An Information Processing Approach to Evoked Set Formation

Thomas L. Parkinson, The Pennsylvania State University
ABSTRACT - This exploratory study investigates the dynamics of evoked set formation from an information processing perspective. Subjects were asked to identify actual evoked sets and to provide the necessary attribute rating data for simulating an evoked set decision utilizing five information processing strategies. When the actual and simulated evoked sets were compared, the best "fits" resulted from application of the unweighted linear compensatory and lexicographic strategies.
[ to cite ]:
Thomas L. Parkinson (1979) ,"An Information Processing Approach to Evoked Set Formation", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 227-231.

Advances in Consumer Research Volume 6, 1979      Pages 227-231

AN INFORMATION PROCESSING APPROACH TO EVOKED SET FORMATION

Thomas L. Parkinson, The Pennsylvania State University

Michael Reilly, The Pennsylvania State University

ABSTRACT -

This exploratory study investigates the dynamics of evoked set formation from an information processing perspective. Subjects were asked to identify actual evoked sets and to provide the necessary attribute rating data for simulating an evoked set decision utilizing five information processing strategies. When the actual and simulated evoked sets were compared, the best "fits" resulted from application of the unweighted linear compensatory and lexicographic strategies.

PROBLEM

With the proliferation of brands that has occurred in the marketplace, it is well known and widely accepted that a consumer only considers a subset of the available brands when making a purchase decision. Howard and Sheth (1969) termed this subset of "acceptable" brands the evoked set, and this concept has received widespread acceptance in marketing in spite of its lack of empirical validation. The research which has been reported concerning the evoked set has been static in nature focusing primarily on the situational and individual factors that influence the evoked set size of an individual in a given purchase situation.

Campbell (1969), Ostland (1974) and Gr°nhaug (1973/74) investigated the relationship of evoked set size and a large number of standard marketing variables, demographic characteristics, and selected individual difference variables. In general, these studies failed to produce significant results, and in most cases only a small percentage of the variance in evoked set size was explained. Only brand loyalty and level of education were found to be related to evoked set size. Using social judgment theory as a basis, Jarvis and Wilcox (1973) investigated the relationship between perceived importance of the product class and evoked set size; and, as hypothesized, demonstrated a positive relationship between the two. These findings coupled with the work of May and Homans (1977) which demonstrated a relationship between the abstractness of an individual's information processing (cognitive style) and evoked set size support the proposition that the size of the evoked set is at least in part an individual variable, related to cognitive style.

The evidence that the size of the evoked set is related positively to education level and degree of abstractness in product perception; together with other studies which have suggested the existence of multiple subsets in addition to the evoked set (Jacoby, 1973; Narayana and Markim, 1975) suggests that the formation of evoked sets is a cognitive process amenable to study from an information processing perspective. Recently, Belonax and Mittelstaedt (1978) suggested such an approach and postulated a two-stage model with evoked set formation as a prerequisite for further brand evaluation. However, they too utilized evoked set size as the dependent variable, and found that set size was inversely related to the number of choice criteria and information variability. Thus in spite of the significance of the evoked set to marketing practitioners, to date there has been no attempt to study the dynamic process of evoked set formation.

Information Processing

This lack of a dynamic approach can, perhaps be traced partly to a lack of a sound theoretical base for such an inquiry. However, recent developments in the field of consumer information processing would appear to provide the tools for this research. Study of human information processing originated in the study of human clinical judgment (Dawes, 1971, 1972; Einhorn, 1970, 1971; Gold-berg, 1968, 1970, 1971; Hoffman, 1960, 1968), and in the attempts to program computers to simulate human problem solving (Newell, Shaw and Simon, 1958; Newell and Simon, 1972; Simon and Newell, 1970). For recent reviews of this research and its applicability to marketing problems see Bither and Ungson (1975), and Jacoby and Chestnut (1977).

One of the major contributions of this research has been the isolation of a number of algorithmic judgment strategies. These strategies notably weighted and unweighted linear compensatory, conjunctive, disjunctive, and lexicographic or elimination by aspects have been hypothesized to represent consumer judgment strategies during purchase behavior, and several methods for measuring them have been developed (Jacoby, Speller, and Kohn-Berning, 1974, 1974b; Einhorn, 1970; Goldberg, 1971; Bettman, 1974; Wright, 1975; Reilly and Holman, 1977).

If one views the formation of an evoked set by a consumer when faced with a multitude of brands as an attempt to simplify the decision making environment (Miller, 1956; Wright, 1974, 1975), then the consumer may use the previously discussed strategies to accomplish this. By only considering a subset of the available brands in a purchase situation, a consumer would only be required to process a fraction of the available product-related information. Thus, the existence of an evoked set for a product category would seem to imply the operation of at least a two-stage decision strategy. In the first stage the individual decides which brands to consider (i.e., forms the evoked set) by the application of a processing strategy. Then, when a purchase situation arises the consumer applies another strategy, or possibly reapplies the same strategy, to the elements of the evoked set to make a decision.

In an actual temporal sense what probably happens is that the consumer considers one brand at a time and makes a binary decision as to whether or not the brand is worthy of future consideration. If the brand is deemed suitable it enters the evoked set. If not, it is rejected. This binary decision is made as soon as the consumer feels that he has enough information to evaluate the brand. The purposes of this research is to investigate which processing strategy is most frequently used by consumer in formulating their evoked set for a given product class. Specifically, the following five strategies will be evaluated for their ability to approximate actual evoked set decisions.

1. Unweighted Linear Compensatory: When utilizing this strategy the individual is seen as summing the attribute ratings for each available brand and then including all brands whose final ratings exceed a certain cutoff in the evoked set.

2. Weighted Linear Compensatory: When utilizing this strategy the individual is seen as summing the attribute ratings, weighted by their importance, for each available brand and then including all brands whose final ratings exceed a certain cutoff in the evoked set.

3. Conjunctive: When utilizing this strategy the individual is seen as including in the evoked set only those brands whose attribute ratings exceed predetermined cutoffs on all attributes.

4. Disjunctive: When utilizing the strategy the individual is seen as including in the evoked set all brands which exceed a predetermined cutoff on at least one attribute.

5. Lexicographic: When utilizing this strategy the individual is seen as rank ordering the attributes of the brands on an importance dimension, and then ranking all available brands based on their rating on the most important attribute. Ties are seen as being broken by using the second most important attribute and so forth. Then only brands which exceed a certain cutoff are included in the evoked set.

METHODOLOGY

In order to investigate the possible application of various information processing strategies to the formation of an evoked set, a written questionnaire was designed and administered to 90 undergraduate students enrolled in three sections of the University's consumer behavior course during the spring term. The instrument first asked the subjects to identify their evoked sets for two product categories and then to provide the additional data necessary for an individual actuarial analysis of the ability of the postulated strategies to correctly duplicate the actual evoked sets of the respondents. The two products selected for inclusion in the design were toothpaste and underarm deodorant. These were selected based on their high level of use and familiarity among students. However, each subject was asked about only one product category due to the length of the questionnaire.

The actual evoked set for each product category was established by presenting the subjects with a list of available brands (8 brands of toothpaste; 12 brands of deodorant), and asking them to indicate those brands with which they were familiar (the awareness set). They were asked to indicate which brands in the awareness set they would consider buying if faced with an immediate purchase decision. This second measure was identified as the actual evoked set for the product category.

Once the actual evoked set was identified, the subjects were asked to evaluate each brand in the awareness set on a number of attributes. These attributes were selected based on previous research (Haley, 1968; Ryan, 1978) and an analysis of current promotional literature. The following attributes were included:

TOOTHPASTE             DEODORANT

Taste                             Smell

Decay Prevention           Ease of Application

Bad Breath Prevention   Effectiveness

Abrasiveness*                 Likelihood of Staining Clothes*

Mouth Refreshment        Length of Protection

Teeth Whitening Ability   Economy in Use

Price*                             Skin Irritation*

                                       Price*

Each of the attributes was measured utilizing a seven point semantic scale with those attributes marked with an * being reverse scored due to their perceived negative influence on purchase behavior.

When all of the familiar brands were evaluated, then the subjects were asked to rate the importance of the attributes themselves with respect to their use in evaluating brands in the product category under consideration. Again, a seven point semantic scale was used. Following the attribute importance determination, the conjunctive and disjunctive cutoff points for each attribute were measured. For this purpose the following question formats were utilized:

Conjunctive Cutoff.

If I found that the Effectiveness of a brand was rated below (Mark the appropriate point)

Effectiveness

Very High ____: ____: ____: ____: ____: ____: ____ Very Low

I would not consider purchasing that brand regardless of how the brand was rated on other attributes.

Disjunctive Cutoff.

If I found that a brand was rated at or above this point with respect to the attribute Effectiveness, I would consider purchasing it regardless of how the brand was rated on all other attributes. (Mark the appropriate point on the scale below)

Effectiveness

Very High  ____: ____: ____: ____: ____: ____: ____ Very Low

                                            OR

The attribute effectiveness alone (regardless of how rated) is not a sufficient condition for considering a brand for purchase -- (check if applicable).

Following completion of the final part of the study, the students were introduced to the theoretical basis for the research and encouraged to ask questions about the study.

ANALYSIS

In order to determine the fit of each information processing strategy, an evoked set decision was simulated utilizing each of the five models and the results compared with the actual evoked sets identified at the start of the questionnaire.

A total of 90 subjects took part in the study divided evenly between the two products. Two questionnaires both dealing with toothpaste were Judged unusable due to missing responses and were discarded. Furthermore only subjects who rated attributes of three or more brands, and identified at least one brand in the evoked set and at least one brand in the non-evoked set were included in the analysis. This latter criterion was made necessary by the inclusion of the compensatory and lexicographic models which do not produce discrete evoked set decisions. This further elimination of subjects from consideration reduced the number of usable responses to 30 to 27 for toothpaste and deodorant, respectively.

The stimulated decision making procedure produced two types of output. In the cases of two linear compensatory and the lexicographic strategies a ranking of the brands was achieved. Brands ranked higher than the size of the actual evoked set were identified as members of the stimulated evoked set. When the conjunctive and disjunctive strategies were applied to the data the result was not a ranking, but a binary decision directly identifying membership or non-membership in the evoked set.

Once the stimulated evoked and non-evoked sets were specified, the results using each processing strategy were compared to the actual evoked set specified at the outset by the subject.

RESULTS

Two measures of "success" were utilized to evaluate the performance of the five information processing strategies and the results are reported in Table 1A and 1B. In Table 1A a "best fit" criterion was applied. The similarity of the simulated and actual evoked sets was evaluated based on the percentage of correct decisions. Those strategies which produced the highest percentages (ties included) of correct answers were credited with a "successful" result. In Table 1B the more demanding "perfect" criterion was used. Only simulated evoked sets which matched actual evoked sets exactly were counted as successes. This more stringent criterion produced absolutely lower success rates, but the relative performance of the different processing strategies remained the same.

DISCUSSION

Any discussion of the results of the current study must first recognize the exploratory nature of the research. Its major contribution lies in the conceptualization of the formation of the evoked set as dynamic, and the adoption of information processing theory as a basis for studying this process. Thus, the critical aspect of the paper is theoretical rather than empirical.

Limitations

There are three limitations of this initial study which may be operating to bias the outcomes. These include the conceptual base, the methodology and the data analysis procedure.

The shortcomings of information processing theory as a conceptual base for studying a decision whether or not to include a brand in the evoked set, are essentially the same as the limitations of information processing theory in the study of other decision processes. Specifically, there is some question concerning the use of information processing strategies as the basic unit of analysis. The idea that individuals in a decision situation use these relatively abstract and complex strategies has never received extensive empirical validation. In reality the decision process may be much more idiosyncratic, and therefore, more difficult to represent algorithmically. On the other hand these processing strategies seem to have a fairly high potential for simplifying inquiry into the decision process due to their ability to approximate a variety of different decisions. Accordingly, the use of these strategies in the current research should not be too debilitating in light of the exploratory nature of the research.

The other limitations of the current study are in the area of data collection and analysis. These shortcomings are, of course, far more significant when interpreting the results, than to the theoretical basis. The major methodological problem arose while attempting to measure the conjunctive and disjunctive cutoffs. These measures appeared to be fairly difficult for the subjects to complete. This could, however, be due in part to the inappropriateness of these strategies in making an evoked set decision, and therefore, may not be a pure methodological problem. In either case, any response errors of this type are probably magnified during the simulation procedure.

TABLE 1A

PERCENTAGE OF SUCCESSFUL EVOKED SET DETERMINATIONS: BEST CRITERION

TABLE 1B

PERCENTAGE OF SUCCESSFUL EVOKED SET DETERMINATIONS: PERFECT CRITERION

In addition, some problems may have arisen due to the fact that a significant number of subjects rated a very limited number of brands. Hopefully, this problem was reduced by including only data from subjects who rated at least three brands. However, when the number of brands rated was small, determination of the best fit between the simulated and actual evoked set was more difficult. Accordingly, care must be exercised in the interpretation of the results.

Findings

It is fairly obvious that the best fitting strategies in the current study were the lexicographic and the unweighted linear compensatory, regardless of which criteria for evaluation was used. Such a result is rather surprising in light of the differences between these two strategies. The lexicographic strategy places supreme weight on the most important attribute and other attributes are used only to break ties. The unweighted linear compensatory strategy on the other hand weights all attributes equally. It seems unusual that these two strategies consistently fit the data best. This result suggests two possible interpretations. One possible explanation for such a result could be that there were two segments of market, differing perhaps on level of involvement, who consistently used different strategies. For one group all of the attributes are equally important, while for the other group the most important dominated the decision.

A second possibility is that the result may be an artifact of the experimental procedure. The tendency of subjects to rate an object on a number of attributes in a manner that is congruent with their overall impression of the object has long been recognized. This tendency, termed halo effects has a long history in both psychological and marketing research (Symond, 1925; Thorndike, 1920; Bingham, 1939; Beckwith and Lehman, 1975, 1976; Johansson, MacLachlan and Yalch, 1976). For recent reviews see Beckwith, Kassarjian and Lehman (1978) and Huber and James (1978). In the current study, it is a defensible position that the success of the unweighted linear compensatory model is due to the subjects' attribute ratings being effected by their previous evaluation of the brands. Clearly, additional research is needed to clarify this issue. One possibility is an experiment involving the introduction of a totally new brand which there is no prior evaluation. The evoked set formation process could then be analyzed without contamination by halo effects.

A second notable outcome is the failure of the weighted linear compensatory, the conjunctive and the disjunctive strategies to accurately simulate the contents of the evoked set. This is particularly surprising for the conjunctive and disjunctive strategies which, by their very nature, are adapted to making a binary decision on the acceptability of an alternative. However, due to the aforementioned limitations of the study it would be hasty to conclude at this point that these strategies are not representative of the decision process by which a consumer generates an evoked set. Again more research is needed before a definitive statement can be made.

Future Research

Assuming that the methodological and theoretical difficulties which hindered the current research can be overcome, and we think that they can; the area of evoked set research is one of great theoretical and practical value. The use of an information processing perspective enables the consideration of a large number of factors which may possibly influence the process of evoked set formation. However, more research is needed to investigate among other things:

1. The individual and situational factors which determine the choice of a processing strategy in the evoked set formation.

2. How the way in which information about a new brand is presented influences the outcome of the evoked set decision, for the particular brand.

3. The way in which a consumer decides to reject a familiar brand currently in the evoked set when confronted with an attractive new brand. This issue is important in light of past research demonstrating the relatively constant size of evoked sets.

REFERENCES

Neil E. Beckwith, Harold H. Kassarjian, and Donald R. Lehmann, "Halo Effects in Marketing Research: Review and Prognosis," in H. Keith Hunt, (Ed.), Advances in Consumer Research, V (1978), 465-467.

Neil 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.

Joseph A. Belonax and Robert A. Mittelstaedt, "Evoked Set Size as a Function of Number of Choice Criteria and Information Variability," in H. Keith Hunt (Ed.), Advances in Consumer Research, V (1978), 48-51.

James Bettman, "Toward A Statistics for Consumer Decision Net Models," Journal of Consumer Research, 1 (June, 1974), 71-80.

W. V. Bingham, "Halo, Involved and Valid," Journal of Applied Psychology, 23 (1939), 221-229.

Stewart W. Bither and Gerardo Ungson, "Consumer Information Processing Research, An Evaluative Review," Working Series in Marketing Research, No. 28, (April, 1975), The Pennsylvania State University.

Brian M. Campbell, "The Existence of Evoked Set and Determinants of Its Magnitude in Brand Choice," Unpublished doctoral dissertation, Columbia University, 1969.

Robyn M. Dawes, "A Case Study of Graduate Admissions: Application of Three Principles of Human Decision Making," American Psychologist, 26 (February, 1971), 180-188.

Robyn M. Dawes, "Slitting the Decision Maker's Throat With Occom's Razor: The Superiority of Random Linear Models to Real Judges." Paper delivered at Seminar on Multiple Criteria Decision Making, Columbia, S. C., 1972.

Hillel Einhorn, "The Use of Nonlinear, Noncompensatory Models in Decision-Making," Psychological Bulletin, 73 (March, 1970), 221-230.

Hillel Einhorn, "The Use of Nonlinear, Noncompensatory Models As A Function of Task and Amount of Information," Organizational Behavior and Human Performance, 6 (January, 1971). 1-17.

Lewis R. Goldberg, "Simple Models or Simple Processes? Some Research On Clinical Judgments," American Psychologist, 23 (July, 1968), 483-496.

Lewis R. Goldberg, "Man vs. Model of Man: A Rationale Plus Some Evidence for A Method of Improving On Clinical Inferences," Psychological Bulletin, 73 (June, 1970), 422-432.

Lewis R. Goldberg, "Five Models of Clinical Judgment: An Empirical Comparison Between Linear and Non-Linear Representations of the Human Inference Process," Organizational Behavior: Human Performance, 6 (June, 1971), 458-479.

Kjelf Gr°nhaug, "Some Factors Influencing the Size of the Buyers Evoked Set," European Journal of Marketing, 7 (Winter, 1973/74), 232-241.

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

Paul Hoffman, "The Paramorphis Representation of Clinical Judgment," Psychological Bulletin, 57 (March, 1960), 116-131.

Paul Hoffman, "Cue Consistency and Configurality In Human Judgment,'' in B. Klernmutz, (Ed.), Formal Representation of Human Judgment, (New York: Wiley, 1968).

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

Joel Huber and William James, "A Measure of Halo," in H. Keith Hunt, (Ed.), Advances in Consumer Research, (1978), 474-476.

Jacob Jacoby, "A Model of Multi-Brand Loyalty," Journal of Advertising Research, 11, 1971, 25-30.

Jacob Jacoby and Robert W. Chestnut, "Consumer Information Processing: Emerging Theory and Findings," in Woodside et al., (Eds.), Consumer and Industrial Buying Behavior, New York: Elsevier North-Holland, Inc., 1977, 119-134.

Jacob Jacoby, Donald E. Speller and Carol Kohn-Berning, "Brand Choice Behavior As A Function of Information Load: Replication and Extension," Journal of Consumer Research, (June, 1974), 33-42.

Jacob Jacoby, Donald E. Speller and Carol Kohn-Berning, "Brand Choice Behavior As A Function of Information Load," Journal of Marketing Research, 1974b, 11 (February, 1974), 63-69.

Lance P. Jarvis and James B. Wilcox, "Evoked Set Size--Some Theoretical Foundations and Empirical Evidence," in T. V. Greer, (Ed.), Combined Proceedings Fall Conference of the American Marketing Association, No. 35, 1973, 236-240.

Johnny K. Johansson, D. J. MacLachlan and R. F. Valch, "Halo Effect in Multi-Attribute Models: Some Unresolved Issues," Journal of Marketing Research, 8 (February, 1976), 414-417.

Frederick E. May and Richard E. Homans, "Evoked Set Size and the Level of Information Processing in Product Comprehension and Choice Criteria," in W. D. Perreault, Jr., (Ed.), Advances in Consumer Research, 4, 1977, 172-175.

George A. Miller, "The Magical Number Seven, Plus or Minus Two: Some Limits On Our Capacity for Processing Information," The Psychological Review, 63 (1956), 81-97.

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

Allen Newell and Herbert Simon, "Elements of A Theory of Human Problem Solving," Psychological Review, 65 (May, 1968), 151-166.

Allen Newell and Herbert Simon, Human Problem Solving, (Engle-wood Cliffs, Prentice-Hall, 1972).

Lyman E. Ostlund, "Evoked Set Size: Some Empirical Results,'' in T. V. Greer, (Ed.), Combined Proceedings Fall Conference of the American Marketing Association, No. 35, 1973, 226-230.

Michael Reilly and Rebecca H. Holman, "Does Task Complexity of Cue Interrelation Affect Choice of an Information Processing Strategy? An Empirical Investigation," in W. D. Perreault, Jr., (Ed.), Advances in Consumer Research, 4 (1977), 185-190.

Michael J. Ryan, "An Examination of an Alternative Form of the Behavioral Intention Model's Normative Component," in H. Keith Hunt, (Ed.), Advances in Consumer Research, V (1978), 283-289.

Herbert Simon and Allen Newell, "Human Problem Solving: The State of The Theory in 1970," American Psychologist, 26 (February, 1971), 145-159.

Percival Symonds, "Notes on Ratings," Journal of Applied Psychology, 9 (1925), 188-195.

E. L. Thorndike, "A Consistent Error In Psychological Ratings," Journal of Applied Psychology, 4 (1920), 25-29.

Peter Wright, "The Use of Phased Noncompensatory Strategies In Decision Between Multi-Attribute Products," Research Paper 223, Graduate School of Business, Stanford University, August, 1974.

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

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