Advances in Consumer Research Volume 17, 1990 Pages 135-143
THE ROLE OF TANGIBLE AND INTANGIBLE ATTRIBUTES IN SIMILARITY AND PREFERENCE JUDGMENTS
Roxanne Lefkoff-Hagius, University of North Carolina at Chapel Hill
Charlotte H. Mason, University of North Carolina at Chapel Hill
Consumer judgments of both similarity and preference are widely used in marketing research. Whereas preference judgments are generally assumed to be subjective and heterogeneous across consumers, similarity judgments are commonly assumed to be objective and homogeneous. However, there is increasing evidence that, like preferences, similarity judgments may be individual specific. In this paper, the results of an experiment examining individual differences in preference and similarity judgments are reported. Using decompositional models for both judgment types, we found substantial heterogeneity across subjects. Moreover, we hypothesized that the "type of attribute" would explain some of the difference found, within subject, between preference and similarity judgments for the same object. Specifically, our results support "intangible" attributes being relatively more important in preference than similarity judgments, whereas "tangible" attributes are relatively more important in similarity judgments.
Understanding how consumers perceive and evaluate products relative to other products is fundamental to marketing strategy. "All aspirins are the sameCwhy pay more?" and 'This car is completely different from any other car in its class" are examples of advertising campaigns that have used similarity and dissimilarity positioning strategies to influence choice (Tversky 1972).
Marketing researchers use both similarity and preference judgments to study consumer behavior. While preference judgments are viewed as individual specific, usually similarity judgments are assumed to be objective, error free, and homogeneous across subjects (Green 1975). In many multidimensional scaling applications, it is common practice to aggregate consumers' similarity judgments in order to formulate "perceptual maps" which are geometric representations of underlying perceptual dimensions. However, aggregating these judgments assumes that subjects perceive the same set of underlying dimensions, subjects make the same judgments about the extent to which objects possess particular attributes, and subjects assign the same importance weights to the attributes in determining the similarities (Jackson 1983).
To relax the last assumption, the INDSCAL multidimensional procedure (Carroll and Chang 1970, Carroll 1972) was developed. It handles individual perceptions by stretching and shrinking the axes of the spatial representation according to the importance of each dimension to the individual. For example, using INDSCAL, Wish (1971) found that subjects have different perceptions about the similarity of various countries. Subjects classified as "hawks" tended to use the political alignment of countries (pro-western versus pro-communist) as the most important dimension, while "doves" tended to use the economic development of countries (underdeveloped versus economically developed) as the most important dimension. Likewise in marketing, INDSCAL has been used to show that consumers perceive the proximity between products differently. However, marketing variables such as familiarity with the produces, personal values, and demographics have not Been successful in explaining-these individual differences in perceptions (Ritchie 1974).
Another common assumption that is central to joint space model configurations (Green and Carmone 1970, Green and Wind 1973) is that consumers' perceptions and preferences are based on the same underlying structural dimensions. There is evidence to suggest that this, too, is an oversimplification. Wish (1971) found that "... the dimensions of liking do not agree so well with the dimensions of similarity...subjects frequently like one and dislike the other of two countries they perceive to be quite similar. These results are somewhat counter to the frequently made assumption that preferences can be directly related to the dimension of a space based on perceptions of the stimuli. " (page 325)
In a marketing study dealing with mothers' perceptions and preferences for toys for their children, Whipple (1976) compared different data collection techniques for similarity and preference measures. He found that "solutions based on attribute ratings/rankings and directly judged proximities are not similar to preference based solutions. " (pages 101-2). This lead him to conclude that "... it is risky to assume that product preference is consistent with overall product similarity. " (page 102).
In this study, we will not assume that individual perceptions are homogeneous across subjects. Instead, the purpose of this study is to attempt to understand, at the individual level, differences between what is important in similarity judgments and what is important in preference judgments. We will explore how "kind of attribute" may account for some of the difference between similarity and preference judgments
TANGIBLE AND INTANGIBLE ATTRIBUTES
Marketing researchers have used a wide variety of attribute descripters to obtain measures of consumers' perceptions and preferences (Myers and Shocker 1981). A comprehensive summary of eleven different attribute typologies is presented by Finn (1985). What is common to all these different approaches is that there is a distinction between the concrete, physical, objective, or TANGIBLE attributes of a product and the abstract, beneficial, subjective, or INTANGIBLE attributes of a product. For example, a car can be described as large, red and luxurious. Large and red are fundamentally tangible attributes as they describe physical properties of the car, while luxurious is primarily an intangible attribute as it describes beneficial and imagery aspects of the car.
It is believed that managers should ultimately formulate their strategies based on consumers' perceptions of the abstract benefits of products (Haley 1968). Yet, in reality, it is necessary to use concrete attributes that are "actionable" or meaningful to managers (Shocker and Srinivasan 1979). Recently there has been some concern and criticism that too much attention has been spent on tangible attributes instead of intangible attributes (Holbrook and Hirschman 1982, Hirschman 1983). In the "layers of meaning paradigm" it is proposed that "the meaning of a product stimulus is a mixture of objective properties and subjective associations" (Hirschman 1980, page 12). Thus, the perceived meaning of a product stimulus is proposed to be a joint construct of both tangible and intangible attributes. Therefore, this study considers perceptions of and preferences for product profiles described by both tangible and intangible attributes. Since attempts to categorize attributes as one or the other may result in fuzzy sets (Myers and Shocker 1981), the distinction between tangible and intangible is viewed as a continuum rather than a dichotomy.
JUDGMENTS OF PREFERENCE AND SIMILARITY
Economists and marketers theorize that consumers do not purchase goods for the goods themselves, but for the satisfaction derived from using the goods (Lancaster 1966 and 1971, Haley 1968, Ratchford 1975). Multiattribute models of consumer preferences partition product satisfaction into satisfaction for individual attributes of the product. Both compositional methods (Wilkie and Pessemier 1973) and decompositional methods (Green and Srinivasan 1978) are based on the notion that attribute utilities underlie consumer preferences. Regardless of the method, the validity of the results depends critically on the choice of appropriate attributes.
If attributes that are important to the consumer are not included, then preference predictions may be suspect. For example, when New Coke was developed, the "taste" attribute was the primary focus. However, loyal Coke drinkers responded negatively to the new product because intangible imagery attributes were probably driving their preferences more than the tangible taste attribute. In general there is a growing realization that, in many product classes, where there are few "meaningful" differences between brands and more so-called "parity" products, image and other intangibles become more important (Business Week, 1983).
There have been a wide variety of approaches to model similarity judgments. These methods include multidimensional scaling, factor analysis, discriminant analysis, clustering (Hauser and Koppelman 1979), and a decompositional approach analogous to conjoint analysis typically associated with preferences (Green and DeSarbo 1978). Some of these approaches use judgments about holistic stimuli whereas others require prior identification of specific attributes. For the latter approach, just as with preference judgments, it is important to identify the appropriate attributes.
While attitude theories (Fishbein 1967, Rosenberg 1956) have provided a strong base for understanding preference judgments in applied marketing, there has not been much theoretical work to understand consumer similarity judgments. Recently there have been some empirical studies to examine how consumers perceive and compare products and product classes (Johnson 1984, Johnson and Fornell 1987, Johnson 1988). There is evidence to suggest that consumers engage in hierarchial processing when making comparison type of judgments. Johnson et al explain that consumers start out with concrete attributes and go to more abstract attributes as alternatives become more noncomparable. Using this hierarchial processing framework, we speculate that if subjects are presented with both tangible (concrete) and intangible (abstract) attributes for the same product class, the tangible attributes would receive more attention and thus be more important in similarity comparison tasks than in preference tasks. Alternatively, because preferences are driven primarily by intangible benefits, we speculate that the intangibles would receive more attention and thus be more important in preference tasks than in similarity comparison tasks.
Based on the above discussion of preference and similarity judgments, we hypothesize the following:
H1a: Intangible attributes will receive relatively more importance weight in preference judgments than in similarity judgments.
H1b: Tangible attributes will receive relatively more importance weight in similarity judgments than in preference judgments.
These hypotheses were tested in the following experimental setting.
43 undergraduate business students at a major university participated in a study dealing with perceptions and preferences for jobs and cars. These product classes were familiar to the subjects and have been used in other studies (Green and Wind 1973; Green, Carroll, and DeSarbo 1981; Montgomery and Wittink 1979). For each product category, a set of four attributes, each at two levels, was developed to form hypothetical full-profile descriptions. Each product was described by two tangible and two intangible attributes. Because the goal of this research was to examine changes in the relative role of tangible and intangible attributes between the two types of judgments, the attributes and levels selected for use in this study were based on limited pretesting. A more comprehensive and formal approach for selecting attributes (Alpert 1971) was not implemented because forecasting choices was not the primary focus of this research. Table 1 shows the attribute levels and Figures 1 and 2 show sample profiles. For both product categories, a full factorial design was used, resulting in sixteen product profiles for each product class.
ATTRIBUTES AND LEVELS
Each subject completed four rank ordering tasks. These involved similarity and preference judgments for each of the two product classes. To counterbalance any order of task effects, subjects were divided into two groups and the order of the judgment tasks was reversed between the groups. Group 1 subjects (N=22):
1. rank ordered the 16 job profile cards according to their similarity to Figure 1.
2. rank ordered the 16 car profile cards according to their similarity to Figure 2.
3. rank ordered the (identical) 16 job profile cards according to personal preference.
4. rank ordered the (identical)16 car profile cards according to personal preferences.
Likewise, Group 2 subjects (N=21) alternated between job and car profiles, but the order of similarity and preference judgments was reversed Group 2 subjects performed tasks 3 and 4 from above and then performed tasks 1 and 2. Thus, Group 2 subjects made preference judgments first and similarity judgments second. The product classes were alternated to provide an intervening task between the similarity and preference judgments for the same product class.
Because different methods of collecting data and estimating attribute importance weights have been found to show low levels of convergence (Jaccard, Brinberg, and Ackerman 1986), this study used the same data collection technique for both similarity and preference tasks. By having subjects perform a rank ordering task for both similarity and preference judgments, differences in the resulting attribute importance weights could then be attributed to differences in the similarity and preference judgments and not be confounded with differences in methods.
Collecting rank orders of product profile cards is a common practice for obtaining preference data. Using conjoint analysis, the importance weights of the attributes and partworth utility functions can be estimated. However, determining the importance weights of the attributes in similarity judgments is not as straightforward. Whereas preferences have meaning for a single item, similarities are only meaningful between pairs of items. For example, a person may have a certain preference value or utility for a 1989 BMW. However, the similarity of the 1989 BMW is meaningful only in relation to another item, say a 1989 Honda. To obtain similarity judgments, comparisons between product profiles were necessary.
JOB PROFILE CARD
CAR PROFILE CARD
There has been limited research applying conjoint analysis to similarity data. Following the approach of Green and DeSarbo (1978) and Green, Rao, and DeSarbo (1978), data were collected to permit additive decomposition of overall similarity judgments into "part-similarity" functions, which are analogous to part-worth functions (Shocker and Srinivasan 1979). To accomplish this, each of the 16 unique profile cards were compared to one referent profile card. While Green et al asked subjects to compare profiles to imaginary, ideal products, instead we asked subjects to compare profiles to a specified, referent profile (which was the same for each subject). This was done so that similarity judgments would not be confounded by any preference insinuations suggested by the term "ideal" product. In addition, the referent profile for both product classes was identical to one of the 16 profiles to be sorted. Every subject specified the card that perfectly matched the referent as the most similar, indicating that subjects understood and followed the directions.
ANALYSIS AND RESULTS
Attribute Importance Weights
Conjoint analysis was used on both the similarity and preference data for each subject. The monotonic analysis of variance (MONANOVA) procedure- was used to estimate attribute importance weights (Kruskal and Carmone 1969). Each subject rank ordered 4 sets of product profiles, so 4 MONANOVA models were computed for each subject, yielding a total of 172 models.
The MONANOVA model for the preference judgments is:
pj = preference of an individual for object j.
fpk = part-worth of kth attribute.
Yjk = level of attribute k for object j.
The MONANOVA model for the similarity judgments is:
Sj = similarity of object j to referent (according to individual).
fsk = part-similarity of kth attribute.
Yjk = level of attribute k for object j (same as above).
SUMMARY STATISTICS FOR ATTRIBUTE IMPORTANCE WEIGHTS
The estimation procedure resulted in an average stress of 10.3 % for the similarity data and 7.3 % for the preference data (0% is best fit and 100% is worse fit). Scatterplots of the best monotonic function of the estimated data values verses the original data values also showed that the preference data had a slightly better fit than the similarity data. Since subjects probably are more familiar with making preference than similarity judgments, this seems reasonable.
The output from the MONANOVA analysis was rescaled so that each set of importance weights would sum to 100 per cent. Summary descriptive statistics for the rescaled importance weights are shown in Table 2. Importance weights ranged widely and, overall, there is no indication that any particular attribute was dominant in either similarity or preference judgments. If other attributes and levels had been used, the estimated importance weights would have been different. However, this research is not concerned with the absolute value of importance weights. Instead, this research is concerned with differences in computed importance weights between the two types of judgment tasks. For the purpose of this research, it does not matter in an absolute sense whether tangibles or intangibles are more important, because this will depend on the particular product class and the particular subject. What does matter is whether importance weights change between similarity and preference judgments.
From the standard deviations of the importance weights, it is clear that the importance weights of the similarity judgments have just as much variance around their means as the importance weights of the preference judgments. This supports the notion that just as preference judgments are individual specific, so too are similarity judgments.
Multivariate Analysis of Variance
A repeated measures multivariate analysis of variance (MANOVA) procedure was implemented to test the hypotheses that the type of attribute would explain some of the difference found, within subject, between preference and similarity judgments for the same object. The between subjects factor was "order" as half the subjects made similarity judgments first, while half the subjects made preference judgments first. Within subject factors were "product" and "task". Products were jobs and cars. Tasks were preference and similarity judgments. The design was crossed because each subject made both similarity and preference judgments for both jobs and cars.
Since the hypotheses concern the weight of tangibles relative to the weight of intangibles, the absolute values of the importance weights themselves are not the focus of this study, as those values will vary from individual to individual and product class to product class. Instead, our interest centers on how the weights vary between similarity and preference judgments. Thus, to define the dependent variables, we focus on the total weight given to tangibles versus intangibles across the different judgment tasks for each of the two product classes.
Four dependent measures were computed for each subject. Each dependent variable is defined as the difference between the weights given to tangible attributes minus the weights given to intangible attributes. For example, Y1 is the sum of importance weights for tangibles in preference judgments minus the sum of importance weights for intangibles in preference judgments for the job product class. Y2 also considers jobs, but deals with the difference between tangibles and intangibles in similarity judgments. Likewise, Y3 considers the preference judgments and Y4 considers the similarity judgments for the car product class.
CELL MEANS AND STANDARD DEVIATIONS
Since Y1, Y2, Y3, and Y4 are defined as the weights given to tangibles minus the weights given to intangibles, if the value of the variable is negative, more weight was given to intangibles. If the value is positive, more weight was given to tangibles. The means and standard deviations for Y1 through Y4 for each group are summarized in Table 3. Plots of the cell means are presented in Figure 3. Of particular interest are differences between Y1 and Y2, and between Y3 and Y4. The mean values for Y1 and Y3 which involve preferences are lower than the means for Y2 and Y4, respectively, which involve similarities. Thus, it appears that intangibles were given relatively more weight in preference judgments compared with similarity judgments. Or, looking at it from the other perspective, it appears that tangibles were given relatively more weight in similarity than preference judgments.
To test the significance of these effects, a repeated measure MANOVA was conducted. The results are presented in Table 4. The only significant factor was the "task" effect (F(1,41) = 6.33, p < 0.016). Thus, the results support the hypotheses. For both jobs and cars, the intangible attributes received relatively more importance weight in the preference judgments than in the similarity judgments. The tangible attributes received relatively more importance weight in the similarity judgments than in the preference judgments. This effect was significant, regardless of the order of the judgment task. None of the interactions were significant.
Although the focus of the study was on the F role of attribute type in similarity and preference i judgments, the results do corroborate the notion that similarity judgments are not homogeneous across people. While subjects read the same attribute information and compared it to the same referent, there was a wide range of resulting attribute importance weights. So overall, there was just as [ much heterogeneity in-similarity judgments as in preference judgments.
Not only was there a difference between : resulting attribute importance weights in similarity t and preference judgments across individuals, but also F within individuals there was a difference. At the individual level, the resulting attribute weights in i the similarity judgments differed from the attribute weights in the preference judgments. The type of attribute was found to be a significant factor contributing to this difference. Intangible attributes received relatively more weight in preference judgments than in similarity judgments, while tangible attributes received relatively more weight in 0 similarity than in preference judgments.
The results presented in this paper are based on an exploratory pilot study using a limited number of subjects and product classes. Further work is planned to extend the analysis by using more subjects and product classes. Another limitation of 0 this study is that attributes were designated as tangible and intangible without verification that the specific subjects performing the tasks agreed with the labeling. A formal manipulation check is planned in future experiments. In addition, we note that since the experiment used a small number of attributes to describe alternatives, the results are not necessarily applicable to real applications involving jobs and cars. However, because the purpose of this study was to understand the relative role of tangibles and intangibles, the number of attributes was limited to four in order to keep the task fairly simple for the subjects and thus maximize internal validity.
The inclusion of intangible attributes into a conjoint framework is not without critics. In a survey of conjoint users in commercial settings (Cattin and Wittink 1982), some respondents reported that they did not use conjoint analyses when the attributes to define the stimulus objects tended to be "soft." While there are some unitary stimuli such as foods, scents, and aesthetic objects that cannot be easily separated into component parts (Huber 1987), many consumer products can be described in terms of both physical characteristics and intangible benefits and there are conjoint studies that do incorporate both kinds of attributes (i.e. Currim, Weinberg, and Wittink 1981). In fact, the results of the most recent survey of conjoint users (Wittink and Cattin 1989) shows there to be an increase in conjoint applications involving services, and intangibles are particularly important in services.
In many marketing applications, data from similarity and preference tasks are used interchangeably. However, the results from this study suggest that, depending on the judgment task, different kinds of information are more or less important. Since the kind of information presented to consumers is a controllable marketing variable, there are interesting methodological and managerial implications. For instance, the tangible-intangible attribute framework could explain the low convergent validity of different approaches for estimating attribute important weights (Jaccard, Brinberg, and Ackerman 1986). The tangible-intangible attribute framework could also explain the low predictive validity of some new product forecasting applications (Alter 1987). Particularly for managers formulating positioning strategies to alter how consumers perceive and evaluate their products relative to the competition, it is important to distinguish between tangible and intangible attributes. "Me too" products that copy only the tangible attributes of a brand leader's product may ultimately fail to get market share if consumer preferences are primarily driven by the intangible attributes.
Alpert, Mark I. (1971), "Identification of Determinant Attributes: A Comparison of Methods," Journal of Marketing Research, 8 (May), 184-191.
Alter, Stewart (1987), "Beecham Sues Saatchi Over Research Forecasts," Advertising Age, July 19. 1.
Business Week (1983), "Marketing: The New Priority, " (November).
Carroll, J. Douglas (1972), "Individual Differences and Multidimensional Scaling, " in Multidimensional Scaling: Theory and Applications in the Behavioral Sciences, Vol 1, eds. Roger N. Shepard, A. Kimball Romney, and Sara Beth Nerlove, New York: Seminar Press.
Carroll, J. Douglas and J.J. Chang (1970), "Analysis of Individual Differences in Multidimensional Scaling Via an N-way Generalization of "Eckart-Young Decomposition," Psychometrika, 35 (September), 283-319.
Cattin, Philippe and Dick R. Wittink (1982), "Commercial Use of Conjoint Analysis: A Survey," Journal of Marketing, 46 (Summer), 4453.
Currim, Imran S., Charles B. Weinberg, and Dick R. Wittink (1981), "Design of Subscription Programs for a Performing Arts Series," Journal of Consumer Research, 8 (June), 67-75.
Finn, Adam (1985), "A Theory of the Consumer Evaluation Process For New Product Concepts," in Jagdish N. Sheth, eds., Research in Consumer Behavior, Vol. 1, Greenwich, CT: JAI Press, 35-65.
Fishbein, Martin (1967), "A Behavior Theory Approach to the Relations between Beliefs about an Object and the Attitude towards the Object," in M. Fishbein (eds.) Readings in Attitude and Theory Measurement, New York: John Wiley & Sons, Inc., 389-399.
Green, Paul E. (1975), "Marketing Applications of MDS: Assessment and Outlook," Journal of Marketing, 39 (January), 24-31.
Green, Paul E. and Frank J. Carmone (1970), Multidimensional Scaling and Related Techniques in Marketing Analysis, Boston, MA: Allyn and Bacon.
Green, Paul E., J. Douglas Carroll, and Wayne DeSarbo (1981), "Estimating Choice Probabilities in Multiattribute Decision Making," Journal of Consumer Research, 8 (June), 76-84.
Green, Paul E and Wayne S. DeSarbo (1978), "Additive Decomposition of Perceptions Data Via Conjoint Analysis," Journal of Consumer Research, 5 (June), 58-65.
Green, Paul E., Vithala R. Rao, and Wayne S. DeSarbo (1978), "Incorporating Group-Level Similarity Judgments in Conjoint Analysis," Journal of Consumer Research, S (December), 187-193.
Green, Paul E. and V. Srinivasan (1978), "Conjoint Analysis in Consumer Research: Issues and Outlook," Journal of Consumer Research, 5 (September), 103-123.
Green, Paul E. and Yoram Wind (1973)* Multiattribute Decisions in Marketing: A Measurement Approach, Hinsdale, IL: Dryden Press.
Haley, Russell I. (1968), "Benefit Segmentation: A Decision Oriented Research Tool," Journal of Marketing Research, 32 (July), 30-35.
Hauser, John R. and Frank S. Koppelman (1979), "Alternative Perceptual Mapping Techniques: Relative Accuracy and Usefulness," Journal of Marketing Research, 16 (November), 495-506.
Hirschman, Elizabeth C. (1980), "Attributes and Layers of Meaning," Advances in Consumer Research, eds. Jerry Olson, 7, 7-12.
Hirschman, Elizabeth C. (1983), "Aesthetics, Ideologies and the Limits of the Marketing Concept," Journal of Marketing, 47 (3), 45-55.
Holbrook, Morris B. and Elizabeth C. Hirschman (1982), "The Experiential Aspects of Consumption: Consumer Fantasies, Feelings, and Fun," Journal of Consumer Research, 9 (September), 132-140.
Huber, Joel (1987), "Conjoint Analysis: How We Got Here and Where We Are," Sawtooth Software Conference Proceedings, 237-251.
Jaccard, James, David Brinberg, and Lee J. Ackerman (1986), "Assessing Attribute Importance: A Comparison of Six Methods," Journal of Consumer Research, 12 (March), 463468.
Jackson, Barbara Bund (1983), Multivariate Data Analysis, Homewood, IL: Richard D. Erwin, Inc.
Johnson, Michael D. (1984), "Consumer Choice Strategies for Comparing Noncomparable Alternatives," Journal of Consumer Research, 11 (December), 741-753.
Johnson, Michael D. (1988), "Comparability and Hierarchical Processing in Multialternative Choice," Journal of Consumer Research, 15 (December), 303-314.
Johnson, Michael and Claes Fornell (1987), "The Nature and Methodological Implications of the Cognitive Representation of Products," Journal of Consumer Research, 14 (September), 214-228.
Kruskal, J. B. and Frank J. Carmone Jr. (1969), MONANOVA, A Fortran Program for Monotone Analysis of Variance, Bell Telephone Laboratories, Murray Hill, NJ.
Lancaster, Kelvin T. (1966), "A New Approach to Consumer Theory," Journal of Political Economy, 74, 132-157.
Lancaster, Kelvin T. (1971), Consumer Demand: A New Approach, New York: Columbia University Press.
Montgomery, David B. and Dick R. Wittink (1979), "The Predictive Validity of Conjoint Analysis for Alternate Aggregation Schemes," in D.B. Montgomery and D.R. Wittink (eds.) Market Measurement and Analysis: Proceedings of the 1979 ORSA/TIMS Conference of Marketing, Cambridge, MA: Marketing Science Institute, 298-309.
Myers, James H. and Allan D. Shocker (1981), "The Nature of Product-Related Attributes," in Jagdish N. Sheth, eds., Research in Marketing, Vol. 5, Greenwich, CT: JAI Press, 211-236.
Ratchford, Brian T. (1975), "The New Economic Theory of Consumer Behavior: An Interpretative Essay," Journal of Consumer Research, 2, (September), 65-75,
Ritchie, J.R. Brent (1974), "An Exploratory Analysis of the Nature and Extent of Individual Differences in Perception, " Journal of Marketing Research, 11 (February), 41-49.
Rosenberg, Milton J. (1956), "Cognitive Structure and Attitudinal Affect," Journal of Abnormal and Social Psychology, 53, 367-372.
Shocker, Allan D. and V. Srinivasan (1979), "Multiattribute Approaches for Product Concept Evaluation and Generation: A Critical Review," Journal of Marketing Research, 16 (February), 159-180.
Tversky, Amos (1972), "Elimination by Aspects: A Theory-of Choice," Psychological Review, 79 (July), 281 -299.
Whipple, Thomas W. (1976), "Variation Among Multidimensional Scaling Solutions: An Examination of the Effect of Data Collection Differences," Journal of Marketing Research, 13 ( February), 98-103.
Wilkie, William L. and Edgar A. Pessemier (1973), "Issues in Marketing's Use of Multi-Attribute Attitude Models, " Journal of Marketing Research, 10 (November), 428441.
Wittink, Dick R. and Philippe Cattin (1989), "Commercial Use of Conjoint: An Update," Journal of Marketing, 53 (July), 91-96.
Wish, Myron (1971), "Individual Differences in Perception and Preferences Among Nations," in C.W. King and D. Tiger, eds., Attitude Research Reaches New Heights, Chicago: American Marketing Association, 312-328.