Effect of Initial Trial of a New Product on Attitude-Behavior Consistency

ABSTRACT - A laboratory experiment was conducted demonstrating that initial experience with a new product reduces the importance of nonattitudinal measures in predicting future behavior. This effect was only slightly moderated by the outcome (good, neutral, or poor) of the initial experience. These findings are discussed in terms of their implications for practitioners and theoreticians interested in the prediction and explanation of behavior across situations.


Richard F. Yalch and Carol A. Scott (1977) ,"Effect of Initial Trial of a New Product on Attitude-Behavior Consistency", in NA - Advances in Consumer Research Volume 04, eds. William D. Perreault, Jr., Atlanta, GA : Association for Consumer Research, Pages: 81-86.

Advances in Consumer Research Volume 4, 1977   Pages 81-86


Richard F. Yalch, University of Washington

Carol A. Scott, The Ohio State University


A laboratory experiment was conducted demonstrating that initial experience with a new product reduces the importance of nonattitudinal measures in predicting future behavior. This effect was only slightly moderated by the outcome (good, neutral, or poor) of the initial experience. These findings are discussed in terms of their implications for practitioners and theoreticians interested in the prediction and explanation of behavior across situations.


Attitude research in the consumer behavior area has been centrally concerned with the investigation of attitude structures and the development of better measurement techniques despite the considerable evidence that attitudes may not always be highly associated with actual behavior (Bass, Pessemier and Lehmann, 1972; Wicker, 1969). In an attempt to address this issue more explicitly, some marketers have recently begun to examine Fishbein's behavioral intention model, sometimes referred to as the extended model (Bonfield, 1974; Harrell and Bennett, 1974; Mathews. Wilson and Harvey, 1975; Weddle and Bettman, 1973). [See Ryan and Bonfield (1975) for a comprehensive review of marketing studies using the Fishbein model.] In the rush to validate and apply this model, however, consumer researchers have ignored studies which point out limitations of the model (Brislin and Olmstead, 1973; Schwartz and Tessler, 1972) and which propose other methods for improving the explanation and prediction of behavior (e.g., Sample and Warland, 1973; Wicker, 1971).

This paper reports the results of an experiment designed to test the most frequently suggested method for enhancing behavior prediction, multiple prediction models developed under experimentally manipulated circumstances. Specifically, the study examines the importance of various predictor variables when individuals are either experienced or inexperienced with the product, and when their initial experience has a good, neutral, or poor outcome.


The frequent finding that attitude measures are only weakly associated with actual behavior has motivated researchers to seek alternative methods of enhancing predictions. In particular, the role of other determinants of behavior has received a great deal of attention. This "other variables" approach is the basis for at least two formal multivariate prediction models (Fishbein, 1967; Wicker, 1971) and for many suggestions of possible predictors. As the partial listing of these additional variables presented in Table 1 indicates, generating likely candidates has not been a problem. Rather, the problem is to select from the many possibilities a set of unique predictors that will yield the most accurate prediction and explanation of consumer decisions.



While necessary and important, consideration of these other variables creates several problems for marketers and other researchers since it demands the development of procedures for selecting the most appropriate predictors and may result in longer and more complicated questionnaires. Therefore, any evidence that most of the predictors provide little additional explanatory or predictive power, at least under certain circumstances, would be of great benefit both to those academic researchers working in this area and to practitioners concerned about ignoring potentially useful predictors.

Unfortunately, few of the proposed predictors listed in Table 1 have been subjected to empirical testing. Further, the tests that have been performed typically consist of an evaluation of only one or two predictors. One exception, however, is a study of the use of low phosphate laundry detergents (Brislin and Olmstead, 1973) which included a comparison of the Fishbein (1967) and Wicker (1971) models. Interestingly, maximum prediction accuracy was achieved by a model consisting of just three predictors, attitude toward the low phosphate detergents, judged influence of extraneous events, [This predictor measures the perceived likelihood that situational factors will influence one's performance of the behavior (Wicker, 1971).] and behavioral intentions. The failure of other variables (e.g., normative beliefs) to contribute significant predictive power suggests that in certain situations individuals may have enough prior experience with an activity or object (their attitudes were assessed after they had selected a laundry detergent) to have sufficient consistency in their dispositions such that only a few variables are needed to determine or predict future behavior.

This proposition received additional support from a study of the association between political dispositions and voting behavior in student elections (Sample and Warland, 1973). Dividing their sample into those who claimed to be certain about their attitude responses and those reporting little certainty, these authors found that attitudes predicted behavior well only when respondents were certain about them. Further, the other predictors contributed significantly to the explanation of behavior only for those subjects who were uncertain in reporting their attitudes.

From the foregoing discussion, it would seem that when individuals have reason to know an object or activity well, their attitudes alone guide their behavior. Otherwise, situational factors and nonattitudinal dispositions play an important role. This raises the question as to what might cause individuals to be certain about their attitude statements. Bem (1968) has proposed that attitude statements may merely be self-descriptions based on observations of prior behaviors, i.e., one would report liking a product because one has used it frequently. It would seem, therefore, that when individuals have recent behaviors by which to evaluate their attitude, they should be more certain that their responses truly reflect their feelings. Under these circumstances, attitudes should be reliable predictors of behavior. On the other hand, when there has been no behavior or the behavior occurred a long time ago, attitudes should be relatively weakly associated with behavior, necessitating the consideration of other predictors.

The effects of experience have been evaluated post hoc in an examination of election survey data (Yalch, 1974). Subjects were contacted either before a preliminary local election or during the time interval between this election and a runoff. Responses to a pre-election interview were used to predict whether or not the subject would vote in whichever election first followed the interview. Among the measures taken were attitudes toward the election (A), normative beliefs (NB), judged influence of extraneous events, (EE), intentions to vote (I), and actual behavior (B) as determined from records maintained by poll watchers.

As shown in Table 2, the pattern of associations derived from these two independent samples is consistent with the proposition that performing a behavior leads to attitude stabilization and greater confidence that these attitudes accurately reflect one's feelings toward the object or activity. The correlation between attitude toward the election (A) and voting (B) nearly doubled (.20 to .38, p < .05) when subjects had some recent experience. Further, the association between the extraneous events measure (EE) and behavior also increased significantly (.02 to .23, p < .001). It appears that one's ability to estimate the influence of these factors is also affected by experience in previous situations. The normative beliefs and behavioral intentions measures, however, were not so affected by experience, perhaps reflecting the stable quality of normative considerations relative to attitudinal ones.

Although these findings are consistent with the proposition advanced, they provide only a weak test since the design was a quasi-experiment. That is, subjects were not recruited and then randomly assigned to participate in the first or second survey. Further, subjects' availability at the interview time and other factors such as the occurrence of elections at different times and the difference between the campaigns involved may have contributed to the differences in the associations.



To provide a more stringent test of the proposition that product experience significantly enhances the association between attitudes and behavior such that additional predictors will contribute little to explaining behavior, an experiment was conducted with subjects randomly assigned to experience or no experience conditions. In addition, the effect of the outcome of this experience on the associations was also studied by randomly assigning subjects to have a good, neutral, or poor outcome. Two predictors shown to be among the best in several previous studies, normative beliefs and influence of extraneous events, were included in the questionnaire. The specific hypothesis was that giving subjects an opportunity to sample a test product would improve the correlation between attitudes, extraneous events and behavior, but would have only a slight effect on the associations with normative beliefs. The net result of these changes would be that the nonattitudinal measures (normative beliefs and extraneous events) would no longer contribute significantly to explaining behavior.

Because there has been little research or theorizing about the effect of positive or negative outcomes on attitude-behavior consistency, only a tentative hypothesis was developed. Based on the belief that an experience that is counter to one's expectations is likely to reduce one's confidence in his attitudes, it was speculated that an outcome inconsistent with the subjects' initial expectations about the product would lead to lower associations with future behavior than consistent outcomes. Since most subjects evinced a slightly negative expectation, the hypothesis was that the attitudes of subjects who tasted an unexpectedly good product would have lower associations with behavior than those of subjects in the neutral or poor outcome conditions.



A 2 x 3 experimental design was used to assess the effects of product experience and the outcome of this experience on attitude-behavior consistency. Subjects' attitudes were assessed either before or after they sampled a new sugar free soft drink and the taste was varied such that it was a good, neutral, or poor experience. Upon entering a market research laboratory, subjects were given a three-part questionnaire and asked to respond to the first set of questions. The pretest questions assessed their general attitudes toward sugar free soft drinks. Next, the subjects were asked to test an unknown or a known brand with the choices set up such that most selected the unknown brand. Before actually sampling the product, they responded to the second set of questions reporting their expectations about its quality. All subjects then tasted the product and finished the questionnaire. During the debriefing procedure, the subjects participated in a lottery with the prize being six cans of soft drink. The requested proportion of the test brand served as the behavioral measure.


Participants were 242 male and female residents of Columbus, Ohio, between the ages of 21 and 60, recruited by a commercial firm for a marketing research study. Subjects were randomly assigned to treatment conditions by means of a color-coded blank sheet inserted in the middle of the questionnaire booklets. By working at tables with side panels, subjects were unable to see the others' booklets and were unaware of any differences. Unusable questionnaires were returned by 16 subjects and 11 subjects elected not to test the new product. Thus, a total of 27 subjects were eliminated and the final analyses were conducted with 215 subjects.


Upon arrival at the research firm, the subjects were escorted in groups of 7 to 20 to a test kitchen. One experimenter introduced the study by telling the subjects that it concerned consumer attitudes toward a new sugar free soft drink. The subjects were asked to read the instructions on the first page of their booklet, to answer some preliminary questions on the second page, and them to wait for further instructions.

Experience/No Experience Manipulation. All subjects reported their expectations and likelihood of buying the test product before actually tasting the product. However, half of the subjects also responded to a page of questions on some nonattitudinal variables, influence of extraneous events, normative beliefs, and motivation to comply with these beliefs, before tasting the product. The other half answered these same questions, but after the taste experience. The experience assignment was accomplished by randomly distributing questionnaires with the nonattitudinal measures in part two or part three of the questionnaire. Again, because of the partitions between the work areas, subjects could not observe the differences in the questionnaires.

Outcome Manipulation. Upon completion of the second portion of the questionnaire, subjects were given approximately 2 oz. of the test product to taste. One-third of the participants were given a popular regular soft drink which was judged as good tasting in blind pilot testing; one-third were given a regionally marketed soft drink that did not taste as good as most soft drinks but did not have the unpleasant aftertaste of most diet drinks; and one-third were given a nationally marketed diet drink judged in previous tests to have a bad taste. In all cases, subjects thought they were tasting the product they had selected and that everyone was tasting the same test brand. The subjects were then asked to complete the last section of questionnaire and to wait until everyone had finished.

Predictor Variables

Attitudes (A). Since prior research has shown that attitudes toward an object are as good a predictor of behavior as attitudes toward an action with the object (Brislin and Olmstead, 1973; Mathews et al., 1975; Schwartz and Tessler, 1972), it was decided to assess attitudes in terms of the subjects' evaluation of the test brand. Subjects responded to a ten-item semantic differential with the test brand as the object before and after they tasted it. The responses were factor analyzed separately but the patterns of loadings were virtually identical. The five items loading on the most important factor, which explained 38% of the variance, were clearly identified as the evaluation dimension and summed to form the subjects' attitude toward the product. The bipolar adjectives were: high quality-low quality; refreshing-not refreshing; satisfying-not satisfying; good aftertaste-bad aftertaste; and desirable-not desirable.

Normative Beliefs (NB). There is little consensus about the best way to operationalize this variable (see Fishbein, 1976; Glassman and Fitzhenry, 1976; and Lutz, 1976). Thus, the statements used should be viewed as exploratory attempts until future research provides better guidelines. The perceived response of significant others to the subjects' purchase or use of a sugar free soft drink was assessed using three 5-point agree-disagree attitude statements:

1. My friends would laugh if they saw me drinking a sugar free soft drink.

2. My relatives expect me to drink only sugar free soft drinks.

3. Other people in a supermarket would notice if I purchased a carton of sugar free drinks.

The scoring of the responses to the first question was reversed and the three responses summed to give an overall measure.

Influence of Extraneous Events (EE). Wicker's (1971) contribution to the behavior prediction problem was operationalized as the perceived likelihood that certain events would increase or decrease their consumption of a sugar free soft drink. The five questions used were:

1. 1.If a friend offered you a glass of the test brand, how likely are you to refuse it?

2. 2.If a vending machine were out of everything except the test brand, how likely are you to select it?

3. If your doctor recommended that you only drink sugar free soft drinks, how likely are you to drink the test brand?

4. If the test brand cost 10% more than other soft drinks, how likely is it that you would buy it?

5. If the test brand cost 10% less than other soft drinks, how likely is it that you would buy it?

The response to the first question was reversed scored and added to the others to determine the total score on this measure. A factor analysis of the responses to each of the nonattitudinal questions was run and the pattern of factor loadings was taken as evidence of the convergent and discriminant validity of these questions.

Criterion Variables

Behavioral Intention (I). The subjects were asked to indicate their likelihood of buying the test brand if it were marketed using a seven-point scale with the end points labeled very likely or very unlikely. The question was asked before and after they sampled the test product.

Behavior (B). After all the questions had been completed, one experimenter thanked the subjects for their cooperation. To enhance the perception that the experiment was over, he told the subjects that they had been tasting a new soft drink, revealed the name, and showed them a labeled sample can. In reality, none of the subjects had tasted the identified product. The experimenter further explained that as a token of appreciation, a lottery would be held for their group and the winner would receive six cans of soft drink which could consist of the test brand, a popular regular soft drink, or any combination of the two. The subjects wrote down which prize they wanted on their questionnaires. The second experimenter collected them, shuffled the questionnaires, and had someone in the group select the winner. The number of cans of the test brand written down on each subject's questionnaire was used as the behavioral measure of their desire to consume the product at a later date.


The intercorrelation matrix for the two groups is reported in Table 3. The taste experience significantly enhanced the correlations between behavior and attitudes (from .18 to .68, p < .001) and between behavior and the extraneous events influence measure (from .33 to .54, p < .05), but as predicted did not affect the association with normative beliefs. The results, therefore, are consistent with those obtained in the field setting using a different questionnaire and behavior. The only exception is the significant increase in the association between intentions and behavior as a result of sampling the product.



It was proposed that the behavioral experience would have such an enhancing effect on the attitude-behavior association that the additional predictors would not contribute much to behavior prediction. This was tested by running multiple regressions for each group and comparing the amount of variance explained by the predictors. The results in Table 4 clearly demonstrate



that the importance of these variables is dependent on the level of experience a respondent has had with the product. With no experience, behavior is better predicted by the nonattitudinal measures, normative beliefs, and extraneous events. But for an experienced group, behavior can be explained solely by the attitude measure. Similar results were obtained when purchase intentions assessed after the taste experience were used as the criterion although the effect was not quite as strong.

The second area of interest was investigating how the attitude-behavior association would vary with the outcome of the initial experience. The contribution of each predictor was determined from multiple regressions run separately for each outcome experienced by members of the after taste sample. The results are reported in Table 5.

The speculation that subjects experiencing a product inconsistent with their expectations would evidence the least amount of attitude-behavior consistency received only marginal support. Attitudes were less important relative to the other predictors for the good taste group, which had the most inconsistent experience, but the differences between some of the other groups was small. Further, since subjects in the neutral taste condition had an experience that most closely matched their initial expectations, [This was confirmed by comparing pretest expectations assessed using a 100-point scale (mean = 34.3) with a similar measure taken after the tasting (mean: good = 51.2; neutral = 35.3; poor = 30.3).] they should have had the



highest attitude-behavior consistency. This occurred only when intentions was the criterion variable. Thus, at this point, it is best to conclude that information garnered from actual product trial greatly enhances the attitude-behavior relationship, and this may occur even when individuals are trying products that do not conform to their expectations.


The results of the experiment support the proposition that the relationship between attitudes and behavior is contingent on an individual's experience with the attitude object. Subjects responding to an attitude survey based on generalizations about similar objects (diet drinks) had such low attitude-behavior consistency that their behavior was better predicted by nonattitudinal measures (e.g., the influence of extraneous events). However, attitude-behavior consistency increased substantially after product trial, and the nonattitudinal variables were not needed to predict behavior accurately. Further, this finding was only slightly conditional on how similar the initial trial was to the subjects' expectations about the product's quality.

The finding of a strong relationship between the level of experience and attitude-behavior consistency supports the belief that attitudes are based on information about prior behaviors. Apparently, one's actions are guided by one's attitudes only when these attitudes are well-formed through previous experiences and, based on the results of the voting study discussed earlier, these experiences are recent. Otherwise, situational and normative considerations appear to be the dominant factors.

These findings have several implications for consumer research. First, researchers investigating attitudes for the purpose of predicting consumer behavior (e.g., to estimate demand for a product) should be certain that responses to attitude measures are meaningful. It may be necessary, for example, to inquire about the respondent's experience with the product and about the certainty with which he answers attitudinal questions.

Further, some check on the reliability of responses obtained could be devised. The consistency of an individual's responses to a series of related questions might be a good indication that the person has well-formed, stable attitudes.

Second, the results raise several questions about some traditional consumer research methods. At the very least, researchers must reconsider the accepted policy of asking respondents to answer all questions regardless of whether or not they are familiar with the product or issue. While this may be appropriate for certain types of research questions, it may be inappropriate for others. If the respondent honestly does not know how he feels, or does not have an attitude toward a particular product or issue, the researcher may gain little by having the person guess. Similarly, the practice of basing product decisions on concept test results [Concept testing is a procedure in which individuals evaluate hypothetical products based on verbal descriptions. See Wind (1975) for a discussion.] may be questioned since these attitude measures are not based on actual experiences and might be poor indicators of how individuals will respond when confronted with the real product.

Finally, academic researchers who are beginning to investigate behavior prediction models should view these results as supportive of the effort to develop multiple predictor models since these models are useful in situations in which subjects have not had much experience. Researchers should be cautioned, however, that predictive accuracy may be contingent upon the subject population, the activity to be predicted, and the situation in which this activity will occur. Attempts to develop simple models expected to yield accurate predictions across many situations (e.g., Fishbein's behavioral intention model) may not be the best approach to this problem. Rather, it is important to understand better the nature of the relationship between the behavior and the predictors since the determination of the best set of predictors appears to vary considerably with the circumstances. In this study, attitudes predicted future consumption best when subjects had tried the product, but measures of normative beliefs and the influence of extraneous events were better predictors when the product had not been sampled. It is not unreasonable to suspect that other circumstance factors exist which also affect the determinants of behavior but have yet to be investigated. The paradigm used here appears to be a fruitful approach to these issues.


A. C. Acock and M. L. DeFleur, "A Configuration Approach to Contingent Consistency in the Attitude-Behav-ior Relationship," American Sociological Review, 37 (December 1972),714-726.

F. M. Bass, E. A. Pessemier and D. R. Lehmann, "An Experimental Study of Relationships between Attitudes, Brand Preferences, and Choice," Behavioral Science, 17 (November 1972),532-541.

D. J. Bem, "Attitudes as Self-Descriptions: Another Look at the Attitude-Behavior Link," in Anthony G. Greenwald, Timothy Brock and Thomas M. Ostrom (Eds.), Psychological Foundations of Attitudes (New York: Academic Press, 1968).

P. D. Bennett and G. D. Harrell, "The Role of Confidence in Understanding and Predicting Buyers' Attitudes and Purchase Intentions," Journal of Consumer Research, 2(September 1975) ,110-117.

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

R. W. Brislin and K. H. Olmstead, "An Examination of Two Models Designed to Predict Behavior from Attitudes and Other Verbal Measures," paper presented at the American Psychological Association Meeting, Montreal, Canada, 1973.

F. Dansereau, J. Cashman and G. Graen, "Expectancy as a Moderator of the Relationship between Job Attitudes and Turnover," Journal of Applied Psychology, 59(April 1974) ,228-229.

J. A. Fendrich, "A Study of the Association Among Verbal Attitudes, Commitment, and Overt Behavior in Different Experimental Situations," Social Forces, 45(March 1967) ,347-355.

M. Fishbein, "Attitudes and the Prediction of Behavior,'' in Martin Fishbein (Ed.), Readings in Attitude Theory and Measurement (New York: John Wiley and Sons, 1967).

M. Fishbein, "Extending the Extended Model: Some Comments,'' in Beverlee B. Anderson (Ed.), Advances in Consumer Research, Volume III, Association for Consumer Research, 1976.

M. Glassman and N. Fitzhenry, "Fishbein's Subjective Norm: Theoretical Considerations and Empirical Evidence,'' in Beverlee B. Anderson (Ed.), Advances in Consumer Research, Volume III, Association for Consumer Research, 1976.

G. D. Harrell and P. Bennett, "An Evaluation of the Expectancy/Value Model of Attitude Measurement for Physician Prescribing Behavior," Journal of Marketing Research, 11 (August 1974) ,269-278.

R. J. Lutz, "Conceptual and Operational Issues in the Extended Fishbein Model," in Beverlee B. Anderson (Ed.), Advances in Consumer Research, Volume III, Association for Consumer Research, 1976.

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

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

J. Sample and R. Warland, "Attitudes and Prediction of Behavior," Social Forces, 51(March 1973) ,292-304.

S. H. Schwartz and R. C. Tessler, "A Test of a Model for Reducing Measured Attitude-Behavior Discrepancies," Journal of Personality and Social Psychology, 24(November 1972) ,225-236.

J. N. Sheth, "An Investigation of Relationships Among Evaluative Beliefs, Affect, Behavioral Intentions and Behavior," Working Paper No. 7, College of Business Administration, Pennsylvania State University, 1970.

D. E. Tarter, "Attitude: The Mental Myth," The American Sociologist, 5(August 1970),276-278.

L. G. Warner and M. L. DeFleur, "Attitude as an Interactional Concept: Social Constraint and Social Distance as Intervening Variables between Attitudes and Action," American Sociological Review, 34 (April 1969), 153-169.

D. E. Weddle and J. R. Bettman, "Marketing Underground: An Investigation of Fishbein's Behavioral Intention Model," in Scott Ward and Peter Wright (Eds.), Advances in Consumer Research, Volume I, Association for Consumer Research, 1973.

A. W. Wicker, "Attitudes versus Action: The Relationship of Verbal and Overt Behavioral Responses to Attitude Objects," The Journal of Social Issues, 15(Autumn 1969),41-78.

A. W. Wicker, "An Examination of the 'Other Variables' Explanation of Attitude-Behavior Inconsistency," Journal of Personality and Social Psychology, 19(July 1971),18-30.

Y. Wind, "A New Procedure for Concept Evaluation," Journal of Marketing, 37(October 1973),2-11.

R. F. Yalch, The Prediction of Behavior: A Field Study of Self-Perception Theory, unpublished doctoral dissertation, Northwestern University, 1974.



Richard F. Yalch, University of Washington
Carol A. Scott, The Ohio State University


NA - Advances in Consumer Research Volume 04 | 1977

Share Proceeding

Featured papers

See More


The Effect of Bariatric Surgery on Delay Discounting for Food and Money: A Longitudinal Study

Ratnalekha Venkata Naga Viswanadham, INSEAD, France
Hilke Plassmann, INSEAD, France
Yann Cornil, University of British Columbia, Canada
Pierre Chandon, INSEAD, France

Read More


Exploring the Intersection of Digital Virtual Consumption and Family Rituals

Linda Tuncay Zayer, Loyola University Chicago, USA
Jenna Drenten, Loyola University Chicago, USA

Read More


Powerful Buy Time: Why Social Power Leads to Prioritizing Time over Money

Myungjin Chung, University of Texas at Arlington
Ritesh Saini, University of Texas at Arlington

Read More

Engage with Us

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