A Grounded Model of Consideration Set Size and Composition

ABSTRACT - This paper presents a model of when brands will enter a consumer's consideration set. While much descriptive work on consideration sets exists, little has been done by way of providing a testable framework to determine how the criteria affecting consideration interact to decide a brand's inclusion in the set. A review of commonly used brand choice models reveals substantial research surrounding evaluation given a consideration set, but little work on the consideration set itself. Assuming a utility-maximizing consumer we calculate the brands of which he or she is aware that should enter the consideration set in terms of their utility, mental processing costs and acceptability.

Citation:

John Roberts (1989) ,"A Grounded Model of Consideration Set Size and Composition", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 749-757.

Advances in Consumer Research Volume 16, 1989      Pages 749-757

A GROUNDED MODEL OF CONSIDERATION SET SIZE AND COMPOSITION

John Roberts, University of New South Wales

[We wish to acknowledge the assistance and ideas that Sue Jenkins, Debbie Gifford, and Rebecca Wolfe of Kellogg Australia have given us during this project. Peter McBurney, Jim Lattin, David Midgley and Pam Morrison have provided considerable input to the paper. Pat Hillary provided invaluable support in the preparation of materials and the typing of the paper.]

ABSTRACT -

This paper presents a model of when brands will enter a consumer's consideration set. While much descriptive work on consideration sets exists, little has been done by way of providing a testable framework to determine how the criteria affecting consideration interact to decide a brand's inclusion in the set. A review of commonly used brand choice models reveals substantial research surrounding evaluation given a consideration set, but little work on the consideration set itself. Assuming a utility-maximizing consumer we calculate the brands of which he or she is aware that should enter the consideration set in terms of their utility, mental processing costs and acceptability.

The model contains two aspects to consideration which may be important to different degrees: both non-compensatory and compensatory factors may influence whether a consumer considers a brand or not. This framework explains some of the conflicting empirical evidence in the field. It is also consistent with work on phased decision rules and provides a nice fit with the concept of inept and inert sets.

The paper concludes with a discussion of management implications of the model in terms of an audit of existing brands, new product management, and product line strategies. A summary of limitations of the research is given, as well as suggestions for future directions.

INTRODUCTION

Most marketing texts describe the consumer decision process as a successive refinement in the number of brands evaluated in the decision process (e.g., Kotler 1988, Figure 6-5). Typical typologies have the total set being divided into those brands of which the consumer is aware and those of which he or she is not aware. Awareness is divided into consideration (and non-consideration) and from that set a purchase decision is made. Each of these stages has attracted research attention. A considerable body of literature in the consumer behavior field examines correlates of the consideration set. However, the work is largely descriptive, rarely attempting to establish how the elements affecting consideration fit together. This paper aims to take phased decision rules from consumer behavior and the marketing science tools of decision analysis to develop a grounded theory of when new brands will enter a consumer's consideration set. This allows forecasts to be made and provides a testable model of the process.

Consideration is an important element to consumer behavior. In many product categories leading brands may derive their predominant market share by entering the consideration sets of a higher proportion of consumers than other brands, while not necessarily having a higher preference amongst considerers. Conversely, low share brands must have t consideration before they can increase their market share. In this situation marketing mix elements aimed at increasing preference at the purchase evaluation stage may be wasted. To quote Ries and Trout (1986, W p. 138) "Out of mind, out of business".

LITERATURE

The concept, an evoked set, or consideration set, as the brands which a consumer would evaluate first appeared in the marketing literature in Howard's text on consumer behavior (Howard 1963). It was empirically demonstrated to be a subset of the set of brands of which a consumer was aware by Campbell (1969) and became an integral part of the Howard-Sheth model. Howard and Sheth (1969) define the evoked set as those brands on which the consumer gathers information.

Since that time the existence of consideration sets as strict subsets of the awareness set has been extensively demonstrated (see Table 1 for a summary of studies. The concept fits nicely with theory in consumer behavior which suggests that for complex decisions or those involving many alternatives a consumer is likely to employ a decision process which can be represented by a phased decision rule (see Bettman, 1979, p. 215 for a review). However, while many consumer behavior studies have demonstrated the existence of the consideration set, marketing modelers have been slow in developing frameworks which specify the way in which determinants of consideration operate. The probability of brand choice (given category purchase) can be thought to have three elements; the probability of being aware of brand j; the probability of considering brand j, given awareness of it; and the probability of choosing brand j, given awareness and consideration. It is the second of these elements that this paper addresses. There is a strong body of research relating awareness to advertising (see Mahajan, Muller, and Sharma 1984 for a review). There are also many individual-level brand choice models which express purchase probability as a function of perceptions and preference, given a consideration set that is some fraction of the entire market (e.g., Assessor, Silk and Urban 1978; Defender, Hauser and Shugan 1983). However, there is a distinct lack of systematic studies to model the process by which aware brands enter the consideration set.

Before looking at work on the modeling of consideration sets it is useful to examine the foundations on which the concept is built. The existence of consideration sets (as distinct from awareness sets) is a logical outcome of theories in economics and psychology and has strong empirical support. Research in the economics of information was initiated by Stigler (1961) and is based on the premise that a consumer will continue to search for information as long as the expected returns from that search (in terms of making a choice of expected higher utility) exceed the marginal cost of further searching. In marketing, Shugan (1980) developed a model of the cost of search which showed that the cost of search was proportional to the number of brands that the consumer evaluates and the difficulty of making comparisons. Ratchford (1980) estimated the expected benefit of searching second, third, and fourth brands for various household appliances.

TABLE 1

STUDIES OF AVERAGE EVOKED SET AND AWARENESS SET SIZES BY CATEGORY

In addition to the economics literature which suggests that the consumer may have little incentive to search numerous brands, the psychology and consumer behavior literature raises questions as to his or her ability to be able to do so. Miller (1956) suggests that our cognitive capacity to evaluate alternatives lies around a maximum of about seven due to limits in ability to differentiate and finite memory span. Researchers in consumer behavior have postulated phased decision rules to represent consumers' methods of coping with this complexity. With these rules the consumer firstly filters available alternatives and then undertakes detailed analysis of this reduced set (e.g., see Wright and Barbour 1977, Lussier and Olshavsky 1979). Bettman (1979, p.215) summarizes the implications of this theoretical and empirical work with his proposition 7.8.iv:

"Phased decision strategies, with an elimination phase and a choice phase, may be found when the number of alternatives is large".

What is large in practice may be as low as four (e.g., Gensch 1987).

Not only is there a sound theoretical basis on which to posit the existence of consideration sets, there is also strong empirical support. Consideration sets have been studied for consumer durables (e.g., Gronhaug 1983; Gronhaug and Troye 1983; Hauser, Roberts and Urban 1983) and packaged goods (e.g., Campbell 1969; Reilly and Parkinson 1985; Narayana and Markin 1975), as well as industrial products (Le Blanc 1981, Kosnick 1986). The majority of this empirical work has been largely descriptive, reporting evoked or consideration set sizes and searching for correlates of the size of an individual's evoked set in terms of attitudes to the category, innovativeness, information search, and sociodemographic information.

A summary of findings on evoked set sizes are presented in Table 1. Average sizes range from 1.3 for mouthwash (Narayana and Markin 1975) to 5.4 for soft drinks (Brown and Wildt 1987). Awareness set sizes are shown where available and these tend to be twice to three times the size of the evoked set. A number of studies on the same category come up with substantially different results (e.g, autos). Reasons for this would appear to come from differences in definition of consideration and the measures used.

A summary of the work on correlates of evoked set size is presented in Table 2. There also appears little consistency between findings regarding determinants of the set, with the exception of the relation between awareness set size and consideration set size. Again, this may well be due to differences in the constructs and their operationalization, as well as different populations.

Varying definitions of consideration or evoking raise the need to discuss nomenclature. "Evoked set" was the term first used by Howard (1963) and Campbell (1969). However, the term "evoked set" has been used in a number of ways. From Howard and Sheth's "brands that the consumer would consider" (1969), Belonax (1979) used it to be brands "acceptable to the consumer". As Myer (1979) points out, these two sets are not likely to be the same. Other authors have included aware but rejected brands in the definition of evoked set, a group which Allaire (1973) calls the relevant set. The relevant set may provide useful measures of perceptions and preference for market research but it has little meaning from a consumer behavior perspective Wright and Barbour (1977) use the term consideration sets and it seems untainted by the ambiguities plaguing evoked sets. Thus it is used in this paper. The definition is the same as that of Howard and Sheth for evoked sets: "the brands that a consumer will consider". This set may change during the information search process and so the consideration set is dynamic. When search is finished and the evaluation phase is undertaken, the final consideration set is termed by some authors the choice set (Kotler 1988, Hauser and Shugan 1983).

Having introduced the concept of consideration sets with a theoretical rationale as well as empirical evidence, we now look at developments in the modeling of consideration sets. The weight of the consumer behavior literature tends to favor noncompensatory models at the consideration phase and then compensatory models at the evaluation stage (e.g., Wright 1975). These have also been observed to give good fits (e.g., Lussier and Olshavsky 1979, Gensch 1987, Lehtinen 1974). In detailed studies of consideration set composition, Brisoux and Laroche (1981) found that a conjunctive model gave better fits than a compensatory model, while Parkinson and Reilly (1979) found the converse. In both cases the difference in fits was very low. In this paper we argue that both formulations are appropriate. Narayana and Markin's (1975) classification of aware but nonconsidered alternatives into inept and inert holds the key to this argument. Inept brands correspond to those that are ruled unacceptable for some reason and a conjunctive (or other non-compensatory) rule would appear to be the best way to model such a decision. Inert brands correspond to those which are acceptable but for which the consumer does not feel a need, that is those that do not offer sufficient utility. A compensatory multiattribute utility model would appear to offer a good representation of deciding between whether a brand is worth considering or not (inert).

TABLE 2

STUDIES OF CORRELATES OF EVOKED SET SIZES

MODEL DEVELOPMENT

The previous section highlighted the theoretical grounding for consideration sets and empirical evidence of them. Analysis of consideration sets will be important where factors influencing consideration are different to those influencing evaluation (e.g., Gensch 1987) or where different decision processes are adopted at the two stages (e.g., Wright and Barbour 1977).

The literature suggests that a grounded model of consideration should include search costs as well as the expected (but uncertain) benefits to search, the flexibility to explain non-consideration in terms of non-compensatory factors as well as compensatory ones, and a framework which allows integration with choice models. Thus, we posit the existence of two types of decision rule at the consideration stage: conjunctive and linear compensatory. This formulation subsumes most of the popular models of consideration, given that disjunctive rules have not been shown to provide good fits (Parkinson and Reilly 1979; Brisoux and Laroche 1981). The major accusation that can be made against it is that it is not parsimonious.

The model for consideration is illustrated as a decision tree in Figure 1. We examine the conjunctive and compensatory aspects of the consideration process separately. For the conjunctive model we calculate whether the brand will enter the consideration set in terms of the underlying conjunctive threshold tolerances for acceptability. For the compensatory aspect we look at the evaluation process which will be undertaken with a given consideration set by a utility-maximizing consumer. Only if a brand passes both of these tests will it be considered. Which of these phases gives the most leverage will depend on situational variables, including the category. For the packaged goods categories examined by Narayana and Markin (1975) the ratio of average number of inept brands to inert brands ranged form 2.4:4.7 for beer to 2.5:2.0 for toothpaste. Within a category this ratio was higher for some brands than for others. Presumably the more specifically a brand is targeted to a segment the higher level of inepts it will get among the population at large and the lower number of inerts. For example, Narayana and Markin showed the smoker-targeted toothpaste brand of Pearl Drops as having 65% inepts and 20% of the population as inerts, while the late entrant generally-targeted Aim had only 35% inept, but 46% inert.

The first step in modeling is to decompose consideration into its two elements; a conjunctive one and a linear compensatory one. Let Aj be the condition that represents the conjunctive part of the model, the brand's acceptability. A_j = 1 if the brand is acceptable, 0 otherwise. Similarly let V_j be the condition that the brand is of sufficient utility to pass the compensatory criterion. Again V_j = 1 if it does and 0 otherwise. Thus a brand will only be considered if both A_j and V_j equal 1.

Conjunctive Model

Aj may be expressed in terms of the underlying constraints (threshold tolerances) that the consumer has. These thresholds may not be the same as the attributes which the consumer uses at the evaluation stage. For example, for breakfast cereals at the conjunctive phase actual sugar levels may have a maximum threshold level, while at the evaluation stage physical sugar levels may not be important, but sweetness might be and although it is positively related to sugar, it may be a desired attribute.

We will denote the thresholds that a brand must meet to pass the conjunctive screen by T_n (n=1,2...N) and measure them so that they represent minimum levels on each constraint (e.g., by multiplying maximum constraints by minus one). The level of product j on constraint n is y_jn and d_jn represents whether j passes the threshold or not.

Thus d_jn = 1 if y_jn $ T_n

= 0 if otherwise

and so EQUATION (1)

Note that Einhorn (1970) suggests that the actual levels of a brand on these constraints may never be discovered and the final section of this paper discusses the case of measurement error or random elements to consumer choice at the conjunctive stage.

Compensatory Model

If we assume that the consumer is aware of brand j and it passes his acceptability criteria, we now examine if it passes a test of sufficient utility. To determine that we first examine the evaluation decision (see Figure 1) following a method proposed by Roberts (1983, pp 219-220).

Using a Fishbein model (Fishbein 1967) we can model attitude towards (or utility for) brand j, Uj, as a weighted linear function of its attributes. Thus,

U_j = S w_k y_jk (2)

where U_j is the utility of brand j,

w_k is the importance weight of attribute k, and

y_jk is the amount of attribute k that brand j has.

McFadden's logit model, a special form of Luce's axiom, may be used to relate Uj, to its probability of purchase. The probability of selecting j given a consideration set K is

EQUATION

It should be noted that with the logit model comes the assumption of the independence of irrelevant alternatives (e.g., Ben Akiva and Lerman 1985). The implications of that to consideration are considered under future research.

FIGURE 1

MODEL OF CONSUMER CHOICE FOR AN INDIVIDUAL BRAND

The expected utility that the consumer will derive form the category given a consideration set k, E(U_k) is given by Ben Akiva and Lerman (1985):

EQUATION (3)

Given this formula it is possible to derive a utility criterion for a brand j to enter the consumer's consideration set w. Brand j will enter the

consideration set if the increase in expected category utility that it causes more than offsets the associated mental and physical transaction costs (C_j).

That is V_j = 1 (brand j is of sufficient utility) if

E(U_kuj)>E(U_k)+C_j   (4)

Substituting in the expression for E(Uk) we can express the criterion either as the maximum search costs that a consumer will bear, given a brand's expected utility, or the minimum level of utility it needs to enter the consideration set, given the physical and psychological costs of evaluating it ("search costs"):

EQUATION (5)

EQUATION (6)

We have expressed the utility needed for brand j to enter the consideration set in terms of the utility of other brands in the set and its search costs. The compensatory nature of this phase of the consideration process may be seen from the fact that individual attributes enter in a compensatory manner (using equation (2)). Note that if C_j = O brand j will be considered on this utility criterion for any U_j>0. A number of implications stem from this model of the compensatory process. These have been expressed as lemmas below.

Lemma 1 As a brand's search costs increase, the necessary utility to maintain consideration increases more than proportionately. This result stems directly from differentiating inequality (6) at its indifference point with respect to search costs:

EQUATION (7)

Note that this is independent of the utilities of other brands in the consideration set. What this means in practice is that when search costs are close to zero then a brand that is only marginally considered will have low utility. If search costs increase by say 10% the brand must increase its utility by substantially more than this to still be considered. As search costs continue to increase the necessary proportionate utility gain to maintain consideration asymptotes to 1.

Lemma 2 Given a consideration set k and EQUATION =k, brand j is preferred for inclusion to brand j' if

EQUATION (8)

Note that the decision is dependent on the risk-adjusted preference of other brands already within the consideration set if search costs are not equal. For equal search cost inequality (8) degenerates to preferring j to j' if U_j >U_j. Inequality (8) says that the extra search costs that a brand with high utility can justify depend on the expected utility that it can add to the category relative to another candidate. That is, the marginal utility of a brand to consideration (its "consideration value") is a function of other members of the set. Lemma 2 suggests that if brand j adds 10% more utility to the existing consideration set than brand j, i.e., (k + exp(U_j)/(k +exp (U_j)) = 1.1 it can afford to have up to a 10% exponentiated transaction cost disadvantage (exp (C_j) < exp (C_j). 1.1).

Lemma 3 Assume the consumer becomes aware of a new product, N. The consumer has a consideration set k. The ongoing search cost of brands already within k is C, while for the new brand it is C_N. The brand will enter the consideration set, i.e., V_N = 1, (given acceptability and awareness) if

i.e. U_N>ln k [e^CN-1] (9)

(increasing the size of the consideration set), or

EQUATION (10)

(displacing an existing member of the consideration set).

Model Summary

To return to the overall model of consideration, the inclusion of brand j in the consideration set occurs it passes both non-compensatory test (A_j=l) and if the compensatory screen (V_j=l). It passes the non compensatory test if it meets each of the attribute thresholds, while it passes the compensatory test if its multiattribute utility exceeds a benchmark utility which is a function of the utilities of brands already in the consideration set and mental transaction costs.

In summary, we have modeled consideration with two elements which to date have been only used separately to model the process. This approach helps explain some of the conflicting evidence and comparable fits of different rules, while sitting well with NaraYana and Markin's concept of inert and inept sets.

MANAGEMENT IMPLICATIONS

The model has been implemented and results from fitting it are included in Roberts and McBurney (1988). In addition to providing a tool for determining levels of consideration of a new product (forecasting), the model provides a framework to investigate why consideration is not taking place. If it is because the brand is perceived as unacceptable then a number of options are available. Firstly, the accuracy of perceptions can be checked. For example, in an application to the Australian ready to eat cereal market, Crunchy Nut Corn Flakes was perceived as having higher sugar levels than it did, disqualifying it from many consumers' consideration sets. Secondly, the offending attribute may be changed or a variant launched; for example caffeine-free Sprite. Thirdly, the threshold may be changed. Samuel Adams and other premium-priced beers have been highly successful in persuading consumers that it is acceptable to pay an extra dollar or so for a beer. Finally, the threshold can be avoided. Of a sample of eighty Australian cereal-eaters, none said that they would reject a brand on the grounds that it was high in energy, while 32% said that they would reject a brand high in calories. Calories are, of course, just a measure of energy. A brand could be positioned as a high energy product to avoid the high calorie conjunctive screen.

Similarly, if brands are acceptable but not considered (that is they fail the compensatory utility test) diagnostic data are available to determine what is necessary to create sufficient utility. The multiattribute utility model outlined in equation 2 can be used to deter nine how attribute perceptions would have to be altered in order to generate sufficient perceived utility to be included in any segment's (or individual's) consideration set.

The above application is an example of the use of consideration sets for a brand audit to examine the health and market potential of existing products. The technique can also be used for new brand forecasting and marketing mix optimization. Gensch (1987) has shown the improvement in forecasting performance by including a consideration stage. Equations 1,2,9,and 10 predict how much consideration will be gained and how this will vary with different product formulations. Attributes important at the consideration stage may be particularly emphasized or (with conjunctive criteria) just satisfied. Examples of this include safety in airline choice, sugar levels in breakfast cereal, and manufacturer reputation in the purchase of industrial equipment (Gensch 1987). These consideration formulae may be used directly or in conjunction with other new product forecasting models, such as Assessor (SiLk and Urban 1978).

A third application to the model is in the field of product line planning. Product line strategies, which attempt to cover the whole of a population's consideration sets with single brands (e.g., Budweiser beer) can be compared to those which segment and aim for higher inepts, but lower inerts (e.g., the Bond Corporation's regional strategy in the U.S. beer market).

SUMMARY

Consideration sets as subsets of the awareness set have been shown to exist. With the increasing level of brand proliferation currently being experienced, consideration may become an increasingly binding constraint on market share. Studying the way in which the determinants of consideration sets are composed seems to be an area of great managerial relevance and academic interest. This paper has presented a model of set composition. It has a number of limitations which include the independence of irrelevant alternatives so that little can be said about the "shape" of the consideration set, no treatment of information uncertainty and its dynamics have been included so changes to the set as search in undertaken have not been addressed, and an error theory which suggests that brands will be probabilistically rather than deterministically considered is lacking. However, we have presented a grounded theory which does explain a number of the conflicting findings in this area as well as providing a testable model of the process.

REFERENCES

Allaire, Y. (1973), 'The Measurement of Heterogeneous Semantic, Perceptual and Preference Structures" unpublished Ph.D. thesis. Massachusetts Institute of Technology, Cambridge, Ma.

Belonax, J. A. (1979), "Decision Rule Uncertainty, Evoked Set Size, and Information Variability," Advances in Consumer Research, 6, 232-235.

Ben, Akiva M., and S. R. Lerman, (1985), Discrete Choice Analysis, MIT Press, Cambridge, MA.

Bettman, J. (1979), An Information Processing Theory of Consumer Choice, Chicago: Addison-Wesley.

Brisoux, J. E. and M. Laroche (1981), "Evoked Set formation and Composition: An Empirical Investigation Under a Routinized Response Behavior Situation," Advances in Consumer Research, 8, 357-361.

Brown, J. J. and A. R. Wildt (1987), Factors Influencing Evoked Set Size, Working Paper 034 87, College of Business and Public Administration, University of Missouri-Columbia, Columbia, MI.

Campbell, B. M. (1969), "The Existence of Evoked Set and Determinants of its Magnitude in Brand Choice Behavior," unpublished Ph.D. dissertation Columbia University, New York, N.Y.

Einhorn, J. (1970), "-Use of Nonlinear Noncompensatory Models in Decision Process Models: A Comparison of Compensatory and Noncompensatory Structures," Transportation Research, 13A, 17-28.

Fishbein, M. (1967), "Attitude and the Prediction of Behavior" in M. Fishbein, ed., Readings in Attitude Theory and Measurement, John Wiley and Sons, New York, N.Y.

Gensch, D. H. (1987), "A Two-Stage Disaggregate Attribute Choice Model," Marketing Science, 6, No. 3, 223-231.

Gronhaug, K. (1973-74), "Some Factors Influencing the Size of the Buyer's Evoked Set," European Journal of Marketing, 7, 232-241.

Gronhaug, K. and S. V. Troye (1983), "Exploring the Content of the Evoked Set in Car Buying," European Research, 11, No.3, 98-104.

Hauser, J. R., J. H. Roberts and G. L. Urban (1983), "Forecasting Sales of a New Consumer Durable," in Zufryden, F. S. (Ed.), Advances in Practices of Marketing Science, University of Southern California, Los Angeles, 115-128.

Hauser, J. R. and S. M. Shugan (1983), "Defensive Marketing Strategies," Marketing Science, 2, No 4, 319-360.

Howard, J. A. (1963), Consumer Behavior: Application of Theory, New York: McGraw-Hill.

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

Jarvis, L. and J. Wilcox (1973), "Evoked Set Size: Some Theoretical Foundations and Empirical Evidence," Combined Proceedings, 35, Chicago, Illinois: American Marketing Association, 236-240.

Kosnik, T. J. (1986), "Flexibility Seeking and Delay: The Effects of Ambiguity of Preference and Experience on Selecting a Short List," Working Paper, Harvard Business School, Boston, MA.

Kotler, P. (1988), Marketing Management: Analysis, Planning, Implementation, and Control, 6th Ed.. 1988. Englewood Cliffs, NJ: Prentice Hall International.

LeBlanc, R. P. (1981), "Organizational Purchase Decision Making: Information-Processing Strategies and Evoked Sets of Qualified Suppliers," Unpublished Ph.D. theses. The University of Arizona.

Lehtinen, O. (1974), "A Brand Choice Model: Theoretical Framework and Empirical Results," Journal of European Research, 2, 51-68.

Lussier, D. A. and R. W. Olshavsky, (1979), "Task Complexity and Contingent Processing in Brand Choice," Journal of Consumer Research, 6, 154-165.

Maddox, R. N., K. Gronhaug, R. E. Homans and F. E. May (1978), "Correlates of Information Gathering and Evoked Set Size for New Automobile Purchasers in Norway and the U.S.," Advances in Consumer Research, 5, 167-170.

Mahajan, V., E. Muller and S. Sharma (1984), "An Empirical Comparison of Awareness Forecasting Models of New Product Introduction," Marketing Science, 3, 179-197.

Miller, G. A. (1956), 'The Magic Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information," The Psychological Review, 63 (March 1956), 81-97.

Myer, J. H. (1979), "Methodology Issues in Evoked Set Formation and Composition," Advances in Consumer Research, 6, 236-237.

Narayana, C. L. and R. J. Markin (1975), "Consumer Behavior and Product Performance: An Alternative Conceptualization," Journal of Marketing, 39, 1-6.

Ostland, L. E. (1973), "Evoked Set Size: Some Empirical Results," in Greer, T.V. (Ed.), Combined Proceedings: Fall Conference of the American Marketing Association, Chicago: American Marketing Association, 226-230.

Parkinson, T. L. and M. Reilly (1979), "An Information Processing Approach to Evoked Set Formation," Advances in Consumers Research, 6, 227-231.

Ratchford, B. T. (1980), "The Value of Information for Selected Appliances," Journal of Marketing Research, 17, 14-25.

Ries, A. and J. Trout (1986), Positioning: The Battle for Your Mind, Singapore: McGraw-Hill.

Reilly, M. and T. L. Parkinson (1985), 'Individual and Product Correlates of Evoked Set Size for Consumer Package Goods," Advances in Consumer Research, 12, 492-497.

Roberts, J. H. (1983), "A Multiattribute Utility Diffusion Model - Theory and Application to the Pre-Launch Forecasting of Automobiles," Unpublished Ph.D. thesis, Sloan School of Management, MIT, Cambridge, MA.

Roberts, J. H. and P. McBurney (1988), "Application of A Grounded Model of Consideration Set Size and Composition," Working Paper, Australian Graduate School of Management, University of New South Wales, Kensington, 20-33.

Shugan, S. M. (1980), 'The Cost of Thinking," Journal of Consumer Research, 7, 2, 99-111.

Silk, A. J. and G. L. Urban (1978), "Pre-Test Market Evaluation of New Packaged Goods: A Model and Measurement Methodology," Journal of Marketing Research, 15, 171-191.

Stewart, D. W. and G. N. Punj (1982), "Factors Associated with Changes in Evoked Set Among Purchasers of New Automobiles," An Assessment of Marketing Thought and Practice, Proceedings of the 1982 Educators Conference, American Marketing Association, Chicago, IL.

Stigler, G. (1961), 'The Economics of Information," Journal of Political Economy, 69, 213-225.

Sutton, R. J. (1984), "Analyzing the Development and Changes in Evoked Sets as a Result of Changes in the Attribute Space," Unpublished Ph.D. thesis, University of Iowa.

Wright, P. R. (1975), "Consumer Choice Strategies: Simplifying vs. Optimizing," Journal of Marketing Research, 12, 60-67.

and P. Barbour. (1977), "Phased Decision Strategies: Sequels to Initial Screening,' in Starr, M. K. and M. Zeleny, (Eds.), North Holland TIMS Studies in the Management Science: Multiple Criteria Decision Making, 6, Amsterdam: North Holland, 91-109.

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