Examining the Diagnostic Utility of the Fishbein Behavioral Intentions Model

Paul W. Miniard, The Ohio State University
ABSTRACT - Evidence and issues relevant to the diagnostic utility of the Fishbein Behavioral Intentions model are discussed. Two experiments are then presented that examine the degree to which the weight associated with the normative component of the model can accurately reflect the absence or presence of salient normative influences.
[ to cite ]:
Paul W. Miniard (1981) ,"Examining the Diagnostic Utility of the Fishbein Behavioral Intentions Model", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 42-47.

Advances in Consumer Research Volume 8, 1981      Pages 42-47

EXAMINING THE DIAGNOSTIC UTILITY OF THE FISHBEIN BEHAVIORAL INTENTIONS MODEL

Paul W. Miniard, The Ohio State University

[Appreciation is extended to Joel B. Cohen for his assistance in the development of the experiments presented in this paper.]

ABSTRACT -

Evidence and issues relevant to the diagnostic utility of the Fishbein Behavioral Intentions model are discussed. Two experiments are then presented that examine the degree to which the weight associated with the normative component of the model can accurately reflect the absence or presence of salient normative influences.

INTRODUCTION

One promising approach to understanding the important motivations underlying a given behavior is represented by the Fishbein Behavioral Intentions (FBI) model. On the basis of a predictive criterion, the model's performance has been substantiated for a variety of behavioral objects such as alcohol (Schlegel, Crawford, and Sanborn 1977), birth control pills (Davidson and Jaccard 1975), female occupations (Greenstein, Miller, and Weldon 1979), financial loans (Ryan and Bonfield 1980), marijuana (Bearden and Woodside 1978), swine flue vaccinations (Oliver and Berger 1979), and toothpaste (Wilson, Mathews, and Harvey 1975). However, the prediction of intention alone offers little pragmatic value. Rather, the promise of any such formulation lies in its potential for generating information that is useful in developing behavioral change strategies. But before the model can be confidently employed for diagnostic purposes, attention should be directed at examining the model's diagnostic utility; that is, the extent to which the model leads to correct inferences about the determinants of intention and behavior. It is to this issue that this paper addresses itself. After briefly reviewing the FBI formulation, discussion turns to the possible threats to the model's diagnostic utility. Two experiments are then presented that examine the model's diagnostic accuracy under conditions that differ in the salience of normative influence.

THE FISHBEIN BEHAVIORAL INTENTIONS MODEL

The FBI model postulates that intention, viewed as the immediate antecedent of behavior, is determined by an attitudinal or personal component and a normative or social component (Fishbein and Ajzen 1975):

B ~ BI = w1 (AB) + w2 (SN),  (1)

where B is the behavior, BI is the behavioral intention to perform behavior B; AB is the attitude toward performing behavior B; SN is the subjective norm; and w1 and w2 are empirically determined weights.

The first component, AB, is the person's attitude toward performing a particular behavior under a given set of circumstances. This attitudinal component can be decomposed into the perceived consequences of performing the behavior weighted by the evaluation of these consequences:

EQUATION  (2)

where b is the belief that performing behavior B leads to consequence i; e is the person's evaluation of consequence i; and n is the number of salient beliefs.

The second component, SN, represents the person's perception of what important others think s/he should do which is proposed to be a function of the person's beliefs about the expectations of important referent others, and her/his motivation to comply with these referents:

EQUATION  (3)

where NB is the normative belief (i.e., the person's belief that reference group or individual j thinks s/he should or should not perform the behavior); MC is the person's motivation to comply with referent j; and n is the number of relevant referents.

The importance of these two components in determining intentions is expected to vary with the behavior, the situation, and individual differences between persons. Some individuals, for example, may be more "sensitive" to social demands and therefore may attach more weight to their normative considerations than other individuals. Similarly, behaviors that carry greater interpersonal significance or that are visible to important others may be more susceptible to social influences than behaviors that have less interpersonal implications or that are less observable. The component weights (i.e. w1 and w2),which are traditionally estimated by multiple regression procedures, presumably reflect both the existence and the relative importance of these two determinants.

CRITERIA FOR ESTABLISHING THE MODEL'S DIAGNOSTIC UTILITY

Before any model that attempts to identify the determinants of behavioral intention can be confidently employed for diagnostic purposes, there are several important criteria that the model should meet. First, a model that decomposes the determinants of intention is useful only to the extent that there exists a strong relationship between intentions and behavior. Second, the causal flow hypothesized by the model should be validated. Third, the model should be fully identified in the sense that exogeneous variables (i.e., variables other than those postulated by the model as the immediate antecedents of intention) will affect intentions only indirectly (t. e., the exogeneous variable' s influence on intention is mediated by one or more of the postulated determinants). Fourth, the weights employed for representing the presence and relative importance of the hypothesized determinants of intention should accurately reflect "reality". Each of these criteria are in cum discussed below with a brief review of the research relevant to each.

The Intention-Behavior Relationship

One major assumption of the FBI model is that "behavioral intentions are the immediate determinants of the corresponding overt behaviors" (Fishbein and Ajzen 1975, p. 372). If behaviors are not predictable from intentions, there would be little value in specifying and validating a model that identifies the determinants of intentions. Thus, establishing the existence of an intention-behavior relationship as well as the factors that influence the magnitude of this relationship is of crucial importance.

Much of the work relevant to this issue is reviewed by Fishbein and Ajzen (1975, pp. 372-381) and thus need not be repeated here. In general, the assumption of an intention-behavior relationship has been supported, although Fishbein and Ajzen (1975, pp. 368-372) do point out that this relationship is dependent upon (1) the degree to which the measure of intention corresponds directly to the observed behavior, (2) the time interval between measurement of intention and the behavior's occurrence, and (3) the degree to which the person is able to act in accordance with his intention or without the assistance of others. A strong relationship between intention and behavior is expected when the intention measure is specific to the behavior of interest, performance of the behavior is temporally close to the measurement of intention, and the behavior is under volitional control.

Verifying the Hypothesized Causal Flow

As implied by Equations 1, 2, and 3, the FBI model postulates that (1) the influence of AB and SN upon B are mediated by BI, (2) the influence of Sbiei upon BI is mediated by AB, and (3) the influence of SNBjMCj upon BI is mediated by SN. For the most part, this hypothesized causal flow has been substantiated. In an examination of BI's mediating role, Ajzen and Fishbein (1970) held BI constant and tested the partials between B-AB and B-NB. This procedure completely eliminated the previously significant B-AB relationship. The B-AB partial correlation was significant, however, in only one of the two statistical tests. Similarly, Ryan and Bonfield (1980) found that initially significant B - Sbjej and B- SNBjMCj relationships became insignificant when BI was partialled out. Further support is provided by Ajzen and Fishbein (1974) where changes in BI were related to changes in AB and SNBjMCj. In a product evaluation setting, Lutz (1977) tested the causal relationships assumed by the FBI model and concluded that the results provided "reasonably strong support for the postulated flow of effects" (p. 206), although several conceptual and methodological concerns regarding the study and the conclusions drawn have been raised (cf. Carnegie-Mellon University Marketing Seminar 1978; Dickson and Miniard 1978; Lutz 1978a, b).

Several investigations have failed to support the postulated causal flow. Miniard and Cohen (1981) found that, contrary to the causal flow involving the normative component, the addition of SNBjMCj into a regression model containing SN provided a significant improvement in the prediction of BI. This finding casts doubt upon SN's ability to mediate the effects of SNBjMCj upon BI. Further, Bentler and Speckart (1979) report that attitudes and past behavior accounted for a significant portion of behavior that was not mediated by intention.

The Model's Mediational Adequacy

One assumption underlying the model that has received considerable attention is that exogeneous variables will affect intentions only indirectly through either component of the model. Thus, any variable related to intentions should also be related to one of the model's components. Further, any relationship between such variables and intentions should be eliminated when AB and SN are statistically held constant (e.g., the addition of an external variable should not provide a significant increment in the prediction of intentions).

Several approaches have been adopted for operationalizing these external variables. Survey-based investigations have examined the model's ability to mediate such exogeneous variables as religion, age, occupational prestige, and assorted personality and social constructs. In an early study involving transplant donations, Schwartz and Tessler (1972) found a subset of these external variables increased the model's ability to predict intentions. Later investigations, however, have supported the model's mediational adequacy. Jaccard and Davidson (1975)concluded from their study of family planning that the addition of external variables did not provide substantive predictive improvements. Similarly, in the realm of adolescent alcohol use, Schlegel, Crawford, and Sanborn (1977)stated that "...present results could hardly be considered to provide a psychologically meaningful basis for any revision of the Fishbein theory" (p. 428). Attention has also been directed at the components' ability to mediate the effects of the traditionally employed attitude toward an object (AO) upon intentions. Several investigations have shown that AO is mediated by the model's components (Ajzen 1971; Ajzen and Fishbein 1970, 1974) although others have found the components unable to completely mediate AO (Jaccard and Davidson 1975; Schwartz and Tessler 1972).

An alternative approach to this issue has been to examine the extent to which the components mediate the effects of assorted experimental manipulations upon intentions. Ajzen and Fishbein (1970) report that while holding AB and NB constant strongly attenuated the effects of the two experimental manipulations upon intentions, one of the manipulations still remained significant. This latter finding was replicated by Songer-Nocks (1976b), although this and related findings have been debated (Fishbein and Ajzen 1976a, b; Songer-Nocks 1976a). Unconditional support for the components' ability to mediate the effects of experimental manipulations upon intentions is provided by Ryan (1977).

To summarize, evidence bearing on the model's mediational adequacy has been mixed, with some investigations supporting the model's ability to mediate exogeneous variables while other research has shed a less favorable light. However, even in those cases where the components have not completely mediated the effects of external variables, they have greatly attenuated the relationship between such variables and intentions.

Validity of the Component Weights

One key validity concern is the model's accuracy in identifying both the existence and relative importance of salient attitudinal and normative influences. If attitudinal and normative influences were equally important in determining intentions, one would hope that the component weights would correctly represent this condition. One should also expect a given determinant to receive a significant weight when the determinant is in fact important in guiding intentions and not to receive a significant weight when it does not influence intentions. If this were not the case, then the diagnostic usefulness of the model would be limited.

Several factors could threaten the validity of the component weights and thus the model's diagnosticity. Multicollinearity between the components, for instance, can lead to situations where salient influences receive an insignificant weight while a significant weight is assigned to nonsalient influences (this latter situation has been discussed by Cohen and Cohen 1975 under the label of suppressor variables). Reason to suspect the model on this basis has been provided by Miniard and Cohen (1979) as they found the attitudinal and normative measures currently used for model implementation to be inadequate in separating these two sources of influence. The weights' accuracy should also be dependent on the measures' ability to fully capture the causal influences that affect intentions. It has recently been shown that SN may not mediate the influence of MC upon intentions (Miniard and Cohen 1981). This implies that the normative component may be less accurate when SN is used for measuring normative influences.

Evidence demonstrating that the measures are not distinct or that they fail to fully mediate influences which they should capture, however, is limited in that it only establishes the potential for inaccurate weights. It may be that, even in the presence of such problems, the weights correctly reflect the salient influences. What is needed, then, is a direct test of the weights' validity.

Two possible criteria for evaluating the validity of the component weights are (1) the weights' ability to represent the absence or presence of salient attitudinal and normative influences and (2) whether the weights differ in their relative sizes in expected ways. Under the first criterion the normative weight, for example, should be significant when normative influences are important determinants of intentions. Similarly, the weight should not be significant when normative influences are unimportant. Failure to confirm these expectations would cast serious doubt upon the model's diagnostic usefulness. The second criterion is less concerned with the accuracy of a particular weight. Rather, the focus is on the relative magnitude of the weights. Drawing upon some conceptual framework, one might hypothesize that a given component should be more important (i.e., have a larger weight) in situation A with the remaining component dominating intentions in situation B. Note that support for one criterion does not automatically imply support for the other. For instance, one might be led to predict that both sources of influence are important (Criterion 1), although the attitudinal component should be more important than the normative component (Criterion 2). If the results yielded a large significant attitudinal weight but a small insignificant normative weight, only the latter prediction would be supported.

Research examining the weights' validity has, without exception, relied upon the second criterion. Using the Prisoner's Dilemma game, Ajzen (1971) and Ajzen and Fishbein (1970) told subjects to consider themselves partners (cooperation condition) or to do better than the other person (competitive condition). It was expected that normative considerations would carry greater weight in the cooperation condition while attitudinal considerations would be more important in the competitive condition. The observed patterns of regression weights supported these predictions. Similar results derived from this experimental paradigm are reported by Songer-Nocks (1976b).

Wilson, Mathews, and Monoky (1972) found partial support for hypothesized differences in the weights' relative magnitudes. As predicted, the attitudinal component received a larger weight when the other person in a buyer-seller dyad was described as being dissimilar to the subject. The attitudinal component was also dominant when the other vas described as being similar, although it was expected that the normative component would be of greater importance.

Evidence failing to support expected differences in the components' relative weights within a marketing context is reported by Ryan (1978). When the model was applied to two brands of toothpaste, social influences were predicted to be more important for Ultra Brite than Crest since the former's promotional activities emphasized social interactions. The results contradicted this hypothesis as the normative weight was larger than the attitudinal weight for Crest while the reverse was true for Ultra Brite.

As stated above, evidence addressing the weights' accuracy in representing the presence or absence of salient normative and attitudinal influences is nonexistent. However, close inspection of the correlational and regression analyses reported in several investigations suggests that the weights may not be accurately reflecting the true importance of a given component. In particular, it is often the case that a component correlates significantly with intentions yet receives an insignificant weight. Fishbein (1966), for example, reported that males' attitude toward premarital sexual intercourse had a significant correlation with their intention to engage in this behavior but received an insignificant weight in the regression equation. The same result with respect to SNBjMCj is reported by Ajzen and Fishbein (1972) where the normative component weight was insignificant although SNBjMCj was significantly correlated with intentions. An extreme instance of these inconsistencies where actual behavior was the criterion is provided by Ajzen and Fishbein (1970). As reported in Table 4 of their results, both AB and NB were significantly related to behavior but neither measure received a significant regression weight. Further examples of significant correlations and insignificant weights can be found in Greenstein, Miller, and Weldon (1979), Schwartz and Tessler (1972), and Warshaw (1980).

Unfortunately, evidence that a given component receives a significant correlation but an insignificant weight can not be unambiguously interpreted as evidence against the model's diagnostic utility. It may be the case that the measure should not correlate significantly with intentions. If the measure used to operationalize the normative component was sensitive to attitudinal influences (i.e., the normative measure also tapped the attitudinal component), the normative component might correlate with intention even when the intention is solely under attitudinal control. In such situations, the lack of a significant normative weight is quite appropriate. However, if a significant correlation between the normative component and intentions is valid in the sense that normative influences do play an instrumental role in determining intentions, the model's failure to assign a significant weight to the normative component would provide a clear indication of the potential for misleading inference. Without knowing whether a given component is in fact important, one cannot easily reconcile inconsistencies between the correlational and regression analyses.

Therefore, the following two experiments will attempt to provide some initial evidence concerning the FBI model's ability to accurately reflect the absence or presence of salient attitudinal and normative influences. In particular, the validity of the normative component's weight will be examined under conditions where (1) the social environment does not influence intentions (Study I) and (2) social influences do affect intentions (Study II).

STUDY I

Method

Subjects and Procedures. A total of thirty male and female undergraduate marketing students participated in the experiment. Subjects were informed that they were participating in a study of "new product introductions" and that they would receive a booklet containing two role playing descriptions involving a new product. Both descriptions began by asking subjects to assume that they had participated in a nationwide marketing research study of consumers' reactions to a new product. In one of the descriptions, the subject was portrayed as being very favorable toward purchasing the product (Attitude-Positive). The remaining description stated that the subject was very unfavorable toward buying the product (Attitude-Negative). The order of presentation of these two descriptions was counterbalanced across all subjects.

For a third of the subjects, this attitudinal statement completed the description. This condition (Others' Expectations-Unknown) thus represented a situation that was devoid of information concerning the social environment (e.g., whether others thought the subject should purchase the product). The remaining subjects, however, received descriptions that contained information concerning the opinions of others. Subjects in the Others' Expectations-Positive condition were informed that most others were very favorable toward the new product and that almost everyone had responded "yes" to the question "Do you think others should buy this product?". Conversely, others were described as being unfavorable toward the product in the Others' Expectations-Negative condition and as answering "no" to the question regarding whether others should purchase the product.

Subjects in these latter two conditions (where others' opinions are known) were then told that: "Of course, what others feel you personally should do here is a matter of no concern to you. Buying or not buying the new product is strictly your own decision." These statements were included in order to create a situation in which the social environment is not an important source of influence. If this is the case (which will be examined below), the question then becomes whether the FBI model can correctly reflect via the normative component weight the unimportance of the social environment.

Measures.  Prior to completing the following measures, subjects were instructed how the response scales were to be used (e.g., the meaning of the scale categories). Subjects first responded to a measure of BI via three 7-point bipolar semantic differential scales (likely-unlikely, probable-improbable, possible-impossible). Subjects then indicated their attitude toward the behavior ("For me, buying the new product would be:") on four 7-point bipolar evaluative semantic differential scales (good-bad, foolish-wise, rewarding-punishing, harmful-beneficial). Subjects next responded to the SN measure (i.e., "Most people who are important to me think I should/should not buy the new product") via a 7-point scale ranging from "I should" (+3) to "I should not" (-3).

Results

The results of the 2 (Attitude) X 3 (Others' Expectations) ANOVA conducted on subjects' responses to the BI, AB, and SN measures are presented in Table 1. Verification that social influences were not an important determinant of BI required testing the effect of the Others' Expectations manipulation upon subjects' responses to the BI measure. If the provision of information concerning others' opinions did in fact play an influential role in the formation of subjects' intentions to purchase the product, then the Others' Expectations manipulation should have a significant impact on BI. This manipulation did not attain significance (p < . 22), although the attitudinal manipulation was highly significant and accounted for 92% of the measure's variation, as calculated by use of Hays' (1963) omega-squared (w2) statistic. The lack of a significant normative manipulation indicates that beliefs regarding others' expectations were not an important determinant of subjects' intentions to purchase the product. Accordingly, the normative component as operationalized by SN should not receive a significant weight.

Examination of the normative weight's ability to reflect the nonsalient social influences existing in the Others' Expectations-Positive and Negative conditions involved conducting a regression analysis pooling (i.e., collapsing across the experimental cells after adjusting for differences in cell means) across these conditions. Consistent with prior research, the model provided a reasonable prediction of intentions (R = .73, p < .001). Support for the normative weight's validity was attained as only the attitudinal component received a significant (p < .01) weight while the normative weight was clearly insignificant. It should also be noted that, in this situation, the regression and correlational analyses were consistent as AB achieved ,a significant (p < .01) correlation (r = .72) with intentions while SN did not (r = .08).

TABLE 1

ANOVA SUMMARY TABLE: STUDY I

STUDY II

[This experiment has been previously reported (Miniard and Cohen 1979, 1981), although the data were not discussed in light of the present concerns. The reader is referred to these articles for a more detailed discussion of the experimental procedures.]

Method

Subjects and Procedures.  A total of 193 male and female undergraduate marketing students participated in the experiment. Each subject read one of four different scenarios in which a hypothetical woman was considering the purchase of a dress and that a female neighbor who was accompanying this woman thought she should purchase the dress. The four scenarios represented the product of a 2 X 2 manipulation of the scenario content. The first manipulation, No Manipulative Intent (NMI)--Manipulative Intent (MI), refers to the presence or absence of information that the woman's neighbor was exerting influence for self-serving reasons. The second manipulation, Agree--Disagree, altered the statement describing the woman's attitude toward the dress from positive to negative. Upon completion of the scenario, subjects were asked to play the role of this woman in responding to a questionnaire containing the measures necessary for operationalizing the FBI model. At this point, a third manipulation was introduced as subjects responded to one of three possible questionnaires that differed in the level of specificity at which MC was assessed (i.e., general, moderate, and situation specific). This factor was included since there presently exists some disagreement as to which level of measure specificity is most appropriate (cf. Ahtola 1976; Fishbein 1976; Glassman and Fitzhenry 1976; Miniard and Cohen 1981).

Measures.  Subjects first responded to one of three alternative measures of MC via 7-point scales ranging from "I want to do" (+3) to "I want to do the opposite of" (-3). Depending upon whether level of specificity was general, moderate, or situation specific, the scale was prefaced by either "In general", "In the area of clothes shopping behavior'', or "In this situation", respectively. Next, NB was assessed on 7-point scales with the endpoints "I should" (+3) and "I should not" (-3). NB and MC were obtained for three referents: the hypothetical woman's neighbor, husband, and parents. Subjects then reported their attitude toward the behavior on four 7-point bipolar evaluative semantic differential scales (good-bad, foolish-wise, rewarding-punishing, harmful-beneficial). Subjects next responded to the SN measure via a 7-point scale ranging from "I should" (+3) to "I should not" (-3). This was followed by an alternative attitudinal measure (PAB) not employed by Fishbein. This measure ("Forgetting what others think you should do and their reaction, you personally feel that buying the dress is:") was designed to minimize the potential for normative contamination by instructing subjects that the influence of others was not to be included in their responses to the measure. The major difference between PAB and AB was that the latter does not suggest that the influence of others is to be ignored. BI was next measured on three 7-point bipolar semantic scales (likely-unlikely, probable-improbable, possible-impossible).

Validation Condition.  In addition to examining the impact of the NHI-MI manipulation on BI for inferring the importance of normative influences, further evidence was gathered in the form of subjects' self-reports. An additional 57 male and female undergraduate marketing students participated in this phase. Each subject read one of the four scenarios, responded to a measure of BI, and was then asked to estimate the impact of the social influences depicted in the scenario upon their intentions to purchase the dress via an 11-point scale with the endpoints "more likely to buy the dress" (11) and "less likely to buy the dress" (1). The scale midpoint of this "directional importance" measure was to be circled if such influences were unimportant. If social influences were an important determinant of intentions, subjects' responses to this measure should then significantly differ from the scale midpoint.

Results

Evidence supporting the importance of normative considerations in determining intentions was attained as the cell means for both the NMI (X = 8.31) and MI (X = 4.14) scenarios significantly (p < .05) differed from the scale midpoint (i.e., 6) Further support for the salience of social influences was also attained as the NMI-MI manipulation had a highly significant (p < .001) effect upon intentions and explained 16% of the variability. These findings suggest that the normative component should receive a significant regression weight.

TABLE 2

POOLED MULTIPLE REGRESSION ANALYSES FOR THE PREDICTION OF BEHAVIORAL INTENTIONS: STUDY II

Given the two alternative attitudinal measures (i.e., AB and PAB) and the four different normative representations (i.e., SN and the three SNBjMCj formulations that differ in the level of MC specificity), a total of eight regression models were tested. The results of these pooled regression analyses for the various attitudinal and normative operationalizations are summarized in Table 2. Focusing on those models employing Fishbein's standard AB measure (i.e., models 1-4 in the table), the FBI model fails to correctly identify the importance of normative influences when either SN (model 1) or a general level of MC specificity (model 2) is employed for operationalizing the normative component as an insignificant normative weight was attained for these models. Interestingly, replacement of AB by PAB (which is significantly less correlated with SN) improves the FBI model's diagnostic performance when SN is employed for representing normative influences. While the normative weight was insignificant for the AB + SN model, the weight attained significance for the PAB + SN model. With respect to the correlational results, all of the normative measures except for the SNBjMCj formulation where MC was assessed at a general level correlated significantly (p < .05) with intentions. Thus, in the one instance where the correlational and regression analyses conflicted (i.e., model 1), the correlational analysis was correct.

CONCLUSION

The ability of the FBI model to validly represent the importance of normative influences in two distinct situations was examined. Although the model correctly identified the lack of salient normative influences, inadequacies were found for certain normative operationalizations when normative influences were an important determinant of intention. Evidence relevant to the appropriate level of MC specificity was also uncovered as the general level failed to reflect the importance of social influences whereas the moderate and situation specific levels did adequately represent such influences. These results can be attributed to the fact that some of the experimental conditions (i.e., the MI conditions) under which MC was assessed involved situations where a generally positive referent became negative. The general level, which does not ask subjects to report the situational importance of a referent, was apparently unable to reflect these situational variations in the referent's influence potential. Since the normative component received an appropriate weight when either of the two remaining levels of MC specificity were employed, these findings suggest that such levels are perhaps most appropriate for measuring MC.

One interesting question that arises is why the normative component, when represented by SN, performs accurately under conditions lacking social influences but is inaccurate under settings involving salient normative influences. One possible explanation is the multicollinearity between the attitudinal and normative components. In Study I, AB and SN were unrelated (r = -.90). In Study II, however, there was a significant correlation between the two measures (as reported in Table 2) thus indicating that the likelihood for diagnostic error increases in situations where the components are not independent.

While the present findings do shed some light on an area that has thus far been ignored, it is clear that further work will be necessary before the FBI model can be confidently used as a diagnostic tool for identifying the importance of attitudinal and normative influences. Possible areas of improvement over the present studies could involve the use of more natural (i.e., non-role playing) settings that allow a test of the weights' ability to represent the absence or presence of salient attitudinal and normative influences within the same experimental framework rather than comparing across data sets. Research that identifies the conditions under which the model will or will not accurately identify the importance of these influences would seem to hold the greatest promise.

REFERENCES

Ahtola, O. T. (1976), "Toward a Vector Model of Intentions," in Advances in Consumer Research, Vol. 3, (ed.) B. B. Anderson, Cincinnati, OH: Association for Consumer Research, pp. 481-4.

Ajzen, I. (1971), "Attitudinal vs. Normative Messages: An Investigation of the Differential Effects of Persuasive Communications on Behavior," Sociometry, 34, 263-80.

Ajzen, I., and Fishbein, M. (1970), "The Prediction of Behavior From Attitudinal and Normative Behaviors," Journal of Experimental Social Psychology, 5, 400-16.

Ajzen, I. (1972), "Attitudes and Normative Beliefs as Factors Influencing Behavioral Intentions," Journal of Personality and Social Psychology, 21, 1-9.

Ajzen, I. (1974), "Factors Influencing Intentions and the Intention-Behavior Relationship," Human Relations, 27, 1-15.

Bearden, W. O., and Woodside, A. G. (1978), "Situational and Extended Attitude Models as Predictors of Marijuana Intentions and Reported Behavior," Journal of Social Psychology, 106, 57-67.

Bentler, P. M., and Speckart, G. (1979), "Models of Attitude-Behavior Relations," Psychological Review, 86, 452-64.

Carnegie-Mellon University Marketing Seminar (1978), "Attitude Change or Attitude Formation? An Unanswered Question,'' Journal of Consumer Research, 4, 271-6.

Cohen, J., and Cohen, P. (1975), Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. Hillsdale, NJ: Lawrence Erlbaum Associates.

Davidson, A. R. and Jaccard, J. J. (1975), "Population Psychology: A New Look at an Old Problem," Journal of Personality and Social Psychology, 31, 1073-82.

Dickson, P. R., and Miniard, P. W. (1978), "A Further Examination of Two Laboratory Tests of the Extended Fishbein Attitude Model," Journal of Consumer Research, 4, 261-6.

Fishbein, M. (1966), "Sexual Behavior and Propositional Control," paper presented at the Psycholomic Society meetings, St. Louis.

Fishbein, M. (1976), "Extending the Extended Model: Some Comments," in Advances in Consumer Research, Vol. 3, (ed.) B. B. Anderson, Cincinnati, OH: Association for Consumer Research, pp. 491-7.

Fishbein, M. and Ajzen, I. (1975), Belief, Attitude, Attention, and Behavior: An Introduction to Theory and Research. Reading, MA: Addison Wesley.

Fishbein, M. (1976a), "Misconceptions About the Fishbein Model: Reflections on a Study by Songer-Nocks," Journal of Experimental Social Psychology, 12, 579-84.

Fishbein, M. (1976b), "Misconceptions Revisited: A Final Comment," Journal of Experimental Social Psychology, 12, 591-2.

Glassman, M., and Fitzhenry, N. (1976), "Fishbein's Subjective Norm: Theoretical Considerations and Empirical Evidence,'' in Advances in Consumer Research, Vol. 3, (ed.) B. B. Anderson, Cincinnati, OH: Association for Consumer Research, pp. 477-80.

Greenstein, M., Miller, R. H., and Weldon, D. E. (1979), "Attitudinal and Normative Beliefs as Antecedents of Female Occupational Choice," Personality and Social Psychology Bulletin, 5, 356-62.

Hays, W. L. (1963), Statistics for Psychologists. New York: Holt, Rinehart and Winston.

Jaccard, J., and Davidson, A. R. (1975), "A Comparison of Two Models of Social Behavior: Results of a Survey Sample" Sociometry, 38, 497-517.

Lutz, R. J. (1977), "An Experimental Investigation of Causal Relations Among Cognitions, Affect, and Behavioral Intention," Journal of Consumer Research, 3, 197-208.

Lutz, R. J. (1978a), "Rejoinder," Journal of Consumer Research, 4, 266-271.

Lutz, R. J. (1978b), "Rejoinder," Journal of Consumer Research, 4, 276-8.

Miniard, P. W., and Cohen, J. B. (1979), "Isolating Attitudinal and Normative Influences in Behavioral Intentions Models," Journal of Marketing Research, 16, 102-10.

Miniard, P. W. (1981), "An Examination of the Fishbein Behavioral Intention Model's Concepts and Measures," Journal of Experimental Social Psychology, in press.

Ryan, M. J. (1977), "Programmatic Research Based on Fishbein's Model," in Consumer and Industrial Buying Behavior, (eds.) A. G. Woodside, J. N. Sheth, and P. D. Bennett, New York: North Holland, pp. 151-66.

Ryan, M. J. (1978), "An Examination of an Alternative Form of the Behavioral Intention Model's Normative Component," in Advances in Consumer Research, Vol. 5, (ed.) H. K. Hunt, Ann Arbor, MI: Association for Consumer Research, pp. 283-9.

Ryan, M. J. and Bonfield, E. H. (1980), "Fishbein's Intentions Model: A Test of External and Pragmatic Validity,'' Journal of Marketing, 44, 82-95.

Schlegel, R. P., Crawford, C. A. and Sanborn, M. D. (1977), "Correspondence and Mediational Properties of the Fishbein Model: An Application to Adolescent Alcohol Use," Journal of Experimental. Social Psychology, 13, 421-30.

Schwartz, S. H., and Tessler, R. C. (1972), "A Test of a Model for Reducing Measured Attitude Behavior Discrepancies,'' Journal of Personality and Social Psychology, 24, 225-36.

Songer-Nocks, E. (1976a), "Reply to Fishbein and Ajzen," Journal of Experimental Social Psychology, 12, 285-90.

Songer-Nocks, E. (1976b), "Situational Factors Affecting the Weighting of Predictor Components in the Fishbein Model," Journal of Experimental Social Psychology, 12, 56-59.

Warshaw, P. R. (1980), "A New Model for Predicting Behavioral Intentions: An Alternative to Fishbein," Journal of Marketing Research, 17, 153-72.

Wilson, D. T., Mathews, H. L., and Harvey, J. W. (1975), "An Empirical Test of the Fishbein Behavioral Intention Model," Journal of Consumer Research, 1, 39-48.

Wilson, D. T., Mathews, H. L. and Monoky, J. F. (1972), "Attitude as a Predictor of Behavior in a Buyer-Seller Bargaining Situation: An Experimental Approach," in American Marketing Association Combined Conference Proceedings, (eds.) B. W. Becker and H. Becker, Chicago: American Marketing Association, pp. 390-5.

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