Schema-Based Planning of Events in Consumer Contexts

Lawrence W. Barsalou, Emory University
J. Wesley Hutchinson, University of Florida
ABSTRACT - Many types of consumer behavior, including information search, decision making, and product usage, are preceded by some type of planning. This paper examines the role of memory-based schemata in such planning. In particular, the idea is advanced that only the most general aspects of schematic knowledge are well-established in memory and are used on most occasions by most people. Other types of information, such as specific schema variables and their instantiations, are much more variable and context dependent. The results of two exploratory studies are briefly described.
[ to cite ]:
Lawrence W. Barsalou and J. Wesley Hutchinson (1987) ,"Schema-Based Planning of Events in Consumer Contexts", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 114-118.

Advances in Consumer Research Volume 14, 1987      Pages 114-118

SCHEMA-BASED PLANNING OF EVENTS IN CONSUMER CONTEXTS

Lawrence W. Barsalou, Emory University

J. Wesley Hutchinson, University of Florida

ABSTRACT -

Many types of consumer behavior, including information search, decision making, and product usage, are preceded by some type of planning. This paper examines the role of memory-based schemata in such planning. In particular, the idea is advanced that only the most general aspects of schematic knowledge are well-established in memory and are used on most occasions by most people. Other types of information, such as specific schema variables and their instantiations, are much more variable and context dependent. The results of two exploratory studies are briefly described.

INTRODUCTION

We have all failed to plan sufficiently at one time or another. For instance, there is the classic tale of the vacationer in an exotic country who wiled away the days buying souvenirs and bargains. He imagined how impressed his friends at home would be. Only on the day of his departure did he discover that he had no way to take them home because he came with overpacked luggage. A hasty final purchase completed his trip. Although few purchases are so directly caused by memory failure during planning, it seems likely that many purchase decisions are framed by the plans that they fulfill.

In general, consumer research has focused on behaviors that are rather directly related to the purchase of a single item from a set of competing alternatives. Thus, there are large literatures on decision making, information search, reactions to advertising, etc. Within this framework, several authors have noted that memory can exert strong effects on choice by limiting either the number of alternatives that are considered (Alba and Chattopadhyay 1985; Baker, Hutchinson, Moore, and Nedungadi 1986) or the types of information that are retrieved during decision making (Bettman 1979; Biehal and Chakravarti 1983, 1986). Moreover, the effect of memory on choice may depend on the type of decision rule that is employed (Lichtenstein and Srull 1985).

This paper examines the effects of memory on choice from a somewhat broader perspective. First, we focus on planning, rather than decision making, per se. That is, the primary concern is with cognitive activities that precede and guide information search and decision making. Second, the problem domain is defined at a more abstract or diffuse level. In contrast with the traditional problem of brand choice within a particular product class, we address the problem of how consumers draw upon memory to structure several interrelated activities.

A central issue in our discussion is the role of schematic knowledge during planning. The exploratory evidence that is reported suggests two general phenomena. First, people do rely heavily on schema-like structures that are well-established in memory. Second, such schematic structures are fairly general and there is considerable variation in the ways in which specific information is incorporated into planning. Moreover, this variation is systematically related to identifiable contextual factors.

Protocol Studies of Event Planning

Barsalou, Usher, and Sewell (1985) performed two protocol studies in which they asked people to plan various kinds of events. Study 1 observed superficial planning of 24 different events (e.g., buying a dog, travelling to visit a friend, giving a surprise birthday party, getting a broken car fixed). Study 2 observed in-depth planning of a single event, namely, taking a vacation. After transcribing subjects' planning episodes, Barsalou et al. developed and applied various coding schemes to the protocols. Reliability in analyses for which it has been computed so far has been around .90. The results from both protocol studies will be presented together to provide a more coherent account of how subjects planned events. A forthcoming paper will present these findings in greater detail and in a more quantified manner.

The Nature of Schemas

When subjects were asked to plan an event, they initially retrieved schematic knowledge containing variables that needed to be instantiated for the event to occur. These variables were "schematic" in the sense that they provided a general description of a kind of event, yet accepted more specific information as instantiations. The view of schemas presented here is closely related to the view proposed by Rumelhart and Ortony (1977).

When subjects planned a vacation in Study 2, for example, they initially retrieved variables such as expenses, temporal parameters, location, accommodations, transportation, activities, COD anions, things to take, and so forth. In many cases, a variable was actually a see of more specific, closely related variables, each of which needed to be instantiated. For example, transportation often included major transportation (e.g., airlines), minor transportation (e.g., airport shuttle), and transportation at the vacation location (e.g., rented car). Similarly expenses often included total cost and source of money. Table 1 provides proportions and average output positions for all the variables instantiated by subjects in Study 2 when planning a vacation. Superordinate variables such as transportation, were counted as instantiated any time one of its subordinates was mentioned (e.g., at vacation location). For this reason, average output position is not noted for superordinates.

TABLE 1

VACATION SCHEMA VARIABLES PRODUCED IN STUDY 2 (N-16)

One might think that a schema is a static package of variables retrieved as a rigid unit during planning to be instantiated. However, there were several reasons why this does not appear true. First, subjects shoved substantial variability in the variables they retrieved and instantiated when planning the same event. Because most proportions in Table 1 were less than 1.0 (p<.05 for all values less than .8), subjects in Study 2 clearly did not retrieve and instantiate a static set of variables when planning a vacation. Similar variability was observed when people planned a wide variety of events in Study 1.

Second, after a given subject had retrieved and instantiated a few variables, he or she often experienced substantial difficulty in retrieving additional variables that needed to be instantiated. Static intact sets of variables did not immediately become active when subjects planned events. Instead much time and effort was often necessary to retrieve all the relevant variables. The later discussion of scanning provides explicit evidence for this claim.

Third, some variables were only retrieved when activated by a particular instantiation of another variable, what we refer to as instantiation triggered variables. For example, instantiating the location variable with a foreign location often caused subjects to consider the health hazards variable. This makes sense because it is often necessary to consider health hazards in order to obtain the necessary immunizations before travelling aboard. In contrast, when subjects selected domestic values for location, they never considered health hazards. This, along with other examples from our data, indicate that certain variables may only be relevant in particular planning contexts, where "context" refers to how a schema's variables have been instantiated so far at any given point in the planning process.

Consequently, schemas do not appear to be retrieved as static sets of variables. Instead, it may be more appropriate to assume that people have loosely-organized schematic knowledge about events, with this knowledge containing variables that vary in accessibility and that are to some extent context-dependent. This does not necessarily imply that there are no stable schemas in long term memory. Rather we propose that long term memory contains a large amount of relatively stable knowledge for a kind of event, but that only a contextually-relevant subset is used in a particular situation and that the same subset is rarely if ever used in two situations. This view is similar to recent characterizations of consumer choice heuristics (Bettman and Park 1980; Biehal and Chakravarti 1986). Barsalou (in press-a) provides a more detailed justification and discussion of this view.

FIGURE 1

PARTIAL EXAMPLE OF A PLANNING SCHEMA

The use of schemas during planning has been studied by a number of other investigators, including Luchins (1942), Byrne (1977), and Wilensky (1983).

Instantiation of Schemas

Once subjects retrieved a variable, they generally attempted to instantiate it. The left side of Figure 1 shows a partial example of how subjects instantiated variables when planning a vacation in Study 2. Subjects' protocols were generally much more extensive and complicated than this example.

Subjects instantiated variables in a number of ways. For some variables, subjects generated a candidate set of instantiations from which they eventually selected one as part of the final plan. When subjects planned a vacation, they often instantiated location in this manner. For other variables, all of the instantiations generated were included as part of the final plan. When subjects planned a vacation, they often instantiated things to take in this manner. Sometimes subjects failed to instantiate a variable after considering it briefly but planned preparations (discussed shortly) that would enable them to instantiate it later. Other times when subjects didn't generate instantiations for a variable, they described ideal or probable characteristics of instantiations (cf. Barsalou, 1985).

The measurement level of variables varied widely. Across the 24 events in Study 1, most variables were instantiated by nominal values (57%). Variables such as companions and things to take, for example, were instantiated by unordered sets of individuals. Variables were also frequently instantiated by values from ratio scales (20%). Many of these variables had to do with the financial and temporal parameters associated with an event (for examples see temporal and financial parameters (n Table 1). Variables were also frequently instantiated by values from binary scales (21%).

Although not shown in Figure 1, the instantiation of a variable was often represented by schematic knowledge itself. In Figure 1, for example, camera could have been represented more completely with the variables that comprise its schematic knowledge (e.g., brand, year, model, cost, size, capabilities). When planning, subjects often generated such variables from the schematic knowledge for an instantiation and then attempted to instantiate these lover-level variables. For example, a subject could have generated the variable of size from schematic knowledge for camera and then instantiated it with pocketsize. Consequently, the plans that subjects constructed can be viewed as a large collection of embedded schemas. Constructing a complete plan requires selecting instantiations for a large number of variables, in a large number of schemas, at many different levels of embedding.

Tagging Instantiations

As subjects retrieved instantiations, they often tagged them in a various ways. One kind of tagging involved an instantiation decision about whether or not an instantiation would be a part of the final plan. In Figure 1, for example, Europe might be explicitly chosen over California and Florida . A second kind of tagging involved stating a necessary condition that had to be met before an instantiation would be incorporated into the final plan. In Figure 1, for example, skis would only be taken if the vacation occurred during Winter Break. Subjects also tagged instantiations in other ways, including evaluations of instantiations and reasons for choosing instantiations. All of these tags appear to become stored in long term memory as part of a person's memory for Planning an event.

Preparations

Subjects did not spend all of their time instantiating variables. They also spent a lot of time planning various preparations that would have to be done between the time at which they were planning and the time at which the planned event would occur. Subjects often spent more time planning preparations than they did instantiating variables.

In Study 2, four preparations accounted for 84% of all the preparations mentioned. They were obtaining something (34%), seeking information (25%), deciding something (15%), and verifying something (10%). Subjects often realized that they didn't have enough information to instantiate a variable acceptably at the time of planning. As a result, they planned to seek information about instantiations at a later time and then decide about final instantiations when more informed. For example, subjects often didn't have enough information at the time of planning to decide on particular accommodations. Consequently, they planned to seek information about accommodations and to make a final decision once better informed.

Subjects also often realized after instantiating a variable that the instantiation chosen either entailed having to obtain something or verify that some state was true, if the plan was to succeed with that instantiation. For example, one planner taking a vacation to the beach didn't have a swim suit and, therefore, planned to obtain one before leaving. Another planner needed to take a passport for foreign travel and, therefore, planned to verify that it was still current.

As can be seen from Figure 1, the preparations in subjects' protocols were schemas themselves. Each preparation schema generally had a characteristic set of variables; these variables were instantiated in the various ways discussed earlier for main schema variables; and instantiations of these variables were tagged with decisions and so forth just as for the instantiations of main schemas variables. In general, these four preparations appeared to be well-learned and highly-articulated cognitive skills that most people possess and that are probably used across a wide range of planning situations. Obtaining things, seeking information, making decisions, and verifying states each appears to be an activity that most people perform regularly, and each appears to be an interesting and important object of study in its own right. Interestingly, only seeking information and deciding something have received much attention in either the consumer or cognitive literatures. Other preparations include informing (3%), constructing (3%), giving (3%), packing (3%), discussing (2%), recommending (1%), and completing (1%).

Rules of Constraint Between Variables

Subjects' schematic knowledge of events also contained another important kind of information, namely, rules of constraint between variables. These rules specified how the instantiations of one variable were related to the instantiations of another variable. Subjects frequently used these rules to select instantiations and to insure that the instantiations they selected were compatible with the instantiations of other variables. In Study 2, we observed subjects using 51 different rules, with many being used frequently. For example, subjects often stated that the instantiation of available money determined the instantiation of location, the underlying principles being that (1) more money enables travelling to more distant locations, and (2) more money enables staying at more expensive locations. Similarly, the instantiations of activities often constrained the instantiation of location, the underlying principle being that certain activities require locations with certain characteristic (e.g., snow skiing requires mountainous locations). Further examples of these rules are shown in Table 2. Most variables appeared to be constrained by rules projecting from many other variables, and a given variable often projected constraint to many variables.

TABLE 2

EXAMPLES OF CONSTRAINT RULES MENTIONED BY SUBJECTS

Subjects also frequently mentioned high-level goals while planning. When planning a vacation, for example, subjects often attempted to maximize safety, to maximize enjoyment, to maximize comfort, to minimize cost, and so forth. These goals often played an important role in how people instantiated variables and appear to be the source of the ideals discussed in Barsalou (1985). It would not be surprising if the high-level goals for an event schema are integrated together with rules of constraint between variables to provide an intuitive theory for that kind of event (Murphy and Medin, 1985).

Scanning Schema Variables

As discussed earlier, subjects often hat difficulty retrieving variables that needed to be instantiated. At these points, subjects often explicitly scanned the schema variables they had already instantiated. Here are several examples:

um... we'd be all set for the plane, for places to stay, travel while we're there, um... eating while we're there, am... what else... where to go while we're there... um... (whispered) God I wonder if that's all there is to it

um... okay, so let's see... um have I planned everything? Where, when, who, money, what I'm gonna do...

well after we have figured out the location, the time, and if we can afford it, then we'll plan what we're going to do

um... let's see, so once I had a time and an agreed-upon group, then we'd work on the specific location.

Although not all subjects scanned explicitly in this manner, it would not be surprising if all were scanning implicitly.

Scanning may perform several functions: First, subjects may scan to insure that they have not forgotten to instantiate any necessary variables. Second, subjects may use scanning as a means to retrieve variables they are having trouble retrieving. Because variables are related to one another by rules of constraint, scanning instantiated variables may activate uninstantiated variables via these relations. Third, subjects may use scanning as a means of propagating constraint. By reminding themselves of how they have instantiated previous variables, subjects may insure that instantiations of future variables will be compatible with already existing instantiations.

Whatever purposes it may serve, scanning clearly demonstrates that people have variables in mind when planning events. This offers support for the existence of schematic event knowledge in long term memory and for its use in planning.

Constraint-Driven Focusing of Schema Variables

When instantiating a variable, subjects often constructed increasingly specific categories of instantiations for the variable, a phenomenon we will refer to as focusing. For example, when one subject attempted to instantiate location, he first focused the set of possible instantiations to places that can be driven to and later focused the first focusing to places that can be driven to that have fishing. We observed an average of 3.44 explicit focusings per subject, and we suspect that a great deal of additional focusing occur ed implicitly.

Interestingly, many of these focusings appeared to be constraint-driven in that subjects used the instantiations of other variables to guide focusing. In the above example, the first focusing (places that can be driven to) appeared to have been driven by the subject having decided to instantiate transportation with drive my car. Similarly, the second focusing (places that can be driven to and that have fishing) appeared to be driven by the subject having decided subsequently to instantiate activity with fishing.

In other words, what subjects often appeared to do when focusing a variable was to conjoin the variable with one or more instantiations of other variables to construct increasingly specific categories of possible instantiations. This is quite similar to the conjunctive and Elimination-By-Aspects strategies that have been discussed in the context of decision making (Payne 1976; Svenson 1979; Tversky 1972; Wright 1975). The present analysis highlights the interactive and context-dependent aspects of such processes.

The Origin of Goal-Derived and Ad Hoc Categories in Planning

These results so far suggest an account of how categories originate during planning. According to this account, categories can originate in two ways: First, some categories originate because they provide sets of instantiations for schema variables. For example, places to go on a vacation provides instantiations for the location variable in the vacation schema. Second, other categories originate during constraint-driven focusing when subjects contain a variable with one or more instantiations of other variables. For example, places to go on a vacation that can be driven to and that have fishing results from conjoining places to go on a vacation with instantiations of the transportation and activity variables. We refer to both kinds of categories as goal-derived categories (Barsalou 1983. 1985).

To the extent a goal-derived category receives frequent processing, it should become well-established in memory. This appears to occur for many of the categories that instantiate schema variables. For example, people seem to have stable knowledge of activities to do on a vacation. This also appears to occur for frequent constraint-driven focusings of schema variables. For example, people seem to have stable knowledge of activities to do on a vacation at the beach.

However, many constraint-driven focusings produce novel categories, what we refer to as ad hoc categories (Barsalou 1983, 1985). These categories become constructed when other variables are instantiated in a novel way such that a novel focusing of a variable occurs (cf. Bettman and Zins 1977). For example, activities to do on a vacation with one's grandparents in Tokyo might be an ad hoc category for someone who has never experienced this combination of constraints before. Of course, the categories that are ad hoc should vary from person to person, depending on their experience.

If this account is correct, it becomes possible to specify the space of categories that can be constructed during planning. Once the variables of a schema are known, and once the instantiations of these variables are known, then the categories that can exist are (1) those that instantiate the schema variables, and (2) those that instantiate the possible focusings of those variables, as constrained by the existing instantiations of other variables. In addition, if rules of constraint between variables are taken into account, then the space of potential categories is reduced substantially. Because these rules specify that many combinations of instantiations are impossible, they rule out all constraint-driven focusings involving those combinations.

By collecting normative data about variables and their instantiations, it becomes possible to map out the space of categories that may be used when achieving a particular goal. For many natural domains, it may be impossible to identify all the possible variables and all their possible instantiations. However, in most cases it appears possible to identify a very large proportion of the relevant variables and instantiations such that the potential space of categories can be mapped out to a large extent. In certain settings that have relatively static and well-defined domains, it should be possible to map out the space of categories completely. In addition, as new variables and instantiations are encountered, the new categories made possible can be easily specified.

Conclusion

The process of planning an event can be viewed as conceptualization. To plan an event, people initially retrieve schematic knowledge about that kind of event in the form of uninstantiated variables. These variables provide a vague conceptualization of the kind of event being planned. Schematic knowledge not only provides the initial conceptualization of an event, it also provides materials - in the form of well-established goal-derived categories--that can be used to construct more specific conceptualizations. By constructing ad hoc categories through constraint-driven focusing, and by eventually selecting specific instantiations, people can increasingly refine their conceptualization of an event.

Besides containing materials from which plans can be constructed, schematic-knowledge also provides guidelines for how to select and combine these materials. The high-level goals and rules of constraint associated with a schema provide a causal model of the event that ides and constrains conceptualization. This model specifies what instantiations are acceptable in particular contexts and how constraints are to be combined once variables in a schema have been instantiated. In addition to schematic knowledge, people also use schema-independent tools--in the form of preparation schemas- to help conceptualize and eventually execute an event. These cognitive skills often serve to increase a person's knowledge of a kind of event and to provide an interface with the physical world such that the plan can be implemented.

Much consumer research has been directed at the process of decision making and the information that directly influences purchase decisions. Most of the research paradigms that have been used impose a specific structure on the decision. Research on the nature of consumer planning should complement this effort by providing an understanding of how consumers structure problems for themselves.

REFERENCES

Alba, Joseph W. and Amitava Chattopadhyay (1985), "The Effects of Context and Part-Category Cues on the Recall of Competing Brands," Journal of Marketing Research, 22 (August), 340-349.

Baker, William, J. Wesley Hutchinson, Danny Moore, and Prakash Nedungadi (1986), "Brand Familiarity and Advertising: Effects on the Evoked Set and Brand Preference," in Advances in Consumer Research, Vol. 13, ed. Richard J. Lutz, Provo, UT: Association for Consumer Research, 637-642.

Barsalou, Lawrence W. (1983), "Ad Hoc Categories," Memory and Cognition, 11, 211-227.

Barsalou, Lawrence W. (1985), "Ideals, Central Tendency, and Frequency of Instantiation," Journal of Experimental Psychology: Learning, Memory, and Cognition, 11, 629-654.

Barsalou, Lawrence W. (in press), "The Instability of Graded Structure: Implications for the Nature of Concepts. In U. Neisser (Ed.)," Concepts and Conceptual Development: Ecological and Intellectual Factors in Categorization. Cambridge: Cambridge University Press.

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

Bettman, James R. and C. Whan Park (1980), "Effects of Prior Knowledge and Experience and Phase of the Choice Process on Consumer Decision Processes: A Protocol Analysis," Journal of Consumer Research, 7 (December), 234-248.

Bettman, James R. and Michael Zins (1977), "Constructive Processes in Consumer Choice," Journal of Consumer Research, 4 (September), 75-85.

Biehal, Gabriel and Dipankar Chakravarti (1983), "Information Accessibility as a Moderator of Consumer Choice," Journal of Consumer Research, 10 (June), 1-14.

Byrne, R. (1977), "Planning Meals: Problem-Solving on Real Data-Base," Cognition, 5, 287-332.

Lichtenstein, Meryl and Thomas R. Srull (1985), "Conceptual and Methodological Issues in Examining the Relationship Between Consumer Memory and Judgment," in Psychological Processes and Advertising Effects: Theory, Research, Application, eds. Linda F. Alwitt and Andrew A. Mitchell, Hillsdales, NJ: Erlbaum, 113-128.

Luchins, A. (1942), "Mechanization of Problem Solving," Psychological Monographs, 54, No. 248.

Murphy, Gregory L. and Douglas L. Medin (1985), "The Role of Theories in Conceptual Coherence," Psychological Review, 92, 289-316.

Payne, John W. (1976), "Task Complexity and Contingent Processing in Decision Making: An Information Search and Protocol Analysis," Organizational Behavior and Human Performance, 16 (August), 366-387.

Rumelhart, David E. and Anthony Ortony (1977), "The Representative of Knowledge in Memory. In R. C. Anderson, R. J. Spiro, and W. E. Montague (Eds.)," Schooling and the Acquisition of Knowledge. Hillsdale, NJ: Erlbaum.

Svenson, Ola (1979), "Process Descriptions of Decision Making," Organizational Behavior and Human Performance, 23 (February), 86-112.

Tversky, Amos (1972), "Elimination by Aspects: A Theory of Choice," Psychological Review, 79 (July), 31-48.

Wilensky, Robert (1983), Planning and Understanding: A Computational Approach to Human Reasoning, Reading, MA: Addison-Wesley.

Wright, Peter L. (1975), "Consumer Choice Strategies: Simplifying versus Optimizing," Journal of Marketing Research, 11(February), 60-67.

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