Unresolved Issues in Identification of Determinant Attributes

Mark I. Alpert, The University of Texas at Austin
ABSTRACT - It has been over a decade since the term determinant attributes was introduced to signify those attributes critical in determining product and brand preferences and choices. Substantial research and conceptual activity have occurred in this period, but problems remain. This paper will discuss issues of definitions, conceptualization, measurement, evoked set, and response set influences on identification of determinant attributes.
[ to cite ]:
Mark I. Alpert (1980) ,"Unresolved Issues in Identification of Determinant Attributes", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 83-88.

Advances in Consumer Research Volume 7, 1980     Pages 83-88

UNRESOLVED ISSUES IN IDENTIFICATION OF DETERMINANT ATTRIBUTES

Mark I. Alpert, The University of Texas at Austin

[The author gratefully acknowledges the helpful suggestions of Andy Mitchell, Jim Myers and Jerry Olson.]

ABSTRACT -

It has been over a decade since the term determinant attributes was introduced to signify those attributes critical in determining product and brand preferences and choices. Substantial research and conceptual activity have occurred in this period, but problems remain. This paper will discuss issues of definitions, conceptualization, measurement, evoked set, and response set influences on identification of determinant attributes.

INTRODUCTION

Determinant attributes have been attracting an increasing amount of interest in recent consumer behavior literature. Articles by Berkowitz, Ginter and Talarzyk (1976); Shimp and Lindgren (1978); Heeler, Okechuku and Reid (1979); Anderson, Cox and Fulcher (1976) (plus rejoinders) have added to our knowledge of how individuals utilize attitudes towards product attributes that "make a difference" in arriving at preferences and choice decisions and how these attributes may be distinguished from those that do not have relevance in these contexts. Every ACR conference involves some papers referring to determinant attributes.

But this same literature has also added to our confusion over some of the basic issues about determinant attributes. Articles still appear in the literature in which salience is used to stand for importance, and/or determinance, and a subsequent paper by Olson, Kanwar, and Muderrisoglu (1979) was written to attempt further clarification of the concepts discussed by Myers and Alpert in 1977. In spite of the clarification attempted with these papers, there remain substantial problems ranging from conceptual, semantic, and measurement confusions among determinance and related constructs, as well as conceptual and operational problems unrelated to semantic confusion. Accordingly, the present paper and this ACR workshop are intended to provide a valid perspective concerning the conceptual foundations for determinant attributes, the relationships between determinance, importance, and salience, and the relevance of each to such matters as multiattribute models and information processing. Hopefully, this position paper can summarize the consensus among consumer researchers who are working in this area, as well as indicate where different schools of thought diverge on conceptual and operational issues. The issues are more than semantic, although differences in semantics are troublesome enough. Researchers who adhere to salience vs. determinance may be pursuing what are thought to be competing theoretical and pragmatic approaches that might logically and productively be viewed as complementary rather than competitive.

After reviewing the history of semantic and conceptual differences regarding determinant attributes (and related terms), it may be useful to consider some improvements in the conceptualization of determinant attributes, as well as how they can be elicited and measured for relative strength in influencing attitudes and behavior. In addition, conceptualization and measurement efforts may require modification to deal with complexities introduced by multi-stage decision-making, threshold-levels for satisfaction, information-processing, and individual differences in cognitions and response styles. The general perspective will be divided into two parts, conceptual/definitional and operational/procedural, with the understanding that issues within each category are related to some extent with those in the other.

Conceptual and Definitional Issues: Salience, Importance and Determinance

At the most general level, semantic and conceptual confusion about these three terms is in part due to their common link to the "importance" of a feature, attribute, or specific attitude in forming an individual's overall disposition toward a product or service (Myers and Alpert, 1977). Ail three terms contend to describe features or related attitudes towards them that "make a difference" in overall attitudes toward and consumption of products or brands. The problem arises in part from loose conceptual and operational definitions, and in part from loyalty towards a particular research "school" that prefers to define and measure critical attitudes/ attributes in a particular way. If all researchers and their readers could agree on the underlying construct, and that multiple measures are frequently needed, semantic problems would not be substantive barriers to communication, effective investigation, and applications. At present there are conceptual conflicts, as well as exaggerated differentiation of research traditions, that stem in part from the preference for various labels and their particular traditions involving different (and perhaps differently effective) operationalizations.

Scholars in this area are hopefully by now familiar with the various constructs and definitions associated with salience, importance, and determinance. What follows therefore is a very brief precis of each, with details available in Myers and Alpert (1977), and Olson, et al. (1979).

Salience

While the term has a military derivation that describes a formation as "standing out" from the ranks, the first attitude definition of salience Myers and Alpert (1977) found comes from Kretch and Crutchfield (1948). They stated (p. 163) "Saliency refers to the fact that not all of a man's beliefs stand out with equal prominence in his cognitive field. He may be more acutely aware of certain of his beliefs than others, they may enter his thoughts more readily, they may be more frequently verbalized--they are, in a word, salient." Most researchers have followed the operationalization suggestions of Fishbein (1971) and have measured saliency by noting the order in which attributes of a product were verbalized when respondents were asked to name the things "that come to mind when you think about buying ..... (specified product)."

The recent paper by Olson, Kanwar, and Muderrisoglu prefers salience (over the other two terms) as the "only abstract, theoretical concept." Viewing it from an in-formation-processing perspective, they define salience as "the activation potential of stored knowledge, that is, the likelihood that an item of knowledge, such as a belief about a product attribute, will be activated from memory for use in some cognitive process." (p. 288). They later state (p. 289) that salient attributes are activated from memory and used in attitude formation, and imply conscious processing, which fits the Kretch and Crutchfield definition. However, they also note that self-reports may not be accurate indicants of which attribute beliefs actually influence attitudes, and that multiple measures may be needed. While these statements agree with assertions by those preferring to work with determinant attributes as a construct (Alpert, 1971, Myers and Alpert, 1968 and 1977), the Olson, et al. paper does not resolve the issue of conscious vs. non-conscious processing. Further, their measurement examples and their first definitions of salience imply a preference for conscious and readily verbalized attitudes being considered as the salient ones. To obtain the benefits of seeing the construct as "those beliefs that causally influence" attitudes, it may be necessary to use a more general term than salience.

Importance

The importance of attributes in attitude formation and choice behavior has not been rigorously defined. Dictionary definitions equate importance with having weight or consequence, or significance. Important attributes would presumably figure heavily in a consumer's decisions regarding preferences and choices among alternatives. As Olson, et al. (1979) wrote, without specification of "importance for what," it is hard to determine what would be meant by relative importance of a feature. However, their specification of "importance for determination of attitude" is not of much help in reducing ambiguity if attitude is dependent on the consumption/ usage situation. For example, one's attitude towards paper plates may be enhanced by beliefs concerning their disposability, in the context of choosing dishes for a picnic. In the context of entertaining guests at home, the polarity of this belief's contribution towards overall attitude may well reverse.

At the conceptual level, the meaning of salience for a specific feature may also vary across situations. For example, "economy" may be mentioned as a salient trait of paper plates, in a picnic context, while "cheap" might be the not-so-equivalent feature that becomes salient in a "dining" context.

Measures of importance have typically used stated importance "in choosing a...(product)," typically without specification of a specific consumption usage situation, although there are exceptions (Alpert and Davies, 1974). In addition to problems of ambiguity concerning the choice context, self-report data may present problems of respondent lack of knowledge and/or willingness to reveal true reasons. Moreover, there is the question of importance vs. determinance.

Determinance

Myers and Alpert defined the term "determinance" to reflect the fact that certain product features or attributes are more influential in determining preferences and choices than are others. They state (1968, p. 13), "Attitudes toward features which are most closely related to preference or to actual purchase decisions are said to be determinant; the remaining features or attitudes--no matter how favorable--are not determinant." Determinance is said to go beyond importance as the latter is generally measured, although it is not different conceptually. A feature such as automobile safety might be seen as highly important by consumers yet have no real effect on choice among competing brands if all are perceived to be equal for this feature. Since many studies have shown that people list as important (and frequently give as salient) features that do not differentiate among alternative brands, determinance was conceptualized to stress both important and perceived differentiating characteristics of products and services. While not restricting operationalization to correlational methods, determinance was said to imply covariation between perceptions of a product's feature(s) and overall attitude/choice vis-a-vis the product. If, for example, persons favor a brand rated higher on certain attributes than those brands they dislike, there is reason to suspect that, ceteris palibus, some role is played by these features (halo aside). For those features they see as equivalent across brands, there is less determinance possible for those features, as long as no differentiation breakthroughs take place.

At the conceptual level, determinance was equated with importance, but was also defined to incorporate the significance of differentiation and avoid the loose usage associating importance with stated importance. Originally, Myers and Alpert intended to use the term "salience'' but were dissuaded by their discovery of Kretch and Crutchfield's definition that stressed top-of-mind awareness and ready cognition. Clearly the relevance of some attributes that determine choices may not be known even to the chooser. ("Motivation research" is persuasive on at least this point.)

Olson, et al. (1979) state that determinance was developed to help predict product choices and was therefore not developed to suit theoretical objectives. However, most studies in determinance have either used overall preference or simulated product choices, as actual choices are frequently confounded by marketing mix variations that occur between attitude measurement and purchase. While predicting choices from attribute evaluations remains a concern of determinant attribute research, its application to attitude formation is intended in both its conceptual development and research literature. Further, there is nothing inherently atheoretical about empirical validation of the relationship between attribute perceptions and choice behavior. While prediction without explanation is not theory, a concern that the theory also have some relevance to predicting choices is hardly a disqualification. Some of the conflict between schools of thought might pertain to the ancient but false "theory vs. practice" dichotomy.

In summary, the conceptual and definitional confusions surrounding determinance, importance, and salience arise in part from lack of specification of what role each term places at the conceptual level of attitude formation and choice behavior. Conceptual refinements should also be sought in linking these three terms and their related concepts. Care should be taken to preserve the meanings that originated with salience, importance, and determinance. Which term should be used to signify those attitudes/attributes that are most closely linked to overall attitude and or behavior? In one sense it does not matter, as long as the conceptual base is built to recognize the broad framework in which attitudes are formed and combined to reach preferences and choices. In this framework, under varying contexts, certain attitudes and/ or attributes will be more closely related to key dependent variables than others. Given the "conscious" connotations of "salience" and weak conceptualization of "importance," "determinance" is probably a more inclusive construct. A determinant attribute of preference or choice may well be salient. However, nonsalient attributes, which the individual cannot or will not consciously recognize, may well be determinants of overall preferences. Given the recent history of attitude research in this area, it may well be that "salience" may continue to be misinterpreted as a concept that includes all attitudes that may "cause" overall attitude and/or preference. No damage would be done by this, provided that salient is redefined and given the breadth of determinance, and provided that it is not operationalized in the limited manner (consistent with its "top-at mind" connotations) as it has been in the past.

UNRESOLVED OPERATIONAL AND PROCEDURAL ISSUES

Semantic/Definitional Issues

Some of the unresolved measurement issues for identification of the determinance of attributes relate to the largely semantic confusions discussed above. For example, given the definition of salient attitudes as known "at least" to the individual, accessible by memory, traditional measures have operationalized salience as order of elicitation of attributes considered by respondents in connection with attitudes and purchases of various stimuli. Similarly, given the traditional interpretations of importance, it has been traditionally operationalized as stated importance of various features (with either aided or unaided recall, each of which presents problems). These measures of salience and importance both suffer from ambiguities when the decision or attitude context is not clearly specified. Further, they both present the well-known problems of respondent ignorance and/or unwillingness to reveal socially unacceptable attitudes or features that may be determinants of choices. Order of elicitation also presents the added problem of low reliability in rank orders and similarity of concepts elicited (Olson and Muderrisoglu, 1979). While perhaps generally higher, reliabilities of stated importance are rarely reported in the consumer psychology literature, and reliability improvement is no doubt a psychometric task of continuing relevance.

A related definitional-measurement problem is the tendency to identify a construct with a particular operational definition. THE MEASURE IS NOT THE CONSTRUCT. Salience in attributes may be less reactively elicited if subjects were to participate in free-access computer-assisted tasks in which individual utilization of product attribute information might provide an information-processing measure of the salience of attributes. The amount of attribute information requested would be one measure of salience, along with the order in which attributes were assessed in the particular decision-context. The usual caveats are needed in designing these studies (Bettman, 1979) to provide realistic and involving choices and to avoid biasing the information acquisition as much as possible. Eye movement cameras and related measures of information acquisition may provide additional measures of salience that are potentially less reactive than computer-assisted procedures, information boards, and self-reports, including verbal protocols. Refinements in each of these methods may enhance our ability to measure salience of attributes, presuming conscious information processing. Creative work in protocol analysis and the design of multi-attribute stimuli for visual scanning might even enable inferences about attributes that may be determinant (or at least involve a lot of information processing activity) yet are not consciously known to the subject. These "revealed determinants'' may not be salient nor likely to be consciously evaluated using "information", if the subject does not know of their determinance.

Determinance has frequently been operationalized by direct dual questioning (Myers and Alpert, 1968), in which stated importance of features has been multiplied times stated perceived differences among alternative choice objects. While reliabilities such as .78 and .86 have been cited for this measure (Alpert, 1968 pp. 129-130) and it has been shown to have higher predictive validity than the single question of stated importance alone (Myers and Alpert, 1977), it too is only one measure of determinance and not the construct. Where largely objective attributes are involved, direct dual questioning may be a reasonable shortcut to more demanding methods, some of which will be indicated below. It may be necessary both to generate attributes and to explore attitudes and choices with a battery of techniques involving observation, experimentation, direct and indirect questioning of respondents, covariate methods, conjoint measurement, and others (Myers and Alpert, 1968 and 1977). Some of these methods are correlational and thus provide measures of determinance such as beta-weights in regression. Even without multicollinearity problems (see Green, Carroll and De Sarbo, 1978, for methods of dealing with multicollinearity) such measures are of course not indicative of causal links between attributes and overall preference/behavior. However, the features so obtained as "determinant" are better candidates for determinance than those not significantly involved in the equations. Corroboration by conjoint measurement, laboratory and/or field experimentation might be additionally desirable for attributes said to be high in determinance by regression, dual questioning, or other methods. Regression users should be mindful of the need to minimize the distortions in measured determinance due to halo effects (Beckwith and Lehmann, 1975). The more complex and costly methods should logically refine determinance measures for those attributes previously "identified" by less costly ones. Further, convergent validation of measures of attribute determinance, to assist in theory construction as well as establish guides for cost-benefit analysis of identification methods is still proceeding (Alpert, 1971; Heeler, Okechuku, and Reid, 1979; McCullough and Best, 1979; Green and Srinivasan, 1978).

Non-Definitional Issues

While the above issues may be viewed as stemming in part from single vs. multiple measures of determinance (as defined and conceptualized in alternative terms), there remain other issues that may bear on the nature of determinant attributes in a variety of contexts, that do not merely reflect essentially semantic differences. Some of these were noted in Myers and Alpert (1977) and still require work regarding the determinance of attributes above and below certain threshold levels, decision-processing rules for evaluating levels of attributes, and dynamics in determinance over time. It is worth noting, however, that for these concerns, as well as for the definitional and measurement issues so far discussed in the present paper, we have assumed a single stage decision process, with a constant level of determinance which each individual would give for each attribute (perhaps excepting the notion of attribute threshold levels). This "individual choice of individual product" paradigm is relevant to a large number of purchase and evaluation situations. However, a substantial amount of attitude and choice formulation may well take place involving: groups of individuals, sequences of choices in a multi-stage hierarchy of evaluations, and sets of products to be chosen together. Other complications may also be induced by the difficulties in adjusting for response traits of subjects along one or more measures of determinance. The following section will discuss important issues to be resolved in the theory and measurement of determinance in these contexts.

Response Set "Bias" and Ipsatizing

The first complexity to be added to the methods discussed previously concerns the tendency of various individuals to respond in a consistently biased pattern towards a set of scales. Measures such as stated importance, dual questioning, and stated ideal points are particularly likely to result in different individuals who will rate all attributes as high and others who may rate all as low. Since benefit segments might be distorted by these response traits (and also discriminant analysis of types of customers, using profiles of these scales) a traditional strategy is to standardize within the individual, so that all persons have the same mean and variance for these scales. Clustering could then proceed on the relative profiles, seemingly leading to greater homogeneity within the segments thus formed. However, there is theory and recent empirical support to suggest that means for stated importance of, in one instance, automobile features were substantially related to respondents' income (positively) and alienation from the marketplace (negatively) measures (Bruce, Alpert, and Allison 1979). Behavioral and attitudinal correlates of these confounding variables are too important to be washed out by ipsatizing, and persons with different mean scores, as well as profiles about these means, should perhaps be placed into different segments. Whether or not to ipsatize is an unresolved issue, and both procedures have advantages and disadvantages that might require their joint use to determine what implications are sensitive to the standardizing procedures and which are not. A seemingly minor and "standard" procedure such as standardizing scales within the individual might require additional research to determine conditions under which the resulting clusters might obscure within-cluster heterogeneities in important marketing variables. How to account properly for these correlates will be especially important when doing cross-cultural research or comparing subcultures within the same country.

Techniques such as conjoint measurement and regression analysis may appear to abstract from this problem, in that they standardize within the individual through co-variation between attribute ratings and overall ratings of brands. However, regression analysis presents the analogous problem of "halo" effects in which beliefs may be uniformly high or low, depending on the overall attitude towards the brand. Subtracting the individual's mean score for the brand, across all beliefs, is analogous to ipsatizing importance weights, but similarly presents problems when the product is genuinely seen as high (or lower) in a substantial number of features. A wealthy person may be more influenced by the handling, ride, safety, and durability of a Mercedes sedan (vs. a Ford Fairmont) and less concerned with its price than would the Fairmont buyer. To give both his "images" of the cars the same average belief rating could severely distort the measures of attribute determinance. Care in instrument design is to be preferred over after-the-fact data adjustments (e.g., rate several concepts along each attribute at a time or pre-specify hypothetical attribute levels and covary with preference or simulated choice).

Unit of Analysis: Individual vs. Group

Rather than belabor the familiar issue of whether to analyze each individual's responses across a set of concepts (data permitting) vs. aggregation into managerial segments (such as media exposure groups), we should comment upon the measurement of determinant attributes considered in group vs. individual decision-making. As has been noted by Wind (1978) and others, decisions made by groups of individuals, whether in a family or organizational context, may involve different decision processes and focus on different features of the choice alternatives than those involved when an individual chooses "more or less" alone. Compared to the battery of measurement methods for individual choice behavior, the array of techniques for studying group processes and inferring relative determinants of features of the products, and especially how these interact with the dynamics of group leadership and power, are considerably less well known in consumer research. Insights from the literature of small groups and management theory, as well as conflict and negotiation study in labor relations and political science, may prove useful in modifying and extending the techniques used for individual-dominated evaluation situations. One thing stands clear: as hard as it is to study attitude formation and choice for individuals in a nonreactive manner, the problems involved in questioning groups concerning their "methods" are considerably more complex. Consider the law inter-subject reliability data noted by Davis and Rigaux for self-reports of household purchase influence (1974). On the other hand, what would be the probable validity and reliability of data obtained by asking a group of decision-makers to represent their criteria via a conjoint measurement technique, with rankings determined by individuals and then compared with the somehow-determined group rankings? Would relative power be manifested by congruence with group rankings (or is this merely conformity), and how do any of these measures predict what would happen in an external validity sense? Extensions of determinant attributes identification methods to group decisions will not be easy, but the importance of understanding these evaluation situations may justify continued research. Discussion of some possible methods may be found in Wind (1976, 1978).

Unit of Analysis: Single vs. Multiple Evaluations

Just as the addition of other purchasers or evaluators complicates the identification of determinant attributes, so it is when an individual (or more complex, group of individuals) must evaluate in order to choose an assortment of products from some evoked set, rather than simply choose one. Leigh McAlister (1978) has argued persuasively that individuals choosing what might be termed product portfolios will evaluate products and their features in the context of others that are tentatively included with them as a set to be chosen. Her data support the notion that the utility of a combination of products is not the simple sum of their individual utilities, and may vary according to the degree of attributes overlap in the set, as well as the subjective probability that a chosen alternative will be received. Given a reasonable method for generating a parsimonious set of product combinations, one might secure ratings of the utility given by a set of products along several attributes, and then correlate this with the overall utility of the "portfolio". Whether this should be done via regression, conjoint, or some other method requires further research. The data demands are no doubt greater than single product choice situations. However, it is worth learning about the relevance of attribute satiation. It is possible that individuals can be studied by assigning some fraction of their budgets to a product category, identifying the joint utility levels of all possible combinations, and then using linear programming to approximate their chosen portfolios. This work, when combined with research on product-market boundaries (Day, Shocker, and Srivastava 1977) can be used to determine when choices may be studied as independent of other products owned, and when substitutes and complements are chosen in some sequence that requires more elaborate procedures of measurement and conceptualization.

Multi-Stage Decisions

The concepts and measurement techniques so far discussed have been implicitly used to describe choices made from simultaneous consideration of the choice options, whether by individuals or groups. Yet it is worth noting that the concept of determinance is dynamic rather than static. As noted by Myers and Alpert (1968) and more recently by scholars such as Shocker and Srinivasan (1979), features that are non-determinant at one stage in a multi-stage decision process may become determinant at a later stage, and vice versa. Myers and Alpert noted that safety may not be a determinant attribute for persons choosing among alternative airlines (important but not usually differentiating among airlines). However, for the choice of travel mode, persons choosing not to fly may view safety as quite determinant, leading Amtrak to different promotional strategies than United Airlines, for some market segments. Moreover, attracting passengers may be a multi-stage process for airlines, who may need to improve the perceived safety of air travel when appealing to non-fliers, and quality of in-flight service (for example) when appealing to fliers.

Given the large variations in attribute levels for some possible choice options, it is surprising that anything useful comes from asking respondents to state the relative importance of features such as economy, when choosing an automobile (for some type of use, although this is frequently unspecified). The same is applicable to inquiries concerning perceived differences among alternative cars, for example, in terms of each feature. Shimp and Lindgren have demonstrated variations in the perceived differences among various evoked choice sets (1978). Thus, for the subcompact buyer who chooses a VW Rabbit over a Ford Fiesta, low purchase price may be less determinant than handling and interior space, given the final choice set. However, it should be clear that he may value low purchase price more and interior room less than someone who chooses a Mercedes sedan over a Chevrolet.

To improve the conceptual understanding of determinant attributes in multi-stage decisions, research might proceed by taking subjects through stages, involving a list of choice options (specification of which is also an unresolved issue), and seeking to identify determinant attributes that move the respondent from one stage of product elimination (or some other decision rule) to the eventual final evaluation of the alternatives that have survived this screening. Linear compensatory models may therefore owe some of their predictive validity to the either conscious or unconscious matching of respondents to brands that are all within a relevant range, along the determinant attributes. The single-stage models may "work" provided the samples are studied with some pre-screening for "compact car buyers," for example, or by specifying the choice options as "fast food restaurants," and obtaining subjects who normally choose at least some of the time from that category. The inclusion of a product or brand that is obviously noncomparable along one or more determinant attributes might evoke a different choice situation and alter the determinance of the product features so measured.

Managerial vs. Theoretical Perspective

As noted in the above and previous sections, there are numerous instances in the determinant attributes literature in which theoretical and managerial considerations may not necessarily be the same. Managers seem to see theoreticians as concerned with "what would be nice to know, but what is not necessary to know." Hence, they might view the multi-stage discussion as irrelevant, provided the sample is properly chosen and given appropriate situational perspective.

The theoretician-purist, represented in this area by the Olson, et al paper (1979), tends to hold that managers and those who feel managerial issues are relevant, are not concerned at all with theory. No doubt, there is insight in this observation, but there is also the same amount of overstatement as pertains to the managerial "perspective" that theory is inherently impractical. As long as organizations such as the ACR continue to provide a not-so-comfortable meeting place between industry and academia (although neither is monolithic), it is likely that some degree of purity vs. relevance debates will transpire. This paper is not the place to attempt resolution of this conflict, but it is appropriate to state this too as an unresolved issue in the determinant attributes area.

Most of this paper has been concerned with prediction AND understanding of the process of attitude formation and choice, and the role of certain attributes or product features that may be termed determinant in these processes. These are not incompatible perspectives, but the conflicts between them are non-trivial. For example, the product portfolio extensions of individual product choices may be important to the theory of consumer behavior. However, the cost-benefit ratios of such inquiry may be hard to justify in an industrial marketing research context if either a) the choices are truly independent of other products owned, or b) they are not, but there is no effective way of targeting messages to persons who already have certain known product assortments. Perhaps if theoretical work could provide a kind of Guttman-scale of product ownership, in which persons who bought one appliance could be positioned in terms of past and probable future selections, managers would be more interested in applying the results of the research to design promotional campaigns aimed at specific purchasers.

The issue is far more complex than this, and cannot be fully explicated, given this paper's scope. However, it is worth noting that well-trained academics are not normally concerned with managerial implications beyond the scope of the sub-specialty, and numerous examples of this could be cited in the literature of determinant attributes and other areas as well. Similarly, while managers also sub-optimize within their areas of expertise, they may have a broader view of what researchers might be able to do for them, yet lack the specialized skills to develop tight theory and measurements that may be needed. The dialogue should continue, and might be aided by consumer researchers attempting to recognize the relevance of questions such as "so how can I use this knowledge?"

The importance of theoretical inquiry, even in the absence of direct application, is not an issue here. Not all researchers should be expected to focus on practical applications of their constructs. It has been suggested that this makes for poor research in some cases, and dilution of skills at any rate. However, at the ACR workshop in which this paper was discussed, it became obvious that managerial and theoretical perspective can lead to focus on issues that are viewed as competitive. Productive interchange may be enhanced by the recognition of differing viewpoints and the complementary nature of many of them.

Similarly, many of the approaches to identification of determinant attributes that have been proposed have been set in opposition to each other. More research is needed to determine the relative utility of each of these methods, under varying situations, and choice sets, numbers of evaluators, and sequences of decision stages. For quite some time, it is expected that multiple conceptual frameworks and identification methods may be more useful than "one best" approach, provided the biases and limitations of each are considered in selecting approaches for study and evaluating their findings. Semantic and philosophical differences need not obscure the issues, provided they are understood and properly dealt with.

REFERENCES

Mark I. Alpert, "Identification of Determinant Attributes: A Comparison of Methods," unpublished D.B.A. dissertation, University of Southern California, 1968.

Mark I. Alpert, "Identification of Determinant Attributes: A Comparison of Methods," Journal of Marketing Research, 8 (May 1971), 184-191.

Mark I. Alpert and C. Shane Davies, "A Decision Sciences Approach to the Marketing of Public Transportation," in Manfred W. Hopfe, ed., American Institute for Decision Proceedings, 6 (Atlanta, Georgia: American Institute for Decision Sciences, 1974), 162-165.

W. Thomas Anderson, Jr., Eli P. Cox, III, and David G. Fulcher, "Bank Selection Decisions and Market Segmentation,'' Journal of Marketing, 40 (January 1976), 40-45.

Nell E. Beckwith and Donald R. Lehmann, "The Importance of Halo Effects in Multi-Attribute Attitude Models," Journal of Marketing Research, 12 (August 1975), 265-275.

Eric N. Berkowitz, James L. Ginter, and W. Wayne Talarzyk, "An Investigation of Alternative Indicators of Attribute Salience and Their Effects on the Size of the Attribute Set in the Multi-Attribute Model," in Kenneth L. Bernhardt, ed., Marketing: 1776-1976 and Beyond, (Chicago: American Marketing Association, 1976), 611-615.

James R. Bettman, An Information Processing Theory of Consumer Choice, (Reading, Massachusetts: Addison-Wesley Publishing Company, 1979).

Thomas V. Bonoma, Gerald Zaltman, and Wesley J. Johnston, Industrial Buying Behavior, (Cambridge: Marketing Science Institute, 1977).

Harry Davis and Benny P. Rigaux, "Perception of Marital Roles in Decision Processes," Journal of Consumer Research, 1 (June 1974), 51-62.

Paul E. Green, J. Douglas Carroll, and Wayne S. DeSarbo, "A New Measure of Predictor Variable Importance in Multiple Regression," Journal of Marketing Research, 15 (August 1978), 356-360.

Paul E. Green and V. Srinivasan, "Conjoint Analysis in Consumer Research: Issues and Outlook," Journal of Consumer Research, 5 (September 1978), 103-123.

Roger M. Heeler, Chike Okechuku, and Stan Reid, "Attribute Importance: Contrasting Measurements," Journal of Marketing Research, 16 (February 1979), 60-63.

David Krech and Richard S. Crutchfield, Theory and Problems of Social Psychology (New York: McGraw-Hill Book Co., 1948).

Leigh McAlister, "Choosing Multiple Items from a Product Class," Stanford University Research Paper No. 447 (Palo Alto, California: Stanford Graduate School of Business, 1978).

James McCullough and Roger Best, "Conjoint Measurement: Temporal Stability and Structural Reliability," Journal of Marketing Research, 16 (February 1979), 26-31.

James H. Myers and Mark I. Alpert, "Determinant Buying Attitudes: Meaning and Measurement," Journal of Marketing, 32 (October 1968), 13-20.

James H. Myers and Mark Alpert, "Semantic Confusion in Attitude Research: Salience vs. Importance vs. Determinance," in William D. Perreault, Jr., ed., Advances in Consumer Research, Vol. 4, (Atlanta: Association for Consumer Research, 1977), 106-110.

Jerry C. Olson and Aydin Muderrisoglu, "The Stability of Responses Obtained By Free Elicitation: Implications for Measuring Attribute Salience and Memory Structure," in William L. Wilkie, ed., Advances in Consumer Research, Vol. 6 (Ann Arbor: Association for Consumer Research, 1979), 269-275.

Jerry C. Olson, Rajesh Kanwar, and Aydin Muderrisoglu, "Clarifying the Confusion Regarding Salience, Importance, and Determinance Concepts in Multiattribute Attitude Research," in Neil Beckwith, et al., eds. 1979 Educators' Conference Proceedings, (Chicago: American Marketing Association, 1979), 286-290.

Terrence A. Shimp and J. H. Lindgren, "Some Methodological Considerations and Empirical Findings Concerning the Use of the Dual Questioning Method to Identify Determinant Attributes," in H. Keith Hunt, ed., Advances in Consumer Research, Vol. 5 (Ann Arbor:. Association for Consumer Research, 1978, 327-332.

Allan D. Shocker and V. Srinivasan, "Multiattribute Approaches for Produce Concept Evaluation and Generation: A Critical Review," Journal of Marketing Research, 16 (May 1979), 159-180.

Rajendra K. Srivastava, Allan D. Shocker and George S. Day, "An Exploratory Study of Situational Effects on Product Market Definition," in H. Keith Hunt, ed., Advances in Consumer Research, 5, (Ann Arbor, Association for Consumer Research, 1978), 32-28.

Yoram Wind, "Preference of Relevant Others and Individual Choice Models," Journal of Consumer Research, 3 (June 1976), 50-57.

Yoram Wind, "Organizational Buying Behavior," in Gerald Zaltman and Thomas Bonoma, eds., Review of Marketing, 1978, (Chicago: American Marketing Association, 1978), 160-193.

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