Emotional States and Decision Making

Haim Mano, Washington University in St. Louis
ABSTRACT - This study investigates the influence of affect's two primary dimensions, Pleasantness and Arousal, on multiattribute choice processes. Subjects described their emotional state and performed two multiattribute choice tasks. The results indicate that subjects in more pleasant mood spent more time deliberating and used more decision-related information. The results are interpreted in terms of (1) a congruency between one's hedonic state and selected decision strategy, and (2) restriction in attentional capacity.
[ to cite ]:
Haim Mano (1990) ,"Emotional States and Decision Making", 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: 577-584.

Advances in Consumer Research Volume 17, 1990      Pages 577-584

EMOTIONAL STATES AND DECISION MAKING

Haim Mano, Washington University in St. Louis

ABSTRACT -

This study investigates the influence of affect's two primary dimensions, Pleasantness and Arousal, on multiattribute choice processes. Subjects described their emotional state and performed two multiattribute choice tasks. The results indicate that subjects in more pleasant mood spent more time deliberating and used more decision-related information. The results are interpreted in terms of (1) a congruency between one's hedonic state and selected decision strategy, and (2) restriction in attentional capacity.

In recent years cognitive, social, and consumer psychology have shown an increased interest in the influence of emotional states on decision making (for reviews, see Gardner 1985; Isen 1987). Two common features of this line of research are that it (i) views affect as a unidimensional construct consisting of good, neutral, and bad mood, and (ii) examines affect's impact on behavior by experimentally inducing good or bad moods and contrasting between resulting behaviors (or with those resulting from neutral moods).

Until recently, both theory and evidence on the relationship between affect and decision making have suggested that people in positive affect will tend to reduce decision complexity by engaging in speedy and simplifying kinds of processing, e.g., shorter decision times, lesser acquisitions of decision related information (Isen and Means 1983). The suggested theoretical explanations were that either (i) positive affect increases the load on working memory, causing subjects to compensate for the increase by cutting elsewhere; or, (ii) if not increasing the load, positive affect somehow makes subjects more sensitive to and avoidant of the cognitive strain demanded when complex strategies are employed (Isen 1987). In particular, for a multiattribute choice task, Isen and Means (1983) found that--compared to subjects in a neutral mood-subjects in a positive mood made faster choices, revealed less information, were less likely to review information they had already looked at, used fewer attributes and ignored more attributes considered unimportant, and were more likely to eliminate from consideration alternatives that did not meet a threshold criterion on an important dimension.

Recently, Lewinsohn and Mano (1989) examined decision behavior of subjects whose moods were not experimentally manipulated. When performing a multiattribute choice task, subjects in naturally occurring pleasant mood spent more time deliberating, examined more decision related information, used more effort requiring interdimensional moves, and ignored fewer product attributes. Lewinsohn and Mano explained the contrasts with Isen and Means' (1983) in terms of the differential impact of affect's two primary dimensions, Pleasantness and Arousal. Two mediating mechanisms were offered to explain affect's influence on decision making, one for Pleasantness and one for Arousal. Regarding Pleasantness, Lewinsohn and Mano suggested a congruency between one's current emotion and cognitive operations (Bower 1980; Isen 1987). Subjects experiencing higher Pleasantness would be in a more playful state of mind, which would lead them to enjoy the decision situation and satisfy their curiosity. Therefore a consequence of one's better mood would be a more thorough and deliberative decision process. This hypothesis, even though not consistent with Isen and Means's (1983) results, it is consistent with the notion that persons in a better mood state are more creative and perform better in problem solving tasks (Isen et al 1987).

The second mechanism, which was also examined and confirmed (Lewinsohn and Mano 1989) suggests that individuals experiencing higher levels of Arousal are more likely to have a narrower attention span (Kahneman 1973); in turn, this attention-triggered selectivity may lead subjects to spend less time deliberating, focus on fewer decision related information pieces, and ignore more of the products' attributes.

Given that Lewinsohn and Mano's (1989) results stand in contrast with those of Isen and Means, the present study will attempt to replicate and extend Lewinsohn and Mano's findings with subtle mood manipulations induced by exposure to television commercials, and to clarify further affect's influence on decision making, in general, and multiattribute choice, in particular.

AFFECT AS A MULTIDIMENSIONAL CONSTRUCT

The focus of much of past research on the relation between affect and cognitive and social behaviors has been on affect's hedonic tone (i.e., pleasantness-unpleasantness). However, a host of evidence has been offered suggesting that the affective experience is not unidimensional. Affect's multifaceted nature has been modeled in varying forms of structural complexity, generally categorized under two paradigms: Dimensional description and Classification.

The dimensional paradigm suggests two principal independent dimensions, Pleasantness and Arousal, as underlying the emotional experience whereby feelings are expressed as points on the perimeter of a circle--called a circumplex (Figure 1)-determined by the two dimensions (e.g., Holbrook and Batra 1987; Russell 1980). The parsimony and generality of two dimensions stands in contrast to the richer view of the Classification paradigm which suggests a number of independent of clusters (e.g., Joy, Sadness, Anger) as the units for classification of emotions (Batra and Ray 1986; Batra and Holbrook 1988). Nonetheless, it is important not to consider the paradigms as opposing but rather as complementary (Havlena & Holbrook 1986; Russell 1980).

FIGURE 1

THE MOOD CIRCUMPLEX

The present study views affect as a two-dimensional construct of Pleasantness and Arousal. By specifying in a more refined manner the decision maker's mood states, the two-dimensional view has important implications for the relationships between moods and decision making. Consider, for example, manipulations used to induce good mood (e.g., watching a comedy film, receiving a small free gift, or receiving success feedback in a motor-skills task). Even though these situations will induce positive affect, the evoked emotions may differ both in the intensity of their pleasantness as well as in the degree of their arousal (cf. Clark, Milberg and Erber 1984). It is important, therefore, to consider-both theoretically and empirically--whether the two central positive feelings, Elation and Pleasantness have the same effects on decision making and, if not, what their differential impacts are (Isen, Daubman and Nowicky 1987).

More generally, given affect's multifaceted nature we should delineate the effects of affect's specific facets on decision making. When examining whether an emotional state influences some behavior, one must examine whether the specific emotional state was active and what other emotional state(s) might have been active when the effect took place; and, what were the intensities of these activations. Otherwise. effects attributed to one domain of affect may, in fact, be stemming from other or additional domains.

HYPOTHESES

(1) It is expected that subjects experiencing a more positive mood will be engaged in more elaborate decision making activities.

(2) Subjects experiencing higher degrees of arousal will have a narrower attention span and therefore will spend less time deliberating, focus on fewer decision related information pieces, and ignore more of the alternatives' attributes

METHOD

The present study attempted to induce subtle emotional states by exposing subjects to TV ads prior to their performing two multiattribute choice tasks. Past research (Aaker et al 1987; Edell and Burke 1987; Gorn 1982) has suggested TV advertising's ability to generate emotional reactions. In order to examine the ads' direct impact on emotions, the ads were not embedded as part of some ongoing TV program. Using a mood questionnaire, emotional states were assessed twice: immediately following exposure to the commercials and after the two choice tasks. The mood questionnaire measured the emotional intensity of each of the octants in affect's circumplex. Analyses concentrated on the impact and relationships among subjects' emotional states and decision processes involved in the choice tasks.

TABLE 1

TOTAL NUMBER OF RESPONSES ON THE SCALES FOR THE FOUR ADS (MANO 1988)

Subjects

Thirty-two subjects were recruited through ads placed in campus offering $2.50 for participation in a decision making experiment that would last about 20 minutes.

Procedure

Subjects were randomly assigned to one of four different sets of ads and were run individually. On arrival, they read an overview of multiattribute choice followed by instructions for information acquisition and choice. Subjects were then trained by choosing from a 3X4 information board, similar to those used in the subsequent main choice tasks. After this practice task, subjects saw one of the four sets of TV ads and proceeded to answer the mood questionnaire. Next, they performed the two main choice tasks and, upon their completion, the mood questionnaire was readministered. Subjects then answered three cognitive-effort related questions. Finally, they were asked a few background questions, were debriefed, and paid.

TV ads

Each of the four groups of subjects was exposed to a set of ads intended to generate a somewhat different emotional impact. The ads were chosen from a previous study (Mano 1988) which assessed the emotional impact of 41 prime-time TV ads. In that study, following exposure to each ad, 18 subjects reported on a checklist format whether they felt emotions associated with each of the eight emotional states (for the list the emotions see "Mood Questionnaire"). Table 1 presents the total number of responses to the four ads on each affect scale.

All four sets started with Ad 1 which was expected to have a weak emotional impact. Set 1 consisted only of Ad 1. Sets 2, 3, and 4 started with Ad 1 followed by Ad 2, 3 or 4, respectively. Ad 2 was expected to have an impact on Pleasantness, Ad 3 on Elation and Pleasantness, and Ad 4 on Calmness and Pleasantness. It was anticipated that subjects exposed only to Ad 1 (Group 1) would retain their initial emotional state. On the other hand, subjects in Group 2, (exposed to Ad 2 after Ad 1), were expected to be induced into a more pleasant emotional state; similarly, Ad 3 (Group 3) was expected to have an impact on Elation and Pleasantness; and Ad 4 (Group 4) to evoke feelings of Pleasantness and Calmness. The rationale for using a mild ad first was based, in part, on recent evidence that preceding ads may become the background against which subsequent ads are evaluated; in particular, some contrast between two subsequent ads may enhance the emotional appeal of the second ad (Aaker et al. 1983).

Mood questionnaire

The questionnaire (Mano 1988) contains twenty four emotion-describing adjectives designed to assess the intensity of each of the eight regions in the Pleasantness-Arousal circumplex (Figure 1). The items contributing to each scale are: 1) Aroused: Aroused, Astonished, Surprised; 2) Elated: Elated, Active, Excited; 3) Pleased: Pleased, Satisfied, Happy; 4) Calm: Calm, At rest, Relaxed; 5) Quiet: Quiet, Still, Quiescent; 6) Bored: Sleepy, Sluggish, Drowsy; 7) Unpleasant: Unhappy, Sad, Blue; and 8) Distressed: Anxious, Fearful, Nervous. The items are presented in random order on 1-point scales (1 "not at all" to 11 "very much"). In Mano (1988), the reliabilities of the scales ranged from .64 to .91 (average .76); in Lewinsohn and Mano (1989), test-retest reliabilities resulted in correlations ranging from .67 to .95 (average .80). The questionnaire's construct validity has been examined by multidimensional scaling and factor analyses which revealed that it adequately describes the mood circumplex (Lewinsohn and Mano 1989; Mano 1988).

Decision making task

Subjects made choices from two information display boards. The information board is a matrix with choice alternatives, e.g., product brands, as rows, and attributes as columns; each entry contains the value of an attribute for a particular brand. The entries are first covered and the subject is asked to choose a brand after revealing as many information entries as desired, one at a time. An entry remained visible until the next one was revealed. The boards were 6X5, i.e., 6 alternatives described on 5 attributes. The product employed was soft-sided luggage described on the following attributes: Volume (A), Handle control (B), Workmanship (C), Impact resistance (D) and Weight (E). Attribute values ranged from 1 (lowest) to 5 (best). The different pieces of luggage were described as equally priced and had to be judged only with respect to the five attributes. The decision making tasks were performed on a personal computer (IBM-PC). All information was administered, monitored, and recorded by the PC and subjects directed their search and made choices with the PC keyboard.

Dependent Variables

The information board methodology allows for the assessment of a number decision related variables. These include: amount of information searched, i.e., total number of revealed entries; decision time; and processing speed, i.e. average time spent per entry. The decision process can be examined in terms of search pattern; each move to a new entry within the same alternative is classified as interdimensional; a move to a different alternative but in the same dimension is classified as intradimensional. The number of times entries were revealed more than once was classified as number of repetitions; the number of attributes that the subject did not reveal any entry for a particular attribute was defined as number of ignored attributes. Cognitive effort and information load play an important mediating role in Isen's (Isen et al. 1987) and Lewinsohn and Mano's (1989) explanations of the influence of affect on decision making. Thus, in addition to the above decision variables that can also be used to assess indirectly effort, a complementary view of subjects' cognitive exerted effort was elicited by self reports. Subjects rated cognitive effort on three eleven-point scales: the amount of effort exerted to perform the choice tasks (Effort 1); how hard they had to push themselves to keep going (Effort 2); and, how much mental energy was required while working on the tasks (Effort 3).

RESULTS

Affect Measurement

The four sets of advertisements had no differential impact on subjects' emotional states; ANOVAs conducted on the eight scales did not reveal any differences between the four groups. Although unexpected, this lack of change can be explained--albeit post hoc--by a number of reasons. First, exposure to one or two short (30 seconds) TV ads may not constitute a strong enough emotional stimulus capable of generating an impact that lasts more than a few additional seconds (cf. Nowlis 1965). Second, the contrast between exposure to the ads and the mood questionnaire--both embedded in the context of a choice task--may have removed any of the ads' emotional impact. Third, a primacy effect could have taken place and only the first ad (with scant emotional content) may have affected the subjects' emotional state. To assess further the ads' emotional impact, the eight scales were compared with those in the study by Lewinsohn and Mano (1989) which examined naturally occurring moods in a similar experimental situation but did not involve mood induction. The comparisons revealed that the two populations differed on only one of the eight scales: subjects in the present study were less bored than subjects in Lewinsohn and Mano; M = 3.13 vs. M = 4.63, t(62)=3.0, p < .01.

Given the lack of major differences between the four groups in their initial emotional states, the homogeneity of the emotions observed here and in Lewinsohn and Mano's (1989) noninduced moods, and the structural convergence of the eight scales (see next section), the four groups were pooled. This aggregation does not affect subsequent analyses that will examine the impact of emotions on decision making. In fact, the reasons that led to the pooling of the four groups are based on one of the notions presented in the introduction; namely, when it is assumed that some manipulation induced an emotional state, a careful and detailed examination must be carried out to find out whether that particular mood was indeed induced.

Changes in Affect and the Structure of Emotions

The intercorrelations of the eight affective scales between the pre- and post-decision task elicitations were: Arousal .79, Elation .86, Pleasantness 0.71, Calmness .60, Quiet .83, Boredom .72, Sadness .91, Distress .66 (all p's <.01). Moreover, comparisons between the pre and post-decision affect scales revealed one significant difference (the degree of Elation decreased following the decision task; t(31)=2.36, p=.02). Thus the experimental task had only a minor effect on subjects' emotional state.

The joint structure of the pre-and post-decision emotional scales was examined by multidimensional scaling (MDS). Before presenting the results, an overview of the method is offered. In MDS of a correlation matrix, each variable is represented as a point in a geometric space whereby a higher correlation between two variables results in the points being closer in the space. To decide on the appropriate number of dimensions, two criteria are employed in conjunction: (1) an "elbow" in stress-, I.e., a diminishing decrease in the levels of stress (or diminishing increase in explained variance), and (2) solution interpretability. The stress coefficients and percent of explained variance for the one to three-dimensional solutions were, respectively, .435 (43%), .11 (92%), .08 (96%). The two dimensional solution met both the stress reduction and interpretability criteria and appears in Figure 2. Visual inspection suggests a high degree of resemblance between the hypothesized circumplex of emotions (Figure 1) and the obtained map.

Decision Making: Effects of Good Mood and Arousal

As a first step in examining the influence of emotions on decision making, subjects were classified according to their intensities in Good Mood and Activation. To that goal, first, the scales that conceptually contribute to these two states were added; i.e.,

Good Mood = Elation + Pleasantness + Calmness,

Activation = Distress + Arousal + Elation

In order to enhance the reliability of the two scales, one of the nine items was removed from each scale, resulting in alphas of .73 and .76 respectively. Then, using median splits, the population was divided into four groups for a 2X2 ANOVA with the two tasks serving as a replication (Table 2).

The results indicate that, despite the loss of power caused by the median split, subjects who were in a better mood worked longer (p<.05) and revealed more entries (p<.05). In terms of within-trials shifts, subjects processed information faster during the second task (p<.01); for processing speed there was also a significant interaction between Good Mood and Activation. Activation also had a significant impact on the three reports of cognitive effort. No other effects were statistically significant.

Overall, the above data replicate Lewinsohn and Mano (1989) and stand in contrast with those of Isen and Means (1983). Even though many effects failed to reach statistical significance, it is nonetheless clear that subjects in better mood made more deliberate considerations of the examined alternatives.

Correlational Analysis

While the preceding analyses indicate the joint effects of Good Mood and Activation on decision variables, they do not fully capture the multifaceted nature of affect or the intervality of the affect scales. A more detailed description of these relationships is suggested by their intercorrelations and MDS description. For these analyses, the means of the decision variables across the first and second choice tasks were used (the (intercorrelations of the decision variables between the two choice tasks were: Decision Time .55, Revealed Entries .57, Seconds per Entry .54, Inter- Moves .76, IntraMoves 0.71, Ignored Attributes .68, and Repetitions .76; all p's<.001). The pre and post-decision measures were also added. The correlations between the decision variables and the 8 affect scales are reported in Table 3.

The joint MDS of affect and decision variables revealed the following stress coefficients for the one to four-dimensional solutions: .474 (31%), 0.259 (58%), .157 (77%) and .120 (86%). The three dimensional solution met both the elbow and interpretability criteria (Figure 3). Visual inspection reveals that the Good-Mood related states (Elation, Pleasantness, and Calmness) were relatively close to one another, and, as a whole, they were close to decision time, number of revealed entries, interdimensional moves and repetitions. Projections on space 2-3 capture the circumplex of affect which is vertical to the two-dimensional space of the decision variables. This spatial analysis provides an alternative and complementary view on the previous ANOVAs. More importantly, however, it offers an interesting perspective for examining affect's construct validity, and the relationships between affect and decision making.

DISCUSSION

One of the original hypotheses was that subjects in a better mood would used more elaborate and time consuming decision strategies (Lewinsohn and Mano 1989). The results obtained here support that hypothesis and stand in contrast with Isen and Means' (1983) results concerning the influence of pleasant mood on decision making. The explanation offered here was that the decision maker chooses a strategy that is congruent with one's current emotion and is associated in memory with a similar to the currently experienced emotional tone (Bower 1981; Isen 1987). In the context of decision strategy selection, it is speculated that pleased subjects tended to employ a more elaborate and time consuming decision strategy because they framed the decision task as "cognitive play". Less leased subjects, on the other hand, framed the same decision situation as a necessity that they had to get rid off. Given the subjects' playful state of mind, their goal was to enjoy the situation and satisfy their curiosity. As a consequence, they adopted a more "creative" approach and exposed themselves to more information.

The results presented here provide only partial support for the hypothesis that level of arousal would influence the attention allocation mechanism which, in turn, would effect decision strategy selection. Nonetheless, the direction of the results does not contradict that hypothesis. An explanation for the lack of strong results for arousal could be related to the possible confusion generated by the experimental manipulation, i.e., the presence of advertisements without a program context and within a decision situation.

These results suggest that Pleasantness and Arousal-do not have an equally strong influence on the decision making process and that Pleasantness may have a more stable and longer term effect than Arousal (Lewinsohn and Mano 1989). Furthermore, Pleasantness may set the quality, direction, and to some extent, the intensity of the chosen cognitive strategy while Arousal allows slack cognitive effort to be allocated to the selected strategy.

Lewinsohn and Mano (1989) and the present study suggested a number of explanations for the seemingly contradictory findings with Isen and Means (1983). One aspect that seems to underlie these contrasting results may be the nature of the employed mood constructs. For example, there could be differences in the intensity of the hedonic tone and level of activation; i.e., due to the manipulation (informing subjects that they performed at the 97th percentile on a perceptual motor task) subjects in Isen and Means'(1983) positive affect group may have experienced higher levels of pleasantness and elation than subjects in the present study. These higher intensities could draw more attention to the experienced emotions and thus decrease the slack of attentional resources required for elaborate information processing.

FIGURE 2

THE CIRCUMPLEX OF EMOTIONS (PRE AND POST DECISION TASK)

TABLE 2

MEANS FOR DECISION VARIABLES BY GOOD MOOD, AROUSAL AND WITHIN TASKS

TABLE 3

INTERCORRELATIONS BETWEEN DECISION AND AFFECT VARIABLES

FIGURE 3

THE THREE DIMENSIONAL SPACE OF AFFECT AND DECISION VARIABLES

Similarly, Isen and Means' experimental induction could have incorporated a motivational component which may have generated a "competition" framing leading to the use of more "efficient" decision strategies. Furthermore, procedural differences between the choice tasks (e.g., information-board sizes, computerized versus card-based information boards) might have also contributed to some of the contrasting results. Finally, because the mood manipulation was not successful, an alternative interpretation is that the present study examined the impact of some enduring personality variable rather than a temporary mood manipulation (cf. Diener and Larsen 1984). Note, however, that whether affective responses are due to transient states or persistent individual traits, the suggested mechanisms mediating affect and decision making are still plausible.

The effect of emotions on decision making is powerful. The evidence that has been accumulated over the course of the last few years suggests that these interrelationships cannot be ignored. The present paper started from the premise that affect is not unidimensional; by examining in greater detail the decision maker's emotional state we might be able to understand better how and why affect influences decision making. Clearly, additional research is needed to clarify and further elaborate these issues.

REFERENCES

Aaker David A., Douglas M. Stayman, and Michael R. Hagerty (1986). "Warmth in advertising: Measurement, impact and sequence effects," Journal of Consumer Research, 12 (March), 365381.

Batra, Rajeev and Morris B. Holbrook (1988), "Developing a typology of affective responses to advertising: A test of validity and reliability," Research Working Paper No. 88-AV-6, Columbia Business School.

Batra, Rajeev and Michael L. Ray (1986), "Affective response mediating acceptance of advertising," Journal of Consumer Research, 13 (September), 234-249.

Bower, Gordon H. (1981), "Mood and memory," American Psychologist, 36, 129-148.

Clark, Margaret S., Sandra Milberg and Robert Erber (1984). Effects of arousal on judgments of other's emotions," Journal of Personality and Social Psychology, 46, 551-560.

Diener, Ed and Randy J. Larsen (1984), 'Temporal stability and cross-situational consistency of positive and negative affect," Journal of Personality and Social Psychology, 47, 871-883.

Edell, Julie. A. and Marian C. Burke, (1987), "The power of feelings in understanding advertising effects", Journal of Consumer Research, 14 (December), 421 433.

Gardner, Meryl P. (1985), "Mood states in consumer behavior: A critical Review," Journal of Consumer Research, 12 (December), 281-300.

Gorn, Gerald J. (1982), "The effects of music in advertising on choice behavior: A classical conditioning approach" Journal of Marketing, 46, 94-101.

Havlena, William J. and Morris B. Holbrook, (1986), "The varieties of consumption experience: Comparing two typologies of emotions in consumer behavior," Journal of Consumer Research, 13 (December), 394-404.

Holbrook, Morris B. and Rajeev Batra (1987), "Assessing the role of emotions as mediators of consumer responses to advertising," Journal of Consumer Research, 14 (December), 404-420.

Isen, Alice M. (1987), "Positive affect, cognitive processes, and social behavior," in Advances in Experimental Social Psychology, Vol. 20, ed. Leonard Berkowitz, 203-253.

Isen, Alice M., Kimberly A. Daubman and Gary P. Nowicky (1987), "Positive affect facilitates creative problem solving," Journal of Personality and Social Psychology, 52, 1122-1131.

Isen Alice M. and Barbara Means (1983), "The influence of positive affect on decision-making strategy", Social Cognition, 2 (1), 18-31.

Kahneman Daniel (1973), "Attention and Effort," Englewood Cliffs, N.J. Prentice Hall.

Lewinsohn, Shai and Haim Mano (1989), "The influence of naturally-occurring moods on decision making processes," Washington University, St. Louis, Manuscript submitted for publication.

Mano, Haim (1988). Assessing the Structure and Intensity of Noninduced and Induced Affect. Washington University, St. Louis, Manuscript submitted for publication.

Nowlis, Vincent (1965), "Research with the mood adjective check list," in Affect, cognition and personality: Empirical studies, eds. Silvan S. Tomkins and Caroll E. Izard, New York: Springer Publishing Company, 352-389.

Russell, James A. (1980), "A circumplex model of affect", Journal of Personality and Social Psychology, 36, 1152- 1168.

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