Toward Articulating Theories, Triangulating Concepts, and Disambiguating Interpretations: the Role of Converging Operations in Consumer Memory and Judgment

ABSTRACT - A variety of general methodologies for the study of consumer memory and judgment are outlined. Although each of them is useful in isolation, it is argued that researchers should strive to combine them, and collect converging evidence from widely divergent paradigms, whenever possible.


Thomas K. Srull (1992) ,"Toward Articulating Theories, Triangulating Concepts, and Disambiguating Interpretations: the Role of Converging Operations in Consumer Memory and Judgment", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 23-27.

Advances in Consumer Research Volume 19, 1992      Pages 23-27


Thomas K. Srull, University of Illinois at Urbana-Champaign


A variety of general methodologies for the study of consumer memory and judgment are outlined. Although each of them is useful in isolation, it is argued that researchers should strive to combine them, and collect converging evidence from widely divergent paradigms, whenever possible.

I have had a passing familiarity with the study of consumer behavior over the past decade and, more than anything else, two things have become apparent. One is that information processing theory has clearly become the dominant theoretical -- or metatheoretical -- approach to the field. Everything from introductory textbooks in consumer behavior to the pages of our most elite journals are filled with its fruits.

There are, of course, very good reasons for this. One is that information processing theory has been enormously successful in revolutionizing basic experimental psychology. A related but still distinct reason is that it has been nearly as successful in revolutionizing the study of consumer behavior as well. Without disparaging the impact of other traditions or viewpoints, the past decade has provided a virtual explosion of new scientific questions and, in many cases, quite subtle and sophisticated answers. The fact that this explosion of new knowledge has coincided with the field's adoption of information processing theory as the major guiding paradigm is no accident.

The second thing that has become apparent to me over the past decade, and this is in no way inconsistent with the first, is that we are still a very immature field in our approach. As I'm sure many of us do, I sometimes marvel at how researchers can tackle such difficult problems in such an elegant and sophisticated way. But then there are other times in which I become distraught over how many loose ends remain untied.

The field of consumer behavior is following a natural evolutionary course, one that has mirrored quite closely that of cognitive psychology. However, since cognitive psychologists have been active much longer, and have progressed much further, it is useful to examine some of the changes in orientation that have been responsible for their progress. In this way, we can avoid some of the mistakes they made and capitalize on some of their more productive achievements.

Anyone who reads both the cognitive psychology and consumer behavior literatures will be struck, above all, by the following: while researchers in consumer behavior typically use one experiment to test five or six hypotheses, cognitive psychologists typically use five or six experiments to test one hypothesis. This difference follows from a very basic orientation concerning how and when scientific inferences can be drawn from observed empirical data.

This point is worth elaborating in a slightly different way. Researchers in consumer behavior typically use a single experimental paradigm, and then draw inferences about several different cognitive processes. This is useful, and even necessary, in the early stages of investigation because meaningful theories require a data base that is both rich and descriptive. Cognitive psychologists working with virtually every psychological process traveled down this same path. Eventually, however, cognitive psychologists discovered that this is a very dangerous strategy in the long run because not nearly enough constraints are imposed on the theory. If one continues to follow this strategy, the theories will remain unspecified and unwieldy.

Having learned their lesson, painfully but well, cognitive psychologists now follow exactly the opposite strategy; they use many different paradigms to provide evidence that converges upon a single type of process or memory representation. As consumer behavior develops into a mature field, we will need to move in the direction of gaining evidence about a single memory representation from many different paradigms and dependent measures.

There is every reason to believe that those measures that have proved to be most useful in cognitive psychology will prove to be equally useful in consumer behavior. In the rest of this paper, I will try to outline in general terms what these measures are and when they can be used most easily. The point to keep in mine, however, is that they can and should be used in conjunction with each other whenever possible.

Free Recall Measures

A common type of experiment in consumer behavior is one in which consumers receive information about an object (e.g., a brand or category) with instructions to use it for a particular purpose (e.g., to learn the information or form an evaluative judgment). They are then asked at some later time to recall as much of the information as possible. A researcher might then compute any of six different indices: (1) the total amount of information recalled, (2) the type of information that tends to be recalled, (3) the order in which the various items are recalled, (4) the number of intrusions that are reported, (5) the latencies associated with the recall of each item type, and (6) the implications of the recalled information for a particular type of judgment or decision.

These indices are typically used to draw conclusions about three separate questions: (1) the extent to which the information was initially learned, (2) the content and organization of the mental representation that is formed, and (3) the particular aspects of the information that are used as a basis for judgments or decisions. These inferences can be made in the context of (at least) four general memory models.

Associative network models. Many models conceptualize the representation of information in memory as an associative network of concepts and relations. Concepts in such a network are represented by nodes, and associations by pathways connecting them. However, the nodes themselves are assumed to have no internal structure. Associations are usually thought to be the result of contemporaneous representations of the concepts in working memory (rehearsing two items together, thinking about the concepts in relation to each other, and so on).

Two general types of retrieval processes have been postulated. One involves a serial directed search of the network. Once the concept represented by a particular node is activated, other concepts are retrieved by progressing from node to node along the pathways that have been established (see e.g., Srull, Lichtenstein, & Rothbart, 1984; Srull & Wyer, 1989). A second type of model assumes a parallel search of the network. This is best exemplified by various spreading activation formulations (see e.g., Anderson, 1976).

Retrieval models. It should be clear that associative network models tend to put all of the conceptual weight on the relations between concepts, and they deemphasize the internal structure of new episodes that are encoded into memory. Retrieval models take nearly the opposite approach.

As a general class, retrieval models make minimal assumptions about how a newly encoded episode is "organized" and represented in relation to other things that are known (see e.g., Ratcliff, 1978). That is, they deemphasize any relations that may exist between concepts. They do, however, make elaborate assumptions about the internal structure of any encoded event. In most cases, a vector or matrix format is used to represent the structure of new events that are encoded into memory. Each event is broken up into a set of elementary features (which include self-generated associations) and the "values" of such features are the components of the vector or matrix that is used to represent the event. Each vector or matrix is stored independently of all the others.

A common element of these models is that a retrieval cue (either presented by the experimenter or self-generated) can also be decomposed into more elementary features and represented in vector or matrix format. The effectiveness of any cue will be a function of the "similarity" of its own vector (matrix) with the vector (matrix) of the to-be-recalled event. This provides one explanation for why a cue is effective only if some or all of its features were thought about at the time of encoding (Tulving & Thomson, 1973).

Reconstructive models. A third type of model is particularly powerful in conceptualizing the dynamics that result in the intrusion of general world knowledge into the recall of specific information. There are two general versions of this model.

One postulates the existence of previously learned knowledge structures (scripts, schemata, etc.) that function as configural units and are used to interpret raw information whose features can be instantiated in terms of them. For example, one may have a particular schema for health food. When specific information is received about a particular health food, this information is not directly stored. Rather, only a "pointer" to the schema is retained, along with equivalence rules for translating specific features of the new information into prototypic ones and back again. This representation is sufficient to permit the new information to be reconstructed later, even though its details are not uniquely stored. However, because of this same process, one would expect to find many intrusions of unspecified features in consumer's recall protocols as well.

A second general type of reconstructive model also assumes that new information is interpreted with reference to a pre-existing knowledge structure. However, it assumes that once the information is organized and encoded in terms of this prior knowledge, it is stored in memory as a conceptual unit rather than as simply a pointer to the prototype used to interpret it (see e.g., Wyer & Srull, 1986).

Amount-of-processing models. A fourth conceptualization places special emphasis on the amount of cognitive activity involved in processing information at the time it is encoded (see e.g., Burke & Srull, 1988). In general, these models assume that more extensively processed information becomes more accessible in memory and therefore is more likely to be recalled. This may be true because thinking more extensively about information leads it to be more elaboratively encoded in terms of previously formed concepts, thereby making these concepts more effective retrieval cues (Jacoby & Craik, 1979). Another possibility is that more extensively processed information becomes more strongly associated with contextual (situational) cues available at the time it is received, and therefore is more likely to be retrieved in any later situation in which those same cues are present (Tulving, 1983).

Cued Recall Measures

One of the most significant influences on the development of an information processing framework for the study of memory and cognition was the primary consideration given to retrieval and computational processes (e.g., Lichtenstein & Srull, 1985, 1987), as well as to instances of retrieval failure (e.g., Lynch & Srull, 1982). The use of cued recall tasks has become increasingly common as more and more theoretical issues concerning retrieval have been raised (see e.g., Keller, 1987, 1991).

It is important to distinguish between cued recall as a process and as an experimental procedure. Several theorists (e.g., Battig & Bellezza, 1979; Voss 1979) have argued that much of one's recall involves the use of self-generated cues. Thus, a free recall task might involved "cued recall" in terms of the underlying processes being activated. As an experimental procedure, however, a cued recall task involves a situation in which the experimenter provides cues (at the acquisition stage, retrieval stage, or both) that may aid the subject in recalling other material.

There are two major advantages to using cued recall as an experimental procedure. First, it can often be used to detect information available in memory that less sensitive methods (e.g., free recall) are unable to detect. The second advantage of cued recall is more subtle but even more important. Many contemporary theorists (see e.g., Raaijmakers & Shiffrin, 1980, 1981) postulate that all information encoded into long-term memory is potentially available from that point on, and "forgetting" is due to the inability to retrieve a given piece of information in a particular context. By experimentally varying the situational context, and therefore the nature of the cues available to the subject, one can examine under what conditions people will and will not be able to retrieve given information.

Recognition Measures

Measures of recognition memory like the free recall and cued recall measures discussed earlier, are often used as an index of learning and retention. However, the conclusions one draws from these measures are probably even more dependent on the theoretical assumptions one makes about the processes that underlie the responses being measured.

In a recognition paradigm, subjects who have previously been exposed to a set of stimulus items are then presented with each of several test items and asked to decide whether each item was among the ones they had seen earlier. Some of the test items were actually presented before, and others ("distractors") were not. Three indices are usually considered: (1) the probability of correctly identifying a stimulus item as having been presented (the hit rate), (2) the probability of stating that a distractor had been presented when it had not (the false alarm rate), and (3) the time required to make these judgments. I will focus here only on the first two measures (for a discussion of how response time can provide additional information in a recognition memory paradigm, see Srull, 1984).

Quite different theories exist concerning the processes that underlie recognition responses. One view is that free recall, cued recall, and recognition all lie along a single continuum. That is, they are not qualitatively different from each other (Watkins, 1979). While in a free recall paradigm there is only one cue and it is very general ("recall everything you can from the list"), in a cued recall paradigm there are many cues and they are necessarily more specific. According to this model, a recognition paradigm is very similar. In recognition, many of the cues that are provided to subjects to aid recall are duplicates of the original items (these are called copy cues). However, the underlying retrieval process is no different than in cued recall, or free recall for that matter. According to this model, the critical determinant of recognition responses (like all others) is the association of the stimulus items stored in memory with the retrieval cues available at the time of recognition. In other words, recognition involves a retrieval process that is no different from that in free recall.

An alternative view is proposed by Gillund and Shiffrin (1984). They postulate that the primary criterion for a recognition response to an item is the item's subjective familiarity. That is, a test item will be responded to affirmatively if it is experienced as "familiar" in the situational and informational context in which the item is presented. According to this model, search and retrieval processes, which play the dominant role in free recall, are simply bypassed when recognition memory measures are used.

A third, less well formalized conceptualization incorporates features of both of the above models. According to this view, a subject who receives a recognition test item will first use its features as cues to probe long-term memory for an item that has these same features. If the search is successful, a "yes" response will be given. However, if the subject does not find an item, he or she will have to "guess" whether the test item was included in the original list. This guessing process will be based on some index of subjective familiarity, which is assumed to be a function of the similarity of the item in question to other items the subject does remember as having been presented.

The conclusions drawn from a given set of recognition data will often depend on which of these general sets of assumptions is implicitly made. Several detailed examples of the interpretive controversies that can result are provided by Wyer and Srull (1988).

True - False Verification Measures

It seems reasonable to assume that the time required to respond to a stimulus is some function of the time required to perform each of the component stages that underlie the generation of this response. Moreover, if one has a precise theory of the stages of processing that underlie judgments in a particular situation, and if one postulates factors that independently affect the different stages of processing, these factors should have independent and additive effects on overall response time (Sternberg, 1969, 1975). If the effects are not additive, it implies that either: (1) the experimental factors do not affect the stages of processing that one assumes, or (2) the stages themselves are nonindependent.

There is often an inherent ambiguity in interpreting response-time data. This is particularly true when the responses being made require assumptions about both: (1) the structure and organization of the cognitive material on which the response is based, and (2) the nature of the retrieval and use of this material. As Anderson (1976) and others have pointed out, it is often impossible to localize the logical source of response-time differences. In other words, a given pattern of response times can be accounted for equally well by: (1) postulating differences in the structure and content of memory, but assuming a single search and retrieval process for accessing it, or (2) assuming a single memorial representation and postulating different search and retrieval processes. In many instances, these alternative interpretations must be evaluated in terms of their relative plausibility or parsimony rather than (simply) their consistency with the data.

These considerations make it clear that the use of verification times as a memory index is highly contingent on the a priori assumptions one wishes to make about the process that underlies them. When one is willing to make a particular set of assumptions, however, some useful conclusions can be drawn. An elegant example is provided by Sentis and Burnstein (1979) in an investigation of subjects' tendency to organize information in memory in a manner implied by cognitive balance theory. Subjects received sets of three affective relations among persons and objects that were theoretically either balanced or imbalanced. Then they were presented with configurations of one, two, or three relations from the original materials, and asked to verify whether they were among the relations they had learned.

Sentis and Burnstein postulated that subjects would store the relations in imbalanced triads independently of one another, and thus the time required to verify a configuration would increase with the number of relations presented. However, subjects were expected to organize the relations contained in balanced sets into integrated units rather than storing them as independent entities. They were therefore expected to verify a balanced configuration more quickly when all three of the relations it comprised were presented than when only a subset was presented. In short, the time required to verify the relations contained in imbalanced sets was expected to increase with the number to be verified, whereas the time required to verify the relations contained in balanced sets was expected to decrease with the number of relations presented. The results supported both hypotheses, and provide an elegant demonstration of the use of this type of methodology.

Ease of Retrieval Measures

A wide range of questions related to cognitive organization can be investigated by examining the ease with which material that is represented in memory can be accessed. The precise experimental paradigms that are used vary greatly; while some resemble yes-no verification tasks, others resemble free recall tasks, and still others fall somewhere in between. Three representative but very different examples are described below to illustrate the continuum.

One question that often occurs within the study of consumer behavior is how closely one item is associated with another (and how the strength of that association might be manipulated). For example, Fazio, Herr, and Powell (1992) examined different advertising executions that presented the same information; the only difference was that the brand name was identified very early in one case, and not until the end of the ad in the other (so-called Mystery Ads). They then presented consumers with a yes-no verification task in which they were to indicate whether a particular brand belonged to the product category designated (e.g., Crest - Toothpaste).

Fazio et al. found that the time required to make these verifications was much less when a novel brand was learned about via a Mystery Ad than in the more traditional executional style in which the brand name is presented first. This suggested that the brand-category associations were much stronger in the former than latter case. Moreover, Fazio (1990) has shown that such verification times correlate highly with how early a brand is listed when consumers are asked to list all of the members of a given product category, providing just the type of converging evidence that one would hope to find more often.

Similar approaches can be used to determine when consumers can retrieve information directly, rather than compute it indirectly on the basis of other, relevant information that has been stored. For example, Kardes (1988) presented consumers with an advertisement in which the conclusion from an argument was stated directly or only implied. He then measured the amount of time required to determine that the conclusion was true.

Consumers who saw the conclusion directly stated could simply retrieve it from memory, and their response times were very fast. Kardes also discovered, however, that consumers who saw the conclusion only implied were equally fast to respond if they were in a high involvement situation. This indicated that these consumers drew the necessary inference at the time they acquired the information, and they too could retrieve the (in this case, inferred) conclusion directly from memory. In contrast, when consumers were in a low involvement situation, and the conclusion was omitted, the time required to respond was much longer, indicating that there was very little inferential activity at the time of acquisition.

Finally, there are many situations in which it is useful to manipulate rather than assess the ease of retrieval. For example, semantic categories that are primed often have a dramatic effect on how subsequent information is processed (for a review, see Wyer & Srull, 1989).

It also appears that the ease with which information is recalled influences the way in which its implications are used. In a very powerful study, for example, Schwarz, Bless, Strack, Klumpp, Rittenauer-Schatka, and Simons (1991) asked subjects to recall either six examples of their own assertive behavior (a very easy task) or twelve examples of their own assertive behavior (a very difficult task). Other subjects did the same with unassertive behavior.

Schwarz et al. found that subjects who recalled twelve examples of assertive behavior actually rated themselves as less assertive than subjects who recalled only six examples. In fact, ratings of assertiveness were higher after recalling twelve unassertive behaviors than twelve assertive behaviors. Based on these and other data, it appears that self-assessments only reflect the implications of recalled content when the ease of retrieval is high. Otherwise examples that are retrieved from memory are discounted.


Cognitive psychologists have developed a wide range of paradigms that can be used to study consumer memory and judgment. An overreliance on any one of them is dangerous because, as history has demonstrated, the data collected from any single paradigm do not impose enough constraints for a complete and well articulated theory. Thus, consumer researchers should strive to collect evidence from different paradigms that converge on a common set of psychological processes. In this way, interpretations can be made less ambiguous, concepts can be triangulated, and theories can be (tentatively) fleshed out in a more bottom-up fashion. If the impressive progress that has been made in our understanding of consumer psychology is to continue, such an approach will need to be taken. There are already some signs that this is occurring, and the purpose of the present paper is to help accelerate that trend.


Anderson, J. R. (1976). Language, memory and thought. Hillsdale, N.J.: Erlbaum.

Battig, W. F., & Bellezza, F. S. (1979). Organization and levels of processing. In C. R. Puff (Ed.), Memory organization and structure. New York: Academic Press.

Fazio, R. H. (1990). A practical guide to the use of response latency in social psychological research. In C. Hendrick & M. S. Clark (Eds.), Review of personality and social psychology (Vol. II). Newbury Park, CA: Sage.

Fazio R. H., Hess, P. M., & Powell, M. C. (1992). On the development and strength of category-brand associations in memory: The case of mystery ads. Journal of Consumer Psychology.

Gillund, G., & Shiffrin, R. M. (1984). A retrieval model for both recognition and recall. Psychological Review, 91, 1-67.

Jacoby, L. L. & Craik, F. I. M. (1979). Effects of elaboration of processing at encoding and retrieval: Trace distinctiveness and recovery of initial context. In L. S. Cermak & F. I. M. Craig (Eds.), Levels of processing in human memory. Hillsdale, N.J.: Erlbaum.

Kardes, F. R. (1988). Spontaneous inference processes in advertising: The effects of conclusion omission and involvement on persuasion. Journal of Consumer Research, 15, 225-233.

Keller, K. L. (1987). Memory factors in advertising: The effect of advertising retrieval cues on brand evaluations. Journal of Consumer Research, 14, 316-333.

Keller, K. L. (1991). Memory and evaluation effects in competitive advertising environments. Journal of Consumer Research, 17, 463-476.

Lichtenstein, M., & Srull, T. K. (1985). Conceptual and methodological issues in examining the relationship between consumer memory and judgment. In L. F. Alwitt & A. A. Mitchell (Eds.), Psychological processes and advertising effects: Theory, research and application. Hillsdale, NJ Erlbaum.

Lichtenstein, M., & Srull, T. K. (1987). Processing objectives as a determinant of the relationship between recall and judgment. Journal of Experimental Social Psychology, 23, 93-118.

Lynch, J. G., & Srull, T. K. (1982). Memory and attentional factors in consumer choice: Concepts and research methods. Journal of Consumer Research, 9, 18-37.

Raaijmakers, J. G. W., & Shiffrin, R. M. (1980). SAM: A theory of probabilistic search of associative memory. In G. H. Bower (Ed.), The psychology of learning and motivation, Vol. 14. New York: Academic Press.

Raaijmakers, J. G. W., & Shiffrin, R. M. (1981). Search of associative memory. Psychological Review, 88, 93-134.

Ratcliff, R. (1978). A theory of memory retrieval. Psychological Review, 85, 59-108.

Schwarz, N., Bless, H., Strack, F., Klumpp, G. Rittenauer-Schatka, & Simons, A. (1991). Ease of retrieval as information: Another look at the availability heuristic. Journal of Personality and Social Psychology, 61, 195-202.

Sentis, K. P., & Burnstein, E. (1979). Remembering schema-consistent information: Effects of a balance schema on recognition memory. Journal of Personality and Social Psychology, 37, 2200-2212.

Srull, T. K., (1984). Methodological techniques for the study of person memory and social cognition. In R. S. Wyer & T. K. Srull (Eds.), Handbook of social cognition (Vol. 2, pp. 1-72). Hillsdale, N.J.: Erlbaum.

Srull, T. K., Lichtenstein, M., & Rothbart, M. (1985). Associative storage and retrieval processes in person memory. Journal of Experimental Psychology: Learning, Memory and Cognition, 11, 316-345.

Srull, T. K., & Wyer, R. S. (1989). Person memory and judgment. Psychological Review, 96, 58-83.

Sternberg, S. (1969). Mental processes revealed by reaction-time experiments. American Scientist, 57, 421-457.

Sternberg, S. (1975). Memory scanning: New findings and current controversies. Quarterly Journal of Experimental Psychology, 27, 1-32.

Tulving, E. (1983). Elements of episodic memory. Oxford: Clarendon Press.

Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80, 352-373.

Voss, J. F. (1979). Organization, structure, and memory: Three perspectives. In C. R. Puff (Ed.), Memory organization and structure. New York: Academic Press.

Watkins, M. J. (1979). Engrams as cuegrams and forgetting as cue overload: A cueing approach to the structure of memory. In C. R. Puff (Ed.), Memory organization and structure. New York: Academic Press.

Wyer, R. S., & Srull, T. K. (1986). Human cognition in its social context. Psychological Review, 93, 322-359.

Wyer, R. S., & Srull, T. K. (1988). Understanding social knowledge: If only the data could speak for themselves. In D. Bar-Tal & A. W. Kruglanski (Eds.), The Social psychology of knowledge. Cambridge: Cambridge University Press.

Wyer, R. S., & Srull, T. K. (1989). Memory and cognition in its social context. Hillsdale, NJ: Erlbaum.



Thomas K. Srull, University of Illinois at Urbana-Champaign


NA - Advances in Consumer Research Volume 19 | 1992

Share Proceeding

Featured papers

See More


Crossing Race and Markets: Introducing the Race in the Marketplace Research Network

Kevin D Thomas, University of Texas at Austin, USA
Sonya Grier, American University, USA
Guillaume D Johnson, Université Paris-Dauphine

Read More


Q4. The notion of self-optimization in context of self-tracking and beyond

Agnieszka Krzeminska, Leuphana University Lüneburg
Dorthe Brogård Kristensen, University of Southern Denmark, Denmark

Read More


When Negative Observations Broaden Generalization of Product Attributes to Novel Products

Rui Chen, Tarleton State University
Marcus Cunha Jr., University of Georgia, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.