Altering Retrieval Sets: When Will Contextual Cues Make a Difference?


Elizabeth J. Cowley and Sharmistha Law (1995) ,"Altering Retrieval Sets: When Will Contextual Cues Make a Difference?", in NA - Advances in Consumer Research Volume 22, eds. Frank R. Kardes and Mita Sujan, Provo, UT : Association for Consumer Research, Pages: 323-327.

Advances in Consumer Research Volume 22, 1995      Pages 323-327


Elizabeth J. Cowley, University of Toronto

Sharmistha Law, University of Toronto

It is known that the retrieval of brand information is a necessary condition for choice. It is also known that retrieval is affected by cues present in the encoding and retrieval environment. This paper examines the role of prior knowledge as a moderator in the relationship between contextual cues and set formation. Results show that retrieval cues are influential in the formation of the retrieval set for the high knowledge consumers, while low knowledge consumers are unaffected by the retrieval cues. Not only are the retrieval sets of the high knowledge consumer altered across retrieval contexts, but the additions to the set are relevant to the new retrieval context.


A great deal of attention has been paid to the advantage enjoyed by experts when solving a problem and making a decision. Experts have been found to more effectively use informal reasoning mechanisms (Voss, Blais, Means, Greene and Ahwesh 1986) and general principles (Chi, Feltovich and Glaser 1982; Larkin, McDermott, Simon and Simon 1980). In a marketing context, their improved performance has been attributed to an increased motivation to process attribute statements (Beattie 1983; Maheswaran and Sternthal 1990; Walker, Celsi and Olson 1987) and incongruent information (Sujan 1985); an ability to reorganize incoming information into a meaningful format (Hutchinson 1983; Srull 1983); or their efficiency in information search (Brucks 1985). Unfortunately, little attention has been paid to the ability of the expert to retrieve the most appropriate set of alternatives given a purchase objective (for an exception see Brucks 1985). This ability is critical in making optimal purchase decisions, as retrieval of the appropriate alternative is a necessary condition for its eventual choice. Sophisticated choice strategies are wasted if conducted on a suboptimal set of alternatives.

Recently, evidence has been offered supporting the idea that retrieval sets are dynamic, changing across choice occasions (Nedungadi 1990). These findings are counter to the previous assumption that retrieval sets are relatively static; that consumers generate and store the sets in order to simplify choice (Howard and Sheth 1969). In the static model of the retrieved set, the inclusion of a brand is decided on the basis of its assessed utility assigned during information search. In the dynamic model of the retrieved set, inclusion of a brand is influenced by the presence of a usage context, purchase objective or buying occasion at retrieval.

The dynamic model has essentially introduced flexibility into the retrieval process. It assumes that the consumer is able to recall different brands on various choice occasions. Given a set of information, people can recall different aspects of the information when presented with different cues (Anderson and Pichert 1978; Pichert and Anderson 1977). In a marketing setting, the ability to retrieve different details about a brands when the usage context varies has been related to the amount of prior knowledge held by the consumer (Cowley 1993). It is not clear however, that the ability to recover a richer description of brands in response to a variation in contextual cues will translate into relevant additions to the set of retrieved brands. A more appropriate retrieval set may result in more optimal decision making.


In memory based choice, the formation of the retrieval set is a precursor to choice. According to Nedungadi (1990), choice is a two stage process. The first stage, the brand consideration stage, includes the retrieval of the set of alternatives. The second stage reduces the consideration set to a set of brands selected for further evaluation. Alba and Chattopadhyay (1985) distinguish the consideration set from the retrieved set conceptually, but use the retrieval set in their investigation of memoryBbased choice. The reasoning for the operationalization is 1] that retrieval is a necessary condition of choice and 2] that membership in the retrieval set is more stable. It is the retrieval set which is of interest here.

The issue critical to the thesis of this paper arises from the comparison of the static model and the dynamic model discussed above. Implicit in the dynamic model is the assumption that the consumer can access new information when the retrieval context changes. This ability may vary with the expertise held by the consumer in a given domain. In other words, prior knowledge may moderate the ability to alter the retrieval set in different contexts. Also, while few researchers would contest the importance of context in the formation of the set of alternatives, explicit attention has not been paid to the impact of the contextual cues existing at the time of encoding and their interaction with the cues present while retrieving, on the retrieval set. This paper addresses these gaps in the literature.



High knowledge consumers are better able to recall information relevant to a subject (Chiesi, Spilich and Voss 1979). Superior recall may result from a more effective memory search which requires the more elaborate knowledge structure associated with experts (Chi, Feltovich and Glaser 1982; Fiske, Kinder and Larter 1983; Mitchell, Dacin and Chi 1993).

Alba and Hutchinson (1987) suggest that the retrieval set for a higher knowledge consumer should include a superior group of brands. In this study, superiority of the set is assessed on the basis of its relevance to the decision context. If the mechanism influencing the formation of the retrieval set occurs when information is recalled, then brands most appropriate to the retrieval context would be present in the retrieval set independent of their appropriateness to the encoding context. In other words, the retrieval explanation predicts that the retrieval set of the consumer should be most appropriate to the retrieval context. It follows that as the retrieval context changes, so will the membership of the retrieved set: brands will be recovered which were inaccessible given a different retrieval context. Thus, the dynamic model will be upheld, and high knowledge consumers will recover new alternatives when the choice context varies.


As predicted by the retrieval explanation, the elaborate knowledge structure of the high knowledge consumer results in flexible retrieval set, similar to the one modelled in the dynamic retrieval framework. The less developed knowledge structures of the low knowledge consumer however, will limit the ability of the decision maker to vary the retrieved set over choice occasions. When brand information is provided to consumers and later retrieved with various usage contexts:

H1: Consumers high in product knowledge will be more likely than consumers low in product knowledge to alter their retrieval sets given differences in retrieval contexts.

H2: Additions to the retrieval set or recovered brands, will be relevant to the new context.



The accumulation of knowledge may occur over time and across different usage contexts. It is therefore, reasonable to expect that consumers' prior knowledge will impact their encoding of new information. In fact, the advantage of the expert at encoding is demonstrated by Johnson and Russo (1984) who report that knowledgeable consumers are able to concentrate on the information relevant to the task at hand (see also Spilich, Voss, Chiesi and Vesonder 1979). Alba and Hutchinson (1987) concur that the ability to interpret the attributes of a product on a taskBrelevant basis is a function of expertise within a product domain, and that high knowledge consumers are more able to identify relevant information (Alba 1983). The high knowledge consumer is thus more likely to guide their effort during encoding. It is predicted that high knowledge consumers will process the information relevant to the encoding context more selectively, and therefore find it more difficult to recover previously inaccessible information relevant to another context.

H3: The presence of a usage context at encoding will reduce the ability of the high knowledge consumer to recover brands when the usage context varies.


In summary, retrieval sets will vary over choice occasions for high knowledge consumers. Retrieval sets will vary less over purchase occasions for the low knowledge consumers when compared to high knowledge consumers. The presence of a context at encoding will affect the ability of the high knowledge consumer to recover different brands when compared to encoding without a usage context.


Subjects and Design

Sixty six [Data for four of the subjects were excluded as they were unable to comprehend or complete the study.] undergraduate students of an eastern university participated in the study. All subjects participated in three sessions: encoding, recall session one and recall session two.

Encoding. During the encoding session subjects were randomly assigned to one of two conditions: no encoding context or encoding context (see the Design diagram). Subjects assigned to the no encoding context condition were simply asked to read the product information slowly and carefully. Subjects in the encoding context condition were given context one which was one of two usage contexts: image or functional (to be described later).

Recall One. During recall session one, subjects were asked to use context one (in the encoding context condition) or provided context one (in the no encoding context condition) when recalling information.

Recall Two. During recall session two, subjects were provided context two prior to retrieving the information.

The order of presentation of the retrieval contexts was counterbalanced across subjects to minimize any systematic bias due to differences in their content. The order of presentation of the contexts was not a significant factor in recovery performance (F=1.55, p>.22). This allays any concern that the content of the contexts drove the ability to recover brands.

Subjects were blocked as high or low in product knowledge on the basis of their subjective and objective knowledge in the domain, familiarity and experience scores. (see Design Diagram).


Brand information was presented as though the reader was taking a walk through a bicycle shop, looking at the merchandise on display. Ten bicycles were described with five bicycles suitable for each usage context: image or functional. Brand information was pretested to ensure that an equal amount of important information for each of the usage contexts. The image context was operationalized by instructing the subjects to imagine that they have recently joined a bicycle club, that the club meets for rides on weekend afternoons. They have been to one meeting, and they like the people in the club very much, but they noticed that everyone had a trendy, stylish bike. They are told that their bike is not trendy or stylish, so they are going shopping for a new bike. The functional usage context was operationalized by instructing the subjects to imagine that they have decided to ride a bicycle to school instead of taking the TTC or driving. They need a bicycle to get them from point A to point B. That they do not need anything fancy, they just something durable and reliable. They are also concerned for their safety on the street.



Bicycles were described with concrete attributes only. In other words, the attributes were quantified in terms of colour, size, durability, brand reputation and warranty specifications. A typical description was "the second mountain bike is blue, the brand name is wellBknown for quality and workmanship. The frame is plain and simple. The frame comes with a three year warranty."

Past research has demonstrated that the complexity of the information affects the influence of expertise on the amount recalled. Specifically, low knowledge consumers remembered at least as much of the simple information, but dramatically less of the complex information (Alba 1983). To avoid comprehension as an alternative explanation, terminology in the information was simple. Any less familiar terms were explained in nontechnical language. For instance, "the bike is an American brand, it has 26" wheels which is a size that is easy to handle".

As expected, the amount recalled, or in this case, the size of the retrieval set did not vary between high and low knowledge consumers (low knowledge=3.9, high knowledge=4.3, t=-0.7837, p>.44). The implication is that low knowledge subjects were able to understand the information as well as high knowledge subjects.

Measures of Expertise

To accurately capture the level of expertise, four measures were taken; experience, familiarity, subjective knowledge and objective knowledge. The experience measure was an indication of how often the subject rides a bicycle (never B 0 and very often B 10). The familiarity measure was an indication of how familiar the subject felt they were with bicycles (novice B 0 and expert B 10). The subjective measure of expertise was an indication of the amount of information the subject held with respect to bicycles (novice B 0 and expert B 10). Finally, the objective measure of knowledge was a raw score on eight multiple choice questions, five definitions and the number of brandnames provided on an unaided test of recall.

The coefficient alpha for the sum of these four measure was .84. All of the correlation coefficients were significant (p<.0001). A median split of this score resulted in 31 high knowledge subjects and 31 low knowledge subjects.


As described earlier, subjects were randomly assigned to one of the conditions: no encoding context, encoding context. Subjects in the no encoding context condition were not given a context prior to encoding, while subjects in the encoding context condition were given context one before reading the brand information. All subjects read the stimulus information. The next twenty minutes were spent on an unrelated task.

During recall session one, subjects in the no context condition were given context one at retrieval. All subjects were using context one to retrieve the information. Subjects were instructed to write down the usage context and to list as many of the bicycles as they could possibly remember.

Following the first recall session one, subjects spent six minutes working at an additional unrelated task. Subjects were then told "that people can sometimes remember information that they thought they had forgotten if they were given a new perspective with which to think about the information". They were provided with usage context two. They were asked to provide the context and to list as many of the bicycles as possible.

Finally, subjects completed the expertise measures, were debriefed, thanked and awarded course credit.

The Dependent Measures

The retrieval sets provided by subjects were compared across the first and second recall sessions. When a new brand appeared in the second set which was not present in the first set, the change was captured as an instance of recovery. Thus, recovery is not just the inclusion of a detail not previously recalled, but remembering a brand not recalled in the first session. The greater the value of the recovery variable, the more substantial the change in the membership of the retrieval set over usage contexts. New members of the retrieval set were designated as relevant or irrelevant to the particular context on the basis of the pretest described earlier.


The Retrieval Explanation

Hypothesis One. The first hypothesis posits that high knowledge consumers would be more likely to recover brands when the retrieval context varies than would low knowledge consumers. This hypothesis was supported. There was a main effect for knowledge level on recovery of previously forgotten brands. A one way ANOVA of knowledge level on recovery found a main effect for prior knowledge(F=18.40, p<.0001). The result holds in each of the conditions: in the no encoding context condition (low knowledge=.20, high knowledge=1.11), and in the encoding context condition (low knowledge=.14, high knowledge=.68). (See Table One).

Hypothesis Two. Hypothesis two asserted that the recovered brands would be relevant to the new context. The hypothesis was supported. Relevance was based on the pretest mentioned in the procedure section. High knowledge consumers recovered brands relevant to the new context on 24 of the 26 recovery occasions (92%). There were only 5 instances amongst low knowledge subjects. The sample was too small to draw reliable statistical inferences.

Summary of Evidence for the Retrieval Explanation. The first two hypotheses support the retrieval explanation for the ability to alter the content of recall for high knowledge consumers. The cues present in the form of usage contexts at retrieval affect the content of the retrieval set for the high knowledge consumer. The same can not be said about the low knowledge consumer.

Hypothesis Three

The Influence of at Context at Encoding. Offering a usage context at encoding was expected to influence the ability to alter recall given a different retrieval context. High knowledge consumers were hypothesized to be less able to recover brands at retrieval in this condition. The demonstration of a significant interaction implies that the context present during encoding influences the dynamic nature of the retrieval set. An analysis of recovery performance indicated that the interaction between knowledge level and condition (presence or absence of a context at encoding) is significant(F=10.31, p<.0001). High knowledge consumers were better able to alter their retrieval sets when they learn the information without a context (encoding context=.68, no encoding context=1.11).


The results of the study suggest that high knowledge consumers were able to alter their retrieval sets in the manner described in the dynamic set models. Low knowledge consumers were not able to recover brands. It is important to note that the changes in the retrieval set for high knowledge consumers, were relevant to the choice at hand. Hence, some of the advantage of the high knowledge consumer in choice may be explained by this ability to alter the retrieval set by recovering relevant brands.

The hypothesis of a positive correlation between the ability to recover brands when the purchase context varies and the ability to choose the most appropriate brand in a particular context is supported by an exploratory measure not reported in the results section. During both recall sessions, subjects were asked to indicate "which, if any, of the brands they had listed might consider purchasing". The measure was intended to investigate whether the recovered brands had a higher probability of membership in the consideration set than other brands. In total, subjects circled less than 20% of the listed brands. Of the recovered brands, 65% were circled by the subject as a brand which would be considered further. This suggests that the recovered brands were not only objectively relevant to the new context (pretest results indicate that they are appropriate given the purchase context), but that the brands were subjectively relevant (the subject indicates that the recovered brand is also a member of a subset of the retrieval set which will be further considered in choice).

It is also important to note that low knowledge consumers were not only less able to recover brands, but in statistic terms, their performance was not significantly different from zero. In each of the two conditions, the recovery performance of low knowledge consumers was not statistically significant: no encoding context (low knowledge consumers=.2, t=1.5, p>.16) and encoding context (low knowledge consumer=.14, t=1.8, p>.085).

Another interesting point is that while the provision of a usage context prior to encoding did not improve the probability of recovery for low knowledge consumers, it did result in an increase in the size of the retrieval set (no encoding context=3.1, encoding context=4.3, t=1.97, p<.059). It may be possible that the increase in the set size was a result of the providing a framework (in the form of a usage context) which facilitated processing (Johnson and Bransford 1972). The provision of a context at encoding did not have the same effect on the retrieval set for high knowledge consumers.

What insights can marketing practioners glean from these results? First, it appears that providing a context for encoding is beneficial to low knowledge consumers. In order to facilitate consumer learning, marketers might therefore consider ways of framing product information in terms of usage situations. Second, the confirmation of the retrieval hypothesis suggests that while low knowledge consumers have static retrieval sets which are not responsive to changes in usage contexts, high knowledge consumers have dynamic sets. The results show that for high knowledge consumers, the recall of brands on a purchase occasion is influenced by the cues at retrieval. Relative to low knowledge consumers, high knowledge consumers are able to alter the composition of their retrieval sets such that they are more relevant to the choice at hand.


The external validity of the empirical results presented here can be extended by the introduction of interfering information (relevant competing material). It is possible that the introduction of advertising clutter could differentially influence consumers of various knowledge levels. Interference may affect consumers of different knowledge levels differentially. Proactive interference may be more damaging to the formation of the retrieval set for low knowledge consumers, while retroactive interference should be more damaging for high knowledge consumers. Future work might investigate the extent to which interfering information influences the ability to alter membership in the retrieval set.

It would be valuable to investigate how the membership of the consideration set might be influenced when the choice occasion is not strictly memoryBbased, but incorporates brand information in the choice environment: a mixed choice task (see Lynch and Srull 1982).


Alba, Joseph W. (1983), "The Effects of Brand Knowledge on the Comprehension, Retention, and Evaluation of Brand Information," in Advances in Consumer Research, Vol. 10, eds., Richard P. Bagozzi and Alice M. Tybout, Ann Arbor, MI: Association for Consumer Research, 577B580.

Alba, Joseph W. and Amitava Chattopadhyay, "The Effects of Context and PartBCategory Cues on the Recall of Competing Brands," Journal of Marketing Research, 22: 340B349.

Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13: 411B454.

Anderson, John R. (1983), "A Spreading Activation Theory of Memory," Journal of Verbal Learning and Verbal Behavior, 22: 261B 295.

Anderson, Richard C. and James W. Pichert (1978), "Recall of previously unrecallable information following a shift in perspective," Journal of Verbal Learning and Verbal Behavior, 17: 1B12.

Beattie, Ann E. (1982), "Brand Expertise and Advertising Persuasiveness," in Advances in Consumer Research, Vol. 9, ed. Andrew A. Mitchell, Ann Arbor, MI: Association for Consumer Research, 336B341.

Bransford, John D. and Marcia K. Johnson (1972), "Contextual Prerequisites for Understanding: Some Investigations of Comprehension and Recall," Journal of Verbal Learning and Verbal Behavior, 11: 717B726.

Brucks, Merrie (1985), "The Effects of Brand Class Knowledge on Search Behavior," Journal of Consumer Research, 12: 1B16.

Chi, Michelene T. H., Paul J. Feltovich and Robert Glaser (1981), "Categorization and Representation of Physics Problems by Experts and Novices," Cognitive Science, 5: 121B152.

Chiesi, Harry L., Spilich, George J. and Voss, James F. (1979), "Acquisition of DomainBRelated Information in Relation to High and Low Domain Knowledge," Journal of Verbal Learning and Verbal Behavior, 18: 257B273.

Cowley, Elizabeth J. (1993), "Recovering Forgotten Information: A Study in Consumer Expertise," in Advances in Consumer Research, Vol. 21, eds. Chris T. Allen and Debra Roediger, Ann Arbor, MI: Association for Consumer Research, 336B341.

Fiske, Susan T., Kinder, Donald R. and Larter, Michael W. (1983), "The Novice and the Expert: KnowledgeBBased Strategies in Political Cognition," Journal of Experimental Social Psychology, 19: 381B400.

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

Hutchinson, J. Wesley (1983), "Expertise and the Structure of Free Recall," in Advances in Consumer Research, Vol. 10, eds., Richard P. Bagozzi and Alice M. Tybout, Ann Arbor, MI: Association for Consumer Research, 585B589.

Johnson, Eric J. and J. Edward Russo (1984), "Brand Familiarity and Learning New Information," Journal of Consumer Research, 11: 542B 550.

Larkin, Jill H., John McDermott, Dorthea P. Simon and Herbert A. Simon (1980), "Models of Competence in Solving Physics Problems," Cognitive Science, 4: 317B345.

Lynch, John G. Jr. and Thomas K. Srull (1982), "Memory and Attentional Factors in Consumer Choice: Concepts and Research Methods," Journal of Consumer Research, 9: 18B37.

Maheswaran, Durairaj and Brian Sternthal (1990), "The Effects of Knowledge, Motivation and Type of Message on ad Processing and Brand Judgements," Journal of Consumer Research, 17: 66B73.

Mitchell, Andrew A., Peter F. Dacin and Michelene T. H. Chi (1993), "Differences by Expertise in the Content and Organization of Knowledge for a Brand Class," Working Paper, University of Toronto.

Nedungadi, Prakash (1990), "Recall and Consideration Sets: Influencing Choice without Altering Brand Evaluations," Journal of Consumer Research, 17: 263B276.

Pichert, James W. and Richard C. Anderson (1977), "Taking Different Perspectives on a Story," Journal of Educational Psychology, 69: 309B315.

Spilich, George J., Gregg T. Vesonder, Harry L. Chiesi, and James F. Voss (1979), "Text Processing of Domain Related Information for Individuals with High and Low Domain Knowledge," Journal of Verbal Learning and Verbal Behavior, 18: 275B290.

Srull, Thomas K.(1983), "The Role of Prior Knowledge in the Acquisition, Retention, and Use of New Information," in Advances in Consumer Research, Vol. 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Arbor, MI: Association for Consumer Research, 572B576.

Sujan, Mita (1985), "Consumer Knowledge: Effects on Evaluation Strategies Mediating Consumer Judgements," Journal of Consumer Research, 12: 31B46.

Voss, James F., Jeffrey Blais, Mary L. Means, Terry Greene and Ellen Ahwesh (1986), "Informal Reasoning and Subject Matter Knowledge in the Solving of Economics Problems by Naive and Novice Individuals," Cognition and Instruction, 3(4): 269B302.

Walker, Beth, Richard Celsi and Jerry Olson (1987), "Exploring the Structural Characteristics of Consumers' Knowledge," in Advances in Consumer Research, Vol. 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT: Association for Consumer Research, 17B21.



Elizabeth J. Cowley, University of Toronto
Sharmistha Law, University of Toronto


NA - Advances in Consumer Research Volume 22 | 1995

Share Proceeding

Featured papers

See More


The Asymmetry between Time and Money Compensation effect when feeling Scarcity: Time helps the Money Poor, but Money doesn’t help the Time Poor

Jane So, University of Washington, USA
Nidhi Agrawal, University of Washington, USA

Read More


Social Sharing of Negative Emotions in Virtual Travel Communities

Clara Koetz, Rennes School of Business
Anke Piepenbrink, Rennes School of Business

Read More


P6. Marginal Cost Consideration

Ethan Pew, Stony Brook University
Hyunhwan Lee, University of Miami, 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.