Product Familiarity and Learning New Information

Eric J. Johnson, Carnegie-Mellon University
J. Edward Russo, University of Chicago
ABSTRACT - What is the relationship between product familiarity and the ability to learn new product information? An experiment shows that product familiarity can lead to increased learning during subsequent purchase decisions. However, this ability strongly interacts with the specific decision task: the monotonic relationship between familiarity and learning holds for judgment but not for choice, The results also show that judgment and choice strategies leave different information in memory. We believe that phased choice strategies account for this difference.
[ to cite ]:
Eric J. Johnson and J. Edward Russo (1981) ,"Product Familiarity and Learning New Information", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 151-155.

Advances in Consumer Research Volume 8, 1981      Pages 151-155

PRODUCT FAMILIARITY AND LEARNING NEW INFORMATION

Eric J. Johnson, Carnegie-Mellon University

J. Edward Russo, University of Chicago

[The authors thank Jim Bettman for hie comments on an earlier draft. The first author is now at the Department of Psychology, Stanford University, Stanford, CA 94305. This work was supported by National Science Foundation Grant #DAR76-81506 to the second author.]

ABSTRACT -

What is the relationship between product familiarity and the ability to learn new product information? An experiment shows that product familiarity can lead to increased learning during subsequent purchase decisions. However, this ability strongly interacts with the specific decision task: the monotonic relationship between familiarity and learning holds for judgment but not for choice, The results also show that judgment and choice strategies leave different information in memory. We believe that phased choice strategies account for this difference.

INTRODUCTION

Product familiarity is an important factor in explaining consumer behavior. Its rule is recognized in both traditional theories (Howard and Sheth 1969, Hansen 1972, Howard 1977) and newer information processing approaches (Bettman 1979, Olson 1978). Familiarity has also received recent attention in empirical research (Olson 1979, Bettman and Park 1980a). With the exception of Bettman's and Park's paper, however, little research explores the effect of familiarity on consumer information processing. The present paper reports an empirical investigation of the effect of familiarity on one important consumer behavior, learning new product information during purchase decisions.

Consumer Knowledge

In this paper, product familiarity refers to prior knowledge of the brands within a product category. Obviously, all consumers start as novices at some point before their first purchase in a product class. As the consumer gains experience, product familiarity grows, and this knowledge affects the acquisition of new product knowledge. To understand this relation between familiarity and learning, a theory of product knowledge is essential. We use a taxonomy of knowledge developed by Russo and Johnson (1980).

The two central components of this taxonomy are shown in Figure 1. The level of inference refers to the relationship between external information and knowledge in memory. This resembles a hierarchy. New knowledge starts with externally available information, the brand-attribute values at Level 5 in Figure 1. The purchase decision process generates additional intermediate knowledge in the natural course of comparisons among brands. Finally, there is knowledge of the best brand, shown at Level 1 at the top of the hierarchy.

Knowledge can also be classified by its inferential basis, the dimension across the top of Figure 1. Consumers can make decisions using either brand-based or attribute-based strategies (Bettman 1979). The resultant product knowledge will then be brand-based or attribute-based. Examples of the interaction between inferential level and inferential basis are found in the cells of Figure 1.

FIGURE 1

A TAXONOMY OF PRODUCT KNOWLEDGE

With this scheme for characterizing product knowledge, both old and newly acquired, we can ask how product familiarity affects learning additional product information. There are two plausible hypotheses.

The first, which might be called the "enrichment hypothesis", derives from cognitive psychology. It has been repeatedly demonstrated that greater prior knowledge facilitates learning. One classic example is the research on chase performed by Chase and Simon (1973). They showed a number of board positions to both chess masters and novice players for five seconds apiece. Knowledgeable subjects remembered much more than novices. However, when shown random patterns of chess pieces, the masters' recall was no better than the novices'. This, Chase and Simon argued, was because the recall of random positions did not use the superior knowledge of the chess masters. Thus, the "enrichment hypothesis" argues that prior knowledge provides experienced consumers with better encoding and recall skills. Therefore, when confronted with new information, greater experience facilitates greater learning, a "rich get richer" hypothesis. This pattern would generate data similar to the exponential curve in Figure 2.

FIGURE 2

ALTERNATIVE HYPOTHETICAL RELATIONS BETWEEN PRODUCT FAMILIARITY AND LEARNING

In contrast, Bettman and Park (1980s) claim that prior knowledge has the "inverted U" effect shown in Figure 2. They are concerned with external information search, but it seems natural to extend their hypothesis to the knowledge that remains in memory after search. In their view, very inexperienced consumers have difficulty understanding new information, and therefore search less. Consumers with a moderate level of knowledge search widely; they can both understand the new information and also benefit from its retention in memory. Finally, very experienced consumers search less. Although they can understand new product information, they have little need of it. This hypothesis was confirmed by Bettman and Park's search data, and the same "inverted U" effect might describe newly acquired knowledge in memory.

Experimental Rationale

To test these two hypotheses, we performed an experiment that used a brand-by-attribute display of information. The matrix contained new information about sub-compact cars that had just become available at the beginning of the model year. Consumer subjects, differing in their familiarity with automobiles, were instructed to evaluate the cars using only this information. A surprise recall of this information provided us with evidence of the role of prior knowledge in learning (remembering) new product knowledge.

Choice Versus Judgment

Besides the effect of different levels of familiarity, we added a second manipulation. We suspected that the strategy used to evaluate products would influence the product knowledge that was acquired and, therefore, remembered. First, consider a conventional choice task, i.e. choosing the best of several alternatives. Bettman and Park (1980b), Wright and Barbour (1977), and Johnson and Russo (1978) all find clear support for the use of phased decision rules in consumer choice. If subjects used phased strategies, some alternatives should be eliminated quickly, on the basis of only one or two pieces of information. Less knowledge of these eliminated brands should be generated and retained in memory. Some support for these claims was found by Johnson and Russo (1978).

In contrast with a choice task, subjects can also judge the overall quality of each alternative. This second task should force subjects to analyze, and remember, equal amounts of information for every alternative. In this experiment, half the subjects performed an ordinary preferential choice task, while the other half rated each alternative on a scale of attractiveness.

METHOD

Subjects

The 54 subjects participating in this experiment were students in the evening master's program of the Graduate School of Business of the University of Chicago. They completed the task as part of a classroom demonstration during the first meeting of a consumer behavior course. The product category, new sub-compact cars, was selected so this subject population would represent at least one segment of typical purchasers. Their demonstrated interest in the task partially confirmed this.

Stimulus

The brand-by-attribute matrix shown to the subjects was an edited copy of an advertisement placed by General Motors' Oldsmobile Division. Originally the advertisement compared one model of Oldsmobile to a number of imported sub-compact automobiles. By removing the Oldsmobile from the choice set, we created a matrix that had no dominant alternative, yet preserved the interrelationships among the attributes that naturally occurred in the marketplace. Additionally, we deleted anything beyond one model per manufacturer to avoid possible confusion during recall. These actions left a matrix of eight alternatives, all smaller imported cars. They were rated on ten attributes: model, transmission, engine (number of cylinders), EPA estimated miles per gallon, fuel tank capacity, estimated driving range, passenger capacity, interior volume, trunk volume, and the manufacturer's suggested retail price.

Procedure

Subjects were run in a single group. At the beginning of the session, they were given a booklet containing all experimental materials. The first page presented the appropriate task instructions and requested that they not proceed until asked to do so by the experimenter.

Judgment Task.  In the judgment condition, the instructions informed the subjects that they would be given a brand-by-attribute matrix of information relevant to automobiles. They were asked to judge each of the automobiles on a seven-point scale, based only on the information provided, and not prior knowledge.

Choice Task.  Subjects in the choice task were similarly informed of the brand-attribute matrix, but were asked to choose the most preferred of the eight automobiles rather than make individual judgments.

When all subjects finished their respective tasks, they were asked to turn the page to a "Demographic Questionnaire." This page had two purposes. First, it was used to gather demographic information, including a self-rating of product familiarity. Subjects were asked to "rate your previous knowledge of automobiles, compared to the rest of the population", on a five-point scale. Second, the recall of demographic facts prevented the retention of product information in short-term memory.

Subjects were then instructed to turn the page and read the instructions for the unexpected recall task. They were told to "try to recall as much as you possibly can." This included not only the information that they were given but also any "observations and judgments about the alternatives and the attributes." These instructions were designed to encourage complete recall, including information at the higher levels of inference. Subjects were then given an unlimited time to record their recall on blank sheets of paper. After the recall was completed, subjects were asked to report the car selected as best, any past experience with each of the eight automobiles, and the strategy they used to perform the choice or judgment task.

Analysis

The written protocols were divided into a series of "complete thoughts" (Newell and Simon 1972) and then coded according to the scheme shown in Figure 1 and described in Russo and Johnson (1980). This scheme is similar to the one used by Bettman and Park (1980a).

Familiarity with the product category wee measured by subjects' self-reports on the scale described above. The distribution of these self-ratings was concentrated at the middle value. Self-ratings at or below 2 were classified in the low familiarity group, between 3 and 4 as the moderate familiarity group, and above 4 as the high familiarity group. The number of subjects in each familiarity condition was 12, 27 and 16, respectively.

The factor of task, judgment versus choice, was crossed with the three levels of familiarity to create a 2 X 3 factorial design (with unequal cell sizes). This two-factor design served as the basis for all analyses reported.

RESULTS

We have claimed that the information presented to subjects represented new knowledge, even to the subjects with greater product familiarity. This was reinforced by task instructions asking subjects to limit their attention to the information presented. However, to check the validity of this claim, we examined the recall protocols for accuracy and intrusions. Inaccurate information could come from two sources: either outside knowledge or a faulty remembrance of the presented information.

A rater blind to our hypo theses, coded each brand-attribute value on a five-point scale. Inaccurate recall was defined as those statements rated 4 (somewhat inaccurate) and above, and accounted for less than 10% of the information recalled. When the numbers of intrusions and inaccuracies are included as covariates in the analyses of variance, our results remain unchanged. Thus, the results that follow are probably not due to the recall of previously remembered information by the highly familiar consumers.

Familiarity

What is the relationship between existing product knowledge and the ability to learn new information? We have considered two possible relationships: the "enrichment hypothesis" and the "inverted U" hypothesis, shown in Figure 2. A test of these two plausible relationships between familiarity and learning can be based on the total amount of knowledge recalled as a function of familiarity. The mean number of statements recalled increases with the level of familiarity. For the low, medium, and high levels, the mean frequencies were 12.0, 16.6 and 21.1, respectively. However, an analysis of variance shows that this effect is only marginally significant (F (2.49) = 2.825, p < .07 ). Nonetheless, the pattern of results mildly supports the "enrichment" hypothesis: increased familiarity does lead to the recall of more new information.

A closer look at chess date reveals that the effect of familiarity interacts with the evaluation task. The mean number of statements recalled is plotted separately for choice and judgment tasks in Figure 3. For judgment, the enrichment effect is large and consistent, with highly familiar consumers recalling two and one-half times as much information as the low familiarity consumers (28.6 versus 11.6 statements per protocol). The consumers who were asked to choose one brand, however, show an "inverted U" relationship. The subjects moderately familiar with automobiles exhibit the greatest recall. The ANOVA shows that this familiarity by task interaction is reliable, F (2,49) = 3.20, p < .05. A priori contrasts indicate that the judgment condition contains a significant linear relationship with familiarity, (F (1,49) = 8.32, p < .001). In contrast, the group making choices exhibits the inverted U or quadratic relationship (F (1,49) = 4.11, p < .06).

FIGURE 3

LEARNING AS A FUNCTION OF FAMILIARITY AND TASK

Thus, while the enrichment hypothesis holds for consumers making judgments, it does not describe the memory of consumers making choices. As Bettman and Park predict, moderately familiar consumers search and remember more information.

Choice versus Judgment

How do these decision strategies affect newly acquired product knowledge? An important manipulation in this study concerns the task instructions. The consumers either chose one alternative or judged the overall quality of each alternative. We suspect that these two tasks require different patterns of information processing.

If, as we believe, consumers use phased choice rules, they first examine the alternatives in an attribute-based manner, using this information to eliminate alternatives. Eventually, the basis of the decision process shifts, and the remaining alternatives are examined more intensively. This later processing is by brand (Bettman and Park 1980b, Bettman 1979, Johnson and Russo 1978). As a consequence of this phased strategy, more information is examined and recalled for the chosen alternatives (Payne 1976, Johnson and Russo 1978). This differential recall is also affected by the level of inference (Russo and Johnson 1980). If a brand is eliminated early in the choice process, few higher level inferences should be made. Thus, the higher the inferential level, the greater the recall advantage of the chosen brand (Johnson and Russo 1978).

In contrast, the judgment task does not permit the use of phased strategies. Because subjects rated all eight automobiles, they were required to examine information for every alternative. Thus, in the judgment task we expect equal recall for all brands, preferred or not.

To test these predictions, we tallied the number of statements in each protocol that referred to the automobile selected by the subject. In the choice condition, this was the most preferred alternative. In judgment, subjects reported which car they judged as the best. The proportion of recalled statements referring to the preferred automobile is .35 for the choice task and .18 for judgment, This difference was highly significant (F (1,49) = 13.84, p < .001). In addition, both values, including the 181 observed in the judgment task, are significantly greater than the 12.5% expected by chance (p < .001). These data provide clear evidence of the use of phased strategies in the choice task, and identify much smaller effects for judgment instructions,

To further examine memory for chosen products, we broke down the proportions of statements referring to chosen products by task and level of familiarity. The means are shown in Table 1.

TABLE 1

PROPORTION OF STATEMENTS REFERRING TO THE PREFERRED ALTERNATIVE

An ANOVA confirms that there is a task-by-level interaction. For judgment there is no effect of familiarity, but the choice task shows a large effect (F (2,49) = 5.82, p < .01). Half the recall of high familiarity subjects in the choice condition referred to the car they chose. This compares to less than a third for the low and moderate familiarity groups. A priori comparisons confirm the reliability of this pattern (F (1,49) = 11.20, p < .005).

Overall, the data indicate that the consumers who made choices used phased rules which eliminate alternatives, while consumers making judgments did not. Familiarity leads to an increasingly selective memory for information about the chosen product caused, we suspect, by the use of highly selective search. The most experienced consumers do not look at all the new information, only that which they consider useful.

Types of Product Knowledge

Let us return to the taxonomy of product knowledge discussed earlier and summarized in Figure 1. Do the above results hold over all levels of inference? To answer this question the recalled statements were partitioned into five inferential levels. The same two-factor (familiarity and task) ANOVA was performed on the number of statements at each level. The only differences attributable to the main effect of familiarity occurred at Level 5, brand-attribute values, (F (2,49) = 3.46, p < .05). This lowest level of product knowledge consists simply of the values in the brand-by-attribute matrix. This is also the only inferential level that shows a significant familiarity-by-task interaction (F (2,49) = 3.65, p < .05). Thus, differences in recall can be attributed entirely to knowledge at the lowest level of processing, the acquisition of the presented brand-attribute values. No other inferential level shows these effects.

HOW DOES FAMILIARITY WORK?

Can these results help explain the differential effects of familiarity on choice and judgment? Recall that Bettman and Park explain their "inverted U" results by claiming that experienced consumers use internal search as a substitute for external search. Because of our earlier analysis of intrusions into recall, we suspect that experienced subjects did not use stored knowledge of the brand-attribute values. Since little of the information recalled deviates from the information presented (less than 10%), internal search of existing knowledge was probably not a factor. Thus, we need another explanation for these results.

An alternate explanation of increased familiarity relies on other aspects of experienced consumers prior knowledge. There is little doubt that experienced consumers have a great deal of information about products stored in memory (Russo and Johnson 1980). This knowledge contains more than just facts about the relations among competing brands. In addition, experienced consumers learn which attributes are most important, and should subsequently focus on these attributes in making a choice. This is a type of higher-level knowledge that goes beyond particular brands. It implies that, with experience, consumers become more selective in their search for information, and use more narrowly focused phased decision rules.

We did not collect decision protocols in this study. This means we cannot examine the decision processes directly. But we did find that differences in memory for chosen and rejected products will be greater for more experienced consumers resulting from their more selective search. This was demonstrated by the significant familiarity-by-task interaction displayed in Table 1. Also, as we have already seen, the effect of familiarity is confined to the earliest, low-level inferences. These inferences come from the initial examination of the matrix which, according to this model, experienced consumers delete from their decision processes.

Is one explanation of familiarity effects superior to the other? We cannot provide a definitive answer with these data. Our "knowledge of the attributes" explanation is very similar to Bettman and Park's, and differs only for information previously unseen by the consumer. But in the real world, both mechanisms probably operate, and increase the effects of familiarity. A knowledge of important attributes, along with a knowledge of brand-specific facts, can limit the search of experienced consumers. So for many real world situations, differences in the explanations may not matter. The unique contribution of the current explanation is limited to new information, whether from new products, or new product classes.

DISCUSSION

What does all this mean for the study of consumer behavior? It is important to see where the present results fit in the growing literature on consumer information processing. A surprisingly consonant view of consumer information processing is emerging from a number of investigators working within different experimental paradigms, and with different products and subject populations. First, recent work calls attention to the predominance of phased strategies in choice (Bettman and Park 1980b, Wright and Barbour 1977, Payne 1976). More work is needed in display formats other than the brand-by-attribute matrix. However, we now know that phased strategies affect not only information acquisition, but also subsequent memory. Second, familiarity with product classes has an effect on decision and search behavior. At first, it leads to an increase in external search and subsequent recall. But with high levels of familiarity, both information search and memory for new information decline (Bettman and Park 1980b).

This paper attempts to make two additional contributions to this area. First, we try to extend the concepts advanced by Bettman and Park to memory for new product information. Second, we offer an alternative explanation for why experienced consumers search less information. We argue that this is not caused by the presence of relevant information stored in memory, but by their higher-level knowledge of the product class and its important attributes. They search less information and use more selective, phased decision rules that delete the preliminary exploration of external information.

These findings have implications for public and private providers of product information. They suggest that information use will be segmented by the familiarity level of the consumer. The most important implication is that providing a brand-by-attribute matrix for a low familiarity segment may do little good. Simply posting information is probably ineffectual, especially for technically sophisticated products and attributes. To help these consumers make better decisions, information about the attributes, their importance, and their relationship to quality must be provided.

REFERENCES

Bettman, James R. (1979), An Information Processing Theory of Consumer Choice, Reading, Mass.: Addison-Wesley.

Bettman, James B. and Park, C. Whan (1980a), "Implications of a Constructive View of Choice for Analysis of Protocol Data: A Coding Scheme for Elements of Choice Processes," in Advances in Consumer Research, Vol. VII, Jerry C, Slash, ed., Ann Arbor: Association for Consumer Research, 148-153.

Bettman, James R. and Park, C. Whan (1980b), "Effects of Prior Knowledge, Exposure and Phase of the Choice Process on Consumer Decision Processes: A Protocol Analysis," Journal of Consumer Research, in press.

Chase, William G. and Simon, Herbert A. (1973), "Perception in Chess," Cognitive Psychology, 4, 55-81.

Hansen, Flemming (1972), Consumer Choice Behavior, New York: Free Press.

Howard, John A. (1977), Consumer Behavior: Application of Theory, New York: McGraw Hill.

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

Johnson, Eric J. and Russo, J, Edward (1978), "What Is Remembered After a Purchase Decision?," working paper, Graduate School of Business, University of Chicago.

Newell, Allan and Simon, Herbert A. (1972), Human Problem Solving, Englewood Cliffs, N, J.: Prentice-Hall.

Olson, Jerry C. (1778), "Theories of Information Encoding and Storage: Implications for Consumer Research," in The Effect of Information on Consumer and Market Behavior, Andrew A. Mitchell, ed., Chicago: American Marketing Association, 49-60.

Russo, J. Edward and Johnson, Eric J. (1980), "What Do Consumers Know About Familiar Products?," in Advances in Consumer Research, Vol. VII, Jerry C. Olson, ed., Ann Arbor: Association for Consumer Research, 417-423.

Wright, Peter L. sad Barbour, Frederic (1977), "Phased Decision Strategies: Sequels to an Initial Screening", in North Holland/TIMS Studies in the Management Sciences, Volume 6: Multiple Criteria Decision Making, Martin K. Starr and Milan Zeleny, eds., Amsterdam: North Holland, 91-109.

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