The Role of Specific-Item Causal Dispersion in Attributional Focus and Confidence Determination

ABSTRACT - Two of the most important determinations in an attribution research setting are those of the attributional focus and attributional confidence. However, for two different reasons, assessing these two constructs has presented problems to researchers. Problems regarding assessment of the former involve both conceptual and operational problems, while problems involving assessment of the latter involve mostly operational problems In this paper a measure of causal dispersion advanced, and empirically demonstrated, as a measure capable of assessing both constructs. In doing so, the distinction between attributions and the attributional focus is highlighted. Finally, implications for attribution researchers are discussed.

Citation:

Donald R. Lichtenstein (1988) ,"The Role of Specific-Item Causal Dispersion in Attributional Focus and Confidence Determination", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 89-95.

Advances in Consumer Research Volume 15, 1988      Pages 89-95

THE ROLE OF SPECIFIC-ITEM CAUSAL DISPERSION IN ATTRIBUTIONAL FOCUS AND CONFIDENCE DETERMINATION

Donald R. Lichtenstein, Louisiana State University

ABSTRACT -

Two of the most important determinations in an attribution research setting are those of the attributional focus and attributional confidence. However, for two different reasons, assessing these two constructs has presented problems to researchers. Problems regarding assessment of the former involve both conceptual and operational problems, while problems involving assessment of the latter involve mostly operational problems In this paper a measure of causal dispersion advanced, and empirically demonstrated, as a measure capable of assessing both constructs. In doing so, the distinction between attributions and the attributional focus is highlighted. Finally, implications for attribution researchers are discussed.

INTRODUCTION

Beyond these matters of social and self-perception, attribution theory is related to a more general field that might be called psychological epistemology This has to do with the processes by which man "knows" his world and, more importantly, knows that he knows, that is, has a sense that his beliefs and judgments are veridical (Kelley 1973, p 107).

Attribution theory deals with the processes by which individuals assign causality to events. When an individual makes an attribution for some effect, s/he is implicitly saying "I know what caused the effect." However, because individuals may make attributions with varying degrees of confidence, even more important than the attribution itself may be that the attributor "knows that s/he knows." Therefore, from an attributional research perspective, two important issues are the measurement of what the perceiver "knows" to be the causal entity and how much s/he "knows that s/he knows." The former issue translates into one of determining the attributional focus, while the latter translates into one of determining attributional confidence.

STUDY PURPOSE

The first purpose of the present paper is to propose an attribution scale-based derivative measure for determining both the attributional focus and attributional confidence in attribution research The proposed measure for measuring both cognitions is a measure of causal dispersion. Causal dispersion can be defined as the variance in perceived causal probability across a given set of possible attributions for an event. As such, the measure is very closely related to Mizerski's (1978) causal complexity measure, but may be better suited for some attribution research studies. The second purpose of the present research is to investigate if a measure of causal dispersion derived from nonabstract "specific-item" attribution scales (e ,., "the ear's poor gas milage is the reason it is being discounted"), performs differentially than a causal dispersion measure derived from more abstract"general-item' attribution scales (e.g, "something about the car in general is the reason it is being discounted")

However, first discussions regarding the importance of attribution focus and confidence determination are provided. Then a measure of causal dispersion is proposed, followed by theoretical predictions between causal dispersion and other constructs. Finally, a measure of causal dispersion derived from specific-item attribution scales, in addition to the one derived from general-item attribution scales, are employed in an attempt to measure both the attributional focus and attributional confidence The validity of the two measures is assessed in accord with each one's consistency with theoretically-based predictions.

Attributional Focus Determination

The attributional focus is the "object" of the attribution process, that is, the entity about which causal beliefs will be formed (Mizerski, Golden, and Kernan 1979). For example, if an individual were to assess the probability as .6 that an advertised discount on an automobile were due to one or more product-related causes (e.g, "because the automobile has poor styling," and/or "because the automobile is a lemon," etc.), and .4 that the discount was due to one or more person-related causes (e g, "because the dealership wants to create the image of a discounter," and/or "because the salesman wants to make quota," etc ), the product would represent the attributional focus because it is the entity which occupies the majority of the perceiver's causal allocation.

In many consumer behavior situations, determination of the attributional focus represents a difficult task. As depicted in the example above, because many consumer behavior situations involve an interaction of perceptual areas (e g, a salesperson (person) making a claim about a product (object)), many times the attributional focus is not clear For example, a consumer may make either of the two product-focused attributions and/or either of the two object-focused attributions mentioned above to account for a price discount Or, a consumer may attribute the discount claim to neither person or automobile, but rather to some particular circumstance, e g., "overshipment from the manufacturer." Or, finally, consistent with Mizerski's (1978, p 221) contention that consumers generally feel that information from or about the marketplace is the result of a mix of stimulus, person, and circumstance causes, the consumer may make a combination of these attributions. Unfortunately, attribution theory docs not provide clear rules for determining the attributional focus (Mizerski et al. 1979) Although framing attributional questions for subjects in an experimental setting may increase the likelihood that the problem will be perceived in the manner framed, it is quite possible that subjects may respond to a person-perception-framed question (e.g, why did the salesman make that claim?) with object-focused attributions, and likewise, to an object-perception-framed framed question (e.g, why is this automobile being discounted?) with person-focused attributions.

For at least two reasons, attributional focus determination is important to attribution researchers. First, and most obvious, a perceiver's attributional focus represents his/her perception of the causal entity of the observed effect Thus, in consumer behavior situations that reflect an interaction of perceptual areas, if the researcher is to determine the perceived causal entity for an effect, it is imperative to determine the attributional focus, because the two are one and the same Second, from a theory application perspective, it is of critical importance to consumer attribution research to identify the attributional focus that consumers use because misspecification will lead researchers into focusing causation on the wrong entity, and concomitantly, into choosing an inappropriate theoretical paradigm (Mizerski et al. 1979, p. 131). For example, if an attributor focuses predominantly on the offering merchant in his/her search for attributions to account for a price discount, the perceiver would be viewing the situation as one of person-perception, and thus, the attributional focus would be the merchant From a research perspective, theories of person-perception capable of addressing questions of merchant disposition (e.g., Correspondent Inference Theory, Jones and Davis 1965), would become more applicable as the causal probability assigned to the merchant increased. Likewise, if the attributor focused predominantly on attributes of the automobile which may have prompted such a price discount claim, the attributor would be perceiving the situation as one of object-perception, and thus, the attributional focus would be the automobile From a research perspective, theories which address merchant disposition become less appropriate as the causal probability assigned to the object increases.

From a theoretical perspective, the Mizerski et al. (1979) distinction between attributions and attributional focus is consistent with Calder's (1977a,b) contention that everyday attributions by individuals are made at low levels of abstraction (e g, "because the car has poor styling"). It is of importance only to the attribution theorist to elevate these attributions to higher levels of abstraction, i e., to the level of the attributional focus (e.g., "because of something about the stimulus") for scientific explanation purposes. Despite this difference, however, it appears that several researchers have attempted to measure nonabstract attributions with measures of the more abstract, attributional focus. For example, many researchers have attempted to use measures such as "How likely is it that the effect was caused by something about the product?" to measure attributions (cf McArthur 1972; Jolibert and Peterson 1976; Major 1980; Spariman and Locander 1980). From a theoretical perspective, these measures are at a level of abstraction equivalent to that of the attributional focus. Thus, such abstract measures are not attribution measures, but rather, are measures of the attributional focus.

On an empirical plane, however, Lichtenstein and Bearden (1986) provided evidence which suggest that these measures even lack validity as measures of the attributional focus. In their study, Lichtenstein and Bearden measured product, person, and circumstance attributions by scales consisting of 10, 6, and 7 specific-item attribution measures, respectively They also measured these three attributions using the single, general-item measures Their results failed to provide evidence of convergent or discriminant validity between the two different measurement procedures. That is, respondents did not engage in the cognitive task of summing multiple nonabstract attributions to arrive at an attributional focus. Further, predictive validity was found between the specific-item attribution measures and theoretically related constructs, but not for the general-item measures. These results led Lichtenstein and Bearden to suggest that the same concepts are possibly not being measured when attribution measurement shifts from lower to higher levels of abstraction

Consequently, questions regarding attribution focus determination remain One way to do this would be to simply ask subjects for their perception as to the cause of some observed effect. The experimenter could then classify these attributions into attribution focus categories. However, while such free elicitation procedures have been successfully employed in measuring the existence of attribution processing (cf. Smith and Hunt 1978), Elig and Frieze (1979) found that when the attributions had to be classified into a taxonomy, attributions measured-via free elicitation lacked both validity and reliability when compared to structured methods of measurement (Elig and Frieze 1979). This is potentially due to the traditional coding and classification problems that accompany all open-ended questions (Peterson 1982) For example, is "because the dealership is in a cash flow bind" a person (e g, poor cash management) or circumstance (e g, unforseen expenses) cause? A researcher can use his/her inference in classifying attributions but the correct classification is the one used by respondents (Weiner 1985).

Elig and Frieze (1979) and Weiner (1985) have suggested, on grounds of both validity and reliability, a two-step approach to measuring attributions In the first step, a pretest sample is asked to provide possible attributions for some hypothetical event. In the second step, these attributions are converted to scale form. One potential problem, though, with this approach is the way in which the attribution problem is framed, i.e., either person- or object-perception, may affect attributions respondents provide. However, this procedure has the psychometric advantages of structured methods (Elig and Frieze 1979) Also, the attribution measures represent those of the respondent rather than those of the researcher (cf Weiner 1985, p 552) This method is also able o take advantage of factor analysis, which is one of the most common mathematical techniques used to analyze the responses of research participants to discover the underlying causal structure (Weiner 1985) Because factor analysis allows looking at the covariance between attributions as perceived by respondents, this method also allows for the classification of an attribution into a category as seen by respondents.

In addition to these attribution measurement advantages, and of particular relevance here, this method also allows for the determination of the attributional focus That is, by calculating a derivative measure of the specific-item attributions, evidence of the attributional focus can be obtained A measure of causal dispersion is proposed here as such a measure

Attributional Confidence Determination

Also of critical importance to attribution researchers is the ability to measure a respondent's attributional confidence This is evidenced by the fact that better than one-third of the research efforts cited in Mizerski et at's review article (1979) employed explicit or surrogate measures of attributional confidence It has also been suggested that there may be a relationship between attributional confidence and actions resulting from attributions, such that the higher the level of attributional confidence, the more likely actions will follow the attributions (Mizerski et al. 1979).

While there seems to be some consensus regarding the importance of attributional confidence, its measurement has created some controversy. The most common method of measuring attributional confidence is to have subjects respond to "confident-not confident" scales regarding their attributions (cf. Calder and Burnkrant 1977). However, there appears to be some controversy regarding the possibility that these confidence measures are redundant with the attribution measures themselves, i.e., attributional confidence is reflected by the extremity of the attribution, and thus, they are equivalent and inseparable measures (cf. Mizerski 1975; Mizerski et al 1979). While controversy remains in this area, it appears plausible to suggest that a measure of attributional confidence that did not mirror the method used for measuring attribution extremity, i.e., scales, may provide a more unique measure of attribution confidence. A measure of causal dispersion is proposed here as such a measure.

THE RELATIONSHIP BETWEEN CAUSAL COMPLEXITY AND CAUSAL DISPERSION

Causal complexity refers to a perceiver's attribution allocation among various possible causes. Those perceivers who are the most causally complex are those that a) recognize more possible causes of an event, and b) weigh the potential causes equally with respect to causal probability (Mizerski 1978). While both criteria are integral in assessing an individual's causal complexity, in some research applications, a less precise but more practical measure may be advantageous. For example, Mizerski (1978) listed five possible attributions for an effect and also an "other" category in which subjects could supply their own attribution. Thus, there were six possible responses. Subjects were to respond to each by supplying the probability that the particular cause was responsible for the effect. The responses had to total to one hundred percent, i.e., the subject had to account for all possible causes for the effect.

From an attribution research perspective, there may be problems with this approach. First, it is very likely that the number of possible attributions for an observed effect in a given situation may be very large, i.e., many more than six As the list of possible attributions grows larger, the cognitive task asked of respondents of dividing one hundred percentage points becomes very difficult. Further, as the list of possible attributions grows larger, many of the attributions may be perceived as very similar by some respondents. Therefore, using a percentage approach may represent "double-counting" some attributions, thereby overstating their probability while understating the probability of other attributions For example, based on the suggestions of Elig and Frieze (1979) and Weiner (1985) on grounds of reliability and validity, Lichtenstein and Bearden (1986) conducted a pretest to elicit attributions to account for why a particular automobile dealership was offering a price discount on one of their automobiles. They obtained thirty-four attributions, some of which were similar In addition to the cognitive task problem that would have occurred if subjects had been asked to divide 100 percentage points between the thirty-four possible causes, because it is likely that for may respondents that at least two of the attributions were perceived as the same, or at least very similar, this procedure would likely have resulted in double-counting Instead, Lichtenstein and Bearden converted the attributions to probable-improbable scales for the main study to assess the probability of each as the causal factor

One might suggest that many of the thirty-four attributions might be similar enough that they could be combined at a higher level of abstraction, thereby limiting the number of possible attributions. This solution would address both the "double-counting" problem and also the problem of the large cognitive task asked of respondents. For example, instead of having both "because the car has poor styling," and "because the car is poor on economy," etc., one might have "because of something about the automobile itself" As a matter of fact, this is what Mizerski (1978) did. However, again, Lichtenstein and Bearden (1986) provided evidence that respondents cannot or do not engage in the cognitive task of summing specific-item attributions into abstract general-item attributions. Therefore, in addition to the causal probability assigned to the general-item measure being questionable, any derivative measure of the general-item attribution scales, e g., a general-item-based causal dispersion measure, must also be questioned

By sacrificing one of the two criteria used by Mizerski (1978), a measure correlated to the causal complexity measure employed by Mizerski can be obtained Causal dispersion is such a measure and is proposed here as more applicable for some attribution research. A measure of causal dispersion can be obtained by employing only the second criteria employed by Mizerski (1978). Thus, given the thirty-four possible attributions used by Lichtenstein and Bearden (1986), an individual's causal dispersion can be measured by the standard deviation of responses across the thirty-four probable-improbable scales. An individual of high causal dispersion (i e., a high standard deviation) would be an individual who responded "very probable" on some scales and "very improbable" on others. These individuals are able to discriminate between likely and unlikely causes for an effect. Those individuals who respond in a similar manner, with respect to probability, to all attributions, whether all high, low, or somewhere in between, would be of low causal dispersion, and thus cannot discriminate between likely and unlikely causes

CAUSAL DISPERSION AND ATTRIBUTIONAL FOCUS/CONFIDENCE

Two hypotheses developed by Mizerski (1978) were used as a basis for the present analysis. First, based on the discounting principle, which suggests the role of a given cause in producing a given effect is discounted if other plausible causes are also present (Kelley 1973), Mizerski (1978) hypothesized that causally simple subjects will tend to form stronger stimulus attributions (i.e., attribute more of the total causation to the product on which the information is given). Mizerski's rationale was that without information to the contrary, stimulus attributions tend to be expected. Therefore, a large number of person or circumstance attributions should discount (in the consumer's mind) the possibility of a stimulus cause This present author agrees with this statement when the stimulus represents the attributional focus. Mizerski et al. (1979) contends that the determinant of the attributional focus should be the "object" of the attributional process, that is, the entity about which causal beliefs will be formed Thus, if the person represents the attributional focus, then it stands to reason that the presence of stimulus or circumstance causes should discount the otherwise expected person cause. Although the attribution question in the present investigation was framed as one of person-perception, it is no guarantee that the person was perceived as the attributional focus. Therefore, in conjunction with the purpose of the present paper (i.e., determination rather than confirmation of the attributional focus), no hypothesis is offered here. Evidence of the attributional focus can be addressed by examining the relationship between causal dispersion and the allocation of causation between product, person, and circumstance attributions and between causal dispersion and correspondent inferences (since correspondent inferences are attributions of intent or disposition to the focal person).

Mizerski (1978) also hypothesized that causally simple subjects will tend to be more confident in their causal allocations based on the rationale that one is more confident in a sharply peaked distribution than a broad one. Based on the same rationale in conjunction with the rationale of Kelley (1967, p. 213) that once attributions are made, they become the basis for making further ones, as- causal dispersion increases, so should confidence in subsequent attributions. Thus, evidence of causal dispersion as a measure of attributional confidence can be addressed by examining the relationship between causal dispersion and confidence in correspondent inferences.

Both the attributional focus and attributional confidence analyses will be conducted for a causal dispersion measure derived from specific-item attribution measures, and a causal dispersion measure derived from three abstract, general-item measures. Consistent with Calder's (1977a,b) contention and the findings of Lichtenstein and Bearden (1986), it is hypothesized that the former measure of causal dispersion will outperform the latter as indicated by the consistency of results within both analyses.

METHOD

Data were collected in a 2x2x2x2 study designed to assess the effects of price discounts on consumer attributions Subjects were exposed to a Volkswagen Rabbit automobile advertisement in which a price discount was offered. They were asked to respond to a series of scales regarding why they thought the merchant was offering the discount and also, what they thought of the merchant. Subjects were 544 undergraduate business majors at a large state university. Because the manipulations are beyond the scope of this paper, they will not be discussed. However, in order to insure that the obtained results were not a product of the manipulations, a pooled within-cell analysis was performed.

Variable Operationalizations

Consistent with the procedures suggested by Elig and Frieze (1979), a pretest sample of 50 undergraduate business majors responded to an open-ended question asking for reasons why a dealership would offer a price discount A total of thirty-four responses were obtained. These responses were then converted to the nine-point probable-improbable statements shown in Table 1 Using data from the experiment, the thirty-four statements were subjected to an unrestricted principle components factor analysis Seven factors had eigen values greater than one, therefore the analysis was reconducted restricting the number of factors to seven. It appeared that four of the factors were person-related factors, two were circumstance-related, and one was product-related. Thus, based on this interpretation and given the a priori notion of Kelley that person, stimulus, and circumstance attributions exhaust the causal space, the analysis was reconducted again restricting the number of factors to three. The factor structure is shown in Table 1. (Similar results with no substantive differences were found using common exploratory factor analysis using a varimax rotation). It does appear that the three factors can be interpreted as representing product, circumstance, and person attributions, respectively. Also, no item interpreted as a person, product, or circumstance attribution in the seven-factor solution loaded on any other factor in the three factor solution, e.g., no person item in the seven-factor solution loaded on the product or circumstance factor in the three-factor solution. A split-half analysis was also conducted and supported the stability of the three-factor solution Thus, the items comprising the respective factors in Table 1 were summed into subscales (cf. Yalch and Yoshida 1983). Coefficient alpha estimates for the three subscales were .81, 78, and 68, respectively In addition to these summed specific-item measures, product, person, and circumstance attributions were also assessed via single, nine-point general-item measures (cf Sparkman and Locander 1980) For example, the general-item person measure was "How probable is it that the discount was due to something about Lott Volkswagen?"

Evaluation of the merchant making the price claim was operationalized in terms of correspondent inferences (Jones and Davis 1965) toward the automobile dealership. A correspondent inference is an inference about individual's intentions and dispositions that follows directly from or corresponds to their behavior. In the present analysis, Lott Volkswagen was personified as the individual and the offering of the price discount represented the behavior. A correspondence in attribution is operationally defined in terms of personality trait attributions (cf Calder and Burnkrant 1977). Thus, correspondent inferences were operationalized by summing nineteen nine-place personality trait semantic differential scales (e g, trustworthiness, greediness, honesty, caring). Coefficient alpha was .90.

A confidence in correspondence inference measure was also assessed After each of the nineteen individual correspondent inference scales was a nine-place confident-not confident scale. Thus, confidence was directly measured for each individual attribution. The nineteen confidence measures were then summed to yield an overall confidence measure.

Two measures of causal dispersion were assessed The first one, which is labeled here as "specific-item causal dispersion," was measured as a respondent's standard deviation across the thirty-four specific-item probable-improbable scales shown in Table 1 Subjects with the highest standard deviations represented those of high causal dispersion, and subjects with low standard deviations represented those of low causal dispersion The second measure of causal dispersion, labeled here as "general-item causal dispersion," was measured as a respondent's standard deviation across the three general-item attribution measures Mizerski (1978) ordered respondents along a causal complexity continuum and conducted analyses on the upper and lower quartiles. A similar partitioning was used here for both causal dispersion measures.

TABLE 1

SPECIFIC-ITEM ATTRIBUTION MEASURES AND FACTOR LOADINGS

RESULTS

Pooled mean scores and correlation coefficients used to address both the attributional focus and confidence issues are presented in Table 2. The pooling process entailed calculating a mean score and correlation coefficient within each of the 16 experimental cells for each variable/relationship shown in Table 2. Weighted averages (based on cell sizes) of means and correlation coefficients were then calculated These weighted averages are the values reported in Table 2. (Statistical significance was calculated within each cell and is referenced by notes c and d in Table 2).

For the specific-item dispersion measure, adjusting for the number of items in each scale, respondents of high dispersion made stronger person attributions than circumstance or product attributions. Also, high dispersion respondents made stronger person attributions than did those of low causal dispersion These results suggest a person attribution focus. For the general-item dispersion measure, subjects of low causal dispersion made stronger product, person, and circumstance attributions than those of high causal dispersion, although those of high dispersion did make stronger person attributions, as opposed to product or circumstance attributions. The correlation coefficients reported at the bottom of Table 2 provide additional evidence. For the specific-item measure, with increasing causal dispersion, there was an increase in person attributions and a concomitant strong decrease in product attributions as predicted For the general-item dispersion measure, all attributions decrease with increasing causal dispersion.

Since correspondent inference attributions are person attributions, it is also appropriate to use them in addressing the attributional focus determination issue. Once again, by looking at the means, for the specific-item measure, high dispersion subjects made much stronger correspondent inference attributions than low dispersion subjects For the general-item dispersion measure, although the difference is very slight, individuals of low causal dispersion made stronger correspondent inferences than those of high dispersion. Again, the correlation coefficients reported al the bottom of Table 2 support this For the specific-item measures, as causal dispersion increased, correspondent inferences increased as predicted Again, this is not the case for the general-item dispersion measure, which does not show any evidence of being correlated at all with correspondent inferences Thus, for the specific-item causal dispersion measure, a consistent pattern of results across all attributional measures for both the mean and correlational analyses provides evidence that the person was perceived as the attributional focus. However, the general-item dispersion measure failed to provide any theoretically-consistent results across the attributional constructs for either analysis The totality of the evidence suggests: (1) the person represents the attributional focus in the present investigation, and (2) only the specific-item dispersion measure was capable of determining it.

TABLE 2

POOLED WITHIN-CELL MEANAND CORRELATION ANALYSES OF THE RELATIONSHIP BETWEEN CAUSAL DISPERSION AND ATTRIBUTIONAL OUTCOMES

The confidence issue can be addressed by considering the confidence in correspondent inference measures also reported in Table 2. For the specific-item dispersion measure, subjects of high causal dispersion did display higher attributional confidence. However, for the general-item dispersion measure, the difference was much smaller Further evidence is provided by the correlation coefficients between confidence and the dispersion measures Confidence was positively correlated to the specific-item dispersion measure, but not the general-item measure Again, these findings are taken as support for causal dispersion as a measure of attributional confidence for the specific-item measure only.

DISCUSSION

A dual purpose of the present study was to propose a measure for determining both the attributional focus and attributional confidence in an experimental setting A measure of causal dispersion was proposed as such a measure Another purpose was to demonstrate that causal dispersion derived from specific-item attributions, and not general-item attributions, was a valid measure. Mean and correlational analyses provided evidence of specific-item, but not general-item, dispersion as a valid measure of both the attributional focus and attributional confidence.

These results have implications for attribution researchers. First, as evidenced by the poor results associated with the causal dispersion measure derived from the general-item attribution measures, valid attribution measurement is a necessary condition for a valid measure of dispersion, and thus, also for attribution al focus/confidence determination. Second, the present study also empirically demonstrates the difference between attributions and the attributional focus. That is, it is quite possible, and even probable in consumer behavior situations, to have a consumer make a mix of product, person, and circumstance attributions to account for some effect in the marketplace. Yet, only that entity which receives the largest summated probability rating, is the attributional focus. Finally, given the evidence provided regarding causal dispersion as a measure of confidence, because dispersion is a derivative rather than a direct scale measure, the criticism of the confidence measure being isomorphic with attribution extremity is not applicable. If the past similarity of measurement procedures is at least partially responsible for the high correlations between attribution extremity and attributional confidence, perhaps dispersion can provide a more accurate reading of a respondent's attributional confidence, which in turn, may allow researchers to be more successful in using confidence as a moderating variable in assessing the impact of attributions on various attributional outcomes.

REFERENCES

Calder, Bobby J. (1977a), "Attribution Theory: Phenomenology or Science," Personality and Social Psychology Bulletin, 3, 612-615.

Calder, Bobby J (1977b), "Endogenous-Exogenous Versus Internal-External Attributions: Implications for the Development of Attribution Theory," Personality and Social Psychology Bulletin, 3, 400-406.

Calder, Bobby J and Robert E Burnkrant (1977), "Interpersonal Influence on Consumer Behavior An Attribution Theory Approach," Journal of Consumer Research, 4 (June), 29-38

Elig, Timothy W. and Irene Hanson Frieze (1979), "Measuring Causal Attributions for Success and Failure," Journal of Personality and Social Psychology, 37 (4), 621-634.

Jolibert, Alain J P and Robert A. Peterson (1976), "Causal Attributions of Product Failure: An Exploratory Investigation," Journal of the Academy of Marketing Science, 4, 446-455.

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

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

Lichtenstein, Donald R and William O Bearden (1986), "Measurement and Structure of Kelley's Covariance Theory," Journal of Consumer Research, 13 (September), 290-296.

Major, Brenda (1980), "Information Acquisition and Attribution Process," Journal of Personality and Social Psychology, 39 (6), 1010-1023.

McArthur, Leslie (1972), 'The How and What of Why: Some Determinants and Consequences of Causal Attribution," Journal of Personality and Social Psychology, 22 (2), 171-193

Mizerski, Richard W (1975), "A Test of the Relationship Between Trait and Causal Attribution," in Advances in Consumer Behavior. Vol 2. ed. Mary J. Schlinger, Chicago: Association for Consumer Research, 471-480.

Mizerski, Richard W (1978), "Causal Complexity: A Measure of Consumer Causal Attribution," Journal of Marketing Research, 15 (May), 220-228.

Mizerski, Richard W., Linda L Golden, and Jerome B. Kernan (1979), "The Attribution Process in Consumer Decision Making," Journal of Consumer Research, 6 (September), 123-140.

Peterson, Robert A. (1982), Marketing Research, Plano, Texas: Business Publications, IDC.

Smith, Robert E. and Shelby D. Hunt (1978), "Attributional Processes and Effects in Promotional Situations," Journal of Consumer Research, S (December), 149-158.

Sparkman, Richard M. and William B. Locander (1980), "Attribution Theory and Advertising Effectiveness," Journal of Consumer Research, 7 (December), 219-224

Weiner, Bernard (1985), "An Attribution Theory of Achievement Motivation and Emotion," Psychological Review, 92, 548-573

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