Attitude Measurement and Behavior Change: a Reconsideration of Attitude Organization and Its Relationship to Behavior

Richard P. Bagozzi, University of California, Berkeley
Robert E. Burnkrant, Ohio State University
ABSTRACT - The validity of a single component model of attitude is assessed and compared with a multi-component attitude conceptualization. Convergent and nomological validity are found for a two component conceptualization of attitude, while the single component model is not supported. The implications of these findings for consumer research are discussed.
[ to cite ]:
Richard P. Bagozzi and Robert E. Burnkrant (1979) ,"Attitude Measurement and Behavior Change: a Reconsideration of Attitude Organization and Its Relationship to Behavior", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 295-302.

Advances in Consumer Research Volume 6, 1979      Pages 295-302

ATTITUDE MEASUREMENT AND BEHAVIOR CHANGE: A RECONSIDERATION OF ATTITUDE ORGANIZATION AND ITS RELATIONSHIP TO BEHAVIOR

Richard P. Bagozzi, University of California, Berkeley

Robert E. Burnkrant, Ohio State University

[Both authors contributed equally to the research.]

ABSTRACT -

The validity of a single component model of attitude is assessed and compared with a multi-component attitude conceptualization. Convergent and nomological validity are found for a two component conceptualization of attitude, while the single component model is not supported. The implications of these findings for consumer research are discussed.

INTRODUCTION

A considerable body of consumer behavior literature has dealt with the attitude-behavior relationship. Almost invariably attitude has been treated as a function of beliefs and their associated values (i.e., Hansen, 1969, Bass and Talarzyk, 1972; Kraft, Granbois and Summers, 1973; Lutz, 1975). These attitudes have been related to behavioral criteria with mixed but frequently disappointing results. It has been pointed out that some of the weaker results may have been due to the use of improper measurement procedures (i.e., Cohen, Fishbein and Ahtola, 1972; Cohen, 1972) or to a lack of correspondence between attitudinal predictors and behavioral criteria (i.e., Ajzen and Fishbein, 1977). The more recent use of attitude toward an act to predict specific intentions or behaviors reflects this concern for correspondence between attitude and behavior (i.e., Ryan and Bonfield, 1975; Wilson, Mathews and Harvey, 1975; Lutz, 1977).

The earlier research applying attitude to consumer behavior was concerned primarily with demonstrating the predictive validity of the attitude construct (cf., Wilkie and Pessemier, 1973). More recently it has been demonstrated that changes in beliefs lead to changes in attitude (Lutz, 1975) and that people combine belief and evaluation multiplicatively in their formation of an attitude (Bettman, Capon and Lutz, 1975). These contributions have undoubtedly advanced our understanding of attitudes and their application to consumer behavior. It is our contention, however, that the basic issue of attitude organization and its consistency with attitude measurement procedures has been largely ignored in this literature.

Two major distinguishable conceptualizations of attitude have flourished in the psychology literature. Those who hold a multi-dimensional view consider attitude to be a complex construct comprised of two or more components. In accordance with this view, Krech, Crutchfield and Ballachey (1962) define attitude as an enduring system of cognitions, feelings and response dispositions centered about a single object. Similarly, Rosenberg and Hovland (1960) regard attitude as a predisposition to some class of stimuli with cognitive, affective and behavioral responses. It is implied by those who hold a multi-component view of attitude that this attitude-behavior relationship would be stronger when the components are consistent than when they are inconsistent (Rosenberg and Hovland, 1960; Rosenberg, 1968). The multi-component view suggests that cognitive, affective and conative evaluations of objects are distinguishable aspects of attitude and that simultaneous consideration of all three components should be most predictive of overt behavior (cf., Greenwald, 1968). Failure to find a consistent direct relationship between attitude and behavior may be due to a failure to measure people's standing on all three components of attitude and to employ these as simultaneous and/or independent predictors of behavior.

The major alternative to the three component view treats attitude as a single affective construct. For example, Thurstone defines attitude as "the affect for or against a psychological object" (1931, p. 261). Fishbein (1967) argues that all attitude scaling techniques have in common the characteristic that they place individuals on a dimension of affect. This affect for or against an object is typically inferred from an assessment of people's beliefs about the object and the evaluative aspect of those beliefs. Therefore, alternative approaches to the measurement of attitude provide alternative measures of the same thing (i.e., affect) and should yield the same results. Obtained differences among alternative instruments in measurement of an attitude, according to this approach, would be due to measurement error and not the assessment of alternative components.

A third and intermediate position is maintained in the present paper. According to this view, attitude is a complex construct comprised of cognitive and affective components. These components simultaneously account for behavioral intentions. These intentions, in turn, lead to overt behaviors.

Katz and Stotland (1959) and Rosenberg (1968) point out that all true attitudes must have both cognitive and affective content, although they need not include a conative component. Similarly, Rosenberg (1968) stresses that, with the exception of cognitive dissonance, most of the consistency theories give only token recognition to the definition of attitude as an internally consistent structure of affective, cognitive and behavioral components; but, in practice, the behavioral component is usually treated as a dependent variable. The two component attitude position taken here recognizes and is consistent with the fact that self-reported behaviors and stated intentions to respond have frequently been treated as dependent effects of affective and/or cognitive variables (e.g., Tittle and Hill, 1967; Warner and DeFleur, 1969, Rogers and Thistlethwaite, 1970). Intentions appear to be at a lower level of abstraction (i.e., closer to observable behavior) than the cognitions and affective feelings on which they are based. We, therefore, propose that attitude be viewed as a two component construct comprised of a cognitive and an affective component. These two attitudinal dimensions are believed to simultaneously account for behavioral predispositions, although they may have a differential impact on them. Behavioral predispositions, in turn, lead to overt behaviors. The purpose of this paper will be to test and contrast this two component view to the single component model.

Attitude research in marketing, in contrast to this variation in conceptualizations, has almost invariably assumed that attitude is an unidimensional construct representing the affect for or against a psychological object, event, or situation. This assumption is implicit in the operationalization of Fishbein's model. It is consistent with the fact that most studies in marketing obtain only one measure of attitude which they relate to behavior (Hansen, 1969). When multiple methods are employed to measure attitude they are treated as alternative measures of the same thing (i.e., attitude) rather than measures of alternative components of attitude (e.g., Lutz, 1975; Lutz, 1977).

Several attempts have been made to provide empirical support for a multi-component treatment of attitude, but they have yielded mixed results. Woodmansee and Cook (1967) failed to find components which could be interpreted as representing affective, cognitive and conative components of attitude when they factor analyzed scales designed to measure attitudes toward blacks. Ostrom (1969) and Kothandapani (1971) employed judgment procedures to develop Guttman, Likert, Thurstone and Guilford self-rating scales for each of the three components. While both researchers claim that their results support a three component model of attitude, Bagozzi's (1978) reanalysis of their results obtained support for a multi-component attitude only with Ostram's data.

The present research involves a reanalysis of data initially presented by Fishbein and Ajzen (1974). In that study, each subject's attitude toward "being religious" was assessed on Thurstone, Guttman, Likert, semantic differential and Guilford self-rating scales. In addition, subjects provided responses to either 100 behavioral intention items or 100 self-reported behaviors. While Fishbein and Ajzen were consistent with a unidimensional approach to attitude in their treatment of all five attitude scales as alternative measures of attitude, it is our contention that the evaluative dimension of the semantic differential and the Guilford self-rating scale represent alternative measures of the affective component of attitude and that the Guttman, Likert and Thurstone scales represent alternative measures of the cognitive component of attitude.

Osgood, Suci and Tannenbaum (1957) define attitude as the projection of a concept on the evaluative dimension of semantic space. This unidimensional definition identifies attitude as the single dimension of semantic space which accounts for the concept's goodness or badness. Katz and Stotland (1959) identify the affective component with attributions of good or bad qualities. Similarly, McGuire (1969) argues that the evaluative dimension of semantic space is a measure of affect. Norman (1975) recently treated the evaluative dimension of semantic space as an operationalization of the affective component of attitude.

The Guilford self-rating scale asked people to provide their attitude toward being religious by checking an 11 point scale ranging from extremely favorable to extremely unfavorable. According to Ostrom (1969, p. 16) the affective component contains "statements representing favorable or unfavorable feelings." Kothandapani (1971, p. 323) made the same statement in his definition of the affective component. Similarly, Norman (1975) used a direct rating of a concept's favorability as an operationalization of the affective component of attitude toward that concept.

"The other three attitude measures were standard religiosity scales based on opinion items" (Fishbein and Ajzen, 1974, p. 62). Elsewhere, Fishbein and Ajzen (1975), p. 12) consider opinions to be cognitions or beliefs, and they distinguish them from affect and co-nation. The standard scales Fishbein and Ajzen (1974) employed included a Likert scale from Bardis (1961), a Guttman scale from Faulkner and De Jong (1969), and a Thurstone scale from Poppleton and Pilkington (1963). The statements employed in these scales are of the following form:

1. "Religious faith is merely another name for belief which is contrary to reason" (Poppleton and Pilkington, 1963, p. 34).

2. "Religious truth is higher than any other form of trust"   (Faulkner and De Jong, 1969, p. 567).

The cognitive component of attitude accounts for the perceived relationships between the object of attitude and other objects or concepts. Ostrom points out that the cognitive component of attitude includes, "beliefs about the object, characteristics of the object, and relationships of the object with other objects" (1969, p. 16). Kothandapani (1971) notes that this component includes beliefs and opinions about the object. Examination of the Guttman, Likert, and Thurstone scales employed by Fishbein and Ajzen shows them to be consistent with this conceptualization of the cognitive component of attitude.

It is possible to distinguish empirically between the single component and the two component view of attitude using Fishbein and Ajzen's data. A single component model of attitude would be supported if convergent validity is obtained when all five attitude scales are treated as alternative measures of attitude. A two component attitude model would be supported if convergence is obtained only when semantic differential and Guilford self-rating scales are treated as alternative measures of the affective component; and Guttman, Likert, and Thurstone scales are treated as alternative measures of the cognitive component of attitude. Furthermore, the two component position requires that both the cognitive and affective components simultaneously account for behavioral intentions and self-reported behaviors. The data do not permit us to distinguish between the prediction that behavior is accounted for by intentions alone and the prediction that behavior is accounted for by the simultaneous consideration of the conative, cognitive, and affective components of attitude. This is due to the fact that behavior and behavioral intention were not both obtained from the same individuals. However, a finding that the cognitive and affective components of attitude simultaneously account for behavioral intention and verbal reports of past behavior would be consistent with the two component view suggested here.

In summary, we expect that a reanalysis of the Fishbein and Ajzen data will obtain convergent validity when the cognitive and affective measures of attitude are treated as separate components, but convergence will not be obtained when all five instruments are treated as alternative measures of the same underlying construct. Furthermore, we expect that each component of attitude will separately account for scaled multiple act criteria, and both components of attitude will simultaneously account for scaled multiple act criteria.

METHODOLOGY

Subjects and Measures

Two samples of respondents were obtained by Fishbein and Ajzen (1974). In the first, 62 male and female undergraduates indicated which items from a set of 100 behaviors they had performed (the self-reported behaviors sample). The set of behaviors consisted of a list of 70 actions dealing with religious matters (e.g., pray before or after meals, donate money to a religious institution) and 30 additional actions in a refusal format selected from the original 70 (e.g., refuse to state religious preference during university registration). The second sample was composed of 63 male and female undergraduates who indicated which items from the set of 100 behaviors they would perform (the behavioral intentions sample).

In addition, all subjects completed five scales measuring attitudes toward religion. A Guilford self-rating scale measured attitudes towards being religious on an 11-point scale ranging from extremely favorable to extremely unfavorable. A semantic differential scale measured evaluations of "being religious" on five 11-point bipolar scales having the following end points: good-bad, harmful-beneficial, wise-foolish, pleasant-unpleasant, and sick-healthy. The Guilford and semantic differential scales are considered in the present study to tap largely the evaluative or affective dimension of attitudes. Support for this contention was provided in the preceding sections of this article.

To measure the cognitive component of attitude toward religion, three standard religiosity scales were employed. These include Likert (Bardis, 1961), Guttman (Faulkner & De Jong, 1969), and Thurstone (Poppleton & Pilkington, 1963) scales.

Method of Analysis

Convergent Validity. Before testing the hypotheses relating attitudes to behavior, it is necessary to establish the validity of the attitude measures. [Although a full test of construct validity would require the use of the multitrait-multimethod research design, this was not possible in the present study because the same multiple measures were not used on each of the two hypothesized attitudinal components. However, because two measures of affect and three measures of beliefs were obtained, it is possible to ascertain the degree of convergent validity for the attitude measures. Further, as additional evidence for construct validity, the data allow also for the determination of nomological validity as developed below.] Figure 1 presents two path diagrams that can be used in this regard. The single factor model of Figure la hypothesizes that the five attitude measures (i.e., the Guilford self-report, SR; semantic differential, SD; Guttman, G; Likert L; and Thurstone T; scales) each indicate a single underlying construct, "attitudes" (i.e., A1+2). Fishbein and Ajzen (1974) did not differentiate the five attitude measures into affective and cognitive dimensions. Rather, the authors treated all five as independent measures of the same underlying "attitude towards religion" construct. Thus, they were implicitly assuming that the scales converged to measure a single attitude construct. This assumption is represented in Figure 1a. The validity of this assumption as well as the bicomponent model will be tested explicitly in the present study before investigating the attitude-behavior hypotheses (see below).

The two factor model of Figure lb posits that attitudes are represented as two conceptually independent, yet empirically related constructs: (1) an affect or evaluative component (A1) and (2) a cognitive or belief dimension (A2).

To test the models of Figure 1, a confirmatory factor analysis methodology can be used (cf., Kenny, 1976). Briefly, the general confirmatory factor analysis model to test is

y = Lx + z  (1)

where y is a vector of p measurements, x is a k < p vector of factors, z is a vector of p unique scores, and L is a p x k matrix of factor loadings. For Figure 1a, p = 5 and k = 1; while for Figure 1b, p = 5 and k = 2. With the assumptions that E(x) = E(z) = 0, E(xx') = f, and E(zz') = y, where y is a diagonal matrix, the variance-covariance matrix of y may be expressed as

E = LfL' + y   (2)

Joreskog (1969) derives a maximum likelihood procedure for estimating the parameters in L, f, and y. Further, the methodology yields an overall X2 goodness-of-fit test. The computer program, LISREL, may be used to test the models of Figure 1 and equations (1) and (2) (Joreskog & van Thillo, 1972).

FIGURE 1

PATH DIAGRAMS FOR DETERMINING CONVERGENT VALIDITY OF ATTITUDE MEASURES

Nomological Validity. As further evidence of construct validity, the degree of nomological validity of the attitude measures was determined. Following Campbell (1960), nomological validity is defined herein as the degree to which predictions from a concept in a theoretical system of concepts are confirmed. Figure 2 presents a path model that can he used to test for the nomological validity of attitudes where, for purposes of illustration, attitudes (A) are shown measured by three scales (y1, y2, and y3) and behavior (B) is indicated by two scales (y4 and y5). The hypothesis is that the nomological validity of the attitude construct (A) will be confirmed if predictions from the model as to the resulting behavior (B) are validated. The proper attitude construct (A) to use is the one achieving convergent validity as determined in the previous analysis. The appropriate behavioral measure (B) is the one constructed from scaled behaviors because these behaviors approach the same level of specificity as the attitudinal items. For a description of the scaling procedures used, the reader is referred to Fishbein and Ajzen (1974).

The hypothesis of nomological validity can be tested using J÷reskog's analysis of covariance structures model (J÷reskog, 1970). That is, the model to test is

B = gA + z  (3)

where it is assumed that all variables are taken to have zero expectations and that the disturbances (i.e., z, d1-d3, el, and e2) are mutually independent and are independent-of their corresponding explanatory variables. The computer program, LISREL, can be used to

EQUATION  (4)

estimate parameters and test the model of Figure 2 and equations (3) and (4). Although the model is illustrated for the case of three attitudinal and two behavioral measures, it can be generalized to the case of attitudinal and m behavioral measures.

FIGURE 2

PATH DIAGRAM FOR DETERMINING THE RELATIONSHIP BETWEEN ATTITUDES AND SCALED MEASURES OF BEHAVIOR

RESULTS

Looking only at the intercorrelations of variables in both samples, Fishbein and Ajzen (1974) concluded that the five verbal attitude scales showed "a high degree of convergent validity." This and the other hypotheses discussed above will be investigated using the structural equation methodology.

Convergent Validity

The findings for the tests of convergent validity using the more rigorous structural equation methodology, are shown in Table 1. Notice first that the hypothesis of convergence for the single factor model must be rejected for both the self-reported behaviors sample (i. e., X2 = 23.91, d.f. = 5, p = .00) and the behavioral intentions sample (X2 = 13.98, d.f. = 5, p = .02). Thus, contrary to the original claim made by Fishbein and Ajzen (1974) and contrary to the assumptions made in most consumer research, one can not accept the hypothesis that the five scales each measure a single underlying attitude construct.

Consequently, the two factor model of attitudes was tested. As shown in the final two columns of Table 1 the hypothesis of convergence receives strong support for the self-reported behaviors sample (i.e., X2 = 2.30, d.f. = 4, p = .68) and adequate support for the behavioral intentions sample (i.e., X2 = 8.30, d.f. = 4, p = .08). [Although the X2 value for the behavioral intentions sample indicates a borderline fit, (i.e., p = .08), when Bartlett's (1951) small sample correction factor is applied, the model reaches acceptable levels of significance (i.e., X2 = 7.65, d.f. = 4, and p = .10).] Thus, convergent validity is established for the two factor affective/cognitive model of attitudes but not for the single factor model. Any further analysis must take this finding into consideration.

TABLE 1

GOODNESS-OF-FIT TESTS AND PARAMETER ESTIMATES FOR CONSTRUCT VALIDITY MODELS

Nomological Validity

Given that two dimensions of attitude have been identified, nomological validity was determined for each. Looking first at the relation between the affective component of attitude and the scaled behavior measures, it can be seen in Table 2 that nomological validity is established for both the self-reported behaviors and behavioral intentions samples. This occurs in either case whether behavior is measured with the Guttman and Likert techniques or the Likert and Thurstone techniques. [Although the X2 value for the behavioral intentions sample indicates a borderline fit, when Bartlett's (1951) small sample correction factor is applied, the model reaches acceptable levels of significance.] Also, as illustrated by the standardized estimates for g, the evaluative attitudes toward the church relate at a high level of magnitude to the scaled behavior measures, as predicted by theory.

Looking next at the relation between the cognitive component of attitude and the scaled behavior measures, it can be seen in Table 3 that nomological validity is again established for both samples and for both pairs of scaled behavior measures. Further, the cognitive components relate at a high level of magnitude to the scaled behavior measures, as predicted by theory.

In sum, based on the results for convergent and nomological validity, some evidence exists for establishing the construct validity of the affective/cognitive model of attitudes.

TABLE 2

GOODNESS-OF-FIT TESTS AND PARAMETER ESTIMATES FOR NOMOLOGICAL VALIDITY MODELS RELATING AFFECT TO BEHAVIOR

TABLE 3

GOODNESS-OF-FIT TESTS AND PARAMETER ESTIMATES FOR NOMOLOGICAL VALIDITY MODELS RELATING COGNITIONS TO BEHAVIOR

To investigate the differential effects of affect and cognitions on behavior, the path model of Figure 3 may be used. The model to test is based on Joreskog's analysis of covariance structure:

EQUATIONS  (5),   (6),  and  (7)

where it is assumed that all variables are taken to have zero expectations and that the disturbances are mutually independent and are independent of their corresponding explanatory variables. Again, LISREL may be used to estimate the parameters in the model and test hypotheses.

Applying LISREL to the data yields the results shown in Table 4 where only the findings for the self-reported behaviors sample are displayed. In general, the model fits the data well (X2 = 13.11, d.f., = 10, p = .22). Further, the affective or evaluative component can be seen to be approximately three times as forceful in its impact on behavior as the cognitive component [The nomological validity models are run with the Guttman/Likert and Likert/Thurstone operationalizations rather than with a single Guttman/Likert/Thurstone operationalization of behavior because inspection of the correlation matrix in Table 1 reveals that the Guttman and Thurstone measures correlate at a lower level than the other two pairs in both samples.] (g1 = .651 vs. g2 = .226). It should be noted, however, that the values for g1 and g2 will be sensitive to the pattern of correlations among the behavior and attitude measures, and problems of multicollinearity should be considered.

DISCUSSION

The results obtained from a reanalysis of the Fishbein and Ajzen data support a two component model of attitude. When Guilford self-rating, semantic differential, Guttman, Likert and Thurstone scales were treated as alternative measures of a single underlying construct (i.e., attitude), the analysis failed to provide convergent validity. When Guilford self-rating and semantic differential scales were treated as alternative measures of an affective component of attitude and Guttman, Likert and Thurstone scales were treated as alternative measures of a cognitive component of attitude, convergent validity was supported. The validity of the cognitive-affective model was supported for both the self-report and the behavioral intentions samples.

Further support for the validity of the two component model is provided by a consideration of nomological validity. It was found that each component of the two component model separately predicted scaled multiple act criteria. Furthermore, both components contributed to the simultaneous prediction of behavior. While both components simultaneously accounted for behavior, the affective component was approximately three times as powerful as the cognitive component. In summary, then, the results provide support for the nomological validity of the two component attitude model. Behavior was predicted from a simultaneous consideration of the cognitive and affective components of attitude.

These results and those obtained from Bagozzi's (1978) reanalysis of Ostrom's (1969) data provide strong support for a multi-component treatment of attitude. Attitude is regarded as a complex construct comprised of affective and cognitive components. These components were found to account for behavior and behavioral intentions.

FIGURE 3

THE IMPACT OF AFFECT AND COGNITIONS ON BEHAVIOR

TABLE 4

GOODNESS OF FIT TESTS AND PARAMETER ESTIMATES FOR A MODEL TESTING THE RELATIVE IMPACT OF AFFECT AND COGNITIONS ON BEHAVIOR

It was pointed out earlier that most attitude research in consumer behavior has been based on measurement of only one component of attitude. Both in research dealing with attitude change and in research considering the attitude-behavior relationship, conclusions are generally made after measuring either the cognitive or the affective component of attitude. It has been maintained that failure to obtain consistent results in research dealing with the attitude-behavior relationship may be due, in part, to the failure to measure more than one component of attitude in any given study. The results obtained from the present research are consistent with the argument that a complete accounting of attitude requires measurement of both the cognitive and affective components.

Greenwald (1968) suggests that the cognitive and affective components of attitude have distinct antecedents. The affective component may be formed most directly through classical conditioning whereas the cognitive component may be more directly affected by communication and cognitive learning. This indicates that attitude change strategies may have differential effects on the components of attitude. Even though a message is quite effective in changing the cognitive component of attitude it may have a relatively weak effect on behavior if the affective component is a more powerful predictor of the given behavior. This issue has received little attention although it may have important implications for the effects of attitude change strategies on behavior change.

In this study, even though both components of attitude were found to simultaneously account for behavior, the affective component was found to be roughly three times as forceful in its impact on behavior as the cognitive component. It may be that the relative impact of the various components of attitude is dependent on the attitude-behavior target under consideration. We may speculate that the affective component may play a stronger role in the purchase of low involvement products such as soft drinks, for instance, whereas the cognitive component may play a larger role in the purchase of products requiring a more complex decision process prior to purchase. The findings of the present study may tend to argue against this since we may expect religion to be highly involving, although the performance of religious behaviors may not involve a very conscious or extended decision process. The point is, however, that both components should be investigated as simultaneous predictors of purchase behavior. The relative impact of the cognitive and affective components may vary in terms of the nature of the product under consideration.

In the Fishbein and Ajzen study the cognitive and affective measures of attitude were highly correlated. This may be due to pressure toward cognitive consistency. Even though the cognitive and affective components of attitude may be formed by distinct processes, the pressure to achieve affective-cognitive consistency probably leads to high correlations among the components of attitude. As McGuire (1960) has pointed out, consistency is likely to increase over time. The relatively weak relationship between cognitive and affective components of attitude shown in the Lutz (1977) study may be due to the fact that, in contrast to Fishbein and Ajzen, Lutz manipulated beliefs and then measured the effects of this manipulation on cognitive and affective measures of attitude. It may be that a stronger relationship would have been obtained if a delayed post-test had been employed. The delay would have provided more time for consistency restoration to take effect.

In summary, the results of this study show that attitude is comprised of two distinct components. These components simultaneously account for behavioral criteria. Proper assessment of the attitude behavior relationship, therefore, requires measurement and use of both of these attitude components as simultaneous predictors of behavior.

REFERENCES

I. Ajzen and M. Fishbein, "Attitude-Behavior Relations: A Theoretical Analysis and Review of Empirical Research," Psychological Bulletin, 84 (1977), 888-918.

R. P. Bagozzi, "The Construct Validity of the Affective, Behavioral, and Cognitive Components of Attitude by Analysis of Covariance Structures," Multivariate Behavioral Research, 13 (1978), 9-31.

P. D. Bardis, "A Religion Scale," Social Science, 36 (1961), 120-123.

M. S. Bartlett, "The Effect of Standardization on an Approximation in Factor Analysis," Biometrika, 38 (1951), 337-344.

F. M. Bass and W. W. Talarzyk, "An Attitude Model for The Study of Brand Preference," Journal of Marketing Research, 9 (1972), 93-96.

J. R. Bettman, N. Capon, and R. J. Lutz, "Multiattribute Measurement Models and Multiattribute Attitude Theory: A Test of Construct Validity," Journal of Consumer Research, 1 (1975), 11-15.

D. T. Campbell, "Recommendations for APA Test Standards Regarding Construct, Trait, or Discriminant Validity," American Psychologist, 15 (1960), 546-553.

J. B. Cohen, "Toward an Integrated Use of Expectancy-Value Attitude Models." Paper presented at the ACRIAMA workshop on buyer/consumer information processing, Chicago (1972).

J. B. Cohen, M. Fishbein, and O. T. Ahtola, "The Nature and Uses of Expectancy-Value Models in Consumer Attitude Research," Journal of Marketing Research, 9 (1972), 456-460.

J. E. Faulkner and G. F. De Jong, "Religiosity in 5-D: An Empirical Analysis." In J. P. Robinson and P. R. Shaver (eds.), Measures of Social Psychological Attitudes. Ann Arbor: University of Michigan, Institute of Social Research (1969).

M. Fishbein, "A Consideration of Beliefs and Their Role in Attitude Measurement." In M. Fishbein (ed.), Readings in Attitude Theory and Measurement. New York: Wiley (1967).

M. Fishbein and I. Ajzen, "Attitude Towards Objects as Predictors of Single and Multiple Behavioral Criteria," Psychological Review, 81, (1974), 59-74.

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

A. G. Greenwald, "On Defining Attitude and Attitude Theory." In A. G. Greenwald, T. C. Brock and T. M. Ostrom (eds.), Psychological Foundations of Attitudes. New York: Academic Press (1968).

F. Hansen, "Consumer Choice Behavior: An Experimental Approach," Journal of Marketing Research, 6 (1969) 436-443.

K. G. J÷reskog, "A General Approach to Confirmatory Maximum Likelihood Factor Analysis," Psychometrika, 34 (1969), 183-202

K. G. J÷reskog, "A General Method for Analysis of Co-variance Structures," Biometrika, 57 (1970), 239-251.

K. G. J÷reskog and M. van Thillo, LISREL: A general computer program for estimating a linear structural equation system involving multiple indicators of unmeasured variables. Educational Testing Service Research Bulletin, (1972), 72-156.

D. Katz and E. Stotland, "A Preliminary Statement to a Theory of Attitude Structure and Change." In S. Koch (ed.), Psychology: A Study of a Science, Vol. 3, New York: McGraw-Hill (1959), 423-475.

D. A. Kenny, "An Empirical Application of Confirmatory Factor Analysis to the Multitrait-Multimethod Matrix," Journal of Experimental Social Psychology, 12 (1976), 247-252.

V. Kothandapani, "Validation of Feeling, Belief, and Intention to Act as Three Components of Attitude and Their Contribution to Prediction of Contraceptive Behavior, Journal of Personality and Social Psychology, 19 (1971), 321-333.

F. B. Kraft, D. H. Granbois and J. O. Summers, "Brand Evaluation and Brand Choice: A Longitudinal Study, Journal of Marketing Research, 10 (1973), 235-241.

D. Krech, R. S. Crutchfield and E. L. Ballachey, Individuals in Society. New York: McGraw-Hill (1962).

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

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

W. J. McGuire, "Cognitive Consistency and Attitude Change," Journal of Abnormal and Social Psychology, 60 (1960), 345-353.

W. J. McGuire, "The Nature of Attitudes and Attitude Change." In G. Lindzey and E. Aronson (eds.), The Handbook of Social Psychology, Vol. 3, Cambridge, Mass: Addison-Wesley (1969).

R. Norman, "Affective-Cognitive Consistency, Attitudes, Conformity, and Behavior, Journal of Personality and Social Psychology, 32 (1975), 83-91.

C. E. Osgood, G. J. Suci and P. H. Tannenbaum, The Measurement of Meaning. Urbana: University of Illinois Press (1957).

T. M. Ostrom, "The Relationship Between the Affective, Behavioral, and Cognitive Components of Attitude," Journal of Experimental Social Psychology, 5 (1969), 12-30.

P. K. Poppleton and G. W. Pilkington, "The Measurement of Religious Attitudes in a University Population," British Journal of Social and Clinical Psychology, 2 (1963), 20-36.

R. W. Rogers and D. L. Thistlethwaite, "Effects of Fear Arousal and Reassurance on Attitude Change," Journal of Personality and Social Psychology, 15 (1970), 227-233.

M. J. Rosenberg and C. I. Hovland, "Cognitive, Affective and Behavioral Components of Attitudes." In M. J. Rosenberg, C. I. Hovland (eds.), Attitude Organization and Change: An Analysis of Consistency Among Attitude Components. New Haven: Yale University Press (1960).

M. J. Rosenberg, "Hedonism, Inauthenticity, and Other Goads Toward Expansion of a Consistency Theory." In R. P. Abelson, et al. (eds.), Theories of Cognitive Consistency: A Source-Book. Chicago: Rand McNally (1968).

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

L. L. Thurstone, "The Measurement of Social Attitudes," Journal of Abnormal and Social Psychology, 26, (1931), 249-269.

C. R. Tittle and R. J. Hill, "Attitude Measurement and Prediction of Behavior: An Evaluation of Conditions and Measurement Techniques," Sociometry, 30 (1967), 199-213.

L. G. Warner and M. L. DeFleur, "Attitude as an Inter-actional Concept: Social Constraint and Social Distance as Intervening Variables Between Attitudes and Action," American Psychological Review, 34, (1969), 153-169.

W. L. Wilkie and E. A. Pessemier, "Issues in Marketing's Use of Multiattribute Attitude Models," Journal of Marketing Research, 10, (1973), 428-441.

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

J. J. Woodmansee and S. W. Cook, "Dimensions of Verbal Racial Attitudes: Their Identification and Measurement.'' Journal of Personality and Social Psychology, 7 (1967), 240-250.

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