Data Collection and Analysis Approaches For Studying Consumer Information Processing

ABSTRACT - Five methods useful for studying the processes involved in choice are examined: information integration, protocol analysis, information monitoring, analysis of eye movements, and the concept of task analysis. Implications of recent work using these methods are discussed, particularly for characterizing consumer decision rules and for understanding how decision rules are developed and implemented.


James R. Bettman (1977) ,"Data Collection and Analysis Approaches For Studying Consumer Information Processing", in NA - Advances in Consumer Research Volume 04, eds. William D. Perreault, Jr., Atlanta, GA : Association for Consumer Research, Pages: 342-348.

Advances in Consumer Research Volume 4, 1977   Pages 342-348


James R. Bettman, University of California, Los Angeles


Five methods useful for studying the processes involved in choice are examined: information integration, protocol analysis, information monitoring, analysis of eye movements, and the concept of task analysis. Implications of recent work using these methods are discussed, particularly for characterizing consumer decision rules and for understanding how decision rules are developed and implemented.


The recent growth in interest in using an information processing perspective in studying consumer behavior phenomena has resulted in a concern with available methodologies for developing and analyzing those models of consumer behavior which explicitly consider information processing. Two general types of models which consider information processing have been examined, structural and process models (Jacoby and Olson, 1976). Structural models focus upon tapping constructs presumed to relate to consumer information processing, usually some measures of psychological states, and then examining the interrelationships among these constructs. A typical study of this sort was the effort by Farley and Ring (1970) to examine the structural relations among constructs underlying the Howard and Sheth (1969) theory of consumer behavior. These structural studies often use notions of simultaneous equation models from econometrics, in some cases combined with causal flow notions. The analyses are usually done at the aggregate level, for groups of consumers. Process models on the other hand, are primarily concerned with the delineation, measurement, and analysis of those processes which are presumed to operate upon information, and which usually accompany changes in psychological states. Process models have been studied with a sometimes bewildering variety of methods, most of which have been developed fairly recently. One of the few features common to many process studies has been a concern with examination of data at the individual level, although aggregate analyses have been used.

This paper does not attempt to compare these two approaches or single out one approach as best; indeed, the approaches are attacking somewhat different problems. Progress in consumer research will be aided more if multiple approaches to research problems in consumer judgment are used (Wright, 1974). Rather, the purpose of this paper is to examine in more detail the approaches to studying process models of consumer choice, particularly some of the new and as yet relatively untapped methods available. Several basic methods for studying processes are outlined below. For each of the these methods, approaches which have been developed to analyze the resulting process data are briefly considered, and finally problems remaining to be solved for each method are highlighted. In this discussion also no attempt is made to choose a "best" approach. Multiple approaches are necessary. The goal is to increase awareness of each approach and the methods available for analysis in each case. The paper closes with a discussion of two recent problem areas in consumer information processing research and shows how the methods outlined can be used to study these problem areas.


Several methods for studying process are discussed below: information integration; protocol methods; information monitoring; and analysis of eye movements. [Note that one of the most typical approaches to studying judgments, the correlational approach, is not considered. This is due to several factors: the approach is well known and needs no detailed review; there are many problems of interpretation and analysis of correlational data on judgments (Bettman, Capon, and Lutz, 1975; Bruno and Wildt, 1975; Anderson and Shanteau, 1977); and most importantly, in many cases one may be studying only the results of processing information, not process itself (Calder, 1975; Kernan, 1974).] In addition, the concept of task analysis is considered as an extremely important tool, which can be used in conjunction with any of the above methods. It is beyond the scope of this paper to give a detailed discussion of each method. More complete discussions exist elsewhere and are referenced for each case.

Information Integration

A typical information integration study begins with the specification of some pieces of information which the consumer can combine to form some judgment. For example, Bettman, Capon, and Lutz (197Sa,b) examine integration of two pieces of information, a belief and an evaluation, to form an attitude. Bettman (1975) studied integration of information on uncertainty and consequences in consumer judgments of perceived risk. Each piece of information is conceived to be a factor in an ANOVA, with several levels (e.g., several degrees of belief, from very unlikely to very likely). Given the specification of the pieces of information, consumers are presented with profiles of data, where the number of items in each profile corresponds to the number of pieces of information to be integrated, and each element in the profile is a particular value for one of the pieces of information. For example, Bettman, Capon, and Lutz (1975a) presented subjects with two item profiles, one item being a specific belief level, and the second being a specific evaluation level for that belief. Note that the subject is thus presented with information which must then be integrated. The subject is not queried about his or her own beliefs. The subject makes some response (e.g., a rating of attitude in Bettman, Capon, and Lutz, or a rating of amount of risk in Bettman) to each profile, and usually receives the set of profiles corresponding to a complete factorial design, often with several replications. For example, in Bettman, Capon, and Lutz (1975a), the belief and evaluation factors each had five levels, and the subjects received two replications of the resulting 5 x 5 factorial design, or 50 profiles in all. Anderson (1971, 1974) has been the major proponent and developer of this information integration approach.

Given the profile rating data, methods of analysis need to be used which can allow one to study the combination rules apparently used by subjects. Fortunately, Anderson (1971, 1974) has developed a relatively complete set of methods, based upon detailed examination of the results of analysis of variance applied to each subject's data individually. Methods for distinguishing adding, averaging, and multiplying models have been developed. Also, more complex models where the weight given to a piece of information depends upon the extremity of the information can be applied.

The major drawback to the information integration approach is the task itself. The factorial task structure and profiles presented may be too simplistic and unreal to subjects. The relatively transparent task may force the results. For further discussion of this approach as applied to consumer research, see Bettman, Capon, and Lutz (1975a,b,).

Protocol Methods

Protocol analysis has been used in several consumer research studies (e.g., Alexis, Haines, and Simon, 1968; Bettman, 1970; Russ, 1971; Payne, 1976a,b). In using this method, the subject is instructed to think out loud as he is actually performing the task of interest, say shopping or choosing among alternatives. This verbal record is termed a protocol. It may be distinguished from introspection or retrospective report in that the subject is verbalizing thoughts as they occur in the course of problem solving. The protocol data are then used to gain insights into the processes being used. The major advantage of the method is that a great deal of data on internal events are made available for inspection. Without these data available, details of the rules used may be lost. Protocol data are then used to develop a model of the processes used by consumers in making Judgments. In most of the research to date, these models have taken the form of decision nets (Bettman, 1974a) in which attributes are arrayed in a branching structure which specifies the specific rules to be applied in judging alternatives.

Many new analysis methodologies have been specifically developed for protocol data and the resulting consumer processing models. Newell and Simon (1972) have pioneered the development of structured methods for analyzing protocols and developing models from them, and Payne (1976a) has been the first to apply these methods in consumer research. Bettman (1974b) has done extensive work in developing measures which allow the researcher to compare and analyze processing models expressed as decision nets. In particular, measures of net structure and measures of similarity for pairs of nets are developed, and issues of fit and model generality are discussed. Recently Takeuchi (1975) has proposed a new measure of the importance of particular attributes in a decision net.

The advantages and problems of protocol data and their analysis are well documented (Bettman, 1974a; Wright, 1974). The major problems are analysis of protocol data in developing a model, measures of fit for model to protocol and model to choice data, and the validity of protocol data given the highly obtrusive nature of the data collection process itself and questions about the ability of subjects, even if willing, to verbalize internal thought processes. It is particularly difficult co obtain verbalization of currently occurring thoughts rather than retrospective rationales when pro-totals are obtained during actual shopping trips.

Information Monitoring Approaches

Information monitoring methods have only recently been used to study consumer processing. Jacoby and his associates (Jacoby, 1975; Jacoby, Chestnut, Weigl, and Fisher, 1976; Bettman and Jacoby, 1976) have been the main developers of this approach, with independent work in parallel done by Payne (1976a,b). The typical approach has been to present information to subjects on an information display board, essentially a matrix (with say brands as rows and attributes as columns) array, with information cards available in each cell of the matrix for the particular brand and attribute appearing in that row and column. [This approach is to be distinguished from the cognitive monitoring approach of Wright (1973). Wright's approach is concerned with measuring the content of thoughts elicited by a communication and then examining the structural relations among the content categories and acceptance of the message.] The subject is asked to choose a brand after selecting as many information cards as desired, one at a time. The sequence of cards selected becomes the major data yielded by the method. Thus a detailed record of the sequence of information examined is made available by directly controlling the selection process for information.

Information monitoring approaches have also led to rapid developments in analysis methodologies. One type of research has examined measures of the structure of the sequence of cards chosen. The first approaches concentrated on analysis of transitions from one card to the next, particularly transitions of two sorts. In the first case, a subject proceeds by choosing a brand, examining several attributes for that brand, choosing a second brand, examining several attributes for that brand, and so on. This may be called processing by brand (PB), and is indicated by sequences of same brand-different attribute transitions. A subject may also proceed, however, by choosing an attribute, determining values for each of several brands for that attribute, choosing a second attribute, and so on. Thus, a series of several different brand-same attribute transitions indicates this processing by attribute (PA) method. Given measures for the extent of brand and attribute processing performed by a subject in the card sequence, the next question addressed was how to classify the subjects into groups according to their sequence structures. Bettman and Jacoby (1976) used a decision net classification scheme which yielded four major groups of subjects: those who used no cards, those who processed by brand, by attribute, or by a hybrid form, labeled feedback processing. Payne (1976b) used an index ranging from -1 (all different brand--same attribute transitions) to +1 (all same brand--different attribute transitions) to classify his subjects into brand and attribute processing groups. Finally, Jacoby, Chestnut, Weigl, and Fisher (1976) developed an ingenious c2 technique to determine degree of fit to each of several pre-specified models and used these degrees of fit to examine relations with other variables. Jacoby et. al. also consider in some detail other possible measures of sequence structure based upon more complex factors than analysis of transitions. In addition, they propose measures for depth (amount of information examined) and content of the information acquired. This area of characterizing the information selected by a subject seems quite fertile for future research.

The major problems to be solved for the information monitoring approach concern the nature of the task. First, it is a relatively obtrusive process, with subjects perhaps biasing their information seeking behavior since it is so obviously under observation. Second, internal processing is not studied directly, but only external information seeking responses. Internal processing of alternatives by decision rules may not be revealed explicitly in the information seeking sequence. Finally, the matrix structure of the information presentation makes it equally easy for a consumer to process by brand or by attribute. This is not like many actual consumer tasks, where information is often organized by brand (e.g., supermarket shelves, commercials), hindering attribute processing. This is discussed further below under task analysis.

Eye Movement Analysis

Analysis of eye movements has also been recently applied to study of consumer choice rules (Russo and Rosen, 1975; Russo and Dosher, 1975; van Raaij, 1976). In using this method, the choice objects are either displayed on a screen in front of the subject, in tabular format (Russo and Dosher, 1975; Russo and Rosen, 1975) or perhaps as separate packages (van Raaij, 1976). The sequence of eye movements used by the subject in examining the choice objects is then recorded using some form of sensing apparatus. To allow the sensing apparatus to give accurate measures, subjects' heads must often be immobilized to prevent large head movements. Finally, to guard against ambiguous interpretations due to subjects' use of peripheral vision, there must be relatively large separations between items in the visual display. The resulting eye movement data can be examined to look for patterns in the sequence of fixations. The main advantage of eye movement data is that detailed trace of the information examination process is provided, and eye movements may be relatively more difficult to censor than verbal protocols. Newell and Simon (1972) report an application of eye movement data by Winikoff (1967) to a problem solving situation, and found that the eye movement data corresponded quite well to protocol data, but also added some important information.

Several new methods have been developed for analyzing eye movement data. As with information monitoring, one of the major issues is developing measures that allow the researcher to make inferences from the eye movement sequences. Russo and Rosen (1975) develop measures for indicating when a subject is making a comparison between a pair of alternatives. Russo and Dosher (1975) also examine measures for attribute and brand processing. Finally, Russo end Rosen and Russo and Dosher used a combined methodology that seemed to be quite effective. In addition to the eye movement data, prompted protocols were obtained, prompted by a replay of the sequence of eye movements to the subject after his choice was made. This type of multimethod approach seems quite valuable, as the various methods have different strengths and weaknesses and can often complement each other.

Eye movement data also have unresolved problems. First, collection of such data requires processing one subject at a time and is very time consuming. Thus sample sizes are usually small. Also, the apparatus often used is quite reactive, since head movement must be restricted and subjects are obviously aware that their eye movements are being monitored (some newer eye movement systems do not require this restriction of the head). The apparatus is also quite complex and expensive to construct. The choice stimuli used in eye movement studies have often been simplistic arrays because of the desire to be able to localize eye movements. More detailed and complex visual stimuli might cause problems. Finally, the fixations are information seeking responses, and do not necessarily reveal the details of internal processing.

Task Analyses in Studying Consumer Choice Processes

A concept which has not been applied to any great extent in consumer research but seems to be extremely useful in conjunction with any of the above methods is the idea of task analysis. Newell and Simon (1972) argue quite forcefully that the structure of the task being undertaken by a subject greatly influences what processes will be used. That is, the task itself imposes certain constraints on what types of processes will work in accomplishing that task. Newell and Simon (1972) argue that a thorough task analysis, a consideration of what the task requires for successful completion, yields a great deal of knowledge about how behavior must be structured to be adaptive in the particular task environment. Much of the difficulty in building models of consumer choice stems from the inability to perform detailed task analyses, given the range of tasks undertaken by consumers. For example, consumer tasks vary a great deal even within the realm of choice among alternatives. Whether a choice is made before arriving at a store or only after actually in the store is important. If the choice is made in the store, then the product display itself serves as an external memory, reducing the memory load on the consumer. This allows the consumer to use recognition rather than recall to access any needed facts stored in his or her own memory. However, this may not always be true. If a consumer tends to choose among alternative product classes for some need as opposed to choosing among brands from one product class, then a physical separation of product classes in the store may eliminate some of the aid provided by the external memory. Also, store displays are arranged by brand, which hinders processing by attribute. Thus, fairly subtle distinctions among the kinds of choice tasks encountered by the consumer can lead to use of very different mechanisms in making these choices. [Another example of task effects is the difference among various tasks related to obtaining information. Attempting to process information from a television commercial, where the rate and sequence of information flow are outside the consumer's control, poses very different task requirements from a print ad. One might hypothesize that products where the main information source is television commercials would be marked by use of recognition strategies and hence in-store processing to a greater extent than those products where more print information is utilized, since processing information for later recognition is presumably easier than attempting to store the information for future recall.]

Thus, it seems crucial to perform a task analysis before studying any consumer choice phenomenon of interest. In addition, task analyses of research tasks presented to subjects can be quite useful in uncovering biases built into those tasks. For example, the information display board task used in most information monitoring studies has several properties which may bias the results obtained. First, the row and column structure makes it equally easy to process by brand or by attribute. This is not congruent with many consumer tasks, where information is organized by brand, thus making attribute processing more difficult (e.g., shelf displays, most advertisements). Second, the row and column structure also makes it easier for the consumer to process sequentially through a row or column, using the organization imposed by the matrix display, rather than imposing his or her own organization on the search process, perhaps based upon specific attributes or brands. By examining these task effects, more insight into adaptive information gathering strategies can be obtained. Bettman and Kakkar (1976) manipulated the ease of processing by brand or attribute and found that consumers processed information in that fashion which was easiest given the display used. If attribute processing was encouraged by the display, that is how the information was acquired, and similarly for brand processing. Bettman and Kakkar also examined the second issue, board structure versus consumer imposed structure, by developing a random display condition, where the display was in matrix form, but with cells randomly placed. This design superimposes two types of tasks. The physical display matrix provides constraints on accessing the entries; the content of the display, however, is randomly arranged, so that if the consumer wishes to seek specific content he or she may depart from row and column access. Thus the access question and the consumer's imposition of structure are separated. The results showed that roughly 30% of the subjects imposed their own structure, roughly 50% used row or column search schemes, and roughly 20% used some other scheme.

Thus, one of the most important areas for future consumer research is how to develop task analyses for consumer tasks. Formal methodologies need to be developed which will allow determination of the major dimensions along which tasks vary with regard to their information processing properties, and the Implications of variation along these dimensions for the specific processes used. This type of analysis seems to be basic for developing process models of consumer choice.

Now two problems of current interest in building consumer process models are examined to illustrate how the process methods discussed above may be useful in developing more detailed models. The first problem considered, building models of decision rules, is one which has seen a great deal of research. The second, dealing with the constructive nature of choice processes, has seen little study.


Consumer Decision Rules

Wright (1975), in characterizing decision rules, argues that two major aspects must be considered. 1) The process by which an alternative is evaluated, or the specification of how an indicator of value is assigned to an alternative. A rule partitions the set of alternatives into classes on an evaluation dimension. This aspect of a decision rule can be viewed as taking beliefs about alternatives as inputs and arriving at an attitude as an output. An attitude is thus seen as the degree of affect toward or against the alternative, and as an outcome of the process of choice. 2) A decision rule must also specify the choice criterion, or the process by which one alternative is chosen from among the set of alternatives which have been evaluated. This choice criterion then adds intention to the evaluation. Wright (1975) discusses the two aspects above, evaluation process and choice criterion. He notes that choice criteria could be 'choose the best,' 'choose the first alternative that is satisfactory,' and so on. However, Wright (1975) omits a crucial third aspect which must also be specified, and that is 3) how alternatives are processed to obtain an evaluation and make a choice. Two basic processing forms were outlined above, processing by brands (PB) and processing by attributes (PA). As noted above, other hybrid forms may be used but are not crucial for the present discussion.

This distinction about form of processing affects the notions presented above. If PA is used, a rule may function to yield an overall evaluation for each alternative, but in a different manner than implied by the discussion above. For a lexicographic choice procedure, some alternatives may be eliminated after the most important attribute has been examined, for example. This group of alternatives which is eliminated may form one evaluation class, with further classes formed as the process continues. It is not clear that this derived evaluation has any relation to what the consumer does, however. If PA is used in a manner where elimination is not used, it is difficult to specify how (or even if) evaluations are assigned to specific alternatives. For example, if differences between alternatives on each attribute are utilized, as in the additive difference model (Tversky, 1969), only a difference in evaluation may be available. Thus the notion of an evaluation process and of an attitude seem based on a view of processing where each alternative is evaluated and then evaluations are compared (PB). As shown below, subjects may not use this PB form.

Eight major decision rules and the properties of each with regard to evaluation process and form of processing are summarized in Table 1. For example, the compensatory formulation of the multiattribute attitude model, whether adding or averaging formulations are used, assumes processing by brand. Each brand is evaluated as a unit. A lexicographic model, on the other hand, explicitly assumes processing by attribute, since all brands are compared on the most important attribute, then on the second most important if there are ties, and so on.

This table suggests that one way to analyze what choice models consumers are using is to examine form of processing, therefore. Many recent studies have examined this question. Russo and Dosher (1975), using eye movement and prompted protocol data, find that most of the choice processing carried out by their subjects was attribute processing, with a heuristic version of the additive difference model seeming to describe their data quite well. Only if differences on dimensions were difficult to compute did processing by object occur. Wright and Barbour (forthcoming) also support a heuristic additive difference model. Beckman and Jacoby (1976) (information monitoring), Russ (1971) (information monitoring and protocols) and Russo and Rosen (1975) (eye movement data and prompted protocols) also support evidence for a good deal of processing by attribute.

However, if the tasks used in these studies are subjected to task analysis, one can argue that the types of display used could have biased the results. Tversky (1969) also argues that the format of presentation of the data has important effects. If alternatives are displayed sequentially (as in advertisements for exam-pie), processing by objects is encouraged. If information on dimensions of several alternatives is available simultaneously, people show strong tendencies to process by attribute. This factor implicitly appears in the above studies. The information displays (mostly visual arrays) used by almost all of the studies make attribute and object processing equally easy, and subjects prefer attribute processing. Svenson (1974), on the other hand, used individual booklets for each alternative and found more processing by brand using protocol data. Van Raaij (1976) used individual packages and eye movement data and found more brand processing. As noted above, Bettman and Kakkar (1976) found strong task effects. It was presumably more difficult for subjects to use attribute strategies in these tasks. Thus, these process studies have shown that form of processing is strongly affected by format of the display, and that attribute processing is used extensively in task environments that do not hinder it. These results imply that the types of models used by consumers will be to some extent contingent upon the information displays available. In typical consumer task environments, brand processing is encouraged, so these models using brand processing would be favored.[This typical organization by brand can be manipulated, of course. If a marketer feels he has a differential advantage on some attribute and wishes to encourage attribute processing, he can use a comparative advertisement or package information. One manufacturer recently followed this strategy, putting a "comparison table" on their package, comparing their brand with several other brands on two attributes. If the information environment is not conducive to the form of processing desired, one can attempt to alter their information presentation format to achieve the desired effect.]



Application of process methods and a focus on form of processing have thus provided important insights into consumer choice models that were not possible with other approaches, and have raised some serious issues. These insights are contingent upon the assumption that internal decision making processes are congruent with the form of the external information seeking responses, which seems to be reasonable. First, the effect of the format of information presentation upon the resultant form of processing shows that which models a consumer uses may be affected by how information on alternatives ia presented. In most past research, there may have been potentially strong biases due to this factor. Correlational methods must develop some predicted evaluation for each alternative, and hence effectively assume brand processing for all rules. Also, in most studies, information on alternatives has been presented by brand, e.g., with information on each alternative displayed on a separate card. Thus, past results would be biased toward use of models which are characterized by brand processing (e.g., linear, conjunctive) and biased against models characterized by attribute processing (e. g., lexicographic, elimination by aspects). [The extent of any such bias is difficult to assess, since linear compensatory models have been shown to be extremely robust in most studies on choice, largely for statistical rather than psychological reasons (Dawes and Corrigan, 1974).] A second major issue is whether an attitude, an overall evaluation of each alternative, is a necessary or even useful component of choice processes. Most of the models cited do not develop a direct evaluation, only a derived one, and attribute processing is not always conducive to forming such an evaluation. 6 The emphasis on attitude may be based on the reliance on linear models, which do develop a direct evaluation, and the past reliance on correlational methods, which also are biased toward direct evaluations and brand processing, as noted above. The important questions are whether the attitude construct too narrowly limits the types of choice processes congruent with it, and whether an evaluation of each alternative is needed at all to depict many choice processes. Attitudes may be developed in parallel with choice, but not as a direct result of the choice process undertaken. Alternatively, a choice may be made using some type of choice process (e.g., lexicographic, additive difference), with attitudes formed after usage experience, and eventually future choices may be made by affect referral. This notion of relation of attitude to choice process is similar to the self-perception notions of Bem (1967). [For an example of a model which does attempt to form a direct evaluation based on attribute comparisons, see Nakanishi, Cooper, and Kassarjian (1974).]

Constructive Aspects of Choice Processes

In contrast to the relatively large amount of research using process methods to study decision rules, there is virtually no research on the constructive nature of choice processes. There are two basic competing characterizations for the type of choice processing that may be used. One is that man has available a repertoire of strategies already existing in memory and a control process which in effect calls these strategies when needed, much in the same manner as a computer program uses subroutines. In the most extreme form of this notion, the strategies are called as complete units from memory and executed. A second basic characterization of choice processing is that it is constructive. That is, rules for processing alternatives are developed from fragments or elements of rules existing in memory. These elements may be beliefs, evaluations, combination rules, and so forth. There is a hierarchy of general plans, and these plans guide the construction of a rule in the situation. Under this conception, the strategies or rules are not stored in their entirety in memory, but exist as fragments, or sub-parts, which are put together constructively at the time of processing. Thus rules may differ from one situation to the next, depending upon how the fragments or elements are combined. The overall process is a cyclical one, with alternating cycles of construction, information search, and possibly changes in goals.

Both of these characterizations are true of choice processing, with the adequacy of each at any point in time depending upon degree of learning, the task environment, and whether the consumer tends to process in the store situation or to do some prior processing outside of the store. One proposition espoused here is that processing is initially constructive when a choice is made for the first time, that in new situations rules tend to be built up from elements. This constructive nature of processing will thus be most evident when there is little knowledge of the particular choice situation and when processing is done in the store situation. Under these conditions, the constructive elements are mainly built up from information available in the specific task environment, and rules for combining these elements are developed. However, if the consumer has had prior experience with a particular choice, elements may exist in memory, and if used over time the elements may become organized into an overall response strategy in memory. In effect, rule elements are "chunked" into a strategy. In this case, the consumer's use of a rule may be approximated by the "subroutine" concept of a rule's being called as a unit from memory. In this sense, use of rules as learned units is seen as the result of habituation in the constructive process. At any point in time, a particular consumer will thus be engaged in both construction and usage of learned rules, the extent of each being determined largely by experience with the choices under consideration. However, the nature of the constructive process is not well understood, since the "subroutine" conception bas dominated consumer research.

How would one study these constructive processes using process methods? The first step might be to merely isolate the phenomenon. Bettman and Zins (1976) attempted to study constructive and other types of processing by using consumer protocol data as an input to a panel of judges, who judged each of many choice "episodes." The results showed that the protocol data were extremely ambiguous and difficult to judge. Of those items where the judges agreed at a reasonable level (about 44% of the items), roughly 25% of the episodes were judged as being constructive in nature, and 75% non-constructive. A second finding was that the protocol data themselves are subject to many biases and limitations for making inferences about the nature of processing carried out by consumers. Experimental approaches were suggested for studying constructive processes.

Process methods can be extremely useful in developing models of consumer behavior. In the above several methods have been summarized, and a few applications of these methods noted. The importance of understanding consumer tasks has been stressed as a preliminary step to studying processing. Despite the array of available methods and techniques, however, many problems remain, both for developing new methods and resolving issues with present methods. More accurate, easier, and faster methods for obtaining process data and for analyzing these data once obtained are most important goals for future research.


Marcus Alexis, George H. Haines, and Leonard Simon, "Consumer Information Processing: The Case of Women's Clothing," in Proceedings of the Fall Conference, Marketing and the New Science of Planning (Chicago: American Marketing Association, 1963), 197-205.

Norman H. Anderson, "Integration Theory and Attitude Change," Psychological Review, 78(March,1971), 171-206.

Norman H. Anderson, "Methods for Studying Information Integration," Technical Report No. 43, Center for Human Information Processing, Department of Psychology, University of California, San Diego, 1974.

Norman H. Anderson and James Shanteau, 'Weak Inference with Linear Models," Psychological Bulletin, (1977), in press.

Daryl J. Bem, "Self Perception: An Alternative Interpretation of Cognitive Dissonance Phenomena," Psychological Review, 74(1967), 183-200.

James R. Bettman, "Information Processing Models of Consumer Behavior," Journal of Marketing Research, 7 (August, 1970), 370-76.

James R. Bettman, "Decision Net Models of Buyer Information Processing and Choice: Findings, Problems, and Prospects," in G. David Hughes and Michael L; Ray; eds. Buyer/Consumer Information Processing (Chapel Hill: University of North Carolina Press, 1974a), 59-74.

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

James R. Bettman, "Information Integration in Consumer Risk Perception: A comparison of Two Models of Component Conceptualization," Journal of Applied Psychology, 60(1975), 381-85.

James R. Bettman, Noel Capon, and Richard J. Lutz, "Cognitive Algebra in Multi-Attribute Attitude Models," Journal of Marketing Research, 12(May 1975a), 151-164.

James R. Bettman, Noel Capon, and Richard J. Lutz, "Multiattribute Measurement Models and Multiattribute Attitude Theory: A Test of Construct Validity," Journal of Consumer Research, 1(March 1975b), 1-15.

James R. Bettman and Jacob Jacoby, "Patterns of Processing in Consumer Information Acquisition," in Beverlee B. Anderson, ed., Advances in Consumer Research, Volume 3 (Chicago: Association for Consumer Research, 1976), 315-20.

James R. Bettman and Pradeep Kakkar, "Effects of Information Presentation Format on Consumer Information Acquisition Strategies," Working Paper No. 43, Center for Marketing Studies, Graduate School of Management, University of California, Los Angeles, 1976.

James R. Bettman and Michael A. Zins, "Constructive Processes in Consumer Choice," Working Paper No. 42, Center for Marketing Studies, Graduate School of Management, University of California, Los Angeles, 1976.

Albert V. Bruno and Albert R. Wild,, "Toward Understanding Attitude Structure: A Study of the Complementarity of Multi-Attribute Attitude Models," Journal of Consumer Research, 2(September, 1975), 137-45.

Bobby J. Calder, "The Cognitive Foundations of Attitudes,'' in Mary Jane Schlinger, ed., Advances in Consumer Research, Volume II (Chicago: Association for Consumer Research, 1975), 241-47.

Robin M. Dawes and Bernard Corrigan, "Linear Models in Decision Making," Psychological Bulletin, 81(February, 1974), 95-106.

John U. Farley and L. Winston Ring, "An Empirical Test of the Howard-Sheth Model of Buyer Behavior," Journal of Marketing Research, 7(November, 1970), 427-38.

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

Jacob Jacoby, "Perspectives on a Consumer Information Processing Research Program," Communication Research, 2(July, 1975), 203-15.

Jacob Jacoby, Robert W. Chestnut, Karl C. Weigl, and William Fisher, "Pre-purchase Information Acquisition: Description of a Process Methodology, Research Paradigm, and Pilot Investigation," in Beverlee B. Anderson, ed., Advances in Consider Research, Volume III (Chicago: Association for Consumer Research, 1976), 306-314.

Jacob Jacoby and Jerry C. Olson, "Consumer Response to Price: An Attitudinal Information Processing Perspective,'' in Y. Wind and H. Greenberg, eds., Moving Ahead with Attitude Research (Chicago: American Marketing Association, 1976), in press.

Jerome B. Kernan, "Comments on the Role of Summated Attribute Models in the Central Processing of Information,'' in G. David Hughes and Michael L. Ray, eds., Buyer/Consumer Information Processing (Chapel Hill: University of North Carolina Press, 1974), 353-355.

Masao Nakanishi, Lee G. Cooper and Harold H. Kassarjian, "Voting for a Political Candidate Under Conditions of Minimal Information," Journal of Consumer Research, l(September, 1974), 36-43.

Allen Newell and Herbert A. Simon, Human Problem Solving (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1972).

John W. Payne, "Heuristic Search Processes in Decision Making," in Beverlee B. Anderson, ed., Advances in Consumer Research, Volume III (Chicago: Association for Consumer Research, 1976a), 321-27.

John W. Payne, "Task Complexity and Contingent Processing in Decision Making: An Information Search and Protocol Analysis," Organizational Behavior and Human Performance, (1976b), in press.

Frederick A. Russ, "Consumer Evaluation of Alternative Product Models," Unpublished doctoral dissertation, Carnegie-Mellon University, 1971.

Edward J. Russo and Barbara A. Dosher, "Dimensional Evaluation: A Heuristic for Binary Choice," Unpublished working paper, Department of Psychology, University of California, San Diego, 1975.

Edward J. Russo and Larry D. Rosen, "An Eye Fixation Analysis of Multi-Alternative Choice," Memory and Cognition, 3(May, 1975), 267-76.

Ola Svenson, "Coded Think Aloud Protocols Obtained When Making A Choice to Purchase One of Seven Hypothetically Offered Houses: Some Examples," Unpublished paper, University of Stockholm, 1974.

Lawrence R. Takeuchi, "The Structural Analysis of a Consumer Information Processing Model," in Proceedings of the Fall Conference (Chicago: American Marketing Association, 1975), 156-61.

Amos Tversky, "Intransitivity of Preferences," Psychological Review, 76(January, 1969), 31-48.

Amos Tversky, "Elimination by Aspects: A Theory of Choice," Psychological Review, 79(July, 1972), 281-99.

W. Fred Van Raaij, "Direct Monitoring of Consumer Information Processing by Eye Movement Recorder," Unpublished paper, Tilburg University, 1976.

Arnold Winikoff, "Eye Movements as an Aid to Protocol Analysis of Problem Solving Behavior," Unpublished doctoral dissertation, Carnegie-Mellon University, 1967.

Peter L. Wright, "The Cognitive Process Mediating Acceptance of Advertising," Journal of Marketing Research, 10(February, 1973), 53-62..

Peter L. Wright, "Research Orientations for Analyzing Consumer Judgment Processes," in Scott Ward and Peter L. Wright, eds., Advances in Consumer Research, Volume I (Chicago: Association for Consumer Research, 1974), 268-79.

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

Peter L. Wright and Frederic Barbour, "Phased Decision Strategies: Sequels to an Initial Screening," Management Science (forthcoming).



James R. Bettman, University of California, Los Angeles


NA - Advances in Consumer Research Volume 04 | 1977

Share Proceeding

Featured papers

See More


Making the Wait Worthwhile: Mental Accounting and the Effect of Waiting in Line on Consumption

Chris Hydock, Georgetown University, USA
Sezer Ulku, Georgetown University, USA
Shiliang Cui, Georgetown University, USA

Read More


Consumers’ Trust in Algorithms

Noah Castelo, Columbia University, USA
Maarten Bos, Disney Research
Donald Lehmann, Columbia University, USA

Read More


A2. A Window to the Brand's Soul: How Models' Eyes Affect Consumers' Attitudes

Maria Giulia Trupia, IESE Business School
Martina Cossu, Bocconi University, Italy
Zachary Estes, Bocconi University, Italy

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.