Beruflich Dokumente
Kultur Dokumente
Catherine A. Cole
Marketing Department
University of Iowa
How and when positive and negative moods affect attitudes, risk perceptions, and choice is a
problem that interests both consumer researchers and practitioners. We propose that the extent
of constructive processing moderates mood effects with stronger effects when constructive processing is higher. In addition, we propose that when consumers have unrestricted versus constrained processing resources, moods are more likely to operate through affect priming and less
likely to operate through the affect-as-information process. The results from 3 experiments
support these hypotheses. We discuss implications of the findings for models of how affect influences judgments and directions for future research.
FEDORIKHIN
EXPERIMENT 1
Overview
To test HI and H2, in this 2 (mood: positive or negative) x 2
(constructive processing: high or low) between-subject experiment, we assessed mood effects on consumer reactions to
a commercial for a new brand of cookies under two constructive processing conditions. We randomly assigned one of the
four mood-task conditions to groups of respondents drawn
from a large participant pool at a major Midwestern university. One hundred thirty-three respondents participated.
Procedure
Initially, respondents learned that they would participate in
two studies: one about advertising and the other about scale
development. To induce respondents to form initial judgments about the target product prior to the mood manipulation, respondents in the low-constructive processing condition did the following: viewed the target commercial for an
unfamiliar brand of cookies, evaluated the cookies on filler
scales, viewed a mood-inducing video segment, completed
the mood scales, viewed a filler commercial, evaluated
cookies on a questionnaire that included new items designed to measure attitudes toward the brand (As,), perceived likelihood of negative outcomes, perceived disutility
of those negative outcomes, and finally, evaluated the filler
product.4 To induce respondents to form their initial evaluations of the cookies after the mood manipulation, respondents in the high-constructive processing condition did the
following: viewed the filler commercial, evaluated the filler
product, viewed the mood-inducing video, completed mood
scales, watched the cookie commercial, and finally evaluated the cookies on the same cookie evaluation questionnaire as the low-construction group. Consistent with others,
to disguise our mood manipulation as a separate scale development study, a different experimenter showed respondents the 4-min mood-inducing movie episode and administered the measures.
Stimulus Development
To manipulate moods, we selected 4-min episodes from two
movies, including Tommy Boy, a comedy about a recent college graduate, for the positive mood manipulation, and The
4The first 42 participants completed paper-and-pencil questionnaires,
but the remaining respondents completed the "Advertising Study" on the
computer. Before pooling the two groups, we tested whether there were differences between the two groups in means and variances for each dependent
variable. We found no significant differences in variances due to mode of administration and no main effects or interactions of mode with other independent variables. Hence, we concluded that mode did not impact the results
of the analyses, and it is not discussed further.
perceived disutility for the participants was their average response to the six disutility items.
Results
Out of the 133 respondents, we excluded 9 questionnaires
from the analyses because the participants were able to connect two parts of the study. Table 1 contains the means and
standard errors for the dependent variables.
Manipulation checks. After the respondents saw the
video episodes, we measured their moods using the PANAS
scale. Positive affect was significantly higher and negative
affect was significantly lower for the positive mood condition
than for the negative mood condition; positive affect: coefficient a = 39, 3.1 versus 2.4; F(1, 114) = 30.6, p < .0l; negative affect: coefficient a = .87, 1.9 versus 1.5; F(1, 113 =
1 0 . 7 ,<
~ .01. We checked whether we manipulated constructive processing in two ways: by analyzing responses to a
two-item self-reported construction effort (SRCE) scale and
unobtrusively measuring response time for participants who
completed the questionnaire on the computer. The SRCE
scale items asked respondents to indicate the extent to which
they agreed with the following two statements: The preceding questions about the cookies really made me think about
them; I really thought a lot about whether I liked the cookies
or not (the two items correlated .71). The mean on this scale
was significantly higher in the high- than in the low-constructive processing condition, 4.9 versus 3.5; F(1, 123) =
27.5, p < .01. Also, those in the high-constructive processing
condition spent more time completing the dependent variable
scales than those in the low-constructive processing condi~ .01.
tion, 128.6 sec versus 110.0 sec, F(1, 97) = 8 . 1 , <
TABLE 1
Means and Standard Errors for Experiments 1 and 2
Positive Mood
Negative Mood
Positive Mood
Negative Mood
4.6 (0.23)
3.7 (0.22)
4.3 (0.21)
3.8 (0.19)
5.5 (0.19)
3.1 (0.23)
3.9 (0.24)
4.2 (0.28)
4.7 (0.29)
3.2 (0.22)
52%
4.7(0.32)
3.1(0.24)
67%
5.4 (0.25)
2.6 (0.24)
71%
4.0 (0.27)
4.0 (0.33)
32%
FEDORIKHIN
EXPERIMENT 2
Hypotheses
In addition to testing H1 and H2, we develop and test H3,
which relates mood and constructive processing to consumer choice. Consumers in positive moods, when compared to those in negative moods, will be more likely to
choose a new product over a familiar one. Compared to
those in a negative mood, consumers in a positive mood not
only perceive lower risk in, and develop more favorable attitudes toward, new products, but they also seek variety
when selecting brands among safe, enjoyable products
(Isen, 2001; Kahn & Isen, 1993). As established in Experiment 1, however, consumers' tendency to favor the new
brand when under the influence of a positive mood will not
5Toconfirm the manipulation of old and new cookies, in a posttest, participants evaluated the actual cookies on a two-item familiarity scale. They
rated the new cookies as less familiar than the old cookies. In Study 2, the
difference in familiarity between the two cookies equaled 2.5, t(17) = 5.3, p
< .05. In Study 3, the difference between the twocookies equaled 3.9, t(17) =
8.4, p < .05.
sampled earlier. At the end of the study, participants answered questions about the study's goals.
Results
We dropped 4 out of 91 questionnaires from the analyses because these participants were able to link the two parts of the
experiment. Means and standard errors for the dependent
variables are in Table 1.
Manipulation checks. To determine whether the mood
manipulation worked, we used the same PANAS scales as in
Experiment 1: positive affect: coefficient a = 39; negative
affect: coefficient a = 22. As expected, respondents reported
higher positive affect, 3.0 versus 2.3, F(l,84) = 22.9, p < .01,
and lower negative affect, 1.7 versus 1.3, F(l,84) = 12.7,p <
.01, in the positive than in the negative mood condition.
To determine whether the task manipulated constructive
processing, we assessed both responses to a three-item SRCE
scale and time needed to complete the dependent variable
scales. The three-item SRCE scale included the two items
from Experiment 1. As expected, both the average on the
SRCE scale (coefficient alpha was .74) and the time spent
completing the scales were significantly higher for the highthan for the low-constructive processing conditions: SCRCE,
5.7 versus 5.1; F(1,82) = 7.6, p < .01; time, 154.5 sec versus
~ .01.
124.2 sec, F(1, 82) = 1 4 . 2 ; <
Hypotheses testing. In the ANOVA analysis with A B ~
(coefficient a = .95) as a dependent variable, the mood effect
and the mood-by-task interaction were significant: mood,
~ .03; interaction: F(1, 80) = 5 . 3 5 , ~< .03.
F(1, 80) = 5 . 2 , <
Consistent with HI, follow up t tests indicated that there was
a significant difference in ABrbetween positive and negative
mood conditions in the high-, 5.4 versus 4.0, t(39) = 3.17, p <
.01, but not in the low-constructive processing condition, 4.7
versus 4.7, t(45) = .Ol , ns.
For the ANOVA with perceived likelihood of losses as a
dependent variable, the mood effect and the mood-by-task
interaction were significant: coefficient a = .68, mood: F(1,
~ .04, interaction: F(l,80) = 6 . 6 , <~ .02. Consis80) = 4 . 6 , <
tent with H2, follow-up t tests showed a significant difference in perceived likelihood of losses between positive and
negative mood conditions in the high-, 2.6 versus 4.0, t(39) =
3.27, p < .01, but not in the low-constructive processing conditions, 3.2 versus 3.0, t(45) = .8, ns.
To test H3 about the effect of mood and constructive processing on choice, we ran a logistic regression with mood,
task, and the mood-by-task interaction as the independent
variables and choice as the dependent variable. This analysis
showed nonsignificant mood and task effects but a significant mood-by-task interaction ( ~ =2 4.24, p < .04). We followed this up with chi-square tests in the high-constructive
processing condition, which showed that as respondents
EXPERIMENT 3
Hypotheses
We propose that the availability of processing resources influences which mechanisms underlie mood effects on attitudes. When motivated to construct opinions, consumers
with unrestricted processing resources will use a relatively
slow piecemeal process to systematically analyze the different attributes of the object and to form an evaluation (Pham et
al., 2001; Sanbonmatsu & Fazio, 1990; Sujan, 1985). The affect-priming process described earlier can color these attribute-oriented thoughts by selectively priming mood-congruent memory nodes and by directing attention to
mood-congruent details (Fiedler, 1991). Consistent with affect priming, Petty et al. (1993) showed that the valence of
cognitive responses to a persuasive message mediates positive mood effects on attitudes in high-elaboration conditions.
When consumers face constrained processing resources,
but are motivated to construct evaluations, their moods may
still bias these evaluations (Petty et al., 1993). However, when
consumers have few processing resources, they tend to form
evaluationsby using a categorizationprocess instead of attribute analysis (Brewer, 1988; Sanbonmatsu & Fazio, 1990;
Sujan, 1985).By classifying anew stimulus into afamiliarcategory, consumers can quickly retrieve and apply the affect associated with the category to the new stimulus. The affect-as-information mechanism described earlier may drive
these mood effects because people may confuse their mood
with the affective tag for the category through a "how do I feel
about it?'heuristic (Petty et al., 1993;Pham, 1998;Schwarz &
Clore, 1983).In summary, when compared to consumers with
unrestricted processing resources, these consumers will be
more likely to report relying on their feelings to make judgments and will have thought protocols that contain more references to categorization and fewer references to attribute-oriented thoughts (Sujan, 1985). As a result, we predict that
H4(a): Under constrained processing resources, consumers will generate more categorization-oriented and
fewer attribute-oriented thoughts and will indicate
higher reliance on feelings in forming judgments
than consumers under unrestricted processing resources.
Accordingly, we expect that different cognitive variables
will mediate mood effects under different resource conditions. Under constrained processing resources, the valence of
categorization thoughts will mediate the mood-attitude relation. Under unrestricted processing resources, the valence of
attribute-orientedthoughts will mediate mood effects on attitudes. This reasoning leads to the following hypotheses
about differences in the mechanisms underlying mood effects under different levels of resource availability:
H4(b): In high-constructive processing conditions, under
unrestricted resources, the valence of attribute-oriented thoughts will be more likely to mediate the
effects of moods on attitude than the valence of
categorization thoughts.
H4(c): In high-constructive processing conditions, under
constrained processing resources, the valence of
consumers' categorization thoughts will be more
likely to mediate mood effects on attitudes than
the valence of attribute-oriented thoughts.
Overview
To test hypotheses H1 through H4, this 2 (mood) x 2 (constructive processing) x 2 (levels of processing resources) between-subject experiment exposed consumers to one of two
different mood-inducing movie episodes, manipulated the
task to induce either high or low levels of constructive processing, and manipulated availability of processing resources
by putting respondents under different levels of cognitive
load. We collected the same dependent variables as in the
TABLE 2
Means and Standard Errors for Experiment3
Unrestricted Processing Resources
Low Construction
Dependant Variable
AB~
L1
Choice
Nattr
Ncat
Feelings
High Construction
Low Construction
High Construction
Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative
5.33(.17)
2.04 (.26)
61%
2.5 (.34)
.67 (. 18)
31%
5.29(.24)
1.6 (.I81
42%
2.2 (.34)
.37 (.13)
11%
5.8(.24)
1.4 ( 7
66%
3.4 (.45)
.61 (.24)
17%
4.3(.40)
2.9 (.52)
25%
3.8 (.44)
.92 (.23)
50%
5.2 (2.1)
2.1 (.28)
52%
1.8 (.44)
.94 (.24)
47%
4.8 (.32)
2.6 (.29)
44%
2.5 (.45)
.65 (.21)
53%
5.37 (.22)
1.56 (.27)
39%
1.5 (.27)
1.I (.22)
57%
4.36 (.35)
2.6 (.38)
29%
2.4 (.39)
1.2 (.21)
64%
--
Note. Ae, = attitude toward the brand; L1= loss likelihood; Choice = % of consumers choosing the test brand; Nattr is the number of attribute-oriented (categorization) thoughts in the protocol; Ncat is the number of categorization thoughts; Feelings = % of respondents mentioning feelings in relationship to use of the
product.
high-constructive processing condition, we found that as respondents moved from negative to positive moods, choice of
the new cookie increased significantly (27% vs. 55%; ~2 =
4.52, p < .04), whereas in the low-constructive processing
condition, mood did not significantly affect choice (43% vs.
57%; = 1.39, ns).
To test H4, two judges independently coded each participant's thought protocol. They classified the valence of each
thought and whether each thought was attribute oriented, categorization related, or other (see Sujan, 1985). Attribute-oriented thoughts were statements related to the evaluation of an
attribute or comparisons of attributes; categorization
thoughts noted how the test cookie was similar to an overall
category (such as cookies, food one takes to a potluck, snack
food).6 In addition, they judged whether respondents made at
least one statement connecting feelings to the use of the new
cookies (e.g., I would be embarrassed to take the new cookies
to a potluck).
Consistent with H4(a) ANOVA analyses on the number of
attribute-oriented and categorization thoughts and chi-square
analysis on the proportion of respondents using feelings-oriented thoughts showed significant effects of cognitive load.
Consumers generated more attribute-oriented and fewer categorization thoughts under unrestricted than under constrained resource conditions (attribute-oriented thoughts: 2.9
versus 2.1, F(1, 126) = 8.0, p < .01; categorization thoughts:
1.0 versus .6, F(1, 127) = 5.1, p < .03. In addition, a larger
proportion of respondents in the constrained (54%) than in
the unrestricted processing resources condition (25%) made
statements connecting their feelings to the use of the cookies
( ~ =2 1 0 . 6 , <
~ .01). We also analyzed participants' responses
to a question: "When evaluating the cookies I relied mostly
on my ... " anchored by feelings or thoughts. Again, consistent with H4(a), those in the unrestricted resources condition
indicated higher reliance on thoughts than those in the constrained resources conditions: 4.77 versus 4.09 accordingly,
with higher numbers indicating higher relative reliance on
thoughts, F(1, 126) = 4.37, p < .04.
To test H4(b), we did a mediation analysis in the unrestricted resources, high-constructive processing condition,
to test whether the proportion of positive attribute-oriented
thoughts (PPAT) mediated the effects of mood on AB~.
(PPAT is the number of positive attribute-oriented thoughts
divided by the total number of attribute-oriented thoughts).
We found evidence of strong mediation. Mood had a significant effect on ABr: mood, F(1, 31) = 10.4, p < .01. In another ANOVA, mood had a significant effect on PPAT, F(1,
27) = 19.4, p < .01. In addition, when PPAT was added to
the first ANOVA on ABn PPAT achieved statistical signifi-
x2
6Examples include positive attribute thoughts (I like the chocolate), negative attribute thoughts (The inside cookies part is too dry), positive category
thoughts (These cookies would be great snacks), negative category thoughts
(I wouldn't take those cookies to a potluck).
10
FEDORIKHIN
GENERAL DISCUSSION
Understanding how moods influence consumer behavior is
important because consumers bring a variety of moods into
decision-making and because the shopping experience itself
can induce different moods. From a theoretical perspective,
these three experiments provide several new insights. Our
work highlights constructive processing as an important
moderator of mood effects on attitude, perceived risk, and
choice in both high- and low-elaboration conditions. Although dual-process models such as the Elaboration Likelihood Model (ELM) do not explicitly incorporate constructive processing, they do allow a variety of peripheral and
central cues to compete with and diminish mood effects in
both high- and low-elaboration conditions (Petty & Wegener,
1999). By studying cognitive markers in high- and low-processing resources conditions, we show that the type of
cognitions that mediate mood effects on attitudes depends on
the cognitive resources that consumers have available.
Through mediation analysis and coding of thought protocols,
we find evidence that moods bias attitudes through an affect-as-information process when consumers have limited
processing resources, but when consumers have unrestricted
processing resources, we find evidence that consumers'
moods bias cognitions in a manner consistent with the affect-priming process.
ACKNOWLEDGMENTS
This article is based on Alexander Fedorikhin's Ph.D. dissertation at the University of Iowa under the guidance of
REFERENCES
Adaval, R. (2001). Sometimes it just feels right: The differential weighting
of affect-consistent and affect-inconsistent product information. Journal
of Consumer Research, 28, 1-17.
Adiga, A,, Birger, J., Carter, A., & Feldman, A. (2001, November). Sudden
impact. Money, 30, 84-97.
Barone, M. J., Miniard, P. W., & Romeo, J. B. (2000). The influence of positive mood on brand extension evaluations. Journal of Consumer Research, 26, 386400.
Bettman, J. R., Luce, M. F., & Payne, J. (1998). Constructive consumer
choice processes. Journal of Consumer Research, 25, 187-217.
Bower, G. H. (1981). Mood and memory. American Psychologist, 36,
129-148.
Brewer, M. B. (1988). A dual process model of impression formation. In T.
K. Srull & R. S. Wyer (Eds.), A dual process model of impressionformation (pp. 1-36). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Coupey, E., Irwin, J., & Payne, J. (1998). Product category familiarity and
preference construction. Journal of Consumer Research, 24, 459468.
Dowling, G. R. (1986). Perceived risk: The concept and its measurement.
Psychology and Marketing, 3, 193-21 0.
Dowling, G. R., & Staelin, R. (1994). A model of perceivedrisk and intended
risk-handling activity. Journal Of Consumer Research, 21, 119-135.
Dunn, M. G., Murphy, P. E., & Skelly, G. U. (1986). Research note: The influence of perceived risk on brand preference for supermarket products.
Journal of Retailing, 62, 204-2 16.
Estrada, C. A., Isen, A. M., &Young, M. J. (1998). Positive affect facilitates
integration of information and decreases anchoring and reasoning among
physicians. Organizational Behavior and Human Decision Processes, 72,
117-135.
Fiedler, K. (1990). Mood-dependent selectivity in social cognition. In W.
Stroebe & M. Hewstone (Eds.), European review of social psychology
(Vol.1, pp. 1-32). New York: Wiley.
Fiedler, K. (1991). On the task, the measures, and the mood in research on
affect and social cognition. In J. P. Forgas (Ed.), Emotion andsocialjudgments (pp. 83-104). Elmsford, NY: Pergamon.
Folkes, V. S. (1988). The availability heuristic and perceived risk. Journal of
Consumer Research, 15, 13-24.
Forgas, J. (1995). Mood and judgment: The affect infusion model (AIM).
Psychological Bulletin, 117(1), 39-66.
Forgas, J. (1999). Feeling and speaking: Mood effects on verbal communication strategies. Personality and Socictl Psychology Bulletin, 25,
850-863.
Gardner, M. P. (1985). Mood states and consumer behavior: A critical review. Journal of Consumer Research, 2, 281-285.
Gay, V. (2001, September). When bad gets worse. Brandweek, 42,
SR3-SR4.
Gorn, G. J., Goldberg, M. E., & Basu, K. (1993). Mood, awareness and product evaluation. Journal of Consumer Psychology, 2, 237-256.
Gorn, G., Pham, M. T., &Sin, L. Y. (2001). When arousal influences ad evaluation and valence does not (and vice versa). Journal of Consumer Psychology, 11,43-57.
11
12
FEDORIKHIN