Ethno: a Methodology For Studying Process Information

ABSTRACT - Using event structure modeling, we examine consumer information processing as originally developed by Bettman. Using the data reported for consumer C1 in Bettman's 1970 paper, we show how applying a formal logic to that data reveal alternative decision sequences (pathways) in that model. We found that the consumer decision structure was more sequentially structured than was shown in the original model. In addition, we introduce facilitating decisions as part of information processing structure and show how this modification refines the understanding of consumer choice as modeled by information processing.


James H. Barnes (1993) ,"Ethno: a Methodology For Studying Process Information", in NA - Advances in Consumer Research Volume 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT : Association for Consumer Research, Pages: 63-69.

Advances in Consumer Research Volume 20, 1993      Pages 63-69


James H. Barnes, University of Mississippi


Using event structure modeling, we examine consumer information processing as originally developed by Bettman. Using the data reported for consumer C1 in Bettman's 1970 paper, we show how applying a formal logic to that data reveal alternative decision sequences (pathways) in that model. We found that the consumer decision structure was more sequentially structured than was shown in the original model. In addition, we introduce facilitating decisions as part of information processing structure and show how this modification refines the understanding of consumer choice as modeled by information processing.


Consumer research has begun to focus increasingly on subjective realities. Led by qualitative researchers who make a substantial research investment to cultivate informants and attain a deep enough understanding of subjects' thinking to report the realities that the subjects cannot or will not articulate themselves, the field is approaching a greater understanding of why consumers buy. This increased emphasis on field observation and data collection may however suffer from several drawbacks. The observer may fail to correctly record the event, overlooking some critical step in the reported sequence and second, the informant may likewise fail to recall all the details of his or her actions. Some of these drawbacks have been circumvented by the introduction of new methods of qualitative analysis that offer systematic, uniform, computer-assisted procedures for data analysis. These computer-assisted procedures ask more questions than are usually put to data and demand extraordinary precision and meticulousness in descriptions of events (Corsaro and Heise 1990).

Our proposed method of analysis requires two kinds of data. First, we need experts' definitions of events and logical relations. Second, we need records of actual event sequences to aid elicitation and to define correct orderings of relevant events[though not necessarily the only permissible orderings] (Corsaro and Heise 1990, p. 5).

In this paper, we will discuss one such methodology and show how it can be used to organize consumer type data and thus result in improved collection and understanding of the process being studied.


The method employed in the current study is derived from the theory of rational action developed in cognitive science and called the theory of production systems. The concept of these systems were introduced by Newell and Simon (1972) and have been continued and applied by psychologists (e.g. Anderson 1983), sociologists (Axten and Skvoretz 1980; Fararo and Skvoretz 1984) and computer scientists (e.g. Waterman and Hayes-Roth 1978). A PC based computer program using this theory for studying events, their logical connection, and for discovering the rules that govern specific actions has been developed by Heise (1988). The program based on the theory of production systems assumes that people's conceptions of the world are logically structured.

This approach to modeling event structures derives from action being governed by if-then rules: if a certain configuration of conditions arises then a certain action occurs. That is, an event cannot occur until all of its prerequisites have occurred. For example, a consumer wishing to purchase product X cannot do so until he or she has arrived at the particular retail outlet that sells the product. Such prerequisites as finding car keys, driving to the mall, parking, and walking to a certain store must all occur before purchase. In addition, this process may cause other consumption events to occur (consume gasoline, for example). Once in the shopping environment, it is not unusual to find consumers purchasing one product because they need it for an impending occasion, another because it was on sale, and a third, purchased impulsively simply because they liked the product. One must thus be able to model these different decisions.

People are, however, not very good at computing long chains of implications (Corsaro and Heise 1990). Thus, if one is questioning a consumer concerning their purchase decision process, the initial facilitating decision to drive to the mall may very well have been forgotten as a part of the process. In addition, the researcher may overlook events and fail to record them though the sequence may be video recorded or otherwise documented. That is, subtle aspects of decisions may be simply overlooked.

The idea that events are primed by if-then prerequisites is a conjunctive process. The required configuration of states for an event may be obtained in some cases through alternative prerequisite events in which case, the event is disjunctively related to its prerequisites. This results in an either-or type of condition as the priming event in a sequence. For example, in our previous example, the consumer may not go to the mall but may instead order product X through a catalog or other direct merchant in which case, other related consumptions, different from the prior example, occur.

A second assumption in production system theory is that occurrence of an event depletes the conditions that primed it. This means that the event structure does not become stuck in a loop with certain states producing the same event repeatedly (Corsaro and Heise 1990). That is, for our consumer to purchase product X again, he or she would have to come again to the retail store where the purchase was made. Events can however be repeatable if the logic of the process suggests that such is true. Our consumer could make multiple purchases of product X while in the shopping environment (one may argue that this would entail a different decision process and should be modeled differently).

A third assumption of production systems is that with exceptions, an event is not repeated until all of its consequences are used up. That is, more and more of something is not produced without being used. For our consumer, he or she would not generally continue to purchase product X without consuming it.


An existing program, ETHNO (Heise 1988, Heise and Lewis 1989), is used to illustrate the computerized analysis of qualitative data using the principles of event structure analysis discussed above.

The program employs a modifiable verbal framework to elicit verbally-defined elements in a desired domain and the logical relations among elements. During elicitation the program employs past answers about implications to minimize further questions (with the presumption of local logic). The program analyzes an event series for consistency with an obtained structure and then again to compute priorities. The program also allows completed models of events structures to be used for simulation (Heise 1989, p. 149).

The data used for this analysis were reported in Bettman (1970, 1979) and formed the bases for an information processing theory of consumer decision making that has gained widespread acceptance in the consumer behavior literature. Since information processing is viewed as a sequential series of activities, it represents an ideal process to subject to the rigors of a formal analytical approach.

Our analysis proceeds in the phases discussed above. Phase one consisting of entering the events and determining their logical relationship. In the present example, this logical relationship will initially be that the ETHNO model duplicates the consumer models developed by Bettman. Phase two consists of analyzing the series of decisions to make it consistent with the implicational structure or other modeling assumptions. Phase three will consist of reanalyzing the series to assess event priorities. The fourth and final phase will consist of demonstrating the revised model for simulating events.

Phase One: Building the Initial Model

The data concerned Bettman's Consumer C1 and represents a mother of five with a husband who has recently finished medical school and currently teaches. "Her decisions were for the most part based on price, but she let the children have some of their favorites" (Bettman 1970, p. 373). The model of decision making for this consumer is shown as Figure 1 in Bettman's 1970 paper and is also discussed in his 1979 book. For purposes of data entry, we assume that the sequence of decisions reported for consumer C1 and numbered X1 through X44 are in fact recorded in time sequence order (Later in our analysis, we find this assumption to be incorrect). To develop an event structure, one enters events in the computer program in time sequenced order. This allows the software to develop the hierarchical structure required for event analysis. Thus, the initial computer entry after defining the problem to ETHNO is event numbered X1. This is followed by event two, etc. Following each entry, the program asks, Is a specific prior event (or something similar) essential for the current event? Specifically, the program is asking the person entering data to define which prior events are prerequisites for the event being entered. Thus, the program itself does not dictate the structure but relies on the researcher to develop it based either on recorded field notes, the researcher's understanding of the process, or in the present case, another researcher attempting to examine or replicate prior findings. By asking questions, the program is forcing the researcher to examine his or her own logic and understanding of the process being modeled.

For each question and for each entered event, the answer "yes" or "no" was given such that the linkages in ETHNO represented those reported by Bettman. The ETHNO graphical model thus developed is shown as Figure 1.

The structure shown in Figure 1 is the same as reported in Bettman (1970). The key to the figure shows the definitions also reported by Bettman. ETHNO however has added a group of special relations. The special relations are shown as a "flip-flop" which is ETHNO's way of reporting commutative relations.

The relation between some events is peculiar in that each is a prerequisite for the other, as in entering and leaving a room: after initially entering, one has to leave to enter again, and one has to enter to leave again. Such events are commutativeCone or the other event happens next depending on which happened last. The events alternate or flip-flop (Heise and Lewis 1988, p. 108)

Once data are entered, we are ready to proceed to phase two of the analysis.

Phase Two: Analyzing the Model

Phase two consists of the very important step of allowing ETHNO to analyze the developed structure considering the rules for event analysis discussed earlier. In the present model, the program began to examine event one (Is this meat or produce?), proceeding to the next event determining its various consequences for other events in the system. This analysis had not proceeded very far when the program stopped and began to ask questions and suggesting ways in which a discovered structural problem in the model could be dealt with. The program suggestions for solving the problem are based on the assumptions of event structures and their relationship one to another as discussed earlier. The first problem encountered by ETHNO concerned the sequencing of X44, X41, X42, and X43 before X4. If the sequence of events had occurred as listed by Bettman in his KEY TO FIGURE 1, consumer C1 would have practically completed her shopping before deciding if the meat or product (X42) was for a specific use. Since it appears that this decision was necessary for the meat or produce decision, our consumer could not complete her purchase choice until the prerequisite events were fulfilled and hence, the model sequence as originally formulated was unacceptable to the structure. In the case of these events, the model suggested that their sequence may have been incorrectly recorded and provided a means for a structural change to place these decisions into their proper sequence. Having no reason to suspect that the structure was indeed different from that suggested by ETHNO, we accepted the program suggestions and modified the model.

Having solved this problem, the analysis proceeded stopping at the eggs decision. At this juncture, none of the solutions offered by ETHNO seemed appropriate given the reporting by Bettman (1970). When this occurs, the program offers as its final option, the ability to add and/or delete events if necessary to make the structure consistent. At this point, the researcher is forced to reconsider the data to determine if errors in recording all events may have occurred. We considered first if the consumer had to decide if the product were meat or produce (X1) before deciding if the product were eggs (X5). Clearly, this does not appear to be a necessary step. The order in which a consumer shops is going to be a function of store layout and the products needed on any particular shopping occasion. A consumer may purchase eggs without purchasing meat or product. Based on the discussion by Bettman of consumer C1, it does appear that independent decisions were made by the consumer. This structure suggests the need for a distinction between facilitating events and decision events. We will define facilitating events as those actions necessary to position the consumer in the correct spatial/temporal situation to make her choice. That is, one must be at the right grocery section to purchase milk and at another section to purchase produce. Varied store layouts may influence the way in which the consumer shops and the way in which he or she processes information. Although Bettman's model does not explicitly make such distinctions, some form of transition is suggested in the model in the form of questions such as X1 and X5. That is, one can assume that the consumer is in fact making her decision either in the produce, meat, or dairy section of the store. The decision to proceed from one section to the other is not recorded or is not reported in the decision structure proposed by Bettman. The shopping decision is not one continuous decision as suggested by Bettman's model but, is in fact several independent decision phases conditioned by the consumer being in the shopping environment. To capture this conditioning process, we added those facilitating decisions such as moving from one store section to another as prerequisite events for purchase decision in each product category. It should be noted however that the events X9 to X37 may not have the correct facilitating structure. Based on a reading of the discussion offered by Bettman, we are unable to determine what and how many different products are actually being purchased. It appears from the model that consumer C1 may have purchased as many as twelve different products on this shopping occasion. Does one infer from the Bettman model that the decision for coffee, flour, sugar, and soup are made in the same structural form? This seems unlikely.



Once the necessary structural changes have been made to the model, the analysis phase is repeated to insure that the new event sequence is compatible with the modeling rules of event analysis. The revised model of Bettman's consumer C1 is shown in Figure 2.

In Figure 2, the decision events Xi are retained and the facilitating events are shown by a series of alpha characters made from the letters of the particular event. This revised structure allows for one to delete some of the events reported in the earlier model. For example, "X44: Is this produce?" may not be needed as a decision step since the consumer is in the produce section. However, depending on the particular store, there may be a number of non-produce items in the produce section: special bottled fruit juices, fruit/vegetable dip mixes, little cupcakes you put fresh berries on top of, etc. From the reported data, one is unable to determine if consumer C's store contained only produce. To be in the produce section and thus to ask if this is produce may or may not be redundant. Another advantage of the ETHNO representation of the decisions processes is that it disentangles the different decisions that the consumer makes and does not force one to represent them as a single decision of many branches.

In addition ETHNO has added additional Special relations. Along with the flip-flop situation described earlier, the term non-depletion has been added. Non-depletion indicates that the occurrence of the superordinate event does not deplete a previous occurrence of the subordinate event. As one can observe, the non-depletions occur at junctions in the decision model that allow for multiple paths through the process. These multiple paths suggest repeated operations on the part of the consumer in question.

Phase Three: Generating Priorities

In the previous section, the software used information in the event series in two different ways. The first use consisted of time ordering the sequence so some conceivable relational questions could be eliminated such as making produce decisions in the produce section and not later in the shopping event. ETHNO assumes that later events cannot be prerequisites for earlier events. The second use of information in the pattern of event series is as a cue that something may be wrong with the original structure. This second phase allows the analyst to adjust the original structure. Once the modified model is congruent with the assumptions, one can now use the event information in a third way by computing priorities.

The grammar(consumer's verbalized decision steps) of an event analysis allows a certain amount of free variation, and this free variation is examined to reveal preferences for certain events over others. The basic question elicited from the analyst and used by the software in its statistical calculations is that when several events are possible, which take priority over the others. An event is given a value of 100 if it always takes priority over alternative events 100% of the time when a choice is available. A value of 0 indicates that the event in question always happens last when nothing else is possible in the series. Other events will thus have values somewhere between these extremes. For the modified model of Bettman's consumer C1, the priority values are shown in Table 1.

By computing the table of probabilities, one can, in some cases, gain insight into the event sequence that may not be apparent from the pictorial representation of the structure. For example in our present shopping case, it is interesting to note that decisions of an external nature such as family preference, prior experience, etc. appear later in the decision sequences and thus have lower probabilities. Another way to view these results is that our subject consumer placed less importance on the decision variables than she placed on other aspects of a more direct evaluation (color, feel, etc.) of the products in question. Alternately, one could argue that each product choice had to pass the earlier criteria before it is considered for family use. The point is that the modeling structure highlights areas that the researcher can go back to the subject and probe for the specific meaning. Without the flag provided by the computer analysis, the deeper meaning may not emerge.

Phase Four: Generate a Series

This final step in our analysis consisted of using the priorities and the model generated by the ETHNO program as an action grammar for producing a new event series. The program began by listing all events that are theoretically possible at the beginning of the data set. The events are listed from high priority to low priority. After selecting the event you want to happen, the program next offers those events that can occur after the selected event. Again as with previous analysis, several alternate pathways appeared and these were modified to produce a final revised model.


In this paper, we have attempted to demonstrate an important methodology for analyzing ethnographic data. It can help researchers systematize their findings and quantify them in interesting ways, without at the same time loosing the richness of ethnographic data. It should however be pointed out here that the data set used in the present case is less than ideal. Specifically, it is hoped that the researcher attempting to use this program would actually take it into the field so that questions of structure could be resolved with the informant. At the very least, the researcher should subject his or her data to the rigors of analysis while the collection experience is still fresh. In the present case, we were working with data that were old and even more limited by what was reported by someone else. Thus, entering the data at this time required the researcher to make a series of highly subjective judgments regarding which prior events are required for the current event to occur. These judgments may have not correctly matched the subject's perception of the world at the actual time. More specifically, one cannot determine at this later date if the problems encountered are that of omission on the part of the subject and/or faulty recording on the part of the observer. Despite these limitations, the use of a computer based analytical approach broadened the application and understanding of the previous work.

Finally, one must keep in mind that using such a program will not ease the data collection burden of the researcher but, may in fact increase his or her task by forcing a greater depth of analysis and understanding. For the research community as a whole, the use of such a tool does provide advantages in that the structure thus developed can be studied, examined, and discussed by different scholars and in different contexts.






Anderson, John R. (1983), The Architecture of Cognition, Cambridge, MA: Harvard University press.

Axten, Nick and John Skvoretz (1980), "Roles and Role-Programs," Quality and Quantity, 14, 547-83.

Bettman, James R. (1970), "Information Processing Models of Consumer Behavior," Journal of Marketing Research, 7 (August), 370-6.

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

Corsaro, William A. and David R. Heise (1990), "Event Structure Models from Ethnographic Data," in Clifford C. Clogg, ed. Sociological Methodology, 1990, Oxford: Basil Blackwell.

Fararo, Thomas J. and John Skvoretz (1984), "Institutions as Production Systems," Journal of Mathematical Sociology, 10, 117-82.

Heise, David R. (1988), "Computer Analysis of Cultural Structures," Social Science Computer Review 6:1 (Spring), 183-96.

Heise, David R. (1989), "Modeling Event Structures," Journal of Mathematical Sociology, 14 (2-3) 139-69.

Heise, David R. and Elsa M. Lewis (1988), Introduction to ETHNO, Raleigh, NC: National Collegiate Software Clearinghouse.

Newell, Auther and Herbert Simon (1972), Human Problem Solving, Englewood Cliffs, NJ: Prentice-Hall.

Waterman, D.A. and Fredrick Hayes-Roth (1978), Pattern-Directed Inference Systems, New York: Academic Press.



James H. Barnes, University of Mississippi


NA - Advances in Consumer Research Volume 20 | 1993

Share Proceeding

Featured papers

See More


F5. Alternative Food Consumption (AFC) Adoption and Low SES Youth Food Well-Being: From Precontemplation to Maintenance

Wided Batat, American University Beirut

Read More


The Impact of Previews on the Enjoyment of Multicomponent Multimedia Experiences

Jayson S. Jia, University of Hong Kong
Baba Shiv, Stanford University, USA

Read More


Placing Identity into the Self-Concept: The Role of Causal Beliefs in Identity-Based Consumption

Stephanie Chen, London Business School, UK
Oleg Urminsky, University of Chicago, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.