Concept Mapping in Marketing: a Research Tool For Uncovering Consumers’ Knowledge Structure Associations

Christopher Joiner, Kansas State University
ABSTRACT - The associations that exist in consumers’ marketing-related cognitive structures have received a great deal of attention across a wide range of consumer research domains. This paper reviews the concept of knowledge structure associations. A qualitative methodology, concept mapping, is then introduced and the results of two studies which demonstrates the usefulness of concept maps in marketing research are presented. The results suggest that concept maps can reliably access individuals’ salient associations for a variety of marketing objects and reflect differences in the knowledge structures of consumers with varying levels of knowledge.
[ to cite ]:
Christopher Joiner (1998) ,"Concept Mapping in Marketing: a Research Tool For Uncovering Consumers’ Knowledge Structure Associations", in NA - Advances in Consumer Research Volume 25, eds. Joseph W. Alba & J. Wesley Hutchinson, Provo, UT : Association for Consumer Research, Pages: 311-322.

Advances in Consumer Research Volume 25, 1998      Pages 311-322

CONCEPT MAPPING IN MARKETING: A RESEARCH TOOL FOR UNCOVERING CONSUMERS’ KNOWLEDGE STRUCTURE ASSOCIATIONS

Christopher Joiner, Kansas State University

ABSTRACT -

The associations that exist in consumers’ marketing-related cognitive structures have received a great deal of attention across a wide range of consumer research domains. This paper reviews the concept of knowledge structure associations. A qualitative methodology, concept mapping, is then introduced and the results of two studies which demonstrates the usefulness of concept maps in marketing research are presented. The results suggest that concept maps can reliably access individuals’ salient associations for a variety of marketing objects and reflect differences in the knowledge structures of consumers with varying levels of knowledge.

INTRODUCTION

The associations in individuals’ cognitive structures have recently received a great deal of attention across a range of consumer research domains. This paper reviews the concept of associations in consumer’s knowledge structures for marketing-related objects. A qualitative methodology, concept mapping, which can be useful for accessing consumers’ associations, is then introduced. Finally, the results of two studies demonstrating the use of concept maps in marketing research are presented.

Cognitive structure is a hypothetical construct referring to the organization of concepts, or associations, in memory. Most widely accepted conceptualizations of cognitive structure involve some type of associative model and spreading activation metaphor (e.g., Anderson 1983; Collins & Loftus 1975). In these models, activation of a piece of information in long term memory spreads to other pieces of information structurally linked to the first, increasing the probability that the second will also come into awareness. Structures in memory influence a wide range of information processing activities, including evaluations, judgments, and inferences (Markus & Zajonc 1985).

Consistent with these accounts, consumers’ knowledge structures (e.g., brand, product, and/or usage situation knowledge) are conceptualized as nodes in memory to which a variety of associations are linked (e.g., Alba, Hutchinson & Lynch 1991; Keller 1993). Keller (1993), for example, defines brand associations as information nodes linked to a brand node in memory, which contain the meaning of the brand for consumers. Marketing practitioners invest heavily in establishing and reinforcing these associations (Broniarczyk & Alba 1994).

Researchers have cataloged a variety of associations that may be linked to the brand in memory, including many non-product-related associations. Although much of the focus in recent marketing literature has been on the cognitive structures and associations of brands, the associations linked to product categories, usage situations, and other marketing-related concepts are important as well.

Consumers’ associations are central to a wide range of consumer behavior. For example, they play an important role in both the evocation and evaluation of alternatives in memory based choice (Alba et al. 1991). To consumers, brand equity is a function of a set of associations for a brand name (cf. Keller 1993). Studies of brand extensions seek to understand how brand associations are transferred from parent brands to extensions, as well as how these associations might influence judgments of extension fit. Specific associations affect a brand’s ability to stretch beyond its initial product (e.g., Broniarczyk & Alba 1994).

ACCESSING ASSOCIATIONS AND KNOWLEDGE STRUCTURES

Given the importance of marketing-related knowledge structures and associations, it is desirable to develop methodologies that can access them. Measuring perceptions of brand image has typically relied on a variety of forced-choice instruments (e.g., traditional approaches to perceptual maping). One of the major limitations with forced-choice measures is that the items provided by the researcher actually structure the information elicited from individuals (Steenkamp, Van Trijp & Ten Berge 1994). For this reason, many researchers have recommended the use of more open-ended, qualitative ways of measuring perceptions of brands (e.g., Park & Srinivasan 1994). An open-ended task may provide better clues to the type of associations people use in responding to brands (Russo & Johnson 1980). It also allows more information to be obtained from those consumers who have more elaborate knowledge structures for the object in question. For example, Olson and his colleagues (Kanwar, Olson & Sims 1981; Olson & Muderrisoglu 1979) developed a free elicitation procedure for identifying salient product attributes and for indicating the content and organization of consumers’ product-knowledge structures. They used probe cues of varying degrees of specificity (e.g., a product category name, a brand name) to trigger or activate a particular cognitive structure in semantic memory. Respondents were asked to state "all the things that come to mind" in response to each probe. In the next section, a concept mapping procedure, similar to the Olson elicitation procedure, but which allows respondents to explicitly link related ideas, is introduced.

CONCEPT MAPPING

Concept maps (CMs) are a tool being widely used in educational and counseling research into cognitive structure (e.g., Novak & Gowin 1984). In the CM methodology, key ideas are represented as nodes that are linked by pathways or relationships. All CM methods consist of respondents putting concept labels on a page and linking them, if appropriate, with lines to show a relationship (Stuart 1985). The technique assumes that information is displayed in such a way that its arrangement on the page reveals something about the relationships inherent in the information represented in a person’s mind. The conceptual basis for the CM methodology is that of activation theory (Collins & Loftus 1975) and the associative network model of memory. It is assumed that, once a structure is activated, a person should be able to report the majority of the structure content.

Research Using Concept Maps. Stuart (1985) examined two assumptions underlying the use of CMs. The first was that CMs are equivalent to some part of an individual’s thinking processes. Although Stuart acknowledged that it may not be reasonable to assume that a CM is a structural representation of how concepts are stored in the mind, she concluded that a CM probably reveals part of the "thinking process" about a topic, especially the connections or relationships that people recognize between concepts. The second was that CMs are able to be "scored" in some way in order to detect differences between individuals and/or between the same person at different times. Results indicated that CM component scores were reliable, based on two indices: (a) measures used to detect differences between individuals and, (b) single-individual pre/post measures of change.

Prior research has also examined differences between the CMs of "experts" and "nonexperts" (Markham, Mintzes & Jones 1994). One study (Markham et al. 1994) indicated that the CMs of more knowledgeable groups were structurally more complex and that the differences in the structural complexity depicted in the CMs were reflected in a sorting task. Individuals’ CMs have also been shown to increase in complexity following exposure o an instructional intervention, providing evidence of the validity of concept mapping in a controlled experimental environment (Wallace & Mintzes 1990). Similar studies have also been conducted in counseling research (cf. Comeau & Hiebert 1991; see also Martin 1985) with the authors concluding that the CM procedure provides innovative access to and visual representation of client conceptualizations.

Finally, the CM methodology has been applied to the study of attitudes in social psychology (Lord et al. 1994). Lord et al. (1994) asked individuals to draw CMs to depict the key ideas and relationships surrounding their perceptions of an abstract social policy. They were also asked to number each concept node as it was "listed" and to add affect tags (evaluations of whether the concept was viewed as positive, negative, or neutral). This latter methodological modification is consistent with discussions of the favorability of brand associations (Keller 1993; Park & Srinivasan 1994), and is appropriate for the analysis of marketing-related objects. Lord et al. (1994) concluded that the CM procedure appears to be as spontaneous as thought-listing tasks, providing valuable insights about the cognitive structure of an individual’s attitude with the addition of a spatial display.

Limitations of the CM Technique. An important assumption underlying CMs is that their generation simply involves a traversal of associative pathways, and therefore, are representative of the underlying structure with respect to both content and organization (Lynch & Srull 1982). However, given the indeterminacy of observed responses, Lynch and Srull (1982) emphasize that both representation and process must be considered in order to draw inferences about memory structure from a recall protocol. In a related argument, Carlston (1991) asserts that it is not certain that free association methods are completely non-reactive or that they can be reliably used to map the structure of a person’s thoughts in any comprehensive way. Because individuals edit activated material (with varying degrees of awareness), Carlston argues that (1) free associations will vary with the situation and mental state of the individual, and (2) people cannot actually engage in passively undirected free association but will instead adopt implicit or explicit response strategies, reflecting such factors as motivation and perceived researcher expectations. One response to this concern is to provide a neutral probe that does not suggest any particular retrieval strategy to the respondents (Shavelson 1974). The instructions used in this research were designed specifically with these limitations in mind.

Summary-Concept Mapping. Prior research and theorizing suggests that the CM procedure can be a useful tool for uncovering the associations making up consumers’ knowledge structures for marketing objects. Olson and Muderrisoglu (1979) stress that determining the content and organization of a consumer’s knowledge structure requires the identification of the "interrelationships or structure among various product knowledge characteristics." The CM methodology appears to have an advantage over other free elicitation procedures in this regard. Concept mapping is also consistent with free association techniques in that it is open and nondirective, which increases the likelihood that the full variety of cognitive associations in a person’s memory structure may be evoked.

The CM task addresses two of the issues Keller (1993) identifies with respect to measuring the characteristics of brand associations. First, the CM procedure seems to adequately capture the favorability (evaluation) of associations. Second, the CM task appears to provide a rather direct means of accessing the relationships among brand associations. Researchers have concluded that CMs complement, but usually do not duplicate, the work of other, more conventional psychometric techniques (Wallace & Mintzes 1990). In addition to noting some of their limitations, Carlston (1991) acknowledges that free association techniques can provide useful data regarding the importance or availability of some of the houghts people have about a topic. These associations clearly convey important information about evaluative reactions to a target. It is perhaps necessary to acknowledge that the major benefit of CM is probably as a qualitative, discovery-oriented methodological tool. As Carlston suggests, other methods and techniques (e.g., priming, reaction times) should also be used to investigate individuals’ knowledge structures and associations. Two studies were conducted in order to examine the use of the CM task in a marketing context. Study 1 examined whether the CM methodology could be used with marketing-related stimuli to elicit salient concepts, and its validity with respect to capturing individuals’ knowledge differences. Study 2 examined the psychometric stability and reliability of the CM data.

STUDY 1

A concept mapping procedure adapted from Lord et al. (1994) was used in this study. Seventy-two undergraduates took part in the study, as partial fulfillment of their course requirements, and were run in small groups of 1 to 5. To provide a "general" frame of reference for the task and minimize potential idiosyncratic context effects caused by respondent-unique interpretations of the task, the survey contained a cover sheet explaining, in as nondirective a manner as possible, the purpose of the study ("to understand what individuals think about, and how they organize information when they think about different marketing objects"). The instructions also stressed that we were looking "for anything that comes to mind when you think about the target object" and that there were no "correct" or "incorrect" concept maps. After respondents had read this introduction to the study, the experimenter explained the concept mapping technique using an example chosen to minimize the priming of any specific types of associations. A CM presented in Lord et al. (1994) was shown as an example of a completed CM. Respondents were randomly assigned to one of the ten stimuli described below and were allowed as much time as necessary to complete an original CM.

Each map began with only the target "node" labeled and participants were instructed to construct a map, numbering the concepts as they wrote them, and adding evaluative labels (+, -, 0). After individuals completed the CM, a series of questions were answered. Individuals’ attitudes were measured with 5 seven-point items describing their attitudes towards the stimulus (favorable/unfavorable, like very much/dislike very much, very positive/very negative, excellent/poor, and desirable/undesirable). Three seven-point questions were used to measure individuals’ knowledge (Alba & Hutchinson 1987) regarding the stimulus target (extremely familiar/extremely unfamiliar, have had a great deal of experience/have had no experience, have had a lot of exposure/have had no exposure). Because the alpha coefficients were sufficiently high for both scales (attitude: .946; knowledge: .792), the attitude (ATTIT) and knowledge (familiarity/experience/exposure; FEE) questions were each analyzed using an average of the scales.

Stimuli It is widely accepted that as a result of differences in expertise, involvement, and other factors, consumers vary in the extent to which they have well-developed, interrelated cognitive structures for a particular concept (Alba & Hutchinson 1987; Sujan 1985). Consequently, consumers are likely to differ in the number of associations stored in memory for a marketing object, the content of hese associations, and their structure. CMs have been shown to be sensitive to these differences in other contexts (Markham et al. 1994). In the current study, two groups of stimuli were chosen on an a priori basis (based on students’ relative experience buying products in each group) so that they would differ with respect to our undergraduates’ knowledge. The two groups chosen were Vehicles and Casual Clothing, with Vehicles expected to demonstrate greater variability in individuals’ knowledge. Within each group there were five different levels of objects: the overall group (Vehicles, Casual Clothing), a more subordinate product class (Automobiles, Jeans), a well-known brand name (Chevy, Levi’s) and two less well-established, goal-derived categories (Barsalou 1983; e.g., "Cars to buy as your first car," "Clothing to wear when exercising"). These different levels were included in order to examine a range of marketing categories that have been studied in past consumer research (hierarchical product category structures and goal-related categories). In contrast to past free elicitation studies (e.g., Kanwar et al. 1981; Olson & Muderrisoglu 1979), participants in the current study completed a CM for a single stimulus in order to avoid the possibility that previous CMs would prime certain types of associations, resulting in 7 or 8 completed maps per stimulus target.

FIGURE 1

CM Measures and Results

In the literature on concept mapping, many measures have been employed in the analysis of CM data (e.g., Martin 1985; Novak & Gowin 1984) including both qualitative and quantitative measures. For the purposes of this study, each of the completed CMs were scored for six conceptually relevant quantitative measures. One measure of the complexity (or differentiation) of a CM suggested in the literature, is the total number of associates in the map (cf. "dimensionality" in Kanwar et al. 1981; see also Markham et al. 1994; Martin 1985). The number of specific examples (Markham et al. 1994; Stuart 1985) in a map has been used as an indicator of the specificity of knowledge reflected in the map. In this context, examples were operationalized as the number of brand names in the map. The total number of brands were counted for each CM and included subbrands (e.g., Corvette, 501s) and the names of retail outlets (e.g., The Gap). Finally, the number of links, or relationships, in a map has been identified as a measure of both the extent of an individual’s knowledge and the conceptual integration and cohesion present in their representation (Stuart 1985; Wallace & Mintzes 1990). Highly integrated knowledge structures are those with a great many connections among all conceptual nodes (Martin 1985). The CMs were scored for the total number of links, the number of direct (first-level) links (cf. "centrality" in Comeau & Hiebert 1991; Lord et al. 1994), the number of second-level links, and the number of third or higher-level links. The number of second and higher-level links provide a measure of the interrelationships among the non-target associations in people’s maps. In this study, all link measures were calculated with respect to the target object. Figure 1 shows a prototypical CM from Study 1 along with its component measure scores. Table 1 lists the means for the CM measures described above for each of the individual targets.

TABLE 1

CELL MEANS FOR CM STRUCTURAL MEASURES

Overall Comparison of Vehicles and Clothing groups. The only measure which was significantly different between the two groups was the number of specific examples (Vehicles, M=3.80, range: 0 to 27; Clothing, M=1.56, range: 0 to 8, t(69)=2.37, p=.021).

CMs and Attitude. An analyis of the evaluative labels in the CMs was conducted in order to assess one aspect of the validity of the CM methodology. If the CMs are actually tapping consumer’s salient associations for the target stimuli, the affectively labeled concepts in the maps should correlate with individuals’ attitudes (Fishbein & Ajzen 1975). Of interest is whether all concepts in the CM are in fact salient. Kaplan and Fishbein (1969) suggest that salient beliefs are elicited first in a free-response task and that the number of beliefs that an individual can associate with any attitude object at any point in time is somewhere between 5 and 9. The use of elicitation procedures such as the CM task may encourage (or "force") individuals to produce associations well in excess of their salient hierarchy. As a result, analyses based upon all of an individual’s elicited associations may actually contain many nonsalient (and perhaps irrelevant) information.

In order to examine this issue, the correlation of ATTIT with the following evaluative measures was computed: the ratio of the number of positive concepts to the total number in the map (POSTOT, r=.48, p=.000), the ratio of the number of positive concepts to the sum of the valenced (i.e., non-neutral) concepts (POSVAL, r=.47, p=.000), and the number of positive concepts in the first five (FIVE, r=.36, p=.004) and first nine (NINE, r=.42, p=.000) nodes. We used both 5 and 9 salient beliefs to be consistent with Kaplan and Fishbein’s (1969) heuristic. When participants elicited less than 5 or 9 beliefs, the estimate was based on the number of beliefs elicited. As these correlations indicate, there is evidence for the validity of the concepts included in respondents’ CMs. All of the measures are positively correlated with the ATTIT measure. Additionally, the fact that the highest correlation with the ATTIT measure is for the POSTOT measure (r=.48) suggests that the CM method is not "forcing" individuals to produce non-relevant associations. It appears that, at least for this study, participants have not gone beyond the point where their associations are starting to lose meaning, suggesting that this unstructured, introspective task can be used to access salient associations.

Knowledge Differences. An additional test of the validity of the CM procedure involved examining quantitative differences in the CMs of knowledgeable versus less knowledgeable individuals (cf. Markham et al. 1994). Valid measures of major characteristics of knowledge structures should reflect differences between individuals with varying levels of knowledge (Kanwar et al. 1981). Differences in respondents’ knowledge for the stimulus targets was measured by the FEE scale described above. Individuals were classified as high (N=28) vs low (N=44) knowledge using a median split of the FEE measure (median=5.67). A series of 2 (Knowledge: high vs low) X 2 (Group: Vehicles vs Clothing) ANOVAs were run to test for potential differences on the six quantitative CM measures described above.

For the total number of concepts, the knowledge main effect was significant (F(1,67)=4.40, p=.04) as was the knowledge X group interaction (F(1,67)=4.01, p=.049). When compared to low knowledge individuals, high knowledge individuals constructed CMs with more total concepts (M=20.41 vs. M=16.64). However, this pattern was only significant for the Vehicles group (25.80 vs 16.68). In the Clothing group the means were not significantly different (17.24 vs 16.58). For the number of brands, the group (F(1,67)=10.09, p=.002) and knowledge (F(1,67)=10.97, p=.001) main effects were significant as was the knowledge X group interaction (F(1,67)=4.52, p=.037). High knowledge individuals included more brands (M=4.19) than low knowledge individuals (M=1.73) and there were more brands mentioned in the Vehicles maps as noted earlier. The interaction was again due to the High-Low difference being significant only in the Vehicles group. There were no significant effects in the ANOVAs for the four links measures. Finally, the correlations between the FEE score and the six CM structure measures were nonsignificant, with the exception of the total number of links (r(72=.27, p<.05).

As expected, there was a significant difference in the average knowledge level between the two stimulus groups (Vehicles: M=4.79; Clothing: M=5.55, p<.03). Inspection of the mean knowledge scores for the high and low knowledge individuals in the previous analysis suggests that there may have been insufficient variance for the Clothing group, with the entire sample being relatively knowledgeable. Specifically, the mean score for the high group (N=16) was 6.37 versus 4.86 for the low group (N=19). In contrast, in the Vehicles group the score for the high group (N=11) was 6.30 compared to 4.10 for the low group (N=24). Additionally, only 4 (11.4%) of the individuals in the Clothing group had scores below the scale midpoint while 12 (34.3%) fell below the midpoint in the Vehicles group. On the basis of this data, we conducted a separate analysis of the CM structure measures for the Vehicles data. Individuals were again divided into high (N=20) and low knowledge (N=16) groups on the basis of a median split. Table 2 lists the means for the high and low knowledge groups for each measure. All differences are in the hypothesized direction and significant with the exception of the number of second level links which approached conventional significance levels (p=.052), and the number of direct links which was not significant (p>.50). [Analyses using the scale midpoint to divide the groups produced virtually identical results, with the number of second level links significant (p=.006) and the number of brands marginally significant p=.07).] The correlations of the FEE measure with these measures are informative as well (Table 3). For the Clothing group none of the correlations are significant.

TABLE 2

MEANS FOR HIGH VS LOW KNOWLEDGE: VEHICLES STIMULI

TABLE 3

FEE CORRELATIONS WITH CM STRUCTURAL MEASURES (N=36)

In summary, for the Vehicles stimulus group, more knowledgeable individuals generated larger, more specific, and more integrated CMs on average. These results did not replicate for the Clothing group, perhaps because of our sample’s high levels of knowledge for these stimuli. Study 1 suggests that the CM task appears to hold significant promise as a methodological tool. However, before accepting the CM methodology, the psychometric stability and reliability of the data need to be established (cf. Olson & Muderrisoglu 1979).

STUDY 2

The goal of Study 2 was to examine the stability of the associates elicited in respondents’ concept maps and the reliability of several quantitative measures (cf. Olson & Muderrisoglu 1979). Stability refers to the occurrence of similar associates in respondents’ maps at two points in time. The reliability of the quantitative measures is indicated by test-retest correlation coefficients. For the purposes of this study, the quantitative measures used in Study 1 were calculated for each completed map at both T1 and T2.

Twenty-two undergraduate marketing students took part in the study as partial fulfillment of their course requirements. Respondents completed the concept mapping task for the target brand name "Haagen-Dazs" on two occasions, exactly one week apart. The concept mapping instructions and procedure were repeated at the second administration of the task. Haagen-Dazs was chosen as the focal brand name on the basis of high brand name recognition among the student population from which the sample was drawn, and positive affect towards the brand among the same group.

Results and Discussion

Three overall indices provide some information about the test-retest stability of the CM methodology. First, the average number of associates produced at T2 that were also mentioned at T1 was 4.6, or 46.4% of the T2 associates. Second, the average percentage of T2 associates that were evaluated the same at T1 was 86.4%, suggesting that the evaluative implications of each of the elicited associates is highly stable. Finally, across the 22 pairs of maps, the rank order correspondence (mean Kendall’s tau) between similar concepts was a moderate, but significant, .577. [These results are similar to those reported by Olson and Muderrisoglu (1979). The degree to which similar associated were elicited in the test-retest task (46.4%) was towards the low end of the range they reported (range: 47% to 60%, mean: 57.6%). However, the rank order correspondence found in Study 2 was higher than those reported by these authors (range: .20 to .48, mean: .349).] Together these analyses suggest that although there is a significant amount of variation between the associates at each time, there is also a relatively stable core of associates elicited by the CM methodology.

TABLE 4

T1 AND T2 MEANS AND TEST-RETEST CORRELATIONS

The quantitative scores for the pairs of CMs provide further evidence of the reliability of the CM measures. The means, and the correlation between the number at T1 and T2 (test-retest reliability), for each of the quantitative measures described above are shown in Table 4. The results suggest that these measures are, for the most part, moderately reliable, with measures of the links in the elicited maps tending to be more reliable than measures of the number of associates.

Finally, it is also possible to look at the most frequently mentioned associates at each of the two time intervals. At the first administration of the CM task, the most frequent single responses were: ice cream (63.6%), chocolate (45%), expensive (40.9%), flavors, tastes good, and dessert (all 27.3%). At T2 they were: expensive (63.6%), ice cream (59.1%), cold (54.5%), chocolate (36.4%), and rich/rich tasting (31.8%). Grouped in broader categories, the number of associates were again very similar, for example, types of products (23 at T1 vs 21 at T2), flavors (24 vs 20), price-related (13 vs 17), and health-related (16 vs 14). This suggests the existence of a reasonably stable component of the elicited CMs. However, the methodology does appear to be somewhat sensitive to immediate context. During the administration of the T2 survey, the wind-chill index was 40 below zero, perhaps as a result, the number of mentions of cold-related attributes and winter-related associations jumped from 10 at T1 to 21 at T2.

The general pattern of results suggests that the CM task appears capable of producing reasonably stable sets of responses. Additionally, quantitative measures of these maps tend to be reasonably reliable. Olson and Muderrisoglu (1979) concluded, "Free elicitation responses may be influenced by a number of variables in addition to the presumed major causal factorCnamely, the underlying cognitive knowledge structure in memory." Acknowledging this, the obtained stability and reliability figures appear promising. [It is important to acknowledge that a major limitation to any test-retest study is the possibility of alternative explanations for the reliability coefficients (Olson & Mudderisoglu 1979).] Further research should now begin to focus on analyzing, in a more qualitative manner, the specific elements of the elicited maps and the relationship between maps for conceptually related stimuli (e.g., brand names, individual branded product). The development of detailed coding schemes that would allow for the interpretation of the rich information elicited in the CM task is a particularly important next step.

CONCLUSION

Clearly, any technique which attempts to access an ndividual’s knowledge structure should permit any belief or association to be elicited (Fishbein & Ajzen 1975). One of the main advantages of both the free elicitation procedure (Olson & Muderrisoglu 1979) and the CM task is that a broad array of associations are elicited. Many elicitation procedures ask consumers to state "product characteristics that come to mind" with the resulting elicited thoughts weighted heavily towards concrete product attributes (Olson & Muderrisoglu 1979). A much broader set of associations is likely to be important to a variety of marketing issues. It appears that the CM task is able to tap into the associations in a person’s knowledge structure and also reveal evidence of the relations between associations. The CM methodology captures some of the structure thought to exist in consumers’ mental representations, and acknowledges the relations which exist in these representations. Feature list representations (and by extension simple thought listing tasks) are restricted in representing relations which provide conceptual coherence for various concepts.

REFERENCES

Aaker, David A. & Kevin L. Keller (1990), "Consumer Evaluations of Brand Extensions," Journal of Marketing, 54 (January), 27-41.

Alba, Joseph W. & J. Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (March), 411-454.

Alba, Joseph W., J. Wesley Hutchinson & John G. Lynch, Jr. (1991), "Memory and Decision Making," in Handbook of Consumer Behavior, ed. Thomas S. Robertson and Harold H. Kassarijian, Englewood Cliffs, NJ: Prentice Hall, 1-49.

Anderson, John R. (1983), The Architecture of Cognition, Cambridge, MA: Harvard Press.

Barsalou, Lawrence W. (1983), "Ad Hoc Categories," Memory and Cognition, 11(3), 211-227.

Broniarczyk, Susan M. & Joseph W. Alba (1994), "The Importance of the Brand in Brand Extension," Journal of Marketing Research, 31 (May), 214-228.

Carlston, Donal E. (1991), "Free Association and the Representation of Complex Cognitive Structures," in Advances in Social Cognition, Vol IV, ed. R.S. Wyer, Jr. & T.K. Srull, Hillsdale, NJ: Erlbaum, 87-96.

Collins, Alan M. & Elizabeth F. Loftus (1975), "A Spreading Activation Theory of Semantic Processing," Psychological Review, 87 (6), 407-428.

Comeau, Nancy & Bryan Hiebert (1991), "Changes in Conceptualizations During Assertiveness Training," Canadian Journal of Counselling, 25 (1), 56-72.

Fishbein, Martin & Icek Ajzen (1975), Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research, Reading, MA: Addison-Wesley Publishing Company.

Judd, Charles M., Roger A. Drake, James W. Downing, & Jon A. Krosnick (1991), "Some Dynamic Properties of Attitude Structures: Context-Induced Response Facilitation and Polarization," Journal of Personality and Social Psychology, 60 (2), 193-202.

Kanwar, Rajesh, Jerry C. Olson, & Laura S. Sims (1981), "Toward Conceptualizing and Measuring Cognitive Structures," in Advances in Consumer Research, Vol. 8, ed. K. Monroe, Ann Arbor, MI: Assoc. forConsumer Research, 122-127.

Kaplan, Kalman J. & Martin Fishbein (1969), "The Source of Beliefs, Their Saliency, and Prediction of Attitude," Journal of Social Psychology, 78, 63-74.

Keller, Kevin L. (1993), "Conceptualizing, Measuring, and Managing Customer-Based Brand Equity," Journal of Marketing, 57 (January), 1-22.

Lord, Charles G., Donna M. Desforges, Steven Fein, Marilyn A. Pugh, & Mark Lepper (1994), "Typicality Effects in Attitudes Toward Social Policies: A Concept-Mapping Approach," Journal of Personality and Social Psychology, 66 (4), 658-673.

Lynch, John G. & Thomas K. Srull (1982), "Memory and Attentional Factors in Consumer Choice: Concepts and Research Methods," Journal of Consumer Research, 9 (June), 18-37.

Markham, Kimberly M., Joel J. Mintzes, & M. Gail Jones (1994), "The Concept Map as a Research and Evaluation Tool: Further Evidence of Validity," Journal of Research in Science Teaching, 31 (1), 91-101.

Markus, Hazel & R.B. Zajonc (1985), "The Cognitive Perspective in Social Psychology," in Handbook of Social Psychology, 3rd edition, ed. Gardner Lindzey & Elliot Aronson, New York: Knopf, 137-230.

Martin, Jack (1985), "Measuring Clients’ Cognitive Competence in Research on Counseling," Journal of Counseling and Development, 63 (May), 556-560.

Novak, John D. and D. Bob Gowin (1984), Learning How to Learn, Cambridge, England: Cambridge University Press.

Olson, Jerry C. & Aydin Muderrisoglu (1979), "The Stability of Responses Obtained by Free Elicitation: Implications for Measuring Attribute Salience and Memory Structures," in Advances in Consumer Research, Vol 6., ed. W. Wilkie, Ann Arbor, MI: Assoc. for Consumer Research, 269-275.

Park, Chan Su & V. Srinivasan (1994), "A Survey-Based Method for Measuring and Understanding Brand Equity and Its Extendibility," Journal of Marketing Research, 31 (May), 271-288.

Russo, J. Edward & Eric J. Johnson (1980), "What Do Consumers Know About Familiar Products?" in Advances in Consumer Research, Vol. 7, ed. Jerry C. Olson, Ann Arbor, MI: Association for Consumer Research, 417-423.

Shavelson, Richard J. (1974), "Methods for Examining Representations of a Subject-Matter Structure in a Student’s Memory," Journal of Research in Science Teaching, 11 (3), 231-249.

Steenkamp, Jan-Benedict E.M., Hans C.M. Van Trijp, & Jos M.F. Ten Berge (1994), "Perceptual Mapping Based on Idiosyncratic Sets of Attributes," Journal of Marketing Research, 31 (February), 15-27.

Stuart, Heather A. (1985), "Should Concept Maps be Scored Numerically?" European Journal of Science Education, 7 (1), 73-81.

Sujan, Mita (1985), "Consumer Knowledge: Effects on Evaluation Strategies Mediating Consumer Judgments," Journal of Consumer Research, 12 (June), 31-46.

Wallace, Josephine D. & Joel J. Mintzes (1990), "The Concept Map As a Research Tool: Exploring Conceptual Change in Biology," Journal of Research in Science Teaching, 27 (10), 1033-1052.

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