Consumer Information Processing: Attributing Effects to Causes

Citation:

Robert B. Settle, John H. Faricy, and Glenn T. Warren (1971) ,"Consumer Information Processing: Attributing Effects to Causes", in SV - Proceedings of the Second Annual Conference of the Association for Consumer Research, eds. David M. Gardner, College Park, MD : Association for Consumer Research, Pages: 278-288.

Proceedings of the Second Annual Conference of the Association for Consumer Research, 1971     Pages 278-288

CONSUMER INFORMATION PROCESSING: ATTRIBUTING EFFECTS TO CAUSES

Robert B. Settle, University of Florida

John H. Faricy, University of Florida

Glenn T. Warren, University of Florida

INTRODUCTION

An important part of the marketer's task is to obtain consumer acceptance of a brand or product through the management of information. Considerable attention has been focused on the management of information by the marketer, particularly in the area of promotion and advertising. The "elements" in the promotional mix have been dissected and analyzed, the form and content of the persuasive message have been studied, and the effect of various media and communicators have been explored. These efforts have certainly not proven useless, and it seems quite likely that continued research in this area will provide further insight into the effective use of information.

Just as marketers manage information, so do consumers. Just as the communicator has a variety of alternative methods to sort and present information, the receiver has several alternative methods to select, evaluate and use the information. The objectives, alternatives, and methods of consumer information management have been studied to a much lesser extent than have those of marketers because the focus of research has been predominantly on the preparation and "sending" of information.

If the goal of information management is successful communication, it seems likely that both marketer and consumer would benefit from further understanding of the receiver's information processing methods. It is the purpose of this study to investigate one mode of consumer information processing; that suggested by attribution theory.

Attribution Theory

Fritz Heider (1958) first described the attribution process in his book, The Psychology of Interpersonal Relations, as ". . . the linking of an event with its underlying conditions. . ." [p. 89]. He noted, ". . . that man is usually not content simply to register the observables that surround him: he needs to refer them as far as possible to the invariances of his environment'! [p. 81]. Heider suggested that the attributions are made on the basis of a ". . . naive factor analysis of action" [p. 123]. He also observed, ". . . correct attributions . . . always serve to build up and support the constancy of our picture of the world" [p. 92]. He does not suggest, however, that the attributions made on the basis of this naive factor analysis will always conform to objective reality. Distortions may result from imperfect knowledge of the conditional antecedents of the action, or from egocentric needs of the individual.

Attribution theory was elaborated and extended by Harold H. Kelley (1967) in an extensive study of the theory in social psychology. Kelley agreed that, basically, the attribution is made on the basis of covariation. The effect is attributed to a causal condition that is present when the effect is observed and absent when the effect is absent. He used a simple three-dimensional cube as an expositional device of the analysis of variance, as depicted in Figure 1.

FIGURE 1

DATA PATTERN INDICATING ATTRIBUTION OF SMOOTH OPERATION TO BRAND A GASOLINE

Validation Criteria

Kelley (1967, p. 197) identified four criteria used by the individual to subjectively validate attributions:

1. Distinctiveness: the impression is attributed to the thing if it uniquely occurs when the thing is present and does not occur in its absence.

2. Consistency over time: each time the thing is present, the individual's reactions must be the same or nearly so.

3. Consistency over modality: his reactions must be consistent even though his mode of interaction with the thing varies. (For example, he sees it to have an irregular outline and he feels it to be rough; or first he estimates the answer to the problem and then he calculates it.)

4. Consensus: attributes of external origin are experienced the same way by all observers.

The more perfectly the individual's attributions fulfill the criteria, the more confident he will be that he has a valid picture of the world. An example may clarify the use of the criteria for subjectively validating an attribution.

If an auto owner observed that his car runs smoothly on Brand A gasoline but knocks and misses with all other brands, he may realize that smooth operation is uniquely associated with Brand A. This situation is depicted in Figure 1. The S for smooth operation appears in several cells of the Brand A layer, but in no other layer of the cube. This indicates that smooth operation is distinctive to Brand A, and to that extent the driver will be confident of his attribution of the effect, smooth operation, to the cause, Brand A.

If the driver finds that Brand A is associated with smooth operation every time he uses this brand he will be more confident, since he has achieved consistency over time with the brand. Similarly, he will be more confident if he finds that this effect is present in both city and country driving. This can be seen as consistency over modality. Note that Figure 1 depicts an S in both city and country modes, all four times the gasoline is used.

Lastly, the car owner will be more confident to the extent that other drivers recognize the same association between the brand and the effect, and make similar attributions. If they communicate their attributions to him, he will have achieved consensus. This condition is indicated in Figure 1 by the symbols appearing in the layers labeled O for other drivers.

In one respect, assurance reached through consensus is quite different from assurance reached through direct personal experience. When another person communicates his experience to the person making the attribution, the communicator's message is, in itself, an effect. That is, something caused the communicator to express himself. The receiver must make some judgment of the validity of the message in view of the communicator's possible motives and intent. The receiver may make these judgments by using the same naive analysis of action cited earlier. He may cast the problem in terms of whether or not the communicator is consistent over time and modality in presenting such a message, whether he is in agreement or disagreement with others, and whether he makes such statements only in relation to the topic under discussion, or in relation to every topic which is brought to his attention. In such a manner, the receiver may attribute the effect, in this case, the message, to such various factors as a genuine interest in him, an ulterior motive of the communicator, or some irrelevant factor-such as the egocentric need of the communicator to influence everyone with whom he comes in contact. The degree to which the receiver achieves assurance through consensus will depend on this attribution concerning the source.

In reference to source factors which affect the attributions of the individual, Kelley (1967, p. 204) comments:

The communicator factors usually considered relevant to [the individual's] acceptance of persuasive communication are [the communicator's] expertness and trustworthiness. These notions are readily reduced to attribution terms. Expertness can be defined as the communicator's contact with or mediation of the relevant external causal factors. . . . Trustworthiness implies the absence of irrelevant causal factors (personal motives, role demands) in the person's statements.

While these communicator factors appear to be amenable to study using an attribution theory framework the focus of this investigation is on the consumer's mode of information processing. Without denying the distinctive nature of the consensus method of validating an attribution, it is possible to see consensus as a third form of consistency. One can then speak of consistency over time, consistency over modality, and consistency over people. It should be noted, however, that consistency over people (consensus) involves a double or higher order attribution: first, an attribution of the message as an effect to some causal factor, and then secondly, an attribution of the content of the message to a causal factor.

Causation and Covariation

While Heider and Kelley formed and used attribution theory in the study of interpersonal relations, it is the purpose of this study to apply Kelley's formulation to the consumer's information processing. The theory is premised on the assumption that the individual will attribute an effect to a cause on the basis of covariation. Thus, if a given effect consistently appears in the presence of one possible cause, and never appears when the cause is absent, the individual should be relatively certain that the causal factor is, in fact, the single causal agent. The less consistent the relationship or the more frequently the effect is associated with other possible causal agents, the less certain the individual should be that he has identified the unique cause of the effect.

An example may clarify this critical concept. Assume that the effect to be considered is a remark by another person that a certain movie is very poor. The individual receiving this information must now use some process to evaluate the effect's implications for his behavior. If the process he uses is that of assessing covariance, as attribution theory suggests, he might proceed in the following manner:

Two possible causes for this effect will be considered: the other person dislikes movies of all kinds, or the movie is such that all kinds of people would dislike it. The person making the attribution would then look at the variation in the communicator's remarks about other movies and also look at other people's remarks about this movie. If he finds that this person rates every movie as poor, while some others rate this movie as good or fair, he will attribute the effect, a poor rating for this movie, to the 'cause" which is present when the effect is present, namely the other person. If, on the other hand, he finds that this particular communicator rates some movies as good, some as fair, and some poor, and if he finds that all others rate this particular movie as poor, the effect (poor rating for this movie) would be attributed to the movie itself, rather than to the other person.

Attribution theory further stipulates that the degree of confidence of the individual will be a function of the degree of consistency. That is, the more consistent the relationship between cause and effect, the more confident the individual will be that he has, in fact, identified the real cause.

Hypotheses

Two principal hypotheses can be drawn from the example cited above:

I. The individual will attribute an effect to a cause on the basis of covariance.

II. The degree of certainty that the attribution is correct is a function of the consistency of the relationship between cause and effect

METHODOLOGY

The hypotheses are subject to empirical test through experimentation. The method of investigation must present the experimental subjects with a data base and measure the direction of the attribution and the degree of confidence in its correctness.

Experimental Design

The basic design for presenting the data base to the subjects (S's) and obtaining the responses is presented in Table 1. Two sets of cells at the lower and right margins of each data matrix are identified by underscoring. and the S's were asked to indicate both their estimate of the rating and their confidence that the rating was "correct.' The estimates were to be made on the basis of the information on the ratings of three movies by three people, as indicated in the body of the table. Table 1, then; depicts the design for both the presentation of a data base and the measurement of response.

All S's in each condition received nine pieces of information concerning how three fictitious people rated three fictitious movies. The S's were then asked to estimate the ratings which a fourth person would assign to the three movies, and the ratings which the three people would assign to a fourth movie. They were asked to indicate their degree of confidence in each of the six estimates, as well.

In Condition I, the three fictitious informants were entirely consistent in their ratings of any one movie.

TABLE 1

DESIGN FOR PRESENTATION OF THE DATA BASE AND COLLECTION OF THE SUBJECTS' RESPONSE

All of the variance in the data base matrix is associated with movies. Attribution theory suggests that the subjects will attribute the effect (the ratings) to the causal factors on the basis of covariation. Thus, the subject might assume that this situation contains a good movie, a fair movie, and a poor movie.

In Condition II, the three fictitious movies were all rated exactly the same by any one person. All the variation is associated with people. Attribution of the effect to a cause on the basis of covariation would lead the subject to the conclusion that this situation contains one person who likes movies, one who is rather neutral, and one who dislikes movies. In other words, the effect (the ratings) is attributed to the people, rather than to the movies as in Condition I.

Test Instrument

The objective of the test instrument was to present the S's with the nine pieces of information, and to obtain six responses for each dependent variable, or a total of twelve responses from each subject. The information and the questions were presented in random order to avoid any systematic order bias.

The S's were given an instruction sheet and a pack of standard data processing cards. This pack consisted of nine information cards and six question cards. The information cards were buff colored, and each contained a simple statement, such as:

ALAN rated the movie "Karelia" as Good.

The question cards were white so that they could be distinguished easily by the S's, and each card contained two questions:

How do you think that DALE will rate the movie "Karelia"

__Good __Fair __Poor

How certain are you that he will rate the movie that way?

___:____:____:____:____:____:___:____:___

90% 80% 70% 60% 50% 40% 30% 20% 10%

Statistical Design

Hypothesis I states that the individual will attribute an effect to a cause on the basis of covariance. If this is the case, a given effect (a rating) should be relatively easy to estimate in the presence of the cause and relatively difficult in the absence of the cause. In Condition I, for example, if the effect is attributed to the movies, the S's should be able to estimate the effect (rating) with relative ease if the cause (the movie itself) is a "known" even though the fourth person is not "known" in the sense that the S's have data on him. On the other hand, an estimate of the rating for an "unknown" movie would be difficult even though the S's have data on a "known" person who is making the rating. Of course, the converse is true for Condition II, if the effects (ratings) were attributed to the persons, rather than to the movies.

The relative ease or difficulty in making the estimates should be reflected in the accuracy of the estimates, if row and column means are used as the criteria for correct estimates. Consequently, the frequency of a "correct" estimate in each cell can serve as the dependent variable to test Hypothesis I. The data were submitted to Chi2 analysis in a contingency table for proportion of correct and incorrect estimates in high and low consistency modes.

Hypothesis II states that the degree of certainty that the attribution is correct is a function of the consistency of the relationship between cause and effect. The dependent variable relevant to this hypothesis is the rating of confidence in each estImate. A significant difference between mean responses of confidence for high and low consistency treatments would support the hypothesis. The confidence data were submitted to analysis of variance to determine significance of differences in means

RESULTS

The frequency of "correct" estimates for each one of the marginal cells, together with the mean confidence ratings for the 45 subjects in each condition, are depicted in Table 2. The frequencies are expressed as both absolute value and as a percentage of total response. Each of the 45 subjects in each condition responded to every question. The scale of confidence ranged from zero to nine.

TABLE 2

FREQUENCY OF CORRECT ESTIMATES AND MEAN CONFIDENCE RATINGS

Frequency of Correct Estimates

The frequency of correct and incorrect estimates of ratings for high and low consistency treatments are shown in Table 3. The high consistency treatment refers to the rows of Condition I and the columns of Condition II, while the low consistency treatment was the columns of Condition I and rows of Condition II. Consequently, each treatment includes both consistency over people (as in Condition I) and consistency over movies (as in Condition II).

The Chi analysis of the contingency table presented in Table 3 indicates that the differences in the distributions of response between correct and incorrect ratings for high and low consistency is highly significant. It can be concluded from the data that subjects had significantly more difficulty in estimating a rating for an "unknown" entity, whether a person or a movie, when all of the variance in the matrix was associated with that class of entities. Conversely, the extremely high proportion of correct responses indicated that subjects had little difficulty in making the estimates when no variance was associated with the class of entity (whether people or movies).

Hypothesis I states that the individual will attribute an effect to a cause on the basis of covariance. The data on estimation of ratings appear to indicate that the subjects are attributing the effects (ratings) which they were given to either the entities being rated (movies) or to the entities doing the rating (people) on the basis of the variance in the information data. As a result, they seem to have exhibited relatively more difficulty in making an estimate of a rating for an unknown entitly when that class of entities, whether people or movies, is seen as the "cause" of the effects. The data support Hypothesis I.

TABLE 3

FREQUENCY OF CORRECT ESTIMATES FOR HIGH AND LOW CONSISTENCY TREATMENTS

Confidence Ratings

The mean confidence ratings for each example of an entity class are depicted in Table 2. The data for each condition were submitted individually to analysis of variance to determine if there were significant difference in mean response for subjects, consistency treatments, and examples within treatments (the three people and movies). The results of the analysis are presented in Tables 4 and 5.

The analysis of variance indicates that there are significant main effects of both subjects and consistency treatments, as well as a significant interaction between these variables. Examples within treatments did not prove to produce significant differences in mean response. The data indicate that confidence ratings are significantly higher for high consistency examples than for low consistency examples, that the ratings are generally higher for some subjects than for others, and that the differences in consistency affects the confidence of some subjects differently than others.

Hypothesis II states that the degree of certainty that an attribution is correct is a function of the consistency of the relationship between cause and effect. The significantly higher expressed confidence in the high consistency treatments supports the hypothesis.

CONCLUSIONS

The assumption of attribution theory, that attributions are made on the basis of covariation, has been supported by the study. It appears that both the attribution and the confidence associated with the attribution are functions of the consistency with which an effect is associated with a cause.

It should be noted that the experimental subjects were presented with the information data base in random order, and were questioned in random order, though they were permitted to arrange the information and questions in any manner they saw fit. During the administration, most could be observed to order the cards in groups of three, and in several cases the subjects actually arranged the cards in matrix form much as it appears in Table 1 for one condition. Their response appears to be very orderly and consistent, even though any such order was entirely their own. They did not appear to have been importantly influenced by the first or last piece of information, or by any single item. Rather, they seem to have regarded at least one column or row of three items and in many cases the entire set of data.

TABLE 4

ANALYSIS OF VARIANCE: CONFIDENCE RATINGS, CONDITION I

TABLE 5

ANALYSIS OF VARIANCE: CONFIDENCE RATINGS, CONDITION II

Attribution a Learning Theory

Those familiar with stimulus-response models of learning will recognize the similarity between the association of an observable effect with a probable cause in attribution theory and the association of an act with a reward in learning theory. There are, of course, a variety of distinctions between the two types of theory; many of them fundamental and important. Of note here is the distinction in the role of each theory for the study of information processing.

Learning theory typically deals with the connection between a single stimulus, one particular response, and a single reward. It may not deny reward value inherent in a variety of other factors in the environment of the subject, however these factors are exogenous to the paradigm. In the complex and multidimensional world in which the consumer must operate, attribution theory offers a mode of connecting a variety of stimuli with a variety of causal factors, assuming that reward value is generalized and inherent in correct and stable attributions. The value of exploration of relatively simple configurations of stimulus-response can not be denied, but it also seems likely that comprehension of complex patterns in a verifiable model has a special value and significance. Attribution theory may constitute such a model.

Implications for Future Research

This study employed a subject population of college juniors and seniors who are relatively familiar with organization techniques and tasks similar to those required of them in this case. Certainly the external validity and generalizability of the experiment would be enhanced by a replication on the general population of adults. Aside from a different population, one might wish to enlarge the data base matrix to determine the degree of incorporation of information in the process and its effect on confidence.

The matrices employed in this experiment represented extreme cases, such that all of the variance was in either a row or a column. Before moving to larger numbers of either examples or dimensions, it seems likely that the most appropriate refinement would be inclusion of a mixed data matrix such that some variance was associated with each entity class. This design would permit assessment of the role of the entire matrix in influencing the confidence in the attribution, as opposed to only single row or column consistency. Thus, it may be possible to measure more accurately the amount of information used by the individual in making and holding an attribution.

Implications for Promotion

If the consumer makes attributions on the basis of covariation, as the theory suggests and the study indicates, the promotion manager may find it useful to tailor the information and communications directed to the consumer to fit the consumer's information processing methods. For example, if all product features and characteristics are lauded with equal intensity, the consumer may be prone to attribute the message to the marketer's desire to sell the product. On the other hand, if the marketer were to single out one or a few particularly strong product features, while recognizing some shortcomings in other areas, the consumer may attribute the effect (the favorable score on the strong feature) to the product and its actual performance.

Similarly, if a TV personality praises every product and brand sponsored, the consumer may conclude that this effect can be attributed to the endorser's paycheck. If, however, the endorser is sometimes critical of a product even though the firm is a sponsor, the consumer would be likely to see the praise which does come forth as varying with the brand or product itself, and therefore attributable to that cause.

These are but two examples of the applicability of attribution theory to marketing information management. The underlying assumption in the study of the consumer information process, and that of the study of all consumer behavior, is the overall value of understanding. It is assumed that marketing, no less than any other profession, will ultimately benefit by a more complete understanding of those being served.

REFERENCES

Heider, F. T psychology of interpersonal relations. New York: Wiley, 1958.

Kelley, H. H. Attribution theory in social psychology. In D. Levine (Ed.), Nebraska symposium on motivation. Lincoln: University of Nebraska Press, 1967.

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