The Extended Fishbein Model: Additional Insights...And Problems


Michael J. Ryan and E. H. Bonfield (1975) ,"The Extended Fishbein Model: Additional Insights...And Problems", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 265-284.

Advances in Consumer Research Volume 2, 1975      Pages 265-284


Michael J. Ryan, The University of Alabama

E. H. Bonfield, The University of Alabama

The model was derived from Dulany's theory of propositional control as a means of studying relationships among attitude, behavior, and other variables. Behavioral intention is viewed as a mediator between behavior and attitude and social influence. Theoretical development and research testing the model are reviewed. In addition to showing support for the model, methodological shortcomings, in particular selection of attitude and social influence measures and use of single item measures, are described. The measures which appear to furnish the best statistical fit may not provide the greatest diagnostic power. A reconceptualization of the model indicating causal links is proposed providing a possibility for both best fit and high diagnostic power.


Much of the discussion and issues involved in the controversy surrounding multi-attribute attitude models has resulted from interest in Rosenberg's (1956) attitude model which has been rekindled largely through the work of Fishbein (1963). Although a great deal of attention has been given the attitude model, as attested by the review of Wilkie and Pessemier (1973), relatively little has been published relative to Fishbein's (1967) extended or intentions model which is an adaptation of Dulany's (1968) theory of propositional control

Interest in Fishbein's extended model is twofold. First, the model provides a basis for studies of the relationships between attitudes and social influence variables relative to behavior. Second, the model may be useful for the prediction of behavior utilizing behavioral intentions as a mediator. Much of the early research based on the intentions model was encouraging since intentions and behavior were frequently correlated at levels above .85 (Fishbein, 1973).

The purpose of this paper is to review the theoretical framework of the extended Fishbein model along with the research concerning it. The primary focus is on the issues associated with use of the model in consumer research, in particular the conceptual antecedents which underlie the model.


The Fishbein model is an adaptation of the central theoretical statement contained in Dulany's theory of propositional control. Fishbein's adaptation basically extended the laboratory work of Dulany to a social psychological framework In order to identify the model's conceptual antecedents Dulany's work is briefly outlined and Fishbein's adaptation explained The operational procedures used in marketing and social psychology are then discussed.

Dulany's Theory of Propositional Control

Dulany's theory forms a network of propositions about the effect of reinforcement parameters on verbal responses. Knowledge, beliefs, and awareness are represented as propositions in the form of two hypotheses: (1) the Response Hypothesis (RH), that is, the individual's hypothesis concerning the expectation of a reinforcement, and (2) the Behavioral Hypothesis (BH), that is, the individual's hypothesis concerning the congruence of a response with group norms. Associated with the Response Hypothesis is a feeling of value called the Subjective Value of a Reinforcer (RSv). Likewise, associated with the Behavioral Hypothesis is an evaluative feeling termed the Motivation to Comply (MC). Dulany contends there are many additional variables which affect behavior. However, these other variables have only an indirect effect. They are exogenous to his model and their influence on behavior is reflected in the model's endogenous variables.

The variables included by Dulany reflect specific actions and situations and are proposed to predict and explain behavioral intention (BI). To the extent that the independent variables are specific to a given act, behavioral intention approximates overt behavior.

Through a lengthy process of deductive reasoning, Dulany models his theory to predict an individual's particular verbal response or class of responses in a given situation as

B = BI = [(RHd) (RSv)] w0 + [(BH) (MC)] w1

where B = overt behavior; BI = behavioral intention; RHd = hypothesis of the distribution of reinforcement, i.e., the degree to which the individual thinks a specific response will lead to reinforcement or reward; RSv = the subjective value of a reinforcer, i.e., the value the individual places on the reward; BH = behavioral hypothesis, i.e., the degree to which the individual believes a particular behavior is expected of the individual by some specified or generalized set of others; MC = motivation to comply, i.e., the degree of the individual's desire to conform to a BH; and w0 and w1 are empirically determined weights.

Dulany's empirical work was carried out in laboratory experiments concerned with verbal conditioning and concept attainment. As such, there was only one RHd and BH and their respective RSv and MC. These variables were generally manipulated by an experimenter who, within the context of the experiment, was the only influence source. Dulany (1968) reports several tests of the model. These tests support the assumption of additivity of independent variables as well as the need to include BI as a moderator; The independent variables accounted for a large proportion of the variance in BI (50-77%) and BI accounted for a large proportion of the variance in B (80-888).

The Fishbein Adaptation

Fishbein restated Dulany's theory as follows:

An individual's intention to perform a specific act, with respect to a given stimulus object, in a given situation, is a function of the following:

(1a.) His beliefs, Bi, about the consequences of performing a particular behavior (in a given situation), that is, the probability or improbability that the performance of behavior x will lead to some consequence Yi

(1b.) The evaluative aspect, ai, of Bi, that is, the S's evaluation of Yi

(2a.) A normative belief, that is, the S's belief about what he should do in this situation (NB).

(2b.) His motivation to comply with the norm, that is, his desire, or lack of desire, to do what he thinks he should do (MC). (Fishbein, 1967, p. 488)

In the above restatement, item (la) or Bi approximates RHd, item (lb) or ai approximates RSv, item (2a) or NB approximates BH, and item (2b) approximates MC. Fishbein thus shows the first component, (RHd) (RSv), of the Dulany theory to be analogous to the summative attitude model associated with him


where Aact = the attitude toward performance of a specific act, and Bi and ai have been previously defined in Fishbein's restatement of Dulany's theory. The Aact model differs from Fishbein's Aov or attitude toward an object model, more popularly described in the marketing literature. In the Ao model, Bi statements refer to concept objects which, in marketing, are usually product attributes Bi in the Aact model, however, is associated with behavioral outcomes. Fishbein (1971) has stated the Aact conceptualization is more appropriate for predicting and understanding purchase intentions and behavior. The argument is analogous to the features-benefit position traditionally taken in marketing (Haley, 1968).

Fishbein's adaptation of Dulany's second component, (BH) (MC), is more general Social normative beliefs (SNB) are described as

...the individual's beliefs about what "society" (i.e., most other people, his "significant others," etc.) "says" he should do (i.e., a social or group norm). (Fishbein, 1967, p. 489)

This conceptualization of the second component is an expansion of Dulany's work since he conceived this variable in terms of pressure exerted by the experimenter.

Fishbein summarizes the entire adaptation

To summarize briefly then, it can be seen that in its adapted form, the theory essentially leads to the prediction that an individual's intention to perform any behavior in a given situation (and thus his actual performance of the behavior) is a function of (l) his attitude toward performing the behavior in the situation, (2) his perception of the norms governing that behavior in that situation. and (3) his motivation to comply with those norms. (Fishbein, 1972a, p. 248)

This formulation is modeled

B = BI = [Aact]w0 + [NB.MC]w1   (1)

where NB = a normative belief, the degree of belief that others-expect the individual to perform a specific act, MC = motivation to comply with the expectations of others. All other elements have been previously defined.

Although Fishbein does not use the term, BI appears to be a moderating variable. It comes between the independent and dependent variables in a causal sequence. It is viewed as a consequence of the independent variables and as a determinant of the dependent variable.

The model purports to predict and explain human behavior. Fishbein (1973) states that the model is appropriate for a number of behavioral criteria. Applications of the model are very restrictive in its present form due to a set of exogenous variables.

Generally speaking, there are three major factors that influence the size of the relationship between intention and behavior: (1) the specificity of the intentional measure; (2) the time between the measure of intention and the behavioral observation; and (3) the degree to which carrying out the intention is completely under the individual's control. (Fishbein, 1973, p. 15).

Fishbein considers the first of these factors to be of primary importance. To put it more simply, the more removed the measure Of intention is from the criteria to be predicted, the poorer will be the prediction. The second and third restrictions have also been considered in the more general consumer behavior models (e.g., Howard and Sheth, 1969; Andreason, 1965) and by surveyors of consumer intentions to purchase durables (Clawson, 1971; Juster, 1966). To date, no attempts have been made to make these variables endogenous to Fishbein's model. Also, the attitudinal and social influences must be situation specific. It is not expected that general attitude or social influence measures will predict behavioral intention.

In addition to form (1), Fishbein has offered two basic alternative formulations of the model. The first alternative form involves a departure from Dulany's conceptualization in which the possibility of adding a third component, personal normative beliefs (NBp), is considered. This component is described as

The individual belief about what he personally feels he should do (i.e., a personal norm or rule of behavior). (Fishbein, 1967, p. 489)

This modified formulation is

B = [Aact]w0 + [NBs MCs]w1 + [NBp MCp]w2   (2)

where p and s refer to personal and social norms respectively. Fishbein's addition of the personal norm component was his initial method of handling the question of whether NB referred to a personal norm, a social norm, or both.

The second alternative form of the extended Fishbein model results from recognition that NBs may result from several sources.

This formulation suggests that it may be necessary to consider many different types of normative beliefs; for example; beliefs about what one's (a) parents, (b) friends, (c) co-workers, (d) religious group, etc., "says" the individual should do. (Fishbein, 1967, p. 490).

This formulation is modeled


where NBj = degree of belief that a specific act is expected by the jth group or person and MCj = motivation to comply with the expectation of the jth group or person.

Model forms (1), (2), and (3) represent the basic extended Fishbein model and the original alternative forms he suggested. Operationalization procedures have lead to additional forms of the model.

The forms of the model discussed state the independent variables combine additively. That is, they are orthogonal. Conceptually, these constructs may have separable effects on BI. On the other hand, these constructs may not be completely independent. The notion of perceptual distortion, for example, is based on an interrelationship between attitude and social influences.


Fishbein's operationalization procedures follow two basic steps: (1) determining the salient Bi outcomes and NBj groups and (2) developing techniques for measuring the constructs.

Determining Saliency

Fishbein (1963) determines salient behavioral outcomes according to procedures developed by Maltzman, Bogartz, and Breger (1958). Subjects are asked to give different associations to the same stimulus words in a free association situation. When Aact measures are being developed, the stimulus words refer to behaviors rather than attitude objects. The technique precludes probing, prompting, or presentation of predetermined lists of behavioral outcomes or potential associations in order to avoid triggering sets of interdependent beliefs based on social desirability rather than true feelings (Cowling, 1973b; Fishbein, 1971). Based on research on information processing and span of attention, Fishbein (1971) states there are probably only 5 to 9 behavioral outcomes that serve as the primary determinants of a person's attitude at any point in time.

...Saliency refers to the fact that the respondent is aware of or conscious of the attribute, that it's on the "tip of his tongue." In other words, it has a high probability of being elicited by the respondent.

...If we really want to know the determinants of attitude we have to know the person's salient beliefs. Unfortunately, there is at the present time no independent way of assessing salience outside of order of elicitation, and even more importantly, a non-salient belief may be just as good as an indicant of a person's attitude as a salient belief. Thus at present, outside of using a direct elicitation procedure, there is no way of telling whether you've obtained salient or nonsalient beliefs. (Fishbein, 1971, pp. 313-314).

The relevant beliefs, or behavioral outcomes, and reference groups, when determined according to the elicitation procedure, are obtained from the same population being studied (e.g., Jaccard and Davidson, 1972). Fishbein believes this procedure is more theoretically sound than either "in depth probing" or using factor or regression analysis to analyze a predetermined list of attributes or outcomes.

Construct Operationalization

Behavior (B), has been generally operationalized as the observation of an individual's choice in a specific situation. For example, Bonfield (1974) used consumer dairies to obtain self reports of fruit drink purchases. Harrell and Bennett (1974) used prescription records provided by a trade association.

The procedures used to measure the remaining constructs have been based on semantic differential techniques (Osgood, Suci, and Tannenbaum, 1957).

Behavioral intention (BI) was usually operationalized with a seven place scale from probable to improbable. For example (Jaccard and Davidson, 1972)

I intend to use birth control pills

probable _:_:_:_:_:_:_ improbable

A disadvantage here is that such one item instruments tend to be unreliable. Lutz (1973b) used multiple item scales such as

When it is introduced onto the market, I intend to try Brand M detergent.

likely _:_:_:_:_:_:_ unlikely

probable _:_:_:_:_:_ improbable

possible _:_:_:_:_:_ impossible

When multiple item scales are utilized, it is possible to sum scores over all scales or take an average as either scoring method will yield the same correlations. Averages are preferable if results are to be compared with studies using single item scales.

Aact has been operationalized as the sum over four semantic differential type items. For example (Ajzen and Fishbein, 1970)

Choosing (alternative X) is

foolish _:_:_:_:_:_:_ wise

good _:_:_:_:_:_ bad

harmful _:_:_:_:_:_ beneficial

rewarding _:_:_:_:_:_ punishing

Bi, ai, NBj, and MCj respetively have usually been operationalized with one item semantic differentials.  For example (Jaccard and Davidson, 1972)

Beliefs about the act (Bi)

Using birth control pills would affect my sexual morals

probable _:_:_:_:_:_ improbable

Evaluation of Consequence (ai)

Having my sexual morals affected is

good _:_:_:_:_:_ bad

Normative Beliefs (NBj)

My mother thinks I should use birth control pills

probable _:_:_:_:_:_ improbable

Motivations to Comply (MCj)

With respect to sexual behavior

I want very much to _:_:_:_:_:_ I want very much not to

do as my mother thinks

A single expectancy times an evaluative or motivation to comply statement has usually been included for each salient consequence and group. That is, there were n sets of cognitive statements and k sets of group statements each set being summed to yield underlying structure scores.


A detailed review of empirical tests of the model in the social psychological literature has been provided by Ajzen and Fishbein (1973). This body of research suggests the model is an accurate predictor of a wide range of behavioral intentions and behavior. The need to include BI as a moderator was supported. The traditional attitude toward an object, AoJ measure was related to overt behavior only to the extent it affected either the attitudinal, Aact, or normative component. In all cases Aact was operationalized as the sum of the EBiai and Aact scores.

The majority of these studies operationalized the normative component as the sum of separate sources of influence, see (3), with two studies (Ajzen and Fishbein, 1969 and 1970) utilizing generalized other as a single normative referent.

The research reviewed by Ajzen and Fishbein does not clearly indicate whether NBp should be included in the model. Ajzen and Fishbein (1973) indicated empirical findings have shown NBp to be an alternative measure of BI. In recent writings Fishbein (e.g. 1973) only described the model in form (3) which excludes NBp. Also, Ajzen and Fishbein (1969) reported significant beta weights for Aact, NBp and NBs supporting the model with three independent variables (2). However, in a marketing application, Bonfield (1974) found NBp to act as a suppressor variable that caused Aact to become statistically insignificant

Although model (3) appears to be the appropriate model for adaptation in marketing, the operationalization of Aact as a four-item semantic differential type scale rather than as EBiai reduces the diagnostic power of the model.

Studies in Marketing

A number of marketing studies testing the model have been conducted in Great Britain (Tuck, 1973). The majority of the work has been done commercially, however, and detailed results have been held as proprietary. Summary results from five of these studies have been published (Cowling, 1973a, b; Bright and Stammers, 1971; Bruce, 1971; and Tuck and Nelson, 1969). The results of these studies suggest elicitation of salient behavioral outcomes for specific segments and products results in higher correlations between Aact and EBiai than using predetermined lists. Aact was more highly correlated with BI than either Ao or EBiai. These findings support Fishbein's contention that Aact is a more appropriate predictor of BI than Ao and that elicitation techniques are necessary. Finally, the British studies showed relative weights associated with attitudinal and normative components of the model vary across products and market segments. Stronger correlations between BI and Aact versus BI and Ao have also been reported by American market researchers (Weddle and Bettman, 1973; Mathews, et. al., 1974).

Nine U.S. studies have been reported in which a form or adaptation of the extended model has been tested in a marketing context. The major results of these studies are summarized in Table I. The model appears to have value in predicting and explaining variance in intentions and behaviors over a wide range of purchase intentions and purchase behavior. Harrell and Bennett (1974) tested the model with respect to physician prescribing behavior, Ryan (1973) with respect to intentions toward purchase of products of automobile manufacturers, Ryan (1974), Bonfield (1974), Lutz (1973), and Mathews, Wilson and Harvey (1972) with respect to consumer convenience goods, Weddle and Bettman (1973) with respect to purchasing underground term papers, and Wilson, Mathews and Monoky (1973) with respect to bargaining in a personal selling situation. Harrell and Bennett (1974), Bonfield (1974), Mathews, et. al. (1974), and Wilson, et. al. (1973) reported correlations between intentions and behavior while others did not report measuring behavior.



The average correlation between BI and B across these studies was .435.   The average multiple correlation of attitude and social influence on BI was .62. Three of these studies (Harrell and Bennett, 1974; Mathews, et. al., 1974; Lutz 1973a) also employed cross validation procedures. Correlations obtained using cross validation were comparable to those above indicating the model is a stable predict-or.

Using R2 as a criterion, the predictive power obtained in the-marketing studies has been generally lower than obtained in the social psychology studies. A possible explanation may be found in the different types of behaviors and attitudes examined. The social psychology studies were predominantly concerned with potentially central attitudes (e.g., attitudes toward behavior involving racial or religious beliefs). The marketing studies examined purchase activities which may have involved noncentral attitudes. Consequently, the marketing behavior may have had smaller associations with attitudes and social influence because of their lack of centrality to the individual. Bonfield (1974) found the deterministic influence was stronger when subjects perceived the product as important. Ryan (1974) also found the deterministic influence was stronger for automobiles than toothpaste. An automobile purchase would be expected to be more central than a toothpaste purchase.

In an attempt to explain why marketing behavioral models have generally accounted for a small proportion of systematic variance, Bass (1974) has demonstrated that purchase behavior is predominantly a stochastic process. The emerging pattern in the studies discussed here suggests the relative influence of deterministic and stochastic elements may vary across products.

Bonfield (1974) also found the relative influence varied across different segments. In addition to perceived product importance, high education and low brand loyalty groups exhibited stronger deterministic influences than low education and high brand loyalty groups respectively. Wilson, et. al. (1974) accounted for more variation in housewives' purchase intentions toward toothpaste brands than did Ryan (1974) among college undergraduates. The housewives may have more carefully considered toothpaste purchase than college students.

These findings suggest claims that behavior is primarily predicted from either deterministic or stochastic influences may be premature. Rather, the relative importance of these influences may be a function of group, individual, or product characteristics. In general, the R2's found in the marketing studies have not shown the model to be a particularly good predictor. Bonfield (1974) found a naive model predicted equally well. Instead, the patterns in relative R2's associated with the model suggest the model has explanatory potential. A synthesis of a much larger body of research is needed to suggest a more definitive pattern of deterministic and stochastic influences among products and segments. The strategic implications are obvious. For example, promotional strategies based on attitude change tactics are inappropriate for products and groups whose purchase behavior is primarily stochastic.

Operationalization Differences

Operationalization of the dependent and independent components of the model has varied considerably in the marketing studies conducted in the United States. Behavior has been measured in field studies in terms of the actual brand most frequently prescribed by physicians over a period running both before and after measures of intention, attitude, and social influence had been obtained (Harrell and Bennett, 1974), brand of fruit drink purchased on the first occasion following the independent component measures (Bonfield, 1974), and choice of a free sample immediately following an interview (Mathews, et. al., 1974). Only Wilson, et. al. (1973) measured behavior as more than a dichotomous outcome, treating the variable in terms of cooperative choices in a prisoner dilemma, negotiation experiment.

Attitude has been measured closely paralleling Fishbein's Aact procedures as well as in terms of substituting other measures. As noted previously, Fishbein has used AoJ zBiai, and Aact (evaluative semantic differential type scales) in various social psychological studies. Aact was operationalized as a multi-item, semantic differential type scale by Mathews, et. al. (1974), Ryan (1973), and Wilson, et. al. (1973); Harrell and Bennett (1974), and Weddle and Bettman (1973) basically operationalized attitudes as EBiai. Ryan (1974) and Lutz (1973b) used both techniques. Bonfield (1974) substituted a factor analytic model as described by Howard Sheth (1969).

A EBiai operationalization of the attitude component appears essential to understanding since it represents the structure underlying attitude. However, operationalization in this form has presented some difficulties. Among the marketing studies using EBiai scales only Ryan (1974) used multi-item B1 and ai scales. One item scales are generally considered unreliable since there is no opportunity for random and specificity errors to average out over a set of scales. Reliability estimates should be incorporated in future correlational studies in order to ascertain whether weak correlations are due to unreliable measures.

Semantic differential type measurements deviate from the interval and ratio characteristics necessary for their applications in expectancy models (Heise, 1969). Although violations of the interval assumptions make little difference when correlating scale scores (Nunnally, 1967, pp. 24-30), multiplication of these scores does require a ratio scale in order to be meaningful (Lord and Novick 1968, p. 21). Schmidt (1973) addressed the multiplicity issue and provided empirical evidence indicating expectancy value models are not robust with respect to violating ratio scale assumptions. Scaling and operational procedures are an important area for further research which consumer researchers are beginning to explore (e.g. Bettman, Capon, and Lutz, 1974). Future studies should incorporate both Aact and EBiai operationalizations in order to provide both appropriate correlation coefficients, beta weights, and understanding the underlYing attitudinal structure.

While there is general agreement the first component of the extended model is attitude, some confusion exists as to the theoretical conceptualization of the second component. Fishbein (1967) originally conceived of it as a social normative component. More recently he has viewed it as a measure of the perceived attitude of others toward performing the behavior (Ajzen and Fishbein, 1972) as has Ryan (1974) in a marketing context. Bonfield (1974) referred to the component as an indicant of reference group or social influence.

Four strategies have been followed in operationalizing the social influence component of the model. Wilson, et. al. (1973), following an experimental strategy paralleling Dulany's work, controlled the situation so there would be only one NBjMCj source Bonfield (1974) and Harrell and Bennett (1974) used single item, generalized other NBjMCjoperationalizations. Mathews, et. al. (1974); Ryan (1973, 1974), Lutz (1970), and Weddle and Bettman (1973) utilized multi-other, NBjMCj scales where each NB-MC represented a specific source of normative influence Ryan (1974) elicited salient others as influence sources. The other studies appeared to use predetermined lists.

Ryan (1974)) following Ajzen and Fishbein (1972), also operationalized the social influence component as perceived attitude of relevant others, AactO, toward performing the behavior Ajzen and Fishbein (1972) have conceived of AactO as the attitude of others toward their, the others, performing the act not their attitude toward the individual, whose intentions are the subject of the model, performing the act Essentially, the Ajzen and Fishbein statement is of the form

Most of the people whose opinion is important to me with respect to this act think this act is

while Ryan (1974) has operationalized the statement in the form

Others think that for me to perform this act is

The same evaluative semantic differential type scales, good-bad foolish-wise harmful-beneficial, and punishing-rewarding, have been used by both Ryan and Ajzen and Fishbein.

There is no apparent behavioral expectation element in the AactO component as operationalized by Ajzen and Fishbein, although the expectation notion appears essential in the antecedent, Dulany theory In addition, their AactO measure is inconsistent with Fishbein's conceptualization of Aact which requires the attitude to be linked to the individual's performance of some behavioral act. Ryan's operationalization is consistent with the expectancy notion in the Dulany theory and the specificity requirement of the Fishbein Aact conceptualization

Disturbingly, operationalizing the second component as AactO and as NB MCj may not be equivalent. Ryan (1974) has shown correlations of less than .00 between the two measures. AactO correlated more highly with BI than did ENBjMC or than AactO correlated with ENBjMCj. Thus, AactO, operationalized as other's attitude toward the perceiver's performance may be an appropriate operationalization of the second component when prediction is the purpose of the model. The NBjMCj method of operationalization, however, provides better diagnostic power.

Additional study of the second component of the model is needed Additional evidence is needed relative to the inclusion of MC, the substitutability of AactO and NBjMCj, and construct validity of these components. Since Aact and Biai measurement; problems also apply here, it is necessary to have the same type of research scrutiny paid the normative or social influence measures.

Additivity of Independent Variables

The forms of the model discussed state the independent variables combine in an additive manner. That is, they are orthogonal. Operational measure of these constructs have been shown to have separate effects on BI. Yet, common sense suggests these constructs are not completely independent. For example, a person who perceives an act as morally correct would be expected to believe others also view this act as morally correct. Thus, on a theoretical basis it is expected that interaction as well as direct effects should be present in the model.

The number of statistically significant beta weights for Aact and ENB MC shown in Table 1 far exceed the number that are insignificant, supporting the contention that each independent variable has direct predictive power. This evidence must be viewed with caution since only Ryan (1974) used measures for both attitude and social influence that were highly and equally reliable. In Ryan's research, the beta weights associated with social influence were not statistically significant.

Evidence suggesting the independent components are not additive is shown by high correlations between independent variables; Bonfield (1974), Mathews, et. al. (1974), and Ryan (1974) found correlations between attitude and social influence which were higher than the correlation of either on BI. In an experimental setting, Ryan (1974) found changes in both Aact and AactO occurred regardless of whether Bi or NB were manipulated, thus suggesting a nonadditive relationship. More research iJs needed to investigate the additivity assumption. In addition nonadditive reformulations of the model based on intuitively appealing conceptual antecedents should be investigated.

If an additive model can be supported, beta weight analysis has implications for marketing strategies as a means of ascertaining whether brand or product purchase intentions are primarily under attitudinal or social influence control. For example, Mathews, et. al. (1974) suggested an attitudinal-social influence continuum exists analogous to an instrumental-expressive continuum used for classifying consumer products. That is, automobiles and clothing were classified as expressive products since they are indicants of social status. Thus, automobiles and clothing would be expected to have higher beta weights associated with social influence than with attitude. Laundry detergents were viewed as instrumental products in that they are relatively homogeneous, generally purchased and used for utilitarian benefits, and do not tend to carry status implications. Therefore, beta weights associated with attitude toward instrumental products would be expected to be greater than the beta weights associated with social influence. Mathews, et. al. caution that the instrumental and expressive dimensions are not mutually exclusive. The results reported by Mathews, et. al. (1974) support their hypothesis. Variation in intentions concerning toothpaste purchases explained by attitude were consistently lower than explained by social influence, but the beta weights associated with attitude for the cosmetic brands (Ultra Brite, MacLeans, and Pepsodent) were lower than the beta weights associated with attitude for the noncosmetic brands and social influence was relatively stronger with respect to the cosmetic brands (See Table 1). Wilson, ct. al. (1'973) found situational influences affected the relative importance of attitudes and social influences. Attitudinal influences predominated, but social influences were stronger among perceived similar buyer-seller dyads.

The Relationship Between Behavior and Behavioral Intentions

Sheth (1974) has defined behavior as a function of behavioral intention and those situational factors that could not be predicted by the individual at the time of verbally expressing his behavioral intention. In the studies where BI and B measures were contiguous or nearly contiguous, BI-B correlations were high (Mathews, et. al. 1974; and Wilson, et. al. 1973).In the studies where measures of B and BI were allowed to vary, the correlations were much lower (Bonfield, 1974; and Harrell and Bennett, 1974) Future studies should examine situational variables intervening between measures of B and BI Perhaps the procedures developed by Sheth (1973) could be adapted to the present model

Three social psychology studies (Darroch 1971; Fishbein, et al 1970; Ajzen and Fishbein, 1970) have furnished evidence indicating the need to include BI as a moderator variable even though situational variables were controlled These findings suggest understanding purchase intentions may be necessary for understanding purchase behavior Consumer researchers should address this issue


The theoretical development of the extended Fishbein model along with the empirical evidence gathered testing it leads to a number of causal inferences Consistent with Fishbein's earlier work, the independent attitude variable is Aact not EBiai. Evaluation and beliefs explain, that is, come before Aact Although Fishbein does not use the term antecedent variable, EBiai fits the use prescribed by Rosenberg (1968) for this term, that is, attempting to discover first causes.

The antecedent variable is a true effective influence; it does not explain away the relationship between the independent and dependent variables but clarifies the influences which precede this relationship (Rosenberg, 1968, p. 66)

Following the same reasoning used to describe the two forms of attitude as antecedent and independent variables, it appears reasonable to suggest ENBjMC represents an underlying structure antecedent to AactO.

Although Fishbein does not use the term, BI appears to be a moderating variable. It comes between the independent and dependent variables in a causal sequence. It is viewed as a consequence of the independent variables and as a determinant of the dependent variable That is, it is a necessary link in the causal chain.

Finally, the empirical evidence reviewed here suggests substantial interdependence among the variables. That is, changes in ENBjMC are not only expected to lead to changes in Aact and BI, but also in Aact while changes in Biai are not only expected to lead to changes in Aact and BI, but also in AactO. In addition, changes in Aact are expected to lead to changes in Aact and vice versa. These causal influences are indicated in Figure 1.

There are three generally accepted demonstrations inferring causal relationships. First, there should be an association between the independent and dependent variables. Second, there should be causal order. That is, the independent variables should be shown to occur prior to the dependent variables. Third, the association should not be spurious. That is, the association between variables should be shown not to disappear when the influences of rival causal variables are removed. One has only to break one of these three links to establish noncausality (Popper, 1959).

In order to test this reconceptualization of the model, two research strategies are suggested. The first involves the use of path analysis as a means of determining whether empirical evidence will support the relationships in Figure 1. Secondly, experimental designs should be utilized to test the causal order assUmptions which underlie path analysis. If the associations and causal order Can be supported, Competing variables should be introduced as antecedents in order to test for spuriousness in the relationships.



One avenue of potential future research has already been suggested in terms of the reconceptualization of the model. In addition, the methodological shortcomings described throughout the discussion of empirical evidence lead to a call for better research conceptualization where possible.

Specifically, whenever the underlying StrUCtUre of Aact and AactO are being studied, it seems necessary to elicit behavioral outcomes and salient others for each behavior and homogeneous target group.

One item scales, being potentially unreliable, should be replaced by multi-item scales for Bi, ai, NBi, and WMCj. In addition, Care should be taken in use of evaluative, Semantic differential type scales to measure Aact and AactO. Only evaluative scales should be used in these measures as ascertained by factor analysis sinCe it is only the evaluative scales that have been shown to correlate With standard attitude measures (Osgood, et. al., 1957). Enough semantiC items should be included to adequately represent evaluative semantic space and provide reliable scales.

Validation of attitude and social influence measUres are needed. Although, .Ract is generally measured Using semantic differential type scales, the measure has not been validated utilizing a standard semantic differential bipolar adjective set. In addition, EBiai measures have never been shown to correlate as high with semantic differential evaulative scales as the latter scales have been shown to correlate With Thurstone equal-appearing interval scales. Unfortunately, validating constructs for the social influence component are not so clearly identifiable sinCe more conceptual work is needed as a starting point.

Few studies have included cross validation techniques in model parameter estimates. Cross validation should be a standard procedure in correlational tests of the extended model in order to check for stability.

Some question still exists relative to the appropriate method for scoring Bi, ai, NB. and MC. scales. The only work reported testing coding schemes has shown bipolar Scales to be superior to unipolar scales while also supporting the multiplication of Biai scales (Bettman, Capon, and Lutz, 1974).

The body of research based on the extended Fishbein model is now considerable and the results of tests suggest the model has value for the understanding of a wide range of purchase behavior and behavioral intentions. However, the validation process has just begun. In addition to the possible avenues of research pointed out, replications are needed and the range of purchase behavior and subject and respondent groups needs to be expanded. Hopefully, future research will build on what has been done in the past so a cohesive body of knowledge will continue to develop.


Ajzen, I. & Fishbein, M. The prediction of behavioral intentions in a choice situation. Journal of Experimental Social Psychology, 1969, 5, 400-416.

Ajzen, I. & Fishbein, M. The prediction of behavior from attitudinal and normative variables. Journal of Experimental Psychology, 1970, 6, 466-487.

Ajzen, I. & Fishbein, M. Attitudes and normative beliefs as factors influencing behavioral intentions. Journal of Personality and Social Psychology, 1972, 21, 1-9.

Ajzen, I. & Fishbein, M. Attitudinal and normative variables as predictors of specific behaviors. Journal of Personality and Social Psychology, 1973, 27, 41-57.

Andreason, A. R. Attitudes and consumer behavior: A decision model. In L. Preston (Ed.), New research in marketing. Berkeley: University of California, Institute for Business and Economic Research, 1965, 1-16.

Bass, F. M. The theory of stochastic preference and brand switching. Journal of Marketing Research, 1974, 11, 1-20.

Bettman, J. R., Capon, N. & Lutz, R. J. Cognitive algebra in multi-attribute attitude models. Working paper no. 10, Center for Marketing Studies, University of California, Los Angeles, February 1974.

Bonfield, E. H. Attitude, social influence, personal norms, and intention interactions as related to brand purchase behavior. Journal of Marketing Research, 1974, in press.

Bright, B. E. & Stammers, H. R. Advertising, promotion and medical detailing. Proceedings: Seminar on pharmaceutical research, European Society of Market Research. 1971.

Bruce, J. First experiences with Fishbein theory and survey methods. Proceedings: Seminar on translating advanced advertising theories into research reality. Amsterdam, The Netherlands: European Society of Market Research, 1971, 25-39.

Clawson, C. J. How useful are 90-day purchase probabilities? Journal of Marketing, 1971, 35(4), 43-47.

Cowling, A. B. Determining and influencing consumer purchase decision. European Research, 1973a, 1, 26-31, 34.

Cowling, A. B. Use of elicitation technique for producing dimension of brand choice. Proceedings. British market research society, 1973b, 139-156.

Darroch, R. K. The prediction of intention and behavior: A comparison of theoretical models of the attitude-behavior relationship using a photographic release signing situation. Unpublished doctoral dissertation, University of Illinois, 1971.

Dulany, D. E. Awareness, rules, and propositional control: A confrontation with S-R behavior theory. In D. Horton and T. Dixon (Eds.), Verbal behavioral and general behavior theory. New York: Prentice-Hall, 1967, 340-387.

Fishbein, M. A. An investigation of the relationships between beliefs about an object and the attitude toward that object. Human Relations, 1963, 16, 233-239.

Fishbein, .M. A. Attitude and the prediction of behavior. In M. Fishbein (Ed.), Readings in attitude theory and measurement. New York: Wiley, 1967, 477-492.

Fishbein, M. A. Some comments on the use of "models" in advertising research. In Proceedings: Seminar on translating advanced advertising theories into research reality. Amsterdam: The Netherlands: European Society of Market Research, 1971.

Fishbein, M. A. The search for attitudinal-behavior consistency. In J. B. Cohen (Ed.), Behavioral science foundations of consumer behavior. New York: Free Press 1972, 245-252.

Fishbein, M. A. The prediction of behavior from attitudinal variables. In C. D. Mortensen and K. K. Sereno (Eds.), Advances in communication research. New York: Harper and Row, 1973, 3-31.

Haley, R. I. Benefit segmentation: A decision oriented research tool. Journal of Marketing, 1968, 32(3), 30-35.

Harrell, G. D. 4 Bennett, P. D. An evaluation of the expectancy value model of attitude measurement for physician prescribing behavior. Journal of Marketing Research, 1974, 11, 269-278.

Heise, D. R. Some methodological issues in semantic differential research. Psychological Bulletin, 1969, 72, 406-422.

Howard, J. A. & Sheth, J. N. The theory of buyer behavior. New York: Wiley, 1969.

Jaccard, J. J. & Davidson, A. R. Toward an understanding of family planning behaviors: An initial investigation. Journal of Applied Social Psychology, 1972, 2, 228-235.

Juster, F. T. Consumer buying intentions and purchase probability: An experiment in survey design. Journal of the American Statistical Association, 1966, 61, 658-696.

Lord, F. M. & Novick, M. R. Statistical theories of mental test scores. Reading, Massachusetts: Addison Wesley, 1968.

Lutz, R. J. A comparison of two alternative models of the attitude-behavior relationship. Paper presented at the 81st Annual Convention of the American Psychological Association, Division 23, Montreal, Canada, August 27-31, 1973a.

Lutz, R J. Cognitive change and attitude change: A validation study. Unpublished doctoral dissertation, University of Illinois at Urbana-Champaign, 1973b.

Maltzman, I., Bogartz, W., & Breger, L. A procedure for increasing word association originality and its transfer effects. Journal of Experimental Psychology, 1958, 56, 392-398.

Mathews, H. L., Wilson, D. T., & Harvey, J. W. An empirical test of the Fishbein behavioral intention model. Working paper no. 19, College of Business Administration, The Pennsylvania State University. February. 1974.

Nunnally, J. C. Psychometric theory. New York: McGraw-Hill, 1967.

Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. The measurement of meaning. Champaign: University of Illinois, Press, 1957.

Popper, K. R. The logic of scientific discovery. New York: Basic Books, 1959

Rosenberg, M. J. Cognitive structure and attitudinal effect. Journal of Abnormal and Social Psychology, 1956, 53, 367-72.

Rosenberg, M. The logic of survey analysis. New York: Basic Books, 1968.

Ryan, M. J. Purchase behavior: An explanatory theory, a predictive model. In R. L. King (Ed.), 1973 proceedings. Southern Marketing Association.

Ryan, M. J. An empirical test of a predictive model and causal chain derived from Fishbein's behavioral intention model and applied to a purchase intention situation. Unpublished doctoral dissertation, University of Kentucky, 1974.

Schmidt, F. L. Implications of a measurement problem for expectancy theory research. Organizational Behavior and Human Performance, 1973, 10, 243-251.

Sheth, J. N. A field study of attitude structure and attitude-behavior relationship. Faculty working paper no. 116, College of Commerce and Business Administration, University of Illinois at Urbana-Champaign, 1973.

Sheth, J. N. An investigation of relationships among evaluative beliefs, affect, behavioral intention, and behavior. In J. U. Farley, J. A. Howard, Q L. W. Ring (Eds.), Consumer behavior: Theory and application. Boston: Allyn & Bacon, 1974, 89-114.

Tuck, M. Fishbein theory and the Bass-Talarzyk problem. Journal of Marketing Research, 1973, 10, 345-348.

Tuck, M. & Nelson, E. N. The relationship between attitudes and behavior. In ESOMAR congress papers. Amsterdam: The Netherlands: European Society of Market Research, 1969, 257-278.

Weddle, D. E. & Bettman, J. R., Marketing underground: n investigation of Fishbein's behavioral intention model. In S. Ward and P. Wright (Eds.), Advances in consumer research Vol. 1. Urbana, Ill.: Association for Consumer Research, 1973. pp. 310-318.

Wilkie, W. L. & Pessemier, E. A. Issues in marketing's use of multi-attribute attitude models. Journal of Marketing Research 1973, 10, 428-441.

Wilson, D. T., Mathews, H. L., & Monoky, J. F. Attitude as a predictor of behavior in a buyer-seller bargaining situation: An experimental approach. American Marketing Association Combined Conference Proceedings, Spring and Fall,1972, Series No. 34. Chicago, Illinois: American Marketing Association, 1973, 390-395.



Michael J. Ryan, The University of Alabama
E. H. Bonfield, The University of Alabama


NA - Advances in Consumer Research Volume 02 | 1975

Share Proceeding

Featured papers

See More


G5. The Phenomenon of Brand Noise and Related Consumer Preferences in the Luxury Industry

Daria Erkhova, University of Bern
Elena Ehrensperger, University of Bern
Harley Krohmer, University of Bern
Wayne Hoyer, University of Texas at Austin, USA
John Zhang, University of Pennsylvania, USA

Read More


G8. How Does Pronunciation Difficulty of Brand Names Influence Consumer Responses? The Role of Self-Construal

Gunben Ceren Aksu, Rutgers University, USA
Yeni Zhou, Rutgers University, USA
Alokparna (Sonia) Monga, Rutgers University, USA

Read More


M8. Nostalgia Increases Healthy Attitudes and Behaviors

Jannine Lasaleta, Yeshiva University
Carolina O. C. Werle, Grenoble Ecole de Management
Amanda Pruski Yamim, Grenoble Ecole de Management

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.