Product Perceptions, Preferences, and Knowledge: Decisions in Conducting Research

George Brooker, Central Washington University
[ to cite ]:
George Brooker (1993) ,"Product Perceptions, Preferences, and Knowledge: Decisions in Conducting Research", in NA - Advances in Consumer Research Volume 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT : Association for Consumer Research, Pages: 229-231.

Advances in Consumer Research Volume 20, 1993      Pages 229-231


George Brooker, Central Washington University

Researchers preparing to test cause-effect relationships are confronted by an extensive number of choices and decisions. Successful completion of the research and drawing appropriate conclusions about relationships will depend on these decisions. While some choices are obvious, others are somewhat less so. The implications of many of these decisions are not always apparent; it is only after the fact that their importance can be seen.


One of the decisions researchers face is whether to pursue a project at all. Simon (1969, pp. 223-226) provides several useful criteria to assess the importance of research, including: Will the study test an important theory? Are results apt to be surprising or unexpected? Does the work address contemporary arguments and open questions, or does it extend or contradict previous work?

Smith's paper provides some interesting conceptual notions on how new and later product introductions may be affected by information-processing issues. It is especially intriguing to consider in the innovation-adoption model perspective. Gatignon and Robertson (1985) present a set of propositions which range broadly across the literature on diffusion, covering most recognized topics. Smith's probing is more constrained, limited to new product entry and its immediate and subsequent effects on processing and differentiation. From the diffusion perspective, her work fits into the narrower confines of the adoption process and, implicitly, innovator characteristics.

While most of her propositions have a strong element of "common sense" and logic to them, there is some limited controversy here. Maheswaran and Sternthal (1990) suggest it is possible to motivate people to process information in detail; Smith suggests product category knowledge will dictate likelihood of processing information. These are not conflicting notions, but it will be interesting to see what the limits are for each.

It would be useful to test Smith's notions in the diffusion model; such tests could add depth to a literature which, thirty-plus years after its inception, still lacks detail in many areas. With Smith's propositions incorporating elements such as type of processing, product knowledge, and motivation, opportunity, and ability to process information from competitors, it seems they have the potential to add refinement to the innovation diffusion mural.

Veryzer gives us an interesting study looking at some elements long-neglected and only beginning to be explored by consumer researchers. The neglect is intriguing, since many products have their aesthetics consciously planned (e.g., cars, furniture, tableware, etc.), and aesthetics even may be the basis for purchase in some (e.g., telephones, silver service, paintings and other art, clothing [Holbrook 1986] and accessories, etc.). The area is one where a number of significant contributions can be made. While, in most cases, aesthetics is not the basis for the advertising appeal which is made (manufacturers seldom refer to their "attractive appliance"), it certainly could be the basis for differentiating brands in a crowded marketplace. It would be useful to explore the relationship between aesthetics and choice conceptually and empirically to define the importance of the topic.

Veryzer's hypothesis on interactive effects is an interesting application of the Gestalt principle. We should be skeptical of the test results on the natural sound machine which, apparently, had problems in the visual representation. A distortion problem in an aesthetics study seems substantial, indeed. With an improved treatment presentation, this is a notion worth reexamination.

Somasundaram's present work leaves several things unclear. As it stands, it is a simple manipulation with two covariates. The exposition left me wondering about the adequacy of some of the measures, of the dependent variable, even of the internal validity of the study. It is not clear that this experiment has tested any of the extant attribution theory models. The results of such a study are of limited value. However, the importance of this paper lies in what it may generate in the future rather than what it is here. The author intends to do a study examining the impact of involvement and product knowledge on causal attributions. Such a study would fit in the attributional bias literature (Folkes 1988).


When formulating research questions, hypotheses, or propositions, researchers must frame them in a way which will render them amenable to testing. There are at least three factors which must be considered: clear specification of relationships, maintaining a consistent perspective on concepts, and recognizing the need for operational definitions.

Unless relationships are specified clearly, there may be uncertainty regarding appropriate testing methods. If concepts are viewed inconsistently, there will be confusion regarding which perspective should be adopted in testing; ultimately, this could confound the operational definitions and operationalizations which are developed.

For example, one of Smith's major constructs is the level of processing used by potential adopters. She suggests the degree of match/mismatch or differentiation between a brand and its product category will influence level of processing, yet some of her propositions suggest level of processing is person-related. This should be clarified.

Veryzer provides a substantial amount of description of his concepts, leading to clearly understandable operationalizations. The relationships are specified clearly, and there is consistency in the use of terminology. The end result is a clear ability to test the hypotheses he presents.

Somasundaram's descriptions also meet the criteria for testability. However, some of his operationalizations need additional thought for a clear test of his hypotheses.


In general, care and thought must be given to the research design as they are to other aspects of research. In addition to the overall design, time can be well spent reflecting about choice of manipulations, representations of concepts, choosing dependent measures, and minimizing potential for demand effects.

Veryzer's research uses a within-subjects design. He tells us few subjects appear to have a conscious awareness of design principles. Conscious awareness may not be necessary to bias a result; Janiszewzki (1991) found evidence of subconscious processing which influenced evaluations of brand names and logos. The added information available to subjects in a within-subjects design certainly has the potential to exert unusual influences. A replication using a between-subjects design would have the potential to provide a "cleaner" result.

Somasundaram's original design here was a pilot study. His presentation of the projected work raises interesting issues on some choices on methodology.

An experimenter must make decisions on the nature of the independent variable. The success-failure manipulation here seems well-defined. A decision must be made on how great the deviation of the "failure" condition must be relative to the "expected" condition to be judged a "failure." Obviously, the deviation must be substantial enough to have the intended effect, but too great a deviation may cause subjects to consider the exercise as obvious, and to treat it trivially. Pretesting of several different "failure" renderings would be desirable.

Related to this notion is the choice of photographs to study the phenomenon. I wonder if the product is important enough to raise questions about failure causes. As Folkes (1984, p. 407) states, "If the product is trivial, consumers may neglect to ask themselves why the failure has occurred." In this case, the subjects did not take the pictures, so they have no investment in them. Beyond this, poor photographs are not a tragedy. A photograph often may be retaken, and processing can be redone. The question then arises whether it is worth the effort to establish causal attributions.

The use of a comparison picture to establish an expectation raises questions about the potential for demand effects. If the "baseline" picture is seen as possible, then the "failure" picture may be perceived as an unnecessary outcome and questions may arise regarding the intent of the experiment. Since Somasundaram is using two buildings in close proximity (which may be identified in captions), perhaps a "baseline" photo is unnecessary. Then, only the "failure" photo need be presented to those in the "failure" condition, with those in the "ambiguous" condition receiving one flawed picture and one perfect picture, and those in the "success" condition receiving two perfect pictures. Again, pretesting would determine the feasibility (and need) for this approach.

Choosing buildings to manipulate situation-specific involvement assumes involvement with the architecture of the buildings. It would help to have a measure to confirm involvement differences. At minimum, manipulation checks are needed; but these are post hoc. Some preliminary testing might help determine, a priori, that the manipulations will have the intended effects. This does not obviate the need to test involvement levels; it only means the manipulations may work.

I have had difficulty determining what the new research design will look like. Somasundaram says 1/3 of the subjects will be put in each of the three conditions. However, each condition seems to have a different number of cells, or treatments. There is no need to balance the number of people in each of the three major conditions. In fact, this will lead to an unnecessarily large number of subjects being used.

It appears that the Success condition is a control; Failure will have four treatments (assuming within-treatment consistency of failures) representing two sources of failure (photographer and processing) and two types of controllability (controllable, uncontrollable). Within-treatment consistency of failure will reinforce the attributional ascriptions; inconsistency may confound them. The Ambiguous condition will have eight treatments (two "success" results X two failure sources X two controllability types). Thus, thirteen cells are needed. The numbers needed per cell can be estimated in pretesting.


Research can have a sterile quality to it. Often, it arises from necessity. As the researcher "purifies" the design in focusing on the effects of the independent variable, and attaining a high level of internal validity, experiments can become distant from "real world" representations. In the early stages of a research program, this may be desirable to see if the research model behaves as expected. However, as the research progresses, it is desirable to see if the findings are robust. In addition, adding mundane reality to the manipulations may change perceptions.

Veryzer's study was designed to look only at two elements involved in aesthetics design - proportion and unity. In that sense, the design was quite "clean" and focused. However, products are not assessed in a vacuum; there will be other elements in the viewing field, and the buyer's schema may include, e.g., a setting in which the product will be used. The focus in this study is on the physical image of the object, in isolation from other stimuli. A manipulation including more mundane reality might change or influence aesthetic perceptions. Kleine and Kernan (1991) present evidence that context of presentation influences perceptions, affecting the meanings ascribed to ordinary objects. Certainly, future research should examine the changes which may be introduced to aesthetic response by context effects.


The dependent variable should have soundly reasoned conceptual ties to the theory or research model being tested. There may be many different representations of the dependent variable which would be appropriate; it may be closely aligned with the manipulation, or it may be more removed (as a second-order effect); it may be behavioral or verbal. However, there is one desirable characteristic of all dependent variables: sensitivity appropriate to the strength of the treatment.

Somasundaram's proposed dependent measures are verbal, based on the Weiner (1986) model of attributional causes: locus, stability, and controllability. Locus and controllability will be manipulated; all three causes will be examined. I wonder if the intended dependent measures (thought listings) will be sensitive enough to show differences. Thought listings assume substantial commitment on the part of the subjects. To the extent that they do commit to the task, such an approach will be useful. However, the use of subclassifications of the thought listings may miss potentially useful experimental effects because the dependent variable fails to reveal them. The approach taken by Folkes (1984), using semantic differential scales, might prove more revealing. Certainly, Somasundaram could use his pilot test to develop sets of possible causes to be evaluated and put them in scalar form. Such an approach assumes less of the subject, with the potential to increase power to test the manipulations.


The need to plan the analysis before conducting an experiment is well known. The analysis must be appropriate for the design.

Veryzer's use of simple ANOVA leaves us in limbo regarding results. LaTour and Miniard (1983) describe the problems arising from use of between-subjects analysis in within-subjects designs. The end result of such use can be an underestimation or an overestimation of the alpha rate, with no way to evaluate the accuracy of the inferences. A more appropriate approach would be to use a repeated measures ANOVA or a repeated measures MANOVA. This is a correction easily made.


Folkes, Valerie S. (1984), "Consumer Reactions to Product Failure: An Attributional Approach," Journal of Consumer Research, 10, 398-409.

Folkes, Valerie S. (1988), "Recent Attribution Research in Consumer Behavior: A Review and New Directions," Journal of Consumer Research, 14, 548-565.

Gatignon, Hubert, and Thomas S. Robertson (1985), "A Propositional Inventory for New Diffusion Research," Journal of Consumer Research, 11, 849-867.

Holbrook, Morris B. (1986), "Aims, Concepts, and Methods for the Representation of Individual Differences in Esthetic Responses to Design Features," Journal of Consumer Research, 11, 337-347.

Janiszewski, Chris (1990), "The Influence of Print Advertisement Organization on Affect Toward a Brand Name," Journal of Consumer Research, 17, 53-65.

Kleine III, Robert E., and Jerome B. Kernan (1991), "Contextual Influences on the Meanings Ascribed to Ordinary Objects," Journal of Consumer Research, 18, 311-324.

LaTour, Stephen A., and Paul W. Miniard (1983), "The Misuse of Repeated Measures Analysis in Marketing Research," Journal of Marketing Research, XX, 45-57.

Maheswaran, Durairaj, and Brian Sternthal (1990), "The Effects of Knowledge, Motivation, and Type of Message on Ad Processing and Product Judgments," Journal of Consumer Research, 17, 66-73.

Simon, Julian L. (1969), Basic Research Methods in Social Science: The Art of Empirical Investigation, New York: Random House.

Smith, Karen H. (this volume), "The Moderating Influence of Depth of Processing on Order of Entry Processing Effects."

Somasundaram, T. N. (this volume), "Consumers' Reaction to Product Failure: Impact of Product Involvement and Knowledge."

Veryzer, Robert W. (this volume), "Aesthetic Response and the Influence of Design Principles on Product Preferences."

Weiner, Bernard (1986), An Attributional Theory of Motivation and Emotion, New York: Springer-Verlag.