The Conceptual Organization of Behavior and Attitude-Behavior Consistency

Mahmood M. Hajjat, Ohio State University
ABSTRACT - The attribute-processing perspective views behavior as being driven primarily by cognitions and as being instrumental to the realization of consumption objectives. Alternatively, the recent surge of interest in the role of affect in buying behavior suggests that behavior may be consummatory (noninstrumental) and driven by emotions. Therefore, the validity of the single-component perspective was tested against a two component (consummatory and instrumental) model of behavior. Additionally, the effect of focusing on an attitude component (affective/cognitive) on attitude-behavior relation was examined. Results indicate that the two-component model achieved discriminant and convergent validity whereas the single-component model did not. Moreover, attitude-behavior correlations were higher when attitude and behavior were influenced by the same attitude component than when they were influenced by different attitudinal components.
[ to cite ]:
Mahmood M. Hajjat (1990) ,"The Conceptual Organization of Behavior and Attitude-Behavior Consistency", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 777-784.

Advances in Consumer Research Volume 17, 1990      Pages 777-784

THE CONCEPTUAL ORGANIZATION OF BEHAVIOR AND ATTITUDE-BEHAVIOR CONSISTENCY

Mahmood M. Hajjat, Ohio State University

ABSTRACT -

The attribute-processing perspective views behavior as being driven primarily by cognitions and as being instrumental to the realization of consumption objectives. Alternatively, the recent surge of interest in the role of affect in buying behavior suggests that behavior may be consummatory (noninstrumental) and driven by emotions. Therefore, the validity of the single-component perspective was tested against a two component (consummatory and instrumental) model of behavior. Additionally, the effect of focusing on an attitude component (affective/cognitive) on attitude-behavior relation was examined. Results indicate that the two-component model achieved discriminant and convergent validity whereas the single-component model did not. Moreover, attitude-behavior correlations were higher when attitude and behavior were influenced by the same attitude component than when they were influenced by different attitudinal components.

INTRODUCTION

A basic tenet of the attribute-processing perspective is that the comprehension of the brand attribute claims leads to a change in consumers' stored cognitions. Thus, consumers are postulated to be problem-solvers and rational buyers who "base their decisions on the persuasive information provided" (Day 1973). The logical flow of the stages of cognitive processing (e.g. attribute comprehension-attitude-behavior-problem solving) assumes causation and assumes that behavior is instrumental to the realization of enduring goals (e.g., problem solving). Although it continues to explain important consumption phenomena, the attribute-processing perspective has failed to produce consistent findings regarding attitude-behavior consistency (e.g. Ajzen and Fishbein 1977).

The restricted conceptualization of attitude and behavior in the attribute- processing perspective has contributed greatly to this failure. For example, it is frequently argued that attitude may not be formed from decomposable origins (e.g. physical attributes) rather it may "involve a gestalt, configural appraisal of the stimulus object, going beyond the assessment of the utility contributed by each individual attribute" (Batra 1986). It is also generally accepted that consumption may not involve solving a problem, or realizing other such enduring goals. It may involve instead some sensory pleasures, joy, and emotional responses (Holbrook and Hirschman 1982). It appears that there are situations in which attitude would be based on affect rather than cognitions and consumption goals would be transient rather than enduring. In such situations, behavior should be viewed as being consummatory (e.g. to obtain pleasure) and driven by affect rather than instrumental (e.g. to achieve enduring goals) and driven by cognitions.

This article has two purposes. The first is to provide empirical tests of the validity of these conceptualizations by contrasting a single-component with a two-component (consummatory and instrumental) model of behavior. The second is to provide empirical evidence as to whether focusing on either attitude component (affective/cognitive) affects attitude-behavior relation.

The next section is a discussion of the conceptual organization of attitude. It is followed by a conceptual analysis of behavior and a discussion of the effects of thought on attitude-behavior consistency. Finally, an experiment designed for achieving the research objectives is outlined and the findings are thoroughly examined.

ATTITUDE ORGANIZATION

The attitude concept was originally defined as being a single-component entity of likes and dislikes (e.g., Bem 1970). This concept was subsequently reconceptualized to include cognition and conation in addition to affect (e.g., Kothandapani 1971, McGuire 1969). The latter definition has received some empirical support. ln an experiment that was designed for predicting contraceptive behavior among low income Negro women, Kothandapani {1971) examined the discriminant and convergent validity of this tripartite classification. He found convincing evidence both to support this classification and to conclude that conation was a better predictor of contraceptive behavior than either of the other two components of attitude (Kothandapani 1971).

However, the inclusion of conation in attitude structure is not universally accepted. This is because the difficulty of directly observing overt behavior has resulted in the use of conation as a surrogate for behavior by many researchers. This conceptual equivalence explains findings of higher conation than affect or cognition-behavior correlations. Moreover, a behavior involving the use of a drug (e.g., contraceptive) that could produce genuine health problems might be more influenced by a medical decision (e.g., prescription) than by affect or cognition. Thus a midpoint position which contends that attitude is composed of only affect and cognition is adopted here. At the same time, conation is considered to be an evaluative response based on the more accessible, more salient of only two attitude components - affect or cognition (e.g., Bagozzi and Burnkrant 1979, Millar and Tesser 1986). The affective component of attitude toward a brand is thought to contain the feelings (e.g., pleasure, happiness, joy) that may be evoked by the brand and the cognitive component is thought to contain the encodings of physical attributes of, and beliefs about the brand (e.g., price, size, effectiveness). Bagozzi and Burnkrant (1979) examined the validity of a single-component model of attitude and compared it to a two-component model. They reported evidence to support an affective-cognitive conceptualization and lack of evidence to support a single-component model of attitude.

BEHAVIOR ORGANIZATION

Much research indicates that all products, no matter how mundane, carry attributes such as packaging, color, or taste that are not central to the objective value (e.g., effectiveness) of the product symbolic features. It is also known that people make decisions so as to experience an emotion or to realize some type of utility. For example, a person may drive to a shopping mall on a nice evening just to enjoy him/herself by looking at store windows while another may do the same thing to maintain good health by buying a needed prescription. Similarly, a customer may purchase a product (e.g., toothpaste) because its symbolic features (e.g., taste) are richer and more salient than its objective attributes (e.g., fluoride) while another may do the same thing for just the opposite reason.

It appears that purchase behavior can be differentiated along customers' expectations from consuming the product (e.g., freshness/healthy gums from consuming toothpaste). Thus, purchase behavior that stems primarily from the pursuit of sensory pleasures, happiness, or any emotional responses consummated coincident with consumption of the product is different from purchase behavior that stems primarily from the pursuit of objectives that last beyond the time of actual consumption. These two types of behavior depend on the primary stimuli, and are called consummatory behavior (primary stimuli are immediate consummatory emotional responses) and instrumental behavior (primary stimuli are enduring objectives).

Figure 1 shows path diagrams of a single-component and a two-component model of behavior. A confirmatory factor analysis using LISREL VI (Joreskog and Sorbom 1984) was used for testing the fit of each model.

The single-component model achieves convergent validity if (a) responses that differentiate consummatory from instrumental behavior exhibit high intercorrelations and (b) an insignificant chi-square is obtained. Discriminant and convergent validity of the two-component model will be achieved if (a) within-component correlations are high, (b) across-component correlations are logically consistent and significantly lower than within- component correlations, (c) rho-statistic (Bentler and Bonett 1980) is within an acceptable range (p>0.90), and (d) root mean square residual (RMSR) is within an acceptable range (RMSR<0.10). A final test of the overall fit of both models based on an inferential evaluation of nested models (Long 1988) could be carried out to determine which model fits the data better. Based on this analysis, the following hypotheses were tested:

H1: A two-component (consummatory and instrumental) model of behavior achieves discriminant and convergent validity whereas a single-component does not.

H2: A two-component (consummatory and instrumental) model of behavior fits the data better than a single-component model.

EFFECTS OF THOUGHT ON ATTITUDE-BEHAVIOR RELATION

In recent years, research in attitude-behavior relation has proliferated and has primarily examined the conditions under which attitude and behavior would or would not correlate (Wilson, Dunn, Bybee, Hymann, and Rotondo 1984). An area that has captured much attention, and is also the focus of this article, is the effect of affective and cognitive focus on attitude-behavior consistency. In one study, Fazio, Zanna, and Cooper (19783 stimulated their subjects to think about the attitude objects (puzzles) by instructing them to empathize with a similar individual on a videotape who was playing with the puzzles. Fazio et al (1978) found that when subjects were instructed to think about the attitude objects (puzzles), the amount of salient information about the attitude objects increased. The attitudes of those subjects who were led to focus on the attitude objects were also found to be better predictors of behavior than subjects' attitudes in the control condition (who did not think about the attitude objects).

In contrast, Wilson et al. (1984), using a similar paradigm, found results that were contradictory to those reported by Fazio et al. (1978). Wilson et al. (1978) predicted that analyzing the reasons-for one's own attitude (thinking in terms of why a person feels the way she does) would shift attitude in a direction away from behavior, thus attitude-behavior correlation would be low. In a series of experiments using puzzles, vacation scenes, and dating mates as attitude objects, Wilson et al. (1978) found that when subjects analyzed the reasons of why they would or would not like the attitude object, their self-reported attitudes accounted for only 1% of the variance of behavior while subjects' attitudes in the control condition were found to account for approximately 30%.

In an attempt to provide a resolution to the apparent contradictory findings, Millar and Tesser (1986) conducted an experiment in which they used puzzles, similar to those used in Wilson et al. (1978), as attitude objects. Millar and Tesser (1986) reasoned that if attitude report and behavior emanated from the same attitudinal component (either affective or cognitive), attitude-behavior correlation would be high. And, if behavior was initiated by an attitudinal component that would be different from that component dictating attitude report, attitude-behavior correlation would be low.

FIGURE 1

PATH DIAGRAM OF A SINGLE-COMPONENT AND A TWO-COMPONENT MODEL OF BEHAVIOR

In Millar and Tesser (1986), the time spent playing with each puzzle, the proportion of each puzzle chosen, and the order of choice in a free-play period were recorded for each subject. The play time for subjects in the cognitive and affective-focus conditions was correlated with their attitude reports. These correlations were higher when behavior and attitude reports were driven by the same attitude component (a match situation) than when each was driven by a different component (a mismatch situation). However, one could argue that some methodological issues in Wilson et al. (1984) and in Millar and Tesser (1986) may have contaminated their findings. In formulating a general evaluative response about an attitude object, some of its salient objective attributes or symbolic features would be more important than others. Thus, it is possible that subjects' attitudes toward the attitude objects (puzzles were used in both studies, and vacation scenes and dating mates were used in the first) might have had more symbolic or hedonic than physical or attribute origins. For example, the joy of triumph or the agony of defeat probably influences one's liking or dislike of a puzzle far more than the puzzle's contribution to one's analytic ability. Thus, since Wilson et al's (1984) subjects did not have a large number of cognitions to begin with, they may have generated a biased set of reasons, just to appear more analytical, that might not have reflected their true attitudes.

Similarly, since Millar and Tesser's (1986) subjects in the cognitive-focus condition did not have many relevant attributes about the puzzles to start with, they may also have generated a biased set of reasons, but to a lesser extent than in Wilson et al. (1984). In fact, Millar and Tesser (1986) reported insignificant differences between affective and cognitive-focus condition correlations when subjects' behavior was driven by cognitions, and the proportion of puzzles chosen and order of choice were used as behavioral measures.

Second, since Millar and Tesser (1986) did not include a control cell to compare with, it is difficult to conclude whether analyzing reasons in the cognitive-driven behavior condition increased attitude-behavior correlations or just did not decrease them. For the same reason, it is also difficult to conclude whether focusing on the affective component of attitude in the affective-driven behavior condition increased attitude-behavior correlations or simply did not decrease them. Third, considering the time span over which Millar and Tesser's (1986) study was completed (a few minutes), it could be argued that salient-attitude reports in the cognitive focus condition might have habituated subjects' behavioral responses when behavior was driven by the same cognitions. Thus, the observed consistency between attitudes based on cognitions and behaviors driven by these cognitions might have been inflated.

While taking these issues into consideration, the conceptual parallel between affective (cognitive) attitude that is based on feelings (cognitions) and consummatory (instrumental) behavior that is stimulated by symbolic (objective) attributes of the product was maintained in the present research. In the course of doing so, a research paradigm that is similar to those used by Wilson et al. (1984) and Millar and Tesser (1986) was utilized. Second, the attitude objects employed in this research were products (toothpaste brands) that would have different symbolic origins (e.g. packaging, color) for the hedonic component of attitude and several physical origins (e.g. fluoride, abrasive) for the cognitive component. Third, a control group was included and attitude-behavior correlations for all experimental cells were compared with those of the control group. Finally, subjects' actual behavior was observed one week after obtaining self-reported attitudes. The following hypotheses were tested:

H3: If attitude report and behavior are influenced by the same attitude component (affective/cognitive), attitude-behavior correlation would be high; whereas if they are influenced by different attitude components, attitude behavior correlation would be low

H4: Focusing on the attitude component that drives behavior would result in a high attitude-behavior correlation; whereas the lack of focus on a specific attitude component would result in a low attitude-behavior correlation.

METHOD

Fifty six male and fifty two female undergraduate students from a marketing course at a major mid-western university completed a short questionnaire that was designed to measure initial attitudes and initial behavioral intentions. After collecting this instrument, subjects were randomly assigned to one of 2 (affective/cognitive attitude focus) / 2 (consummatory/instrumental behavior) cells and a control condition. Subjects completed another questionnaire (ostensibly for a different experiment), were thanked, and promised a gift (was intended to measure actual choice) in the coming days.

Procedure

On the first questionnaire, subjects provided some demographical data; recalled as many brands as they could -from four different product classes including toothpaste, ranked the recalled brands, and indicated their preferences. The purpose of this instrument was to facilitate the contrast of subjects' initial to post-experimental attitude-behavior correlations. The second questionnaire contained the following regarding each of five toothpaste brands: (a) a black-and-white picture of the brand, (b) a 9-point bipolar scale for measuring the attitude toward brand, (c) eight 9-point bipolar scales for measuring behavioral responses that, based on a pretest, were found to differentiate consummatory from instrumental behavior, (d) a thought-listing question for measuring the degree of affective and cognitive focus, and (d) a purchase intention question. This instrument was designed to measure the differential effects of experimental manipulations.

Subjects were approached by an experimenter who, as a cover story, claimed that the purpose of the first questionnaire was to measure market share for the brands that subjects could recall. They were given two minutes for each product category and were instructed not to turn pages unless they were told. After collecting the first instrument, participants were approached by another experimenter who distributed the second instrument and asked them to read and follow all instructions carefully. They were given all the time they needed for examining the picture of any brand and for answering the questions.

Subjects were asked to keep a little numbered post-it note from the first questionnaire, attach the note to the second questionnaire, write the number of the note on a little card attached to the second questionnaire, and keep the card to claim the gift. When gifts were claimed a week later, each subject deposited his/her claim card in the same box from which the toothpaste tube was chosen. Finally, the researcher matched the questionnaires and the card to isolate the responses of each individual.

Experimental Manipulations

An important issue in this research was whether to use fictitious (novel) or familiar and popular brands as attitudinal objects. The selection of well- known over fictitious brands was governed by a number of practical and theoretical considerations. These were: (a) an attempt to resemble a marketing situation as close as possible, (b) an attempt to measure overt behavior under the volitional control of subjects and correlate this index of observed behavior with attitude so as to properly illustrate attitude-behavior consistency, (c) an attempt to equate all recall cues across brands and subjects (the brand picture was the only cue), (d) an attempt to stimulate subjects to report real sentiments from their past experiential consumption episodes, and (e) an attempt to eliminate the likelihood of generating artificial (biased) inferential information, thoughts, and sentiments that would be otherwise increased if written attributes, needed with novel brands, were provided.

Subjects in the affective focus condition were instructed to focus on how they felt about each brand while they-were viewing its picture, whereas those in the cognitive focus condition were instructed to think about the reasons that made them feel the way they did. The purpose of the brand picture was either to remind subjects of the sentiments and feelings they experienced whenever they used the product in the past or to aid in recalling the stored attributes of, and beliefs about the product. Subjects in the consummatory behavior condition were led to believe that they would receive a social sensitivity test (how well they would fit with others) at the end of the experiment, whereas those in the instrumental behavior condition would receive a brand and attribute-recall test (how much information they could recall). Neither group was actually tested. To get maximum help on the anticipated tests, subjects in the consummatory behavior condition should have heightened their stored feelings and emotions upon viewing the pictures. While those in the instrumental behavior condition should have accessed their stored belief and brand-attribute information. Subjects in the control condition were not given attitude focus and behavior manipulation instructions.

Attitude and Behavior Measures

Each subject indicated on 9-point bipolar scales how much s/he liked or disliked each brand. One on the scale indicated dislike very much and 9 indicated like very much. Behavior was observed one week after completing the questionnaires; each subject chose and received one tube of toothpaste from the five brands used.

Affective and Cognitive Focus Measures

Subjects listed all the thoughts that occurred to them while they were examining the picture of each brand. These thoughts were later coded by two judges into feelings and reasons. The two coding sets were highly correlated (r=0.93). The feelings represented the cognitive focus.

Responses Differentiating Consummatory From Instrumental Behavior

Each subject indicated on 9-point bipolar scales how- much s/he agreed or disagreed with four consummatory and four instrumental behavior statements regarding each brand. The representativeness of these statements to consummatory and instrumental behavior was established by a pretest of twenty subjects from the same population. Consummatory behavior statements tapped the degree of happiness (happy), joy (joy), cheerfulness (cheer), and refreshment (refresh) felt by each subject if the brand of toothpaste under consideration was used. Instrumental behavior statements measured the extent of belief that the brand of toothpaste might contribute to gum health (health), cavity reduction (cavred), teeth brightness (bright), and tartar buildup reduction (tarred). On all scales, 1 indicated completely disagree and 9 indicated completely agree.

ANALYSIS AND RESULTS

Confirmatory Factor Analysis

For the single-component model to achieve convergent validity and provide a good fit, the variables differentiating consummatory from instrumental behavior should exhibit high intercorrelations and an insignificant chi-square should be obtained. The test of this model using LISREL VI gave zero correlations and a significant chi-square (%28=3629, P=0.0). Thus, the single component model did not achieve convergent validity and it could not be accepted without testing a two-component model.

In the two-component model, the first set of these variables (happy, joy, cheer, and refresh) was hypothesized to have high loadings (l11-l41) on consummatory behavior and zero loadings (l51-l81) on instrumental behavior. The second set (health, cavred, bright, and tarred) was hypothesized to have high loadings (l52-l82) on instrumental behavior and zero loadings (l12-l42) on consummatory behavior. The correlation (f21) between the two components of behavior should not be high.

The correlation matrix used in the analysis (not included) showed that variables loading on consummatory behavior were highly correlated (r=0.70 to r=0.94) and variables loading on instrumental behavior were highly correlated (r=0.61 to r=0.83) while across-component correlations were low. LISREL estimates showed high loadings from the first set of variables on consummatory behavior (happy=0.962, joy=0.972, cheer=0.956, and refresh=0.735), high loadings from the second set of variables on instrumental behavior (health=0.918, cavred=0.877, bright=0.731, and tarred=0.800), low correlation between consummatory and instrumental behavior (f21=0.37), high Bentler and Bonett's rho statistic (p=0.90), and a low root mean square residual (RMSR=0.1). These results indicate that the two-component model (consummatory and instrumental) achieved discriminant and convergent validity.

FIGURE 2

MEAN FEELINGS AND MEAN REASONS GENERATED IN THOUGHT LISTING PROTOCOL

The last goodness-of-fit test conducted here was an inferential evaluation of nested models. Since the free parameters in the single-component constitute a subset of the free parameters in the two-component model, it was possible to test a null hypothesis that the single-component fits the data better than the two-component model. If the null hypothesis could not be rejected (by obtaining an insignificant chi-square 1, then the single-component would be more desirable than the two-component model because the first would be more parsimonious than the second. The test resulted in a significant chi- square (X9=3472, p=0.0), thus the null hypothesis was rejected and it was concluded that the two-component model of behavior (consummatory and instrumental) fits the data better than the single-component.

Attitude-Behavior Relation

The number of reasons for each subject was analyzed in a 2 (consummatory vs. instrumental) / 2 (affective vs. cognitive) / 2 (male vs. female) ANOVA. The only significant source of variation was thought focus, F( ,86)=9.45, P=0.0028. Subjects in the cognitive focus condition produced more reasons (M=0.75) than did subjects in the affective focus condition (M=0.39), F(1,86)=4.58, P=0.0051. The number of feelings produced by subjects was also analyzed in a similar ANOVA and thought focus was the only significant source of variation, F(1,86)=17.77, P=0.0001. Subjects in the affective focus condition generated more feelings (M=1.15) than did subjects in the cognitive focus condition (M=0.47), F(1,86)=6.96, P=0.0003 (see figure 2).

Two measures of behavior were constructed and correlated with subjects' self-reported measures of liking and dislike of the five brands of toothpaste, actual choice of toothpaste and the tube size. Correlations were transformed to z-scores and analyzed in two separate 2 (consummatory vs. instrumental) / 2 (affective vs. cognitive) / 2 (male vs. female) ANOVAs. The two ANOVAs gave identical results. The-only significant interacting effect was type of behavior / type of focus, F(1,86)=16.9, P=0.0001. Subjects in the instrumental behavior condition had higher attitude-behavior correlations- when they focused on their cognitions (M=0.66) than when they focused on their feelings (M=0.38), F(1,40)=2.77, P=0.05. Similarly, subjects in the consummatory behavior condition had higher attitude-behavior correlations when they focused on their feelings(M=0.63) than when they focused on their cognitions (M=0.31), F(1,44j=3.03, P=0.004. It is interesting to note that subjects in the control condition had attitude-behavior mean correlations of M=0.40 (see figure 3).

DISCUSSION

This article has demonstrated that consumers behavior can be differentiated according to consumers' expectations from consuming the product. A two- component model of behavior achieved discriminant and convergent validity whereas a single-component model did not. Focusing on the attitude component (affective/cognitive) that drives behavior (consummatory/ instrumental) increased attitude behavior correlation above that in the control condition (when subjects were not instructed to focus). Alternatively, focusing on the attitude component that does not drive behavior decreased attitude-behavior correlation below that in the control condition.

FIGURE 3

MEAN CORRELATIONS OF ATTITUDE AND BEHAVIOR

It is possible that subjects' age and purchase experience may have contributed to the observed differences in attitude-behavior correlations. This possibility was investigated in a number of additional analyses. First, a 2 (type of behavior) / 2 (type of focus) / 2 (sex) / 4 (purchase frequency) / 5 (age group) ANOVA indicated that significant differences between mean correlations at 0.08 significance level still existed and the only two-way interaction to reach significance was type of focus / type of behavior, F(1,86)=5.36, P=0.02. Second, a 5 (age group) / 4 (purchase frequency) / 2 (sex) ANOVA indicated that no significant main or interacting effect was produced by any of these variables, F(25,62)=1.41, P=0.14. Third, mean correlations of attitude and behavior, attitude and behavioral intention, and initial attitude and initial behavioral intention for experimental conditions, were compared to those for the control condition. This analysis showed that there were no significant differences between mean correlations of the five groups (F<1.87, P>0.15) except between attitude-behavior mean correlations. These findings clearly indicate that factors other than the type of focus and the type of behavior do not explain the observed difference in attitude- behavior correlations

CONCLUSION

Though the attribute processing perspective has yielded valuable insights into consumer attitude-behavior relation, it falls far short of completely explaining the consumption phenomena. The findings of this research provide partial support to the notion that purchase behavior is largely dependent on the expectations of the consumer in consuming the product. Current findings are also consistent with findings in social psychology, and are apparent in some of the current advertising practices. For example, AT&T appeals to both behavioral components by airing two different advertisements, "reach out and touch some one" and the "price of the call" ads. Toothpaste commercials are usually diversified to appeal to the majority of consumers by emphasizing the benefits (e.g. fewer cavities) andlor stressing the consumption experience (e.g. fresh breath) consumers can expect from using the brand. Even beef is advertised as a red, tasty, and delicious meat and as a good source of necessary nutrition.

REFERENCES

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

Bagozzi, R. and Burnkrant, R. (1979), "Attitude Organization and the Attitude-Behavior Relationship," Journal of Personality and Social Psychology, Vol.37, No.6, 913-29.

Batra, R. (1986), "Affective Advertising: Role, Processes, and Measurement," in The Role of Affect in Consumer Behavior: Emerging Theories and Applications, R. Peterson, W. Hoyer, and W. Wilson (eds.), Lexington, MA: D.C. Heath and Company.

Bem, D. J. (1970), Beliefs, Attitudes, and Human Affairs, Belmont, Calif.: Wadsworth Publishing Company, Inc.

Bentler, P. and Bonett, D. (1980), "Significance Tests and Goodness-of-Fit in the Analysis of Covariance Structure," Psychological Bulletin, 88, 588-606.

Day, G. (1973), 'Theories of Attitude Structure and Change," in Consumer Behavior: Theoretical Sources, Ward and Robertson (eds.), Engle Wood Cliffs. NJ: Prentice-Hall.

Fazio, R., Zanna, M., and Cooper, J. (1978), "Direct Experience and Attitude-Behavior Consistency: An Information Processing Analysis," Personality and Social Psychology Bulletin, Vol.4, No. l, 48 -51.

Holbrook, M. and Hirschman, E. (1982), 'The Experiential Aspects of Consumption: Consumer Fantasies, Feelings, and Fun," Journal of Consumer Research, Vol.9, 132-40.

Joreskog, K. and Sorbom, D. (1984), LISREL VI: User's Guide, Mooresvill, Indiana: Scientific Software, Inc.

Kothandapani, V . (1971), "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, Vol.19, No.3, 321-33.

Long, J. (1988), Confirmatory Factor Analysis, Sage University Paper Series on Quantitative Applications in the Social Sciences 07-033, Beverly Hills and London: Sage Pubns.

McGuire, W. J. (1969), 'The Nature of Attitudes and Attitude Change," in Handbook of Social Psychology, Vol. m (2nd ed.), G. Lindzey and E. Aronson (eds.), Reading, Mass.: Addison-Wesley Publishing Company, Inc.

Millar, M. and Tesser, A. (1986), "Effects of Affective and Cognitive Focus on the Attitude-Behavior Relation," Journal of Personality and Social Psychology, 51, 270-76.

Wilson, T., Dunn, D., Bybee, J., Hyman, D., and Rotondo, J. (1984), "Effects of Analyzing Reasons on Attitude-Behavior Consistency," Journal of Personality and Social Psychology, Vol.47. No.l. 5-16.

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