On the Nature of Product Attributes and Attribute Relationships

Michael D. Johnson, University of Michigan
ABSTRACT - Consumer researchers have come to regard product attributes as relatively concrete, objective properties while using a number of ill-defined and overlapping terms to label a product's more abstract qualities. The present paper argues for a unification of these distinctions into a single concrete to abstract attribute continuum along which attributes are instrumentally, reflectively, or vicariously related. The proposed framework emphasizes the complex structure of consumers' product knowledge over a labeling of knowledge content. A general methodology and a potential scaling technique for studying attribute relationships are briefly outlined.
[ to cite ]:
Michael D. Johnson (1989) ,"On the Nature of Product Attributes and Attribute Relationships", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 598-604.

Advances in Consumer Research Volume 16, 1989      Pages 598-604

ON THE NATURE OF PRODUCT ATTRIBUTES AND ATTRIBUTE RELATIONSHIPS

Michael D. Johnson, University of Michigan

ABSTRACT -

Consumer researchers have come to regard product attributes as relatively concrete, objective properties while using a number of ill-defined and overlapping terms to label a product's more abstract qualities. The present paper argues for a unification of these distinctions into a single concrete to abstract attribute continuum along which attributes are instrumentally, reflectively, or vicariously related. The proposed framework emphasizes the complex structure of consumers' product knowledge over a labeling of knowledge content. A general methodology and a potential scaling technique for studying attribute relationships are briefly outlined.

INTRODUCTION

Consumers describe products using a vast array of attributes. Although attributes are commonly defined to include inherent characteristics and ascribed qualities, consumer researchers tend to adopt a narrow view of attributes as concrete and objective. The more abstract characteristics or qualities that consumers associate to products have, in turn, been described using a wide range of loosely defined and overlapping terms, including characteristics, functional consequences, psychosocial consequences, benefits, instrumental values, experiential aspects, and terminal values.

Recent research suggests that this proliferation of terminology is both unnecessary and limiting. The present paper argues for a unification of the existing range of concepts and terminology into a single concrete to abstract attribute continuum. The paper begins by contrasting alternative views of product attributes and then, using attribute abstraction as an organizing principle, describes a framework for studying the inherently complex structure of product attribute relationships. The paper ends with a brief discussion of a general methodology for examining the framework and a potential scaling technique for identifying attribute relationships.

ATTRIBUTES, CONSEQUENCES, AND VALUES

Our narrow view of attributes as concrete properties can be traced to at least two developments in consumer research. One was the proliferation of process tracing research in the 1970's (see Bettman 1986 for a review). These studies, many of which relied on brand by attribute matrices or information boards (Jacoby et al. 1976), examined how consumers used relatively concrete attributes, such as the cooking levels and capacity of microwave ovens (Bettman and Park 1980), to evaluate and choose among brands. It was natural for these studies to focus, at least initially, on consumers' everyday processing of concrete, brand-level information. However, an unintended consequence of this research was the equating of attributes with the concrete, objective properties of products (Holbrook and Hirschman 1982).

Another important development has been the proliferation of terminology designed to distinguish and describe a product's more abstract qualities. In marketing, for example, Haley (1968) tabbed these abstract properties as benefits. In economics, Becker (1976) used the term characteristics to describe the qualities that products provide. The greatest proliferation of terms occurred under the general label of means-end chains. This research draws on Rokeach's (1973) concept of human values, or our enduring beliefs regarding modes of conduct and end states of existence. Rokeach dichotomized "modes of conduct" and "end-states" into instrumental and terminal values, respectively. Instrumental values are considered more concrete values of "doing" while terminal values are are abstract values of "being."

The goal of a means-end chain is to link presumably concrete product attributes with the consequences of those attributes and, in turn, to link consequences with consumers' values or desired end-states (Gutman 1982; Gutman and Reynolds 1979; Olson and Reynolds 1983; Vinson, Scott and Lamont 1977; Young and Feigin 1975). Olson and Reynolds (1983), for example, postulate six levels of concrete to abstract attributes, consequences, and values: concrete attributes > abstract attributes > functional consequences > psychosocial consequences > instrumental values > terminal values. Within this approach, lower levels of the chain are instrumental in affecting higher levels. Typically, attributes are restricted to a very concrete, objective level (for exceptions see Cohen 1979 and Geistfeld, Sproules, and Badenhop 1977).

There are at least three major problems with this approach. First, whether values, consequences, or desired end-states, the distinctions made in a means end chain are not well defined. Just where do attributes- end and consequences begin? Similarly, where do consequences end and values begin? And just how conceptually and empirically distinct are instrumental and terminal values? It is difficult to find clear, working definitions for many of these terms. Second, attributes, consequences, and values are linked instrumentally from the concrete to the abstract. Qualitatively different types of relationships are typically ignored, such as the perceived correlation among attributes at the same level of abstraction (Brunswik 1943) or the reflective, concrete implications of a product's abstract attributes (Cohen 1979). By their very nature means-end chains describe unidirectional relationships from means to ends. And last but not least, means-end chains, by concentrating on the content of knowledge, focus attention away from the structure of knowledge. As a result, means-end chains fail to provide a solid foundation for theoretical development. Although they may offer significant pragmatic value (see, for example, Gutman and Alden 1985), their ability to advance our theoretical understanding of the nature of consumer product knowledge is limited.

Theory development in psychology, particularly in the areas of verbal learning (Paivio 1971), memory (Quillian 1968; Collins and Loftus 1975), and categorization (Rosch 1975; Rosch and Mervis 1975; Rosch et al. 1976), has adopted a very different view of attributes. This view is consistent with the general definition of attributes as inherent characteristics or ascribed qualities. Accordingly, attributes are viewed as varying along a continuum from the concrete to the abstract. Abstractness, in this context, is defined as the inverse of how directly an attribute denotes particular objects or events, and is equated with the specificity-generality of terms and the subordination-superordination of category labels (Paivio 1971; Rosch 1975; Rosch et al. 1976). Concrete attributes describe some particular aspect of a product while more abstract attributes more generally and completely describe a product. Similarly, more abstract or superordinate category distinctions encompass larger, more general product groupings.

A central thesis of the present paper is that it is advantageous to simply view attributes as lying on a continuum from the concrete to the abstract, a continuum that encompasses characteristics, consequences, benefits, and values. At least two other constructs are fundamentally tied to this concreteness-abstractness continuum. One is the inherent objectivity-subjectivity of an attribute. Consistent with the above definition of abstractness, concrete attributes are the more objective, direct aspects of a product, while abstract attributes are more indirect and necessarily subjective. Attribute concreteness-abstractness is also critically linked to an attribute's inherent feature-dimensionality (Johnson and Fornell 1987). While abstract attributes often resemble continuous dimensions, concrete attributes may be more akin to dichotomous features. Central to their similarity is the property of inclusion that underlies the two constructs. Abstraction implies a summation or concentration of information. Thus a single abstract attribute (the safety of an automobile) may capture several more concrete product attributes (seat belts, air-bags, brakes). Similarly, a single dimension may be viewed as a set of nested features (Gati and Tversky 1982).

There are several arguments in favor of deemphasizing the vast collection of terms that consumer researchers have borrowed or spawned in favor of a single, concrete to abstract attribute continuum. An attribute continuum avoids the use of overlapping and ill-defined distinctions. Abstraction, meanwhile, is a relatively well-defined organizing principle for theory development (cf. Bettman and Sujan 1987; Howard 1977; Johnson 1984, 1986; Johnson and Fornell 1987; Sujan 1985). A continuum also allows for other than simply instrumental attribute relationships. Finally, there is growing empirical support for a continuum in a consumer products context (Johnson 1984; Johnson and Fornell 1987).

Presuming a continuum of concrete to abstract attributes does not imply that consumers think in terms of a continuum. Operationally, consumers distinguish and describe products at particular levels of the underlying continuum. The important point is that the concept of abstraction can be used to organize the attributes that consumers themselves ascribe to products without imposing attribute, consequence, and value distinctions. Product knowledge may still be meaningfully classified. At a general level, for example, product knowledge includes a product's attributes, the meaning of those attributes, and a product's temporal and spatial information (Anderson 1983; Brucks 1986; Wyer and Srull 1986). Yet there is limited value in imposing a number of ill-defined distinctions on any one aspect of product knowledge, such as product attributes.

A Hierarchical View of Product Attributes

Postulating a single concrete to abstract attribute continuum allows consider research to focus on an important theoretical issue: the inherently complex structure of attribute or propositional knowledge (Anderson 1983, Oden 1987). In particular, Howard's theory of buyer behavior (Howard 1977; Howard and Sheth 1969) uses a concreteness-abstractness continuum to begin to address the critical question of attribute structure.

Following the basic principles of categorization (cf. Rosch 1975), Howard views consumers as systematically grouping and distinguishing products, on the basis of similarity, into hierarchies. Consumers make choices at different levels of these hierarchies or different levels of abstraction. Category-level choices occur at more abstract levels of a hierarchy, while brand-level choices occur at a more concrete level.

An important component of Howard's model is that consumers also form hierarchies of attributes, from the abstract to the concrete, that correspond to their product hierarchies. A major prediction of the model is that consumers select and process attributes at a level of abstraction of their attribute hierarchies that corresponds to their level of choice. Howard thus predicts a direct relationship between the level of abstraction of a choice (e.g., category versus brand) and the abstractness of the choice criteria. Boote (1975), using Rokeach's instrumental and terminal values as proxies for concrete and abstract choice criteria, reports a study that supports this hypothesis. Recently, Johnson and Fornell (1987) found support for a generalized version of Howard's hypothesis. They found the abstractness of descriptive product attributes, not just choice criteria, varying directly with the abstractness of the product. This result is consistent with existing views of human memory (Anderson 1983) and, in particular, the concept of cognitive economy (Collins and Loftus 1975). Another aspect of Howard's attribute hierarchies that is well supported is the notion that descriptive attributes generally decrease with abstraction (Boote 1975; Johnson 1984). Abstraction, by definition, implies a concentration or summation of attribute information.

However, a hierarchical view of attributes and abstraction may be overly restrictive. A strict hierarchy presumes that each lower level attribute maps into or affects but one abstract attribute. At the other extreme, a product's concrete attributes may map into each of several relevant abstract attributes (Rokeach 1973). Johnson (1986) recently modeled the attribute abstraction process as lying somewhere at or between these two extremes. It is quite natural, for example, to expect an automobile's size to affect not only its safety but its economy and prestige.

The next section of the paper advances a framework for studying attributes and their relationships. The goal is to integrate qualitatively different attribute relationships, from Howard's notion of an attribute hierarchy to Brunswik's notion of vicarious functioning, into a single attribute relationship framework.

ATTRIBUTES AND ATTRIBUTE RELATIONSHIPS

The proposed framework is designed to capture the structure of consumers' product attribute knowledge. This use of the term knowledge is similar to both Anderson's (1983) notion of propositional knowledge and Wyer and Srull's (1986) notion of referent bins. Attribute structure is taken to mean the relationships among a product's associated attributes, either within an associative network (Anderson 1983) or an integrated product schema (Wyer and Srull 1986). (The differences between a general associative network and schema models of memory are not critical to our discussion. Both tend to reduce to network relationships (Oden 1987) and are not necessarily incompatible (Wyer and Srull 1986).)

The framework's main organizing principle is the concrete to abstract continuum along which attributes may be described. This continuum is hierarchical in the sense that concrete attributes generally map into more abstract attributes, and the number of relevant attributes decreases with abstraction. At the same time, lower level attributes may map into several higher level attributes, and higher level attributes may infer lower level attributes. Finally, attribute relationships may exist at the same level of abstraction. In other words, product knowledge is only loosely rather than strictly hierarchical.

Within this hierarchy, three qualitatively different types of attribute relationships are postulated to explain the associative and inferential structure of product knowledge: instrumental relationships, reflective relationships, and vicarious relationships. An instrumental relationship, drawing on Rokeach's terminology, exists whenever a relatively concrete attribute is instrumental in affecting or changing a more abstract attribute. Decreasing the size of an automobile may, for example, have a direct, positive effect on fuel consumption which may, in turn, increase economy. Recall that means-end chains are conceptually limited to this type of instrumental relationship.

Reflective relationships, meanwhile, capture the concrete consequences that result from a change in the level of a more abstract attribute. Reflective relationships capture many of our abstract-based inferences regarding a product's concrete attributes, or what Cohen (1979) refers to as derived beliefs. Although noise level may not be instrumental in affecting an auto's economy, it may be very reflective of economy cars. In many cases, reflective relationships are the direct result of one or more instrumental relationships. Size, engine size, and weight may, instrumentally, increase economy which may, in turn, be reflected by an increase in noise level. In contrast to instrumental relationships, reflective relationships run from the abstract to the concrete.

Both instrumental and reflective attribute relationships are directional, or propositional (Anderson 1983), in nature. They capture the perceived change in or existence of an attribute resulting from the change in or existence of some higher or lower level attribute. In contrast, vicarious relationships capture the perceived covariance among attributes. Brunswik (1943; 1956) conceptualized the existence of vicarious attribute or cue relationships within a lens model framework. Given the redundancy among certain attributes, much of the information in one may be inherent in another. This allows individuals to process attributes vicariously in order to achieve their processing objectives. Brunswik coined the term "vicarious functioning" to describe this substitutability of attributes. In our automobile example, consumers may perceive a relationship between an auto's size and its color. Although there may be no direct, functional relationship connecting size and color, many large cars are black while many small cars are red.

There are two important distinguishing characteristics of vicarious attribute relationships. First, while instrumental relationships flow from the concrete to the abstract, and reflective relationships flow from the abstract to the concrete, vicarious relationships exist at the same general level of abstraction (e.g. size and color). This is in keeping with Brunswik's initial concept of vicarious functioning. Second, in many cases, vicarious relationships are the indirect result of some combination of instrumental and/or reflective relationships. While, for example, a large size and a black color may contribute to a car's prestige, a small size and red color may contribute to a car's sportiness. A vicarious relationship between size and color may, therefore, result from their common instrumental effects on more abstract attributes.

These three relationships are hypothesized to capture the majority of a consumer's product attribute knowledge. They operate within and across the levels of abstraction of a generally hierarchical structure. As with any associative model, individual relationships will vary in strength depending on the consumer's experience (Howard 1977) and the consumption situation (Belk 1975). Finally, instrumental and reflective relationships between any two attributes at different levels of abstraction are not mutually exclusive. That is, the association between any two attributes varying in abstractness may be a combination of instrumental and reflective relationships.

The major advantage of the present framework is that it is designed to flexibly capture the entire range of concrete to abstract product attributes and their relationships. The framework is also very parsimonious. It presumes a single continuum of attributes that may be directly or indirectly related via one or more of three possible relationships. The range of ill-deemed and problematic terms used in previous approaches is avoided. And consistent with existing psychological views of memory and cognition (Anderson 1983; Oden 1987), the complexity of consumer knowledge is captured by a complex system of product attribute relationships.

A GENERAL METHODOLOGY

Four general stages are required to operationalize the structure of product attribute knowledge using the proposed framework. In stage one, the researcher must specify the product, product-situation, or schema of interest That is, what part of the consumer's associative network (Anderson 1983), or what referent bin (Wyer and Srull 1986), is relevant to the research? For example, one might be interested in how a change in a beverage's concrete attributes affects it perception on the more abstract attributes that beverages compete on across categories. Does, say, adding calcium to orange juice make it a more attractive thirst quencher or more nutritious breakfast drink? Alternatively, one might simply be interested in gaining a general understanding of a product's attribute associations and their relationships.

In stage two, the range of relevant product or schema related attributes must be generated. Depending on the research focus, these attributes may, for example, be obtained from the protocols of a choice task (Johnson 1984) or the thought listings of a memory probe (Rosch and Mervis 1975). In stage three, these attributes must be operationalized along the concreteness-abstractness continuum. Johnson (1984) describes a procedure whereby judges rate attributes on a scale from O (very concrete) to 10 (very abstract) in order to produce reliable concreteness-abstractness values (see also Johnson and Fornell 1987). Finally, the important relationships must be identified A scaling technique with the potential to capture these relationships is described below.

Scaling Attribute Relationships

Additive tree scaling is one particularly promising method for identifying or uncovering attribute relationships access levels of abstraction. Additive tree procedure such as Sattath and Tversky's ADDTREE (1977). are free tree scaling methods with a path length metric. One of the advantages of additive trees over traditional scaling techniques (i.e., multidimensional scaling and hierarchical clustering) is their ability to capture "nearest neighbor" relationships, where certain stimuli are more central within a stimulus set (Tversky and Hutchinson 1986). According to Tversky Id Hutchinson, centrality exists when a stimulus is more similar to the other stimuli in a set than those stimuli are to each other. Borrowing one of these authors' examples, a fruit is more similar to either an orange or a banana than an orange and banana arc to each other. Notice that centrality is a common property of stimulus sets that combine relatively concrete stimuli with more abstract, superordinate level stimuli. Additive trees capture centrality relationships by placing the more central or abstract stimuLi on shorter branches that are closer to the root of the tree.

Because additive trees are superior at capturing hierarchical stimulus relationships, they may provide considerable insight into hierarchical attribute relationships. To illustrate the possibilities, a sample additive tree scaling for the choice attributes of televisions, estimated using ADDTREE (Sattath and Tversky 1977), is presented in the Figure. The perceived similarity of twenty television attributes that varied widely in their level of abstraction served as input. These attributes were obtained from a previous study in which subjects made choices involving televisions (Johnson 1984) and were selected on the basis of frequency of mention within different ranges of attribute abstraction. A convenience sample of six subjects was asked to provide pair-wise similarity judgments for all possible pairs of the twenty attributes ( 1=190). These input judgments were then scaled using ADDTREE. The attribute concreteness-abstractness ratings (from Johnson 1984) are provided in parentheses.

There are several interesting observations regarding the ADDTREE solution. First, the solution captures the attribute similarities better than either a two- or three-dimensional multidimensional scaling solution. Although additive trees use approximately the same number of parameters as a two-dimensional MDS solution (Carroll 1976), and fewer parameters than a three dimensional solution. Kruskal's stress for the ADDTREE was .10 compared to .21 and .13 for the two- and three-dimensional MDS solutions respectively. Second, there is indeed a relationship between the level of the attributes in the ADDTREE solution and their rated level of abstraction. The distance from the root of the tree to the attribute nodes is significantly negatively correlated with the attribute abstractness ratings (r=-.57, p<01). That is, more abstract attributes are generally closer to the root of the additive tree. Finally, the solution provides insight into potential attribute relationships. Remote control, for example, appears instrumental in affecting the luxury and pleasure derived from a television. Alternatively, a television's warranty may reflect its reliability and durability. Although a superior warranty may not be instrumental in affecting reliability or durability, it may be very reflective of these qualities.

This illustration is by no means a test of the proposed framework. A simple similarity measure is naturally limited Although similarity may be a good indication of the strength of a relationship, it does not delineate the nature of the relationship. A natural direction for future research is to collect both similarity and directional attribute judgments (i.e., for product X, does attribute A imply attribute B?), in conjunction with concreteness-abstractness ratings, in order to delineate instrumentaL reflective, and vicarious attribute relationships.

FIGURE

ADDTREE SOLUTION FOR TELEVISION ATTRIBUTES

CONCLUSIONS

A major argument of the present paper is that attribute abstraction provides a logical organizing principle for the study of product attributes and their relationships. Abstraction is a relatively well defined alternative to the numerous attribute, consequence, benefit. and value distinctions in the consumer behavior literature. A parsimonious framework is described for studying the structure of product attribute knowledge. This framework integrates previously separate perspectives in order to capture instrumental, reflective, and vicarious attribute relationships across and within levels of abstraction. Finally, additive scaling is presented as an alternative to traditional scaling techniques which may aid in uncovering these relationships.

REFERENCES

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

Becker, Gary S. (1976), The Economic Approach to Human Behavior, Chicago: The University of Chicago Press.

Belk, Russell W. (1975), "Situational Variables and Consumer Behavior," Journal of Consumer Research, 2 (December), 157-164.

Bettman, James R. (1986), "Consumer Psychology," Annual Review of Psychology, 37, 257-289.

Bettman, James R. and Jacob Jacoby (1976), "Patterns of Processing in Consumer Information Acquisition," in B. Anderson (ed.), Advances n Consumer Research, 3, 315-320.

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

Bettman, James R. and Mita Sujan (1987), "Effects of Framing on Evaluation of Comparable and Noncomparable Alternatives by Expert and Novice Consumers," Journal of Consumer Research, 14 (September), 141-154.

Boote, Arthur S. (1975),"An Exploratory Investigation of the Roles of Needs and Personal Values in the Theory of Buyer Behavior," unpublished doctoral dissertation, Columbia University.

Brucks, Merrie (1986), "A Typology of Consumer Knowledge Content," in R. J. Lutz (ed.), Advances in Consumer Research, 13, 58-63.

Brunswik, Egon (1943), "Organismic Achievement and Environmental Probability," Psychological Review, 50 (3), 255-272.

Brunswik, Egon (1956), Perception and the Representative Design of Experiments, Berkeley: University of California Press.

Carroll, J. Douglas (1976), "Spatial, Non-spatial, and Hybrid Models for Scaling," Psychometrika, 41 (4). 439-463.

Cohen, Joel B. (1979), The Structure of Product Attributes: Defining Attribute Dimensions for Planning and Evaluation," in Allan D. Shocker (ed.), Analytic Approaches to Product and Marketing Planning, Cambridge, MA: Marketing Science Institute, 54-86.

Collins, Alan M. and Elizabeth Loftus (1975), "A Spreading-Activation Theory of Semantic Processing," Psychological Review, 82 (6), 407-428.

Gati, Itamar and Amos Tversky (1982), "Representations of Qualitative and Quantitative Dimensions," Journal of Experimental Psychology: Human Perception and Performance, 8 (2), 325-340.

Geistfeld, Loren V., George B. Sproules, and Suzanne B. Badenhop (1977), 'The Concept and Measurement of a Hierarchy of Product Attributes," in W. D. Perreault (ed.), Advances in Consumer Research, 4, 302-307.

Gutman, Jonathan (1982), "A Means-End Chain Model Based on Consumer Categorization Processes," Journal of Marketing, 46 (Spring), 60-72.

Gutman, Jonathan and Scott D. Alden (1985), "Adolescents' Cognitive Structures of Retail Stores and Fashion Consumption: A Means-End Chain Analysis of Quality," in J. Jacoby and J. Olson (eds.), Perceived Quality, Lexington, MA: Lexington Books, 99-114.

Gutman, Jonathan and Thomas J. Reynolds (1979), "An Investigation of the Levels of Cognitive Abstraction Utilized by Consumers in Product Differentiation," in J. Eighmey (ed.), Attitude Research Under the SGAS Chicago: American Marketing Association, 128-150.

Haley, Russell I. (1968), "Benefit Segmentation: A Decision Oriented Research Tool," Journal of Marketing, 32 (July), 30-35.

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.

Howard, John A. (1977), Consumer Behavior: Application of Theory, New York: McGraw-Hill.

Howard, John A. and Jagdish N. Sheth (1969), The Theory of Buyer Behavior, New York John Wiley & Sons.

Jacoby, Jacob, Robert Chestnut, Karl Weigl, and William Fisher (1976), "Prepurchase Information Acquisition: Description of a Process Methodology, Research Paradigm, and Pilot Investigation," in B. Anderson (ed.), Advances in Consumer Research, 3, 306-314.

Johnson, Michael D. (1984), "Consumer Choice Strategies for Comparing Noncomparable Alternatives," Journal of Consumer Research, 11 (December), 741-753.

Johnson, Michael D. (1986), "Modeling Choice Strategies for Noncomparable Alternatives," Marketing Science, 5 (1), 37-54.

Johnson, Michael D. and Claes Fornell (1987), "The Nature and Methodological Implications of the Cognitive H Representation of Products," Journal of Research, 14 (September), 214-228.

Oden, Gregg C. (1987), "Concept, Knowledge, and Thought," Annual Review of Psychology, 38, 203-221.

Olson, Jerry C. and Thomas J. Reynolds (1983), "Understanding Consumers' Cognitive Structures: Implications for Advertising Strategy," in L. Percy and A. B. Woodside (eds.), Advertising and Consumer Psychology, Lexington: Lexington Books, 77-90.

Paivio, Allan (1971), Imagery ant Verbal Processes, New York: Holt, Rinehart & Winston.

Quillian, M. R. (1968), "Semantic Memory," in M. Minsky (ed.), Semantic Information Processing, Cambridge, MA: MIT Press.

Rokeach, Milton (1973), The Nature of Human Values, New York: John Wiley & Sons.

Rosch, Eleanor (1975), "Cognitive Representation of Semantic Categories," Journal of Experimental Psychology: General, 104 (September), 192-233.

Rosch, Eleanor and Carolyn B. Mervis (1975), "Family Resemblances: Studies in the Internal Structure of Categories," Cognitive Psychology, 7 (October), 573-603.

Rosch, Eleanor, Carolyn B. Mervis, Wayne D. Gray, David M. Johnson, and Penny Boyes-Braem (1976), "Basic Objects in Natural Categories," Cognitive Psychology, 8 (July), 382-439.

Sattath, Shmuel and Amos Tversky (1977), "Additive Similarity Trees," Psychometrika, 42 (3), 319-345.

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

Tversky, Amos and J. Wesley Hutchinson (1986), "Nearest Neighbor Analysis of Psychological S paces," Psychological Review, 93 (1), 3-22.

Vinson, D. E., J. E. Scott, and L M. Lamont (1977), 'The Role of Personal Values m Marketing and Consumer Behavior," Journal of Marketing, 41 (April), 44-50.

Wyer, Robert S. and Thomas K. Srull (1986), "Human Cognition in Its Social Context," Psychological Review, 93 (3), 322-359.

Young, Shirley and Barbara Feigin (1975), "Using the Benefit Chain for Improved Strategy Formulation," Journal of Marketing, 39 (July), 72-74.

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