Evoked Set Formation and Composition: an Empirical Investigation Under a Routinized Response Behavior Situation

Jacques E. Brisoux, University of Quebec at Three-Rivers
Michel Laroche, Concordia University
ABSTRACT - The purpose of this study is to test some hypotheses regarding evoked set formation under routinized response behavior for regular beer drinkers under a specific scenario. By using a decomposition approach of the perceptual product space, and individual configurations, it is found that the best decision model is the conjunctive one, and that all brands included in the evoked set are previously tried.
[ to cite ]:
Jacques E. Brisoux and Michel Laroche (1981) ,"Evoked Set Formation and Composition: an Empirical Investigation Under a Routinized Response Behavior Situation", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 357-361.

Advances in Consumer Research Volume 8, 1981      Pages 357-361


Jacques E. Brisoux, University of Quebec at Three-Rivers

Michel Laroche, Concordia University


The purpose of this study is to test some hypotheses regarding evoked set formation under routinized response behavior for regular beer drinkers under a specific scenario. By using a decomposition approach of the perceptual product space, and individual configurations, it is found that the best decision model is the conjunctive one, and that all brands included in the evoked set are previously tried.


Evoked Set

The proliferation of brands occurring in the consumer goods markets create information processing problems for the consumers. They must often devise means to simplify their purchase decisions. One of the results of this simplification process is the existence of what Howard has referred to as "evoked set," which include the brands the buyer considers when he contemplates purchasing a unit of the product class (Howard 1963, Howard and Sheth 1969). First empirically studied by Campbell (1969), the concept of evoked set has caused a growing interest from researchers in consumer behavior (Ostlund 1973, Jarvis and Wilcox 1973, Gr°nhaug 1973-1974, May and Homans 1976, Williams and Etzel 1976, Maddox et al. 1977, Belonax and Mittelstaedt 1977, Parkinson and Reilly 1979).

Inert and Inept Sets

Narayana and Markin (1975) have suggested a more complete conceptualization of the output of the consumer's simplification process: The individual may not be aware of all the brands available (total set) at time of purchase. This leads to the concepts of awareness set and unawareness set. Within the former any brand is categorized as either evoked, inept or inert. According to Narayana and Markin, a brand would belong to the evoked set if it were considered for purchase by the consumer and evaluated positively. It would be in the inept set if it were rejected from the consumer's purchase consideration and evaluated negatively. Finally, it would be in the inert set if it were neither accepted nor rejected by the consumer and evaluated neither positively nor negatively (p. 2). The measurement of the first two sets is done through direct questioning only on the buying consideration dimension, while the inert set is inferred by deduction of the first two from the awareness set (p. 3). As they noted later, the evaluations attached to the different sets are a precondition to the consumer choice behavior (Narayana and Markin 1976).

Evoked Set Formation and Composition

In his review of the papers by Belonax (1979), May (1979), and Parkinson and Reilly (1979), Myers (1979) raises a number of conceptual and methodological issues concerning the temporal aspects of evoked sec formation and composition. He suggests two alternative conceptualizations of these processes:

a.  "Evoked sets are formed over time by considering each new entry into the awareness set whenever it comes along. It would seem that a conjunctive or a disjunctive processing model would be much more appropriate for this type of evaluation." (p. 236).

b.  Evoked sets are formed by considering all brands in the awareness set at the same point in time, and in this case the most appropriate decision rules are the linear compensatory and lexicographic models. If a new brand enters the awareness set, the whole process is repeated. The results of Parkinson and Really are congruent with the second conceptualization, although the authors call for caution in interpreting their results due to a number of methodological factors: Low sample sizes, student respondents, halo effects, and measurement problems (1979 p. 229).

Myers (1979) describes a phasing model by which consumers may establish their evoked set by using one cutoff model (i.e., conjunctive or disjunctive), followed by an evaluation for each brand within this set for purposes of choice by using a compensatory model. This conception is congruent with the findings by Best (1976), and Pras and Summers (1975), the propositions of Belonax and Mittelstaedt (1977) and Belonax (1979), as well as the paradigm developed by May (1979). This paradigm combines the information processing approach proposed by McGuire (1976) with the learning theoretical approach developed by Howard (1977). In particular, May suggests that Howard's three stages of concept learning, i.e., concept formation, attainment and utilization, correspond to the three levels of information processing complexity related to the exposure search and retrieval steps. The highest level of information processing complexity is required for the first learning stage (concept formation). The third learning stage (concept utilization) requires the least complexity.

Also, according to Howard (1977), at the concept formation stage, consumers identify their evoked set in terms of untried brands only. At the concept attainment stage they identify it in terms of mixed brands (i.e., untried and tried ones). At the concept utilization stage (Routinized Response Behavior), they identify it in terms of tried brands only.

The purpose of this research is to test what Myers called a phasing model, especially its first stage (i.e., a cutoff model is more appropriate), and to test some part of May's conceptualization of combining both learning and information processing hypotheses, including Howard's proposition of the evoked set composed of tried brands only in an RRB situation.


In order to carry out this research and provide some control over possible sources of variation, a number of strategic decisions were made.

Purchase Scenario and Ideal Universe

The product class which vas selected was beer to be purchased for regular and individual consumption by male beer drinkers in the evening before dinner within one major metropolitan area of Quebec (Canada), namely the Three-Rivers area. The scenario was selected because it is typical of beer consumption in that area. This decision is believed to be critical since the choice of a decision rule may be affected by situational factors (Hansen 1976), and prior familiarity (Park 1976). The area selected is relatively homogeneous in terms of sociodemographic characteristics, and the ideal universe is defined as male drinkers residing within the region at the time of the study and consuming at least three 12 oz. bottles of beer a week. This operationalization of a regular drinker corresponds to the industry's definition. Finally, beer is a product at the maturity stage for most regular drinkers, and the purchase decision may be classified as Routinized Response Behavior (Howard 1977).

Sampling Method and Sample

The operational universe is restricted to the male regular beer drinkers, French speaking, 18 years old or more residing in the Three-Rivers, Shawinigan Metropolitan area. This area includes the cities of Three-Rivers, Cap-de-la-Madeleine, Shawinigan and Grand-MOre in the Province of Quebec and is located 90 miles east of Montreal. Its total population is about 150,000 people. In accordance with the rules of area sampling, residential streets were randomly selected from each city. On each street residences were then systematically selected. In each housing unit one regular male beer drinker was personally interviewed. In the case of several regular drinkers within a unit, only one was interviewed by randomly selecting by age. A total of 732 male subjects were contacted. Of these subjects 453 qualified as regular beer drinkers and 373 questionnaires were usable for the purpose of a much larger study than the one reported here and dealing with the content and size of evoked set (Brisoux 1980).

Data Collection

The data collection was done through personal structured interviews at the interviewee's residence. Each interviewer had an individual two hours training period and his work was checked regularly by a supervisor. Interviews were conducted mainly during the evening, and lasted about 80 minutes.


The parts of the interview of concern here consisted of getting essentially three kinds of information:

(1)  Brand perceptions on selected scales: The selection of specific scales was made from the results of a factor analysis of the brand perceptions collected previously in a large scale (private) survey which used 37 scales. The decision to limit to 11 the number of scales was due to the fact that respondents had to rate 13 brands of beer available plus an ideal brand. Respondents were asked to rate all brands they were aware of one attribute at a time on a 9-point bipolar scale like the following one:


The other labels used were


No/Much after taste

Easy/Difficult to digest

Does not create/Creates some discomfort (headache, stomach ache, etc.)

Has less/more calories than the average beer

Very/Little pronounced taste

Very/Little tart taste

Becoming less/more popular

Would/Would not serve it to my friends

Less/More salty taste than the average beer

Finally the wording of each scale was based on the most common semantic structure of regular beer drinkers (Allaire 1972).

(2) Brand classifications: During the interview, each respondent was given a set of thirteen cards, in random order, each with a different authentic beer label (aided recall procedure). The respondent was then asked to indicate the brands of which he was aware, end the others were discarded. Next, using the remaining cards, he was asked through a sequence of questions to classify each brand into one of three categories (Narayana and Markin 1975):

(a) The evoked set consisting of the brand which he would consider buying for his personal consumption in the evening before dinner, plus other acceptable brands to be bought for the same purpose;

(b) The inept set consisting of the brands the con-tamer would refuse to buy for the same consumption situation;

(c) The inert set consisting of the remaining brands, i.e.. those which he would not consider nor reject.

(3) Prior experience: The respondent was given a list of brands and asked to indicate which ones he had tried before,


The purpose of the analysis was to develop a perceptual space for each segment within our sample and decompose this space according to some compensatory and non-compensatory decision models. The analysis proceeded in steps:

Step 1: By using the data on brand perceptions, determine if there are any segments of homogeneous perceptions.

Step 2: For each segment, derive a product space by using Johnson's procedure (1971). In this procedure a product space is built by means of multiple discriminant analysis. This technique permits to find linear orthogonal combinations of attributes which best separate the brands, maximizing the ratio of between-brand to within-brand variances. Average scores on each discriminant function permits to position each brand in the discriminant space.

Step 3: For each individual configuration, i.e., following a micro approach, decompose the product space a posteriori according to six decision models; three non-compensatory models: Conjunctive, disjunctive, lexicographic; and three compensatory models: One linear additive and two geometric ones (cantered on the ideal point or the preferred brand). Figure 1 graphically represents how each model works: The shaded area would be the evoked set zone if each model were true. For each case, the evoked set zone is defined as the smallest area containing all evoked brands. A decision rule is considered successful if that zone does not contain any brand other than the evoked ones. If not, find out to which category these "offenders" belong.

Step 4: Determine the composition of the individual evoked sets in terms of prior trial of the brands.


The findings will follow the four steps defined in the previous section.

Step 1: Identify segments with homogeneous perceptions. In order to identify clusters of respondents on brand per-captions, two different clustering techniques were used, that is the Howard-Harris cluster analysis, and the BMDP-2M hierarchical cluster analysis. The Howard-Harris program uses the criterion of minimum within-group variance at each level of clustering to form groups of objects using an objects-by-variable matrix (Howard & Harris 1966). The BMDP-2M procedure is also a hierarchical method. The two cases having the shortest distance between them are amalgamated and treated as one case and then, in turn, clustered with others. The procedure continues until all cases and clusters are amalgamated into one cluster.



Both techniques led to a single cluster: The whole group of 375 respondents, and, thus, the rest of the analysis is done on the total group.

Step 2: Generate a perceptual product space. The data on brand perception is then used for a multiple discriminant analysis of scale for 375 respondents, following the method developed by Johnson (1971). This method is particularly suited to our needs in terms of stability, uniqueness of the solution and Euclidean distances (Johnson 1971). It was found that three discriminant functions are significative and explain respectively 77.8%, 17.5% and 2.3% of the total variance, for a cumulative percentage of 97.6% of total variance. The first function may be interpreted as the social dimension of beer drinking (before), the second as the strength of beer (during), and the third as another taste dimension (after) concerning physiological effects (after taste, easy to digest).

Step 3: Determine the best decision rule for evoked set formation. In order to find out the best decision for each individual, we derived individual scores on the three significative discriminant functions. For each respondent, and each decision rule, we determined whether or not a particular decision rule leads to a zone containing only evoked brands. For example, in the geometric compensatory rule cantered on the ideal point (Best 1976), the evoked set zone is defined as the sphere centered on the ideal point and the radius of which is the evoked brand farthest away. We then looked to see if one or more brand(s) other than an evoked one is contained within the zone (sphere). For the cutoff models, the cutoff point is inferred by taking the evoked brand with the lowest score on that dimension. The evoked set zone is then the area represented in dark in Figure 1, and an examination of that zone reveals whether or not it contains brands other than evoked ones.

The percentage of cases for which each rule is successful is presented in Table 1 as well as the classification of the brands belonging to each sec violating that rule. It is clear that the best rule is the conjunctive model for which no other brand was found within the centroid defined by the three cutoff points for 72.0% of all respondents. Most violations of the rule occurred with inert brands, with only 5.3% of the cases for which an inept brand was found within the centroid. This result can also be viewed as follows: The percentage of success increases dramatically as we move from the largest area, i.e., the disjunctive rule with 9.6% success, to a smaller area, i.e., the lexicographic rule with 46.1% success, to an even smaller area, i.e., the conjunctive rule with 72.0% success. A similar reasoning with the three compensatory models reinforces the conclusion that the process of evoked set formulation is, in our case, one of inclusion based on cutoff points for all three major dimensions of the product space. This conclusion is consistent with Myers (1979) expectations, as well as those by Pras and Summers (1977). It is also consistent with the findings of Best (1976). On the other hand, our findings do not confirm those of Parkinson and Reilly (1979) although they followed a composition approach. They do report that the questions directly measuring the conjunctive or disjunctive cutoffs were fairly difficult for the subject to complete, and that these response errors were probably magnified in their simulation procedures.

Step 4: Determine prior trial of evoked brands. A comparison was made between the brands in the evoked set on one hand, and those which the respondent had indicated as tried brands. It was found that in 95.4% of all cases, all brands in the evoked set had been previously tried. The remainder of the cases, i.e., 4.6%, may be considered as an error term if we set a level of significance of a = 5%, in which case the critical value is Pc = .069. Therefore we fail to reject the null hypothesis that all brands in the evoked set have been previously tried. This is consistent with Howard's theory of concept learning in a Routinized Response Behavior Situation (Howard 1977, May 1979).


A first limitation of the present study deals with the problem of controlling for the homogeneity of respondents since many factors may affect the choice of the decision rule as reviewed by Belonax (1979, p. 232). Our decision to limit our research to a RRB situation, a specific scenario, regular male beer drinkers, in a relatively homogeneous metropolitan area as well as two attempts at clustering respondents provided some measure of control. But other factors may be introduced such as breadth of categorization (Pettigrew 1956) which is one dimension of cognitive style, as well as perceived risk (Bettman 1973).



A second limitation may lie in our procedure of decomposing the product space into areas based on a specific decision rule on one hand and actual brand classification on the other hand. In effect, we tested the logic of applying a specific rule more than reconstructing the rule itself as in Parkinson and Reilly (1979).

Third, the boundaries of our evoked set zones were approximated by looking at the smallest area containing all evoked brands. There may be some value in doing some sensitivity analysis by varying the cutoff points to allow for a margin of error. Finally, we only looked at minimum cutoff points, but not at maximum ones. This may apply to only our second and third dimensions since the first dimension is the social image of beer which accounted for 77.8% of total discrimination.

Finally, the study is within a static context, and not in a dynamic one as suggested by Myers (1979).


This research has attempted to focus on specific questions regarding Myers' phasing model, as well May's contention that both learning theory and information processing theory are necessary to explain evoked set formation and composition.

We selected a product in the maturity stage and consumers in an RRB situation, faced with a specific scenario, and homogeneous in their perception of the available brands. For each respondent, we generated his perceptual space, and we used a decomposition approach to test three non-compensatory rules and three compensatory rules. In a perfect criterion basis, the best rule was found to be the conjunctive model, which is consistent with previous expectations and findings. Finally, all brands in the evoked set were found to have been previously tried, which, is consistent with Howard's learning theory.

The decomposition approach followed by this study allowed us to examine and compare the output at the maturity stage (RRB situation) from both information processing activities and concept learning. From an information processing viewpoint, we gathered brand perceptions on several attributes and constructed individual product spaces. From a concept learning viewpoint, all brands in the evoked set were previously tried by most respondents. Thus a comparison of brand classifications and product space configurations leads to a perfect match for 72% of all respondents for the conjunctive rule. Thus, we may see here a convergence of both theories as hoped by May (1979).


Allaire, Yvan (1972), "The Measurement of Heterogeneous Semantic, Perceptual and Preference Structures," unpublished Ph.D. dissertation, Massachusetts Institute of Technology.

Belonax, Joseph J. (1978), "Decision Rule Uncertainty, Evoked Set Size, and Task Difficulty as a Function of Number of Choice Criteria end Information Variability," in Advances in Consumer Research, Vol. 6, ed. William L. Wilkie, Urbana, IL.: Association for Consumer Research, 232-235.

Belonax, Joseph J. Jr. and Mittelstaedt, Robert A. (1978), "Evoked Set Size as a Function of Number of Choice Criteria and Information Variability," in Advances in Consumer Research, Vol. 5, ed. H. Ketch Hunt, Urbana, IL: Association for Consumer Research, 48-51.

Best, Roger J. (1976), "The Predictive Aspects of a Joint-Space Theory of Stochastic Choice," Journal of Marketing Research, 13, 198-204.

Bettman, James (1973), "Perceived Risk and its Components: A Model and Empirical Test," Journal of Marketing Research, 10, 184-190.

Brisoux, Jacques E. (1980), "Le PhenomFne des Ensembles EvoquTs: Une Etude des Dimensions Contenu et Taille," unpublished Ph.D. dissertation, UniverstitT Level.

Campbell, Brian M. (1969), "The Existence and Determinants of Evoked Set in Brand Choice Behavior," unpublished Ph.D. dissertation, Columbia University.

Gr°nhaug, Kjell (1973), "Some Factors Influencing the Size of the Buyer's Evoked Set," European Journal of Marketing, 7, 3, 232-241.

Hansen, Fleming (1976), "Psychological Theories of Consumer Choice," Journal of Consumer Research, 3, 117-142.

Howard, John A. (1963), Marketing Management, Analysis and Planning, Irwin.

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

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

Jarvis, Lance P. and Wilcox, James B. (1973), "Evoked Set Size - Some Theoretical Foundations and Empirical Evidence," in Combined Proceedings, ed. Thomas V. Greer, Chicago: American Marketing Association, 236-240.

Johnson, Richard M. (1971), "Market Segmentation: A Strategic Tool," Journal of Marketing Research, 5, 13-18.

Maddox, R. Neil, Gr°nhaug, Kjell, Homans, Richard E., and May, Frederick E. (1978), "Correlates of Information Gathering and Evoked Sets Size for New Automobile Purchasers in Norway and the U.S. ," in Advances in Consumer Research, Vol. 5, ed. H. Keith Hunt, Chicago: Association for Consumer Research, 167-170.

May, Frederick E. (1979), "Evoked Set Formation and Composition: The Learning and Information Processing Hypotheses,'' in Advances in Consumer Research, Vol. 6, ed. Wilkie, Urbana, IL: Association for Consumer Research, 222-226.

May, Frederick E., and Homans, Richard E. (1976), "Evoked Set Size and the Level of Information Processing in Product Comprehension and Choice Criteria," in Advances in Consumer Research, Vol. 4, ed. William D. Perreault, Jr., Urbana IL: Association for Consumer Research, 172-175.

McGuire, William J. (1976), "Some Internal Psychological Factors Influencing Consumer Choice," Journal of Consumer Research, 2, 302-319.

Myers, James R. (1979), "Methodological Issues in Evoked Set Formation end Composition," in Advances in Consumer Research, 6, ed. William L. Wilkie, Urbana, IL: Association for Consumer Research, 236-237.

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

Narayana, Chem L., and Markin, Ron J. Jr. (1976), "An Explanation,'' Journal of Marketing, 40, 60.

Park, Whan C. (1976), "The Effect of Individual and Situation-Related Factors on Consumer Selection of Judgmental Models," Journal of Marketing Research, 13, 144-151.

Parkinson, Thomas L., and Reilly, Michael (1979), "An Information Processing Approach to Evoked Set Formation," in Advances in Consumer Research, Vol. 6, ed. William L. Wilkie, Urbane, I1.: Association for Consumer Research, 227-231.

Pettigrew, Thomas P. (1956), "The Measurement and Correleates of Category Width as a Cognitive Variable," Journal of Personality, 6.

Pras, Bernard, and Summers, John (1975), "A Comparison of Linear and Nonlinear Evaluation Process Models," Journal of Marketing Research, 12, 276-281.

Ostlund, Lyman E. (1973), "Evoked Set Size: Some Empirical Results," in Combined Proceedings, Series No. 35, ed. Thomas V. Greer, Chicago: American Marketing Association, 226-230.

Sherif, Caroline W., Sherif, Muzafer and Nebergall, Robert E. (1965), "Attitude and Attitude Change: The Social Judgement - Involvement Approach, Philadelphia: Saunders.

Williams, Robert G. and Etzel, Michael J. (1976) "An Investigation and Extension of the Evoked Set Concept Applied to Consumer Durables," Proceedings of the Southern Marketing Association.