Models of Memory: Implications For Measuring Knowledge Structures

Andrew A. Mitchell, Carnegie-Mellon University
ABSTRACT - Although there is considerable research interest in understanding how knowledge affects information processing activities, we must first develop valid measures of knowledge. It is argued that a set of different hypothetical constructs will be required to measure knowledge within a domain and that different sets of hypothetical constructs may be required to explain different information processing activities. Finally, different methodological approaches and procedures for measure knowledge structures and the relationship between these structures and information processing activities are discussed.
[ to cite ]:
Andrew A. Mitchell (1982) ,"Models of Memory: Implications For Measuring Knowledge Structures", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 45-51.

Advances in Consumer Research Volume 9, 1982      Pages 45-51

MODELS OF MEMORY: IMPLICATIONS FOR MEASURING KNOWLEDGE STRUCTURES

Andrew A. Mitchell, Carnegie-Mellon University

[This research was partially funded by a grant from the Department of Transportation. The author thanks Eric Johnson for his comments on an earlier version of the paper.]

ABSTRACT -

Although there is considerable research interest in understanding how knowledge affects information processing activities, we must first develop valid measures of knowledge. It is argued that a set of different hypothetical constructs will be required to measure knowledge within a domain and that different sets of hypothetical constructs may be required to explain different information processing activities. Finally, different methodological approaches and procedures for measure knowledge structures and the relationship between these structures and information processing activities are discussed.

INTRODUCTION

It is generally acknowledged that consumers' knowledge about a product category will affect their purchase behavior; however, it is only recently that we have attempted to measure consumers' knowledge structures (e.g., Olson and Muderrisoglu, 1980) and examine how differences in knowledge affect information processing activities (e.g., Edell and Mitchell, 1978; Johnson and Russo, 1981; Bettman and Park, 1980). One major reason for the increase in interest in these areas in consumer behavior has been the rapid advances made in cognitive psychology in both theory and methodology in the study of memory.

Initial indications are that research on consumers knowledge structures for product related information will become an important area of research in consumer behavior. There are a number of critical issues that still need to be examined that have both theoretical and applied importance. For instance, "How do consumers organize information about products and brands in long term memory?" "How does knowledge affect information search strategies?" "How does product class knowledge affect the evaluation of new products from that product class?" and "How does product class and brand knowledge affect the processing of information from advertisements?" There is a great danger, however, that we may never learn the answer to these questions if we are not careful about the validity of our procedures and measures of knowledge structures.

Currently a number of different studies have used the hypothetical construct "product familiarity," as a measure of a consumer's knowledge about a product category (e.g., Bettman and Park, 1980; Johnson and Russo, 1981). At least three different measures of this construct have appeared in the literature. These include the number of purchases within the product category (e.g., Park, 1978), self-report measure of familiarity (e.g., Bettman and Park, 1980), and self-report measures of relative familiarity (e.g., Johnson and Russo, 1981). Little attempt, however, has been made to validate these measures or to even determine if the different measures are even measuring the same construct. Conceptually, these may be poor measures of a consumer's knowledge about a product category. For instance, some consumers may purchase an item numerous times without knowing very much about the product, while other consumers may know a lot about a product without ever purchasing it. Some consumers, for instance, may know less today about automobiles after owning a number of them than a teenager that has yet to purchase his first automobile. Al so, consumers may not be able to accurately access their knowledge about a product category. Some consumers may think that they know a lot about stereos, but next to an electrical engineer they probably know very little.

The issue, then, is how to develop valid measures of knowledge structures. It will be argued that in order to develop valid measures, you must first have a theory of memory (e.g., how knowledge within the domain of interest is organized and the retrieval mechanism operating on this knowledge) and a conceptual definition of the hypothetical construct you are attempting to measure. This argument is, of course, circular because in order to have a theory of memory you need valid measures to test this theory; however, without a theory you cannot have valid measures. To get around this problem, we will argue for the use of a number of different procedures and measures in hopes that although each will be flawed, in unison that will all point in a similar direction.

A second point that will be made in this paper is that since product class knowledge includes a diverse set of information, a single construct, such as product familiarity, probably will not provide an adequate representation of this knowledge. For instance, one consumer may have knowledge about many different brands in a product category but may know little about the technical aspects Of the product. The opposite may be true for another consumer. By using only a single measure of knowledge these subtle differences, which may have a major effect on sane information processing activities, will be lost. Therefore, we will need to develop a number of different hypothetical constructs to adequately represent this knowledge. Examples of different sets of hypothetical constructs are those proposed by Kanwar, Olson and Sims (1981) and Brucks and Mitchell (1981).

A third, and related point, is that the hypothetical constructs that are developed should depend on the purpose of the research. For instance, although a self-report measure of knowledge may be a poor measure of the actual knowledge that a consumer has relative to other consumers, it may be a better predictor of information search behavior than a measure of actual product knowledge.

In general, then, what is required is the development of a number of hypothetical constructs that will provide measures of a consumer's knowledge structure and procedures for measuring these hypothetical constructs. As mentioned above, the development of these constructs should depend on the type of information processing activates that are to be explained. In other words, in developing these hypothetical constructs, some thought should be directed at how these measures of knowledge will effect information processing activities. After measurement procedures for the sets of hypothetical constructs have been developed and validated, the next step is to examine the linkages between the measures and various information processing activities.

The purpose of this paper is to discuss alternative procedures for measuring knowledge structures and for examining the relationship between knowledge and different information processing activities. In the next section, theories of memory are discussed. Then alternative methodological procedures for examining the relationship between knowledge and different information processing activities are reviewed. Finally, alternative procedures for measuring knowledge structures are discussed

THEORY

The most commonly accepted model of memory is an associative network model (e.g., Wickelgren, 1981). Within this model, the nodes of the network represent concepts while the arcs are linkages between concepts. The arcs may vary in strength, so that some concepts are more strongly associated. In propositional models, the arcs also define the relationship between concepts (Figure 1).

FIGURE 1

NETWORK MODEL OF MEMORY

It is currently believed that our knowledge is organized into packets of information. A number of different types Of information packets have been hypothesized to exist in memory. -These include schemata (e.g., Rumelhart and Ortony, 1977), scripts (Schank and Abelson, 1977; Abelson, 1981), frames (Minsky, 1975). prototypes and exemplars (Rosch, 1977), and MOP's (memory organizational packets), and TOP's (thematic organization points) (Schank, 1980, 1981).

Schemas are packets of information centered around concepts. Consumers, for instance, probably have schemas for specific brands such as Chevrolet automobiles. These schemas contain organized information about Chevrolets; however,they may not contain all the information that the individual has about Chevrolets. Scripts contain our generalized knowledge about specific events such as visiting an automobile dealer or fixing a flat tire. Within these scripts the temporal sequence of actions which occur within an event are stored. Frames are structures about classes of objects such as automobiles. Within a particular frame are "slots" which are used in understanding specific instances of a particular class of objects. For instance, an automobile frame may contain" slots" for the general shape and the appearance of the front seats of particular automobile.

A prototype is an abstraction of a particular class of objects while an exemplar is a specific instance that is considered representative of a set of objects. For instance, an individual may have generalized information about a particular class of automobiles such as small fuel-efficient automobiles or that individual may consider the Volkswagen Rabbit to be the best representation of that class of objects. Finally, MOP's are similar to scripts in that they contain generalized knowledge about specific events; however, they also contain linkages to specific events. For instance, we may have an eating at a restaurant MOP that may contain links to specific events that occurred while eating at restaurants - events such as an especially good meal, an especially poor meal or the time you forgot your wallet. According to Schank, specific events become linked to MOP's when the events do not meet the expectations formed from the MOP. For instance, the poor meal may have been linked to the MOP because the restaurant may have been expensive and you had expected a very good meal. Thematic Organization Points (TOP's) represent a higher order organization of events. For instance, individuals may have higher order structures for certain types of plots in plays (e.g., boy meets girl, boy loses girl, boy wins back girl) and these structures are used to identify instances of a general theme

These different packets of information guide our interpretation and processing of information from the environment. Consequently, individuals with different packets of information may interpret new information differently. For instance, someone with a packet of information about microwave ovens will interpret information about a new microwave oven differently than someone that doesn't have this packet of information.

These ideas also suggest that the distinction between episodic and semantic memory by Tulving (1972) may not be as useful as it once was thought. Originally, this distinction was made to differentiate the results of simple verbal learning experiments from research examining generalized knowledge structures. Although this distinction is probably still important we should not necessarily think of episodic and semantic memory as two different systems. In general, I believe that Schank's notion that specific episodes are linked to generalized knowledge is correct. For instance, we may have a generalized knowledge structure for beach parties, and linked to this structure there will be information about specific beach parties. At the same time, I believe that we probably have a memory for specific events that is recorded temporally. For instance, I. may not be able to directly retrieve what I had for lunch yesterday; however, by reconstructing the events that happened yesterday in temporal sequence, I may be able to retrieve that information.

It is currently believed that the retrieval of information involves an activation process (Collins and Loftus, 1976). Within a network model, specific nodes of the network are activated and the activation then spreads to linked nodes. If the amount of activation that spreads to a particular node exceeds some threshold level, then the node will become activated. Which nodes become activated depends on the strength of the association between the activated node and the linked nodes. Those nodes that are the most strongly associated are the most likely to be activated. For instance, in the ACT model developed by Anderson (1976), the probability of a particular linked node becoming activated is given by the strength of that association divided by the sum of all the strengths of the links to the activated node. In addition, recent research has indicated that information may be structured by topic and individuals can limit the spread of activation to a particular topic (Smith, 1981). For instance, a consumer may have knowledge about the appearance and performance of a particular automobile. These two types of information may be subdivided in the knowledge structure for this particular automobile (Figure 2). When this structure is activated, the individual has control over whether the appearance or the performance portion of the structure is activated.

FIGURE 2

SUBDIVIDED MEMORY STRUCTURE

This conceptualization of memory may be used to interpret the results of the Olson and Mudderisoglu (1980) study. In this study, memory probes were used for different product categories (e.g., toothpaste) and brands within the product category. The same probes were used on two occasions the week apart. After the subjects were given the probe, they were asked to repeat whatever came to mind. According to the theory of memory outlined above, each probe activated a node in memory (e.g., the toothpaste node) and activation then spread to connected nodes. This spreading activation directed attention to the connected nodes and the subject then gave the experimentor the contents of these activated nodes.

The results of the study indicated that, on average, six concepts were elicited per memory probe and the reliability of the concepts elicited on the two occasions was around .5. The number of concepts elicited with each probe is consistent with the results of studies in psychology where subjects are asked to give as many instances of a particular category (e.g., Graesser and Mandler, 1978). For this task, subjects typically give up to six instances per probe. Graesser and Mandler (1978) suggest that this may be due to short-term memory limitations at either encoding or retrieval.

The reliability of concepts elicited on each occasion may 0 be given a number of different interpretations. First, it ! might be taken as an indication that the subjects did not I have tightly-knit organized structures of knowledge. If they did, there would be a strong association between concepts in the structure which would result in the same A probe eliciting the same concepts at the two points in time. Note that this interpretation would also imply that only six concepts would be linked to any given knowledge structure. An alternative explanation is that a large number of concepts may be linked together in a knowledge structure and retrieval at any given time is simply a probabilistic process. Therefore, different concepts would be elicited at two different points in time. Under this g latter interpretation, when the same memory probe is used a number of different times, the frequency with which a particular concept is elicited might be used as a measure of the strength of the association between the probe and the concept (e.g., Chi and Koeske, 1981).

Finally, it should be realized that much of our current thinking about memory is based on a computer metaphor (Roediger, 1980). The use of this metaphor leads us to think of long-term memory as a giant filing cabinet. Within this cabinet, information is stored at different locations, related information is stored together, and information retrieval essentially involves searching different locations. An alternative metaphor is the tuning fork metaphor. With this metaphor information on a particular topic may be distributed at different locations in memory and retrieval essentially involves a parallel search through memory. A probe is sent out (e.g., a particular note) and all the portions of memory that respond to that probe are activated. In other words, different parts of memory behave like a tuning fork which becomes activated by a particular probe. According to Anderson and Hinton (1981), this type of model corresponds more closely to what we know about the physiology of the brain. The different portions of memory may be thought of as different neurons that fire given a certain type of probe. Although this metaphor provides an interesting alternative to the computer metaphor, it is not clear at this point how our procedures for measuring knowledge structures would differ if this metaphor was more appropriate.

Characteristics of Memory

Two different characteristics of memory need to be taken into account in measuring knowledge structures. First, most cognitive models suggest that the organization of information in memory is primarily the result of active control processes (e.g., Hastie, 1980; Srull, 1981). Two individuals, for instance, may experience roughly the same events, yet organize the resulting information from these experiences very differently. One individual may do considerable thinking about these experiences and consequently have a number of generalizations from these experiences stored in memory. Another individual may do little active thinking about these experiences and consequently will have little generalized knowledge. We are currently developing a model based on these ideas (Smith, Mitchell, and Meyer, 1982). Consequently, if the researcher is interested in obtaining the measures of what experiences are stored in memory, care will need to be taken to insure that the measurement procedure will tap this information.

Second, context seems to have a strong effect on information retrieval. These context effects are illustrated in experiments by Thompson and Tulving (1973), Anderson and Pritchard (1972), and Bower (1981). In Thompson and Tulving experiments, it was found that the probability of recognizing whether a particular word was learned depended on the similarity of the semantic context in the learning and recognition task. a e learning of the word "bank" in the context of "cash" affected the recognition of "bank" in the context of "river." Anderson and Pritchard (1978) found that different recall probes caused subjects to recall different information about a passage that they had read. Finally, in the Bower experiments, it was shown that if the subjects were in the same mood (e.g, happy) during learning and retrieval, they could recall more information than if they were in different moods. Hood also affected the type of information recalled. SubJects in a happy mood tended to recall primarily happy experiences while subjects in a and mood tended to recall and experiences. These types of context effects can be explained with a network model by assuming that the different contexts activate different nodes. For instance, an individual's node for a particular automobile may be activated at the repair shop and at a cocktail party. In each case the schema for the particular automobile is activated, but the two contexts also activated different nodes. This may result in different portions of memory becoming activated and the recall of different information about the automobile. Consequently, if the researcher is interested in examining the relationship between knowledge and information processing activities, he or she will need to insure that the context for the measurement procedure and the information processing task are as similar as possible.

METHODOLOGY

In this section, we will discuss general research strategies for examining the structure of consumers memory for product information and the affect of knowledge on consumer decision strategies. In addition, we will discuss specific procedures for measuring memory.

General Research Strategies

There are six general strategies that may be used to examine the research issues discussed here. The first approach involves identifying individuals that differ in knowledge within a particular domain and then examining how their performance differs on specific tasks. This has been the general approach used in consumer behavior (e.g., Bettman and Pack, 1980; Johnson and Russo (1981) and cognitive and social psychology (e.g., Markus, 1977). For instance, in a series of studies by Splich, Vesonder, Chiesi and Voss (1979), subjects with differing levels of knowledge about baseball were given a description of a baseball game to read and then asked to recall as much as they could about the specific game. The results indicated that subjects in the high knowledge condition could recall more information, organized it in larger chunks and made more elaborations. Although this procedure allows for the examination of the effect of knowledge on information processing activities, it does not provide an understanding of how specific aspects of this knowledge affect these activities. In order to determine this, it would be necessary to obtain measures of the different hypothetical constructs. The subjects would then be given one or more tasks and statistical techniques would be used to examine the relationship between the measures of knowledge structure and their behavior in the task. This is the second approach. With this approach, it is necessary to use a heterogeneous set of subjects with respect to the knowledge measures to create enough variance on the independent measures and to use tasks where product class knowledge would be the primary determinant of the behavior in the task. The problem with this approach is that the resulting measures of knowledge may be somewhat highly correlated so it may be difficult to disentangle the effects of the different types of knowledge.

A third approach is to create different types of knowledge structures in the laboratory. Here subjects with little knowledge about the product category would be given different types of knowledge. This procedure allows for the development of knowledge structures that are reasonable orthogonal in terms of the measures to be used. This was the approach used by Edell and Mitchell (1978). Here, one group of subjects was given information about silver polish, and their performance in processing information about different brands of silver polish was compared to a group that had little knowledge about silver polish. The danger with this procedure is that the resulting knowledge structures built up in the laboratory may be quite different from those that develop naturally. These differences may then be reflected in the subjects' performance on a task.

Fourth, subjects might be given a task involving new brands within a product class (e.g., learning information about the brands or selecting a particular brand to purchase) and then recall measures might be used to measure what information is recalled and how it is organized (e.g., Biehal and Chakravarti, 1981). Here the results obtained are likely to be highly dependent on the task used. Individuals generally w e different information processing strategies in different tasks and what is recalled and how it is recalled may be highly dependent on the information processing strategy used. Consequently, different tasks will probably cause differences in the organization and structure of information in memory, and we may not be able to generalize across tasks. Also, it is not clear that the resulting structure from a single artificial task will be similar to the memory structures that occur naturally from information acquisition from many different sources (e.g., advertising, neighbor, etc.).

A fifth procedure that might be used is an interference paradigm. Here, subjects might be given information about two types of automobiles (e.g., small compact car and a four wheel drive station wagon) and then a recall task. If subjects tended to recall information about one type of automobile when given a memory probe about the second, this might be a indication that the information about the two automobiles is stored in similar locations. A similar procedure was used by Gunter, Clifford and Berry (1980) to examine whether individuals stored information about different types of television news programs in similar locations.

The final approach is to examine how subjects execute a series of different tasks (e.g., evaluate new products from advertisements) and then try to infer what types of knowledge they need to have in order to execute the task. This procedure has been quite successful in building models of language comprehension. Here researchers examine conversations and determine what type of knowledge is required in order to understand a conversation. This particular approach has suggested the different types of information packets that were discussed earlier.

Measurement Procedures

There are number of different measurement procedures that may be used for measuring knowledge structures. The first procedure is an elicitation procedure (e.g., Olson and Muderrisoglu, 1980). Here the researcher uses a memory probe (e.g., Coca Cola) and the subJect mentions everything that comes to mind. This approach is based on the associative network model of memory. By giving a specific probe, the researcher activates a particular node in memory and the subject then gives you all the concepts that are linked to that node.

If the researcher is interested only in what information is stored in memory, he or she must be careful that the subject provides only the information that is activated from the memory probe and does not attempt to construct new information. In order to insure that this occurs, the subject should be kept in a passive state so they are not actively constructing information. In general, if stored information is elicited, the information will probably be retrieved in chunks with around five chunks of information elicited with each probe. For instance, as discussed earlier, Graesser and Mandler (1978) found that subjects would elicit three to five members of a category (e.g., animal) at a time and Olson and Muderrisoglu (1980) found that subjects elicited around five or six concepts with product category or brand probes. If the subject starts speaking in complete sentences, then they may be constructing information instead of giving stored information. [In some situations, when the information is well rehearsed, it may be stored in complete sentences. In these situations, this criteria is not applicable.] The major problem with this approach is that we do not really know exactly what probes to use in order to determine a subject's knowledge in a particular domain. If we knew how subjects organized product information, then we would know which probes to use. Unfortunately, we don't.

A second procedure is to give subjects a particular task that requires retrieval of information from long-term memory (e.g. , Russo and Johnson, 1980). These might range from a simple "tell me everything you know about coffeemakers" to "what would you tell someone who just moved into your community about the different supermarketS in your neighborhood?" This type of task will produce mostly constructive processes as opposed to tapping actual stored information. Different tasks, and the subject's interpretation of the task, may produce different information. In addition, probably no single task will yield all of a subject's knowledge within a specific domain

The third procedure is to use a questionnaire to obtain measures of memory. Scott (1980), for instance, has developed a series of questions to obtain different measures of knowledge structures within a particular domain. Examples of these questions are categorizing objects within a domain and listing the attributes of objects within that domain. Responses to these questions are then used to provide measures of the organization of the information within the domain.

Finally, response times may be used to measure knowledge structures. The problem here is that most response time measures require a theory of memory in order to be interpretable (Gardner, Mitchell and Russo, 1977, Johnson and Russo, 1977). Currently, our understanding of how individuals store information within a domain does not appear to be complete enough to use this approach. We may, however, be able to be used it in a more basic way to understand what information individuals have associated with a particular concept. For instance, a particular node in memory might be primed first and then the subject would be asked a question. If the information required to answer the question was linked to the primed node, subjects should be able to respond quicker than if the information was not linked. This procedure is similar to the procedure used by McKoon and Ratcliff (1980) to examine knowledge structures created in text comprehension.

SUMMARY

In this paper, I have discussed current theories of memory and discussed alternative procedures for measuring knowledge structures and for examining the relationship between knowledge and different information processing activities. It was argued that a number different hypothetical constructs will be required to obtain measures of knowledge within a domain and an understanding of how these hypothetical constructs may affect information processing activities. In addition, it is argued that the measures that are developed should depend on the type of information processing activity that we want to explain. For instance, one set of measures of knowledge might be more appropriate for explaining the generation of counterarguments and support arguments during exposure to an advertisement and another set of explaining information search.

Finally, it was argued that all the procedures for measuring knowledge structures depend on a theory of memory. For instance, elicitation procedures require an understanding of what schemata exist within a particular domain. Unfortunately, we do not currently know this, so if elicitation procedures are used to tap knowledge within a particular domain, only subsets of this knowledge may be tapped if particular schemata are not probed. It is hoped that different procedures may provide converging evidence as to how consumers organize knowledge about products and services and how this knowledge affects different information processing activities.

REFERENCES

Abelson, R. P. (1981) "Psychological Status of the Script Concept." American Psychologist, 36, 715-729.

Anderson, J. A. and G. E. Hinton (1981) "Models of Information Processing in the Brain." In G. E. Hinton and J. A. Anderson (eds.), Parallel Models of Associative Memory. Hillsdale, NJ: Lawrence Erlbaum Associates.

Anderson, J. R. (1976) Language, Memory and Thought. Hillsdale, NJ: Lawrence Erlbaum Associates.

Anderson, R. C. and J. W. Pritchard (1978) "Recall of Previously Unrecallable Information Following a Shift in Perspective," Journal of Verbal Learning and Verbal Behavior, 17, 1-12.

Bettman, J. R. and C. W. Park (1980) "Effects of Prior Knowledge, Exposure, and Phase of the Choice Process on Consumer Decision Processes: A Protocol Analysis." Journal of Consumer Research, 7, 234-248.

Biehal, G. and D. Chakravarti (1982) "Information Presentation Format and Task Goals as Determinants of Consumers' Memory Retrieval and Choice Processes." In A. Mitchell (ed.), Advances in Consumer Research, V. 9, Ann Arbor: Association for Consumer Research.

Bower, G. H. (1981) "Mood and Memory." American Psychologist, 36, 129-148.

Brucks, M. and A. A. Mitchell (1981) "Knowledge Structures, Production Systems, and Decision Strategies." In K. Monroe (ed.), Advances in Consumer Research, Vol. 8, Ann Arbor: Association for Consumer Research, 750-765.

Chi, M. T. H. and R. D. Koeske ( 1981) "Recall and Network Representation of a Child's Dinosaur Knowledge." Working Paper, Learning Research and Development Center. University of Pittsburgh. PA.

Collins, A. A. and E. F. Loftus (1975) "A Spreading Activation Theory of Semantic Processing." Psychological Review, 82, 407-428.

Edell, J. and A. A. Mitchell (1978) "An Information Processing Approach to Cognitive Responses." In S. C. Jain (ed.), Research Frontiers in Marketing: Dialogues and Directions, Chicago: American Marketing Association, 178-183.

Gardner, M., A. A. Mitchell and J. E. Russo (1 978) "Chronometric Analysis: An Introduction and an Application to Low Involvement Perception of Advertisements." In K. H. Hunt, (ed.), Advances in Consumer Research, Vol. 5, Ann Arbor: Association for Consumer Research, 581-589.

Graesser A. and G. Mandler ( 1978) "Limited Processing Capacity Constrains the Storage of Unrelated Sets of Words and Retrieval from Natural Categories," Journal of Experimental Psychology: Human Learning and Memory, 4, 86-100.

Gunter, B.9 B. R. Clifford and C. Berry (1980) "Release from Proactive Interference with Television News Items: Evidence for Encoding Dimensions Within Television News," Journal of Experimental Psychology: Human Learning and Memory, 6, 216-223.

Hastie, R. (1980) "Memory for Information Which Confirms or Contradicts a General Impression." In R. Hastie, et. al., eds., Person Memory: the Cognitive Basis of Social Perception. Hillsdale, NJ: Lawrence Erlbaum Associates.

Hinton, G. E. and J. A. Anderson (1981) Parallel Models of Associative Memory. Hillsdale, NJ: Lawrence Erlbaum Associates.

Johnson, E. J. and J. E. Russo (1981) "Product Familiarity and Learning New Information." In K. Monroe (ed.), Advances in Consumers Research, Vol. 8, Ann Arbor: Association for Consumer Research, 151-155.

Johnson, E. J. and J. E. Russo (1977) "The Organization of Product Information in Memory Identified by Recall Times." In K. H. Hunt, ed., Advances in Consumer Research, Vol. 5. Ann Arbor Association for Consumer Research. 79-86.

Kanwar, R., J. C. Olson, and Laura S. Sims (1981) "Toward Conceptualizing and Measuring Cognitive Structures." In K. Monroe (ed.), Advances in Consumer Research, Vol. 8, Ann Arbor, MI Association for Consumer Research.

Kintsch, W. ( 1974) The Representation of Meaning in Memory. Hillsdale, NJ: Lawrence Erlbaum Associates.

Kintsch, W. (1981) "Semantic Memory: A Tutorial." In R. S. Nickerson (ed.), Attention and Performance VIII, Hillsdale, NJ: Lawrence Erlbaum Associates.

Markus, H. (1977) "Self-schemata and Processing Information About the Self," Journal of Personality and Social Psychology, 35, 63-78.

McKoon, G. and R. Ratcliff (1980) "Priming in Item Recognition: The Organization of Propositions in Memory for Text." Journal of Verbal Learning and Verbal Behavior, 19, 369-386.

Minsky, M. A. (1975) WA Framework for Representing Knowledge." In P. H. Winston ( ed . ), The Psychology of Computer Vision. New York: McGraw-Hill.

Mitchell, A. A. (1978) "A Information Processing View of Consumer Behavior." In S. C. Jain (Ed.), Research Frontiers in Marketing: Dialogues and Directions: Chicago: American Marketing Association, 188-197.

Olson, J. C. and A. Muderrisoglu (1979) "The Stability of Responses Obtained by Free Elicitation: Implications for Measuring Attribute Salience." In W. Wilkie (ed.) Advances in Consumer Research, Vol. 6, Ann Arbor: Association for Consumer Research. 269-275.

Roediger, H. L. (1980) "Memory Metaphors in Cognitive Psychology." Memory and Cognition, 8, 231-246.

Rosch, E. (1978) "Principles of Categorization." In E. Rosch and B. B Lloyd (eds.), Cognition and Categorization. Hillsdale, NJ: Lawrence Erlbaum Associates.

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

Rumelhart, D. E. and A. Ortony (1977) "The Representation of Knowledge in Memory." In R. C. Anderson, R. J. Spiro, and W. E. Montague (eds.), Schooling and the Acquisition of Knowledge. Hillsdale, NJ: Lawrence Erlbaum Associates.

Schank, R. C. (1980) "Language and Memory." Cognitive Science, 4, 243-284.

Schank, R. C. (1981) Dynamic Memory: A Theory of Learning in Computers and People.

Schank, R. C. and R. P. Abelson (1977) Scripts, Plans, Goals, and Understanding. Hillsdale, NJ: Lawrence Erlbaum Associates.

Scott, W. A. (1969) "Structures of Natural Cognitions." Journal of Personality and Social Psychology, 12, 261-278.

Smith, T. R., A. A. Mitchell and R. Meyer (1982) "A Computational Process Model of Evaluation Based on the Cognitive Structuring of Episodic Knowledge." In A. Mitchell, (ed.) Advances in Consumer Research, Vol. 9, Ann Arbor: Association for Consumer Research.

Smith, E. E. (1981) "Organization of Factual Knowledge." In J. H. Flowers ted.), Nebraska Symposium on Motivation: 1980. Lincoln, NE: University of Nebraska Press.

Spilich, G. J., G. T. Vesonder, H. L. Chiesi, and J. F. Voss (1979) "Text Processing of Domain Related Information for Individuals with High and Low Domain Knowledge," Journal of Verbal Learning and Verb Behavior, 18, 275-290.

Srull, T. K. (1981) "Person Memory: Some Tests of Associative Storage and Retrieval Models ," Journal of Experimental Psychology: Human Learning and Memory, 7, 440-463.

Tulving, E. and D. M. Thompson (1973) "Encoding Specificity and Retrieval Processes in Episodic Memory," Psychological Review, 80, 352-372.

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