Causal Relationships in the Fishbein Behavioral Intention Model

ABSTRACT - One potentially viable diagnostic tool for developing sound behavioral change strategies is the Fishbein behavioral intention model. The soundness of employing the model for such goals is dependent on the validity of the causal relationships specified by the model. The results of an experimental study of the model's formation processes led to rejecting the causal network hypothesized by Fishbein in favor of an alternative causal model.

Citation:

Paul W. Miniard and Thomas J. Page, Jr. (1984) ,"Causal Relationships in the Fishbein Behavioral Intention Model", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 137-142.

Advances in Consumer Research Volume 11, 1984      Pages 137-142

CAUSAL RELATIONSHIPS IN THE FISHBEIN BEHAVIORAL INTENTION MODEL

Paul W. Miniard, The Ohio State University

Thomas J. Page, Jr., University of Wisconsin-Madison

[The authors thank Peter Dickson, Alan Sawyer, and anonymous reviewers for their useful comments and suggestions. Financial support for this research was provided to the senior author by an American Marketing Association dissertation grant and the Marketing Faculty Research Fund, OSU.]

ABSTRACT -

One potentially viable diagnostic tool for developing sound behavioral change strategies is the Fishbein behavioral intention model. The soundness of employing the model for such goals is dependent on the validity of the causal relationships specified by the model. The results of an experimental study of the model's formation processes led to rejecting the causal network hypothesized by Fishbein in favor of an alternative causal model.

INTRODUCTION

Behavioral intention models represent one orientation to understanding the reasons for behavior. The Fishbein behavioral intention formulation, which hypothesizes a dual component model of intention consisting of attitudinal and normative constructs, has received the greatest amount of attention in the consumer behavior literature. One fundamental concern underlying the model's usefulness as a diagnostic tool is the hypothesized causal relationships among the model's constructs. Confirmation of these relationships would lend support to using the model as a framework for devising sound behavioral change strategies.

While there is presently a large body of evidence relevant to the causal relationships underlying the Fishbein model, this literature is limited in several respects. First, research examining the entire set of model constructs with appropriate measures has yet to appear. Many studies, for example, have not considered behavior in examining the model's causa] system while investigations that include behavior have omitted other model constructs. Second, the majority of attention has been focused on the attitudinal portion o' the model. Relatively little emphasis has been given to the normative chain of the model, despite the fact that this model component has been and remains the most problematic. Third, as elaborated below, tests of hypothesized relationships within the normative component have usually occurred within situations that may have biased the results. Fourth, recent advancements in the analytical techniques 'or causal modeling have not been reflected in the analyses undertaken in many investigations. Finally, the causal network assumed by the model has rarely been tested against competing causal configurations. Thus, while a study may provide reasonable support for the model, the question concerning whether alternative causal systems would receive even stronger support is rarely addressed.

In the following research, the causal relationships hypothesized by Fishbein among the entire set of model constructs were test;d and compared with a competing causal configuration. This was done within an experimental setting involving treatment conditions designed to influence the antecedents of the attitudinal and normative components, thus providing a test of causal relationships among the model's endogenous variables under formation conditions. Before presenting the experiment and results, the model and causal issues relevant to this investigation are discussed.

THE FISHBEIN BEHAVIORAL INTENTION MODEL

Figure 1 contains Fishbein's conception of the causal relationships among the model constructs. With the exception of the Sb_ie_i <-> SNB_jMC_j relationships, all paths are consistent with Fishbein's and Ajzen's (1975, p. 16) original diagrammatic presentation of the model. As noted by Ryan (1982), Fishbein's original causal diagram was not completely consistent with his subsequent discussions of the model's causal properties. In particular, Fishbein and Ajzen's (1975, Chapter 7) discussions concerning inferential belief formation indicate that Sb_ie_i and SNB_jMC_j may in fact influence each other (represented in Figure 1 by the double-headed arrow).

According to Fishbein, behavioral intention (BI), viewed as the immediate antecedent of behavior (B), is determined by the person's attitude toward the behavior (AB) and by the person's perception of social pressures which is represented by subjective norm (SN). AB and SN are, in turn, decomposed into specific cognitive and motivational constructs. Attitude is viewed as a function of the beliefs (bi) about the behavior's consequences weighted by the evaluation (ei) of these consequences (i.e., Sb_ie_i AB). Similarly, SN is proposed to be a function of the normative beliefs (NBj) about referent expectation weighted by the motivation to comply (MCj) with these referents (i.e., SNB_jMC_j -> SN).

FIGURE 1

FISHBEIN'S CAUSAL MODEL

Although research testing the entire causal model is nonexistent, there is a considerable amount of evidence pertaining to the individual causal paths in Figure 1. There is an abundance of correlational (Bonfield 1974; Bowman and Fishbein 1978; Jaccard, Knox, and Brinberg 1979; King 1975; Pomazal and Jaccard 1976; Ryan and Bonfield 1980) and experimental (Ajzen 1971; Ajzen and Fishbein 1970, 1974; Songer-Nocks 1976) data to support the intention-behavior relationship. Similarly, attitudes and subjective norms have been shown to accurately predict intentions across a variety of behaviors ranging from family planning (Davidson and Jaccard 1975) to product choice (Bonfield 1974; Ryan and Bonfield 1975; Warshaw 1930), and experimental manipulations on intentions (Ajzen 1971; Ajzen and Fishbein 1970, 1972, 1974, Lutz 1977; Ryan 1977; Songer-Nocks 1976). There is substantial correlational evidence for the postulated relationships between Sb_ie_i and AB (see Fishbein and Ajzen 1975, Chapter 6) and between SNB_jMC_j and SN (Glassman and Birchmore 1974; Glassman and Fitzhenry 1976; King and Jaccard 1973; Pomazal and Brown 1977; Pomazal and Jaccard 1976). Few studies, however, have examined whether changes in the hypothesized determinants of AB and SN generated predictable effects. Lutz (1975) and Ryan (1977) show that manipulations of bi influence AB as expected, although Lutz reported somewhat disconfirming findings regarding the effects of changes in ei. Ryan (1977) also found that varying NB produced expected variations in SN. Finally, support for the postulated relationship between Sb_ie_i and SNB_jMC_j is reported by Ryan (1982).

On the basis of our analysis of the model's conceptual and operational structure and recent empirical findings, we propose an alternative causal configuration which is presented in Figure 2. The causal model expressed in Figure 2 differs from the prior one in that (1) SNB_jMC_j is expected to have a direct effect on AB, (2) SNB_jMC_j is not expected to have a direct effect on SN, and (3) iB is expected to have a direct effect on SN. The rationale and evidence for these modifications are discussed below.

FIGURE 2

A PROPOSED CAUSAL CONFIGURATION

Within Fishbein's framework, AB is conceptualized and operationalized as representing one's overall evaluation of the behavior. Thus, AB is intended to reflect one's behavioral evaluation based on both personal and normative considerations. We would expect, then, behaviors to be evaluated more favorably when they are consistent with referents' expectations with whom one is motivated to comply. The potential for SNB_jMC_j to have an indirect effect on AB through Sb_ie_i is recognized by both causal configurations. We believe, however, that the full range of social influences (i.e., compliance, identification, internalization) that MCj presumably captures may not be completely reflected in terms of the behavioral consequences contained within Sb_ie_i. Consequently, we hypothesize that SNB_jMC_j will have a direct effect on AB. Evidence for such an effect has recently been provided by Ryan (1982).

Threats to the validity of the postulated relationship between SN and SNB_jMC_j have been raised by a number of researchers (Ahtola 1976; Lutz 1976; Miniard and Cohen 1981; Warshaw 1980). Of particular relevance to the present investigation is the role of MCj in the formation of SN. As presently operationalized, SN requires the person to indicate whether "most people" who are important to her/him think s/he should or should not perform the behavior. While one can easily see how responses to this perceptual measure would be based on some aggregation of perceptions concerning. specific referents (i.e., SNB_j), it is not clear how or even why the SN measure would reflect the motivational forces captured by MCj. As pointed out by Ahtola (1976), suppose the person perceived her/his referent others to favor her/his performing the behavior. In this case, both SNB_j and SN would be positive. When the person is motivated to comply with others (i.e., MCj is positive), then SNB_jMC_j and SN would provide consistent results as both would yield positive scores. If, however, MC- is neutral or negative, then SN and SNB_jMC_j would produce conflicting results (i.e., SN would be positive while SNB_jMC_j would be neutral or negative). Accordingly, we hypothesize that SNB_jMC_j will not have a direct effect on SN unless MCj approaches a positive constant.

As previously noted, a statistically significant and frequently substantial correlation between SN and SNB_jMC_j has often been reported in the literature. However, without knowing the correlation between SNBj and SN, one cannot evaluate the need for weighting NB3 by MCj in predicting SN. A significant correlation between SS and SNB_jMC_j would occur if SN and SNBj were highly related and MCj approached a positive constant. Since most behavioral settings in which others are influential are likely to involve only positive referents (i.e., one is motivated to comply with the referents), it is probably that MCj will often be positive with little variation across referents. To overcome this potential bias, Miniard and Cohen (1981) examined MC's role in the SNB_jMC_j -> SN relationship under experimental conditions designed to induce sufficient variation in motivation to comply. In this instance, weighting NB; by MCi decreased the prediction of SN which suggests that MCj is not an antecedent of SN. Accordingly, as detailed below, the causal relationships among the Fishbein model constructs are examined in this study under experimental conditions that should produce the variation in MCj necessary for an unbiased test.

Recall that SN represents the person's perception of what "most important others think I should do." In many cases, a person will possess either direct (perhaps through face-to-face communication) or indirect (perhaps through observing others' behavior) knowledge of others' expectations. Responses to SN-type measures should then be based on information retrieved from memory. However, neither direct nor indirect knowledge of these expectations is likely to exist for novel behaviors (such as the one examined in this study). Both Miniard and Cohen (1981) and Warshaw (1980) have suggested that, in such situations, responses to SN-type measures will be based on one's attitude toward the behavior (i.e., AB -> SN). That is, the person may infer that most important others would favor her/his performing behaviors that s/he favorably evaluates but not to perform those that are viewed as unattractive.

It should be noted that Fishbein and Ajzen (1981) have acknowledged the potential for subjective norms to be inferred from one's attitude toward an object (Ao). However, Fishbein has consistently distinguished Ao from AB with Ao being "...treated like any other variable external to the theory" (Fishbein and Ajzen 1975, p. 317). Consequently, the AB -> SN path was not included in the Fishbein causal configuration.

The final causal issue addressed in this research is the potential for AB to have a direct effect on behavior. Bentler and Speckart (1979, 1981) have questioned whether a cognitive construct such as behavioral intent can fully mediate the influence of affect on behavior. Although their findings substantiated the existence of an AB -> B path, both Bagozzi (1981) and Ryan (1982) identify limitations stemming from the use of retrospective self-reported behavior and the failure to employ Fishbein's standard measurement procedures in assessing these constructs. In a study that did not rely on self-reported behavior and which did use acceptable AR and BI measures, Bagozzi (1981) found BI to mediate the influence of AB on behavior. Bagozzi (1982) has recently suggested that AB may directly influence behavior under certain conditions. None of these conditions, however, apply to the present investigation. Thus, we hypothesize that AB will only affect behavior indirectly (i.e., through BI).

METHOD

Overview

Subjects were led to believe that they were participating in a study of factors (e.g., communication patterns, group cohesiveness) that influence group decision making in which they would recommend one of several alternative brands for market introduction. Subjects were randomly assigned to the cells of a 2 (Attitude) x 2 (Normative) between-subjects factorial design. The first manipulation, Attitude, was intended to vary subjects' attitude toward recommending a given brand by altering the brand's suitability for market introduction. The second manipulation, Normative, attempted to alter the group's ability to exert normative influence by varying the group's reward power. Each subject, then, was exposed to a particular combination of attitudinal and normative influence.

Subjects

Forty-four male and female undergraduate marketing students participated in the experiment as partial completion of class requirements. The task was designed specifically for these subjects in order to heighten involvement and relevance. Accordingly, marketing students were placed in a "case analysis" type of setting involving the introduction of a new product.

Procedure and Independent Variables

Subjects were informed that they would act as a marketing consulting group for a company that was introducing a new brand of dog food. Their goal was to recommend one of seven alternative brands for market introduction. The brand chosen (i.e., recommended to the group) for introduction represented the critical behavior to be explained by the model.

After reviewing the procedures, subjects were placed in separate rooms and read the materials describing the case situation. The first "marketing research" report given to subjects in the Attitude-favorable (Attf) conditions indicated that brands A and D were identical to one another and superior to the remaining brands. In addition, subjects believed that they alone (in their group) were given a second report substantiating the superiority of A and D. Subjects in the Attitude-unfavorable (AttU) conditions first received a report suggesting that A and D, again identical, were similar in overall attractiveness to brand F and superior to the remaining brands. However, the second report (which again they believed that only they received) clearly revealed that F was more likely to be successful. A "group leader," who was supposedly chosen at random from the group, subsequently recommended brand A in all conditions. Subjects in the AttU conditions, then, have no reason to be favorably disposed toward the leader's recommendation, outside of possible normative considerations.

In order to examine the normative chain of the Fishbein model, it was necessary to provide the group with a social power basis for influencing subjects' brand choice behavior. This was accomplished by providing the group with reward power, a power basis that has previously been employed (Kelman 1974) for inducing social compliance. Subjects in the Normative-strong (Norms) conditions were informed (with appropriate rationale) that they had an opportunity to win a $20.00 prize to be awarded by the group to the member displaying the highest level of "team spirit." It was clear from the description of this award that supporting the leader was essential if they desired to be rewarded by the group. These subjects were also told that their winning the award was enhanced by their position in the group's communication structure (i.e., they would be the last group member to send a written message to the "other members"). Subjects in the Normative-weak (Normw) conditions were not informed of any monetary prize (i.e., just the award itself). The lack of financial reward power should substantially lower the group's normative influence potential in the Normw conditions relative to the Norms conditions.

Questionnaire

Following a page and a half of instructions and examples concerning how the 7-point bipolar response scales employed for assessing the various constructs should be used, subjects first responded to the measures necessary for operationalizing the Ebiei index of attitude. Beliefs (bi) that recommending a particular brand for market introduction would lead to two behavioral consequences, meeting the needs of the target market and winning the group member award, were assessed on "likely-unlikely" scales. Attitude toward recommending each brand was derived from the summed score of four evaluative semantic differential scales (good-bad, foolish-wise, rewarding-punishing. harmful-beneficial).

In accordance with Fishbein's (1976) recommendations, motivation to comply with the group was assessed on a scale ranging from "I want to do" to "I want to do the opposite of." Normative beliefs concerning whether or not subjects perceived the group to favor their recommending a given brand were assessed for each brand on "I should-I should not" scales This scale was also employed for measuring subjective norms ("most people who are important to me think I should/I should not recommend brand A") for each brand. Intention to recommend each brand was measured on "likely-unlikely" scales. Subjects then recorded the letter of the brand they would recommend to the group.

RESULTS

Impact of Manipulations on Motivation to Comply

It was anticipated that the experimental manipulations would induce the variation in MCj necessary for examining its role in the formation of SN. It would appear that the manipulations were successful in this regard as the means for MCj ranged from 1.38 in the Attf/Norms condition to -0.60 in the Attu/Normw condition. Further, the results of a 2 (Att) x 2 (Norm) ANOVA on MCj revealed a significant (p < .023) Att main effect while the Norm main effect was marginally significant (p < .079). Subjects were more motivated to comply with the group when (1) the group possessed financial reward power (i.e., Norms) than when it lacked such power (i.e., Normw) and (2) the brand recommended by the group (brand A) was an optimal choice (i.e., Attf) than when it was a suboptimal selection (i.e., AttU).

Analysis of Causal Configurations

The data were analyzed by performing path analysis using LISREL IV (Joreskog and Sorbom 1978). A major advantage of using LISREL over other path analysis approaches is that it provides a statistical measure of the entire causal model's tenability. This is the chi-square goodness of fit measure described in Joreskog and Sorbom (1978). If the obtained chi-square value is statistically nonsignificant, the model being analyzed fits the data well (i.e., the model generated correlation matrix corresponds closely to the raw data correlation matrix). On the other hand, if the obtained chi-square value is statistically significant, the model does not fit the data well, and alternative models should be investigated. Bentler and Bonett's (1980) index of incremental fit (A) was also calculated for each causal model. An index value above .90 indicates that there is little room for improvement in model fit compared to a null model of-complete independence.

Subjects' choice behavior during the experiment involved either brand A or brand F. Thus, it was possible to examine the competing causal models for both brands. Table 1 contains the parameter estimates and the corresponding critical ratios for the individual paths of the causal model hypothesized by Fishbein (hereafter referred to as the standard model) as depicted in Figure A. An observed critical ratio equal to or exceeding a value of 1.65 would indicate that the path is statistically significant at the .05 alpha level (one-tail test). The chi-square test indicated that the standard model did not provide an acceptable representation as it was rejected for both brand A [x2(9) = 36.53, p < .001, t = .74] and brand F [x2(9) = 17.06, p < .049, E = .86] Interestingly, the SNB_jMC_j -> SN path did not attain statistical significance for either brand

TABLE 1

PARAMETER ESTIMATES FOR STANDARD MODEL

Contrary to the findings involving the Fishbein standard model, the proposed causal model provided an acceptable fit for brand F [X2(8) = 8.45, p > .39, G = .93]. This was not the case, however, for brand A as the proposed model was rejected [X2(8) = 25.23, p < .C02, a = .82]. Nonetheless, a comparison of the chi-square values for the competing configurations revealed that the proposed model was a superior representation of the causal relationships for brand A [X2(l) = 11.30, p < .001] and brand F [X2(l) = E.61, p < .01].

TABLE 2

PARAMETER ESTIMATES FOR PROPOSED MODEL

While the preceding findings indicate that the changes in model relationships contained in the proposed configuration as a group improved the model fit, it is also useful to examine the results for the individual paths. These results are summarized in Table 2. Recall that the two competing configurations differ considerably in their treatment of relationships involving AB. Specifically, the potential for SNB_jMC_j to have a direct effect on AB as well as the possibility of AB having a causal impact on SN are recognized only by the proposed model. Evidence concerning the former relationship is mixed as the path estimate attained statistical significance (p < .05) for only brand F. Clear support, on the other hand, was provided for the AB y SN path which was significant (p < .01) across both brands.

The proposed model, unlike the standard model, does not include a path between SNB_jMC_j and SN. Support for this alteration was provided by the results for the standard model (Table 1) as this path did not attain statistical significance for either brand. Similarly, the addition of this path to the proposed configuration did not produce a significant improvement in model fit for either brand A [X2(l) = 0.57, p > .1] or brand F [X2(l) = 0.55, p > .1].

In order to further explore the validity of the previously raised arguments that SN does not incorporate MCj, the relative accuracy of SNB_jMC_j versus SNBj in the prediction of SN was compared for each of the seven brands. If the SNB_jMC_j -> SN relationship is valid, then weighting NBj by MCj should not decrease the prediction of SN. Table 3 contains the correlations between SN - SNB_jMC_j and SN SNBj for each brand. Whereas the correlation between SN and SNB_jMC_j approached statistical significance (p < .1) for only brand F, the SN - SNBj correlation was statistically significant (p < .05) for every brand. Further, the results of a z test for the difference between two correlations from the same sample (Roscoe 1975) revealed that weighting NBjMCj decreased (p < .05) the prediction of SN for six of the seven brands (Table 3). These findings support our hypothesis that MCj is not an antecedent of SN.

The final causal issue examined was the direct effect of AB on behavior. Analysis of the proposed model including the AB -> B path yielded conflicting results concerning the need for this relationship. Inclusion of this path led to a level of fit that approached statistical acceptance for brand A [x2(7) = 12.34, p < .091, t = .91] and produced a significant improvement in model fit relative to the proposed configuration [X2(l) = 12.89, p < .001]. In contrast, the fit for brand F involving the model version including the AB + B path [x2(7) - 6.1G, p > .52, L = .95] did not significantly [x2 (1) = 2.35, p > .13 differ from that obtained for the proposed model. Consistent with these findings, the AB b B path estimate was significant (p < .05) for only brand A. Thus, our expectation that A8 would not have a direct effect on behavior was only partially supported. Possible reasons for this result are discussed in the following section.

TABLE 3

CORRELATIONS BETWEEN THE NORMATIVE MEASURES

CONCLUSION

Limitations of the Study

Several caveats should be raised with respect to the present research. First, the results are based on relationships observed under formation conditions in a highly controlled experiment. These conditions were designed to induce the variation in motivation to comply necessary for attaining a valid test of the relationship between SNB_jMC_j and SN. Whether similar findings will occur in different settings requires further research. One can anticipate that future studies examining the SNB_jMC_j -> SN relationship under conditions where the person is motivated to comply with referent others are in fact unlikely to replicate our findings involving this normative relationship. We have argued, however, that our procedures provide greater insights into the relationships underlying the model's constructs.

Second, in comparison to other causal investigations, the sample size of the present study was relatively small, which lowers the likelihood that a causal model will be rejected. This consideration should be tempered by the fact that Fishbein's causal model was rejected in favor of an alternative configuration. In addition, the incremental fit index suggested that there was little room for improvement in the proposed causal model.

Finally, with the exception of Ag, there was only one measure of each construct. Consequently, the potential influence of measurement error could not be controlled such as in the case of the analysis of latent variables. This measurement limitation may in fact explain the direct effect AB had on behavior for one of the two brands. While BI was assessed by a single scale, AB was measured by the summed response to four scales. Thus, it is likely that AB possessed less measurement error relative to BI. Consequently, our findings concerning the AB -> B relationship should be viewed as tentative.

Summary and Implications

This research reports the first overall test of the entire causal system hypothesized by Fishbein. The present findings indicate that Fishbein's causal model, when compared to the proposed configuration, provides an inferior representation of the relationships occurring in this study. These inadequacies in the Fishbein configuration involved both the lack of needed relationships and the inclusion of inappropriate relationships. Concerning the former, our results suggest the need for AB -> SN and SNB_jMC_j -> AB paths, although the failure to obtain statistically significant estimates across both brands for the latter path does constrain the strength of this recommendation.

Perhaps most compelling is the evidence regarding the SNB_jMC_j -> SN relationship. Contrary to Fishbein's position, this path was not statistically significant for either brand. Further, while NBi was significantly correlated with SN, weighting NBj by MCj decreased the prediction of SN. This result indicates that SCj does not play a role in the formation of SN.

Although the proposed configuration "outperformed" Fishbein's causal model in the present study, we do not believe that this result is generalizable across all behavioral settings. Indeed, our proposed model was developed strictly for behaviors possessing certain key properties (e.g., novel behaviors) which were generated by the experimental setting and procedures. We cannot, of course, determine whether these behavioral characteristics are, in fact, responsible for the superiority of the proposed model since we did not systematically manipulate these characteristics and track their effects on the causal relationships. Nonetheless, we believe that the search for a single causal model that is "universally applicable" holds little promise. What seems more likely is that several causal models will emerge, each of which providing a valid representation of the causal relationships among Fishbein's constructs under a particular set of conditions. It is toward this end that we would encourage future research.

REFERENCES

Ahtola, Olli T. (1976), "Toward a Vector Model of Intentions," in Advances in Consumer Research, Vol. 3, ed. Beverlee B. Anderson, Cincinnati, Ohio: Association for Consumer Research, 481-484.

Ajzen, Icek (1971), "Attitudinal vs. Normative Messages: An Investigation of the Differential Effects of Persuasive Communications on Behavior," Sociometry, 34, 263-280.

Ajzen, Icek and Martin Fishbein (1970), "The Prediction of Behavior from Attitudinal and Normative Variables," Journal of Experimental Social Psychology, 6, 466-487.

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

Ajzen, Icek and Martin Fishbein (1974), "Factors Influencing Intentions and the Intention-Behavior Relation," Human Relations, 27, 1-15.

Bagozzi, Richard P. (1981), "Attitudes, Intentions and Behavior: A Test of Some Key Hypotheses," Journal of Personality and Social Psychology, 41, 607-627.

Bagozzi, Richard P. (1982), "A Field Investigation of Causal Relations Among Cognitions, Affect, Intentions, and Behavior," Journal of Marketing Research, 19, 562-583.

Bentler, Peter M. and Douglas G. Bonett (1980), "Significance Tests and Goodness-of-Fit in the Analysis of Covariance," Psychological Bulletin, 88, 588-606.

Bentler, Peter M. and George Speckart (1979), "Models of Attitude-Behavior Relations," Psychological Review, 86, 452-464.

Bentler, Peter M. and George Speckart (1981), "Attitudes 'Cause' Behavior: A Structural Equation Analysis," Journal of Personality and Social Psychology, 40, 226-238.

Bonfield, E.H. (1974), "Attitude, Social Influence, Personal Norm, and Intention Interactions as Predictors of Brand Purchase Behavior," Journal of Marketing Research, 11, 379-389.

Bowman, Carole H. and Marvin Fishbein (1978), "Understanding Public Reactions to Energy Proposals: An Application of the Fishbein Model," Journal of Applied Social Psychology, 8, 319-340.

Davidson, Andrew R. and James J. Jaccard (1975), "Population Psychology: A New Look at an Old Problem," Journal of Personality and Social Psychology, 31, 1073-1082.

Fishbein, Martin (1976), "Extending the Extended Model: Some Comments," in Advances in Consumer Research, Vol. 3, ed. Beverlee B. Anderson, Cincinnati, Ohio: Association for Consumer Research, 491-497.

Fishbein, Martin and Icek Ajzen (1975), Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research, Reading, MA: Addison-Wesley.

Fishbein, Martin and Icek Ajzen (1981), "On Construct Validity: A Critique of Miniard and Cohen's Paper," Journal of Experimental Social Psychology, 17, 340-350.

Glassman, Myron and M. Birchmore (1974), "The Relationship Between Subjective Norms and Normative Beliefs," Unpublished manuscript, University of Illinois.

Glassman, Myron and Nancy Fitzhenry (1976), "Fishbein's Subjective Norm: Theoretical Considerations and Empirical Evidence," in Advances in Consumer Research, Vol. 3, ed. Beverlee B. Anderson, Cincinnati, Ohio: Association for Consumer Research 677-680.

Jaccard, James, Richard Knox, and David Brinberg (1979), "Prediction of Behavior from Beliefs: An Extension and Test of a Subjective Probability Model," Journal of Personality and Social Psychology, 37, 1239-1248.

Joreskog, Karl G. and Dag Sorbom (1978), Lisrel IV: Analysis of Linear Structural Relationships by the Method of Maximum Likelihood, Chicago: National Educational Resources.

Kelman, Herbert C. (1974), "Compliance, Identification, and Internalization: Three Processes of Attitude Change," in Readings in Attitude Change, eds. Samuel Himmelfarb and Alice H. Eagly, New York: John Wiley, 218-227.

King, George William (1975), "An Analysis of Attitudinal and Normative Variables as Predictors of Intentions and Behavior," Speech Monographs, 42, 237-244.

King, George William and James J. Jaccard (1973), "The Relationship between Behavioral Intention and Attitudinal and Normative Variables," Paper presented at the Speech Communication Association, New York.

Lutz, Richard J. (1975), "Changing Brand Attitudes Through Modification of Cognitive Structure," Journal of Consumer Research, 1, 49-59.

Lutz, Richard J. (1976), "Conceptual and Operational Issues in the Extended Fishbein Model," in Advances in Consumer Research, Vol. 3, ed. Beverlee B. Anderson, Chicago: Association for Consumer Research, 469-476.

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

Miniard, Paul W. and Joel B. Cohen (1981), "An Examination of the Fishbein-Ajzen Behavioral Intentions Model's Concepts and Measures," Journal of Experimental Social Psychology, 1 7, 309-339.

Pomazal, Richard J. and James D. Brown (1977), "Understanding Drug Use Motivation: A New Look at a Current Problem," Journal of Health and Social Behavior, 18, 212-222.

Pomazal, Richard J. and James J. Jaccard (1976), "An Informational Approach to Altruistic Behavior," Journal of Personality and Social Psychology, 33, 317-326.

Roscoe, J.T. (1975), Fundamental Research Statistics for the Behavioral Sciences, New York: Holt, Rinehart & Winston.

Ryan, Michael J. (1977), "An Experimental Test of the Expanded Behavioral Intention Model," Unpublished manuscript, Columbia University.

Ryan, Michael J. (1982), "Behavioral Intention Formation: A Structural Equation Analysis of Attitudinal and Social Influence Interdependency," Journal of Consumer Research, 9, 263-278.

Ryan, Michael J. and E.H. Bonfield (1975), "The Fishbein Extended Model and Consumer Behavior," Journal of Consumer Research, 2, 118-136.

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

Songer-Nocks, Elaine (1976), "Situational Factors Affecting the Weighting of Predictor Components in the Fishbein Model," Journal of Experimental Social Psychology, 12, 56-54

Warshaw, Paul R. (1980), "A New Model for Predicting Behavioral Intentions: An Alternative to Fishbein," Journal of Marketing Research, 17, 153-172.

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