Consumers’ Attitudinal Profiles: an Examination At the Congruence Between Cognitive, Affective and Conative Spaces

Citation:

Jacques-Marie Aurifeille, Fabrice Clerfeuille, and Pascale Quester (2001) ,"Consumers’ Attitudinal Profiles: an Examination At the Congruence Between Cognitive, Affective and Conative Spaces", in NA - Advances in Consumer Research Volume 28, eds. Mary C. Gilly and Joan Meyers-Levy, Valdosta, GA : Association for Consumer Research, Pages: 301-308.

Advances in Consumer Research Volume 28, 2001     Pages 301-308

CONSUMERS’ ATTITUDINAL PROFILES: AN EXAMINATION AT THE CONGRUENCE BETWEEN COGNITIVE, AFFECTIVE AND CONATIVE SPACES

Jacques-Marie Aurifeille, The University of La Reunion

Fabrice Clerfeuille, University de Nantes

Pascale Quester, The University of Adelaide

INTRODUCTION

From its early presentation by Rosenberg and Hovland (1960), the three component view of attitudes has been the object of a prolific research, following two major directions. Some researchers have examined the question of whether these components are co-linear or autonomous, whilst others have debated whether attitudes result mainly from situational elements or from more durable consumer characteristics (Anand, Holbrook and Stephens 1988).

Researchers favouring the influence of situation conceptualise attitudes as resulting from situational variables such as involvement, purchase environment or product type (Park and Young 1985). For instance, the involvement profiles developed by Laurent and Kapferer (1985) correspond to ten possible combinations of cognitive (interest), affective (pleasure) and conative (probability of mistake, importance of consequences) factors.

Other researchers have suggested that attitudes are more permanent and based on consumers’ style or profile. First introduced in marketing by Bieri (1961), cognitive styles have been the objects of many subsequent developments (Schaninger and Sciglimpaglia 1981; Alba and Hutchinson 1987, Bagozziand Foxall 1996). In relation to attitudinal components, styles based on affective and cognitive elements have been categorised by Sojka and Giese (1997). In a 2-dimensional space (cognitive/ affective) they identified four types of consumers. 'Thinking Processors’ favour cognitive processes rather than affective ones whereas 'Feeling Processors’ do the opposite.

Several reported studies have provided support for the interrelated nature of affective and cognitive attitudinal components (i.e. Breckler 1984, Breckler and Wiggins 1989 or Miniard and Barone 1997). The contribution of this study is its examination of the outcomes of such inter-relatedness. In particular we believe that both affective and cognitive components of attitudes contribute, with varying degrees of congruence, to an overall attitudinal space which may differ between individuals and/or product category. Furthermore, we believe that the conative component also influences this overall attitudinal space. Thus, whilst cognitive, affective and conative components of consumers’ attitudes are a fundamental focus of much theoretical development in marketing research, this paper aims specifically at providing a new direction for their empirical examination. As a result, these concepts are defined in this study from an empirical perspective. Cognitive aspects of consumers’ attitudes refer to their beliefs and knowledge about products as well as to their perceptions of differences between products (Cooper 1983). Affective aspects comprise the "emotions, moods, feelings and motives" that reflect the matching of consumers’ motives with products (Lelkoff-Hagius and Mason 1993) and are more commonly measured via stated preferences (Creusen and Schoormans 1997). Conative aspects, on the other hand, relate to behavioural consequences and are more easily understood via intention to buy or actual purchase behaviour.

The autonomous nature of conative elements and the opportunity to combine affective and conative elements independently from the cognitive one have already been examined, particularly with respect to retailing. Hirschman and Holbrook (1992) introduced the concept of 'hedonic consumption’ whereby the emotive stimulation derived from the act of purchase itself equals or exceeds that expected from the product consumption. Similar effects were found for 'purchase involvement’ -as opposed to product involvement- (Ohanian and Tashchian 1992, Lockshin, McIntosh and Spawton 1998). In the case of this latter study, the particular purchase or usage situation of the product (wine), seems to add an affective component (pleasure to buy), to that resulting from the product consumption (pleasure to drink).

OBJECTIVES AND HYPOTHESES OF THE STUDY

Current research appears to favour the identification and segmentation of typical profiles whereby consumers’ affective and cognitive predispositions are examined simultaneously (eg. Sojka and Giese’s four classes 1997). This is also the approach adopted in this study where all three elements of attitudes (cognitive, affective and conative) are taken into account, using a quantitative process providing a more precise understanding of both conceptual and operational aspects of the problem.

Our objective is to demonstrate the existence of consumer segments based on attitudinal profiles and to test how such profiles can be related to consumer or product characteristics. The methodology used involves first the determination of a shared space between cognitive, affective and conative consumers’ attitudinal structures using a non-parametric multi-dimensional analysis 'INDSCAL’ (Caroll and Chang 1970), with the contribution of the dimensions of the shared attitude space to each of these structures reflecting the manner in which the structures combine to provide a profile. By shared space, we mean a reduced-size space that accurately reflects the cognitive, affective and conative structures.

Multi-dmensional scaling (MDS) enables the representation of differences between objects as Euclidean distance in a n-dimensions space. The algorithm on which it is based (Torgerson 1952) has since been extensively modified, for example to reflect the ordinal nature of the data (Young and Hamer 1987). An important adaptation of the initial MDS algorithm, INdividual Difference SCALing (INDSCAL), enables the identification of the space common to several dissimilarity matrices (Caroll and Chang 1970) and the estimation of the weight representing the degree to which a dimension of the common space contributes to the representation of the matrix.

In this study, INDSCAL is used to examine simultaneously for each consumer, three matrices representing respectively his or her cognitive, affective and conative predisposition towards a product category. Assuming that m₁, m₂ and m₃ represent the cognitive, affective and conative attitudinal matrices of a consumer, translated in a H-dimensional space, each matrix would be characterised by H parameters w_mh measuring the contribution of a dimension h to the representation of the matrix M in the shared space. Matrices would be found to be congruent when their w_mh parameters are highly correlated ie. when they share equally the dimensions of the common space. The congruence of a matrix with other matrices translated in the same space could then be measured by the correlation between the w_mh of the matrix and the average w_mh across all other matrices. This correlation coefficient, ranging from 0 (high correlation) to 1 (low correlation), is known as the 'weirdness index’ and is available in most statistical softwares allowing INDSCAL analysis (such as SPSS).

TABLE 1

CONGRUENCE AND WEIGHTS OF THREE MATRICES IN A 3-DIMENSIONAL SPACE OF DIMENSIONS h₁, h₂, AND h₃

TABLE 2

CONGRUENCE AND WEIGHTS OF THREE MATRICES IN A 3-DIMENSIONAL SPACE OF DIMENSIONS h₁, h₂, AND h₃

In this paper, the congruence of a matrix with the shared space is simply defined as (1BWeirdness index), and the consumer’s attitudinal profile is the vector of the congruence values of his or her matrices. In this study, therefore, the attitudinal profile has always 3 values, corresponding to the cognitive, affective and conative matrices, even though the number of dimensions of the shared space may be less or more than 3. For example, a consumer whose weights are shown in Table 1 exhibits a totally non-congruent profile {0, 0, 0}, indicating a complete autonomy of the three attitudinal matrices and illustrating an extension of the cognitive/affective independence hypothesis (Zajonc 1968) to all three attitudinal components. Conversely, Table 2 shows a case where the consumer’s three attitudinal matrices are perfectly congruent since each was equally reflected by the dimensions of their common space, hence creating a perfectly congruent profile {1, 1, 1}. Such co-linearity of the three attitudinal components is consistent with researchers who have argued in favour of hierarchical models. These two examples are represented in Figure 1, along with 3 other possible typical attitudinal profiles (congruent matrices are shown as belonging to the same rectangle).

The first profile shows the totally non-congruent example described in Table 1, whereas the second one illustrates the totally congruent example described in Table 2. Profile No 3 shows a consumer whose affective and conative components are congruent but autonomous from his or her cognitive component, consistent with Zajonc’s view of the behavioural and the affective being linked independently from beliefs and knowledge which the consumer might have about the product. Profile No 4 is that of a consumer whose conative and cognitive components are congruent but autonomous from affective processes, consistent with the business-to-business paradigm where rational or collegial decision-making is thought to preclude affective considerations (Brown and Brucker 1990). Finally, Profile No 5 represents a consumer whose conative element (intention to buy) is autonomous from both affective and cognitive components. Intuitively, such instances would seem more likely in the case of impulse purchases or stock-out situations (Vaughn 1980).

Clearly, however, observed attitudinal profiles would be less neatly defined than those 5 typical cases. Their identification and lassification should rely on a segmentation process requiring that discrimination between observed profiles be achieved, using such operational variables as consumer demographics or product category. With this operational aim, and to illustrate how such variables could be associated practically with attitudinal profiles, several hypotheses were developed:

H1: A single reduced-size space can reflect consumers’ cognitive, affective and conative attitude structures (matrices) towards a product.

H2: Attitudinal profiles vary according to some consumer characteristics (eg. demographics, loyalty, etc)

H3: Attitudinal profiles vary by product category and/or brand

H1 will be tested by examining the adjustment indices of the INDSCAL solutions. Stress will be calculated according to Kruskal’s formula (Kruskal 1964). H2 will be tested by comparing the average age, gender and brand loyalty across segments of attitudinal profiles, H2 being rejected if the corresponding F-tests indicate that no differences exist between profile clusters. H3 can be tested by analysing whether consumers are grouped in the same profile clusters when different Fast Moving Consumer Goods (FMCG) are considered, H3 being rejected if the corresponding Chi-Square test yields a less than .05 probability of no difference.

FIGURE 1

FIVE TYPICAL ATTITUDINAL PROFILES

DATA COLLECTION METHOD

The choice of products for this study was driven by the need to examine product categories within which consumers would be familiar with many (10 or so) brands. This is the result of the positive degree of freedom constraint imposed by the need to extract a non-random space with a large enough number of dimensions. In line with previous work using MDS, our study was based on soft drinks and confectionary bars and relied on the use of a student sample (Green, Carmone and Smith 1989, Hoolbrook, Moore and Winer 1982). Several studies have demonstrated the importance of all three attitudinal components on the buyer behaviour relating to these products (Clayes et al. 1995, Vaughn 1980), and students represent a core target market for both products.

MDS is often used to measure brand perceptions when evaluative attributes are unclear or hard to quantify. Thus, our data collection involved measuring perceived brand differences using pair comparisons, the most respondent-friendly and reliable method for collecting such data (Whipple 1976). With this method, respondents appear to adopt different perspectives according to the pair of objects between which they are asked to estimate a difference (Malhotra, Jain and Pinson 1988). As a result, several latent contexts may underpin the dissimilarity space (Lehman 1971, Hustard, Mayer and Whipple 1975) corresponding for example to different purchase or consumption situations. Therefore, one can expect to find multi-dimensionality, not only in the cognitive element but also in the affective and conative ones, since both preference and inclination to buy may vary according to the context embedded in the particular pair of objects under consideration. Involvement profiles (Laurent and Kapferer 1985) are similarly linked to the concept of purchase situation. Their multi-dimensionality demonstrates the degree to which cognitive (or interest), affective (pleasure) and conative (probability of mistake) are combined differently according to the specific purchase situation faced by the consumer.

The questions used in this study were designed to measure the attitude components for each pair of 10 brands of soft drinks and each pair of 10 brands of confectionary bar. Cognitive structures were measured by asking respondents the differences they perceived for 10 brands of soft drinks and 10 brands of confectionary bars. For each pair of product in the same category, the following question was asked:

"Please note with a cross the degree of difference which you perceive between the two following products": (eg Coca Cola- Pepsi Cola)

Not at all different:____:__X_:____:____:____:____:____: Completely different

The affective structure was measured in such a way as to enable the researchers to test its dimensionality. In a similar study, Moore, Pessemier and Little (1979) had used the 'dollarmetrics’ approach (Pessemier 1975). However this approach is unsuited for the inexpensive products used in this study. As a result, a two-step approach was adopted. First, consumers were asked to rank brands by order to preference. Based on this, pairs (A, B) were determined where A is preferred over B. Whilst preferences may also, in part, result from a cognitive process (Zajonc and Markus 1982, Miniard and Barone 1997), the INDSCAL design would indicate how the affective and cognitive structure are reflected along the dimensions of the common space. Hence, the following question was asked about each of the identified pairs of products:

"Please indicate with a cross to what extent you prefer A to B":

Very Little: ____:____:____:____:__X_:____:____: Enormously

For the conative structure, purchase frequency was deemed a better indicator than purchase intentions because the products selected were in the low involvement/frequent purchase category. Therefore, for each pair (A,B) of products within the same category, the following question was asked:

"Please compare your usage frequency of brand A with your usage frequency of brand B and mark with a cross the corresponding answer":

A much less often than B:___:_X_:___:___:___:___:___:  A much more often than B

In order to test H2 and H3, other variables were also collected concerning consumers’ age, gender and brand loyalty. Furthermore, questions were rotated to avoid any order effects and subsequent tests concerning the potential effect of question order showed that its contribution to explaining the results was negligible.

The sampl comprised 150 consumers. In order to ensure that respondents with experience with the chosen product categories were selected, undergraduate and postgraduate students were used, with 75 males and 75 females aged between 19 and 48. The 10 brands used for each product category were selected on the basis of the top-of-the-mind awareness they achieved in preliminary research and included, for softdrinks: Canada Dry, Coca-Cola, Fanta, Gini, Orangina, Pepsi Cola, Ricles, Schweppes, Seven-up and Sprite and for confectionary bars: Bounty, Kinder, Kit Kat, Lion, Mars, Milky Way, Nuts, Snickers, Sundy and Twix. The total data base, therefore, consisted in 900 matrices (150 respondents x 3 matrices x 2 products).

METHOD OF ANALYSIS AND RESULTS

Hypothesis concerning a common space

To examine H1, that a common space existed between the three attitudinal component of any individual, it was first necessary to check that each of the component spaces had a structure. This was achieved by undertaking a MDS analysis for each of the 900 collected matrices. This analysis was undertaken in 4 dimensions, the larger dimensionality possible with a positive degree of freedom, given the number of products used in the study.

As noted previously, the stress function to minimise was that based on Kruskal formula. According to Kruskal, solutions providing a stress level inferior to .1 should be accepted. However, when the number of dimensions and objects is high, a lower level of stress might be expected. Indeed, Spence and Oglivie (1973) suggested that when undertaking a MDS analysis of 10 objects in 4 dimensions, the maximum level of stress allowing the rejection of the null hypothesis with a 95% confidence level is .07. This value was therefore retained in this study.

In order to avoid a degenerescence of the solution, the number of iterations was limited to 20 and the dissimilarity/distance diagrams (Green 1975) were systematically controlled. In total, only 3 matrices were rejected after this process, requiring the removal of the three corresponding respondents from the sample and leaving a total of 882 useful matrices (147x3x2) for further analysis. Thus, all matrices exhibited a non-random structure when represented in a reduced-size multi-dimensional space. This supports the hypothesis that consumers bring a multi-faceted perspective to pair evaluations and that they hold a multi-dimensional view of each attitudinal space.

The next step involved a direct examination of H1, that a common space exists between the three components of an individual’s attitude. In order to perform this test, two distinct INDSCAL analyses were undertaken, one for each product category. The process is similar to the one described above except that the one-matrix approach of MDS is replaced by the three-matrices approach afforded by INDSCAL. Furthermore, in order to better represent the three initial spaces, the analysis was undertaken in 5 dimensions. The recommended maximum level of stress for analyses of this type is .07 (Spence and Oglivie 1973).

When compared with the previous analysis, INDSCAL generated higher levels of stress, indicating that, as expected, the differences in each matrix was less perfectly reflected in the common space. However, stress remained in a confidence interval (at 95%) of [.0377, .0393], well below the threshold of .07. Only 12 matrices exceeded marginally the .07 mark. However, considering how close to the acceptable range they were, and to avoid removing 12 individuals, these were kept in the data set.

In summary, these results show that attitudinal components share a non-random reduced-size space which is both non-random and multi-dimensional. The existence of a common space (ie. a space of reasonable dimension reflecting accurately the non-random structures of the attitudnal components) enables us to propose a definition of attitudinal profiles comprising the congruence scores of the consumer’s cognitive, affective and conative matrices within their common space.

Hypothesis relating to consumer characteristics and product influence

In order to examine H2 and H3, a cluster analysis was undertaken for each of the two product categories, using the k-means algorithm and the Euclidean distance criterion of Ward. The results of this analysis, for solutions comprising 2 to 6 clusters, are presented in Table 3. Internal validity was assessed using Biserial correlation (Milligan 1981) which varies between 1 (very efficient clustering) to -1 (very inefficient clustering). This measure was deemed the most reliable (Milligan 1981) and is equivalent to other commonly used indices (Klastorin 1983). Whether the means of congruence varied significantly between clusters was assessed using ANOVA which results are included in the column so-titled, indicating the proportion of variables for which the congruence means differed significantly between clusters (P<.05).

At the consumer’s level, the INDSCAL analysis of the three matrices (cognitive, affective and conative) defines a common space reflecting the observed distances in the three matrices. Each matrix is characterised by a weight, one for each of the dimensions of the common space. The greater the weight of a matrix on one dimension, the more it determines this dimension. Two matrices of equal weights would therefore influence equally all dimensions, and be described as congruent. The congruence of a matrix can therefore be measured by comparing its weights to the average of the weights of the three matrices. The greater the difference (weirdness index) and the less congruent the matrix. Formally then, congruence was calculated by the correlation of the weights of one matrix with the average weights of all matrices.

For each product, the number of clusters and their respective size (number of individuals per cluster) provided some indication of the operational validity of the analysis. As Table 3 shows, the solutions with 2 or 3 clusters are better for soft drinks whereas the solution with 2 clusters is clearly superior for confectionary bars. In these solutions, a clear distinction is found between those individuals whose attitudinal components are strongly congruent and those whose attitudinal components exhibit little congruence. In the former group, all three components are restituted equally by the space dimensions, consistent with the hierarchical effect model whereby all three components interact in sequence. For example, in cluster 1 of the 3-clusters-solution for soft drinks, consumers are congruent (.863, .861, .862) whereas consumers in cluster 3 exhibit components that appear to contribute to different dimensions (.596, .450, .605). The same observation applies to confectionary bars. In general, clusters with individuals exhibiting more congruent profiles tended to be larger, suggesting the predominance of dependant effects models.

Partial congruence, where congruence of two of the three components is observed, was only found in higher number solutions. For example, in the 5-clusters solution for confectionary bars, cluster 3 is characterised by a clear autonomy of the affective component in relation to the cognitive and conative ones (.452 against .673 and .691), supporting the notion that the affective component is autonomous. In the same cluster solution (cluster 4), cognitive and affective are little congruent (.559 and .523 respectively) while the affective component contributes to their respective dimensions (.641), suggesting an instance where purchase is autonomous from the perceptions of the product and where both are associated with a specific affective effect. As previously noted, simultaneous product and purchase involvements (Ohanian and Tashchian 1992, Lockshin, Macintosh and Spawton 1998) may characterise a type of festive consumption situation associated with chocolate products.

TABLE 3

2-TO 6-CLUSTERS SOLUTIONS OBTAINED: SOFT DRINKS AND CONFECTIONNARY BARS

The internal and face validity of the profiles ientified above must be associated with some operational validity in order to assist marketing decision making. In particular, H2 and H3 relate these profiles to consumer characteristics and to product category. Table 4 summarises the results obtained from those cluster analyses offering the best internal validity (similar results were achieved when examining higher-number cluster solutions). Only one significant relationship is revealed by this analysis, in the case of confectionary bars and only for gender, so that female respondents exhibited a more congruent attitudinal profile than men with respect to this product category. Thus, H2, strictly speaking, cannot be rejected.

The analysis summarised in Table 4 also suggests that attitudinal profiles vary according to product category. However, H3 can be tested more formally by comparing the clusters derived from the data concerning soft drinks and confectionary bars. A Chi-Square test indicated that consumers allocated to a cluster for one product would not be allocated to the same cluster for the other product, hence providing further support for H3. In other words, in explaining a given consumer’s attitude towards two products, the cognitive, affective and conative components would play different parts.

TABLE 4

SIGNIFICANCE BETWEEN GENDER, AGE, LOYALTY AND ATTITUDINAL PROFILES

TABLE 5

DISCRIMINANT ANALYSIS OF BRANDS, ACCORDING TO CONGRUENCE SCORES

Hypothesis concerning a relationship between brands and attitudinal profile

Once a relationship is established between product category and attitudinal profiles, a natural extension of this hypothesis involves examining whether such a relationship also exists with regards to brands. In order to answer this question, consumers’ scores were examined to determine whether they allowed a discrimination amongst brands. Table 5 presents those brand pairs most clearly discriminated by the analysis.

As Table 5 shows, attitudinal profiles of consumers of one brand may differ markedly from that of consumers of another brand. For example, attitudinal profiles of Snickers’ consumers (.62, .62, .62) are clearly less congruent than that of other brands’ consumers (eg Bounty: .75, .74, .74). Likewise, Coca Cola consumers’ attitudinal profiles (.70, .69, .68) are significantly less congruent than that of Schweppes consumers (.78, .77, .79).

Since a lack of congruence indicates that cognitive, affective and conative contribute to different dimensions, consumers of Snickers or Coca Cola may require a more varied mix, emphasising differently cognitive, affective and conative aspects, depending on their particular attitudinal profiles. For example, advertising messages may be included jointly in the communication strategy that appeal to cognition or to emotions while, at the same time, packaging and point-of-sale material may provide the conative element.

CONCLUSION

This study examined the issue of autonomy or congruence of the cognitive, affective and conative attitudinal structures by using a novel approach based on the concept of attitudinal profiles. We observed empirically that a consumer’s cognitive, affective and conative structures were multi-dimensional and that they could be represented in a reduced-size common space. The restitution of these structures by the dimensions of the common space enabled the determination of attitudinal profiles which, when grouped into clusters, exhibited high internal validity and could be significantly related to some consumer and product characteristics.

A majorityof consumers were found to be 'congruent’, consistent with traditional views whereby cognitive, affective and conative predispositions all contribute similarly to overall attitudes. However, other consumers were identified who exhibited a much less congruent profile and whose attitudinal structures were more autonomous, consistent with more recent theories of attitude.

For marketing managers, these findings are interesting since attitudinal profiles may enable a segmentation of consumers that can be related to both personal and product characteristics. Furthermore, the relationship found to exist between attitude profiles and brands suggests that an attitudinal profile segmentation may assist in revisiting previous positioning strategies by differentiating the dimensions of the attitudinal space according to their restitution of consumers’ cognitive, affective and conative structures. Thus, alternative persuasion strategies might be open to marketers who would be able to position their products in terms of cognitive, affective or conative elements as well as segment their consumers in terms of attitudinal profiles.

Of course, this exploratory research has a number of limitations. In particular, student samples are often suspect in research. However, given their competence and importance as a target market for both products used in this study, the use of a student sample appeared justifiable here.

Attitudinal profiles such as the ones identified here in the case of two product categories provide only a limited picture of what this approach might be able to do. Clearly, future research should explore further the potential contribution of this type of approach to the examination of consumers’ attitudinal components. For instance, other consumer variables may provide useful descriptors of consumer segments based on attitudinal profiles. Whilst this study suggested only one discriminant factor (gender), it may well be that for other products, other operational variables such as income, education or household size could prove useful for marketers wanting to reach target consumers exhibiting a given attitudinal profile.

REFERENCES

Alba J. and Hutchinson J.W. (1987), "Dimensions of Consumer Expertise", Journal of Consumer Research, 13, 411-455.

Anand P., Holbrook M.B. and Stephens D. (1988) "The formation of affective judgements: the cognitive-affective model versus the independance hypothesis", Journal of Consumer Research, 15, 386-391.

Bagozzi R. P. and Foxall G.R. (1996) "Construct validation of a measure of adaptative-innovative cognitive styles in consumption", International Journal of Market Research, 13, 201-214.

Bieri J. (1961)"Cognitive Simplicity as a personality Variable in Cognitive and Preferential Behavior", In D.W. Fiske and S.R. Maddi eds, Functions of Varied Experiences, Homewood, Ill., The Dorsey Press, 355-379.

Breckler S.J. (1994) "Memory from the experience of donating blood:Just how bad was it?", Journal of Basic and Applied Social Psychology, 14(4), 467-488.

Breckler S.J. and Wiggins E.C. (1989), "Affect vs evaluation in the structure of attitudes", Journal of Experimental Social Psychology, 25(3), 253-271.

Brown H. and Brucker (1990)" Charting the Industrial Buying Stream", Industrial Marketing Management, 19, 55-61.

Caroll J.D, et J.J Chang (1970), "Analysis of individual differences in multidimensional scaling via an n-way generalization of " Eckart-Young " decomposition ", Psychometrika, 35, 238-319.

Claeys C.,A. Swinnen, et P.V. Abeele (1995), "Consumer’means-end chains for "Think " and "Feel " products ", International Journal in Marketing, 12, 193-208.

Cooper L.G. (1983), "A Review of Multidimensionnal Scaling in Marketing Research ", Applied Psychological Measurement, 7, 427-450.

Creusen M.E.H., J.P.L. Schoormans (1997), "The nature of differences between similarity and preference judgements, a replication with extension ", International Journal of Marketing Research, 14, 81-87.

Green P.E. (1975) " On the Robustness of MultiDimensionnal Scaling Techniques", Journal of Marketing Research, 12, 73-81.

Green P.E., Carmone F.J., Smith S.M. (1989), Multidimensional Scaling: concepts and Applications, Allyn and Bacon, Boston.

Hirschman E.C. and Holbrook M.B. (1982), "Hedonic Consumption:emerging concepts, methods and propositions", Journal of Marketing, 46, 92-101.

Holbrook M.B., Moore W.L. et Winer R.S., "Constructing joint spaces from pick-any data: A new tool for consumer analysis, Journal of Consumer Research, 9, 99-105.

Hustard T.P, Mayer C.S. and Whipple T.W. (1975), "Consideration of Context differences in Product Evaluation and Market Segmentation", Journal of the Academy of Marketing Science, 3, 435-439.

Klastorin T.D. (1983), "Assessing Cluster Analysis results", Journal of Marketing Research, 20, 1, 92-98.

Kruskal J.B. (1964), " Multidimensional scaling by optimizing goodness of fit to a non-metric hypothesis", Psychometrika, 29, 1-27.

Laurent G. and Kapferer, J-N (1985) "Measuring Consumer Involvement Profiles", Journal of Marketing Research, 22, 41-53.

Lefkoff-Hagius R., C.H. Mason (1993), "Characteristics, beneficial, and image attributes in consumer judgements of similarity and preference ", Journal of Consumer Research, 20, 100-110.

Lehman D.R. (1971) "Television Show Preferences: Application of a Choice Model" Journal of Marketing Research, 8, 47-55.

Lockshin L.S., Macintosh G. et Spawton, T. (1998), "Using Product Brand and Purchasing Involvement for Retail Segmentation", Journal of Retailing and Consumer Services, 4, 1, 171-183.

Malhotra N.K., Jain A.K. and Pinson C. (1988), "The Robustness of MDS Configarations in the Case of Incomplete Data", Journal of Marketing Research, 25, 95-102.

Milligan G.W. (1981), "A Monte Carlo Study of Thirty internal validity criterion measures for cluster analysis", Psychometrika, 46, 2, 40-47.

Miniard P.W. and Barone M.J. (1997), "The Case for Noncognitive Determinants of Attitude: A Critique of Fishbein and Middlestadt", Journal of Consumer Psychology, 6(1), 39-44.

Moore W.L., Pessemier E.A. and Little T.E (1979), "Predicting Brand Purchase Behaviour: Marketing Application of the Schonemenn and Wang Unfolding Model" Journal of Marketing Research, 9, 206-210.

Ohanian, Roobina and Armen Tashchian, (1992), "Consumers’ Shopping Effort and Evaluation of Store Image Attributes: The Roles of Purchasing Involvement and Recreational Shopping Interest," Journal of Applied Business Research, 8 (6), 40-49.

Park C.W. et M. Young (1985), " Consumer responses to television commercials: The impact of involvment and background music on brand attitude formation", Journal of Marketing Research, 2, 11-24.

Pessemier E.A. (1975)" Market Structure Analysis of New Product and Market Opportunity" Journal of Contemporary Business, 4, 35-67.

Rosenberg M.J. and Hovland L. 1960) Attitude Organisation and Change: An Analysis of Consistency Among Attitude Components, Yale University Press, New Haven.

Schaninger C.M. and Sciglimpaglia D. (1981) "The Infleunce of Cognitive Personality Traits and demographics on Consumer Information Acquisition", Journal of Consumer Research, 8, 208-216.

Sojka J.Z. et J.L. Giese (1997), "Thinking and/or Feeling: An Examination of Interaction Between Processing Styles ", Advances in Consumer Research, 24, 438-442.

Spence I. and Oglivie J.C. (1973) "A Table of Expected Stress Values from Random Rankings in Nonmetric Multidimensional Scaling", Multivariate Behavioural Research, 8, 551-558.

Torgerson W.S. (1952), " Multidimensional scaling: I. Theory and method ", Psychometrika, 17, 401-419.

Young F.W., R.M. Hamer (1987), Multidimensional scaling: History, theory, and applications. Hillsdale, N.J.: Lawrence Erlbaum Associates.

Vaughn R. (1980)," How Advertising Workds: A Planning Model", Journal of Advertising Research, 20, 27-33.

Whipple T.W. (1976), "Variation Among Multidimensional Scaling Solutions: An Examination of the Effect of Data Collection Differences", Journal of Marketing Research, 13, 98B103.

Zajonc R.B. (1968) "Attitudinal effects of mere exposure", Journal of Personality and Social Psychology, 9, 2(2), 1-28.

Zajonc, R.B. et H. Markus (1982), "Affective and Cognitive Factors in Preferences ", Journal of Consumer Research", 9, 123-131.

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