Representation of Information in Memory: Some General Issues

John R. Rossiter, Columbia University
[ to cite ]:
John R. Rossiter (1980) ,"Representation of Information in Memory: Some General Issues", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 429-430.

Advances in Consumer Research Volume 7, 1980     Pages 429-430


John R. Rossiter, Columbia University

My purpose in this discussion paper is to make some general observations about the representation of information in memory with reference to the four papers delivered under this heading by Hirschman (1980), Gutman (1980), Russo and Johnson (1980) and Chestnut and Jacoby (1980). I intend to evaluate the papers but only in relation to the central question of where this line of consumer research is leading us.


Too often we overlook the issue of why it makes any difference whether information is represented in one way rather than another in memory. Take the familiar "multiattribute" brand by attribute representation as an example. Does it matter whether or not consumers have "matrices in their heads" that correspond to the matrix of brands by attributes identified by the multiattribute model? Surely the matrix representation is an external representation for the convenience of the researcher rather than an internal representation of information storage. Moreover, the model may work quite well in predicting such events as brand choice regardless of how information is "in fact" internally represented.

It is incumbent on researchers to specify why information representation would make a difference in terms of the behavior they are trying to explain or predict. Otherwise, the study of information representation becomes an esoteric exercise.


The likelihood is that consumers learn various connections between brands and attributes, between brands and ideal profiles or prototypes of brands, between attributes and attributes, between attributes and situations, and so forth. It is quite possible that, given appropriate instructions and motivation, consumers could reproduce these responses externally in matrix form, list form, prototype comparisons, or whatever. We see this all the time in marketing research. All we really know is that consumers make responses of one type or another in various circumstances. It is a big leap to infer that these responses mirror the way information is represented in memory because it is simply too easy to get consumers to generate different responses which would lead to entirely different inferences. Let me briefly illustrate this point with reference to the three experimental papers in the session.

As one of the tasks in his experiment, Gutman (1980) had subjects sort 18 written names of breakfast cereals into "as many categories as they felt needed." He then used number of categories observed to infer that those subjects who employed more categories (i.e., who had narrower equivalence ranges) had "more highly developed cognitive structures." The apparent circularity of this inference was reduced somewhat by showing that equivalence ranges generalized to other experimental sorting tasks.

But the tasks themselves were unusual. When, for instance, do consumers ever have to sort cereal names on slips of paper; what if actual cereal boxes had been used instead? How often do consumers evaluate 18 or so cereals at one time? What about the rather vague criterion of "as many categories as they felt they needed"? Does this task really tell us anything about how consumers categorize cereals in the real world? This task, and the others employed in the study, seem to reveal very little about consumer behavior--or about memory representation.

Russo and Johnson (1980), in an otherwise very well conducted experiment, also used a somewhat questionable task to infer memory representation. They asked subjects to "imagine that they had an English speaking friend from a foreign country who needed to know about the products available in the United States. The subjects' task was to tell their friend everything she or he would need to know to make purchases in a specified product class. The friend...knew nothing about American brands (but did have a basic knowledge of common products like automobiles, headache remedies, etc.)." (emphasis added)

In the resulting descriptions, the investigators found a predominance of brand-based information. This is not too surprising given the instructional emphasis on the hypothetical listener's lack of brand knowledge but assumed attribute knowledge. Different descriptions may have emerged, for example, had subjects been asked to describe automobiles, headache remedies, etc. to an English-speaking Martian visitor with no knowledge of either brands or attributes. I happen to agree that consumers' product knowledge is probably organized by brand rather than by attribute except, perhaps, in product categories where a particular attribute has been dominantly emphasized, such as the tar and nicotine content of cigarettes. However, the potential "brand bias" in this laboratory task renders the demonstration less than conclusive.

Chestnut and Jacoby (1980) asked college freshmen and sophomores to make a simulated choice between permanent and term life insurance. The investigators were surprised by the students' lack of knowledge about these important products. But why should people be expected to remember life insurance concepts? Is there any functional relevance, unlike for frequently purchased products, in being able to produce this knowledge on demand? People may learn enough at the appropriate time to make a choice and then forget the details of the choice. Or, as the investigators note, people may assume that others are competent to engage in this learning and choice for them, e.g., a trusted insurance agent.

In their experiment, the investigators found that subjects who showed better comprehension of two 1,400-word product descriptions more often made the "correct" choice in a simulated insurance purchase situation. But this seems to say less about memory than about motivation and short-term problem solving in a laboratory setting.

In short, we should more carefully consider the external validity of measures of internal representation. Isolated tasks of questionable realism are likely to lead to questionable accounts of memory representation.


Models of consumer representation of information in memory must allow for the frequent presence of visual information which can influence choice behavior (Rossiter and Percy 1979). It can easily be shown that visual versus linguistic information representation is a distinction that does have important implications, unlike some of the other representational characteristics proposed in previous studies. Brand awareness, for example, is a complex response which depends in some choice situations on visual recognition, as when a prearticulate child recognizes a cereal box in a supermarket, and in others on verbal recall, as when an adult thinks of the names of several insurance companies to call regarding policy quotations.

None of the session papers takes explicit account of visual information representation. Hirschman (1980) focuses heavily on concept representation in semantic memory whereas it seems likely that episodic images of past or anticipated consumption or usage situations also form an important aspect of product concepts which can influence brand choice. Gutman (1980) used cereal names to simulate cereal brand stimuli whereas in actual choice situations these stimuli are accompanied by a rich variety of visual cues which may draw on past imagery and engender evaluations quite different from those produced by brand names alone. Russo and Johnson (1980) had subjects give verbal descriptions of products such as automobile brands which are difficult to characterize adequately without visual information. Even in the Chestnut and Jacoby (1980) study, where the product was life insurance, it could be argued that brand images derived through visual information, e.g., Prudential's "rock" or Allstate's "good hands," exert considerable influence on choice.

Research on information representation will nearly always be incomplete if it focuses exclusively on linguistic information and ignores visual information.


This brief critique offers three observations about theory and research on the representation of information in memory. Firstly, it is incumbent on theorists in this area to show how and why information representation makes a difference in terms of consumer behavior outcomes. Many theoretical models, for example multi-attribute models, seem to be quite capable of accounting for consumer choices without requiring any assumptions about the way in which information is "actually" represented in the consumer's memory. Secondly, empirical efforts to map the structure and content of information representation must take care to employ ecologically valid tasks. It is simply too easy to generate task-dependent responses which confound short-term problem-solving activities with information retrieval and utilization as these processes occur in the real world. Lastly, there still remains a bias in our memory models and research procedures toward linguistic information and verbally oriented measures with a consequent neglect of visual information which so obviously and so often exerts an important influence on consumer behavior.


Chestnut, R. W. and Jacoby, J. (1980), "Product Comprehension: The Case of Permanent Vs. Term Life Insurance," Advances in Consumer Research, 7,.

Gutman, J. (1980), "Equivalence Range in Categorization," Advances in Consumer Research, 7.

Hirschman, E. C. (1980), "Concept Formation, Product Conceptualization and Cognitive Development," Advances in Consumer Research, 7.

Rossiter, J. R. and Percy L. (1979), "Visual Communication in Advertising," Research Working Paper No. 231A, Columbia University, Graduate School of Business.

Russo, J. E. and Johnson, E. C. (1980), "What Do Consumers Know About Familiar Products?" Advances in Consumer Research, 7.