The Representation of Consumer Information in Memory

Thomas K. Srull, University of Illinois
ABSTRACT - After briefly discussing several individual experiments, the current state of consumer research related to the representation of information in memory is assessed. It is suggested that the field is currently characterized by a series of isolated experiments and is largely phenomenon-driven rather than theory-driven. It is argued that future progress will heavily depend on the development of systematic programs of research that are logically directed toward narrowing down the range of plausible alternative explanations that exist for any single array of data.
[ to cite ]:
Thomas K. Srull (1982) ,"The Representation of Consumer Information in Memory", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 499-501.

Advances in Consumer Research Volume 9, 1982      Pages 499-501

THE REPRESENTATION OF CONSUMER INFORMATION IN MEMORY

Thomas K. Srull, University of Illinois

ABSTRACT -

After briefly discussing several individual experiments, the current state of consumer research related to the representation of information in memory is assessed. It is suggested that the field is currently characterized by a series of isolated experiments and is largely phenomenon-driven rather than theory-driven. It is argued that future progress will heavily depend on the development of systematic programs of research that are logically directed toward narrowing down the range of plausible alternative explanations that exist for any single array of data.

DISCUSSION

Bozinoff, Conover, and Rethans and Hastak have each provided provocative papers. I will begin by discussing some of the factors that need to be considered in evaluating each of these three papers separately. Then I will turn my attention to the topic of the session--The Representation of Information in Memory--more generally, and provide a few caveats from the perspective of an experimental psychologist who also happens to be interested in consumer behavior.

The paper by Bozinoff applying script theory to the study of energy conservation is, in my opinion, a very interesting approach to an important problem. It seems to me there are at least three conceptual reasons for why consumer researchers should consider Bozinoff's paper quite carefully. First, it at least addresses the fact that there are very important unconscious determinants of behavior. This is something that is sometimes lost in investigations of consumer behavior, despite the fact that more and more theoretical weight is being placed on underlying cognitive processes in our attempt to understand consumer behavior.

Perhaps the most succinct general definition of cognition is "any computational activity that operates on some input to produce some output." It is important to realize, however, that most of this computational activity is extremely rapid and performed below the level of conscious awareness. George Mandler has cogently made this point by noting that, despite our intuitions, consciousness does not represent the process of cognition but the result of a long line of prior cognitive activity.

It might also be noted that this sort of unconscious computational activity operates at all levels of information processing. It operates at the featural level, it operates at the perceptual level, it operates at the conceptual level; in fact, it operates all the way up to the so-called "higher-order" inference processes. In general, this is one aspect of consumer information processing that has not received the attention it deserves, and the Bozinoff paper should help to correct that

The second aspect of the paper that is worth emphasizing is that it addresses the issue of automaticity in behavior. There was a time not too long ago when most psychologists were very skeptical that the research on automatic and controlled processing would generalize to very much beyond a letter identification task. However, recent research has convincingly demonstrated that automatic processes can also operate at the level of semantic categorization and in category search tasks. The possibility has also been suggested that such automatic processes can be cascaded into "unboundedly complex" systems. All of these issues are likely to become very important in the consumer domain where there are many repetitive, routine sorts of behaviors that are easily classified.

Finally, although the author does not explicitly discuss this aspect of his work, his conceptualization is theoretically quite compatible with the notion of production systems, and this should allow the research to make contact with the more general principles of "if, then" relations. That is, if conditions X, Y, and Z hold, then certain behaviors (or thoughts) will be automatically elicited. This should be quite appealing to those researchers interested in building formal models of consumer information processing.

Briefly, there are also two cautionary notes that need to be considered in evaluating Bozinoff's paper. First, like most schema theorists, Bozinoff tends to define the schema (in this case, a script) in terms of the effects it is postulated to produce, rather than in terms of its structural properties, as assessed independently of these effects. There is a strong precedent for proceeding in this way in the psychological literature as well, but there may be a great danger in doing so. For example, the author writes, "The finding that consumers do not actively seek energy information when buying appliances indicates that a 'gee energy information' action is not in consumers' BUYING APPLIANCES SCRIPTS." What evidential basis is there for making such a statement? Presumably, the only answer is that consumers don't actually seek such information. Or, to take another example, consumers tend to seek price information when purchasing large appliances. Why? Because they have a "gee price information" action in their script. How do we know this? Because they tend to ask for such information. Obviously, this line of reasoning is entirely circular because the same behavior that is initially meant to be explained is also used to validate the theoretical construct that is used to explain it. One could easily substitute "habit" or "drive" or "motive" or even "instinct" into the theory with no apparent loss of explanatory power. Needless to say, one can get into some very serious logical problems by proceeding in this way.

The final issue of concern with the Bozinoff paper is related to the level of abstraction at which the scripts were assessed and activated. This also tends to be a problem with all schema or script theories. Briefly, one could only expect to find a relationship between self-reported scripted activities and actual behavior if indeed the scripts were assessed at the appropriate level of abstraction. Bozinoff found that these relationships were generally weak, but there is no guarantee that this was done in the study. It is quite possible that an assessment of more abstract or generalized scripts, combined with a series of criterion measures, would have led to much better prediction. This is similar to what Fishbein is getting at when he discusses his Multiple Act Criterion. At any rate, it would seem that this possibility also needs to be investigated, and energy conservation would appear to provide an ideal context for it.

Rethans and Hastak argue in their paper that it may be dangerous for memory researchers to continue to concentrate on linguistic information (as opposed to visual images). That may be so, but their case is far less convincing. Again, several factors need to be considered when evaluating their arguments. First, even though Rethans and Hastak criticize the majority of researchers in the field for relying on "linguistic information," they appear to do exactly the same thing. That is, even though they have the subject "read off" a particular image, the subject still needs to perform some linguistic translation of the (possibly verbal) code before it can be reported to the experimenter. Thus, subjects are not operating on the visual image directly, but (at best) on a linguistic translation of something that is stored in a visual code.

It is also important to make a distinction between the content and format of any particular mental representation. In particular, such visual images need not correspond to some sensory information that is picked up at an earlier time, even though the authors seem to assume that they must. For example, subjects asked to close their eyes and walk down a long hallway often report visual images. However, these images tend to be of themselves walking down the hallway. It turns out that many images have this quality of us seeing ourselves engaged in a particular activity, and it's obvious that they are not simply a reflection of previously acquired sensory information. Rather than being "snapshots" of episodic events, these images tend to have a very large semantic (or real-world knowledge) component to them.

Rethans and Hastak suggest that because it took longer to report personal episodes than semantic information, the former can be thought of as more important than the latter. I personally would find that a very difficult position to defend. It simply will take a person more time to describe the one instance in which he/she almost cut off a foot with a power lawnmower than it will to make the general statement that power lawnmowers can be dangerous. This does not necessarily mean that one is more important than the other, and simply counting the number of words spoken seems to be an extremely crude method of determining the importance of these types of thoughts .

There is, of course, a related problem. And that is importance for what? A critically important oversight of the paper is that the authors never discuss what these thoughts are supposed to be important for. What exactly is their functional significance? It seems to me that most of their empirical questions should have been directed at precisely this issue. If the frequency, or proportion, or vividness of these thoughts could be shown to have an effect on judgments, or purchasing decisions, or use, or attitudes, or anything of psychological or behavioral significance, the research would have been much more impressive from both a conceptual and practical point of view.

There is one final point that needs to be considered, and this is one on which the authors and I simply disagree. Rethans and Hastak suggest that episodic memory is basically a veridical account of past events. I would submit that this is potentially a very misleading view. This is particularly true in the area of product hazards, because here there are going to be very important motivational factors that are likely to affect either a participant's or observer's memory for any given event. One-need only observe a heated argument (or any affectively arousing event) to realize that people often have very different recollections of what was said (or done) in any given context.

Ironically, I agree with the authors that personal experiences with a product are very important in making judgments about potential hazards, but I think this has more to do with the nature of the judgment process (and the fact that people may think their memories are veridical) than the likelihood that they are in fact veridical in the sense of there being consensual validation.

Conover has also asked several interesting questions, but has obtained results that are difficult to account for with his conceptualization. I have only two comments to make about Conover's study. First, his finding that greater degrees of product familiarity are associated with more dimensional knowledge structures does not strike me as at all surprising. Although he predicts a curvilinear relationship on the basis of the Hayes-Roth theory, I believe he has misinterpreted the empirical implications of the theory. She does not suggest (at least to me) that product familiarity will be associated with simplicity in the knowledge structures but, rather, with a more systematic organization. In fact, Conover has found some modest support for this.

The only surprising thing to me is that Conover was able to find such support with the type of paradigm employed. In general, the most sensitive methodology for examining pre-existing structures is some type of reaction time procedure. In fact, Hayes-Roth has used reaction time procedures extensively in her own work, and others interested in this area might consider doing so as well.

It might also be noted that many of these same points have been made in the literature on the so-called "paradox of the expert." In general, the more facts there are associated with a given note, the more interference there is and the more time it will take to access any one (in a reaction time paradigm) or the less likely it is to retrieve any one (in a free recall paradigm). There comes a point, however, when the person is able to organize and integrate previously disparate facts into higher-order units. This, in large part, is what allows one to become expert in a Riven area.

A nice illustration related to this is found in a paper presented by Johnson and Russo at the 1980 ACR meetings. These investigators found the type of curvilinear relationship between product familiarity and free recall that would be predicted by Conover when subjects were asked to choose between various products. However, they also found a linear relationship between product familiarity and free recall when subjects rated each product independently. High familiarity subjects appeared to have more complex knowledge structures than low familiarity subjects, and the curvilinear relationship that was found in the choice task appeared to be a byproduct of the high familiarity subjects simply knowing which attributes were irrelevant and/or nondiscriminating. This allowed them to selectively attend to the most relevant attributes (a fact confirmed in supplementary analyses). Conover's analysis, however, seems to confuse simplicity with increased organization, and this is a distinction that should be considered in future empirical and theoretical work.

Perhaps it would be useful to make a few final observations. This was, after all, a session on the Representation of Information in Memory and one would expect to find a few common threads woven throughout the three papers. I would like to mention only two of these threats. Although these are not the types of things that are normally discussed, I would like to do so in the spirit of constructive criticism and with an eye toward the future. First, all three papers report investigations that might gratuitously be referred to as single experiments (with the emphasis on "single" as much as on "experiment"). Given this fact, it is not terribly surprising that they have all produced largely uninterpretable results. That is, in fact, why I have concentrated in my discussion on the conceptual aspects of the papers rather than on the data.

Bozinoff, for example, cannot be sure whether there really is no relationship between self-reports of behavior and inclusion of those same activities in a script or whether he simply measured the scripts at the wrong level of abstraction Rethans and Hastak do not know whether product hazard information is truly represented in a visual code or whether their results are an artifact of subjects forming images after the fact and/or responding to the demand characteristics of the study.

Conover, as he acknowledges, cannot determine whether there simply is not a curvilinear relationship between product familiarity and dimensionality or whether truly "high familiarity" subjects were not represented in the study. All of these very basic questions remain unanswered.

I would suggest it is time we honestly begin to examine the utility of such isolated experiments. Many discussants of past ACR sessions have commented on the lack of coherence between the various papers and the fact is that very few of these studies are ever followed up. Before we really begin to make progress in understanding the Representation of Information in Memory (or any other content domain within the field of consumer behavior), we are going to need systematic programs of research that seriously begin to narrow down the range of plausible alternative explanations that exist.

Despite what we might like to believe, and what we often imply to our students, contemporary philosophers of science are continually reminding us that there really is very little we can learn from a single experiment. On the other hand, the type of "research programs" that have been discussed by Lakatos and Lauden, as well as other philosophers of science, can be extremely powerful methods for answering many of the theoretical questions that are important in contemporary investigations of consumer behavior.

The second common theme is that all three papers appear to contain an implicit apology for not leading to any important advance. Interestingly, all three authors end up sounding amazingly similar. Conover writes, "This study is clearly exploratory in nature...." Rethans and Hastak write, "This exploratory study...." Bozinoff continues with the same theme by writing, "The study described below serves only as an initial exploratory investigation...." One wonders whether there has ever been an ACR paper that has not contained such a disclaimer.

Although exploratory investigations can be criticized on a variety of grounds, I generally try to remain sympathetic to such endeavors, as they are clearly necessary for the development of the field. The fact that they tend to produce uninterpretable results does not particularly concern me, as long as they are recognized to be so. The three papers presented at the present session are a good example of this. All three sets of data are ambiguous in the sense that there are multiple interpretations of them but all three authors appear to recognize this and (generally) offer the appropriate caveats.

The fact that exploratory investigations tend never to be followed up concerns me only to a minor degree. As noted above, I am convinced that notable advances will only be attained with systematic programs of research rather than isolated experiments on various topics. On the other hand, generating a large number of ideas is not always as important as being able to quickly reject the bat ones. A quick glance through past volumes of the ACR proceedings will produce many examples of exploratory investigations that have apparently never been followed up, and probably never deserved to be.

The one thing that concerns me a great deal is that by calling a particular study "exploratory" in nature allows one to avoid addressing the very difficult issue of why that particular question is an important one in the first place. There are phrases like, "I was interested in..." "It would be interesting to discover whether...", or "I was curious about..." that we are beginning to hear with an alarming frequency. The problem with such an approach is that it implicitly assumes that all ideas are created equal and one hypothesis is as good as another. The potential result of such a situation is a disjointed, disorganized, and utterly chaotic discipline where "anything goes."

I would like to argue that it is in choosing the right question to investigate, rather than even the technical sophistication with which the study is conducted (since that will surely be improved upon), that the major advances in consumer behavior will take place. This is particularly true in an area like The Representation of Information in Memory, but it is doubtless true in other areas as well.

As noted above, all of these comments are offered in the spirit of constructive criticism. Although the three papers presented in the present session score fairly high along these dimensions, they are issues that are well worth considering in general. In particular, it is hoped that future investigators will begin to give very serious consideration to the formulation of the questions that are going to be investigated because, I believe anyway, that the ultimate progress of the field is going to depend on it.

A final possibility worth considering is that one is able to learn a great teal about the expectations of a discipline by examining the type of questions it chooses to investigate. Of course, a discipline like consumer behavior simply does not exist outside of the realm of its individual investigators. In this regard, it may be instructive for each of us to quietly reflect on our own research questions, where we honestly expect them to lead, and whether our own research programs have truly progressed or simply "changed directions" over the years.

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