Using the Repertory Grid to Assess the Complexity of Consumers' Cognitive Structures
ABSTRACT - Cognitive complexity has sometimes been used as a personality variable and, from this perspective, has had a relatively minor impact on consumer behavior research. From a different perspective, when viewed as a cognitive structure variable, this same concept may have important implications for consumer behavior research. In order to investigate the applicability of cognitive complexity for consumer behavior, the results from two studies are reported. The first study indicates that the Repertory Grid is a reliable instrument for measuring cognitive complexity, and the second study indicates the presence of a generalizable, as well as a context specific, component of cognitive complexity.
Citation:
George M. Zinkhan and Abhijit Biswas (1988) ,"Using the Repertory Grid to Assess the Complexity of Consumers' Cognitive Structures", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 493-497.
Cognitive complexity has sometimes been used as a personality variable and, from this perspective, has had a relatively minor impact on consumer behavior research. From a different perspective, when viewed as a cognitive structure variable, this same concept may have important implications for consumer behavior research. In order to investigate the applicability of cognitive complexity for consumer behavior, the results from two studies are reported. The first study indicates that the Repertory Grid is a reliable instrument for measuring cognitive complexity, and the second study indicates the presence of a generalizable, as well as a context specific, component of cognitive complexity. INTRODUCTION In the last decade, there has been increasing interest in cognitive complexity as an individual difference variable. For example, Henry (1980) has related cognitive complexity to information processing accuracy; Klippel, Anderson and Densmore (1976) relate cognitive complexity to brand loyalty; Moschis and Churchill (1979) relate cognitive complexity to teenage consumer skills; and Durand (1979, 1980) relates cognitive complexity to attitude formation. Cognitive complexity refers to the structural complexity of an individual's cognitive system (Kelly 1955). That is, it has to do with the organizing schemes which individuals use for coding and storing information in memory. Cognitive structures serve a selective function in human information processing by limiting the input of information from the environment. They also act as organizing and moderating variables which control and affect behavior (Bieri 1971). As such, an individual's cognitive structure represents an enduring framework for understanding the environment. In this sense it differs from the ever-changing content of cognition such as attitudes and beliefs (Durand and Gur-Arie 1979). The purpose of this paper is twofold. (1) The first purpose is to investigate the reliability of several measurement procedures which have been proposed for operationalizing cognitive complexity. This is an important step, since most indicators of cognitive complexity must be adapted or altered to be relevant for consumer behavior research. (2) A second purpose of this paper is to determine the degree to which cognitive complexity is a domain specific phenomenon, rather than a generalizable phenomenon. ID other words, if a consumer is cognitively complex with respect to sports fishing, to what extent is it likely that this same consumer is cognitively complex with respect to automobile engines? To explore these two objectives, the results of two different studies are reported. GENERALIZED COGNITIVE COMPLEXITY Research indicates that a portion of cognitive complexity is domain-specific (Linville 1982a). In other words, a person might have a complex cognitive structure for organizing information concerning one domain (e.g. cameras), but have a much simpler cognitive structure for organizing information concerning another domain (e.g., automobiles). Degree of complexity may depend, at least in part, upon the type and amount of familiarity and experience regarding a specific domain (Linville 1982a). There also appears to be a generalized component of cognitive complexity (Tan and Dolich 1980; Wallendorf and Zinkhan 1980; Tan and Lim 1982), which derives from experience in general. On the one hand, extensive experience with cameras, for example, could lead to formation of a complex cognitive structure with respect to cameras. On the other hand, once this cognitive structure is formed for cameras, it may be possible to organize knowledge about other products using-a similar cognitive structure. In some situations, structures or dimensions used in one domain may be transferred to other domains. That is, although the content of knowledge in different domains may be substantively different, the structures used to organize these different knowledge bases may be very similar. Thus a cognitive structure constructed for one domain may be transferable to another domain. In this sense, cognitive complexity can be viewed as having the potential for a generalized content. Going beyond experience in a particular domain, exposure to a wide variety of types and forms of stimulation, including formal education, develops in the individual not only domain-specific cognitive complexity, but also a set of broad evaluative criteria and problem-solving skills which can be used-in many domains. The individual who is exposed to a wide variety of stimuli is therefore more likely to be cognitively complex in a broad array of domains. In - addition the person is likely to develop what may be called generalized cognitive complexity or a set of complex cognitive structures useful in organizing information in a wide variety of domains. The theoretical notion put forward here is that cognitive complexity derives from generalized exposure to new information as well as specialized or domain-specific experience. In order to tap both of these components empirically, it is most productive to measure cognitive complexity as a domain-specific phenomenon. In order to isolate a generalized component of cognitive complexity, it is necessary to compare an individual's level of cognitive complexity in several diverse domains. HYPOTHESES It is important to determine whether the notion of domain-specific cognitive complexity originally developed by psychologists interested in the domain of social interaction (Kelly 1955; Bieri et al. 1966) can be appropriately and accurately applied in the domain of a particular product class. Thus, the first hypothesis is concerned with the applicability of a frequently used measure of cognitive complexity (called the Repertory Grid test) as consumer products replace social interaction as the domain of interest. H1: The Repertory Grid measure of cognitive complexity is applicable in product domains as well as other domains. The second issue to be addressed concerns the extent to which there appears to be a generalized component of cognitive complexity which is transferable across domains. It is the purpose here to determine if cognitive complexity exists across domains. That is, can cognitive structures used in one domain be transferred across domains? This will indirectly address the issue of whether complex cognitive structures are based primarily or even solely on expertise built through experience in a particular domain or whether they are more flexible and n.lid H2: Complex cognitive structures m one domain will be generalized such that they will contribute to cognitive complexity in other domains. MEASUREMENT INSTRUMENT: THE REPERTORY GRID Sometimes researchers are interested in creating an in-depth measurement instrument that can tap many dimensions of a proposed construct. Other times, however, researchers are interested in creating a brief scale which, while retaining the flavor of more detailed instruments, can be used in conjunction with other measurement procedures so that full networks (with several endogenous and exogenous variables) can be investigated. It is with this second purpose in mind that we consider the present measurement procedure. The goal is to measure cognitive complexity quickly and efficiently, while at the same time reflecting the theoretical richness which underlies the set of longer instruments that have been proposed. A REPERTORY GRID One instrument for measuring cognitive complexity is the Role Construct Repertory Grid Test (Rep Grid). Using a modified version of the Rep Grid, it is possible to derive 7 different measures of cognitive structure each based on a slightly different theoretical rationale (Seaman and Koenig 1974). Under the original formulation, the Rep Grid is administered as follows. Respondents are asked to think of a person of the same sex whom they admire. Once the respondent has a clear idea of this admired person in mind, the subject proceeds to rate that admired person along 8 six-point scales with bipolar endpoints. Next, the respondent provides similar ratings with respect to a person of the opposite sex who is not admired. This process continues until 8 types of people have been rated across all 8 bipolar scales. Four of the people rated are positive figures and four are negative figures. A completed Rep Grid is shown in Table 1. Seaman and Koenig (1974) outline a method by which seven different measures of cognitive complexity can be obtained from the Rep Grid. Three of these measures are based on Fiedler's (1967) work on leadership; three of the measures are based on the work of Bieri, et al. (1966) on personality; and one measure is based on Scott's (1962) work on information theory. Each of these measures is explained briefly. Bieri et al.'s Positive (Negative) Construct complexity [+CC(-CC)] reflects the subject's ability to use the eight bipolar construct continuums as independent dimensions when rating positive (negative) stimulus objects. Thus, +CC (-CC) is calculated by counting the number of tied attribute ratings for each of the positive (negative) role figures. Ties indicate that the dimensions are not used independently. Thus the higher the +CC (-CC) score, the lower the person's positive (negative) construct complexity. Bieri et al.'s measure of Total Construct Complexity (TCC) is calculated as the sum of +CC and -CC. Fiedler's Most (Least) Preferred score [MPP(LPP)] reflects the subject's tendency to ascribe negative as well w as positive traits to a positive stimulus object. Thus, MPP (LPP) reflects cognitive complexity as indicated by a person's tendency to see stimulus objects which are regarded positively (negatively) as also possessing some negative (positive) characteristics. MPP (LPP) is calculated by summing the ratings within each column of positive (negative) figures and then adding together these column totals and computing their average. Fiedler's two measures are combined into an overall measure, ASO, which is calculated by subtracting LPP from MPP. The final measure used in this research is Scott's R, which represents the number of distinctions (in bits) that the subject is making. The formula for calculating this measure is: R = [log2 N - 1 / N (Sni log ni) ] / log2 N (1) Where N is the total number of attributes and ni is the number that appears in a particular combination of groups. R is derived from information theory and identifies the number of independent dimensions present within an individual's cognitive domain. This measure may be interpreted as the minimum of independent binary dimensions needed to reproduce a subject's original ratings along the Rep Grid (Linville and Jones 1980). The use of Scott's R does not commit the researcher to the untenable assumption that consumers actively think in terms of independent binary categories. This measure is simply a statistical indicator of the complexity of a set of ratings (Linville 1982b). PROCEDURE Two research studies were completed. The first study is concerned with measurement issues, and, in the interest of devising an improved measurement procedure for cognitive complexity, reliability issues are addressed (see hypothesis one). The second study is relevant for hypothesis two and is designed to investigate the extent to which cognitive structure is generalizable, as opposed to domain specific. FIRST STUDY Method If all of the measures from the Rep Grid are tapping a single underlying construct, then it is possible to standardize these scores and combine them additively to form a single measure of cognitive complexity. The advantage of combining all the measures is that convergent validity can be assessed. By operationalizing complexity with multiple measures which are based on different computational formulae and different theoretical foundations, the advantages of a multi-method approach are added (Campbell and Fiske 1959). Thus, using one Rep Grid, an overall measure can be obtained without any increase in the magnitude of the task for the respondent. However, the reliability of this composite scale (sum of standardized scores) must still be assessed. This is the first task of the first study. In order to assess reliability, five of the seven complexity measures are combined. Only five are considered, since the remaining two are redundant. That is, TCC is calculated as the sum of +CC and -CC; likewise, ASO represents the difference between MPP and LPP. For this reason TCC and ASO, as redundant measures, are excluded from the reliability calculations because their inclusion would unjustifiably innate any estimates of reliability. The Rep Grid was administered to three groups of subjects: two groups of college students (n = 80 and n = 106) and one group of adults (n = 82). The first group completed two Rep GridsCone for the domain of social interaction (similar to Kelly's original conception of cognitive complexity) and one for the domain of athletic shoes. In this way, it is possible to compare the original form of the Rep Grid with the adapted form which is more appropriate for consumer behavior applications. The second group of students and the group of adults completed a Rep Grid in the domain of calculators. This facilitates comparison of the adapted form across groups and is an important step by allowing the determination of whether using the Rep Grid on a less homogeneous sample (the adults rather than the college students) produces a similar level of reliability. That is, when a research effort expands beyond a sample of students, the measured reliability of the instrument may decrease merely because the sample is less homogeneous. To the extent that such a decrease occurs, the instrument is less reliable than would be indicated using a student sample alone. Results To calculate coefficient alpha, the five relevant complexity scores (+CC, -CC, MPP, LPP and R) were standardized and summed to form a composite index. The resulting estimates for coefficient alpha are displayed in Table 2. ESTIMATES OF COEFFICIENT ALPHA In three out of four cases, these estimates are above .70 and in all cases are above .60. In terms of reliability, the consumer behavior formulation of the Rep Grid using product brands rather than people as the stimulus objects has a reliability coefficient of .729. This seems to be at least as good as Kelly's original formulation of the Rep Grid which has an alpha coefficient of .684 as measured here. In addition, there is little absolute difference in the reliability coefficients between the student group (.788) and the non-student group (.786) when using calculators as the stimulus object. In summary, the reliability of the composite measure of cognitive complexity appears to be satisfactory. SECOND STUDY Method In the second study, the domains of popular literature and cameras are studied in addition to the more frequently studied domain of social interaction. The purpose of this second study is to examine cognitive structure across all three domains and to determine whether cognitive complexity has a generalized component as well as domain-specific component. Three Rep Grids were administered to 126 undergraduate students. Each Rep Grid measured the respondent's cognitive structure with respect to a different domain. The domains (popular literature, cameras, and social interaction) were chosen to represent maximally dissimilar content. The domain of social interaction was chosen because it is the domain originally investigated by Kelly (1955) in his research on cognitive structure. By examining correlations across domains, it becomes possible to search for a generalized component of complexity. Results When comparing correlations across domains, it does not make sense to include the redundant measures. For example, if +CC for cameras correlates highly with +CC for social interaction, then likewise TCC for these two contexts would also correlate highly. For this reason both TCC and ASO, as redundant measures, are excluded from the analysis across domains. Table 3 presents the correlations between the cognitive complexity measures in all three domains. With respect to social interaction and popular literature, significant coefficients (p < .05) are observed among the Bieri measures (4 out of 4), and the Bieri measures are also significantly associated with 3 other indicators. Altogether, 11 of 25 coefficients achieve statistical significance; the largest coefficient is .27. Notice also that +CC, -CC and MPP correlate negatively with other indicators since they represent inverse measures of cognitive complexity. With respect to the domains of cameras and popular literature, ten of the 25 coefficients achieve statistical significance (< .05); the highest coefficients are observed in the .20 to .33 range. Bieri's measures correlate highly among themselves (3 out of 4 significant at the .05 level); and Fiedler's measures also correlate highly among themselves (2 out of 4 significant at the p < .05 level). In addition, Scott's information measure correlates well with both a Bieri and a Fiedler measure. COMPLEXITY CORRELATION BETWEEN DOMAINS The most encouraging results are obtained when camera cognitive complexity measures are compared with social interaction cognitive complexity measures (see Table 3). In this case, 15 out of 25 coefficients achieve statistical significance (p < .05), and the largest of these are in the .28 to .41 range. The Bieri measures are significantly related to all other CC measures. The same pattern is observed for the Scott and Fiedler measures. When looking within the Fiedler measures, all 4 out of 4 correlations are high. In summary, some support is found to indicate that the cognitive complexity has a generalized component. Between 4% and 16% of the variance in the cognitive complexity measures seems to be accounted for by this generalized component. One reason for this low amount of variance explained may be that many of the respondents in this survey (the college students) had high education levels, and this may have restricted the range of the cognitive complexity measures. Especially encouraging are the results associated with social interaction and cameras. Social interaction represents the original conceptualization and cameras represent the type of domain for cognitive complexity that is most relevant for consumer researchers. Also, since the Rep Grid measures seem to behave in a similar fashion across all three domains, this procedure appears to be appropriate and applicable to consumer behavior research problems. DISCUSSION In the psychological literature cognitive complexity has often been discussed as a personality variable (Vannoy 1965). Within this perspective, cognitive complexity has sometimes been viewed as a quite general trait, pervading all realms of cognitive functioning. In this sense cognitive complexity, as a psychological concept, has suffered from some of the same criticisms which have limited the applicability of personality variables to consumer behavior research (Kassarjian 1971). In contrast, other researchers (e.g., Scott 1963; Zajonc 1960) have viewed complexity as a somewhat less enduring state applying only to a particular cognitive domain. This latter view seems to hold the most promise for consumer behavior researchers. The results reported here suggest that it is appropriate to use an existing measure of cognitive complexity for consumer research. Based upon the reported evidence, it appears that the Rep Grid is reliable when used in both product and social interaction domains. Coupled with the objective scoring procedures associated with the Rep Grid, these findings suggest that the Rep Grid is a useful tool for investigating cognitive complexity in studies of consumer behavior. It should be noted that these conclusions are consistent with those in earlier studies (Tan and Dolich 1980; Tan and Lim 1982). The study of knowledge structures can be important since these structures may guide the processing of information, evaluation of stimuli, and choice processes. The Repertory Grid, as modified here, may prove to be a valuable research tool for exploring one property of cognitive schema: namely, schema complexity. In particular, this research extends the methodology employed in much of the schema literature by treating schema complexity as a continuous rather than a discrete variable. In the future, cognitive complexity should continue to be viewed as a domain-specific phenomenon, although there is a small component of cognitive complexity that may be attributable to a generalized element. That is, there may be some transfer of structure between domains. The exact nature and process of this transfer remain unexplored. Investigation of other domains may expand the usefulness of this concept as an individual difference variable and predictor of consumer behavior. REFERENCES Bieri, James (1971), "Cognitive Structures in Personality," in Harold M. Schroder and Peter Suedfeld, eds., Personality Theory and Information Processing, USA: Ronald Press Company, 178-208. Bieri, James, Alvin L. Atkins, Scot Briar, Robin L. Leaman, Henry Miller and Tony Tripodi (1966), Clinical and Social Judgement: The Discrimination of Behavioral Information, New York: Wiley. Campbell, Donald T. and Donald W. Fiske (1959), "Convergent and Discriminant Validity by the Multitrait-Multimethod Matrix," Psychological Bulletin, 56 (March), 81-105. Durand, Richard M. (1979), "Cognitive Complexity, Attitudinal Affect, and Dispersion in Affect Rating for Products," The Journal of Social Psychology, 107, 209-212. Durand, Richard M. (1980), "The Effect of Cognitive Complexity on Affect Ratings of Retail Stores," Journal of Social Psychology, 110, 141-142. Durand, Richard M. and Oded Gur-Arie (1979), "Cognitive Differentiation: A Moderator of Behavioral Intention," in Neil Beckwith et al. (eds.), Educators' Conference Proceedings, (Chicago:AMA), 305-308. Fiedler, F. (1967), A Theory of Leadership Effectiveness, New York: Norton. Henry, Walter A. (1980), "The Effect of Information-Processing Ability on Processing Accuracy," Journal of Consumer Research, 7 (June), 42-48. Kassarjian, Harold H. (1971), "Personality and Consumer Behavior A Review," Journal of Marketing Research, 8 (November), 409-418. Kelly, G. A. (1955), The Psychology of Personal Constructs, New York: Norton. Klippel, Eugene R., Robert L. Anderson, and Max L. Densmore (1976), "An Experimental Investigation of Individual Cognitive Complexity as a Predictive Measure of Brand Loyalty," Proceedings of the Midwest AIDS Conference, (Detroit: AIDS), 57-58. Lintille, Patricia (1982a), "Affective Consequences of Complexity Regarding the Self and Others," in M. S. Clark and S. T. Fiske (eds.), Affect and .Cognition, Hillsdale, NJ.: L. Erlbaum and Associates. Linville, Patricia (1982b), "The Complexity Extremity Effect and Age-Based Stereotyping," Journal of Personality and Social Psychology, 42, 193-211. Linville, Patricia and Edward E. Jones (1980), "Polarized Appraisals of Out-Group Members," Journal of Personality and Social Psychology, 38, 689-703. Moschis, George P. and Gilbert A. Churchill, Jr. (1979), "An Analysis of the Adolescent Consumer," Journal of Marketing, 43 (Summer), 40-49. Scott, William A. (1962), "Cognitive Complexity and Cognitive Flexibility," Sociometry, 25, 405-414. Scott, William A. (1963), "Conceptualizing and Measuring Structural Properties of Cognition," in O. J. Harvey (ed.), Motivation and Social Interaction: Cognitive Determinants, New York: Ronald Press. Seaman, Jerrol M. and Fedrick Koenig (1974), "A Comparison of Measures of Cognitive Complexity," Sociometry, 37, 375-391. Tan, Chin T. and Ira J. Dolich (1980), "Cognitive Structure in Personality: An Investigation of its Generality in Buying Behavior," Advances in Consumer Research, Vol.7, 547-551. Tan, Chin T. and Hui H. Lim (1982), "Cognitive Structure in Buying: Its Generality in Another Culture," in Bruce J. Walker et al. (eds.), Educators' Conference Proceedings, (Chicago: AMA), 80-83. Vannoy, Joseph S. (1965), "Generality of Cognitive Complexity-Simplicity as a Personality Construct," Journal of Personality and Social Psychology, 2, 385-396. Wallendorf, Melanie and George Zinkhan (1980), "Individual Modernity and Cognitive Complexity as Conceptual Bases for Marketing," in Lamb and Dunne (eds.), Theoretical Developments in Marketing, (Chicago: AMA), 59-63. Zajonc, R. B. (1960), "The Process of Cognitive Tuning in Communication," Journal of Abnormal and Social Psychology, 61, 159-167. ----------------------------------------
Authors
George M. Zinkhan, University of Pittsburgh
Abhijit Biswas, University of Houston
Volume
NA - Advances in Consumer Research Volume 15 | 1988
Share Proceeding
Featured papers
See MoreFeatured
Placing Identity into the Self-Concept: The Role of Causal Beliefs in Identity-Based Consumption
Stephanie Chen, London Business School, UK
Oleg Urminsky, University of Chicago, USA
Featured
Psychological Reactions to Human Versus Robotic Job Replacement
Armin Granulo, Technical University of Munich
Christopher Fuchs, Technical University of Munich
Stefano Puntoni, Erasmus University Rotterdam, The Netherlands
Featured
Communicating a Promise of Change: Visual Steps Enhance Process Imagery
Luca Cian, University of Virginia, USA
aradhna krishna, University of Michigan, USA
Chiara Longoni, Boston University, USA