The Theory of Reasoned Action: Examination of the Sufficiency Assumption For a Television Viewing Behavior

Barbara Loken, University of Minnesota
ABSTRACT - The sufficiency assumption of Fishbein and Ajzen's theory of reasoned action was examined for a television viewing behavior with a sample of 53 students. Findings indicated that variables external to the model such as attitudes toward objects. did not contribute to behavioral intentions significantly over and above behavioral attitudes and subjective norms. The relationship between cognitive variables proposed to underlie behavioral attitudes and external variables was also examined. Finally intentions were found to be moderately good predictors of behavior, and better predictors for nonintenders than for intenders.
[ to cite ]:
Barbara Loken (1983) ,"The Theory of Reasoned Action: Examination of the Sufficiency Assumption For a Television Viewing Behavior", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 100-105.

Advances in Consumer Research Volume 10, 1983      Pages 100-105

THE THEORY OF REASONED ACTION: EXAMINATION OF THE SUFFICIENCY ASSUMPTION FOR A TELEVISION VIEWING BEHAVIOR

Barbara Loken, University of Minnesota

[The author is grateful to Ron Hinkle for his helpful comments.]

ABSTRACT -

The sufficiency assumption of Fishbein and Ajzen's theory of reasoned action was examined for a television viewing behavior with a sample of 53 students. Findings indicated that variables external to the model such as attitudes toward objects. did not contribute to behavioral intentions significantly over and above behavioral attitudes and subjective norms. The relationship between cognitive variables proposed to underlie behavioral attitudes and external variables was also examined. Finally intentions were found to be moderately good predictors of behavior, and better predictors for nonintenders than for intenders.

INTRODUCTION

Ajzen and Fishbein's (1980) theory of reasoned action has increasingly captured the interest of marketing researchers. This interest appears to lie both in its theoretical extension of previous attitude research and in its practical us as a tool for investigating factors that may influence any particular behavior. The present study attempts to examine the use of the model for a behavior of interest in marketing as well as to investigate the "sufficiency assumption" of the model using hierarchical models.

Assumptions of the Model

Overview of Model Components. According to the theory, most behaviors of interest to behavioral scientists are volitional control, and may be predicted by obtaining people's intentions to perform the behavior. A person's intentions to perform a particular behavior (BI) in turn are determined by two factors: (1) a personal or "attitudinal" factor (AB) and (2) a social or "normative" factor (SN). The model further predicts that a person's attitude toward performing a particular behavior is a function of his/her beliefs about the consequences of performing that behavior and his/her evaluations of those consequences.

EQUATION    (1)

where AB = the attitude toward performing the behavior. bi = the behavioral belief that performing the behavior B leads to outcome i, ei = the person's evaluation of outcome i, and" = the number of salient beliefs the person holds about performing behavior B. A person's subjective norm (SN) is also proposed to be a function of his/her beliefs, in this case one's normative beliefs about what important others think he or she should do and his/her motivation to comply with these others.

Sufficiency Assumption. The "sufficiency assumption" states that variables other than attitudes toward performing behavior and subjective norms, such as demographics, personality characteristics. and other attitudes. are "external" to the model. and affect behavioral intentions and behavior only to the extent that they are mediated by the normative and attitudinal components of the model. Furthermore external variables should influence one attitudes toward performing the behavior only to the extent that they affect one's behavioral beliefs and/or outcome evaluations. In a similar fashion, variables that influence the subjective norm should also affect its determinants. The general model that illustrates this chain of processes is diagrammed in Figure 1.

FIGURE 1

THE EFFECTS OF EXTERNAL VARIABLES ON INTENTIONS AND BEHAVIOR

Past Research

Using correlational data the expectancy-value model of attitude and the extended model of behavioral intentions have received support in consumer (Ryan & Bonfield 1975, Wilkie & Pessimier 1973) and social-psychological (Fishbein 1979. Fishbein & Ajzen 1975) research. This sufficiency assumption has also been supported (Ajzen & Fishbein 1970, 1974, Davidson & Jaccard 1975, Loken & Fishbein 1980).

Recently investigators have attempted to focus on validation of the causal relations proposed by the mode]l(e.g. Lutz 1977, Bentler & Speckhart 1979, 1981, Bagozzi 1980). While some of this research tends to support the model (e.g.. Lutz 1977) other finds limited or partial support (e.g. Bentler & Speckhart 1974, 1981). For example, using a structural equation analysis Bentler and Speckhart (1979) found that drug usage was directly rather than indirectly affected by attitudes toward using the drug. That is behavioral intentions did not mediate this attitude behavior relationship. For a different set of behaviors (Bentler & Speckhart 1981) behavioral attitudes again had a direct affect on behavior i.e.. they were not solely mediated by intentions. In contrast Bagozzi (1981) found that behavioral attitudes influenced behavior indirectly through their impact on intentions. Ryan and Bonfield (1980) also find support for model components using regression analyses. These latter authors concluded that the cognitive structures proposed to underlie AB (i.e Ebjei) and SN (i.e. Ebjmj) influenced behavior only through their effects on behavioral intentions.

Relevant to the sufficiency assumption of the model Bentler and Speckhart (1979, 1981) examined the effects of past behavior on (present) behavior. They found a direct effect between these two variables that was not mediated by behavioral intentions. On the other hand. Bagozzi (1981) has argued that retrospective self-reports of behavior may have biased the data in favor of this result and, further, that the behaviors studied may have contained a habitual component.

These findings suggest only limited support for the sufficiency assumption as they pertain to past behaviors performed. However, the effects of other types of external variables. such as attitudes toward objects. have not been examined within this same type of framework. These effects may be particularly interesting for consumer research. which relies heavily on the use of brand attitudes as a dependent variable.

Multi-Attribute Consumer Models

Most multi-attribute models have assumed that purchases are a function of the consumer's perceptions of the product's favorable and unfavorable attributes. Thus. for example. if a television program is perceived as having more favorable than unfavorable attributes, this should lead to a higher probability of viewing behavior. On the other hand, the theory of reasoned action would argue that one's attitude toward a television program (like an attitude toward a product) is an "external" variable, with no necessary or direct relationship to viewing behavior. Thus. according to the theory one's attitude toward a television program may be quite different from one's attitude toward watching the program.

Of course in many instances the two attitudes. toward a program and toward watching the program. could be related. For example, program attributes that lead one to form a favorable or unfavorable attitude toward the program may also be the primary determinants of one's attitudes toward watching the program. In other instances the two attitudes (toward a program and toward watching the program) may be unrelated. For example, during certain hours of the day, or for certain individuals, program attributes may have little or no effect on viewing attitudes and behavior. At those times, or for those people, the traditional multi-attribute models may show lower predictive ability.

These same possibilities exist with regard to attitudes toward persons. For example. in some cases more than others. a viewer will show a stronger tendency to watch a television program if it features a well-known (and liked) actor or actress than if it does not. As pertains to purchase decisions, a popular actor may be hired to promote a particular product, presumably to increase the product's sales. These examples demonstrate the effect that one's attitude toward a person is expected to have on purchase behavior or viewership. However, in many cases, attitudes toward a program or person will not affect behavior with respect to this program or person. According to the Fishbein and Ajzen model, this is not unexpected since attitudes toward persons are not direct determinants of behavior, i.e., attitudes toward persons are "external" variables.

Study Objectives

The present research examines the effects of attitudes toward objects and other external variables on the model components, using structural equation analyses. Furthermore, the relationships between the individual model components are validated, using the traditional correlational and regression analyses frequently employed in tests of the model. Included is a test of the intention-behavior relationship for the behavior under investigation. A relatively short time period (i.e., one week) is used to maximize this relationship. Finally, an additional like in the model, the cognitive structure underlying behavioral attitudes, is measured to suggest possible loci of effects of external variables on behavior. The behavior investigated was watching a particular television program ("The Rockford Files") in the next week.

METHOD

Subjects

Subjects were 56 undergraduates who participated in the experiment as part of a class project in an introductory marketing research course. The class was asked to choose between four different topics for a questionnaire. The most frequently selected topic. "my watching reruns of 'The Rockford Files' on TV in the next week," was then used for the entire group.

Elicitation Pretest

A set of "modal salient beliefs" was acquired through an elicitation procedure (see Ajzen & Fishbein 1980. for a description of this procedure). Respondents were asked to list chose advantages and disadvantages of performing the behavior in question that easily came to mind. as well as anything else they associated with the behavior. Each item was listed on a separate line.

Responses were content analyzed. and the the most frequentLy mentioned responses (53% or the total) were used as modal salient beliefs (see Table 1). Respondents to the elicitation questionnaire were the same persons who completed the main questionnaire. The elicitation items were completed about ten days prior to the main questionnaire.

TABLE 1

MODEL SALIENT BELIEFS

Model Components. The main questionnaire included measures of (a) behavioraL intentions, (b) attitudes toward performing the behavior. (c) subjective norms. (d) 8 behavioral beliefs (i.e. 8 modal salient beliefs) about the outcomes of performing the behavior and (e) 8 outcome evaluations. All variables were measured on seven-place scaLes from +3 to -3. Behavioral intentions subjective norms and behavioral beliefs were measured on likely-unlikeLy scales evaluations were measured on good-bad scales. and attitude toward performing the behavior was measured on four evaluative scales (good-bad enjoyable-unenjoyable, pleasant-unpleasant and nice-awful). Furthermore measures were personalized to the respondent (e.g.. My watching... rather than general in format (Watching... ).

To compute the expectancy-value measure of attitude. i.e.. Ebiei. The score for each belief statement was multiplied by the score for the corresponding evaluation and summed for all 8 beliefs. The four attitude scales were summed to give an independent measure of AB ranging from +12 (very favorable) to -12 (very unfavorable). Factor analyses verified that these four scales were highly intercorrelated (the average intercorrelation was .82) and Loaded highly on the same evaluative dimension.

External variables. Attitudes toward objects variables external to the model. incLuded attitudes toward (1) "The Rockford Files," (2) watching Late night TV. and (3) the actor James Garner who plays Rockford. These three attitudes were each measured on the same four evaluative semantic differentials noted for the behavioral attitude and scored by summing responses to these four scales.

Three additional measures were included at the end of the questionnaire. These questions asked (1) whether the respondent had a television in his or her household (to ensure that all respondents had access to a television). (2) the number of hours a day (approximately) the respondent watched television and (3) the number of people who presently lived in the household where the respondent resided.

Behavioral Measure. One week following completion of the main questionnaire. respondents were asked whether or not they had performed the behavior during the last week. and if so, the number of times they had viewed the show during that week (the program aired 4 times). Thus. two behavioral measures were obtained. a dichotomous yes-no response and a continuous measure.

Only 53 respondents completed the behavioral measures and could be successfully matched with previous questionnaire data. Only these 53 respondents scores were used in calculations reported.

RESULTS

Intention-Behavior Relations

The first link to be tested in the chain of processes described in Figure 1 is the intention-behavior relationship. Two zero-order correlations were obtained between behavioral intention measures and self-report measures of behavior. The first. between the continuous measure of intention to watch reruns of The Rockford Files on TV in the next week and the dichotomous behavioral response, was .53, p < .01. The second between the continuous measure of intention and the number of times the respondent viewed the program (ranged from 0 to 3 from a possible range from 0 to 4), was .61. p <.01. These relationships were at least moderately high. The continuous behavioral measure was a more accurate measure than the dichotomous measure. although not significantly better. Although the behavioral intention was measured with regard to whether or not a behavior was intended. the measure may have implicitly incorporated the notion of quantity (i.e.. the number of times the behavior would be performed). Thus. the more frequently the behavior was performed perhaps the greater the likelihood was of intending to perform it. However the higher correlation associated with the continuous variable could also have been a function of the reliability of the measures.

Frequency data showed three findings of interest. First intentions predicted behavior for 78% (39 out of 50) of the respondents. (Three people stated they were neither likely nor unlikely to perform the behavior.) Of the 10 people who watched the show, 80% (8) had intended to do so (responses +1, +2, +3). Of the 40 people who did not watch the show, 78% (31) had intended not to watch it (responses -1, -2, -3). However. a second finding was that of those 11 subjects whose intentions did not correspond to their behavior, 9 (82%) had intended to perform the behavior. That is a breakdown in intention-behavior relations was more likely to occur for intenders than for nonintenders. Davidson and Beach (1981) who found the same tendency for a family-planning behavior. suggest an "inertia" hypothesis, that performing a behavior is more likely than not performing it to break a norm or status quo. Finally, frequency data pertaining to intentions and behavior suggest a tendency for subjects to not perform the behavior investigated, resulting in a somewhat skewed distribution of this measure.

Model Components

Continuing to the second link in the network shown in Figure 1, according to the theory one or both of two components, AB and SN, predict intentions. A multiple correlation was computed, with AB and SN (both measured in the one week time frame) as predictors and BI (measured in the one week time frame) as the criterion. Consistent with the theory. the multiple correlation was significant. R = .68, p < .001. such that the two components together accounted for 46 of the variance in intentions. Standardized regression coefficients suggest that both attitudes (w = .45, p < .001) and norms (w = .36, p <.01) were important for prediction of intentions. The zero-order correlations between AB and BI. SN and BI. and AB and SN. were .60, .55, and .42, respectively.

Sufficiency Assumption

According to the sufficiency assumption. variables external to the model should contribute to intentions (and behavior) only through their effects on attitudes and/or norms (both are considered behavioral determinants based on above findings). Thus. other attitudes such as attitudes toward The Rockford Files attitudes toward watching late night TV. or attitudes toward the star of the show should not contribute to prediction of intentions over and above AB and SN.

To test these assumptions. two models shown in Figure 2 were compared by applying a maximum likelihood estimation procedure for examining linear structural equation systems (LISREL IV) to correlational data (Joreskog 1973. Joreskog & Sorbom 1978). This procedure simultaneously evaluated both a measurement and a structural model when multiple indicators are used for the constructs in the model. in the present case. since single indicators were used the measurement model was not tested. and the system of equations reduced to a series of regressions. Nevertheless. LISREL provides an estimate of the overall goodness-of-fit of the models tested and enables one to determine whether an external variable contributes to behavior over and above its effects on attitudes toward performing the behavior and subjective norms.

FIGURE 2

MODELS DESCRIBING RELATIONS BETWEEN EXTERNAL VARIABLES AND BEHAVIOR

The coefficients and chi-square goodness-of-fit statistics for each of the six external variables is shown in Table 2. As shown for each of the external variables. the coefficients linking AB and SN to BI, and the coefficient linking BI to behavior, were significant. Furthermore the coefficients linking 4 external variables to AB and the coefficients linking 2 external variables to SN were significant. However, the coefficients linking each external variable directly to behavior were nonsignificant.

TABLE 2

STANDARDIZED PARAMETER ESTIMATES AND CHI-SQUARE GOODNESS-OF-FIT TESTS FOR MODELS 1 AND 2

Moreover the difference in chi-square values for Models 1 and 2 (Figure 7) which is itself distributed as a chi-square was nonsignificant for each of the external variables. Since the sample size was quite small (N = 53) indices of incremental fit comparing to a null model specifying zero covariances among all terms of the model (see Bentler & Bonett 1980) were computed. Chi-squares associated with the null model are shown in Table 2. When the chi-squares of the null model and the proposed model 7 are contrasted the proposed model is a significant improvement in did for each of the external variables. However indices of incremental fit showed estimates above .90 for only two of the five external variables attitude toward the program and attitude toward the star of the program. It should also be noted that the fit was greatest for those external variables most strongly related to behavioral attitudes and subjective norms.

These findings are generally consistent with the assumption that variables external to the components of the model (including other attitudes ) do not add to the prediction of intentions and behavior over and above their effects on AB and SN. However the lack of additional effect of three external variables may be due in part. to the small sample size. so caution in interpretation should be extended to these three variables.

Underlying Belief Structures

The Locus of effect of an external variable may also be estimated by analyzing the sample s underlying belief structure which in the case or Ebiei. was verified to be highly related to AB (i.e.. r = .79 p <.001). For example. the theory predicts that whenever an external variable is related to Ag it should also be related to Ebiei and to one or more of the individual beliefs and evaluations underlying attitude.

To estimate these linkages correlations were performed between each of the four external variables that were significantly linked to AB (as shown in Table 2) and the individual behavioral beliefs and outcome evaluations proposed to underlie attitudes (and. ultimately intentions and behavior). These correlations. shown in Table 3 suggest the possible loci of effect of the external variables on attitudes.

TABLE 3

SIGNIFICANT (P < .05) CORRELATIONS BETWEEN EXTERNAL VARIABLES AND BEHAVIORAL BELIEFS

Thus for example, the more favorable one s attitude toward the program itself or toward its star the greater the belief that watching the show would be entertaining relaxing and a nice break from studying and that the program and its characters were enjoyed. Not surprisingly then. one s attitude toward the program was related to one's attitude toward watching the program primarily via program attributes. In addition. however favorable attitudes toward the program and its star were inversely related to beliefs that other perhaps better programs in the same time slot would be missed and that watching the show was a waste of time.

Likewise the two attitudes were related to severaL outcome evaluations. People with favorable attitudes toward the program and its star were less likely to negatively evaluate being kept from sleeping and missing other programs in the same time slot. Furthermore people with favorable attitudes toward the program less negatively evaluated wasting their time. The number of behavioral beliefs and outcome evaluations that were significantly related to program and person attitudes was quite high. Consistent with the model these two attitudes were therefore also related to AB.

DISCUSSION

Findings and Implications

The present results support the sufficiency assumption of Fishbein and Ajzen's (1975; Ajzen & Fishbein 1980) theory of reasoned action for a selected set of external variables. The effects of attitudes toward objects end other external variables on viewing intentions and behavior appear to have been mediated by effects on the attitude and normative components of the model. As noted earlier the sufficiency assumption concerning attitudes toward objects has received support based on correlational (including partial correlations) and regression analyses (see e.g. Ajzen & Fishbein 1980). The present results extend that support for a television viewing behavior by analyzing the overall goodness-of-fit of hierarchical models.

Past research has generated mixed results concerning the effects of past behavior on present behavior (Bentler & Speckhart 1979, 1981, Bagozzi 1980). In the present study, past viewing behavior for the particular behavior investigation was not measured. However, it was found that effects of past viewing behavior in general (i.e. estimated number of hours per week spent watching TV) appeared to be mediated through behavioral attitudes. More thorough investigations of the types of external variables that directly rather than indirectly affect behavior are needed.

Other correlational data between model components generally supported the assumptions of Fishbein and Ajzen's model. Respondents' viewing intentions were well predicted from their viewing attitudes and subjective norms, and viewing intentions were usually related to actual viewing behavior. Data also suggested, though, that a breakdown in the intention-behavior relationship was more likely to occur for intenders than for non-intenders. An "inertia" hypothesis (see Davidson & Beach 1981) to explain such effects was discussed, although other factors, such as ability to carry out the behavior, may also contribute to this effect. Furthermore, if conditional probabilities had been measured, (e.g., the probability of an intender watching the program given that he/she watched television at that time) prediction among intenders may have increased (see Warshaw 1980). A comparable difference between intenders and non-intenders may exist for purchase decisions, although this phenomenon or its possible cause has not been widely researched. Intentions to not buy a product may better predict purchase behavior than intentions to buy the product, an hypothesis consistent with past research data (e.g.. Banks 1950, Katona 1960).

Finally, results showed that the evaluative semantic differential measure of attitude (AB) tended to be related to respondents' overall behavioral belief structure weighted by outcome evaluations (Ebiei). Although correlational data are reported here, the theory proposes a causal connection between the two variables. Clearly. more research is necessary to determine whether such causal relations exist. However. to the extent they do, certain implications should be noted. By examining these belief structures, the theory suggests a means for identifying the matter in which attitudes toward persons or objects (e.g., attitude toward the actor. attitude toward the program) may influence behavioral intentions and behavior indirectly. Since respondents' viewing intentions were primarily related to their viewing attitudes, and their viewing attitudes were well predicted by considering their behavioral beliefs and outcome evaluations, the effects of external variables on intentions may be located to some extent within this cognitive structure. For example, attitudes toward the program and its star tended to be related to features of the program itself. Thus, for this particular program. enjoyment of the program's characters and the program itself may have played a strong role in decisions to watch the program. However, results suggest that other behavioral beliefs, including scheduling and time-saving considerations were also salient to respondents. Knowledge of these latter types of variables may also be important for understanding the factors that affect viewing intentions and behavior. Furthermore, although the role of the normative component was less than that of the behavioral attitude, the effects of norms were present for two of the external variables. Salient normative beliefs and motivations to comply, i.e., the cognitive structure proposed to underlie subjective norms, were not measured in the present study. However to obtain a more complete picture of the locus of effect of these variables on behavior, the model proposes that these variables be measured as well.

Other types of external variables, such as demographic and life-style characteristics that have frequently been used by marketers to try to understand consumer behavior, could be similarly investigated (see Figure 1). Analogous to arguments pertaining to attitudes toward objects the model suggests that demographic lite-style and other variables should affect intentions only to the extent that they influence the belief structures underlying intentions. Future research is needed to validate these assumptions. Directing advertisements toward high income groups or toward north central states. or toward all persons aged 18-34. has practical value; these segments would probably be easier to identify than a segment that holds a particular belief structure. However knowledge of the mediating factors that underlie consumer behavior should assist researchers in understanding why such relationships between external variables and behavior exist. and when these relations will be most likely to occur .

Study Limitations

Several qualifications should be noted that pertain to these results. First as noted already. correlation and regression analyses are unable to provide unequivocal conclusions about causal relations between variabLes. Rather the data may be said to be consistent with the notion that attitudes toward performing the behavior and subjective norms mediate the effects of other variables on behavior and behavioral intentions. In addition. the sample size (N = 53) was somewhat small high. for certain external variables. may have biased results in favor of support of the model tested (see Bentler & Bonett 1980). However an inspection of chi-squared difference tests shown in Table 2 shows that the differences between the null model and the proposed model were quite large. generally supporting the proposed model. The incremental fit indices for the two A0 variables suggest that the sample data covariance matrix is well represented by the Fishbein and Ajzen model. Since fit indices for the other three variables were lower larger samples may be needed to determine whether the same conclusions would hold for these three variables.

Another qualification of these results involves self-report behavioral measures which may have introduced unknown sources of error. Furthermore the particular behavior investigated was selected by the sample as a whole and as such. may have been quite familiar to many of the respondents. To this extent the cognitive structures pertaining to this behavior may have been well-developed conceivably resulting in better prediction of the proposed model.

Finally mans important external variables such as past behavior or those shown in Figure 1, were not measured in the present study. The mediating effects of these variables and their sources of influences on behavior, the two components of the model. and underlying belief structures is an important topic for future research.

REFERENCES

Ajzen I. & Fishbein M. (1970) "The Prediction od Behavior from AttitudinaL and Normative Variables", Journal of Experimental Social Psychology, 6, 466-487.

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

Ajzen I. & Fishbein M. (1980) Understanding Attitudes and Predicting Social Behavior, Englewood Cliffs, N.J.: Prentice-Hall.

Bagozzi, R. P. (1980). Causal Models in Marketing. New York: John Wiley & Sons.

Bagozzi, R. P. (1981). Attitudes. Intentions. and Behavior: A Test of Some Key Hypotheses," Journal of Personality and Social Psychology 41 607-62;.

Banks S. (1950) The Relationship Between Preference and Purchase of Brands Journal of Marketing 15 (October) 145-1;7.

Bentler P. M. & Bonett D. G. (1980) Significance Tests and Goodness of Fit in the Analysis of Covariance Structures Psychological BuLletin 88. (November) 588-606.

Bentler P. M. & Speckhart (1979) Models of Attitude-Behavior Relations Psychological Review 86 (September) 452-464.

Bentler. P. M. & Speckhart (1981) Attitudes Cause Behaviors: A Structural Equation Analysis Journal of Personality and Social Psychology 40 (February) 726-238.

Davidson A. R. & Beach. L. R. (1981) Error Patterns in the Prediction of Fertility Behavior Journal of Applied Social Psychology 11 675-488.

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

Fishbein M. (1979) A Theory of Reasoned Action: Some Applications and Implications in H. Howe & M. Page. eds. Nebraska Symposium on Motivation Lincoln: University of Nebraska Press.

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

Joreskog. K. (1973) A General Method for Estimating a Linear Structural Equation System in A. S. Goldberger & O. D. Duncan eds. Structural Equation ModeLs in the Social Sciences New York: Seminar Press, pp. 85-112.

Joreskog S. and Sorbom D. (1978) LISREL IV: Analysis of Linear Structural Relationships by the Method of Maximum Likelihood Chicago: National Education Resources , Inc.

Katona. G. (1960) The Powerful Consumer. New York: McGraw-Hill Book Co., pp. 80-83.

Loken, B. & Fishbein M. (1980) An Analysis of the Effects of Occupational Variables on Childbearing Intentions Journal of Applied Social Psychology 10, 202-223.

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

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

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

Warshaw P. R. (1980) Predicting Purchase and Other Behaviors from General and Contextually Specific Intentions Journal of Marketing Research 17 (February), 26-33.

Wilkie, W. L. and Pessimier E. A. (1973) Issues in Marketing's Use of Multi-Attribute Models Journal of Marketing Research. 10 November 428-441.

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