Attribution, Self-Perception, Salience, and Weird Interactions

ABSTRACT - Aided in part by the papers written by Reingen and Bearden and by Allen and Dillon, it is becoming increasingly clear that self-perceptions can influence behavior. This paper reviews briefly the above work as well as the paper by Yalch and Yoshida. Of particular relevance is their finding of a surprising fourth order interaction. Other such strange interactions involving the Kelley ANOVA model are noted and discussed.

Citation:

John C. Mowen (1983) ,"Attribution, Self-Perception, Salience, and Weird Interactions", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 56-58.

Advances in Consumer Research Volume 10, 1983      Pages 56-58

ATTRIBUTION, SELF-PERCEPTION, SALIENCE, AND WEIRD INTERACTIONS

John C. Mowen, Oklahoma State University

ABSTRACT -

Aided in part by the papers written by Reingen and Bearden and by Allen and Dillon, it is becoming increasingly clear that self-perceptions can influence behavior. This paper reviews briefly the above work as well as the paper by Yalch and Yoshida. Of particular relevance is their finding of a surprising fourth order interaction. Other such strange interactions involving the Kelley ANOVA model are noted and discussed.

INTRODUCTION

Each of the three papers in the session dealt with attributional processes involving the issue of how people gain knowledge concerning themselves. The Allen and Dillon and the Reingen and Bearden papers investigated labeling and its effect on verbal ratings. Reingen and Bearden went a step further and assessed its impact on consequent behavior. Yalch and Yoshida utilized the Kelley ANOVA model as an approach to understanding and predicting how consumers come to develop confidence in their evaluations of new products or retail stores. Each study was well executed and possessed a high degree of internal validity. All make some contribution to the marketing literature on attribution theory.

Paper 1--Salience and Labeling

Reingen and Bearden investigated the construct of salience and its impact upon the efficacy of labeling in influencing behavior. The idea be ind the labeling hypothesis is that providing someone with a label helps them to develop a self-perception about the type of person they are. Subsequently, the individual uses the label as a guide to his behavior. Building on the work of Tybout and Yalch (1980), Reingen and Bearden tested the hypothesis that the efficacy of the labeling technique would depend upon its salience.

The concept of salience is slippery. Indeed, in the psychology literature evidence of the effects of salience is mixed (Taylor and Thompson 1982). Nisbett and Ross (1980) stated that something tends to be salient if it is of emotional interest, is highly concrete, or has proximity (either temporal, spatial, or sensory). Tybout and Yalch viewed the salience of the labeling cue as mediated by two factors--its consistency with the individual's self-concept and the availability of other relevant ones.

In their study Reingen and Bearden operationalized salience in an interesting way. Subjects were asked to comply with either a relatively large request or a relatively minor task. They were then given the label of "helpful" people. The authors hypothesized that the label would be more salient in the high initial request condition, due to the operation of Kelley's (1973) discounting principle. That is, in the small request condition, subjects would probably feel that "everybody would have done this- and not make an internal attribution that they were indeed helpful. In the large request condition, an "augmenting" effect would occur resulting in an internal attribution. Such an internal attribution, then, would make salient the label. The question is, 'is this a manipulation of salience or a manipulation of something else? My view is that the manipulation was not of salience but of an attributional state in which the person makes either an external or internal attribution for his behavior. The perception of the locus of causality then either reinforced or canceled the labeling manipulation. The key point is that salience is a hypothetical construct which is difficult to measure directly.

The results of the Reingen and Bearden study partially supported their predictions. Verbal compliance rates were highest when a label was given and the subjects had said yes' to the large first request. However, behavioral compliance was no greater in this condition than the other experimental conditions. Five potential explanations were provided by the authors for the failure of the behavior to follow verbal commitment. Particularly relevant to the salience concept are the possibilities that the passage of time reduces the salience of the label or that the credibility of the labeler may influence the effects of the label.

Paper 2--Self-perception Mediates Choice?

Allen and Dillon's paper makes the excellent point that little direct evidence has been found of a link between self-perception development and decision making. For example, if someone is given the label of a cost conscious consumer is it the label or some other factor which later influences buying behavior? Their paper focused on providing evidence for a link between self-perception and the developments of importance ratings on product performance attributes.

In two separate analyses, one using covariance analysis and the other using causal analysis, the results provided indirect evidence that a proposed self-concept measure did appear to mediate the impact of changes in the type of promotional appeal and subjects' ratings of the importance of energy efficiency in a refrigerator4 freezer. In the study, subjects viewed one of four different promotional appeals. One of the appeals labeled the American consumer as a willing participant in solving the energy problem. The results revealed that the labeling appeal influenced the subjects such that the energy use attribute was rated significantly more important in that condition. Follow-up analyses supported the hypothesis that self-perception formation mediated the changes in importance ratings.

The key question in the Allen and Dillon study was whether their dependent measure assessing the self-concept had construct validity. The two items used to assess the self-concept were: "I've changed the way I purchase and use products as a result of the nation's energy problem" and "When I buy products, I seldom consider how my use of them will contribute to reducing the severity of the nation's energy problem." The statements do deal with "knowledge of reasons for buying,-but is this relevant to the self-concept?

Allen and Dillon also tested a second index, labeled "general feelings of effectiveness." Only one of the five items in the scale was mentioned in the paper--"energy is really not my problem because there is simply nothing I can do about it." This item reads remarkably like items in Rotter's internal-external locus of control scale. This scale and similar others assess the individual's feelings of whether the events which occur to him are caused by uncontrollable outside forces or by his own efforts and abilities. If the scale is a measure of locus of control, it would be highly unlikely that viewing the television ads would affect ratings. Feelings of control are extremely difficult to change even in intensive programs designed for that purpose.

An interesting question spurred by the Allen and Dillon study concerns the relationship between self-perception and the self-concept. My guess is that only over a long period of time will changes in self-perceptions influence the self-concept. This indicates that for a self-attribution inducing stimulus, like labeling, to influence behavior it must tie into the individual's existing self-concept. Notably, this is the very effect investigated and found by Tybout and Yalch (1980). Whether the effect results from salience as Tybout and Yalch suggest or from another mechanism, such as acceptance of the label, the result is the same. For labels to be effective they must fit the individual's already formed self-concept.

Paper 3--The Kelley Cube

The Yalch and Yoshida work tested Kelley's ANOVA model as an approach to understanding how consumers develop confidence in their attributions. As has happened to this author several times in the past, the ANOVA model was only minimally supported.

The most intriguing aspect of the results of the study was the 4-way interaction among the independent variables--distinctiveness, consistency over modality, consistency over time, and consensus for the major dei>- pendent variable of decision confidence. Interpreting 4-way interactions post hoc is extremely difficult, and after a couple of tries this author gave up. Yalch and Yoshida did not discuss the interaction. The problem is that interactions override main effects and should be looked at prior to examining the main effects.

The appearance of unanticipated interactions when the Kelley ANOVA model has been applied to marketing phenomena has plagued this author in research investigating the perception of celebrity endorsers. Mowen and Brown have conducted several studies on the topic and consistently obtained complex interactions involving distinctiveness, consensus, and the age of the subject (Mowen and Brown 1980; Mowen, Brown, and Schulman 1979). On an individual basis, we were able to develop post hoc explanations for the effects, but only through a great deal of imagination--not necessarily a desirable quality in interpreting results. However, when the studies are looked at together, the results are completely inconsistent.

On an intuitive level the Kelley ANOVA model makes sense. Tests of the model in the sterile laboratory environment support it. However, when applied to marketing phenomena, it becomes extremely hard to manipulate only distinctiveness, consensus, etc., and not inadvertently simultaneously create a confounding variable. For example, in one of the .Mowen and Brown studies an attempt was made to manipulate distinctiveness by varying whether the endorser spoke for only one product or also endorsed several others. Subjects over 25 years of age responded as expected. They rated the endorser, who spoke for only one product, as doing the ad because he believed in the product in comparison to the endorser who spoke for several products. In contrast, subjects under 25 years of age rated higher the endorser who spoke for several products. The problem was that later studies not only failed to confirm this finding, but also went on to yield other even more "weird interactions.

As a summary statement, this author views applications of the Kelley ANOVA model to marketing as difficult. The studies require the use of covariates, the careful selection of adult, not student, consumers (because of the age interactions), and the highly precise manipulation of the key variables. I don't believe the theoretical model should be abandoned, but it certainly should not be taken yet as representing how consumers make attributions about others or themselves in consumption situations.

Thoughts on Future Research

A number of thoughts come to mind regarding directions for future research. One area concerns just what labeling does to increase the likelihood of compliance to requests. The traditional hypothesis is that labeling creates a self-perception which later mediates behavior

An alternative explanation of the labeling phenomenon is that it creates a positive mood state. When one examines how labels are assigned, it almost invariably involves giving the subject a compliment. For example, in the Reingen and Bearden study, they told the subjects, 'You are a very helpful person and I wish more of the people we call were as helpful as you are.' Previous research by Isen and her colleagues has shown that the induction of a positive mood statement, often by giving a compliment, increases the likelihood of compliance to a request.

The test for the mood state hypothesis is quite simple. The experimenter creates an "irrelevant" label control group. These individuals are provided with a very positive label which is irrelevant to the next request. Another possibility is to give a negative label. That is, tell the subJects that they are "unhelpful" or some such thing. The labeling hypothesis predicts that compliance should decrease. However, the work on moot indicates that compliance should increase (Isen, Clark, and Schwartz 1976). That is, both negative and positive moods have been shown to increase the likelihood of compliance.

A second area of research involves the potential for boomerang effects when giving labels. Jones (1964) discussed a tactic for augmenting power in interpersonal interactions called "ingratiation." The idea is to gain power by making oneself more attractive to the other. One method of doing this is to compliment the other. The problem with ingratiation tactics is that the ingratiator can be caught with a consequent boomerang effect resulting. That is, the target person may realize that the complimenter had ulterior motives and react very negatively.

The process of labeling can be viewed as a subtle form of ingratiation. If the target of the labeling effort suspects that the experimenter has ulterior motives, the process may boomerang. It has been noted that the foot-in-the-door technique is difficult to implement in applied settings. I suspect that a similar problem will be found to occur with the labeling technique because of the "ingratiator's dilemma."

The three papers presented in the session nicely represent what Kuhn has called "normal science.' They attempt to elaborate and extend previously developed theories or hypotheses. This work should continue in the attribution theory area. Three additional issues raised by the papers deserving further research are: (1) Why does one obtain interactions when the Kelley ANOVA model is operationalized in a marketing setting? The theory only predicts main effects. (2) How are self-perceptions related to the self-concept? How should one measure the self-concept when investigating its relationship to consumer behavior? (3) Just how do you define salience or vividness? Is salience a property of the stimulus, the person, or the person interacting with the environment. Is salience related to involvement? Is a high involvement product salient?

REFERENCES

Isen, A. M., M. Clark, and M. Schwartz (1976) "Duration of the Effect of Good Mood on Helping: Footprints on the Sands of Time," Journal of Personality and Social Psychology, 34, 385-393.

Jones, E. E. (1964) Ingratiation: A Social Psychological Analysis, New York: Appleton-Century-Crofts.

Kelley, H. H. (1973) "The Process of Causal Attribution," American Psychologist, 28, 107-28.

Mowen, J. C., S. Brown, and M. Schulman (1979) "Theoretical and Empirical Extensions of Endorser Effectiveness," 1979 Educator's Conference Proceedings, American Marketing Association, Neil Beckwith et al. (editors), 258-262.

Mowen, J. C. and S. Brown (1980) 'On Explaining and Predicting the Effectiveness of Celebrity Endorsers," Advances in Consumer Research, Volume VIII, Kent Monroe (editor), Association for Consumer Research, 437-441.

Nisbett, R. and L. Ross (1980) Human Inference: Strategies and Shortcomings of Social Judgment, Prentice-Hall, Inc., Englewood Cliffs, N.J.

Taylor, S. and S. Thompson (1982) "Stalking the Elusive 'Vividness' Effect," PsychologicaL Review, 89, 155-181.

Tybout, A. M. and R. R. Yalch (1980) "The Effect of Experience: A Matter of Salience?" Journal of Consumer Research, 6, 406-13.

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