: Association for Consumer Research. 122-127.

Kelly, George A. (1955), The Psychology of Personal Constructs (2 volumes), New York: Norton.

Marks, Larry J., and Olson, Jerry C. (1981), "Toward a Cognitive Structure Conceptualization of Product Familiarity," in Advances in Consumer Research Vol. 8 (ed.) Kent B. Monroe, Ann Arbor, MI: Association for Consumer Research, 145-150.

Mazis, Michael B. (1973), "Cognitive Tuning and Receptivity to Novel Information," Journal of Social Psychology, 9, 307-19.

Olson, Jerry C., and Muderrisoglu, Aydin (1979), "The Stability of Responses Obtained by Free Elicitation: Implications for Measuring Attribute Salience and Memory Structure," in Advances in Consumer Research Vol. 6, (ed.) William L. Wilkie, Ann Arbor, MI: Association for Consumer Research, 269-275.

Raju, P. S., and Reilly, Michael D. (1980), "Product Familiarity and Information Processing Strategies: An Exploratory Investigation," Journal of Business Research, 187-212.

Tan, Chin Tiong, and Dolich, Ira J. (1981), "The Moderating Effects of Cognitive Complexity and Prior Product Familiarity on the Predictive Ability of Selected Multi-attribute Choice Models for three Consumer Products," in Advances in Consumer Research, Vol. 8 (ed.) Kent B. Monroe, Ann Arbor, MI: Association for Consumer Research, 140-144.