Feelings About Feeling-State Research: a Search For Harmony

Citation:

Carl Obermiller and April Atwood (1990) ,"Feelings About Feeling-State Research: a Search For Harmony", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 590-593.

Advances in Consumer Research Volume 17, 1990      Pages 590-593

FEELINGS ABOUT FEELING-STATE RESEARCH: A SEARCH FOR HARMONY

Carl Obermiller, Seattle University

April Atwood, University of Washington

The '80's will be seen as the decade of feelings in consumer research. In the '70's, our focus was attitude, and the paradigm was a cognitive one, the multi-attribute model of structure. But the coldly cognitive framework, however useful, was clearly limited. Explained variance in global evaluation measures or behaviors rarely exceeded 20%, and was more often less than 10%. Moreover, like the more general rational expectations framework in decision making, multi-attribute attitude models assume a degree of elaboration that seems implausible for observed consumer behavior.

Early warnings were sounded by Krugman (1965) and Robertson (1971) who cautioned that much, if not most, consumer behavior is characterized by low involvement decision making and that low involvement results in subtler evaluation processes. Wright (1973) introduced the concept of affect referral to describe the role of feelings about brands in consumer evaluation. Wright proposed that uninvolved consumers would be insufficiently motivated to engage in an elaborate evaluation rule and would merely access a global feeling associated with each alternative. Wright, however, regarded those feelings to be the result of some prior cognitive processing, now stored separately in memory, rather than direct affective responses .

Not until Zajonc's (1980) landmark article on feelings as an independent, primary, and compelling evaluative force did consumer researchers begin to make dramatic changes in their conceptualizations. Throughout the decade, Zajonc has argued that affective responses are an independent processing system, which may respond to perceptual processes but may be as separate from cognitive evaluation as to originate in the muscular system. Gorn (1982) renewed interest in classically conditioned emotional responses. Mitchell and Olson (1981) helped to develop a stream of research on attitude-toward-the-ad, that has, in turn, acknowledged affective responses to ad experiences as important. Holbrook (c.f. Holbrook and Batra 1987) has lately done considerable work on the conceptual side of emotion in consumer responses. Isen (c.f. Isen, et al 1978) introduced new effects of feeling states on psychological processing, and Gardner (1985) followed by introducing consumer researchers to the effects of moods.

The preceding brief review is intended to emphasize the importance of feelings as a topic in consumer research for the past decade. A much longer paper would be required to describe how much progress we have yet to make before we can claim to understand feelings nearly so well as we have come to understand their cognitive counterparts, beliefs. We stand in a paradox; the ways we think and feel seem quite natural and simple; yet, the ways we think about the way we u?link and feel seem quite artificial. Good science exhibits a variety of pleasing harmonies that our research into consumer feelings has yet to display: The harmony of mysteries revealed, when theory represents internally consistent concepts that generate surprising, yet compelling hypotheses. And the harmony of methods that are consistent with the goals implied in the theory or acknowledged by the scientist. And, finally, the harmony of analyses, appropriate to the methodology, and yielding interpretations appropriate to the theoretical or applied goals of the researcher. Such harmony within and among the aspects of scientific research has been elusive in our attempts to understand affective processes, and the three papers of this session are not exceptions. Notwithstanding their valuable contributions, mostly at the conceptual level, all of the papers exhibit disharmonies: Theoretical reasoning is internally inconsistent; conclusions are discordant with existing theory; methods are out of rhythm with research purpose; analyses are out of sync with the designs that generated them.

"Product Type: A Neglected Moderator of the Effects of Mood" by Gardner and Scott is admirable for its attempt to develop theory. The paper takes the logical step of refining known relationships between mood and outcome states--product information availability, product evaluation, and product use. Product type is proposed as a mediator of the mood-outcome relationships, which are premised upon structural and motivational effects of mood. The structural effect, termed by Gardner and Scott accessibility, is most associated with Isen's work showing a bias in information availability in memory, consistent with current mood. The motivational effect, which Gardner and Scott refer to as mood management, is associated with Mischel's work showing that behavior is consciously guided by desire-to attain or sustain good moods and avoid or terminate bad moods. Gardner and Scott propose four product types as mediating conditions--"feel-good" and "feel-bad" products that, respectively, result in or cause good moods and bad moods, "try-not-to-feel" products, which imply consequences sufficiently important that consumers attempt to suppress mood when considering them, and "no-feel" products, which neither imply nor evoke an emotional response.

Two quite general concerns strike us at the outset. First, if we are to borrow the principle of mood management, we ought to examine it further. The authors paraphrase Mischel (1973), "individuals develop expectations for the outcomes of various behaviors, which, in turn, direct their own behavior." This may be the core of the mood management perspective, but it is little different from a cognitive instrumental learning perspective. Our behaviors are guided by the constant motivation to seek instrumental outcomes. It is unclear why we need to invent a new theoretical perspective specifically for moods when an old standby will do as well. Moreover, the major flaw in instrumental learning "theory" is still evident in the mood management perspective--it can explain behavior post hoc but cannot predict. We have little confidence in our knowledge of the mood expectations consumers attach to product use.

Second, and related to the first issue, any attempt to construct theory based on product typology is limited by the idiosyncratic nature of responses to products. One consumer feels good about mortgage insurance, focusing on the security it brings; another feels bad, focusing on possible disaster; another tries not to feel, suppressing emotions because of the importance of the decision; still another is completely uninvolved and feels nothing. Gardner and Scott acknowledge this problem but should discuss it further. It is unclear that product type can be identified with enough accuracy to make it useful for predictions.

Aside from the general concerns, we find several internal inconsistencies in the logic of the propositions. While we fault the theoretical structure here, we suggest that these problems might be fruitful research questions. The first is a minor inconsistency in definitions. For P3, Gardner and Scott exclude effects on product use with an explanation in the form of an instance--consumers cannot let mood dictate when their cars will be repaired. The example is irrelevant, however, since, according to the definition, car repair is not a feel-bad product. Car repair implies immediate, not future, needs. Preventative car maintenance would qualify as a feel-bad product, and it should show the same biases in patterns of product use as in information availability and product evaluation. The exclusion of use from P3 is not consistent with the logic and definitions that are developed.

Two other incontinencies suggest contradictory effects of the structural and motivational aspects of mood. The logic of P1 flows from mood management, but it contradicts the accessibility perspective. Information availability has been shown to be consistent with current mood, so when consumers are in a negative mood, they should exhibit a negative bias toward feel-good products. The logic of P5, on the other hand, appears to flow from the accessibility perspective-the availability of try-not-to-feel product information, because such products are defined as feeling-suppressed, is unaffected by mood. Yet, mood management logic suggests that, because the outcomes of such product decisions are important and mood relevant, we should expect to see positive biases associated with both positive and negative mood states. Whether motivational or structural effects of mood predominate for feel-bad and try-not-to-feel products may be an area for investigation.

Gardner and Scott attempt to refine theory by proposing mediating variables; Mano, in "Emotional States and Decision Making," attempts to refine the essential construct--feelings. Borrowing from the work of Mehrabian and Russell, Mano questions the finding that positive mood state results in faster, less complete decision making (Isen and Means, 1983). His argument is cogent; positive mood state is multidimensional, including, at least, both pleasantness and some degree of arousal. The combination of pleasantness and high arousal (elation) should reduce attention relative to pleasantness and low arousal (calm). His aim was to manipulate degree of arousal within a narrow range of pleasantness to demonstrate that changes in decision making may result from both dimensions of mood.

Our objections to Mano's work are in the methodology, not in the conceptualization, which we find persuasive and fruitful. First, when one proposes a rival explanation of a prior finding, a claim against the construct validity of the original manipulation (as Mano does with respect to Isen and Means 1983), one should begin with the procedure not the result. Mano is correct when he claims that his "theory" can account for the Isen and Means data and for his own hypothesized outcomes, but the reader-would be more persuaded by a careful contrast of the procedures. We are asked to assume that Isen and Means naively confounded pleasantness and arousal; we should be shown how their manipulation did so.

Second, and more fundamental, Mano's experiment falls into the crack between applied and theory testing research. The objective appears to be a theory test; yet, the manipulation procedure (exposure to "real world" commercials) is inadequate to the requirements of construct validity for theory testing. It is not mere hindsight that supports this criticism. There is no reason to expect exposure to commercials to be a strong pure manipulation of mood and no reason to trade off a weak manipulation for any external validity concerns. After all, there is no external validity in forced exposure to ads in a lab setting anyway. When we want to manipulate moods, we should design manipulations that do so, reliably and forcefully. Whether and how commercials affect moods in mundane conditions is a separate, very different question, to be answered with separate, very different methodology.

Because of the inadequate manipulations, Mano's research was reduced to a correlational analysis. Our final criticism is a minor one, referring to the nature of the analysis. Mano presents the results of an overall correlation in Table 3. He also divides the sample into low and high levels on the mood dimension (Table 2) and conducts analysis of variance. What is the justification for the anova, which has substantially lower power? Moreover, the data in the two tables show contradictory effects of activation (arousal). In Table 2, decision time, revealed entries, intra- and inter-moves, repetitions, and effort all increase from low to high activation; yet, the correlations in Table 3 are negative for decision time, revealed entries, and repetitions. Given that this is the central research question, this discrepancy should be discussed. If it is an artifact of the split sample anova, it is further reason for not conducting that analysis.

Dube-Rioux, in 'The Power of Affective Reports in Predicting Satisfaction Judgments," tackles feelings simultaneously as independent and dependent variable. Her contribution is to the conceptualization of satisfaction. It is typically characterized as a cognitive appraisal of attribute performance versus attribute expectations. Dube-Rioux argues for a significant independent contribution of affective response, in particular when satisfaction is, in her terms, "emotionally intense."

We are receptive to the claim that satisfaction should be reconceptualized and that feelings should be acknowledged as part of that construct. We were puzzled somewhat by Dube-Rioux's mediating condition--the emotional intensity of satisfaction. It is unclear what is gained by the claim that for "emotional" satisfaction, measures of emotional response play a larger part than they do for "unemotional" satisfaction. Perhaps some further work can be done to characterize the conditions under which emotion plays a greater or lesser role in satisfaction. One suggestion is to consider the product typology of Gardner and Scott. Affective response measures may be better predictors of satisfaction with feel-good or feel-bad products; worse for the other types.

The use of regression analysis of the contributions of affective and cognitive components raises several questions. First, the two measures are highly correlated, so evidence that only one component is statistically significant in an equation is not compelling evidence that the other has no explanatory power, especially with a small sample size. Second, a comparison of beta weights within an equation assumes that subjects interpreted the affective and cognitive metrics in the same way. In general, a comparison of affective versus cognitive contributions is a comparison of apples and oranges. It begs questions of measurement validity that cannot be answered satisfactorily. A better approach that is available in Dube-Rioux's framework is a comparison of the contribution of either component under different conditions. The appropriate test is whether the beta weights for affect or cognition change between emotional and neutral satisfaction conditions. Such a comparison, however, requires assumptions of certain equalities across samples that might not be met.

Beyond the use of theory, method, and analysis that are in harmony within a study, researchers in this area would do well to strive to conduct research that is clearly tied to the growing body of literature on the effects of feelings and moods. One important step is to define with care the constructs under study. Previous work has attempted to differentiate among affect, emotion, mood, and feelings on conceptual grounds (c.f., Gardner, 1985; Holbrook and Batra, 1987); current and future research should make clear which constructs are of interest and should indicate the operational definitions that are being used. Beyond a clear definitional statement, the research should describe the underlying process of interest and should specify the role that moods or emotions assume. This trio of papers examines affective responses or moods in a variety of roles: Gudner and Scott treat mood as an independent variable whose effects on consumer behaviors are moderated by the type of product under consideration, while Mano examines emotional state as a moderator of consumer decision making, and Dube-Rioux, in her conceptualization of satisfaction as an emotion-based construct, considers feelings as dependent variables in the process under consideration. Some cataloging of which conceptual role the affective responses or moods are filling may help to organize the broad existing literature on moods and would help clarify the handling of the construct in each prospective study.

A second step toward better integration of new studies with existing work in the field can be taken in careful choice of measures. Researchers should consider consolidating, refining, and standardizing past measures of feelings before creating new ones. Ambiguity among constructs is exacerbated by a never-ending series of new operational measures

A third and final general comment is to repeat a criticism of Mano's paper and issue a general call for more appropriate mood induction manipulations. If the aims of a study are to examine moods or the effects of moods in mundane consumer settings, then effects-testing procedures are appropriate and mood manipulations high in mundane realism would be called for. If, however, the aims of the research are theory-testing, then strong experimental manipulations or inductions of mood are appropriate. James Russell, a psychologist who has done considerable work in the area of conceptualizing mood states and examining their effects (c.I. Russell, 1980), reports that a relatively long (twenty minutes) audio tape using musical selections has been effective in inducing mood states (positive or negative); such an approach, while not resembling most real world situations, might be especially useful for theory-testing oriented studies of feeling states. We need to separate the research that aims to answer theoretical questions that are proposed as relevant to consumer behavior from the research that aims to describe the effects- that occur in everyday consumer behavior.

REFERENCES

Holbrook, Morris and Rajeev Batra (1987), "Assessing the Role of Emotions as Mediators of Consumer Responses to Advertising," Journal of Consumer Research, 14 (December), 404420.

Gardner, Meryl (1985), "Mood States in Consumer Behavior: A Critical Review,: Journal of Consumer Research, 12 (December), 281-300.

Gorn, Gerald (1982), 'The Effects of Music in Advertising on Choice Behavior: A Classical Conditioning Approach," Journal of Marketing 46 (Winter), 94- 101.

Isen, Alice, T. E. Shalker, M. Clark, and L. Karp (1978), "Affect, Accessibility of Material in Memory and Behavior: A Cognitive Loop?," Journal of Personality and Social Psychology, 36, 1-12.

Isen, Alice, and Barbara Means (1983), "The Influence of Positive Affect on Decision Making Strategy," Social Cognition, 2, 18-31.

Krugman, Herbert (1965), "The Impact of Television Advertising: Learning Without Involvement," Public Opinion Quarterly, 29 (Fall), 349-356.

Mischel, W. (1973), "Toward a Cognitive Social Learning Reconceptualization of Personality," Psychological Review, 80, 252-283.

Mitchell, Andrew, and Jerry Olson (1981), "Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitude?" Journal of Marketing Research, 18 (August), 318-332.

Robertson, Thomas (1976), "Low-Commitment Consumer Behavior," Journal of Advertising Research, 16 (April), 19-24.

Russell, James A. (1980), "A Circumplex Model of Affect," Journal of Personality and Social Psychology, 39 (6), 1161-1178.

Wright, Peter (1973), "The Cognitive Processes Mediating Acceptance of Advertising," Journal of Marketing Research, 10, 53-62.

Zajonc, Robert (1980), "Feeling and Thinking: Preferences Need No Inferences," American Psychologist, 35 (February), 151-175.

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