Presidential Address the Value of Theory in Consumer Research

Alice M. Tybout, Northwestern University
[ to cite ]:
Alice M. Tybout (1995) ,"Presidential Address the Value of Theory in Consumer Research", in NA - Advances in Consumer Research Volume 22, eds. Frank R. Kardes and Mita Sujan, Provo, UT : Association for Consumer Research, Pages: 1-8.

Advances in Consumer Research Volume 22, 1995      Pages 1-8

PRESIDENTIAL ADDRESS

THE VALUE OF THEORY IN CONSUMER RESEARCH

Alice M. Tybout, Northwestern University

I owe a great debt to my colleagues and doctoral students (present and past) in the marketing department at Northwestern University for providing a supportive and intellectually stimulating environment. Special thanks are given to Bobby J. Calder and Brian Sternthal, who shared in the development of many ideas expressed in this address and who provided detailed feedback on earlier versions of this paper.

INTRODUCTION

Presidential addresses to this body have traditionally called for setting our sights higher, achieving more, exhibiting greater creativity and, in general, having a greater real world impact. These are appropriate topics for a forum such as this one and, during this conference, you will no doubt hear a variety of opinions about how these goals might be best achieved, for one constant in ACR is a diversity of points of view. My remarks will focus on the value of theory in advancing understanding of consumer behavior.

Many factors have figured in the development of ACR. One factor was the interest in theory that emerged in the wake of scathing critiques of business school curricula that were published in 1959. Funded by the Ford Foundation and the Carnegie Foundation, these critiques instructed business schools to "stop teaching descriptive material and start emphasizing theory and research" (Kernan 1994, p.2; also see Shimp 1993). In response, behavioral scientists interested in understanding the processes underlying human behavior began to focus their attention on consumer behavior. ACR was created as a forum for these academicians to interact with each other and with practitioners and public policy makers in their quest to better understand consumers.

Twenty-five years later seems to be an appropriate time to ask, where has the quest for theory led? Some observers contend that progress has been negligible. They echo sentiments expressed earlier in ACR's history when when such critics charge that we have borrowed theories without enriching them with insight unique to the consumer behavior context in which we operate, that our theories are impoverished and microscopic in their scope, and that the theories we do have are ill-tested and, as a result, are of little practical value. Still other observers have gone so far as to question whether theoretical, scientific knowing is distinct from other types of knowledge.

Before we heed any call to abandon the pursuit of scientific theory and to embrace some other approach, we need to scrutinize the source of our dissatisfaction. Much of this dissatisfaction is due to the confounding of the logic or philosophy of science with the sociology and psychology of science; we fail to distinguish between the progress that science has the potential to make and the circuitous route by which science often proceeds because of the behaviors of scientists. Although progress has not been linear, and consumer research can be accurately characterized as focusing more narrowly than might be desirable, there is evidence of impressive theoretical progress. Moreover, if progress is to remain our goal, it is an illusion to think that there is a viable alternative to the scientific pursuit of theory. What are sometimes presented as alternatives are, in reality, simply weaker versions of the traditional scientific approach. Greater progress in the next 25 years requires not abandoning theory, but rather requires a better understanding how to judge and apply theory.

As a starting point in developing this thesis, consider the following five true-false items about theory and procedures for testing theory.

T or F1) Progress can be increased by abandoning theory or by pursuing a different level of theory than we have sought in the past.

T or F2) It is possible to render a unique explanation for a simple (two-level) main effect.

T or F3) Certain measures and data collection procedures must be followed when testing a theory if theoretical progress is to be made.

T or F4) Predictions constitute more rigorous tests of theory than postdictions.

T or F5) A well-designed study will provide theoretical insight and will allow generalization of the effects observed to a real world situation of interest.

I'll return these items later in my remarks. For the moment, let me indicate that there is substantial disagreement about whether the correct response to these statements is true or false. To resolve this disagreement, let me begin by discussing theory and theoretical progress in consumer research.

THEORY AND THEORETICAL PROGRESS IN CONSUMER RESEARCH

Confusion about theory is reflected in the paradoxical uses of the term. Theory refers to specific, scientific principles that are used to explain a set of real-world phenomena. Theory also refers to guesses or conjectures that may be in opposition to reality (as in the case of statements such as, "well that's true in theory, but not in practice"). It is no wonder that misunderstandings of theory abound!

For the purpose of today's remarks, I will use the term theory with reference to a hypothesized set of relationships or a nomological network linking abstractions known as constructs. The constructs and the relationships between them cannot be observed. Instead, the adequacy of a theory is assessed in relation to observations about real-world phenomena, with preference given to the theory that can account for phenomena most parsimoniously. Thus, the goal of theory is ultimately a highly practical one; to enable us to explain and to predict the world in which we live.

Although a theory may gain acceptance, it always remains a "work in process," because new observations have the possibility of contradicting a preferred theory and prompting its revision or replacement. It is in this manner that scientific knowledge is seen as offering the possibility of progress.

While progress is a possibility, it has recently been argued the sociology of science makes this possibility remote at best. The contention is that, in reality:

Ego-involved Theorists persevere indefinitely in the face of Theory-disconfirming results. Theory 'tests' are so imperfect that they can always be written off. When consumer behavior Theories are 'tested,' they do not get better or even change. And, Theory-oriented consumer researchers neither replicate their findings nor systematically investigate the range and limits of their work. (Wells 1993, p. 497).

FIGURE 1

To assess the validity of this analysis, let us turn to the specific body of theory that has been the focus of much attention within ACR. A substantial portion of consumer research has focused upon how individuals acquire and use information in making consumption decisions. Has theorizing on this topic evolved in the past 25 years?

In 1970, Berlyne proposed a two-factor theory that predicted a nonmonotonic relationship (inverted U) between stimulus familiarity and liking. This outcome is predicated on the assumption that early exposures to a stimulus lead to positive habituation due to a reduction in uncertainty, whereas later exposures result in tedium due to satiation.

In 1979, Cacioppo and Petty offered a more detailed view of the process underlying the nonmonotonic relationship proposed by Berlyne. They reasoned that exposures affect the time to think about the stimulus and, thus, the content of thoughts that people generate. This initial Elaboration Likelihood Model offered a robust explanation, accounting for many findings in the psychology literature and in the consumer behavior literature.

In the past five years, the Elaboration Likelihood Model has been extended and modified by consumer researchers. While the elaboration construct offers an explanation of persuasive message processing and attitudinal effects, it does not enable anticipation of when elaboration will enhance persuasion and when it will undermine it. To address this limitation the notion of resources as the engine that fuels elaboration was introduced. The idea is that persuasion is maximized when the resources required for message processing are matched by those available: too few resources inhibit message processing and too many prompt idiosyncratic thinkingCin both instances reducing the processing of the message and thus its persuasive impact (Anand and Strenthal 1990). The theory that elaboration results in the greatest persuasion when resource requirements and demands are matched provides a cogent account for a wide range of repetition effects reported in the literature.

But progress did not stop here. There is a recurrent finding of a null effect of repetition when complex stimuli composed of ads and programming or editorial material are used. These outcomes have been construed by some investigators as evidence that theory fails to offer meaningful application. However, there is emerging evidence that we can explain the moderating role of contextual factors on repetition effects by viewing elaboration not as a single notion but as being of two distinct types. One type of elaboration bolsters the retrieval of the target object and its category membership and, a second type of elaboration helps distinguish the target object from its alternatives.

Thus, in the past 15 years we have extended the initial idea that persuasion is based on elaboration first by introducing antecedent resource conditions to anticipate the impact of elaboration and, more recently, by distinguishing two types of elaboration that can account for the effects of context on the persuasiveness of a message. This is impressive theoretical progress.

And just as important, from a practical perspective, we have progressed from being able to account for the effect of the repetition of a single ad to being able to explain the impact of repetition and a host of others variables that represent the same construct in a complex settings that reflect everyday situations! Consider as an example the industry practice of theater testing to assess the effectiveness of alternative advertising executions. The typical theater test exposes people to three repetitions of alternative ads and compares the effectiveness of these ads in terms of recall and brand preference. The assumption is that is that the ad that results in greater recall and brand preference in the theater test also will be the ad that is the more effective under natural viewing conditions, even though natural conditions may involve higher levels of repetition. However, current theorizing about information processing suggests that such an assumption is only justified if the two ads are similar in their comprehension difficulty. If, instead, one ad is more complex than the other, the simplier ad may dominate at moderate levels of repetition and this pattern may be reversed when repetition is higher, as illustrated in the graph (Figure 1). Thus, theory informs practitioners about factors that must be taken into account in constructing and interpreting research intended for specific application.

FIGURE 2

LEVELS OF ABSTRACT - ION IN RESEARCH

In addition to enhancing understanding of the relationship between repetition and persuasion, theorizing about elaboration and cognitive resources also can explain phenomena such as when and why vivid messages have their impact, when the use of color in ads is likely to have greater impact than black and white ads and when the reverse is true (Meyers-Levy and Peracchio, in process), and when self-reference is an effective message device and when it is not (Burnkrant and Unnava, in press). And, the theory has been enriched by these efforts to expand the set of phenomena for which it accounts.

Of course, these advances in understanding consumer information processing have not been the result of one researcher or one study. Rather, such progress reflects the effectiveness of the larger, scientific process.

While you might agree that the stream of work just described does indeed represent significant progress, it is still reasonable to ask why we haven't made more progress both within the specific realm consumer information processing and within the more general domain of understanding consumers' behavior and consumption experiences. My view is that the preoccupation with variables and procedures in judging theoretical explanations of phenomena has been a major impediment to our progress.

ASSESSING THEORY

To elaborate on this view, it is necessary to distinguish three levels of abstraction that may be employed in interpreting research. As depicted in the diagram (Figure 2), the first, most concrete level reflects an interest in the particular outcome(s) observed. The theater testing of ads described earlier represents such a situation. The ad sponsor is interested in determining which ad is more effective so that the more effective ad can be aired. Although some implicit (or even explicit) theoretical notions may have contributed to the design and to the testing of the alternative ads, it is the outcome observed in the theater test, and not any theoretical notion, that the researcher seeks to generalize.

This type of ground-level effects generalization is typical in many business settings. Firms and managers adopt procedures and techniques that have been shown to be effective in specific case examples often with little scrutiny of the underlying assumptions and the appropriateness of these assumptions to the situation of application. Thus, firms have embraced downsizing, out-sourcing, TQM, benchmarking, re-engineering, and delighting customers (though not necessarily in that order! see Iacobucci, Grayson, and Ostrom 1994).

An alternative, more abstract interpretation of research occurs at the level of variable relationships. Here, the goal is to identify those variable relationships that are robust and to confirm a particular interpretation of such relationships.

Historian Frank Sulloway's research on the role of birth order in historical events has been featured in the general media recently and can serve as an illustration of this approach (Stiff 1994). Sulloway's thesis is that birth order is a critical determinant of one's willingness to embrace radical ideas. The priviledged status of first borns is argued to result in their being conservative and favoring the status quo. By contrast, latter borns, who are less advantaged, are reasoned to be more open-minded, innovative, and rebellious than first borns. In an effort to confirm this hypothesis, Sulloway analyzed a database profiling 6,000 influential historical figures, including Voltaire, Darwin, Lenin, and Einstein. His findings indicate that birth order is superior to other variables, such as social class, in accounting for how people respond to radical new ideas. This confirmation is viewed as especially powerful because it is based on so many observations.

Many studies in consumer research also reflect an interest in identifying and interpreting general variable relationships. For example, research might be conducted to examine the relationship between the presence or absence of pictures and persuasion. The notion might be that pictures reinforce an auditory message, enhancing learning and, thereby, increasing persuasion. This hypothesis could be confirmed by a study demonstrating that an ad with a picture resulted in more persuasion than an ad without a picture. Support for the hypothesis would be bolstered, however, if a manipulation check revealed that the picture was perceived to reinforce the message, if recall measures showed improved performance when a picture was present, and if cognitive responses indicated more positive thoughts when a picture was present. These additional opportunities to confirm would be viewed as making the interpretation rendered more convincing.

Interest in variable relationships is understandable; variables are the currency of theory. However, efforts to equate a variable with particular construct are inappropriate. If a variable were always related to a particular construct, there would be no need for constructs. A single variable may operationalize a variety of different constructs or different levels of a single construct as a function of the people, setting, and time being examined. For example, a picture may serve to repeat or reinforce a message, but a picture may also usurp resources and prompt idiosyncratic thoughts or emotions.

To illustrate the complexity of the variable-construct relationship, let us return to Sulloway's birth order research. Birth order is an intriguing variable because it has a substantial impact on one's life experience. However this impact implies that birth order captures many concepts; not merely one's investment in the status quo. Indeed, Sulloway acknowledges this fact in offering an explanation for the finding that first borns, who are argued to be conservative, dominated the French Revolution. He attributes this outcome to a second construct, first borns' tendency to be "tough-minded," and suggests that tough-mindedness is a factor when "extended revolutions lead to terror." Thus, the challenge is to identify the most appropriate interpretation of variable relationships. This requires moving to the level of general theory and adopting parsimony, rather than confirmation, as the criterion against which theory tests are judged.

Despite the requirement that explanation be in terms of nomological networks of constructs and not in terms of variable relationships per se, there is much resistance to moving to this, highest level of abstraction. Why? From a sociological perspective, remaining at the variable or near-variable level enables numerous experts or leaders to coexist, albeit each in very narrowly defined areas. This serves a practical purpose of accommodating the tenure review process, which at most major institutions requires that a candidate be a leader or expert in some defined area. Thus, our field includes as experts in variables such as repetition, source credibility, attitude-toward-the-ad, and so on.

From a psychological perspective, focusing on variables and procedures fosters the (comforting) illusion that one can know the value of a study before the data are collected. It also provides an easy means for evaluating others' research; if the interpretation of the variables is similar to that in prior research and if all the manipulation checks and so-called process measures are in place, the study is viewed as lacking any "fatal flaw." A practical problem is that studies examining trival or near-tautological hypotheses, as well as ones in which the outcomes are dictated by the procedures (i.e., studies in which subjects seem to have no degrees of freedom), will yield confirmations of these hypotheses (see Wallach and Wallach 1994). This implies that an additional criterion must also be invoked. A common approach is to assert that the research somehow be "interesting." This too is an easy judgment to render; one simply introspects on whether the research results are reconcilable with, but not identical to, the point of view held prior to reading the research. Notice, however, that such judgments are idiosyncratic because they are tied to one's personal knowledge base. It is impossible for researchers to know what will be viewed as "interesting" by a set of reviewers in advance of receiving their opinions.

In short, focusing on variables and confirmation may serve sociological and psychological needs, but in so doing, we depart from the logic of science and, thereby, abandons the possiblity of progress. Instead of confirming an interpretation or attempting to divine what will be "interesting" to an as yet selected set of reviewers, we should strive to select the best explanation available at a point in time. This implies the criterion parsimony.

A parsimonious interpretation of data can only be achieved through the application of convergence or triangulation procedures. These procedures require multiple variable-level operationalizations of each construct being examined. While the need for triangulation is well-known and widely endorsed, in practice researchers often present construct-level interpretations that rest on a single operationalizations. This cannot be justified.

While multiple variables are required, convergence procedures are indifferent to whether these measures include manipulation checks and so-called process measures. Nor is it an issue whether the experimenter measured or manipulated the independent variables. Convergence can be achieved either by multiple dependent measures or by multiple independent variables.

What matters is whether the pattern of results on whatever measures are available (from new research and evidence previously reported) can be interpreted uniquely at a theoretical level. This can only be determined after the fact and by examining the data, not by examining the procedures and measures per se. (Unless, of course, the hypothesis is a tautology.) If a uniquely parsimonious interpretation is offered, then the research makes a contribution to knowledge and the theory is accepted until such time when a more parsimonious alternative or an equally parsimonious but more encompassing alternative becomes available. (See Sternthal, Tybout, and Calder 1987 for an extended discussion of these issues.)

An interesting and, to some, a counterintuitive by-product of how theoretical progress is made is that prediction holds no advantage over post hoc explanation (See Sternthal et al. 1987; 1994; Brinberg, Lynch, and Sawyer 1992). Indeed, just the opposite may be argued to be the case; post-hoc explanation is to be preferred.

When a prediction is tested, the opportunity to assess whether alternative explanations offer as good accounts for the data as the favored view is limited. When tests are post hoc, rival explanations have had such an opportunity. Brush (1989) illustrates this point when discussing two types of evidence for Einstein's theory of relativity: light bending which was predicted before it was observed, and Mercury's orbit which was explained after it was observed:

rather than light bending providing better evidence because it was predicted before the observation, it actually provides less secure evidence for that very reason....Because the Mercury orbit discrepancy had been known for several decades theorists had already had ample opportunity to explain it...and had failed to do so...Light bending, on the other hand, had not previously been discussed theoretically...but now that the phenomenon was known to exist one might expect that another equally or more satisfactory explanation would be found (p.1126).

Recognizing the irrelevance of when an explanation for data is developed, prominent psychologists such as Daryl Bem (1987, 1991) argue that, when writing a research report, it is perfectly appropriate, and may even be desirable from a communication standpoint, to present as hypotheses notions that were developed after the data were collected. To some this seems heretical; it misleads the reader by implying prediction when the reality is postdiction (Kerr 1994). But this view rests on a false assumption, fostered by a confirmation orientation, that a variable represents a particular construct and, therefore, the failure of a variable to behave as anticipated is newsworthy. If the fact that variables can represent many constructs is acknowledged, then deducing the construct operationalized by the variable after the fact is necessary and certainly does no harm to theoretical progress. Moralistic or other non-scientific concerns must be invoked in any attempt to justify a preference for prediction.

Thus, I believe that one reason that we have not seen more theoretical progress because we have not paid enough attention to what theory requires. Too often we have been focused on variables and procedures per se and not on how these variables and procedures inform us about abstract relationships between constructs. For greater progress we must shift our focus to a more abstract, theoretical level and hold our research to the single criterion of identifying a uniquely parsimonious explanation, at least for the moment. We should strive for theories that are universal in the sense that when we become aware of the limits of our theories this should be taken as an opportunity to modify or replace the theory with one that is again encompassing.

It would be naive to ignore the sociological and psychological pressures that sometimes run counter to the logic of science. In fact, it may even be worthwhile to make responses to such pressures the subject of theoretical examination, for they are nothing more than additional data points to be explained. But recognition of these influences in no way implies that they offer a viable alternative to the logic of science in pursuing the goal of progress. If we are seeking progress, there is no alternative to parsimony in judging our theoretical explanations.

You may have observed that I have made only passing reference to how the phenomena that we attempt to explain are documented. Perhaps your awareness of my own style of research and the language of "theory testing" have created the impression that my remarks are primarily applicable to experimental data. I do not believe this to be the case. The criterion of a uniquely parsimonious interpretation can be applied to explanations of ethnographic data, scanner data, and so on. Indeed, as indicated earlier, I am not aware of any other criterion that could be applied appropriately if our goal is progress in understanding.

THEORY IN USE

Beyond changing the way in which we value and assess theory, our progress will be enhanced by broadening the phenomena we attempt explain, as several previous ACR presidents have argued eloquently. This is because failures are more likely than successes to spur theory revision or the development of new theory. (Though, we cannot know which new phenomena will be especially enlightening in this respect a priori.)

But, in attempting to apply existing theory to specific circumstances that may be of interest, it is important to recognize that the central benefit of theory, its relevance across specific circumstances, implies that theory can only anticipate general patterns of relationships. And, translating these relationships into predictions relevant to a particular set of circumstances will likely require numerous assumptions about which constructs and the level of these constructs that are represented by the variables of interest. Further, if assumptions about how the variables map onto constructs are erroneous, even a good theory will make inaccurate predictions. The establishment of a construct-variable link necessarily must rely on empiricism.

These limitations are at the core of many criticisms regarding the value of theory. What good is theory if all it can offer is imprecise (relational) prediction that rests on assumptions about variable-construct relationships, which could be wrong? Theory doesn't seem to get us very far. Why not go "direct," stay at the level of the variables, and skip the abstraction to theory.

The answer is, of course, that there is no escaping the need for theory. How does the researcher know which aspects of any given situation should be studied and which might reasonably be ignored? The researcher's logic for believing that selected variables might affect purchase behavior constitutes a theoretical explanation of the variable relationships. This theory too often remains implicit and untested when more explicit consideration would seem to be beneficial. As evidence for this point, recall the theater testing of commercials discussed earlier.

Thus, when substantial resources are at stake, theory application is served by additional research that examines the calibration of the theory the situation. In such research, the theory is:

... unfolded or re-expanded; that is, the compressed general statement that constitutes the theory must be supplemented with detailed information about the special case (Gell-Mann 1994, p.77).

While using theory may be seen as requiring too many assumptions, assumptions are unavoidable. Theory forces us to be more explicit about these assumptions and thereby, increases the likelihood that the reasonableness of those assumptions will be considered in light of prior research and with concern for the real-world conditions to which the theory is being applied.

It has been suggested that there is an intermediate ground of theory somewhere between abstract, universal theory and variable-level description (Wells 1993). Indeed, this is probably descriptive of the theories that we actually have. But application is not served by striving for explanation at this level. When limits to a theory are identified, future application is better served by revising the theory in the direction of universality than by delimiting its boundaries more narrowly. For example, if one attempts to explain persuasion in terms of consumers' ability and motivation to engage in various types of cognitive elaboration, one has no basis for saying that the relationship applies to, say, advertisements presented on television but not via direct mail advertisements or to advertisements for nondurables but not to advertisements for durables. Such a view confuses variables with constructs. And if the constructs cannot account for phenomena within the domain of the theory, then the theory is of questionable value on the very grounds that it is not universal (i.e., it cannot explain relevant phenomena).

I believe that our field would benefit from more efforts to develop and test theory-based interventions or applications (see Calder, Phillips, and Tybout 1981). In constructing such interventions, attention would center on the variable-construct mapping, with the construct-to-construct mapping assumed to be accurate. To avoid extreme particularism, such applications would strive to address problems that might emerge in a variety of situations. For example, one might compare alternative, theory-based strategies for combatting various types of rumors or one might explore alternative campaigns for charitable donations.

Separating theory testing and theory application is predicated on the notion that it is difficult to design research that is optimal for both goals. The goal of rigorously assessing theoretical deductions and the goal of generalizing to a particular real world situation typically are served by different choices with respect to the research setting, subjects, and independent and dependent variables (see Calder et al. 1981). Thus, it is necessary to give priority to either theoretical or problem-solving goals in a particular study. Contrary to what we are told or may wish; we can't have it allCat least not all at once.

In this regard, I am reminded of an article that I read several years ago. In it the author discussed the interaction between the various roles that women play; worker, wife, mother, daughter, friend etc. She noted that the roles allow women to express different, complementary aspects of their personality. But the roles can also compete for attention and, thereby impair performance. This point was illustrated by describing a situation in which a woman was physically in the workplace but mentally elsewhere during an important staff meeting. Where was she? She was busy making a mental list of tasks related to her other roles, as wife/mother/daughter/friend and so on. Few among us, regardless of gender, have escaped this experience. It is simply a fact of busy rich lives. But don't we enjoy what we are doing more when we immerse ourselves in it for the moment? And might not this idea be applied to our research? We would do well to focus sometimes on assessing theory and other times on using theory in an effort to solve important problems in the real-world.

Thus, I suggest that the goal of better theory is also served by distinguishing between studies that have as a primary goal theory testing and studies that have as their primary goal theory application. Each has value, each may inform the other, but it difficult to design research that optimizes both. More generally, I borrow the words of Speck:

Science is served in many ways: By intelligent discussion and fresh proposals, by the extension or completion of previously presented theories, by the fair-minded and unflinching evaluation of current proposals, by justly protesting, blowing the whistle, and pointing out that this kingly theory or that is not wearing a shred of evidence, by sometimes synthesizing and sometimes isolating, by daring to be explicit andCironicallyCby daring to be suggestive. It is when scientists and philosophers of science cannot make up their minds as to which role they are playing orCwhat is worseCtry to fill several roles at one, that matters go awry. Then the Ivory Tower and the Tower of Babel sound disturbingly alike. (Paul Surgi Speck as quoted in Hunt 1993, p. 176)

THEORETICAL CREATIVITY

Whether focusing on a theory or on solving a problem, we need to foster creativity. As discussed earlier, too often we are hide-bound by our desire for concreteness and specific procedures; we define the phenomena to be explained and/or what would constitute an adequate explanation far too narrowly. A classic puzzle illustrates this well. If you have never seen this puzzle take a moment to try it.

Draw no more than four straight lines (without lifting the pencil from the paper) which will cross through all nine dots.

FIGURE

Once you know the trick, that there is no reason to be confined by the imaginary boundary around the perimeter of the puzzle, the solution becomes easy (see a, Figure 3) If fact, it is possible to solve the puzzle with only three lines if you also reject the notion that the lines must go through the center of the dots (b). And if you are willing to fold the paper in a complex manner (c) or cut the paper and paste the dots in a row, (d) one line will do quite nicely. Or, easier still, as one ten year old suggested, just get a very fat writing "apparatice (sic)" (Adams 1986, 24-32).

The problem is that too often we fail to think outside of the box when defining the phenomena to be explained or the problem to be addressed. Evidence of this is everywhere. Ph.D. students who collect data for the first time are often stumped when they obtain significant but unanticipated results. It is not that the results are uninterpretable or uninteresting. Often they are at least as interesting as what was expected and they can be explained. But the explanation cannot emerge until one abandons the original frame or expectations.

Thus, we must ask ourselves, what are the real constraints we face in our research? When developing an explanation, the phenomena to be explained (not the procedures and measures per se) represent the only relevant constraint. When applying theory, the theory itself, as well as any practical constraints on implementation warrant consideration.

How can we eschew inappropriate constraints on our theoretical creativity? Some insight is suggested by reflecting on the people and situations in which we observe creative insight. I have noticed that, in seminars, the most profound questions are not necessarily asked by the people who are most knowledgeable about the particular topic. An intelligent, highly motivated audience member from another, related field is often able to seize on the key issues or assumptions that should be debated while the so-called experts are buried in details. In a very different realm but a similar manner, children's lack of detailed knowledge enables them often to pose profound questions, as any parent has experienced.

Should we turn our research programs over to people in other areas or our children? Perhaps that's not such a bad idea. But before we do that, we might try training ourselves to think more abstractly and relationally, seeking parallels across seemingly disparate phenomena. Many exercises and techniques exist that encourage more abstract thinking (e.g., Adams 1986). No doubt you are generally familiar with many of these tools and may even have some colleagues who supplement their faculty income by training groups of executives to solve problems through brainstorming, synectics, and the like. Yet too often, creativity training (ironically) sees its primary application in business settings. When we attempt theoretical research we run for the safe "box" of previously used variables, procedures, and explanations. We decry the absence of more strict replications (which, in fact, are technically impossible to conduct and are inherently less informative than conceptual replications). To make more progress, must break out of our theoretical boxes.

CONCLUSION

In summary, let me return to the true-false items presented earlier. As you have probably inferred by now, I believe that the correct answer to each item is "false," that these are myths about theory and procedures for testing theory.

Myth 1: Progress can be increased by abandoning theory or pursuing a different level of theory than we have sought in the past.

Reality: There is no alterative to theory, there are only alternative theories.

Myth 2: It is possible to render a unique explanation for a simple (two-level) main effect.

Reality: Theoretical explanations cannot be tied to any single variable, they are inherently abstract and can only be inferred by employing convergence or triangulation procedures.

Myth 3: Certain measures or data collection procedures must be followed when testing a theory if theoretical progress is to be made. The absence of these measures or procedures constitutes a "fatal flaw."

Reality: While multiple measures are required for convergence or triangulation on theoretical constructs, any set of multiple measures, collected by any combination of methods can be adequate for assessing theory. The only fatal flaw is the presence of rival, equally parsimonious interpretation. (Though an inspection of procedures may offer insight as to why you failed to arrive at a unique explanation.)

Myth 4: Predictions constitute more rigorous tests of theory than postdictions.

Reality: When a theoretical explanation is developed is immaterial to its rigor. Parsimony can occur at any point in time.

Myth 5: A well-designed study will provide theoretical insight and will allow generalization of the effects observed to a real-world situation of interest.

Reality: Theory testing and generalization typically imply different choices regarding the subjects, the setting, and the selection of independent and dependent variables. Thus, a particular study should give clear priority to one of these goals.

FIGURE 3

Although I contend that we have made significant theoretical progress in certain areas of consumer research, junior researchers have no reason to fear that all or even most of the important questions have been answered. At the same time, there is reason to believe that we can make substantial progress in the next 25 years. A starting point for this journey is to abandon the myths that I have outlined. Instead, focus on the true nature and value of theory.

Consumer research offers a rich context in which to assess the robustness of explanations developed elsewhere. And, there is less need to focus on a narrow set of variables than there might be in a more mature, more densely populated discipline (for sociological reasons). We have greater opportunity to pursue integration of theories and convergence through creative application of methods.

To harness our potential, it may be opportune for teams of consumer researchers focus on specific consumption related issues or problems in the marketplace, with the goal of designing and testing theory-based interventions. There have been impassioned calls for such efforts (Hirschman 1989, Andreasan 1991), but to date progress has been limited to an exchange of ideas in ACR sessions and has not been translated into any hands-on projects. Such projects would be one means of operationalizing the theoretical creativity that I referred to earlier, a Consumer Behavior Odyssey of a different sort. Possible topics need not be limited to, but might include, exploring the impact of technology on consumer behavior or examining programs intended to reduce dysfunctional consumption behaviors.

One final comment. After my reference to Sulloway's birth order findings, some of you may have been ruminating about whether I am a first born or a latter born. On the one hand, you might reason that some of my views, say that post hoc explanations have the same status as a priori ones, are unconventional, implying that I am a latter born. On the other hand, you might reasonably interpret the general tenor of these remarks as an appeal to traditional values of theory and science, and, thereby, infer that I must be a first born. Therein lies the problem. It is possible for me to hold both views that are unconventional and ones that embrace the status quo, but I cannot be both first born and latter bornCat least not in this lifetime. Thus, birth order (or any other variable) can never provide a satisfactory explanation for my attitudes and behaviors...... however, for the record, I'm a first born.

Thank you for this opportunity to speak to you.

REFERENCES

Adams, James L. (1986), Conceptual Blockbusting, Reading MA: Addison-Wesley.

Anand, Punam and Sternthal, Brian (1990), "Ease of Message Processing as a Moderation of Repetition Effects in Advertising," Journal of Marketing Research, 27 (August), 345-53.

Andreasen, Alan (1992), "A Social Marketing Research Agenda for Consumer Behavior Researchers, " in Advances in Consumer Research, Vol. 20, eds. L. McAlister and M. L. Rothschild, Provo, UT: Association for Consumer Research, 1-5.

Bem, Daryl (1987), "Writing the Empirical Journal Article," in The CompleatAcademic: A Practical Guide for the Beginning Social Scientist, eds M. Zanna and J. Darley, New York: Random House, 171-201.

Bem, Daryl (1991), "Writing the Research Report," Research in Methods in Social Relations, (6th Edition), eds. C. Judd, E. Smith, and L. Kidder, Fort Worth, TX: Holt 453-476.

Berlyne, Donald E. (1970), "Novelty, Complexity, and Hedonic Value," Perception and Psychophysics, 8, 279-89.

Brinberg, David, Lynch Jr., John G., and Sawyer, Alan G. (1992), "Hypothesized and Confounded Explanations in Theory Tests: A Bayesian Analysis," Journal of Consumer Research, 19, 139-54.

Brush, Stephen G. (1989), "Prediction and Theory Evaluation: The Case of Light Bending," Science, 246, 1124-9.

Burnkrant, Robert E. and Unnava, H. Rao (in press), "Effects of Self-Referencing on Persuasion," Journal of Consumer Research.

Cacioppo, John T. and Petty, Richard E. (1979), "Effects of Message Repetition and Position on Cognitive Response, Recall, and Persuasion," Journal of Personality and Social Psychology, 37 (January), 97-109.

Calder, Bobby J., Phillips, Lynn W., and Tybout, Alice M. (1981), "Designing Research for Application," Journal of Consumer Research, 8, 197-207.

Gell-Mann, Murray (1994), The Quark and the Jaguar, New York: W. H. Freeman and Company.

Hirschman, Elizabeth (1990), "Secular Mortality and the Dark Side of Consumer Behavior: Or How Semiotics Saved My Life" in Advances in Consumer Research, Vol.18, eds. R. H. Holman and M. Soloman, Provo, UT: Association for Consumer Research, 1-4.

Hunt, Shelby (1993), Modern Marketing Theory: Critical Issues in the Philosophy of Marketing Science, Cincinnati, OH: South-Western Publishing Co.

Iacobucci, Dawn, Grayson, Kent, and Ostrom, Amy (1994), "Customer Satisfaction Fables," Sloan Management Review, Summer, 93-96.

Kernan, Jerome (1994), "Declaring a Discipline: Reflections on ACR's Silver Anniversary," in Advances in Consumer Research, Vol. 22, eds., F. Kardes and M. Sujan, Provo, UT: Association for Consumer Research.

Kerr, Norbert (1994), FRCRP Project, Department of Psychology, Michigan State University, East Lansing, Michigan 48824.

Meyers-Levy, Joan and Peracchio, Laura (in process), "Understanding the Effects of Color: How the Correspondence Between Available and Required Resources Affects Attitudes," Journal of Consumer Research.

Shimp, Terence (1993), "Academic Appalachia and the Discipline of Consumer Research," in Advances in Consumer Research, Vol. 21, D. Roedder John and C. Allen, Provo, UT: Association for Consumer Research, 1-7.

Sternthal, Brian, Tybout, Alice M. and Calder, Bobby J. (1994) "Experimental Design: Generalization and Theoretical Explanation", in Principles of Marketing Research, ed. R.P. Bagozzi, Cambridge Mass.: Blackwell Publishers, 195-223.

Sternthal, Brian, Tybout, Alice M. and Calder, Bobby J. (1987), "Confirmatory versus Comparative Approaches to Judging Theory Tests," Journal of Consumer Research, 14 (June), 114-25.

Stiff, David (1994) "Blame the Birth Order for History's Revolts, This MIT Scholar Says," Wall Street Journal, August 23, 1994 1, 8.

Wallach, Lise and Wallach Michael A. (1994), "Gergen Versus the Mainstream: Are Hypotheses in Social Psychology Subject to Empirical Test?," Journal of Personality and Social Psychology, 67, 2, 233-242.

Wells, William D. (1993), "Discovery-oriented Consumer Research," Journal of Consumer Research, 19 (March), 489-504.

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