Effects of Information on Product-Related Perceptions

Michael B. Mazis, The American University
ABSTRACT - The four studies examined use laboratory methods to investigate relationships between information and consumer behavior. As a group, they show significant concern for internal validity issues and they are strongly tied to psychological theory. There is a need, however, to devote greater attention to properly specifying variables, using multiple measures and employing more sophisticated data analysis. Overall, this research makes an important contribution to the field.
[ to cite ]:
Michael B. Mazis (1980) ,"Effects of Information on Product-Related Perceptions", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 538-540.

Advances in Consumer Research Volume 7, 1980     Pages 538-540


Michael B. Mazis, The American University


The four studies examined use laboratory methods to investigate relationships between information and consumer behavior. As a group, they show significant concern for internal validity issues and they are strongly tied to psychological theory. There is a need, however, to devote greater attention to properly specifying variables, using multiple measures and employing more sophisticated data analysis. Overall, this research makes an important contribution to the field.


The papers presented in this session reflect the growing sophistication of consumer research since the first Association for Consumer Research Conference held a decade ago. In all four studies, research hypotheses are tied directly to psychological theory and there is significant attention to research design issues.

Several comments about the investigation should be made before examining each study individually. First, they are stimulus-response studies, which focus on manipulating information input and on observing consumer perceptions or behavior. AS such, these efforts can be criticized for devoting insufficient attention to the processing activities which contribute to responses. Greater focus on internal processes is needed to provide insight into the reasons for the observed behavior.

Second, the investigations use laboratory methods to test theoretical propositions. As might be expected from controlled studies, there is considerable concern for internal validity. On the other hand, there should be greater attention to using more than one method of representing a variable and to using multiple dependent measures. Additional emphasis on convergent validity is needed.

Third, the studies rely on theory to develop predictions. Although this emphasis is laudable, frequently theoretical constructs are improperly translated into research variables. Additional pretesting, manipulation checks and multiple operations can be used to correct this shortcoming.

Fourth, while data analysis used in the reported research is generally appropriate, more sophisticated procedures may be warranted in some cases to isolate effects. In addition to using more powerful analytic procedures, the "sin" of selectively examining statistically significant findings, while non-significant findings, must be avoided.


The paper by John C. Mowen investigates an important issue--consumer reaction to a company making a product recall. Given the massive number of recalls faced by industry in recent years, au understanding of responses to this form of negative marketplace information is needed. Professor Mowen shows great concern for relating his research to psychological theory by generating predictions from Heider's balance theory. While this theory is over twenty years old, it has not heretofore been applied to the recall problem. The author's use of balance theory is quite helpful since a framework is provided for hypothesis generation. Overall, the theoretical section of the paper is its greatest contribution.

The paper develops several hypotheses based on attribution theory. For example, Correspondence Inference Theory suggests that if consumers believe a company recalled a product of its own volition, the firm will be viewed more favorably than if outside pressures, such as a government agency, forced a product recall. Another attribution theory prediction is that if other manufacturers were found to have similar defects, consumers may perceive the corporation as less responsible for the defect. This issue is referred to "consensus information on judgments of causality."

While the author's use of these theoretical propositions is quite appropriate given the problem, one might question the usefulness of a set of theories which makes such obvious predictions. In fact, when these propositions are unsupported by empirical evidence (as in the current study), a reviewer can only question whether the investigator has properly translated the theoretical constructs into appropriate manipulations. One is forced to reason that the failure to reject the null hypotheses must be due to some oversight by the researcher.

There appear to be some problems with the manner in which the variables are specified. The manipulation of the "Correspondent Inference" variable is based on whether the Consumer Product Safety Commission (CPSC) had or had not determined whether the recall is warranted. As specified in the study, this variable refers more to the laxness of CPSC than to the degree of outside pressure on the firm. Given that 373 people had received burns from defective coffee percolators, the threat of law suits may have prompted a recall regardless of CPSC's efficiency in carrying out its law enforcement responsibilities.

Similarly, the failure of source variable (Corning or XYZ Corporation) to generate predicted effects could have been due to other information provided to subjects. The firm was identified as an "upstate New York company that has built a billion dollar business out of glass." Subjects could infer that the XYZ company was Corning or some other very large firm thereby "swamping" the source credibility manipulation. The use of manipulation checks and pretesting would have helped insure that subjects "received" the manipulation.

While the researcher devotes considerable attention to selecting independent variables, the dependent measures are not tied to theory. There is no explanation as to the link between the independent and dependent variables. In the absence of strong concern for any single criterion measure, multivariable ANOVA rather than multiple univariate ANOVA's is the appropriate analytical techniques. A multivariate approach would help reduce the experimentors' alpha error.

In the current design, there are three independent variables and six dependent measures. Therefore, 18 separate F-tests are conducted. These multiple operations result in the expectation that (with alpha = .05) one statistically significant findings will be obtained by chance.

The reported findings indicate only two significant effects for different independent variables. As a result, we cannot have much confidence that any of the main effects had an important impact on consumer perceptions.

In summary, the study investigates an important problem. There is a reliance on psychological theory to generate predictions. However, the design could be improved by more careful specification of variables, greater concern for dependent variable selection and improved analytical procedures.


While the preceding study concentrates on the manipulation of consumer information to observe changes in consumer response, the Walton and Berkowitz paper emphasizes response structure and devotes less attention to stimulus factors. The strength of the study is its attention to response variables, which are frequently ignored by consumer researchers. In addition, the researchers use the appropriate analytical technique--factor analysis--to investigate response structure.

The primary contribution of this research is the finding of differential structures for convenience and shopping goods, a subject which should be explored through additional research. Future research should, however, develop and test hypotheses based on psychological theory. While the current research is useful for hypothesis generation, greater attention to theory-based predictions is suggested for subsequent efforts.

In addition to generating theoretical predictions of structure, there is a need for using theory to select the variables included in the analysis. The authors' selection of response dimensions is based on a collection of measures used in previous price-related studies. No empirical or theoretical rationale is provided for variable selection. Theoretical constructs should be stated and multiple measures should be used to represent each construct.

The use of only one product (either a headache remedy or camera) to represent convenience and shopping goods presents external validity problems. Can the results be generalized beyond this study?

There is reason to believe that results for the convenience good (Excedrin) advertised at particular price discount may not be replicated for gum or adhesive bandages. Similarly, the camera advertised (Kodak) may have produced findings unique to that brand. The high credibility and pleasing commercials may create a product image which is not generalized to other shopping goods. Future studies might consider using fictitious brands or matching known brands across important characteristics. In addition, price discounts should be varied since the price discount presented might affect consumer perceptions.

Finally, while the use of factor analysis is useful for exploratory purposes, concern for the stability of factor structures is needed. Cross-validation procedures could be implemented by conducting factor analysis for two separate sub-samples.

Overall, the study investigates an interesting problem and generates results which should be followed-up in future research. There is a need, however, for development of a theoretical framework and testing hypothesis for several brands in each product category.


The paper by Weinberger and Dillon investigates a similar problem--negative information--as the study conducted by John C. Mowen. Both papers are laboratory studies and show considerable attention to controlling possible sources of bias.

The Weinberger and Dillon study is exemplary in its concern for both internal and external validity. Adjectives are pretested so that the polarity of ratings is equivalent. Four products and four services are used so that a form of multitrait operations is employed. A between-measures design is used on the rating type variable so that subjects are given only favorable or unformable information. In addition, subjects are given information from only one source, either neutral, consumer-dominated, or marketer-dominated. The use of a between-measures design is useful since it does not artificially heighten differences between treatment levels.

This carefully conducted study has only one potential flaw--specification of the rating source variable. This difficulty may have accounted for the failure of the research to provide more definitive results. While multiple measures were used for the product/services variable only one exemplar for market-dominated, consumer-dominated and neutral rating source is used.

The possible problem of variable specification is apparent when the theoretical construct is compared with the variable descriptions provided to subjects. The consumer-dominated variable is defined as "friends, acquaintances, peers and family." Product ratings supplied by "local housewives" is used to operationalize this variable. Perhaps local housewives does not connote the degree of familiarity as envisioned by the theoretical construct.

In addition, the marketer-dominated variable is specified by referring to ratings supplied by a "local trade or professional group." This description does not comport with the usual understanding of a marketer-dominated information source. Pretesting, manipulation checks and multiple measures can be used to overcome this potential deficiency.

Another fundamental problem is the choice of variables for inclusion in the study. Rather than selecting products or services and rating sources for manipulation, investigation of the underlying constructs is needed. In this case, riskiness of the decision and source credibility rather than products/services and marketer-dominated/consumer-dominated/neutral sources appear to be the fundamental dimensions. Manipulation of these variables might have produced more satisfactory results.

In sum, this study is outstanding in many respects. It shows great attention to controlling for possible sources of bias. Insufficient attention is devoted to the underlying variables which are tied to differential responses resulting from positive and negative information. Overall, this study represents a fine contribution to the field.


The study conducted by Wheatley, Walch and Chiu is also a stimulus-response study since it manipulates information and money given to consumers and observes the choice behavior of consumers without investigating central processing. This is a fascinating study because it is one of the few consumer behavior studies which introduces a monetary constraint. By simulating market conditions, results are likely to closely monitor actual consumer behavior.

Overall, this research shows great concern for internal validity. The design is carefully executed and it represents an important addition to the price-quality literature. There are a number of ways, however, in which the study and subsequent presentation might be improved.

The authors construct an economic "strawman," which they proceed to dismember. Fortunately, economists, starting with Stigler in the early 1960's, have abandoned the idea that the consumer has perfect information. Modern economists, led by Rothchild, Stiglitz and Salop, assume that consumers make cost-benefit calculations as to whether to acquire additional information and believe consumers use cues to infer certain information about products, although these signals may be inaccurate. Therefore, the authors' contention that economists assume consumers "are knowledgeable about the products they buy" is no longer accepted in the economic literature.

In addition, the statistical tests used may have insufficient power to detect significant effects. Rather than t-tests, dummy variable multiple regression should be used. Since each individual's initial belief as to whether there is a quality difference among cola brands may greatly affect subsequent responses to the experimental variables, this measure should be used as a co-variate in the regression analysis.

The most controversial issues in the experiment involve providing consumers an opportunity to taste the cola brands and the low risk nature of the product selection task. Since consumers were given two cans of the same cola although they were led to believe the colas presented were different brands, their subsequent choice may have been effected. Consumers who can discriminate among cola brands, and are likely to prefer national brands, may have become intrigued by the disconfirmation of their expectations. This fact might have prompted a switch to low-priced cola as a result to the taste experience. While this phenomena is equally likely to occur across the three cells of the design, it could have reduced the power of the statistical tests to detect significant treatment effects. The taste testing aspect is not needed to investigate the propositions under study.

In addition, the low-risk nature of the product class may have encouraged consumers to select equal quantities of the two cola brands to continue the taste tests in their own homes. The investigators indicate that a substantial number of consumers selected equal proportions of the two colas so that they could "test the differences between the two samples by themselves." This form of "game playing" by subjects may have "swamped" the experimental treatments.

In summary, this is an intriguing experiment which should be replicated. Subsequent studies might consider using other information, higher risk products and larger monetary constraints.