An Investigation Int0 the Differential Effects of Causally Simple and Complex Attributions

Richard Mizerski, Arizona State University
ABSTRACT - This study examines a proposed measure of consumer causal attribution that is based upon the stochastic assumptions of Information Theory. Given the same product information, significant differences were found between causally simple and complex subjects in their attributions, beliefs, and affect toward the stimulus products.
[ to cite ]:
Richard Mizerski (1976) ,"An Investigation Int0 the Differential Effects of Causally Simple and Complex Attributions", in NA - Advances in Consumer Research Volume 03, eds. Beverlee B. Anderson, Cincinnati, OH : Association for Consumer Research, Pages: 176-183.

Advances in Consumer Research Volume 3, 1976      Pages 176-183

AN INVESTIGATION INT0 THE DIFFERENTIAL EFFECTS OF CAUSALLY SIMPLE AND COMPLEX ATTRIBUTIONS

Richard Mizerski, Arizona State University

ABSTRACT -

This study examines a proposed measure of consumer causal attribution that is based upon the stochastic assumptions of Information Theory. Given the same product information, significant differences were found between causally simple and complex subjects in their attributions, beliefs, and affect toward the stimulus products.

ATTRIBUTION IN INFORMATION PROCESSING

The concept of causal attribution, and the study of its impact on the interpretation of information, is a relatively new and expanding area of study by social psychologists. Since information-processing is a critical area in many phases of marketing, it was only natural that researchers would investigate the feasibility of applying these same concepts to help explain consumer processing of product or service information in the marketplace. The few marketing studies that have been conducted (e.g., Mizerski, 1974a; Robertson and Rossiter, 1974; Settle, 1973; Settle and Golden, 1974) show that the concepts can be generalized to marketing, and may provide new perspectives for understanding the consumer. Nonetheless, research in both marketing and social psychology has been severely limited in examining the individual's total attribution process. The first problem is that the causal attributions in these studies are seldom measured directly. Instead, they are inferred using such measures as confidence and credibility. This procedure follows from Jones and Davis (1965), who suggested an isomorphic relationship between confidence and extremity of the attribution. However, research by Newtson (1973) and Mizerski (1974b) seriously questions the validity of this approach. Second, present attribution measures do not allow for an examination of the number of perceived causes, nor the allocation of attribution among each cause. Since consumers generally feel that information from or about the marketplace is the result of a mix of internal and various external causes (Mizerski, 1974a; 1974b), improvements in the elicitation and measurement of this mix should provide for more precise tests and applications of attribution principles. Toward that end, this paper presents a new measure of causal attribution, causal complexity, that is based upon the stochastic assumptions of Information Theory.

ATTRIBUTION AND THE CONCEPT OF CAUSAL COMPLEXITY

Attribution Theory studies the processes whereby people make causal explanations about the information they receive. Then, based upon the causal attribution chosen, it attempts to predict how those individuals will make inferences about their environment. In terms of processing information in the marketplace, consumers can contribute product information to either an internal, external, or a mix of both types of causes. For example, suppose that a consumer received information that a particular brand or model of automobile delivered good gas mileage. If the receiver believed that the information was the result ("is caused by") the information source actually getting good gas economy and then telling the receiver about it, he would attribute the information to an internal cause--a cause internal to the stimulus object. In many cases, however, the consumer may have some doubt about whether the information was "caused" by actual product performance. The receiver may suspect that other causes such as the source's conflict of interest (e.g., the source is selling his own automobile), or some other individual bias (e.g., the source always prefers to speak favorably about the brand) prompted the information. In these two latter situations, the receiver attributes the information to "external causes" and generally finds the information much less suitable for making inferences about the product.

However, the degree to which a consumer makes an internal attribution is not necessarily the only factor of interest. Consumer perception of what is suitable/useful information is largely determined ". . . by the configuration of factors that are plausible causes for that information" (Kelley, 1973, p. 108). In other words, the type and mix of external causes the consumer believes exists can also tell us a great deal about how he or she will interpret product information.

Unfortunately, present scaling techniques can make only very simple distinctions between causal attributions. As an example of this limitation, suppose there are four plausible reasons (causes) for a source to transmit product information (Figure 1). Cause #1 is the internal attribution (the product itself "caused" the information transmission), while causes 2 through 5 are alternate, non-product causal attributions (e.g., source bias, peer pressure to say something favorable or unfavorable about the brand). Note that all three hypothetical receivers allocate 20 percent of the cause to the product (internal cause), with the remainder of the allocation going to various external causes. Present techniques will examine only part of each receivers' attribution domain by measuring the degree of internal attribution (A = 20 percent, B = 20 percent, and C=20 percent) or evaluate attributions made between an internal and a surrogate for all external causes (A = 20 percent internal, 80 percent external; B = 20 percent internal, 80 percent external; and C=20 percent internal, 80 percent external). While this approach may be effective for some types of research, it tends to view the causal attribution of receivers A, B and C as identical. It would seem, however, that there are very different attributional processes involved, and that the "complexity" of each subject's causal allocation (i.e., in terms of number dimensions or causes the attributors employ) may be an important factor that affects how he or she will react to product information (Kelley, 1972).

FIGURE 1

HYPOTHETICAL CAUSAL ALLOCATION OF PRODUCT INFORMATION

Figure 2 provides another example of the attribution of information by a number of receivers. Receiver A attributes the information to only one cause while receivers B and C both perceive four plausible causes for the same effect (the product information). Given information about (1) the number of perceived causes; (2) the allocation of causality; and (3) how a specific cause affects the interpretation of an effect, a marketer could predict receiver A's interpretation and use of the data with more certainty than for either receiver B or C. While receiver C sees more plausible causes than A, he attributes the great bulk of causation to the same factor (#1), and should form very similar product perceptions with the information. Therefore, the marketer would be only slightly less certain about receiver C's information processing. Receiver B, however, feels that each of the causes are equally responsible for the information. This causal allocation provides, in effect, equally weighted explanations about how the consumer will react, and makes the marketer least certain about predicting B's resultant interpretation and use of the product information. It should be noted that this uncertainty is also operating within the attributor himself. The receiver finds situations where many causes are equally plausible more difficult to interpret than when an effect can be attributed to one or a few causes (Jones and Davis, 1965). Therefore, the degree of uncertainty for both the attributor and the marketing observer is a direct function of the dispersion of attribution among the set of plausible causes.

FIGURE 2

HYPOTHETICAL CAUSAL ALLOCATION OF PRODUCT INFORMATION

MEASURING CAUSAL COMPLEXITY

How may one get an objective overall measure of an individual's causal complexity? From the example in Figure 2, it should be apparent that both the number of perceived causes as well as the allocation of cause must be included, since neither completely describes the tm-certainty in a receiver's causal array. Fortunately, the question of how to gauge the complexity of an array of elements has already presented itself in such diverse contexts as psychology, statistics, and communications, and has spurred development of an index of uncertainty called "H" or entropy. This technique was initially proposed by Shannon and Weaver (1949) to judge the amount of uncertainty [To the information theorists ". . .uncertainty and information are interchangeable commodities . . . (with) the amount of information in an event exactly equal to the amount of uncertainty residing in that event before its occurrence, and this uncertainty, in turn, is a direct function of the number of possible events that could have occurred" (Bieri, et. al., 1966, p. 51).] in an array of symbols by measuring the rectilinearity of their distribution. If one considers a subject's decisions on the number and allocations of cause to be a causal array the application of H to measure causal complexity is rather straight-forward. Array complexity is calculated by accounting for the number of distinctions made about a domain in terms of the grouping system used by the subject. The following is the formula for H:

EQUATION

where:

H is the individual subject's causal complexity score;

Pi is the allocation to cause i; and

K is the number of perceived causes.

Applying this formula to the attributions in Figure 2 would designate receiver A as the most causally simple (least uncertainty) and receiver B as the most complex (most uncertainty). The complexity continuum would read, from causally simple to complex; A, C, D, E, and B. Complexity, and thus the highest H score, is greatest with equal allocation of causal attribution among the largest number of perceived causes; while the least complexity would be reflected in the tendency to attribute to one cause.

SUGGESTED DIFFERENCES DUE TO CAUSAL COMPLEXITY

Given a measure of causal complexity, previous research suggests that a number of differences should exist between causally simple and complex individuals. [This categorization should not be interpreted as a personality variable or anything beyond a situation specific attribution decision at this point in the constructs development. Future research must establish the extent of the generalizability across product and information processing situations.] First, the more complex the causal domain, the less confident the individual should be in his or her attribution about the information. This concept forms the basis for the proposed measure, that individuals are less certain/confident in a "broad distribution" (causal allocation) than they are in a "sharply peaked one" (Jaynes, 1957; Jones and David, 1965). Second, given credible product information, causally simple subjects should tend to be more extreme in their causal attributions. [The type of attribution (internal versus external) depends on a number of consumer and situational factors (e.g., the source of the information).] This follows from the "discounting principle" that suggests "the role of a given cause in producing a given effect is discounted if other plausible causes are also present" (Kelley, 1973, p. 113). Without information to the contrary, however, the internal attribution tends to be expected (that the information about the product was caused by the product). Therefore, a large number of external attributions should discount (in the consumer's mind) the possibility of an internal cause.

Third, given the same product information, causally simple individuals should form more extreme beliefs about the product. In an approach consistent with other researchers, Fishbein (1965, p. 107) has stated that "Y a belief about an object may be defined as the probability or improbability that a particular relationship exists between the object of belief and some other object, concept, value, or goal." This definition of a belief appears to be one result of an attribution (Ajzen and Fishbein, 1975; Mizerski, 1974a). Kelley (1973, p. 107) notes that attribution is a part of the process ".. . by which man 'knows' his world, (and) has a sense that his beliefs and judgments are veridical." It would seem to follow that if an individual made a strong attribution about information concerning product characteristics to an internal cause, this would manifest itself in a belief that a relationship existed between the product and those characteristics. In Fishbein terminology, the internal attribution attaches a higher probability that the information (e.g., product attribute ratings) was related to the product. The stronger the internal attribution, the more extreme the belief. If causal complexity follows the discounting principle, causally simple individuals would make stronger internal attributions that should result in more extreme beliefs. If the simple subjects have a greater degree of confidence in their attributions (the first proposed difference), that should further augment a differential in belief strength.

Finally, causally simple and complex subjects should differ in terms of the extremity of their attitude toward a product, given the same stimulus information. A substantial amount of research on attitude models (see Bass and Wilkie, 1973, for a review) suggests that beliefs, defined in the Fishbein sense, are one of the major components in forming an individual's attitude toward a product. One suggested algebraic relationship between the two (Fishbein, 1965, p. 117) is the following:

EQUATION

where:

A = measure of affect;

Bi = the strength of belief i about the attitude of object o, that is the probability or improbability that o is related to some other object xi;

ai = the evaluative aspect of Bi, that is, the evaluation of x.--its goodness or badness;

N = the number of beliefs.

While other expectancy-value models differ somewhat from Fishbein's, each uses some measure of instrumentality or belief strength as a basis for predicting an attitude. If causally simple subjects form more ex-creme beliefs, the differential in extremity should manifest itself in a more extreme affect toward the product.

HYPOTHESIS

In order to test for the suggested difference between causally simple and complex individuals, the following hypotheses were developed:

Given the same stimulus product information;

1. Causally simple subjects will tend to be more confident in their causal allocations.

2. Causally simple subjects will tend to be more internal in their causal attributions.

3. Causally simple subjects will tend to form more extreme beliefs about a product.

4. Causally simple subjects will tend to produce more extreme affect toward the product.

RESEARCH METHODOLOGY

Data Collection

An experiment was performed on 300 upper division undergraduate and graduate students at the University of Florida College of Business Administration. [For the analyses, 30 subjects were deleted for missing data and for improper use of a scale.] The subjects were told that the experiment involved testing how people evaluate, and are affected by, the opinions of others. They then received evaluative information on three salient attributes (chosen by the Fishbein and Raven, 1962, procedure) of either a fictitious automobile or motion picture. Attributes used were maintenance costs, comfort, and gas mileage for the automobile; and acting, plot, and photography for the movie. The information was presented in the form of personal ratings that were supposedly made by another student, who was randomly chosen to test and evaluate the product. Two of the three attributes were given a neutral rating, with only one attribute rated either favorably or unfavorably. [Two information treatments were used in order to conduct another experiment that evaluated the disproportionate influence of unfavorable information (Mizerski, 1974a; 1974b).] The modifiers used for rating were scaled for equal polarity and opposite affective meaning by the methods used in Myers and Warner (1968). Subjects were randomly assigned to the attribute (3 levels) x information (2 levels) x product (2 levels) treatments.

The Measure of Causal Complexity

Following the information treatment, the subjects were asked to allocate the cause for the bogus "rater's" opinion (the information treatment) with the following question concerning attribution complexity:

How much do you feel that each of the following contributed to the student's opinion about gas mileage? (note: any of the following could account for 0% to 100% of the opinion)

1. The automobile itself.........__%

2. The influence of other peoples' opinions.........__%

3. An effort to please or antagonize the viewer.........__%

4. The personality of the tester (natural tendencies to be critical or complimentary).........__%

5. A general bias for or against automobiles or the brand.........__%

6. Other reasons (if any)? __________

________________________________________TOTAL  100%

The external causes (numbers 2 through 6) were determined in two pretest group sessions with subjects similar to those in the main experiment. Each subject's causal complexity score was produced by applying the H transformation to their responses about causal allocation. The respondents were then ordered by complexity, with analyses performed on subjects in the upper and lower quartiles (most and least causally complex). A seven point scale ranging from "no confidence" (#1) to "complete confidence" (#7), recorded the subjects' confidence in the causal allocation for each treated attribute.

The Measure of Internal Attribution

A second question was developed to establish an independent measure of the subject's degree of attribution to an internal cause. [See Mizerski (1974b) for a validation and other uses for this scale.]

To what extent do you feel that other reasons--reasons having nothing to do with the automobile tested--influenced the student's opinion about...

                  other reasons had no                                            other reasons were the

                  effect on the opinion                                             only cause for the opinion

gas mileage             1               2               3              4              5               6              7

The smaller the scale value chosen, the more internal ("caused" by the product only) the attribution.

The Measure of Belief

A third question elicited the subjects' belief about the relationship between the treated attribute and the product, and randomly appeared either before or after the two attribution questions (the order of the attribution questions were also randomized). The format for the belief scale follows that developed by Fishbein and Raven (1962), and has been used in a number of marketing studies (e.g., Mazis and Klippel, 1973; and Mizerski, 1974a).

How likely is it that this automobile....

                                             unlikely                                                                              likely

has low gas consumption    -4         -3        -2         -1        0         +1        +2         +3        +4

The Measure of Attitude

Finally, subjects were asked a question about their overall affect toward the stimulus product.

How much would the automobile tested appeal to you?

Extremely Low    Low              Mildly Low                          Mildly High           High        Extremely High

Appeal                  Appeal        Appeal             Neutral        Appeal                  Appeal      Appeal 

1                         2              3                 4              5                     6             7

RESULTS

Mean scores for the measure of the subjects' confidence in their attribution allocation are presented in Table 1. As hypothesized, causally simple subjects, those who perceived relatively few causes- for the stimulus information, were more confident (XS= 5.16) than complex subjects (XC = 4.54) in their causal allocation.

TABLE 1

MEAN CONFIDENCE SCORES

In order to detect if these differences were significant, an unweighted means analysis of variance (ANOVA) was performed on the data (Table 2). The ANOVA revealed a main effect of complexity (F = 6.73, p < .01) which supports the first hypothesis.

TABLE 2

ANALYSIS OF VARIANCE: CONFIDENCE IN THE CAUSAL ALLOCATION

The second hypothesis, that causally simple subjects would be more internal in their attribution, was also supported. Mean scores for the independent measure of internal attribution are presented in Table 3.

TABLE 3

MEAN INTERNAL ATTRIBUTION SCORES

Since a low mean score reflects an internal attribution, the difference between causally simple and complex subjects are in the predicted direction. An ANOVA on the mean scores (Table 4) reveals a main effect of causal complexity (F = 8.55, p < .01), and an interaction of product and information (F = 3.79, p < .05).

TABLE 4

ANALYSIS OF VARIANCE: INTERNAL CAUSE ATTRIBUTION

The main effect of causal complexity shows that the internal attribution of the causally simple subjects were significantly larger than those of the complex group, and thus supports the second hypothesis. The interaction of product and information was independent of the subject's complexity and does not affect the hypothesis (see Mizerski, 1974b, for a discussion of this interaction).

Analyses of hypothesis three, that causally simple and complex subjects would differ in terms of their extremity of belief given the same stimulus information, required some initial data transformation. Since differences in absolute belief strength were of interest, the sign of the beliefs formed in the unfavorable information treatment (usually 0 to -4 on the -4 to +4 scale) were reflected. [The term "reflected" refers to the procedure of changing the sign of the value (e.g., -4 becomes +4). The method used in this study was to multiply the belief score of subjects in the unfavorable information treatment by a B1.] The mean absolute belief scores for all subjects are presented in Table 5. As predicted,

TABLE 5

MEAN BELIEF SCORES

the overall mean belief score for the simple subjects (XS = 2.81) was more extreme than the causally complex group's (XC= 2.34) beliefs. The high and low complexity groups also differed in terms of the extremity of their beliefs formed about the treated automobile attribute, with simple subjects forming much stronger beliefs (XS = 3.19) than the complex individuals (XC = 2.25). Applying an ANOVA to these scores (Table 6 reveals both a significant main effect of causal complexity (F = 4.63, p < .05) and an interaction of the subject's complexity and the product. In order to detect the source of the complexity x product interactions, a Newman-Keuls Multiple range test was performed (Table 7). The results substantiate that there were significant differences between the simple and complex subjects' beliefs about the treated automobile attribute. This supports the hypothesis that causally simple subjects will form more extreme beliefs, and further notes the significant effect that the nature of the product can have on the attribution/belief formation process.

TABLE 6

ANALYSIS OF VARIANCE: BELIEF SCORES

TABLE 7

RESULTS OF A NEWMAN-KEULS MULTIPLE RANGE TEST

The final hypothesis predicted that, given the same information, causally simple subjects would form more extreme attitudes or affect toward the product. This hypothesis was based largely upon the suggested existence of differences between the two causal styles in attribution confidence, perceived internal causation, and extremity of beliefs. Since these suggested differences did occur, support for the last hypothesis was anticipated.

Since approximately half of the subjects received unfavorable information about one or the other of the products, the original affect scale (#1: extremely low appeal to #7: extremely high appeal) could not be used to test for differences in extremity. Therefore, each score was transformed to a -3/+3 scale value, and affect ratings of those subjects who received unfavorable information were reflected. Mean affect scores are reported in Table 8.

TABLE 8

MEAN AFFECT SCORES

As hypothesized, causally simple subjects tended to form a more extreme attitude toward both products (XS = 0.91, XC = 0.44). Analysis of variance on affect scores (Table 9) show main effects for both causal complexity (F = 5.24, p < .025) and the information treatment (F = 10.09, p < .01). While the latter results do not constitute a test of the hypothesis, they provide additional evidence that unfavorable information is more influential, and that this disproportionate influence seems to be linked to a difference in the attribution process (for an extensive discussion of this phenomenon, see Mizerski, 1974a and 1974b). The significant differences in affect extremity, with causally simple subjects forming more extreme attitudes, supports the last hypothesis.

TABLE 9

ANALYSIS OF VARIANCE: EXTREMITY OF ATTITUDE/AFFECT

DISCUSSION

Results of the experiment support the validity of causal complexity, as well as suggest a number of potentially important cognitive differences between the two causal types. Causally simple subjects tended to be more confident in their attributions, and more apt to attribute the information to an internal cause (the product itself). They also formed more extreme beliefs and attitudes about the stimulus product; the first time attribution has been empirically linked to these latter two measures. It should also be noted that none of the results, except for the product x information interaction for beliefs, were significantly affected by the product on which information was presented. While future research should gauge the extent to which these results can be generalized, the two products used represented very different degrees of social visibility and desirability (in many ways representing a good versus a service), and suggests that the subjects' causal complexity may be rather insensitive to the specific product being evaluated.

A particularly difficult question is establishing where causal complexity fits into the sequence of consumer information-processing, since a number of relationships are plausible. Causal complexity may operate as a personality trait, remaining consistent across many product classes. This relationship required that the subjects complexity scores would have to be largely independent of the product information. That appears doubtful in this experiment since the subjects' complexity scores were significantly affected by one variable in the experimental treatment. An ANOVA on these scores (Table 10) shows a main effect of product attribute (F = 8.87, p < .001); yet it is interesting to note that neither the product nor the information treatment (favorable or unfavorable product information) prompted significant differences in subjects' complexity.

TABLE 10

ANALYSIS OF VARIANCE: CAUSAL COMPLEXITY

Mizerski (1974b) found similar main effects of product attribute related to differences in subjects' attribution of internal causation about product information. It was suggested that the different degree of objectivity of judgment between the attributes caused this effect. In other words, the more an attribute lends itself to a subjective evaluation (e.g., styling vs. gas mileage), the greater the possibility that the subjects will perceive external causes for that information. In terms of this study, the subjectivity of an attribute's evaluation may trigger the subjects' causal complexity, seemingly a very situation specific reaction.

If causal complexity is situational, it could act as a mediating or contributing factor in the determination of the consumer's beliefs and attitude toward a product. Complexity's differential response (compared to the belief and affect scores) to the experimental treatments, however, tends to cast doubt that it operates as the sole criterion. Rather, causal complexity may be formed by the information, then the resultant complexity, situational variables, and the information itself, all prompt relevant beliefs and attitudes.

Of course, another alternative explanation is that the product information prompts the subjects to make causal attribution, which in turn establishes their causal complexity. While the attribution literature suggests that causal perception lends to attribution, this explanation cannot be ruled out at this time (if ever). The determination of causal complexity's place in consumer information-processing will require a measure of complexity independent of the treatments.

Future research could also examine the generalizability of this study's results by introducing such factors as the subjects' interest and experience with the product and the impact of various sources of information (e.g., consumer groups, trade associations, government agencies) on the complexity of the attribution. It should also be noted that the size of the subjects' causal domain was in some ways fixed in this study. It would be interesting to see if the use of a relative rather than an absolute measure of complexity would provide a better measure, outweighing the operational problems involved. [The H formula can be made a relative measure of causal complexity (R) by the following modifications. R = H/Hmax  Where Hmax is a calculation for each subject when subjects perceive different numbers of plausible causal categories. R is calculated by dividing the subject's H by the maximum H possible with that number of categories.

By using a relative, rather than an absolute measure of causal complexity, the experimenter could elicit the subject's causal domain with a free-response questioning technique. This would overcome the problem of possible confining some subjects as to the size of their relevant causal choices. The trade-off is that difficult, if not impossible, subjective evaluations must be made determining if each subject's causal categories are independent.]

The data also suggest plausible ties with another area of behavior, cognitive complexity. While the two constructs theoretically represent different psychological processes, they show similar results in a number of areas (see Streufert, 1972, for a review). Both causally and cognitively simple individuals tend to perceive more internal attributions in certain situations (Harvey and Ware, 1967), as well as form more extreme affect toward a stimulus object or person (Campbell, 1960). Other aspects, such as a differential in the type and extent of information search the two cognitive styles exhibit, may also be applicable to causal complexity. The cognitive complexity literature is rich in empirical research and could prove helpful in further explaining and applying the concept of causal complexity.

The study's results suggest several possible marketing applications. The measure may enable better prediction of consumer response to product information. Kelley (1972, 1973) has developed a framework called a causal schema which is a conceptual representation of how two or more causal factors (e.g., internal vs. external) interact in relation to a particular type of effect (e.g., transmitted product information). He suggests that individuals learn through experience to link effects with possible causes, thus building a ". . . repertoire of abstract ideas (schemata) about the interaction of causal factors" (Kelley, 1972, p. 152). These schemata provide the individual with an economical and fast attribution framework for the fitting of partial information so that reasonably good causal inferences can be drawn. If the marketer knows that a specific schema represents a particular causal problem, he can then make a reasonably good prediction as to how the consumer will interpret the information provided. Although Kelley has developed rather elaborate and complex schemata that integrate many types of causal attributions, the necessary behavioral information required to test and implement these constructs has not been available. Thus, present applications have been limited to representing rather simple attribution decisions (main effects, not causal interactions) which severely limits predictability in more complex market situations. Kelley notes that an important unanswered question concerning the ".. . complexity, in terms of the dimensions or causes . . . attributors can and do employ in their schemata." Without this knowledge we are ". . . uncertain about the kind of causal schema being used" (Kelley, 1972, p. 156). Causal complexity is a more inclusive measure of the individual's causal domain, and may provide the necessary information about the consumer's complexity, and ultimately the information-processing strategy he or she is using.

REFERENCES

Icek Ajzen, and Martin Fishbein, "A Bayesian Analysis of Attribution Processes," Psychological Bulletin, 2 (1975), 261-277.

Frank M. Bass and William L. Wilkie, "A Comparative Analysis of Attitudinal Predictions of Brand Preference" Journal of Marketing Research, 10 (1973), 262-269.

S. Bieri, A. L. Atkins, S. Briar, R. L. Leaman, H. Miller and T. Tripodi, Clinical and Social Judgment (New York: Wiley, 1966).

V. N. Campbell, Assumed Similarity, Perceived Sociometric Balance and Social Influence, unpublished doctoral dissertation, University of Colorado, 1960.

Martin Fishbein, "A Consideration of Beliefs, Attitudes, and Their Relationship," in Ivan D. Steiner and Martin Fishbein, eds., Current Studies in Social Psychology/, (New York: Holt, Rinehart, and Winston, Inc., 1965).

Martin Fishbein and Bertram H. Raven, "The AB Scales: An Operational Definition of Belief and Attitude," in Fishbein, ed., Readings in Attitude Theory and Measurement (New York: Wiley, 1962).

O. J. Harvey and R. Ware, "Personality Differences in Dissonance Resolution," Journal of Personality and Social Psychology, 2 (1967), 227-230.

E. T. Jaynes, "Information Theory and Statistical Mechanics,'' Psychology Review, 106, 620, (1957).

E. E. Jones and K. E. Davis, "From Acts to Dispositions: The Attribution Process in Person Perception," in L. Berkowitz, ed., Advances in Experimental Social Psychology, Vol. 2 (New York: Academic Press, 1965).

Harold H. Kelley, "Causal Schemata and the Attribution Process," in Edward E. Jones, et. al., eds., Attribution: Perceiving the Causes of Behavior (Morristown, N.J.: General Learning Press, 1972).

Harold H. Kelley, "The Processes of Causal Attributions" American Psychologist, 28 (1973), 107-128.

Michael Mazis and Eugene Klippel, "A Comparison of Four Expectancy-Value Formulations in the Prediction of Consumer Attitudes," unpublished paper, University of Florida, 1973.

Richard Mizerski, "Attribution Theory and Consumer Processing of Unfavorable Information about Products," unpublished doctoral dissertation, University of Florida, 1974a.

Richard Mizerski, "A Test of the Relationship Between Trait and Causal Attribution," in J. Schlinger, ed., Advances in Consumer Behavior, Vol. 2, Proceedings of the Association Consumer Research, 1974b.

James H. Myers and Gregory Warner, "Semantic Properties of Selected Evaluation Adjectives," Journal of Marketing Research, 5 (1968), 409-412.

Darren Newtson, "Attribution and the Unit of Perception of Ongoing Behavior," Journal of Personality and Social Psychology, 28 (1973), 28-38.

Thomas S. Robertson and John R. Rossiter, "Children and Commercial Persuasion: An Attribution Theory Analysis," The Journal of Consumer Research, 1 (1974), 13-20.

R. B. Settle, "Attribution Theory and Consumer Behavior,'' in H. H. Kassarjian and Thomas S. Robertson, eds., Perspectives in Consumer Behavior (Glenview, Illinois: Scott, Foresman and Co., 1973).

R. B. Settle and Linda Golden, "Attribution Theory and Advertiser Credibility," Journal of Marketing Research, 9 (1974), 181-185.

C. Shannon and W. Weaver, The Mathematical Theory of Communication (Evanston: University of Illinois Press, 1949).

Susan C. Streufert, "Cognitive Complexity: A Review," Technical Report NO. 2, Office of Naval Research Organization Effectiveness Research Programs, 1972.

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