The Effect of Familiarity on Consumers' Choice Agendas

ABSTRACT - Consumers' choice agendas have a major impact on the information they seek and the choices they make. This paper presents empirical evidence of the top-down and bottom-up agendas described by Hauser (1986), and identifies subtypes of those agendas. Hauser's familiarity hypothesis, that bottom-up agendas will be favored at higher levels of familiarity and top-down agendas will be favored at lower levels of familiarity, may be supported for some subtypes of agendas, but not for others. In general, consumers' use of agendas, like other choice strategies, seems to be adaptive.

Citation:

Judy A. Wagner and Noreen M. Klein (1993) ,"The Effect of Familiarity on Consumers' Choice Agendas", in NA - Advances in Consumer Research Volume 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT : Association for Consumer Research, Pages: 209-214.

Advances in Consumer Research Volume 20, 1993      Pages 209-214

THE EFFECT OF FAMILIARITY ON CONSUMERS' CHOICE AGENDAS

Judy A. Wagner, Virginia Polytechnic and State University

Noreen M. Klein, Virginia Polytechnic and State University

ABSTRACT -

Consumers' choice agendas have a major impact on the information they seek and the choices they make. This paper presents empirical evidence of the top-down and bottom-up agendas described by Hauser (1986), and identifies subtypes of those agendas. Hauser's familiarity hypothesis, that bottom-up agendas will be favored at higher levels of familiarity and top-down agendas will be favored at lower levels of familiarity, may be supported for some subtypes of agendas, but not for others. In general, consumers' use of agendas, like other choice strategies, seems to be adaptive.

INTRODUCTION

"Look for the union label." "Holiday Inn. Stay with someone you know." Marketers expend considerable effort trying to change people's agendas, or the sequence of constraints they place on their decisions. Agendas are strategies of categorization and information selection that determine which attributes and alternatives are evaluated. Agendas impose a degree of organization on the buying problem, and thus make it easier for the buyer to narrow the consideration set down to the chosen alternative.

Hauser (1986) defines an agenda as a "sequence of constraints", and distinguishes between two types: top-down and bottom-up agendas. Consider his example of choosing a restaurant. The consumer may classify restaurants in terms of their cuisine - Indian, Chinese, and Italian. With a top-down agenda, the consumer constrains the choice set on the basis of that classification (e.g. deciding to eat Indian food that night), and evaluates only the restaurants that fall within a certain category. Other criteria, such as prior experience or price, may then be used to make a final selection.

A bottom-up agenda also assumes an initial categorization of alternatives. However, no single category is eliminated from consideration. Instead, a favorite is chosen from each category, and the final choice is then selected from the set of these favorites. In the example used above, the best of each of the Indian, Chinese, and Italian categories would be identified, and the final choice made from those three. Note that the alternatives in the Chinese and Italian categories have an increased probability of being chosen, compared to when a top-down agenda was used to eliminate these categories without looking at specific alternatives.

Changing the agenda structure, for instance by changing the initial constraint from cuisine to price level, may have a major impact on the final choice. In addition, the manner in which an agenda is processed (top-down versus bottom-up) alters both the probability of an alternative being evaluated and the set of alternatives to which it is directly compared. Both the structure of the agenda and the way it is processed should therefore affect choice.

From this intuitive restaurant example, one can imagine just how frequently some form of agenda is used by consumers in simplifying their purchase decisions. It is also easy to see how understanding agendas presents an opportunity for marketers. Because agendas change the choice probability of specific alternatives, the ability to influence how consumers partition and constrain their alternative sets could influence the marketplace success of a particular product or service. The advertising examples given above illustrate marketing efforts to constrain alternatives in a manner that shows the focal product to advantage.

Given the importance of agendas as a means of simplifying and influencing consumer choice, the following issues are of interest to both academicians and practitioners.

1. How closely do agendas used in consumer choices approximate the types formally modeled by Hauser (1986)?

2. What individual and situational factors lead to the use of top-down versus bottom-up agendas?

3. What is the impact of an agenda on the objective (accuracy and efficiency) and subjective (perceived difficulty and confidence) outcomes of a choice?

4. What marketing strategies most effectively influence agendas?

5. In what situations is a consumer's agenda most subject to external influence?

A series of programmatic studies is needed to answer these questions. In the current paper, we focus on two specific research objectives. These are:

1. Provide empirical evidence of bottom-up agendas, and investigate how closely natural agendas, both top-down and bottom-up, conform to the models analyzed by Hauser (1986).

2. Empirically test the familiarity hypothesis: that bottom-up agendas are favored when familiarity is high, and top-down agendas are favored when familiarity is low.

Research on Agendas

The formal models of agendas include well known choice models such as elimination-by-aspects (EBA; Tversky 1972), the Hierarchical Elimination Model (HEM; Tversky and Sattath 1979), and the Generalized Elimination Model (GEM; Hauser 1986). Hauser (1986) analyzed the effects of changes in agendas for these decision models, demonstrating how the probability of a particular choice outcome can be enhanced or reduced, depending upon the type of agenda used. Kahn, Moore, and Glazer (1987) and Glazer, Kahn, and Moore (1991) examined the effect of imposing external constraints (partitions of brands not normally used by the consumer) on top-down agendas, and found different effects on choice than those predicted by HEM. Other work shows that cutoffs that form top-down agendas are influenced by preference structures and contextual factors (Klein and Bither 1987; Huber and Klein 1991).

Both formal mathematical models and empirical investigations have focused on top-down agendas, and their use by consumers to simplify choice has been well documented. Much less is known about the frequency or form of bottom-up agendas. Given the fact that a bottom-up strategy requires the evaluation of all alternatives, it may not be frequently used. Although Hauser (1986) analyzed simple agendas which were completely top-down or bottom-up, he does acknowledge that a mixed agenda is possible. For example, a consumer might quickly eliminate all Italian restaurants in a top-down fashion and then use a bottom-up agenda to compare favorites from the remaining categories. The use of adaptive strategies, in which the consumer changes the basic choice strategy as the decision progresses, has been well documented in process-tracing studies of consumer choice (Payne, Bettman, and Johnson 1988).

What determines which type of agenda buyers use: top-down, bottom-up, or mixed? Hauser (1986) proposed that one key factor is the consumer's ability to predict the utility of particular alternatives. If one can easily evaluate how much an alternative will be liked, then the series of bottom-up comparisons will be easier and tend to yield a high quality choice. If the more intensive evaluations of a bottom-up process still leave a large degree of uncertainty about which alternative is best, then a top-down agenda is advantageous. A top-down process is less effortful, because many alternatives are quickly eliminated, and it insures that a preferred value on the partitioning attribute is achieved.

Hauser (1986) proposed that a top-down agenda is more likely to be used when one is unfamiliar with the alternatives, when the cost of search is high, or when uncertainty is high. Thus, in our restaurant example, when consumers are new in town and know little about the restaurants from which they have to choose, they will tend to simplify the choice by first selecting a preferred cuisine category. On the other hand, individuals who are very familiar with the restaurants in the choice will find it relatively easy to select a favorite restaurant for each cuisine, and compare them for a final selection. This familiarity hypothesis has never been empirically tested. The following study was designed to identify both top-down and bottom-up agendas and test the familiarity hypothesis.

METHOD

Product Selection and Sample

Selecting a context for measuring and evaluating agendas was a nontrivial problem. Several criteria were important. First, respondents had to have a wide range of familiarity with available alternatives in the product category. For both theoretical and practical reasons, the product context had to provide commonly used categories that could be easily verbalized by respondents. Respondents had to be sufficiently interested in the product, and find the choice to be of sufficient complexity, that agendas were likely to be used. After extensive pretesting, a video rental was selected as the choice context. This product was appealing because video stores arrange films by categories (e.g. drama, comedy), so that respondents are used to seeing choice alternatives within a category structure.

Undergraduate students were recruited, using extra credit and cash incentives. Although thirty students responded to the initial questionnaire in the first stage of the study, the time consuming process of creating individualized stimuli for the second stage and tracing the choice processes limited the sample in the second stage to eight respondents.

Research Design

We conducted a within-subjects laboratory experiment, in which we created three levels of familiarity (high, moderate and low). In the High Familiarity condition, the choice alternatives were relatively recent movies that the respondents had already seen. Movies in the Moderate Familiarity condition were relatively recent movies that respondents had not yet seen. Movies in the Low Familiarity condition were somewhat obscure movies from twenty to thirty years ago, which pretesting showed were not recognized by respondents. At each of the three levels of familiarity, respondents made two choices in succession from different sets of movies. This repetition of the condition allowed for adaptation and greater respondent confidence regarding the level of familiarity. In all, respondents made six choices.

The selected choice context allowed us to make a strong manipulation of familiarity. Respondents could easily make preference judgments about movies they had seen (in the High Familiarity condition), which is essential in facilitating bottom-up agendas. In contrast, they should be highly uncertain about their evaluations of the obscure movies in the Low Familiarity condition; this uncertainty should lead to greater use of the movie category information, as in a top-down agenda.

Procedure

Data collection took place in two stages. In the initial classroom questionnaire session, respondents answered questions about 140 movies. Twenty movies were listed in each of seven categories - drama, action, romance, horror, comedy, family, and science fiction. These movies were recent films which had received national distribution and promotion, and which might be generally recognized even if they had not been seen. The rated movies were clustered into categories, with the category name provided at the beginning of each cluster for easy reference. Subjects indicated on the questionnaire whether or not they had seen each of the movies. Respondents also rated their preferences for the seven movie categories on a seven-point scale. We also measured other perceptions about the movies and movie categories, but these data are not relevant to this paper.

The responses from the questionnaire were used to create individualized choice sets for the three familiarity conditions. In the second stage of the study, each respondent came to an individual session in which they were presented with six choices, two for each of the three levels of familiarity (high, moderate and low). The order of presentation of choice sets at each of the three levels of familiarity was varied across respondents. However, the two choices from a particular level of familiarity were presented in succession to allow the respondent to adapt to that level of familiarity.

Although movies in seven categories had been rated, it was not always possible to find enough titles in each category to create the familiarity conditions (e.g. someone may never have seen any of the horror films). Thus, one respondent made choices from five movie categories; all other respondents chose from six categories.

The choice materials consisted of an information board for each movie category in the choice. The front of each board was labeled with the category name, and the reverse side provided places to insert five cards, each printed with a movie title. The boards for a particular choice were placed in random order on a long table, with only the category name showing. Just as consumers in an actual video store could only be physically present in one section of the store at a time, respondents in the study were told that they could "enter one section," or turn over one information board at a time. They had to return the board to its original face-down position before proceeding to another category. Categories could be returned to as many times as the respondent wanted. Also, just as a consumer might carry one or more video boxes with them as they move about a store, respondents were told that they had the option of removing and "carrying along" any of the cards for movies that they were considering. Cards could also be "returned to the shelf" if they were no longer being considered, by placing them on top of the relevant board.

As each of the six choices was made, a concurrent protocol was tape recorded. Respondents were instructed to think aloud while making a choice of one movie to rent. Subjects were also requested to consider only their own preferences in making their decision, rather than the preferences of some other party with whom they might normally watch a movie. Before each choice, respondents were told the kind of movies from which they would be choosing (i.e. "movies that you have indicated you have seen", "recent movies you have indicated you have not seen", and "older movies that you probably haven't heard of"). This was done because pretesting had revealed a natural preference for unseen, current movies, which would actually occur in only one condition (moderate familiarity). Unless the content of the choice was made clear to respondents, they tended to execute a search for such movies and failed to evaluate the movies actually available until their search had proved fruitless. The instructions allowed them to set appropriate expectations for the level of familiarity.

Measures

Concurrent protocols were necessary to trace the choice process in sufficient detail to allow accurate identification of agenda types. Early pre-testing revealed the difficulty of determining the existence, structure, and type of agenda that respondents were using in their choices. Various formats were pretested in an effort to elicit agendas that could be clearly identified. Format proved to have a very strong effect. When alternatives were not presorted by categories (such as movie type), the use of category structure was idiosyncratic and often not spontaneously verbalized. Allowing respondents to enter only one category at a time, and allowing them to remove films from the category as one would in a video store, greatly facilitated the identification of agendas.

To clarify even further what process was being used, a log was kept to supplement the protocols. The experimenter recorded the order in which categories were entered and which titles were removed for further consideration or returned to the shelf. This information was later added to the protocol transcriptions so that it would be clear what stimuli were in front of the respondent at each point in the choice protocol.

Manipulation checks and perceptions of the choice process were measured after each choice. These included perceived knowledge of and familiarity with the alternatives in the choice, confidence and satisfaction with the choice, its difficulty and how much the respondent wanted to see the video selected.

RESULTS

Manipulation Checks

Responses to the two items measuring familiarity with and knowledge of the choice alternatives had a correlation of .86, and so were averaged to test whether familiarity was successfully manipulated. A 3 (familiarity) X 2 (repeated choice) within-subjects ANOVA showed a significant main effect for familiarity (p<.001, eta=.94), indicating a successful manipulation. These averaged ratings had means of 6.6, 4.8, and 1.5, for the high, moderate, and low familiarity conditions. Unexpectedly, respondents also felt more familiar and knowledgeable about the movies in the second choices made within each level of manipulated familiarity (mean=4.6) than they did with the first choices (mean=4.0, p=.009, eta=.80). We believe that in the process of making the first choice within each familiarity level, respondents became more comfortable with the type of movie presented in that choice set. This may have created a sense of greater familiarity.

Coding Agendas

A total of forty-eight protocols were analyzed, six protocols for each of the eight respondents. The protocols were first coded for type of agenda, with classifications of: 1) bottom-up, 2) top-down, and 3) mixed agendas. As expected, respondents used many variations of the two agenda types identified by Hauser (1986). In the initial coding, a strategy was labeled top-down if it involved the elimination of alternatives based on their category, rather than on individual merits. A strategy was coded as bottom-up if it focused on the evaluation and disposition of specific alternatives. Combinations of these two processes in one choice were labeled mixed agendas. Coding by two independent judges showed inter-rater agreement in 90 percent of the cases.

As shown in Table 1, only two agendas revealed a totally top-down strategy, while 19 (forty percent) were coded as strictly bottom-up. The majority (fifty-six percent) represented mixed agendas which combined top-down eliminations and bottom-up processing.

TABLE 1

FREQUENCY OF AGENDA TYPES

Although many of the agendas were classified as bottom-up or top-down, only in the case of the two top-down agendas did the strategies exactly match the definitions offered by Hauser (1986). For every protocol coded as bottom-up, the definition had to be relaxed in some manner. For instance, some respondents selected more than one movie from a category, rather than picking only the best. Alternatively, some respondents chose favorites from only a subset of the categories, even though they examined all of them. Instead of comparing the best of each category, respondents sometimes used an ongoing elimination process in which possible choices were compared to the currently favored movie. In general, few agendas followed the elegant formulation of bottom-up strategies presented by Hauser (1986). It appears that agendas are flexible, and may be constructed as the choice evolves.

The most common pattern seen in mixed agendas was the use of a top-down elimination of one or more categories followed by some approximation of a bottom-up agenda. This finding agrees with research that describes phased strategies, with early elimination of categories of alternatives followed by more intensive evaluation of remaining alternatives (Bettman 1979).

Due to the limited sample size, few statistical tests of agenda types were conducted. However, the data were examined with respect to how often each of the two types of processing occurred at different levels of familiarity, as shown in Table 2. Note that a mixed agenda would be represented as having both types of processing. There appears to be no support for the familiarity hypothesis in these limited data. Bottom-up and top-down processing are used with similar frequency in all familiarity conditions.

TABLE 2

FREQUENCY OF PROCESSING TYPE BY LEVEL OF FAMILIARITY

Agenda Subtypes

In the course of coding protocols, it became apparent that there were certain identifiable variations on the top-down and bottom-up agendas. To gain a better understanding of these patterns, all agendas were coded with respect to subtypes. Inter-rater agreement was .94. Table 3 shows the frequency of these subtypes at the three levels of familiarity. Two forms of top-down strategies seemed distinct. A top-down agenda was coded "Favorite Category" when the respondent first made a choice about a particular category from which to select a movie, and then evaluated the specific movies in that category. In a top-down "Elimination" agenda, the respondent eliminated one or two less preferred categories, and then did further processing on remaining categories. The basic distinction between the two subtypes is the focus on preferred, as opposed to undesirable, categories.

TABLE 3

FREQUENCY OF TOP-DOWN AGENDA SUBTYPES BY FAMILIARITY LEVEL

There is no significant relationship between familiarity and subtype (X²=2.6, p=.27), the small numbers in the Favorite Category subtype are problematic. Still, it is interesting that different patterns emerge for the two subtypes across familiarity conditions, and this should be investigated in future research. According to the familiarity hypothesis, a greater number of top-down strategies should appear in low familiarity conditions. The elimination of disliked categories does not vary across familiarity levels.

This suggests that a general editing process occurs regardless of the difficulty of evaluating specific alternatives. This editing might be for simplification purposes alone. However, it is also plausible that certain categories of movies are easily identified as having sufficiently low utility that evaluation of specific alternatives is unnecessary (e.g. all horror films are disgusting). Although the specific alternatives in the high and moderate familiarity condition may be easy to judge, category information may make such judgments unnecessary. This type of opportunistic editing through a top-down agenda is unrelated to familiarity.

The data on the Favorite Category agenda follows the pattern suggested by the familiarity hypothesis, but the sample size prevents any reliance on that pattern at present. However, note that a Favorite Category agenda reduces the cost of evaluation more sharply than an Elimination agenda, because more categories are ignored. When familiarity is low, it is logical to avoid this evaluation cost because it is unlikely to pay off in a higher quality choice. Further research should explore whether decision makers seek a favorite category more often when alternatives are hard to judge.

The two subtypes of bottom-up agendas shown in Table 4 clearly had divergent relationships to familiarity (X²=12.6, p=.003).The first subgroup is labeled Bottom-up Comparison agendas. In these strategies, one or more movies were chosen from several categories. In some cases, these favorites were evaluated against each other to reach a final choice. This agenda type was most similar to Hauser's model of bottom-up agendas. In other bottom-up comparison agendas, the respondent conducted ongoing eliminations; if a new contender was found, it was compared to the current favorite. If one seemed sufficiently less desirable, it was returned to the shelf. In some cases, respondents did both ongoing eliminations and an evaluation of a final comparison set. Table 4 shows some support for the familiarity hypothesis for this subtype; there are fewer Bottom-up Comparison agendas when familiarity is low. This follows Hauser's logic that when it is difficult to predict the utility of specific alternatives, there will be less incentive to intensively evaluate them.

TABLE 4

FREQUENCY OF BOTTOM-UP AGENDA SUBTYPES BY FAMILIARITY LEVEL

The second type of bottom-up strategy was labeled Bottom-Up, Satisficing. As with the other bottom-up agendas, this processing involved specific alternatives rather than categories. However, in these strategies the respondent's processing did not involve comparisons of favored alternatives but rather a search for one that was minimally acceptable. In thirteen of the fifteen instances of this subtype, respondents searched all movie categories but found only one potentially acceptable movie, or none at all. In the latter case, a new strategy, such as returning to a preferred category, was likely to be tried.

The familiarity hypothesis asserts that bottom-up agendas are most likely when consumers can easily evaluate the utility of the alternatives. Yet in Table 4 we see that many respondents tried to evaluate specific alternatives even when they had little or no way of doing so, rather than resort to a top-down category-based strategy. Respondents who took this approach often expressed great frustration at not being able to find a movie that they recognized and at not being able to evaluate the relative desirability of the low familiarity movies. It appeared that to a large extent, categories were irrelevant, and that each movie was being compared to some minimum standard for recognition.

One handicap in understanding these bottom-up agendas is the difficulty of determining how important a role the categories played in them. Many respondents who used a bottom-up, satisficing agenda focused on titles without commenting on or obviously using category information. Given that Bottom-up, Satisficing processing tended to produce no more than one viable option (if that), it is difficult to say whether a category-based agenda was operative. The fact that two different patterns emerged for the frequency of these subtypes across familiarity conditions warrants further investigation.

DISCUSSION

What conclusions can be drawn from this exploratory study of agendas? We must first acknowledge the danger of overinterpreting the results from a small sample. With that caveat, we think that these results may be useful.

Bottom-up Agendas

Our first objective was to provide some empirical evidence for the existence of bottom-up agendas. While no strategy that was observed followed the exact form of the model provided by Hauser (1986), the essential elements of the bottom-up agenda strategy were present in the 33 protocols that contained some bottom-up, comparison processing. This provides preliminary evidence of bottom-up agendas in a situation which was created to make them most apparent.

It is also quite striking that few of the protocols matched the normative models presented by Hauser (1986). Respondents seemed quite comfortable with flexible strategies. When two movies in the same category seemed attractive, both were taken for further consideration. Some respondents did pairwise comparisons for some pairs of movies they had selected from the sections, while they delayed other (more difficult?) comparisons until all categories had been searched. The finding that agendas are flexible, and that consumers appear to adapt them to the particular choice set at hand, fits in well with rapidly accumulating evidence of adaptive choice strategies (Bettman, Johnson and Payne 1991).

The difficulty of constructing a laboratory situation in which bottom-up agendas are measurable and likely to occur leads inevitably to the question of when such agendas are likely to occur naturally. Hauser (1986) suggests that bottom-up agendas will be favored when familiarity is high, uncertainty about alternatives is low, and search cost is low. However, we discovered in pretesting that unless the category structure was made very salient, with alternatives presorted, respondents did not tend to choose and compare category favorites. Products with well learned, regularly used category structures (colas versus non-colas, fast-food versus regular restaurants) seem the most likely candidates for bottom-up processing. Top-down processing, on the other hand, requires only that you distinguish preferred from less preferred levels of any attribute, making clear categories less essential. For example, you might decide not to consider cars priced higher than $12,000, even though you do not normally group cars on that basis.

Another factor that seems likely to affect the incidence of bottom-up processing is the variance in utility, both between categories and between alternatives within categories. As discussed above, even when the movies were highly familiar, a large enough variance in the expected utility of movies within different categories may be sufficient reason for a top-down agenda. If you hate horror films, there is no sense in picking your favorite. Also, when there is little variance in the utility of alternatives within a category, picking your favorite becomes more difficult and may be avoided in some categories by top-down eliminations. In contrast, when it is relatively easy to specify your favorite, and categories are similar in utility (e.g. I like both dramas and comedies), finding the best of the best in each category makes sense.

Finally, using a bottom-up agenda may be more natural when favorite alternatives for the salient categories are stored in memory. For example, when a decision is made about which restaurant to patronize, someone who is highly familiar may automatically consider only their favorite Chinese restaurant, their favorite Italian restaurant, etc. The costly alternative comparisons within categories are avoided, and bottom-up processing becomes more feasible. These remembered favorites constitute a consideration set. Of course, it could be argued that processing such a set, apart from the original within-category selections, does not constitute an agenda.

Top-down Agendas

More is known about the natural occurrence of top-down agendas. These common editing strategies are used to simplify complex choices and have been identified in many process-tracing studies. In some respects, it was surprising not to find a greater use of these agendas; 40 percent of the strategies in this study had no top-down processing of categories. One possible reason is that respondents found the task novel and entertaining, and may have searched for specific alternatives out of curiosity as to what we had chosen as alternatives. Another explanation relates to a possible weakness of our product choice. Although the categories used to sort movies were very realistic in terms of respondents' experiences in video stores, the categories were not always mutually exclusive. For example, a film can be both a drama and a romance, both a comedy and a family picture, and so on. This means that the categories may not have been very diagnostic about the desirability of specific alternatives they contained. Category information may therefore have been slighted as a basis for decision making, and more information about specific movies may have been sought.

A final explanation for the lack of top-down processing is that a respondent's natural category structure for videos may have diverged from the imposed categories. For example, if movie type is unimportant to a respondent, there is little incentive to use the category to narrow the choice set. Instead, he or she may have used their knowledge of specific alternatives to impose other top-down constraints on the choice set. These constraints may have appeared to be bottom-up processing in our analysis. However, we found no evidence in the protocols of the systematic use of other criteria.

The Familiarity Hypothesis

The second objective of this study was to provide an empirical test of the familiarity hypothesis: that bottom-up agendas are used more frequently as familiarity increases, and that top-down agendas are used more often as familiarity decreases. The results seem to depend on the manner in which agendas are defined and protocols are coded. If all bottom-up and top-down strategies are considered together, familiarity appears to have no relationship to agenda type. However, bottom-up comparison agendas did conform to the familiarity hypothesis. Further descriptive research on agendas will have to contend with definitional and classification problems.

Other Issues

Pretesting demonstrated the importance of format in decision making. During pretesting, respondents increased their report of category use (constraints) when alternatives were presented in categorized format. Additionally, respondents used the exact category names provided for them. When choices were made with no categories or alternatives suggested, the apparent use of agendas was very limited. This finding indicates the potential of marketing efforts to set agendas for customers. Explicitly guiding consumers in devising constraints on alternatives, either through advertising or a specially tailored sales approach, seems more likely to be effective in influencing choice than simply providing unstructured information. Providing the appropriate format for information could have great impact on the shaping of decisions.

Additional work on agendas is warranted. Coding and definitional issues need to be resolved, and the methodology developed for tracing the agenda process needs to be tested in a variety of contexts. There are challenging issues related to both naturally occurring and externally imposed agendas. There appears to be no simple relationship between agenda type and familiarity, perhaps because there are few simple agendas. Greater understanding of this topic will require further normative and descriptive analyses.

REFERENCES

Bettman, James R. (1979), An Information Processing Theory of Choice, Reading, MA: Addison-Wesley.

Bettman, James R., Eric J. Johnson, and John W. Payne (1991) "Consumer Decision Making", in Handbook of Consumer Behavior, eds. Thomas S. Robertson and Harold H. Kassarjian, Englewood Cliffs: Prentice-Hall, 50-84.

Glazer, Rashi, Barbara E. Kahn, and William L. Moore (1991) "The Influence of External Constraints on Brand Choice: The Lone Alternative Effect," Journal of Consumer Research, 18 (June), 119-127.

Hauser, John R. (1986), "Agendas and Consumer Choice," Journal of Marketing Research, 23 (August), 199-212.

Huber, Joel and Noreen Klein (1991), "Adapting Cutoffs to the Choice Environment: The Effects of Attribute Correlation and Reliability," Journal of Consumer Research, 18 (December), 346-357.

Kahn, Barbara, William L. Moore, and Rashi Glazer (1987), "Experiments in Constrained Choice," Journal of Consumer Research, 14 (June), 96-113.

Klein, Noreen M. and Stewart W. Bither (1987), "An Investigation of Utility-Directed Cutoff Selection," Journal of Consumer Research, 14 (September), 240-256.

Payne, John W., James R. Bettman, and Eric J. Johnson (1988) "Adaptive Strategy Selection in Decision Making," Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 534-552.

Tversky, Amos (1972) "Elimination-by-Aspects: A Theory of Choice," Psychological Review, 79 (July), 281-299.

Tversky, Amos and Shmuel Sattath (1979), "Preference Trees," Psychological Review, 84 (July), 327-352.

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