|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
BEHAVIORAL CONSEQUENCES OF AFFECT: COMBINING EVALUATIVE AND
EDUARDO BITTENCOURT ANDRADE
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
Eduardo Bittencourt Andrade
This dissertation is dedicated to my parents.
First of all, I would like to thank Joel Cohen, who supervised this dissertation with
an incredible sense of balance between academic guidance and intellectual freedom. I
also extend thanks to Chris Janiszewski, Richard Lutz, and Margaret Bradley, for their
support. I thank also Alan d'Astous (from HEC Montreal), who introduced me to the
Marketing Department at the University of Florida. The Warrington College of Business
Administration provided me with financial assistance in completing this dissertation, for
which I'm grateful. Even far away from Gainesville, my family and close friends
managed to provide me the emotional support needed to keep the discipline and resolve
throughout these long years. I thank them all.
TABLE OF CONTENTS
A C K N O W L E D G M E N T S ................................................................................................. iv
L IST O F T A B L E S ......................... ........ ...................... .. .. .. ...... ............. .. vii
LIST O F FIG U RE S .......................................................................... viii
ABSTRACT .............. .......................................... ix
1 IN TR OD U CTION ............................................... .. ......................... ..
A ffe c t ................................................................................. 3
M ediational Processes ........................................ .......................... .4
2 THEORETICAL BACKGROUND .................. ...... ...........................
2 THEORETICAL BACKGROUND.......................................7
Affective Evaluation (Affect's Informational Role)................. ............................8
Affect-Regulation (Affect's Motivational Role) ...................................................9
3 INTEGRATIVE MODEL OF AFFECTIVE BEHAVIOR (IMAB)..........................12
4 INITIAL EVIDENCE SUPPORTING THE MODEL: THE IMPACT OF AFFECT
ON HELPING, RISK TAKING, AND EATING BEHAVIOR..............................20
H elp in g ............................................................... ................ 2 0
Positive A effect and H helping ........................................ .......................... 21
N negative A effect and H elping..................................... ......................... .. ......... 22
Theoretical Integration ............................................... ............................. 25
R isk -T making ................................................... .............................. 2 6
Negative Affect and Risk-Taking...................... .... ......................... 27
Positive A effect and Risk-Taking ........................................ ...... ............... 30
E ating B eh av ior ................................................................... ................ ... 3 1
N negative Affect and Eating Behavior............................................................... 32
Positive Affect and Eating Behavior ..........................................................34
S u m m ary ...................................... .................................................. 3 6
5 TESTIN G IM AB : PRED ICTION S .................................. ..........................................39
6 EXPERIM ENT 1 .................................... ..... .......... ....... ..... 42
M eth o d .................. ...................................................................................4 2
Subjects an d D esign .................................................................................42
Procedure ................ ............................... 42
A effect M manipulation ................................................... ............................... 43
R ep lic ate s ............................................................4 4
Product B benefits M anipulation......................................... ......................... 45
D ep en dent M easu re ......... .........................................................................4 6
Results .............................................. ....... .................... 46
M anipu lation C h eck s........................................... ...........................................4 6
Purchase Intention ............................. .................... .. ........ .. .............47
Discussion ......................... ...................... 49
7 E X PE R IM E N T 2 ......... ......... ...... ....................................................... 51
M eth o d .............. ...................................................................................... 52
Subjects an d D esign .................................................................................52
Procedure ................ ............................... 53
Results .............................................. ....... .................... 54
Manipulation Checks.................... ..............................54
Purchase Intention ........................... ...... .. .. ...... ...............54
Affective Evaluation Mechanism ................... ........... ................. 56
A ffect-R regulation M echanism ........................................ ......................... 58
D discussion ......... ............................................................................................... 59
8 EXPERIM ENT 3 .................................... ..... .......... ....... ..... 61
M ethod .............. .......................................................................................... ......... 62
Subjects an d D esign .................................................................................62
Procedure ................ ............................... 62
Results .............................................. ....... .................... 64
M manipulation C hecks ......... ................ ........... ....... ................... ............... 64
E atin g B eh av io r ............ ................................................................ ......... ......... 6 5
A ffect-R regulation M echanism ........................................ ......................... 66
D discussion ......... ...............................................................................................69
9 GENERAL DISCU SSION .......................................................... .............. 71
L IST O F R E FE R E N C E S ............................................................................. .............. 77
B IO G R A PH IC A L SK E TCH ...................................................................... ..................87
LIST OF TABLES
1. Influence of affective evaluation on behavioral intention .......................................... 57
2. Influence of acknowledged use of chocolate to lift mood on behavioral intention.......58
3. Protective influence of overeating-based guilt feelings on behavior.............................67
LIST OF FIGURES
1. Behavioral consequences of affective states...................................... ....................... 2
2. Predictions for experim ents 1 and 2....................................... ......................... 40
3. Behavioral intentions toward coffee and cereal (collapsed)........................................47
4. Behavioral intentions tow ard chocolate................................... ......................... 55
5. A m ount of chocolate consum ed......................................................... ............... 65
6. Amount of chocolate consumed within the positive affect /mood-threatening condition
the impact of expected feelings of guilt after overeating ............. ...............68
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
BEHAVIORAL CONSEQUENCES OF AFFECT: COMBINING EVALUATIVE AND
Eduardo Bittencourt Andrade
Chair: Joel B. Cohen
Major Department: Marketing
The impact of affect on consumer behavior is well documented; however, no
unique pattern of behavior can be expected from a valenced affective state. While one
can find support for the intuitive congruency-type of hypothesis in which positive mood
facilitates action or increases purchase intention and negative mood inhibits action or
decreases purchase intentions, it is now clear that significant exceptions exist. A negative
mood may well encourage action whereas positive mood can inhibit action. Most
importantly, the models available in the literature have focused primarily on explaining
one or two effects (e.g., when negative affect encourages behavior), leading to a lack of
an overarching theory capable of accounting for the sometimes mitigating and sometimes
encouraging impact of positive and negative affective states on behavior and behavioral
An integrative model of affective behavior (IMAB) is therefore proposed, in which
the informational and regulatory properties (mechanisms) of current and anticipated
affective states simultaneously mediate the impact of affective states on behavior. A
broadly-based review of the affect-behavior literature demonstrates that relying on a
single mechanism produces, in the aggregate, an ambiguous and often conflicting account
of affect's role in guiding behavior. Three research streams well known in the literature
(helping, risk-taking, and eating behavior) are reviewed, and their apparent
inconsistencies are elucidated under the proposed model. Finally, in a series of three
experiments, critical moderating variables associated with each of the two mechanisms
(affective evaluation and affect-regulation) are investigated, providing strong initial
support for the model.
The impact of affect on consumer behavior is well documented (see Bagozzi,
Gopinath, and Nyer 1999; Cohen and Areni 1991 for reviews). However, no unique
pattern of behavior can be expected from a valenced affective state. While one can find
support for the intuitive congruency-type of hypothesis in which positive mood facilitates
action or increases purchase intention, and negative mood inhibits action or decreases
purchase intentions, it is now clear that significant exceptions exist. A negative mood
may well encourage action, whereas positive mood can inhibit action. For example,
positive feelings increase purchase intentions (Brown, Homer, and Inman 1998), but also
decrease risk-taking if the odds are too high (Isen and Geva 1987). Bad moods mitigate
consumers' willingness to go to a movie when they have a hedonic goal in mind (Pham
1998), but also increase impulsive consumption (Tice, Bratslavsky, and Baumeister,
2001). As Bagozzi and colleagues (1999) summarized, "Sometimes emotions spur one
onto action; at other times emotions inhibit or constrain action. But only recently have
researchers devoted much attention to studying how this occurs" (p. 199).
Starting with the assumption that affect-behavior relationships can be, at the
aggregate level, categorized into four groups (Figure 1), it becomes clear that the current
most cited models available have focused primarily on explaining one or two cells of this
2x2 matrix. For instance, mood-congruency and affect-as-information conceptualizations
can explain when a positive (negative) mood facilitates (inhibits) action (Cells 1 and 4),
but cannot account for the impact of negative mood on purchase/consumption increases
(Cell 3). Zillmann's (1988) mood-management theory can explain the latter, suggesting
that people in a bad mood are more likely to act in an attempt to feel better. However, it
does not explain, for instance, why people in a bad mood are sometimes less likely to act,
even if the behavioral activity is not expected to put them in a worse mood (Cell 4).
Advertising and Purchase Intention (Brown
et al. 1998)
Hedonic Goals and Purchase Intention
Mood-Congruency Hypothesis (Isen et al.
1978; Bower 1981)
(Schwarz and Clore 1983)
Mood-Maintenance Hypothesis (Isen 1984,
2000; Clark and Isen 1982)
Impulsive Behavior (Tice et al. 2001)
Music and Purchase Intention (Alpert and
Mood-Management Theory (Zillmann
Negative State Relief Model (Cialdini et al.
Risk-taking with low prob. of winning
(Arkes et al. 1988)
Helping when negative stimuli are salient
(Isen and Simmonds 1978)
Mood-Maintenance Hypothesis (Isen 1984,
2000; Clark and Isen 1982)
Hedonic Contingency Hypothesis
(Wegener and Petty 1994)
Helping Children (Kenrick and Cialdini
Food Intake Men (Macht et al. 2002)
Mood-Congruency Hypothesis (Bower
1981; Isen et al. 1978)
(Schwarz and Clore 1983)
Figure 1. Behavioral consequences of affective states
Our study proposes a parsimonious integrative model with three main goals in
mind. First, it enables us to explain the inhibiting and stimulating consequences of both
positive and negative affective states on behavior, which has not been yet seen in the
literature. Second, based on the assumption that two main interdependent mechanisms
(i.e., affective evaluation and affect-regulation) are under operation, our study identifies
the critical moderating variables that mitigate/stimulate each mechanism and eventually
guide behavior. As the literature has focused more often on the evaluation aspect of
current affective states (Bower 1981; Gorn, Pham, and Sin 2001; Pham 1998; Schwarz
and Clore 1983), special attention is given to the affect-regulation mechanism, its
assumptions, and its internal and environmental contingencies. Finally, the model also
demonstrates when regulation-type hypotheses (e.g., mood-maintenance) seem more or
less appropriate as a theoretical account for the behavioral consequences of affect.
Three major components define the scope of analysis of the model to be advanced:
the independent variable (affect), the mediating processes (affective evaluation and
affect-regulation), and the dependent variable (behavior and behavioral intentions).
Affect is defined as "positively or negatively valenced subjective reactions that a
person experiences at a given point in time" (Wyer, Clore, and Isbell 1999, p. 3). Thus,
affect represents the conceptual umbrella for both mood and emotions. Although
distinctions between mood and emotions vary somewhat, researchers tend to agree that
the source of the affective experience represents a critical distinction. While subjects
experiencing emotions are consciously aware that emotions emanates from some source,
subjects experiencing moods are not. For example, whereas people are in a bad or good
mood, they are angry at someone/something (Schwarz and Clore 1996). Secondly,
emotions have also been categorized in terms of in intensity, duration, and/or cognitive
participation (Ortony, Clore, and Collins 1988). Moods, however, represent a unique
(usually indivisible) positive or negative state. This implies that specific types of
emotions are likely to trigger different sets of behavior, depending on their arousing or
cognitive properties (Raghunathan and Pham 1999). For instance, depressed people
(compared to subjects who are simply in a bad mood) may behave differently when a
mood repair opportunity presents itself, since the chronic properties of the former
probably mitigate the impact of affect-regulation on behavior. Although it is beyond the
scope of our model to predict the impact of specific types of emotions, the proposed
model does recognize the uniqueness of each type of emotion in producing behavioral
We believe it is useful to make the simplifying assumption that affect can
potentially mediate evaluative and behavioral patterns at three different levels of
processing (Cohen and Areni 1991; Pham et al. 2001). At the most basic level (Type I-
Affect), affective information is conveyed via sensory-motor programs critical to bio-
regulation. For example, bodily information is captured by the peripheral nervous system
and sent to the central nervous system, which sends back signals that help regulate organs
and biorhythmic activities. Subjects are usually unaware of these automatic mechanisms.
A mid level of analysis (Type II-Affect) refers to basic affective reactions learned
through conditioning, such as fear responses or alertness triggered by danger
identification. This type of information may follow a "low road," departing from the
thalamus (where sensory information is processed) to the amygdala (responsible for
triggering emotional/fearful responses). As minimal cortical processing takes place,
individuals have no more than a rough representation of the stimulus, but are capable of
reacting fairly quickly to it (LeDoux 1996). Finally, affective information can be
processed at a higher level (Type III-Affect), involving subjective appraisal of the
stimulus. In this case, affective information requires significant participation of the
neocortex, where most of the cognitive functions operate, before any behavioral activity
Though recognizing the importance of all three levels of psychophysiological
processes, our model focuses on the impact of affect on behavior via a deliberate
cognitive process, rather than an automatic affective reaction. The propositions and
empirical evidence to be discussed focus on cognitive processes in which individuals
deliberately use affect as a signal to evaluate the environment around them (affective
evaluation mechanism) and/or to regulate their affective experiences (affect-regulation
Finally, two major factors led to the adoption of behavior and behavioral intentions
as the main dependent measures. First, our primary objective is to understand the impact
of affect on behavior. While there is extensive research on the impact of affect on
judgments/attitudes, the impact of affect on behavior (and behavioral intentions) has
received considerably less attention (Forgas 2002). Moreover, as we will see, the theories
rooted in affective evaluations (Isen et al. 1978; Bower 1981; Schwarz 1990), cannot by
themselves account for the impact of affect on behavior. Second, since compelling
evidence suggests strong correlation between intentions and behavior (Eagly and
Chaiken 1993), and many of the available studies have investigated the impact of affect
on behavioral intentions rather than on final action, both types of dependent measures are
Historically, the relationship between affect and its behavioral consequences has
been examined from three different perspectives, each highlighting one major property of
affect. The first, and probably the oldest tradition, stresses the disrupting properties of
affect on judgment and behavior. Brown and Farber (1951) advocated that hunger
produced an affective impact (frustration) if the goal were not reached. Two drives were
hypothesized to operate in parallel, hunger-the "relevant drive," and frustration-the
"irrelevant drive." People were seen as striving for the relevant goal while avoiding the
negative effects of frustration, the irrelevant drive. Using the same rationale, action
control theory (Kuhl and Beckmann 1984) considered frustration a "competing
tendency" that must be controlled for. Finally, negative emotions have been hypothesized
to disrupt "rational" behavior and promote "self-regulation failure" (Leith and
Baumeister 1996). Affect (typically negative affect) is therefore reduced to a disrupting
psychological mechanism leading to negative cognitive and behavioral outcomes.
A critique of the disruptive perspective begins with the fact that feelings are
essentially treated as maladaptive or irrational (and therefore "to be avoided")
components of human nature. This perspective implies that individuals' goals can best be
met by controlling for internal and environmental "emotional temptations." This
conception overlooks the functional aspects of feelings, and constrains our understanding
of the importance of feelings to judgment and behavior (Carver and Scheier 1996; Fridja
1999; Pham et al. 2001). Although it is clear that under certain circumstances affect does
have disruptive and non-adaptive consequences, feelings are usually indispensable for
optimal decision-making, and .. the absence of emotions and feeling is no less
damaging, no less capable of compromising the rationality that makes us human. ... ."
(Damasio 1994, p. xii).
More recently, multiple theoretical accounts have proposed adaptive-type
explanations for the impact of affect on behavior and behavioral intentions. Mood-
congruency (Bower 1981), affect-as-information (Schwarz 1990; Schwarz and Clore
1983), mood as input (Martin et al. 1993; Martin and Stoner 1996), affect infusion
(Forgas 1995), mood-maintenance (Isen 1984; Isen 2000), mood-management (Zillmann
1988) and hedonic contingency (Wegener and Petty 1994) are among the theories
focusing on understanding the mediating processes through which current and anticipated
affective states influence judgment, intentions, and/or behavior. Although varying in
terms of predictions and scope of analysis, it is possible to categorize them at a more
basic level. Based on the main underlying mechanism that links affect to behavioral
consequences, the theories above can be classified into two groups: affective evaluation
Affective Evaluation (Affect's Informational Role)
The first group emphasizes the evaluative properties of affect. It incorporates
theories in which affect has been assumed to influence evaluative judgments, and
eventually behavior, in either a direct or indirect fashion. The affect-as-information
hypothesis (Schwarz 1990, Schwarz and Clore 1983) and the mood-congruency
hypothesis (Isen et al. 1978; Bower 1981) are probably the most cited and investigated
accounts. 1Affect-as-information proposes that individuals directly assess their current
affective states and deliberately use them during any evaluative process. The mood-
congruency hypothesis proposes an indirect influence of affect, suggesting that affective
states make effectively congruent material more accessible in memory, which leads to
changes in evaluation. Evidence has supported both affect-as-information (Pham 1998;
Pham et al. 2001) and mood-congruency (Goldberg and Gorn 1987; Isen et al. 1978; see
also Forgas 1995).
Most importantly, although the proposed mediating process varies, the affect-as-
information and mood-congruency hypotheses make similar predictions. In general,
positive affect is expected to lead to a more favorable evaluation of the environment,
which stimulates proactive activities (increased purchase intention); whereas negative
affect is expected to lead to a less favorable evaluation of the environment, which inhibits
action (decreased purchase intention). In summary, a general affective evaluation
mechanism (affect-as-information and/or mood-congruency) can explain Cells 1 and 4 of
Figure 1. But it does not explain a negative (positive) mood increasing (decreasing) such
behavioral intentions (Cells 3 and 2).
Affect-Regulation (Affect's Motivational Role)
Theories based on affect-regulation assume that positive affect represents a goal, a
desired state unto itself; and that people spontaneously attempt to achieve this "ideal"
affective state or protect it once the state has been attained. For example, Zillmann's
1 Affective states are also known to influence evaluation through changes in processing style. Positive
mood leads to top-down processes while negative mood triggers bottom-up processes (Martin and Clore
2001), which may have strong influences on evaluation (Barone, Miniard, and Romeo 2000). The present
research does not focus on this particular issue. Thus, a deliberate attempt is made throughout the
experiments (e.g., experimental control over the amount of time people are exposed to product information)
to minimize the impact of processing style on the effects of interest.
mood-management theory (1988) suggests that respondents may be willing to act in
anticipation of the mood-lifting consequences of such behavior (Zillmann 1988). Indeed,
Tice et al. (2001) showed that people deliberately use food consumption as a tactic for
mood repair. Similarly, motivation for upward affect-regulation has also been suggested
as a main motivator for self-gifting (Mick and DeMoss 1990a, 1990b; Luomala and
Laaksonen 1997), pro-social behavior (Bagozzi and Moore 1994), and difficult trade-offs
avoidance (Luce 1998). Therefore, based on the framework proposed in Figure 1, mood-
management has focused mostly on explaining when negative affect encourages action
(Cell 3 of Figure 1).
A regulatory explanation is also used to explain when respondents in a good mood
become conservative (that is, less willing to act). Research on helping, and especially risk
taking, shows that when the situational cues threaten respondents' affective states, those
in a positive mood are less likely to take risks or help others (Cell 2 of Figure 1). As
happy people have more to lose, they become more protective of their current affective
states than those experiencing a more neutral affective state (see Isen 2000 for a review).
Good mood has in fact been shown to lead to a protective type of behavior when the
negative features of the product were made salient (Kahn and Isen 1993), when risks of
losing were too high (Isen and Geva 1987), and when the helping task could produce
negative feelings (Isen and Simmonds 1978). Wegener and Petty's (1994) hedonic
contingency hypothesis provides similar predictions with a slightly different (i.e., event
sampling) rationale. Their theory suggests that since happy people are already in or closer
to the ideal affective state, there are fewer (more) behavioral options in the environment
that can make them feel better (worse). As a result, they become less willing to take
chances and, therefore, more protective. In a series of studies, the authors showed that
when instructed to make a choice among several movies, happy respondents were more
attentive to the mood-lifting attributes of the movies than were those in neutral or
negative affect conditions.
Affect-regulation, in essence, predicts when people are likely to move to a more
positive affective state; and also explains when they are likely to try to protect a currently
positive affective state. That is, a combined version of existing affect-regulation theories
(i.e., mood-management and mood-maintenance or hedonic contingency) can explain
when a bad mood stimulates action (Cell 3) as well as when a good mood inhibits action
(Cell 2), the two cells of Figure 1 unaccounted for by theories based on an affective
evaluation mechanism. It is worth noting that the regulatory properties have also been
used to account for the impact of positive affect on behavioral encouragement (Cell 1).
The mood-maintenance hypothesis (Clark and Isen 1982) claims that people may perform
a mood-lifting behavior (e.g., to help) in an attempt to keep the good mood. However, no
direct evidence of this mediating process has been provided, and this effect may be due to
evaluation instead of regulation. This issue is further addressed later.
It is clear therefore that integrating the evaluative and regulatory properties of
affect is essential to account for all four categories of the affect-behavior matrix shown in
Figure 1. The Integrative Model of Affective Behavior (IMAB) attempts to accomplish
INTEGRATIVE MODEL OF AFFECTIVE BEHAVIOR (IMAB)
The first step to developing a model that addresses the four combinations of the
affect-behavior relationship shown in Figure 1 is to integrate (at the mediational level)
the evaluative and regulatory properties of affective states. The integrative model of
affective behavior (IMAB) proposes that affective evaluation and affect-regulation
operate in tandem as soon as a valenced affective state is activated.
Proposition 1: People 's affective states influence behavior continuously via two
parallel mechanisms: affective evaluation (both direct and indirect processes) and affect-
Strong evidence in the literature supports the behavioral influence of each
mechanism (Isen 1984; Forgas 2002; Martin and Clore 2001). However, the literature is
relatively silent regarding their parallel effects (but see Gendolla 2000 and Nygren et al.
1996 for exceptions).
The affective evaluation mechanism assumes that current affective states (positive
or negative) are likely to bias any evaluative judgment, and eventually behavior. Three
rather complementary hypotheses have emerged to account for the processes underlying
affective evaluations: one direct process (affect-as-information) and two indirect
processes (mood-congruency and processing styles). First, the affect-as-information
hypothesis proposes that affect itself may provide unique information that will be directly
retrieved during evaluation (Schwarz and Clore 1983). Individuals ask themselves "How
do I feel about it?", and use this information to make evaluative judgments. Recently,
Pham et al. (2001) moved a step further, indicating that affect is more than a heuristic
cue, as the initial hypothesis would propose. According to these authors, affect-as-
information is potentially faster, more reliable, and more predictive of subsequent
thoughts compared to "cold," reason-based types of information. Second, the mood-
congruency hypothesis states that concepts congruent with an individual's current
affective state may become more accessible (Bower 1981). As evaluation typically
requires a retrieval process, the likelihood of using mood congruent concepts during an
evaluative judgment increases, thereby biasing judgment. Third, the processing style bias
suggests that affect influences people's information-processing approaches, such as focus
of attention, categorization, and analytical vs. creative processing (Barone, Miniard, and
Romeo 2000; Forgas 1995; Schwarz and Clore 1996). For instance, positive affect fosters
the use of accessible information, relying more on top-down, expectation-driven
processes; whereas negative affect leads people to rely on more data-driven aspects
(Clore et al. 2001). Similarly, those in a good (bad) mood form broader (narrower)
categories (Isen 2000). As a result, in some circumstances, the impact of affect on
evaluation may occur simply as a result of an individual's changes in processing style
during the evaluative process.
In short, affect is assumed to influence evaluation directly (as information) or
indirectly (via changes in the accessibility of mood congruent material or through
variations in processing styles). Notice that the mood-congruency hypothesis and the
affect-as-information hypothesis provide similar predictions (Martin and Clore 2001).
Positive (negative) affect leads to more positive (negative) evaluations, thereby mediating
subjects' intentions and behavior (Pham 1998). However, this is the case as long as no
misattribution error is made salient. In fact, the affect-as-information hypothesis suggests
that as soon as people become aware of their current affective states, they tend to control
for any affective influence on judgments; whereas the mood-congruency hypothesis
would predict the identical pattern across levels of awareness.
Although all three paths may lead affect to influence behavior, the model proposed
here emphasizes the more reliably demonstrated affect-as-information and/or mood
congruent effects (for reviews, see Forgas 1995; Schwarz 1990). As the predictions are
usually the same across these two mechanisms, we do not attempt to disentangle these
competing/complementary explanations, since our objective is merely to fully incorporate
affective evaluation effects.
Whereas the affective evaluation mechanism focuses on the informative value of
affect and its mood congruent effects, the affect-regulation mechanism focuses on the
motivational aspect of specific affective states. For example, a current (or expected)
negative (positive) mood may motivate individuals to improve (or protect) their actual
affective state. In other words, affect has informational and goal properties that may
influence behavioral decisions.
Several research streams have devoted attention to this phenomenon. Tice,
Bratslavsky, and Baumeister (2001) demonstrated that people deliberately use food
consumption as a tactic for mood repair. Luce (1998) showed that in an attempt to
regulate a current or expected negative affective state, people systematically avoided
difficult tradeoffs, thereby making less-optimal choices. Similarly, motivation for upward
affect-regulation has also led subjects in a bad mood to increase helping (Bagozzi and
Moore 1994; Schaller and Cialdini 1990) and self-gifting (Luomala and Laaksonen 1997;
Mick and DeMoss 1990). Finally, presumably in an attempt to protect a current affective
state, subjects in a good mood were less likely to take risks compared to respondents in a
control condition (Isen and Geva 1987).
In short, we assume that there are two different pathways through which affect can
influence or bias behavior. However, further assuming simultaneous activity of both
types and interaction between these two processes before action, we seek to identify the
circumstances that lead one mechanism to be more likely to predominate. What are the
situational and internal cues that stimulate (vs. lessen) the impact of affect within these
Proposition 2: The affective evaluation mechanism assumes that affect can acquire
the properties of information as well as alter the accessibility of mood congruent
information in memory. Its impact is therefore contingent people 's current affective state
and on the availability and use of competing diagnostic information about the target
during the evaluative process.
As affect can acquire informational properties, the quantity and quality of
competing information available (as well as the context in which this information is used)
influences the extent to which affect influences evaluation, and eventually behavior
(Martin and Stoner 1996). As the amount of other available diagnostic information
decreases, the impact of affect on behavior via the affective evaluation mechanism
strengthens (Martin and Clore 2001). There is some evidence in the literature that the use
of affect-as-information varies based on its diagnosticity. Pham (1998) showed that the
use of affect is clearly context-dependent, varying also in terms of its quality (i.e.,
representativeness and relevance) during the evaluative process. He showed that subjects
primed with a hedonic (vs. utilitarian) motive were more willing to use affect as
information while assessing purchase intentions. Moreover, the quantity of competing
information has also been shown to produce a significant impact on the affective
evaluation mechanism. Siemer and Reisenzein (1998) showed that by mitigating people's
access to additional information through a time constraint and/or a secondary parallel
task, the effects of mood on judgment increased. Finally, ambiguous information has also
led to an increase in the impact of affect on evaluation and/or behavioral intention (Gorn
et al. 2001; Isen and Shalker 1982; Miniard, Bhatla, and Sirdeshmukh 1992).
Although affect may act as a heuristic (Schwarz 1990) by providing unique
information to rely on when no additional information is made available or when the task
is too complex, affective influence can occur via substantive changes in content. If the
task promotes substantive thinking, it may well increase people's likelihood of accessing
mood congruent material, thereby leading to a strong affective evaluation change (Forgas
1995). In this case, mood congruent retrieval (instead of affect-as-information) may best
represent the operative affective evaluation mechanism (Forgas 1992; Forgas 1993).
In summary, the IMAB proposes that the affective evaluation mechanism
intensifies as subjects' current affective state is directly used as information during the
evaluative process (due to ease/speed of access or lack of more diagnostic information)
and/or indirectly used (via recall of mood congruent material).
Proposition 3: The affect-regulation mechanism assumes that affect can acquire
the properties of a goal. Its impact on behavior is therefore contingent on people's
current affective states, on their anticipated affective states as a result of action, on the
presence of competing/complementary goals, and on people 's willingness and skills to
achieve the goal.
The model assumes that affect-regulation (i.e., pursuit or protection of positive
feeling states) is potentially active and may even overcome the affective evaluation
mechanism depending on internal and environmental contingencies. First, people's
current affective state is critical for affect-regulation to operate, since those experiencing
negative affective states will have a stronger motivation to regulate (upward) their
unsatisfactory feelings (Parrott and Sabini 1990), whereas those experiencing positive
affect will have a stronger motivation to protect their current pleasant feelings (Isen and
Geva 1987). The model implies, therefore, that valenced states (i.e., strong signals) are
more likely to trigger the affect-regulation mechanism than neutral states (i.e., weak
signals). For people to be strongly motivated to change or protect their current feelings,
there must be a strong affective signal, which indicates that the individual is experiencing
a desired (positive) or undesired (negative) affective state. Cohen and Andrade (2004)
provided initial evidence of such an effect. Second, people must anticipate an affective
change (upward or downward) as a result of action. In other words, the behavioral
activity must be expected to lift or damage people's current feelings (Cialdini and
Kenrick 1976; Isen and Geva 1987; Raghunathan and Pham 1999). Notice that this
assumption implies that affect-regulation is a function of people's intuitive theories about
the affective consequences of specific behaviors (e.g., "I believe that eating chocolate
makes me feel better"). Third, competing (or cooperative) goals work in tandem to
determine the impact and direction of an individual's current affective state. If more
personally relevant or salient goals are operative, individuals may delay upward affect-
regulation and retain a negative affective state in order to achieve a competing goal
(Cohen and Andrade 2004) or neutralize negative and positive affect in an attempt to
enhance response flexibility and improve performance (Erber and Erber 2001). Finally,
individuals must have the willingness and skills to regulate their current affective state.
Studies examining chronic affective states, for instance, have shown that depressed
people do not perceive themselves as capable of regulating upward their current negative
affective states (Davidson et al. 2002; Kanfer and Zeiss 1983), which may prevent any
mood-lifting attempt. If these conditions are met, affect-regulation is likely to direct
behavior, potentially overcoming the opposing effects of the affective evaluation
In summary, the models' main predictors for individuals in a positive affective
state, are that, (1) the affective evaluation mechanism should stimulate action (e.g.,
producing an increase in purchase intention), particularly when no strong competing
diagnostic information about the behavior/environment is available; however, (2) when
the behavior presents a potential threat to people's positive feelings (and the other
previously mentioned contingencies are met), then the affect-regulation mechanism is
likely to dominate the informational properties of the initial positive state. In this case,
positive affect should inhibit action (e.g., producing a decrease in purchase intentions) as
people attempt to protect their current affective states.
For individuals experiencing a negative affective state, the model predicts that (3)
the affective evaluation mechanism should inhibit action (e.g., producing a decrease in
purchase intentions), particularly when no strong competing diagnostic information about
the behavior/environment is available; however, (4) when the behavior presents a
potential mood-lifting benefit to people's negative feelings (and the other previously
mentioned contingencies are met), then the affect-regulation mechanism is likely to direct
behavior. In this case, negative affect should stimulate action (e.g., producing an increase
in purchase intentions) as people attempt to improve their current feelings. 2
2 Remember that as the IMAB focuses initially on general negative and positive states, it does not make
predictions for specific types of emotions. However, it does recognize that emotions vary in terms of affect-
regulation and affective evaluation tendencies, which eventually bias behavior. For instance, it is known
that whereas sadness increases risk-taking, anxiety decreases it (Raghunathan and Pham 1999). Similarly,
while helping increases as a function of sadness (Baumann, Cialdini, and Kenrick 1981), it may well
decrease as a result of frustration.
INITIAL EVIDENCE SUPPORTING THE MODEL: THE IMPACT OF AFFECT ON
HELPING, RISK TAKING, AND EATING BEHAVIOR
The impact of affect on behavior has been investigated in several different research
streams within the psychology literature. Thus, to provide initial evidence for the model
basic's propositions, the impact of affect on helping, risk taking, and eating behavior is
reviewed in an attempt to assess IMAB's ability to account for the results and resolve
some of the apparent inconsistencies.
Prevalent finding in the helping literature is that current affective states influence
individuals' willingness to help. However, the effects do not follow a single pattern (for
reviews, see Batson 1990; Salovey, Mayer, and Rosenhan 1991; Schaller and Cialdini
1990). Researchers tend to agree that the relationship between positive mood and helping
is, in general, well established; and that positive mood increases people's propensity to
help (Isen, Clark, and Schwartz 1976; Isen and Levin 1972; Levin and Isen 1975).
However, there is some evidence that the opposite may also be true; thus a decrease in
helping due to individuals' positive feelings (Isen and Simmonds 1978). The impact of
negative affect on helping is also bi-directional. Negative mood sometimes increases
helping (Cialdini et al. 1973; Cunningham, Steinberg, and Grev 1980; Manucia, Bauman,
and Cialdini 1984) and sometimes decreases helping (Berkowitz 1972; Berkowitz and
Connor 1966; Isen 1970). Several hypotheses have been proposed to account for these
"inconsistent" patterns, however the underlying mechanisms seem to vary almost as
much as the results themselves: positive mood-maintenance, guilt reduction, negative
state relief, aversive arousal reduction, positive affective priming, and negative affective
priming (Batson 1990; Salovey et al. 1991). The main findings and their respective
explanations are reviewed below. Then we show how the proposed IMAB offers a more
integrative and parsimonious account of this body of work. The framework proposed in
Figure 1 is used to categorize the effects.
Positive Affect and Helping
In a field study, Isen and Levin (1972) showed that subjects who found a dime in
the coin return of a public telephone were subsequently more willing to pick up papers
dropped off in front of them by a confederate (Study 2). Similarly, after manipulating
mood through false feedback, Isen (1970) showed that happy (sad) students were more
(less) willing to give money to the "Junior High Air-Conditioning Fund". Indeed, the
positive impact of good mood on prosocial behavior is quite robust (Aderman 1972;
Berkowitz and Connor 1966; Levin and Isen 1975; Moore, Underwood, and Rosenhan
1973). Two underlying mechanisms leading to this effect have been advanced; one
cognitive (i.e., priming effects), and one motivational (i.e., positive mood-maintenance).
The cognitive/priming explanation is based essentially on mood-congruency effects (Isen
et al. 1976; Isen, Clark, Shalker, and Karp 1978), through which positive information
became more accessible during evaluation and influenced behavior (Clark and Waddel
The competing motivational explanation for the effects of positive affective states
on helping adopts a regulatory process approach. It has been proposed that people in a
good mood try to remain in their current affective states and, therefore, will be more
willing to help (Isen 1984; Levin and Isen 1975). This hypothesis, however, has not
found direct empirical support in the literature. As Schaller and Cialdini (1990)
summarized "evidence is scarce that happy subjects help as a means to maintain positive
moods" (p. 282). Indeed, when mood-maintenance plays a role, it may reduce rather than
increase helping (since the beneficial affective consequences of trying to help may not be
clear). Although direct empirical evidence of the actual mediating processes is still
lacking, researchers tend to agree that biases in evaluative judgment can play a role on
people's propensity to help (Batson 1990; Salovey et al. 1991; Schaller and Cialdini
1990). However, as no study has contrasted affect-as-information vs. mood-congruency
mechanisms, and both predict the same effects, it seems premature to claim which
process (if not both) is responsible for the impact of positive mood on helping.
Little evidence is available showing that being in a good mood can decrease
helping. However, Isen and Simmonds (1978) found that when the helping scenario
displays situational cues that threaten subjects' current positive mood (e.g., a demanding
helping task), these individuals were indeed less likely to help than subjects in a neural
mood. The authors suggested that a challenging helping task may have led happy subjects
to anticipate negative affect and triggered a self-protective regulatory mechanism. This
type of effect will be further elaborated under the risk-taking literature review, where the
impact of positive affect on behavioral discouragement is well established.
Negative Affect and Helping
Studies showing that negative affect increases helping have generated several
related hypotheses to account for the underlying mechanisms, such as guilt reduction
(Carlsmith and Gross 1969; Regan, Williams, and Sparling 1972), negative mood relief
(Bauman, Cialdini, and Kenrick 1981; Cialdini et al. 1973; Cialdini and Kenrick 1976;
Manucia et al. 1984), and aversive arousal reduction (Piliavin et al. 1981, 1982).
Although adopting different research approaches, they all share the basic assumption that
upward affect-regulation is at the core of people's disposition to help. Helping is
conceived to be an affect-regulation strategy aiming at achievement of this somewhat
Cialdini and colleagues were among the first to categorize helping as a mood repair
strategy. Cialdini et al. (1973) showed that subjects in bad mood were more likely to help
in response to another person's request than those in control conditions. Most
importantly, as soon as rewarding hedonic benefits were interposed between the mood
manipulation and the help request (i.e., an unexpected monetary reward or approval for
task performance), the effects of negative affect on helping disappeared. The authors
asserted that helping, monetary reward, and positive feedback perform a similar
functional goal, bad mood relief (see also Baumann et al. 1981). Manucia et al. (1984)
provided further, and perhaps even more compelling, evidence implicating an upward
affect-regulation strategy as the mediating mechanism linking being in a bad mood to
helping. After instantiation of positive, neutral, or negative affective states, subjects were
asked to take a placebo pill. Half the subjects were informed that this pill would "freeze"
their current affective states for a while. The authors predicted that if helping was used
as a mood repair strategy, subjects in the "frozen" bad mood condition should help no
more than those in control conditions and significantly less compared to the "non-frozen"
bad mood condition. The results confirmed the predictions. Subjects in a bad mood who
were told about the "freezing" effects of the drug helped much less (compared to those in
the "non-frozen" bad mood condition), and similar to those in the neutral condition.
There is also evidence that negative affect can also decrease helping under certain
circumstances (Cialdini and Kenrick 1976; Isen 1970; Isen, Horn, and Rosenhan 1973;
Moore et al. 1973). However, purely cognitive approaches have been used to account for
operative mediating processes. Similar to positive moods, negative moods are also known
to prime congruent thoughts: "Thoughts of deprivation, helplessness, and uselessness
may become especially available, rendering such sad and self-focused individuals less
likely to help. ." (Salovey et al. 1991, p. 222). In summary, whereas upward affect-
regulation is typically identified as responsible for instigating people to help (Cell 3),
affective evaluation is identified as playing a major role when opposite results are found
(Cell 4). At best this strikes us as a "marriage of convenience" rather than a systematic
and balanced explanation.
Moreover, there has been a failure to find evidence consistent with affect-regulation
(i.e., helping increase) in certain studies where negative mood had been induced.
However, in studies where a negative mood decreased helping, children were used as
subjects (Cialdini and Kenrick 1976; Isen et al. 1973; Moore et al. 1973). It is possible
that for children helping is simply not perceived as an effective affect-regulation strategy.
In that case, affective evaluation should have a stronger impact. Indeed, Cialdini and
Kenrick (1976) showed that age and levels of socialization are critical moderating
variables. In one experiment, the authors found an interaction between age (6-8, 10-12
andl5-18 years) and mood on helping. Fifteen to eighteen year old subjects in a bad
mood helped more than those in the other two conditions, who turned out to behave
similarly to one another. As the authors predicted, individuals whose socialization
process is still incipient do not perceive helping as self-gratifying. In short, young
subjects in a bad mood do not help because altruistic behaviors are not perceived as a
viable affect-regulation strategy. As a result, the negative thoughts elicited by a bad mood
operate to reduce helping.
The helping literature shows that positive and negative affect can stimulate or
discourage helping depending on situational cues available in the environment (i.e., the
four Cells of Figure 1), however, a combination of several theories have been required to
account for these effects. IMAB's main claim (proposition 1) is that, under the basic
assumption that two parallel mechanisms underlie the impact of affect on behavior,
situational cues and internal affective signals determine which mechanism will prevail.
As a result, a single model can reconcile these effects.
Positive affect leads to helping increase (Cell 1) via an affective evaluation
mechanism Proposition 2. That is, a positive mood biases (positively) subjects'
evaluations of the helping task either via affect-as-information and/or mood-
congruency -, both of which should increase subjects' willingness to help (Isen 1970;
Isen and Levin 1972; Moore et al. 1973). However, IMAB also requires us to consider
affect-regulation Proposition 3-, so that when situational cues lead subjects to anticipate
negative affect, affect-regulation becomes the dominant mechanism, and behavior is
discouraged (Cell 2). The reason for such protective reaction is that subjects in a positive
mood have more to lose compared to control conditions. That was the case in Isen and
Simmonds' (1978) study, where the helping task was mood-threatening, probably leading
subjects to speculate about the negative consequences of helping on their current positive
affective state. Although this type of effect is rather sporadic in the helping literature, the
study of risk-taking, as we will see, offers consistent theoretical and empirical evidence
of the impact of anticipated negative affect on behavior.
For subjects in a negative mood, upward affect-regulation is usually a reasonably
important motivator and is likely to dominate the impact of the affective evaluation
mechanism. As a result, individuals attempt to improve their current negative affective
states (Cell 3), with several studies providing evidence of a helping increase for sad
subjects (Bauman et al. 1981; Cialdini et al. 1973; Cialdini and Kenrick 1976; Manucia et
al. 1984). However, as with any other goal, affect-regulation is contingent on other
moderating variables, such as subjects' recognition of the stimulus as an effective upward
affect-regulation strategy Proposition 3. When sad subjects were incapable of
perceiving the mood lifting benefits of helping, affect-regulation was mitigated, and the
affective evaluation mechanism (i.e., negative evaluation of the environment) led to a
decrease in helping (Cell 4). That was the case when children were used in the
experiments (Cialdini and Kenrick 1976; Moore et al. 1973).
In short, the proposed IMAB accounts for the bulk of effects in the helping
literature, by suggesting that (1) positive affect increases helping via affective evaluation
(i.e., priming effects and/or affect-as-information); (2) positive affect decreases helping
via affect-regulation when accompanied by mood-threatening cues; (3) negative affect
increases helping via (upward) affect-regulation when mood-lifting benefits are made
available; and (4) negative affect decreases helping via negative affective evaluation
providing subjects are unable to perceive the mood-lifting benefits of helping.
Understanding the impact of affect on risk perception and risk-taking can thus
provide additional insights into the behavioral consequences of affect and its mediating
processes. Johnson and Tversky (1983) found that when asked to evaluate the subjective
probability of positive future events, subjects in positive moods reported a higher
subjective probability (compared to control respondents), and a much higher subjective
probability compared to subjects in a negative mood. The opposite was true when they
were asked to evaluate the subjective probability of negative future events. In this case,
subjects in negative moods reported the highest subjective probability (compared to those
in neutral moods), and were much higher than subjects in a positive mood. After tracking
for cognitive processes (thought listing), Wright and Bower (1992) showed that
individuals focused more on mood congruent information during the assessment of
subjective probabilities. The correlation between affective state and expected outcomes is
currently well established (Loewenstein et al. 2001).
Thus, based on prevailing evidence and on the assumption that people will act
"rationally", one would think that subjects in bad mood, who tend to perceive a situation
as riskier, should be less inclined toward risk-taking. The opposite should be true for
subjects experiencing a positive affective state. Individuals in good moods, who usually
perceive a safer environment, should be more prone to risk-taking. Yet, findings in the
literature do not fully confirm either of these two predictions. Although the results are
rather consistent as to the impact of affect on risk perception (Constans and Mathews
1993; Johnson and Tversky 1983; Mayer et al. 1992; Pietromonaco and Rook 1987;
Wright and Bower 1992), yet to be resolved is why the impact of affect on risk-taking
does not follow the predicted "rational" pattern.
Negative Affect and Risk-Taking
Negative affective states have been shown to increase risk-taking (Gehring and
Willoughby 2002; Leith and Baumeister 1996; Mano 1992, 1994; Mittal and Ross 1998;
Raghunathan and Pham 1999), though this seems counterintuitive (since we know that
negative affect leads to an increase in risk perception, Johnson and Tversky 1983; Wright
and Bower 1992). Two sets of hypotheses have been developed to explain such effects:
affect-regulation (e.g., mood repair Raghunathan and Pham 1999) and affective
disruption (e.g., restricted attentional capacity Mano 1992).
Leith and Baumeister (1996) adopted an "affect as disruption" perspective to
account for this pattern of results. Since negative feelings may disrupt people's ability to
properly/rationally make accurate evaluations, such interference could lead subjects to
choose the "poorer"/riskier option. In a series of six studies the authors showed that
embarrassment, anger, and unpleasant arousing feelings led to an increase in risk-taking.
Since, in one of the studies, sadness did not differ from the neutral conditions, the authors
used this null effect to conclude that the "risky tendencies are limited to unpleasant
moods accompanied by high arousal" (p. 1250). This hypothesis has not found much
support in the literature.
To assess the impact of different emotions on risk-taking, Raghunathan and Pham
(1999) investigated the impact of sadness (low arousal) and anxiety (high arousal).
Contrary to the "affect disruption" prediction, when presented with two gamble options
in a consumer decision task (low risk-low payoff vs. high risk-high pay off), sad subjects
preferred the riskier alternative with a higher payoff (compared to anxious subjects), who
turned out to be strongly risk-averse. The authors suggested that different goals are
primed for sad vs. anxious people: sad subjects focusing on reward replacement (mood
repair) whereas anxious subjects focused on uncertainty reduction. Sad subjects thus
perceived the high risk-high payoff option as more attractive (i.e., mood-lifting), whereas
anxious subjects preferred the low risk-low payoff alternative (i.e., "it can reduce my
uncertainty"). This rationale is in line with the Eysenck and colleagues studies in which
anxiety has led to attentional and interpretational biases. Based on Eysenck's (1992)
cognitive theory of trait anxiety, it has been found that highly anxious people have an
attentional bias toward threat-related words and also interpret ambiguous information as
more threatening (Eysenck, MacLeod, and Mathews 1987). Derakshan and Eysenck
(1997) also found that highly anxious people display an interpretative bias for their own
behavior in social situations the behavior is perceived as more anxious.
Raghunathan and Pham's and Eysenck and colleagues' findings converge with the
IMAB and highlight a critical assumption of the model, the interdependence of the
affective evaluation and the affect-regulation mechanisms Proposition 1. Anxious
people appear to reinterpret any risky action, making it more negatively arousing and
causing whatever mood-lifting benefits that might be associated with a high-risk bet to
dissipate, thereby mitigating the impact of the affect-regulation mechanism.
Simultaneously, anxious subjects arrive at a rather pessimistic and threatening assessment
of the environment, which further strengthens the impact of the affective evaluation
mechanism. The impact of strong negative affective evaluation combined with the
absence of upward affect-regulation forces leads to risk-averse behavioral patterns (Cell
4). However, when people experience sadness, the mood-lifting benefits of similar risk-
taking may remain stable or even intensify, offsetting the negative impact of affective
evaluation on risk perceptions, and leading people to choose more risk-prone behaviors
(Cell 3). Thus, the type of affective state being experienced may well produce different
interpretations of, or attention to, upward affect-regulation opportunities that may be
available Proposition 2, making the affect-regulation mechanism either more or less
influential Proposition 3.
Positive Affect and Risk-Taking
Kahn and Isen (1993) showed that being in a positive mood (compared to a neutral
mood) stimulated individuals to seek more variety among otherwise safe and enjoyable
food products. Arkes and colleagues (1988) also demonstrated that happy respondents
(compared to subjects in a neutral mood) were more willing to pay for lottery tickets
(Study 1). These and similar results show that people in a good mood are apparently more
prone to risk-taking. Since we know that people in a good mood are more optimistic
(Johnson and Tversky 1983; Wright and Bower 1992), this pattern might be labeled as
rather intuitive. However, not all findings have shown this risk-prone behavior among
happy people. Kahn and Isen, for example, showed that the increase of variety-seeking
behavior for happy subjects disappeared as soon as a product's negative features were
included or made salient in the choice context. Similarly, Arkes and colleagues also
showed that whereas happy subjects (vs. a control group) displayed risk-prone behavior
in a pleasure-seeking situation (i.e., buying lottery tickets), they exhibited a risk-averse
pattern in a loss avoidance situation (i.e., buying insurance). It has been hypothesized that
subjects in a positive mood are more risk seeking than subjects in a neutral mood
providing the potential losses are not salient or too high (Nygren et al. 1996). Research
on this topic has used a motivational rationale to account for the findings: people in a
good mood facing mood-threatening stimuli become more self-protective of their current
feelings, thereby discouraging risky behaviors that may lead them to feel bad.
The pattern of results is fully consistent with IMAB, which predicts that affect-
regulation is activated not only when subjects experience negative feelings, but also when
subjects anticipate negative feelings Proposition 3. Thus, when losses are likely (e.g.,
high-risk condition), people in a good mood face a greater relative loss than people in a
neutral mood (Isen and Geva 1987). Such anticipatory negative emotional reactions
reduce the likelihood of engaging in risky behavior to achieve affect-regulation goals,
and hence counteract the impact of positive affect-based evaluations. Nygren et al. (1996)
used the seemingly contradictory expression "cautious optimism" to underscore the dual
and, here, opposing mechanisms at work. They summarize their first study by saying that,
on the one hand, "positive affect participants significantly overestimated the
probabilities...", but on the other hand, "...were less likely to gamble than were controls
when a real loss was possible..." (p. 59). In IMAB terms, whereas "optimism" is a result
of affective evaluations (i.e., affective priming effects), "caution" represents a
consequence of affect-regulation, triggered by anticipated negative affect.
In summary, being in a good mood may promote both risk-averse (Cell 2) as well
as risk seeking behavior (Cell 1). The outcome depends on mediating effects linked
primarily to affect-regulation, which have been shown to be contingent on the presence of
mood-threatening stimulus and its subjective likelihood of triggering downward affect-
regulation. When no "threats" are made salient affective evaluation leads to risk prone
behaviors Proposition 2 -, whereas when environmental cues signal threats affect-
regulation goals are activated Proposition 3 promoting negative mood avoidance
through risk-averse behaviors.
The impact of emotion on eating behavior has been widely investigated (for
reviews, see Canetti, Bachar, and Betty 2002; Christensen 1993; Greeno and Wing 1994).
Researchers' interests vary significantly; from the effects of stress on psychopathological
behaviors (e.g., obesity and bulimia) to normal influences of mild mood swings on food
preferences (e.g., cravings for sweets, carbohydrates, etc.); from tail-pinch stressors and
animal eating responses to unpleasant movies and human propensity to eat snack food.
As our analysis and proposed model focuses on the impact of mild affective states
on everyday behavior, we will concentrate on how negative and positive affective states
influence normal food intake. Consistent with the evidence reviewed above, the first
conclusion to be drawn from this body of research is that affect does not lead to a unique
behavioral consequence. Positive and negative affective states may well stimulate or
discourage food intake.
Negative Affect and Eating Behavior
There are a far greater number of studies dealing with the impact of negative affect
(compared to positive affect) on eating behavior. The underlying assumption in most of
the literature is that food acts as mood-regulator, lifting subjects' current affective state
after intake (Bruch 1973; Kaplan and Kaplan 1957; Morris and Reilly 1987; Polivy and
Herman 1976; Thayer 1989). Thus, (compared to a control condition), negative affect is
expected to encourage eating behavior. However, the results have been shown to vary as
a function of several moderating variables such as gender and food type.
Until the 1990s, the general impact of negative affect, especially stress, had been
surprisingly limited to animal research. Those dealing directly with humans concentrated
more on the interaction between affect and individual differences on eating behavior (see
Greeno and Wing 1994 for a review). Most recently, however, new studies have emerged
in which broader and more fundamental conclusions can be drawn about the
consequences of negative affect on people's eating behavior.
Grunberg and Straub (1992) exposed subjects to a film about industrial accidents
(negative affect) or a pleasant travelogue (control) while having snack foods available in
the room (dependent measure). They found that eating consumption increased as a result
of negative affect, but only among women. The results actually reversed for male
subjects, who reduced the amount of food intake as a consequence of negative affect.
Although the authors did not advance a systematic theoretical explanation for the
mediating effects, the results have proven quite robust. Whereas negative affect tends to
increase food intake among women (Macht 1999; Patel and Schlundt 2001; Weinstein,
Shide, and Rolls 1997; Willner et al. 1998), this effect is either canceled (Pine 1985) or
reversed among men (Abramson and Wunderlich 1972; Macht, Roth, and Ellgring 2002;
Reznick and Balch 1977).
A potential explanation for such variation is that strategic affect-regulation through
food intake is stronger in women than men (Macht 1999; Steptoe, Pollard, and Wardle
1995). Men may not become more attracted to food, or at least certain types of food, as
their affective states worsen. Although a variety of biological and psychological
explanations have been offered to explain why food intake does not increase among men,
they cannot explain why bad feelings lead men to reduce food intake.
As we have shown in the other streams of research, pursuing a single explanatory
mechanism may be at the root of the apparent inconsistency. In this case, however, the
main explanatory mechanism has been affect-regulation rather than affective evaluation.
Once again, IMAB proposes that understanding the interaction between affective
evaluation and affect-regulation is critical to explain the bi-directional pattern. Whereas
upward affect-regulation accounts for the increase in food intake as a result of negative
affect (Cell 3), negative affective evaluation is likely to explain food intake inhibition
(Cell 4). According to IMAB's third proposition, affect-regulation, as a goal, is
contingent on the people's underlying theories about the affective consequences of action
(i.e., "Is this behavior mood-lifting/threatening?"). If men (vs. women) are less likely to
perceive certain types of food as mood-lifters (e.g., Eating chocolate does not make feel
better"), the impact of affect-regulation will be mitigated and affective evaluation will be
most likely to drive the effects (i.e., reduce eating). Our multiple mediator analysis also
implies that if the type of food is not perceived as a "mood-lifter" negative affect is
expected to decrease eating. Consistent with Proposition 3, Oliver and Wardle (1998)
showed that whereas stress increased the consumption of snack-type foods (perceived
both as "quick energy" products and "treats") it decreased the consumption of typical
meal-type foods (fruits and vegetables, meat and fish). Relatedly, Willner and Healy
(1994) showed that after negative affect induction subjects lowered their own evaluation
of cheese in terms of pleasantness and desirability (see also Macht et al., 2002), again
suggesting that affective behavior toward food with no subjective mood-lifting attributes
is mostly directed by the affective evaluation mechanism. So, as bad feelings produce a
worsening evaluation of focal objects such as food, eating should decline.
Positive Affect and Eating Behavior
Since eating disorders (obesity, binge eating, bulimia, anorexia, etc.), which are
normally associated with negative affect, have been at the forefront of research done in
the field from the 70s through the 90s, only recently have researchers devoted attention to
the consequences of positive affect. The general pattern of results suggests that positive
mood stimulates eating (Cools, Schotte, and McNally 1992; Macht 1999; Macht et al.
2002; Patel and Schlundt 2001 upper-left corer of Figure 1), though null effects have
also been reported (Frost et al. 1982; Schmitz 1996). Based on two-week food diaries,
Patel and Schlundt (2001) found that (compared to a control condition) obese women
increased food intake while experiencing both positive and negative affect. Contrary to
the authors' expectations of an interaction between mood and social context (eating alone
vs. eating in a social context), the impact of valenced moods occurred under both social
context scenarios. No explanation was provided to account for the results. Macht and
colleagues (2002) provided a compelling mood congruent explanation for such effects.
They showed that male subjects experiencing positive (vs. negative) affect provided
higher ratings on two general dimensions for chocolates they were eating: affective
responses to chocolate (e.g., taste pleasantness) and motivation to eat (e.g., appetite). The
authors suggested that the positive impact of affect on food intake was probably a result
of mood congruent effects during subject's internal and external evaluation.
Consistent with IMAB's basic propositions, our analysis of prior research suggests
that positive affect probably stimulates eating behavior via the affective evaluation
mechanism, though little has been done to isolate mood congruent effects from affect-as-
information. Finally, to the best of our knowledge, no study has shown that people in a
good mood reduced food intake (Cell 2). However, according to IMAB, this pattern of
results is likely when negative consequences of eating become salient. For instance,
happy people (compared to a neutral mood condition), should be less likely to eat
chocolates if negative nutrition facts are highlighted (i.e., fat product and/or subsequent
feelings of guilt). In summary, as with research on helping, the potential negative
consequences of positive affect are largely unexplored in the eating behavior literature.
Our review shows that IMAB's three propositions can account for the observed
consequences of affect on behavior and behavioral intentions across three different
bodies of literature: helping, risk-taking, and eating behavior.
Behavioral stimulation for people experiencing positive affect (Cell 1) may occur
mostly as a result of the affective evaluation mechanism affect-as-information and/or
mood congruent effects- and there is a paucity of evidence to support mood-
maintenance models of affect-regulation. As long as no aversive or threatening cues
become salient in the environment, happy people perceive a safer environment (i.e., bring
positive thoughts to mind) and therefore become more likely to help, to take risks in
gambles, and to eat. When mood-threatening cues are made salient, the affect-regulation
mechanism activates and may lead to behavioral discouragement for people experiencing
positive affect (Cell 2). As happy people are more sensitive to potential negative affective
consequences, since they have more to lose, and bad feelings can be anticipated,
behavioral discouragement takes place when negative aspects or consequences become
salient. This explains a decrease in helping when the task is mood-threatening as well as
risk avoidance when the odds are too high. Happy people thus seem more motivated to
avoid a negative mood than to maintain positive feelings, since there is no salient threat
likely to instantiate this motivation.
On the negative side of the affective spectrum, the affective evaluation mechanism
seems to drive the impact of negative affect on behavioral mitigation (Cell 4). People
perceive a more intimidating environment (i.e., bring more negative thoughts to mind)
and become less likely to help, to take risks, and to eat. This is most likely to be the case,
however, only when the affect-regulation mechanism is inactive or blocked. Blocking or
mitigating effects can be a result of subjects' inability to perceive specific behavior as an
effective mood-lifting opportunity. That was the case among children (vs. adults) facing a
helping opportunity, anxious (vs. sad) people facing a risky-high payoff opportunity, and
men (vs. women) facing an eating opportunity. When the behavior is perceived to be an
effective upward affect-regulation strategy, there are no stronger competing goals in the
environment, and people are willing/capable of mood improvement, enactment becomes
more likely for those experiencing negative affect via the affect-regulation mechanism,
often counteracting the impact of the affective evaluation mechanism (Cell 3). In such
situations people attempt to improve their current negative affective states, and this
accounts for observed increases in helping, risk-taking, and eating behavior under these
In short, combining the hitherto separately considered affect evaluation and affect-
regulation mechanisms in the integrative model of affective behavior provides the
following parsimonious account of this substantial literature. Behavioral stimulation for
subjects in a positive mood and behavioral mitigation for subjects in a bad mood seems to
be mediated mainly by the affective evaluation mechanism (affect-as-information and/or
mood-congruency effects), whereas behavioral stimulation for subjects in a bad mood
and behavioral mitigation for subjects in a good mood is likely to be mainly mediated by
the affect-regulation mechanism.
Such conclusion is, however, premature, since, to the best of my knowledge, no
single study has addressed the impact of positive, neutral, and negative affective states on
behavior, while providing a direct assessment of the two main mediating processes under
operation. In the next chapters, the results of three experiments are presented in an
attempt to accomplish this goal and offer stronger empirical support for the integrative
model of affective behavior.
TESTING IMAB: PREDICTIONS
To test IMAB propositions, a series of three experiments was conducted.
Experiments 1 and 2 show (1) that positive (negative) affect stimulates (inhibits)
purchase intentions via the affective evaluation mechanism (Cells 1 and 4 of Figure 1),
and (2) that negative affect stimulates purchase intentions via affect-regulation (Cell 3 of
Figure 1). Experiment 3 addresses the remaining effect (Cell 2 of Figure 1), and
demonstrates the circumstances in which positive affect may inhibit action (i.e.,
consumption). It also replicates Cell 1 (i.e., when positive affect increases consumption)
with an actual behavioral measure. All three experiments focused on two critical
moderating variables: people's current affective states and the availability of mood-lifting
(experiments 1 and 2) or mood-threatening (experiment 3) cues in the environment. The
other situational contingencies are held constant throughout all the conditions across the
three experiments. Below, we describe the predictions for experiments 1 and 2.
Predictions for experiment 3 will be described in Chapter 8.
In a consumption scenario where people do not possess enough information about
the product and the environment does not provide people with an opportunity to
upwardly regulate their affective states, IMAB predicts that the affective evaluation
mechanism should influence purchase intentions. Affect congruency principles it is
hypothesized that people in a negative (positive) mood will have a more negative
evaluation of the environment and/or the product, which makes them less (more) likely to
purchase a particular product, compared to a control condition (Figure 2 white bars).
However, if the product presents mood-lifting benefits, those experiencing negative affect
have an opportunity to upwardly regulate their current affective state, which can lead
them to increase their purchase intentions despite the opposing influence of the affective
evaluation mechanism. On the positive side of the affective spectrum, people are already
experiencing something approaching the "ideal" affective state. Thus, the mood-lifting
benefits of a product are not expected to have a strong effect on purchase intention.
However, as the affective evaluation mechanism is also active, people are expected to
provide a more positive evaluation of the product/environment, which should increase
purchase intentions. In short, when they are presented with the opportunity to purchase a
mood-lifting product, the combination of affect-regulation for those experiencing
negative feelings and affective evaluation for those experiencing positive feelings, leads
to a "U" shape pattern (Figure 2 black bars).
4 u mood-lifting benefits
3 no mood-lifting benefits
Negative Neutral Positive
Figure 2. Predictions for experiments 1 and 2
Notice that those in the neutral affect condition are not expected to be strongly
sensitive to the mood-lifting benefits of the product. As they are also at some distance
from the ideal affective state, one could expect these respondents to be willing to improve
their current feelings in order to achieve the ideal affective state. However, affect-
regulation has been shown to be a function of both the affective discrepancy between
current and ideal states and, importantly, the strength of the affective signal (Cohen and
Andrade 2004). Thus, since neutral affect provides weak signals, affect-regulation goals
may not be instantiated, and the regulatory process may not unfold. IMAB incorporates
Marketers have recently highlighted their products' mood-lifting benefits in an
attempt to persuade consumers. Kellogg's, for instance, displays a web page called
"Breakfast and Mood", where it describes the affect-regulation benefits of eating
breakfast cereal in the morning. If the impact of mood-lifting cues are contingent on
people's current and anticipated feelings, as proposed by IMAB, varying respondents'
current feelings (negative, neutral, positive) and providing them with an opportunity to
purchase a mood-lifting versus a non mood-lifting product should allow us to test
IMAB's predictions described above. Experiment one adopts this procedural approach.
Subjects and Design
Two hundred eighty-four undergraduate students from a southeastern university
participated in the experiment in exchange for course credit. The study adopted a two
(product benefits: mood-lifting vs. non mood-lifting) by three (affective state: negative
vs. neutral vs. positive) by two (replicate: coffee vs. cereal) between subjects design.
Respondents were randomly assigned to one of the twelve conditions.
After they entered the lab, respondents were instructed to choose one of the
computers and start the first of the two independent short studies they were about to
perform. "Study 1" represented the affect manipulation while "Study 2" captured the
product benefit manipulation, followed by the main dependent measure. After presenting
students with an informed consent form, the first screen of study 1 introduced the cover
story. Respondents were told that as the number of Internet classes was increasing, the
university had decided to investigate the impact on memory of material transmitted over
the web. Respondents were instructed to watch a video on the web and describe a real life
experience similar to that watched in the film (affect manipulation). Then, they were
asked to rate the video (manipulation check). As soon as they finished study 1 a new
screen introduced the cover page of the "second" study. An informed consent form was
once again provided to the students in order to convey the idea of two independent
studies. The next screen indicated that the study attempted to assess people's willingness
to buy a specific product in a given consumption scenario. They were presented with a
consumption scenario in which the product (coffee vs. cereal), and its benefits (mood-
lifting vs. non mood-lifting) varied across conditions. After 15 seconds spent reading the
information, respondents were provided with a nine-point scale where they indicated their
purchase intentions (dependent measure). On the next page, they were asked about the
information earlier presented in the purchasing scenario (manipulation check for the
product benefits manipulation). Finally, they described the purposes of studies 1 and 2
(hypothesis guessing check) and were later debriefed regarding the affect manipulation in
order to eliminate any residual effects.
To vary respondents' affective states, a combined technique was used (movie plus
personal real life description). This same technique has been successfully used in both the
general affect literature and in consumer research on affect (Cohen and Andrade 2004;
see also Westermann et al. 1996 for a meta-analysis). Respondents watched five minutes
of either a sad sequence of the movie Top Gun (negative affect), a documentary about
Italy (neutral affect) or a happy sequence of the movie American Pie 2 (positive affect).
The second step of the induction asked respondents in the valenced affect conditions to
describe a real life experience that produced the same feelings as those derived from the
video. In the neutral affect condition respondents were asked to describe a real life
experience that was similar in content to the scenes watched in the video. Finally,
respondents were asked to give their opinion about the video (manipulation check). Ten
items were presented in a nine-point semantic differential scale format. Three of the 10
items were designed to assess respondents' current affective states (I felt sad-I felt happy,
It's depressing-It's upbeat, Created a negative mood-Created a positive mood). The order
of the items was randomized to avoid order effects.
The products were selected based on published findings, consumer beliefs, and
companies' attempts to link their products to mood-lifting qualities. Caffeine and
carbohydrates have been shown to produce a positive impact on mood and/or arousal
(Smith, Clark, and Gallagher 1999), and descriptive studies have shown, for instance, that
drinking coffee or other caffeinated products is one of people's common mood repair
strategies (Thayer 1996). Also, companies selling caffeine (e.g., coffee shops) and
carbohydrates (e.g., cereal companies) have explicitly mentioned their products' mood-
lifting attributes. As already mentioned, one of Kellogg's websites, for instance, contains
a section called "Breakfast and Mood", in which the mood-regulating properties of
breakfast cereals are highlighted: "Several studies have shown that consuming high levels
of carbohydrates is associated with better mood. As most breakfast cereals are high in
carbohydrates, this offers one possible reason why eating a high carbohydrate breakfast is
associated with improved mood."
Product Benefits Manipulation
Respondents were presented with a short scenario in which they had to imagine
themselves in a purchasing environment where, after reading the information available
about the product, purchase intentions were to be indicated. For those exposed to the
coffee scenario, they were asked to imagine themselves in a coffee shop with a friend
who planned to buy a cup of coffee. While they waited in line, he/she (the respondent)
started reading a brochure about the benefits of caffeine. In the mood-lifting condition,
the brochure stated that ". Popular culture is in fact correct. A research study
conducted at UCLA showed that caffeine works as an effective mood regulator. Taken in
small doses-like a cup or two a day-, caffeine is a very effective mood lifter." In the
no mood-lifting condition respondents were presented with the following statement "...
A recent study conducted at UCLA showed that coffee is indeed a healthy product. Taken
in small doses like a cup or two a day -, caffeine reduces chances of heart failures as
well as neurological diseases." Those exposed to the breakfast cereal were asked to
imagine themselves in a nearby supermarket. In the cereal aisle one particular cereal box
had a message printed on the back. In the mood-lifting condition, the message was the
following: "Several studies have shown that consuming high levels of carbohydrates is
associated with better mood. As cereal is high in carbohydrates, people who eat cereal
have a 'happier diet!' In the condition where no mood-lifting benefits were
highlighted, the message contained the following statements: "It is widely known that
carbohydrates are an important nutrient. As cereal is high in carbohydrates, don't miss
this opportunity to get those much needed nutrients every morning!" Across all
conditions students no information about the brand was provided.
Students had to wait at least 15 seconds before the purchase intention scale
appeared on the screen, which allowed some control for unwanted processing style
effects. As respondents in a negative (vs. positive) mood are usually more careful in their
analysis (Schwarz 2001), we wanted to make it easy for respondents to read the
information regardless of their affective states. After reading all the information about the
product benefits, they were presented (at the bottom of the screen) with a nine-point scale
on which they indicated their willingness to buy the product (9 = I would definitely buy
Eight students provided a rudimentary guess of the purpose of the study. They were
deleted from the sample. After checking for reliability (a = .92), the three affect-related
items were collapsed to form the affect index. The affect manipulation produced a
significant main effect on people's affective states (F(2, 273) = 301.70,p < .001).
Pairwise comparisons showed that compared to the neutral affect condition (M= 6.2),
respondents in the negative affect condition experienced more negative feelings (M= 2.9;
F(1, 190) = 339.49, p < .001), whereas respondents in the positive affect condition
evaluated their affective state more positively (M = 7.5; F(1, 185) = 40.29, p < .001).
To assess the product benefits manipulation, respondents were asked, at the end of
the second study, to indicate (using a nine-point scale item) the extent to which they
agreed with the statement that the product contained mood-lifting benefits. As expected,
those in the mood-lifting condition perceived the products on average as more mood-
lifting (M= 5.4) than those in the non mood-lifting condition (M= 4.0; F(1, 274) =
33.21, p < .001). Moreover, the pattern remained the same across all affect conditions (F
Product replicate did not interact with product benefit and affective state (F(2, 264)
=.99, p = .37), product benefit (F(1, 264) = 1.06, p = .30) or affective state (F(2, 264) =
.52, p = .59), so data were collapsed across product replicates (i.e., coffee and cereal). All
subsequent analyses followed a three (affect: negative vs. neutral vs. positive) by two
(product attributes: mood-lifting vs. no mood-lifting) between respondents design.
As predicted, respondents' affective state interacted with the product benefits
highlighted in the shopping scenario (F(2, 264) = 11.38, p < .001)(Figure 3).
o 5.18 5.125.12
S4.22 mood-lifting benefits
4 [0O non mood-lifting benefits
Negative Neutral Positive
Figure 3. Behavioral intentions toward coffee and cereal (collapsed)
According to IMAB, the impact of affect on behavior via the affective evaluation
mechanism should be strongest when respondents do not have enough diagnostic
information to make their evaluations and when there are no affect-regulation
opportunities in the environment (i.e., non mood-lifting condition). In this type of
scenario, a monotonic increase in behavioral intentions should emerge when feelings
move from negative, to neutral, to positive, as a result of the use of affect-as-information
and/or mood-congruency during the evaluation process. The results confirmed the
model's predictions. When respondents were presented with a brand of coffee or cereal
with no mood-lifting benefits, there was a monotonic increase on purchase intentions as a
function of experienced affective state (F(1, 143) = 9.57, p < .005). Pairwise comparisons
showed that respondents experiencing negative affect were more reluctant to buy the
product (M= 3.7) compared to respondents in a neutral affect condition (M= 4.9; F(1,
101) = 8.08, p = .005), as well as compared to respondents experiencing positive affect
(M= 5.1; F(1, 85) = 9.60,p < .005). However, there was no difference between the
neutral and positive affect condition (F < 1).
IMAB hypothesizes that a "U" shaped pattern should emerge when respondents
are asked to evaluate a product with mood-lifting benefits. In this case, respondents in a
negative mood faced an affect-regulation opportunity. As a result, the affect-regulation
mechanism was expected to increase purchase intentions despite the fact that the affective
evaluation mechanism, by itself, would tend to lower purchase assessment judgments. As
expected, when respondents were presented with a brand of cereal or coffee with mood-
lifting benefits, their purchase intentions exhibited a quadratic form (F(1, 127) = 4.66, p
< .05). Pairwise comparisons showed that when facing a product with mood-lifting
benefits, respondents experiencing negative affect were more willing to buy the product
(M= 5.2) than those in a neutral mood (M= 4.2; F(1, 87) = 3.78, p = .05). Respondents
experiencing positive affect (M= 5.1) marginally increased their purchase intentions
compared to a neutral condition (F(1, 83) = 3.12, p = .08).
The findings provide initial support for our hypotheses. The expected monotonic
increase in purchase intentions towards an unknown brand of coffee or cereal emerged as
respondents' feelings improved. This pattern, importantly, is contingent on the
availability of affect-regulation opportunities. When the scenario highlighted the mood-
lifting benefits of the product, then the predicted "U" shape curve emerged. As expected,
respondents in a valenced affective state (positive or negative) were more willing to
purchase the product than those in a neutral affective state, but for different reasons.
When respondents experienced negative affect there was a tendency toward upward
affect-regulation, and respondents approached the stimulus that might help them
accomplish this goal. When respondents were in a positive affective state, they also
increased their purchase intention, most likely due to a more positive evaluation of the
product/purchase scenario. Notice that the theoretical accounts available in the literature
cannot explain this combination of effects, which incorporate three out of the four
categories of the affect-behavior relationship described in Figure 1 (Cells 1, 3, and 4).
It is worth noting that although the results provided general support for the
proposed model, positive affect (compared to the neutral affect condition) led to an
increase in purchase intentions within the mood-lifting condition, but not within the non
mood-lifting condition. Therefore, drawing the conclusion that respondents in a positive
affective state were insensitive to mood-lifting benefits is premature. On the one hand, it
is possible that affect-regulation may have also played a role on the positive side of the
affective spectrum, since the difference between neutral and positive affect condition was
significant only among those facing a mood-lifting opportunity. On the other hand, it is
also possible that a "heavy handed" mood-lifting manipulation could have made
respondents within the mood-lifting conditions more sensitive to their feelings, thereby
making affect-mediated effects more likely. A less intrusive manipulation, where no
explicit description of the mood-lifting properties of the product is highlighted, should
eliminate this "focus of attention bias" and produce similar effects for the mood-lifting
and non mood-lifting conditions when positive and neutral affect conditions are
Also, as experiment 1 employs only an indirect assessment of mediating processes,
it is difficult to detect when the two mechanisms operate simultaneously. More direct
assessments may help to disentangle the effects of evaluation and regulation when both
can predict the same outcome (i.e., increase in purchase intentions for those experiencing
positive affect). By the same token, we cannot determine the extent to which affective
evaluation and affect-regulation interacted when sad people faced an affect-regulation
opportunity. One might wonder whether heightening the activity of the affect-regulation
mechanism "switches off' a less intense affective evaluation effect or, on the other hand,
whether affect-regulation simply increases the weight of the product's mood-lifting
Experiment 2 attempted to replicate the first experiment, while addressing the
concerns discussed above. First, to address a possible focus of attention bias (i.e., the
likelihood that respondents in the mood-lifting conditions became more responsive to
feelings), we looked for a scenario in which people's intrinsic perceptions of mood-lifting
properties of the product could vary without any sort of explicit manipulation. The
helping literature has provided precedent for such a procedure. Cialdini and Kenrick
(1976) showed that whereas adults in a bad mood are more willing to help others
(probably in an attempt to regulate their negative mood), the effects reverse among
children. Their explanation is that adults are more likely to perceive the mood-lifting
benefits associated with helping, whereas children are yet to learn such associations.
Following the same rationale, research on eating behavior shows that men and
women have different beliefs about food intake in general, and chocolate in particular.
First, women are more likely to use food consumption as a mood-lifting alternative than
are men (Macht 1999; Steptoe, Pollard, and Wardle 1995). Also, evidence suggests that
women are more likely than men to perceive chocolate as a mood-lifting product
(Benton, Greenfield, and Morgan 1998; Grunberg and Straub 1992). If this is the case,
then women in a negative affective state should be more willing to try a piece of
chocolate (compared to a neutral affect condition), as a result of the affect-regulation
mechanism. Notice that this manipulation relies on one of the basic tenets of the model,
in which the impact of the affect-regulation mechanism is contingent on people's
intuitive theories about the affective consequences of their behavior (e.g., the mood-
lifting benefits of chocolate).
Among men, who are not as likely to perceive the affect-regulation opportunities of
chocolate, negative affect should lead to a decrease in behavioral intentions compared to
the neutral affect condition (via the affective evaluation mechanism). For those already
experiencing positive affect, both women and men should be equally willing to taste a
piece of chocolate (via the affective evaluation mechanism). In other words, across the
three levels of affect, a "U" shape pattern is expected among women due to a
combination of affect-regulation within the negative affect condition and affective
evaluation within the positive affect condition, whereas a monotonic increase is expected
among men due mainly to the impact of the affective evaluation mechanism.
Second, to provide more direct evidence about the two mediating processes
hypothesized by IMAB, an open-ended question asked respondents to explain their
indicated behavioral intentions. Additional items at the end of the experiment also
examined the independent impact of the evaluative and regulatory mechanisms. Finally, a
new neutral affect condition was used to generate a more symmetrical difference across
the three levels of affective states.
Subjects and Design
One hundred fifty-one undergraduate students from a southeastern university
participated in the experiment in exchange for course credit. The study adopted a two
(gender: men vs. women) by three (affective state: negative vs. neutral vs. positive)
between subjects design. Respondents were randomly assigned to one of the six
The procedure was similar to that presented in experiment 1. A two-studies cover
story was introduced. "Study 1" manipulated respondents' affective states. A new video
("Documentary about John Nash") was presented to students in the neutral affect
condition to get a more "neutral point" on the scale. The other two videos were the same
as in the first experiment. In the "second study" respondents were informed that a foreign
company was about to introduce a new chocolate in the American market. The gist of the
story was that as sampling promotions in supermarkets turn out to be quite expensive, the
company had decided to use a new marketing tool, a so-called "Virtual Sampling
Promotion". Respondents were then instructed to imagine themselves in a real sampling
promotion scenario. A picture of chocolate bars was presented, and they were asked to
indicate the extent to which they would try the product. To attach some sort of cost to the
behavioral activity, since these were free products, respondents were told to imagine that
they would have to answer a 6-minute survey if they decided to taste it. After indicating
their willingness to try the product along a 9-point scale (9 = I would definitely try it),
respondents were asked to explain their indicated behavioral intentions. Respondents
were then presented with a series of items, which assessed the manipulations and the
impact of the mediating processes.
To confirm the assumption that women are more likely than men to perceive eating
chocolate as an affect-regulation opportunity, people's instrumental use of chocolate was
assessed with a single item "I eat chocolate to feel better". Another item assessed the
impact of affective evaluation on the cost associated with the behavior (i.e., answering
the survey) "I was a bit concerned that it might take too long to answer the
questionnaire". Finally, to check for potential confounding, additional items checked the
extent to which participants were hungry and thirsty, whether or not they had heard of the
particular brand before, and if the product had made them think of other products that
could have influenced their evaluations.
Six students provided a rudimentary guess of the purpose of the experiment. They
were deleted from the sample. After checking for reliability (a = .91), the three affect-
related items were collapsed to form the affect index. The affect manipulation produced a
significant main effect on people's prior affective state (F(2, 142) = 124.69, p < .001).
Pairwise comparisons showed that compared to the neutral affect condition (M= 5.2),
respondents in the negative affect condition experienced more negative feelings (M= 3.3;
F(1, 98) = 46.76, p < .001), whereas respondents in the positive affect condition
evaluated their affective state more positively (M = 7.8; F(1, 98) = 84.98, p < .001).
Finally, the assumption that women are more likely to perceive chocolate as a mood-
lifting product was supported. Female respondents were much more likely to
acknowledge eating chocolate to feel better than were male respondents (Momen = 5.7 vs.
Amen= 2.7; F(1, 143) = 59.34, p < 001).
As hypothesized, gender and affective state produced a significant interaction (F(2,
139) = 5.09, p < .01). Most importantly, breaking down the analysis by gender, a
monotonic increase was observed among men F(1, 80) = 15.29, p < .001), whereas a "U"
shape pattern emerged among women (F(1, 59) = 9.97, p < .005)(Figure 4).
54.79 O Men
Negative Neutral Positive
Figure 4. Behavioral intentions toward chocolate
IMAB proposes that as men are less likely to perceive the mood-lifting benefits of
eating chocolate, the affective evaluation mechanism should direct their behavior, leading
male respondents in positive (negative) affect to increase (decrease) their behavioral
intentions, compared to the neutral affect condition. Pairwise comparisons indicated that,
when facing an opportunity to try a piece of chocolate, men in a sad mood (M= 4.8)
seemed less willing to do so than those in the neutral affect condition (M= 5.9; F(1, 54)
= 2.76, p = .10). The opposite was true for those in the positive affect condition, who
indicated higher intentions (M= 7.2) to try the chocolate in the same hypothetical
scenario (F(1, 57) = 5.6,p < .05).
Among women, the affect-regulation mechanism was hypothesized to have a strong
influence on behavioral intentions, since this group is more likely (as confirmed by the
manipulation checks) to perceive eating chocolate as an affect-regulation opportunity. In
accordance with IMAB's predictions, women experiencing negative affect indicated
stronger intentions to taste the chocolate (M= 7.4) compared to those in the neutral affect
condition (M= 6.0; F(1, 42) = 5.8, p < .05) Women experiencing positive affect also
exhibited stronger behavioral intentions compared to those in the neutral affect condition
(M= 7.7; F(1, 39) = 6.9, p < .05). IMAB proposes, however, that the latter difference is
due to affective evaluation rather than regulation. Since male respondents, who reported
not using chocolate as an affect-regulation device, displayed similar patterns on the
positive side of the affective spectrum, it seems plausible to suggest that evaluation
(instead of regulation) represented the main mediating process for women under positive
Affective Evaluation Mechanism
Further evidence regarding mediating processes was obtained with the open-ended
question and the additional items. IMAB proposes that the affective evaluation
mechanism leads people experiencing negative (positive) affect to perceive the costs/risks
associated with any behavior as more (less) negative. Remember that across all
conditions an element of cost, orthogonal to the product, was included. Those choosing
to taste the chocolate were told they had to answer a 6-minute survey. Therefore, if the
affective evaluation mechanism is operating, respondents experiencing negative affect
should consider this cost element more carefully than those experiencing positive affect.
As already mentioned, one item assessed people's concern about the length of the survey.
The results show that affect produced a main effect on respondents' concerns about costs
(i.e., survey length) associated with the tasting the chocolate (F(2, 142) = 3.28, p < .05).
Although they were explicitly told that the survey would take precisely six minutes, sad
respondents (M= 6.4) were more concerned that the survey "might take too long" than
those in the neutral (M= 5.2) and positive (M= 4.9) affect conditions (F(1, 98) = 4.12, p
< .05; F(1, 88) = 6.23, p < .05, respectively). The difference between neutral and positive
affect conditions was non significant (F < 1). Also important is the fact that no interaction
between gender and affect emerged (F < 1). Thus, men and women seemed similarly
influenced by the affective evaluation mechanism.
The open-ended question also permitted tracking the impact of the affective
evaluation mechanism by counting the number of people who mentioned the survey in
their justification of their behavioral intentions. The impact of affect was quite evident.
Only 9% of respondents in the positive affect condition mentioned the survey in the
justification, compared to 33% of respondents in the neutral affect condition (Z= 3.14, p
< .01, one tailed test) and 38% in the negative affect conditions (Z= 3.45,p < .01, one
tailed test). Once again, the pattern remained the same across gender (table 1).
Table 1. Influence of affective evaluation on behavioral intention
GENDER Negative Neutral Positive
Women 33.3%a 43.5%a 5.3%b
Men 41.7%a 21.9%* 11.1%
Gender Collapsed 37.8%a 32.7%a 8.9%b
Note: Different superscripts indicate significant differences at p < .05. The asterisk
indicates marginal significance (p < .10).
Combined, these results suggest, first, that affective evaluation mitigates
(enhances) the perceived costs associated with the survey for respondents in a positive
(vs. negative) affective state, which can explain why both men and women were more
inclined to try the chocolate while experiencing positive affect. Second, these results also
imply that if this mechanism were to operate in isolation, women in the negative affect
condition should be less inclined to try the chocolate. Since women experiencing
negative affect were, on the contrary, more willing to try the chocolate than those in the
neutral affect condition (and equivalent to those experiencing positive affect), another
opposing mechanism must have reversed the effects. According to the IMAB, affect-
regulation can produce such effects.
According to the model, affect regulation is contingent on people's beliefs about
the affective consequences of a particular behavior (e.g., "I eat chocolate because I know
it will make me feel better"). Although, in general, women were found to eat chocolate to
feel better much more often than men, there was a certain degree of variance within each
gender condition. Thus, we should expect a decrease in behavioral intentions for those
women who acknowledged they were less likely to eat chocolate to feel better, and an
increase in behavioral intentions for those men who acknowledged they were more likely
to eat chocolate to feel better.
A series of 6 bivariate Pearson correlations (one per condition) between
respondents' acknowledged use of chocolate as a mood-lifting product and their
respective intentions to taste the chocolate was conducted. The correlation should be high
(low) in the conditions where affect-regulation is expected to have a strong (weak)
influence. Table 2 summarizes the results.
Table 2. Influence of acknowledged use of chocolate to lift mood on behavioral intention
GENDER Negative Neutral Positive
Women 0.49** 0.04 0.04
Men 0.37* 0.28 0.14
Note: Double-asterisks indicate Pearson correlation significant atp < .05 level (2-tailed).
Single asterisks indicate Pearson correlation significant atp < .10 level (2-tailed).
As can be observed, the strongest effect (i.e., the only significant result at p < .05)
emerged when female respondents were experiencing negative affect, the condition in
which affect-regulation was hypothesized to direct behavioral intentions. In other words,
intentions to try the chocolate were lowered for female respondents who acknowledged
not using chocolate to regulate their affective states. A marginal effect was also found
among men in the negative affect condition, suggesting that those men who are more
likely to eat chocolate to lift their mood were more willing to try the chocolate. Among
those experiencing neutral or positive affect, there were no significant correlations, which
rules out regulatory influences of affect beyond the negative side of the affective
spectrum. These results not only challenge the mood-maintenance type of explanation for
those experiencing positive affect but also support IMAB's assumption that affect-
regulation is less likely within neutral affect conditions.
By relying on people's perceptions of the product's mood-lifting properties rather
than using a more "heavy handed" manipulation, experiment 2 successfully replicated the
earlier results while eliminating the focus of attention bias inherent in the first
experiment. Most importantly, it also shed additional light on the two mediating
processes assumed to link affect to behavioral intentions.
First, among respondents experiencing positive affect, the affective evaluation
mechanism seems to have a predominant role, since the impact of the survey (i.e., cost
associated with the behavior) weakened as respondents' feelings improved. Moreover,
affect-regulation seems inactive for happy respondents since, unlike for sad respondents,
(1) there was no correlation between respondents' acknowledged use of chocolate as a
mood-lifting option and their respective behavioral intentions and (2) intentions to try the
chocolate also increased among male respondents, who, based on self reports, were in
general unlikely to eat chocolate in an attempt to regulate their mood.
Second, affect-regulation seems to have a predominant role among the women who
experienced negative affect, since their behavioral intentions increased (compared to
those in neutral conditions), and there was a significant positive correlation between the
use of chocolate as a mood-lifting product and their behavioral intentions. In other
words, behavioral intentions were higher for women who were more likely to perceive
eating chocolate as an affect-regulation opportunity. Notice that these results also support
the assumption that as neutral affective states trigger weak signals, the discrepancy
between the current and the ideal affective state becomes less salient, which results in
minimal affect-regulation tendencies.
Finally, the interaction between the two mechanisms can be observed in negative
affect conditions. Independent of gender, the costs associated with the survey increased in
the negative (positive) affect condition, which indicated the negative impact of the
affective evaluation mechanism. Nonetheless, female respondents were highly willing to
taste the chocolate due to its regulatory properties, as indicated by the correlational
analysis. Such opposing patterns suggest, therefore, that affect-regulation does not
necessarily "switch off' the affective evaluation mechanism, but instead dominates its
opposing effect, probably by increasing the weight given to a product's mood-lifting
So far, IMAB has demonstrated and explained three out of the four types of effects
presented in Figure 1. The potential inhibitory consequences of positive affect (i.e., Cell
3) remain to be examined. Can positive affect mitigate people's willingness to eat
chocolate? According to model, the affect-regulation mechanism is more likely to guide
behavior if people can anticipate the affective consequences of the behavioral activity.
When people in a bad mood anticipate that their behavior will make them feel better, they
become more likely to act. Experiments one and two confirmed that hypothesis. It
follows that when people in a good mood anticipate that their behavior will make them
feel worse, they should become less likely to act than those in the control condition,
simply because they have more to lose. Research in risk taking has provided support for
this rationale (see Isen 2000 for a review).
Experiment 3 focuses therefore on the facilitatory (Cell 1) and the inhibitory (Cell
2) behavioral consequences of positive affect. A major procedural change (compared to
the previous experiments) is that a direct behavioral measure rather than intention is used
as the main dependent measure. Therefore, respondents faced actual, instead of projected,
behavioral consequences throughout the experiment. Similar to experiment 2, chocolate
represents the target stimulus; however, in this final experiment actual consumption is
assessed. Popular culture as well as the scholarly research largely support the assumption
that chocolate consumption can be associated with aposteriori negative affect, notably
feelings of guilt (Benton et al. 1998; Cramer and Hartleib 2001; Macdiarmid and
Hetherington 1995). Thus, to create a mood-threatening cue condition that may lead
people experiencing positive affect to be less likely to act, a group of respondents is
reminded of the negative consequences of the particular behavior (i.e., fat and fat calories
associated with chocolate consumption). IMAB predicts that happy respondents facing a
potential affective threat (subsequent feelings of guilt) will eat less chocolate (compared
to the respective neutral affect condition), due to the protective response of the affect-
regulation mechanism. When no mood-threatening cues are made salient, happy
respondents are expected to consume more chocolate than the control condition as a
result of the affective evaluation mechanism (i.e., a better evaluation of the environment).
Since the fat components of chocolate are likely to be perceived as a negative attribute by
both men and women, gender should not interact with the other two factors (i.e., affect
and mood-threatening cue).
Subjects and Design
One hundred sixty-seven undergraduate students from a southeastern university
participated in the experiment in exchange for course credit. The study adopted a two
(affective state: neutral and positive) by two (mood-threatening cue: salient vs. not
salient) between subjects design. Respondents were randomly assigned to one of the four
To keep the affect manipulation similar to the previous studies, the cover story
focused on the impact of video on memory, but it added information identifying potential
interactive effects of food consumption. As in experiments 1 and 2 respondents were
presented with a five-minute program (an episode of"Friends" vs. a documentary about
John Nash), and were asked to recall a personal experience related to film (affect
manipulation) as well as to evaluate the video. Three items embedded into the scale
assessed respondents' current affective state (affect manipulation check). Then
respondents were instructed to open an envelope placed to the left of the computer
monitor. Inside was 89-gram package of M&Ms. A five-minute sequence of the previous
movie was then presented on the screen and respondents' task was simply to watch the
video while eating "as many or as few M&Ms" as they wanted. Food consumption while
watching a movie represents a common data collection procedure in the eating behavior
literature (Cools, Schotte, and McNally 1992; Grunberg and Straub 1992). Each M&M
package was weighed before and after the experiment to ensure an accurate measure. The
amount of M&Ms consumed became the dependent variable.
To manipulate the mood-threatening cue associated with eating behavior an
additional instruction page was given to one subgroup before they started watching the
video and eating the chocolate candies. These respondents were informed that as different
experimental groups tasted different products, the experimenter had to control for the
number of calories consumed. Therefore, respondents should look for the nutrition facts
table on the M&M package and enter the number of calories and fat calories of this
particular product on the screen. In the control group, no nutrition information about the
product was requested from respondents.
After watching the second part of the video, respondents were instructed to put the
M&M package back into the envelope and answer some final questions/fill out some
scales. An additional memory type of question was asked to complete the cover story
followed by another affect manipulation check, also embedded into the movie evaluation
scale. Then, in an attempt to gather additional insights into people's motivations, an
open-ended question about the potential reasons that stimulated or impeded their actions
(i.e., chocolate consumption) followed. Similar to experiment 2, respondents also
responded to items referring to their general consumption habits and perceived
relationships between mood and chocolate consumption. To assess the impact of potential
covariates, a Restraint Questionnaire (Herman and Polivy 1975; Polivy, Herman, and
Warsh 1978), was presented, which assesses people's chronic attempt to keep strict
control over their eating behavior. Respondents' body mass index (BMI) was also
calculated, and the time of data collection was also recorded. All experimental sessions
took place between 2:00 pm and 4:30 pm.
Two students provided a rudimentary guess as to the purpose of the experiment.
They were deleted from the sample. After checking for reliability (a = .88), the three
affect-related items were collapsed to form an affect index. The affect manipulation
produced a significant main effect on respondents' prior affective state. Respondents
exposed to first five-minute exposure of the video plus the description of a personal
experience were happier when they watched an episode of friends (M= 7.9) than when
they watched a documentary about John Nash (M= 5.6; F(1, 163) = 161.32, p <.001).
The mood-threatening cue manipulation was assessed by the number of respondents who
entered the calories and fat calories components of their respective M&M package in the
requested blank spaces. As expected, all respondents in the salient mood-threatening cue
condition correctly followed the instructions. Neither the body mass index nor the
Restraint Scale produced any impact on eating behavior and were not incorporated in
further analyses. Similarly, eating behavior remained the same regardless of the time of
the experiment (2pm, 3pm, or 4pm). Gender produced a marginal simple main effect on
chocolate consumption, with male respondents eating more (M = 26.3) than female
respondents (M= 21.8; F(1, 163) = 3.70, p <.10). However, as predicted, no three-way
interaction was observed (F(1, 157) = 2.59, p >.10). The interactions between gender and
affective state and gender and mood-threatening cue were also non significant (F<1).
As hypothesized, mood-threatening cue (salient vs. not salient) and affective state
(positive vs. neutral) produced a significant interaction (F(1, 161) = 8.25,p =.005)(Figure
S25 23.33 23.52
20 18.05 neutral affect
5 20 18.05
non salient mood salient mood
threatening cues threatening cues
Figure 5. Amount of chocolate consumed
When respondents were asked to eat as much or as little chocolate as they wanted
and no mood-threatening cues associated with the product/behavior were highlighted,
respondents experiencing positive affect ate more M&Ms (M=30.9) than those in the
neutral affect condition (M = 23.3; F(1, 81) = 4.42, p <.05). The results reversed when
the calories and fat calories associated with M&Ms (i.e., the mood-threatening cues of the
product/behavior) were made salient prior to consumption. In this case, those in the
positive affect condition ate significantly fewer M&Ms (M= 18.0) than did those in the
control condition (M= 23.5; F(1, 80) = 3.96, p =.05). This interaction not only replicates
the impact of positive affect on action stimulation (Cell 1) with a direct behavioral
measure, but also, and most importantly, demonstrates that positive affect can also lead to
inhibitory actions (i.e., less consumption) when the behavior threatens people's current
affective state (Cell 2).
IMAB proposes that people's anticipated (post behavior) affective state can
promote or inhibit consumption via the affect-regulation mechanism. Therefore,
respondents experiencing positive feelings, having more to lose than those in a control
condition, should eat less chocolate when the mood-threatening consequences of the
behavior are made salient. This rationale is based on the assumption that chocolate can
trigger guilt feelings aposteriori and that people can anticipate them (Benton et al. 1998;
Cramer and Hartleib 2001; Macdiarmid and Hetherington 1995), which makes happy
people more conservative in their actions, in an attempt to protect the current pleasant
To assess the impact of guilt, respondents were asked (at the end of the experiment)
to indicate the extent to which they generally had guilt feelings after overeating (a four
point scale from never to always). This item was extracted from the Restraint
Questionnaire (Herman and Polivy 1975; Polivy et al. 1978). A correlational analysis
(within each of the four cells) between the grams of chocolate consumed and people's
acknowledgement of experiencing guilt feelings after overeating was then performed.
Table 3 summarizes the results.
Table 3. Protective influence of overeating-based guilt feelings on behavior
MOOD-THREATENING Ne l
Yes -.12 -.30*
No -.01 .07
Note: The asterisk indicates Pearson correlation significant atp < .10 level (2-tailed).
When the mood-threatening cues associated with the product/behavior were not
made salient (i.e., cells at the bottom row), respondents' guilt feelings after overeating
did not correlate with the amount of chocolate consumed. Similarly, when people
experienced neutral affect, thereby having little to lose and/or a weak signal (i.e., cells at
the left column), the correlation was also non significant. However, when respondents
experiencing positive affect, thereby having a lot to lose and/or a strong signal, are given
the mood-threatening cues for chocolate consumption, a marginally significant negative
correlation emerged (r = -.30, p = .064). In other words, within this condition, those who
often or always feel guilty after overeating (M= 16.2) tended to eat significantly less
chocolate compared to those who rarely or never experience these negative feelings after
overeating (M= 21.3, F(1, 36) = 3.10, p < .10)(Figure 6).
Never Rarely Often Always
Feelings of Guilt after Overeating
Figure 6. Amount of chocolate consumed within the positive affect /mood-threatening
condition the impact of expected feelings of guilt after overeating
Within the positive affect no mood-threatening cue condition, an increase in
chocolate consumption emerged compared to the respective neutral affect condition.
Although IMAB predicts this impact to be mostly driven by the affective evaluation
mechanism, others (Clark and Isen 1982) suggest that mood-maintenance (i.e., affect-
regulation) could also be responsible for the effects. In experiment 2, this alternative
explanation was ruled out by showing that within the positive affect conditions there is no
correlation between respondents' acknowledgement of the use of chocolate as a mood-
lifting strategy (i.e., I eat chocolate to feel better) and their willingness to try the
chocolate. In experiment 3 two items were used to test this potential correlation (i.e., "I
eat chocolate to cheer myself up"; "Eating chocolate when I'm happy helps me to
maintain my good mood"). If a positive correlation emerged between chocolate
consumption and these two items, then one can claim there is initial evidence for affect-
regulation mechanism. If no correlation exists, affect-regulation probably did not play a
major role. None of the items correlated with chocolate consumption (p > .50 for both
items), which replicates the results presented in experiment 2 using an actual behavioral
measure. In other words, the extent to which respondents consume chocolate as a mood-
lifting or mood-maintenance strategy did not lead them to eat more or less. Therefore, as
predicted by IMAB, the affect-regulation mechanism seems not to have a major impact
on the increase of chocolate consumption in the absence of mood-threatening cues.
Instead this effect was probably guided by a more positive evaluation of the
The main purpose of the third experiment was to assess the extent to which positive
affect can inhibit behavior (Cell 2 of Figure 1). IMAB predicts that this can occur once
mood-threatening cues are made salient, thereby leading happy individuals to adopt a
more conservative response in an attempt to protect their current enjoyable feelings.
Using actual chocolate consumption, it was shown that respondents experiencing positive
affect ate less chocolate (compared to the neutral affect condition), once the fat and fat
calories of the product were made salient prior to consumption. Moreover, this effect
seemed to intensify (vs. weaken) among respondents who acknowledge being more (less)
likely to feel guilty after overeating. Notice that, similar to experiment 2, this effect is
based on IMAB's assumption that the affect-regulation mechanism is contingent on
people intuitive theories about the affective consequences of their respective behavior. In
experiment 2, those within the negative affect condition who expected to feel better after
the behavior were more likely to try the chocolate. In experiment 3, the reverse rationale
applied. Those within the positive affect condition who expected to feel worse after the
behavior were less likely to eat chocolate.
When the mood-threatening cues (i.e., calories and fat calories of M&Ms) were not
highlighted, happy respondents ate more chocolate than did those in the neutral affect
condition. Moreover, within this particular Cell there was no correlation between
people's acknowledged use of chocolate for mood-lifting or mood-maintenance purposes
and the grams of chocolate consumed. Thus, similar to experiment 2, and as predicted,
affect-regulation was not responsible for the effect, suggesting that affective evaluation
was the operative mechanism.
The behavioral consequences of affect can be categorized into four groups: positive
affect encouraging action, positive affect inhibiting action, negative affect encouraging
action, and negative affect inhibiting action. The several theories and hypotheses
developed since the early 1970's have focused prominently on one, or at most, two of
these four potential effects (Figure 1), thereby providing no parsimonious explanation of
how affect influences behavior. To fill this theoretical gap an integrative model of
affective behavior (IMAB) was proposed, in which two well established properties of
affect (i.e., evaluative/informational and regulatory/motivational) represent the
potentially opposing mediating mechanisms responsible for the behavioral consequences
of individuals' current affective states. Moderating variables attached to each of these
mechanisms determine which mediating process prevails and consequently guides
In a series of three experiments, it was demonstrated that people's current affective
state represents a critical moderator for both the affective evaluation and affect-regulation
mechanisms, where valenced affective states tend to lead to stronger behavioral effects
compared to control conditions. Most importantly, the expected mood changing
characteristic of the behavioral activity (i.e., the mood-lifting cues in experiments 1 and 2
and mood-threatening cues in experiment 3) and its interaction with people's current
affective state help to determine the extent to which the affect-regulation mechanism will
prevail over the affective evaluation mechanism, and eventually direct behavior.
Rooted in the mood-congruency and affect-as-information hypotheses, the affective
evaluation mechanism predicts that people will congruently evaluate any behavioral
activity. As a result of a congruent evaluation, happy (sad) people become more (less)
likely to act, especially when there is no competing, more diagnostic information
available in the environment. Experiment 1 (within non mood-lifting conditions) and 2
(among men) confirmed the effect. Experiment 3 (within non mood-threatening
conditions) further replicates it using a direct behavioral measure. These affective
evaluation-mediated effects characterize Cells 1 and 4 of Figure 1.
The affect-regulation mechanism, however, may come into play as a counteracting
force if certain conditions are met. For those experiencing negative feelings while facing
a potential mood-lifting opportunity (i.e., having their affective state salient in memory
(Cohen and Andrade 2004) and having much to gain effectivelyy) via the behavioral
activity, affect-regulation encourages action, overcoming the opposing impact of the
affective evaluation mechanism. Experiments 1 (within the mood-lifting conditions) and
2 (among women) confirm these predictions. Respondents under such conditions
indicated they were more likely to purchase coffee and breakfast cereal (experiment 1),
and to taste chocolate in an imaginary situation (experiment 2) than those in the
respective neutral affect conditions. For those experiencing positive affect while facing a
mood-threatening scenario, (i.e., having their affective state salient in memory, but also
having more to lose), affect-regulation discourages action. This can overcome the
stimulating effects of the affective evaluation mechanism. Experiment 3 confirms this
prediction by showing that when happy respondents are required to monitor the calories
and fat calories associated with chocolate prior to consumption they eat less chocolate
than those in the neutral affect condition. These affect-regulation-mediated effects
characterize Cells 3 and 2 of Figure 1.
IMAB incorporates the basic rationale of two groups of theories already well
established in the literature, but that have followed independent routes. One the one hand,
it fully incorporates the basic assumptions advanced by the affect-as-information
hypothesis into the affective evaluation mechanism. By the same token, it combines
central aspects of several of the affect-regulation types of theories into a newer and more
comprehensive affect-regulation mechanism. As a result of such integration IMAB has
the ability to account for a broader array of effects described in the literature. However, it
does not necessarily invalidate other accounts. For instance, the fact that the affect-as-
information or mood-congruency hypotheses cannot explain the impact of negative affect
on the increase of purchase intentions does not make these theories inaccurate in any
sense; it simply shows that there are some effects that go beyond the scope of these
models. Put simply, Cells 2 and 3-types of effects do not provide empirical evidence
against theories focused on explaining the effects of Cells 1 and 4. In this sense, IMAB
complements more than competes against the current models, and it may be preferable on
the basis of offering a more parsimonious and internally consistent account.
Nevertheless, the proposed theory does provide a competing explanation for
behavior that has been linked to several existing theories, such as the mood-maintenance
hypothesis (Clark and Isen 1982). Mood-maintenance proposes that people's willingness
to maintain their current affective state will lead them not only to protect themselves
against mood-threatening behavior (Cell 2), but also to promote mood-lifting behavior,
such as helping (Cell 1). Whereas IMAB recognizes and empirically confirms the former,
it raises questions as to the latter. It was shown that the increase in intentions
(experiment 2) and actual consumption (experiment 3) of a mood-lifting product among
happy people (compared to the neutral affect condition) were most likely driven by
people's positive assessment of the environment (i.e., affective evaluation), rather than a
systematic attempt to act in order to keep a current positive affective state (i.e., affect-
regulation). In experiment 2, it was shown that men in the positive affect condition were
more likely to eat chocolate than in the control condition, despite the fact that they did
not perceive chocolate as a mood-lifting product. Also, a correlation between the extent
to which chocolate is used as a mood-lifting strategy and respondents willingness to try
the product did not emerge for both men and women within the positive and neutral
affect conditions, but did appear for both men and women within the negative affect
condition. Finally, in experiment 3 there was an increase in actual chocolate consumption
from neutral to positive affect conditions. However, people's self-rated use of chocolate
to keep their good mood or to improve it did not correlate with actual consumption.
Since existing empirical evidence for mood-maintenance under Cell 1 (i.e.,
promote a mood-lifting action to keep a good mood) is entirely consistent with the IMAB
and does not require a separate mood-maintenance heuristic, IMAB dominates mood-
maintenance as an explanation for this type of behavior. As shown throughout this
research, a better/safer evaluation of the environment rather than the need to keep a good
mood seems more likely to lead people experiencing positive affect to be more active.
However, when it comes to mood-threatening scenarios, IMAB fully converges with the
mood-maintenance hypothesis (Isen 2000 for a review), in which people become more
sensitive and more protective of their behavior (compared to a control condition),
because they have more to lose (Cell 3).
In integrating these two mechanisms, the model also advances theory also by
explicitly identifying the critical moderating variables attached to each mechanism,
particularly for affect-regulation. IMAB is also able to account for the weak/null effects
of neutral affect compared to polarized affective states. The three experiments show that
people in a more neutral affective state are usually weakly sensitive to available mood-
lifting or mood-threatening cues. Based on recent evidence and theoretical propositions
(Cohen and Andrade 2004), IMAB suggests that affect-regulation is not only a function
of the discrepancy between the current and desired affective state, but also of the strength
of the affective signal, which makes the existing discrepancy more or less salient. As
neutral affective states provide weak signals, the discrepancy is not highlighted and the
motivation for change becomes much weaker compared to polarized affective states.
Once the discrepancy is made salient, expected affective changes (i.e., from people's
intuitive theories as to the mood-lifting and mood-threatening properties of the
product/behavior) are critical to affect-regulation and follow the same theoretical
rationale for both negative and positive affective states.
Finally, IMAB points to a tempting conclusion that, compared to a control
condition (i.e., neutral affect), the affective evaluation mechanism is more likely to
influence behavior when we find happy people more willing to act (Cell 1) and sad
people less willing to act (Cell 4), whereas the affect-regulation mechanism is more likely
to have influenced behavior when we find sad people more willing to act (Cell 3) and
happy people less willing to act (Cell 2). However, it is possible that under certain
circumstances, regulation and evaluation may combine to promote or inhibit behavior,
although as to which contingencies are required remains an open question. Also, the
presumed interdependence of affective evaluation and affect-regulation implies that some
moderating variables can well have an impact on both mechanisms at the same time. For
instance, lack of diagnostic information about the behavior/environment tends to
instantiate the affective evaluation mechanism. However, it may also mitigate the impact
of the affect-regulation mechanism if this information modifies the perceived mood
changing properties of the behavior. Moreover, some of the moderating variables
described by the model (e.g., availability of competing goals, people's beliefs about their
affect-regulation skills) are also to be explored. Finally, more specific emotions can also
be incorporated into the model. Raghunathan and Pham (1999) showed that anxious
people are less likely to take risks compared to sad people. In IMAB terms, the reason is
simply that anxious people do not perceive risk-taking as a potential upward affect-
regulation opportunity whereas sad people do. In short, specific emotions may also be an
interesting way of enhancing or inhibiting the impact of the affective mediating processes
LIST OF REFERENCES
Abramson, Edward E. and Richard A. Wunderlich (1972), "Anxiety, Fear and Eating: A
Test of the Psychosomatic Concept of Obesity," Journal ofAbnormal Psychology,
79 (June), 317-321.
Aderman, David (1972), "Elation, Depression, and Helping Behavior," Journal of
Personality and Social Psychology, 24(1), 91-101.
Arkes, Hal R., Lisa T. Herren, and Alice M. Isen (1988), "The Role of Potential Loss in
the Influence of Affect on Risk-taking Behavior," Organizational Behavior and
Human Decision Processes, 42 (October), 181-193.
Bagozzi, Richard P., Mahesh Gopinath, and Prashanth U. Nyer (1999), "The Role of
Emotions in Marketing," Journal of the Academy ofMarketing Science, 27 (2),
Bagozzi, Richard P. and David J. Moore (1994), "Public Service Advertisements:
Emotions and Empathy Guide Prosocial Behavior," Journal ofMarketing, 58
Barone, Michael J., Paul W. Miniard, and Jean B. Romeo (2000), "The Influence of
Positive Mood on Brand Extension Evaluations," Journal of Consumer Research,
26 (March) 386-400.
Batson, C. Daniel. (1990), "Affect and Altruism," in Affect and Social Behavior Studies
in Emotion and Social Interaction, ed. Bert S. Moore and Alice M. Isen, New
York: Cambridge University Press, 89-125.
Baumann, Donald J., Robert B. Cialdini, and Douglas T. Kenrick (1981), "Altruism as
Hedonism: Helping and Self-Gratification," Journal ofPersonality and Social
Psychology, 40 (September), 1039-1046.
Benton, David, Karen Greenfield, and Michael Morgan (1998), "The Development of the
Attitudes to Chocolate Questionnaire," Personality and Individual Differences, 24
Berkowitz, Leonard and William H. Connor (1966), "Success, Failure, and Social
Responsibility," Journal ofPersonality and Social Psychology, 4(6), 664-669.
Bower, Gordon H. (1981), "Mood and Memory," American Psychologist, 36 (February),
Brown, Judson S., and Farber, I.E. (1951) "Emotions Conceptualized as Intervening
Variables with Suggestions toward a Theory of Frustration," Psychological
Bulletin, 48, 465-495.
Brown, Steven P., Pamela M. Homer, and J. Jeffrey Inman (1998), "A Meta-Analysis of
Relationships Between Ad-Evoked Feelings and Advertising Responses," Journal
of Marketing Research, XXXV (February), 114-126.
Bruch, Hilde (1973), Eating Disorders: Obesity, Anorexia Nervosa and the Person
Within, New York: Basic Books.
Canetti, Laura, Eytan Bachar, and Eliot M. Berry (2002), "Food and Emotion,"
Behavioural Processes, 60 (November-December) 157-164.
Carlsmith, J. Merrill, and Alan E. Gross (1969), "Some Effects of Guilt on Compliance,"
Journal ofPersonality and Social Psychology, 11(3), 240-244.
Carver, Charles S., John W. Lawrence, and Michael F. Scheier (1996), "A Control-
Process Perspective on the Origins of Affect," in Striving and Feeling: Interactions
among Goals, Affect, and Self-regulation, ed. Leonard L. Martin and Abraham
Tesser, Mahwah, NJ: Erlbaum, 11-52.
Christensen, Larry (1993), "Effects of Eating Behavior on Mood: a Review of the
Literature," International Journal ofEating Disorders, 14 (September) 171-183.
Cialdini, Robert B., Betty L. Darby and Joyce E. Vincent (1973), "Transgression and
Altruism: a Case for Hedonism," Journal ofExperimental Social Psychology, 9
Cialdini, Robert B. and Douglas T. Kenrick (1976), "Altruism and Hedonism: A Social
Development perspective on the Relationship of Negative Mood State and
Helping," Journal ofPersonality and Social Psychology, 34 (May), 907-914.
Clark, Margaret S. and Alice M. Isen (1982), "Toward Understanding the Relationship
Between Feeling States and Social Behavior," in Cognitive Social Psychology, ed.
Albert H Hastorf and Alice M. Isen, New York, NY: Elsevier North-Holland, 73-
Clark, Margaret S., and Barbara A. Waddell (1983), "Effects of Moods on Thoughts
about Helping, Attraction and Information Acquisition," Social Psychology
Quarterly, 46 (March), 31-35.
Clore, Gerald, Robert S. Wyer Jr., Bruce Dienes, Karen Gasper, Carol Gohm, and Linda
Isbell (2001), "Affective Feelings as Feedback: Some Cognitive Consequences," in
Theories of Mood and Cognition, ed. Leonard L. Martin and Gerald L. Clore,
Mahwah, NJ: Erlbaum, 27-62.
Cohen, Joel B. and Eduardo B. Andrade (2004), "Affective Intuition and Task-
Contingent Affect Regulation," Journal of Consumer Research, forthcoming
Cohen, Joel B. and Charles S. Areni (1991), "Affect and Consumer Behavior," in
Handbook of Consumer Behavior, ed. Thomas S. Robertson and Harold H.
Kassarjian, Englewood Cliffs, NJ: Prentice Hall, 188-240.
Constans, Joseph I. and Andrew M. Mathews (1993), "Mood and Subjective Risk of
Future Events," Cognition and Emotion, 7 (November), 545-560.
Cools, Joseph, David E. Schotte, and Richard J. McNally (1992), "Emotional Arousal
and Overeating in Restrained Eaters," Journal ofAbnormal Psychology, 101
Cramer, Kenneth M. and Mindy Hartleib (2001), The Attitudes to Chocolate
Questionnaire: A Psychometric Evaluation," Personality andIndividual
Differences, 31 (October), 931-942.
Cunningham, Michael R., Jeff Steinberg, and Rita Grev (1980), "Wanting to and Having
to Help: Separate Motivations for Positive Mood and Guilt-induced Helping,"
Journal Personality Social Psychology, 38 (February), 181-192.
Damasio, A. (1994), Descartes' Errors: Emotion, Reason, and the Human Brain, New
York: G.P. Putnam's Sons.
Davidson, Richard J., Diego Pizzagalli, Jack B. Nitschke, and Katherine Putnam (2002),
"Depression: Perspectives from Affective Neuroscience," Annual Review of
Psychology, 53 (1), 545-574.
Derakshan, Nazanin and Michael W. Eysenck (1997), "Interpretative Biases for One's
Own Behavior and Physiology in High-trait-anxious Individuals and Repressors,"
Journal ofPersonality and Social Psychology, 73 (October), 816-825.
Eagly, Alice H., and Shelly Chaiken (1993), The Psychology ofAttitudes, Orlando, FL:
Erber, Maureen W., and Ralph Erber (2001), "The Role of Motivated Social Cognition in
the Regulation of Affective States," in Affect and Social Cognition, ed. Joseph P.
Forgas, Mahwah, NJ: Erlbaum, 275-289.
Eysenck, Michael W. (1992), Anxiety: The Cognitive Perspective, Hove, England:
Eysenck, Michael W., Collin MacLeod, and Andrew Mathews (1987), "Cognitive
Functioning and Anxiety," Psychological Research, 49 (August), 189-195.
Forgas, Joseph P. (1992), "On Mood and Peculiar People: Affect and Person Typicality
in Impression Formation," Journal ofPersonality and Social Psychology, 62
(1993), "On Making Sense of Odd Couples: Mood Effects on the Perception of
Mismatched Relationships," Personality and Social Psychology Bulletin, 19
(1995), "Mood and Judgment: The Affect Infusion Model (AIM)," Psychological
Bulletin, 117 (January), 39-66.
(2002), "Feeling and Doing: Affective Influences on Interpersonal Behavior,"
Psychological Inquiry, 13 (January), 1-28.
Frijda, Nico (1999), "Emotion and Hedonic Experience," in WellBeing: The Foundations
ofHedonic Psychology, ed. Daniel Kehneman, Ed Diener, and Norbert Schwarz,
New York: Russell Sage, 190-212.
Frost, Randy O., Gail A. Goolkasian, Robin J. Ely, and Fletcher A. Blanchard (1982),
"Depression, Restraint and Eating Behavior," Behavioral Research and Therapy,
20 (2), 113-121.
Gehring, William J., Adrian R. Willoughby (2002), "The Medial Frontal Cortex and the
Rapid Processing of Monetary Gains and Losses," Science, 295 (March 22), 2279-
Gendolla, Guido H. E. (2000), "On the Impact of Mood on Behavior: an Integrative
Theory and a Review," Review of General Psychology, 4 (December), 378-408.
Goldberg, Marvin E. and Gerald Gorn (1987), "Happy and Sad TV Programs: How They
Affect Reactions to Commercials," Journal of Consumer Research, 14 (December),
Gorn, Gerald, Michel T. Pham, and Leo Y. Sin (2001), "When Arousal influences Ad
Evaluation and Valence Does Not (and Vice Versa)," Journal of Consumer
Psychology, 11 (July), 43-55.
Greeno, Catherine G. and Rena R. Wing (1994), "Stress-induced Eating," Psychological
Bulletin, 115 (May), 444-464.
Grunberg, Neil and Richard O. Straub (1992), "The Role of Gender and Taste Class in
the Effects of Stress on Eating," Health Psychology, 11 (2), 97-100.
Herman, C. Peter and Janet Polivy (1975), "Anxiety, Restraint, and Eating Behavior,"
Journal ofAbnormal Psychology, 84 (December), 666-672.
Isen, Alice M. (1970), "Success, Failure, Attention, and Reaction to Others: the Warm
Glow of Success," Journal ofPersonality and Social Psychology, 15 (4), 294-301.
Isen, Alice M. (1984), "Toward Understanding the Role of Affect on Cognition," in
Handbook of Social Cognition, Vol. 3, ed. Robert Wyer, Thomas Srull, and Alice
Isen, Hillsdale, NJ: Erlbaum, 179-236.
Isen, Alice M. (2000), "Positive Affect and Decision Making," in Handbook of Emotions,
eds. Michael Lewis Jeannette M. Haviland-Jones, 2nd ed., New York, NY:
Isen, Alice M., Margaret Clark, and Mark F. Schwartz (1976), "Duration of the Effect of
Good Mood on Helping: Footprints on the Sands of Time," Journal Personality
and Social Psychology, 34 (September), 385-393.
Isen, Alice M., Margaret Clark, Thomas E. Shalker and Lynn Karp (1978), "Affect,
Accessibility of Material in Memory and Behavior: a Cognitive Loop?," Journal of
Personality and Social Psychology, 36 (January), 1-12.
Isen, Alice M. and Nehemia Geva (1987), "The influence of positive affect on acceptable
level of risk: the person with a large canoe has a large worry," Organizational
Behavior and Human Decision Processes, 39 (April), 145-154.
Isen, Alice M., Nancy and Horn, and David L. Rosenhan (1973), "Effects of Success and
Failure on Children's Generosity," Journal ofPersonality and Social Psychology,
27 (August), 239-247.
Isen, Alice M. and Paula R. Levin (1972), "Effect of Feeling Good on Helping: Cookies
and Kindness," Journal ofPersonality and Social Psychology, 21 (March), 384-
Isen, Alice M. and Thomas E. Shalker (1982), "The Effect of Feeling State on Evaluation
of Positive, Neutral, and Negative Stimuli: When You 'Accentuate the Positive,'
and Do You 'Eliminate the Negative'," Social Psychology Quarterly, 45 (March)
Isen, Alice M. and Stanley F. Simmonds (1978), "The effect of feeling good on a helping
task that is incompatible with good mood," Social Psychology, 41 (December),
Johnson, Eric J. and Amos Tversky (1983), "Affect, Generalization, and the Perception
of Risk," Journal ofPersonality and Social Psychology, 45 July, 20-31.
Kahn, Barbara and Alice Isen (1993), "The Influence of Positive Affect on Variety
Seeking among Safe, Enjoyable Products," Journal of Consumer Research, 20
Kaplan, Harold I. and Helen S. Kaplan (1957), "The Psychosomatic Concept of Obesity,"
Journal of Nervous and Mental Disease, 125, 181-201.
Kuhl, Julius and Juergen Beckmann, J. (1984), Action control: from Cognition to
behavior, Berlin, Germany: Springer-Verlag.
LeDoux, Joseph. (1996). The Emotional Brain: the Mysterious Underpinnings of
Emotional Life, New York: Touchstone.
Leith, Karen P. and Baumeister, Roy F. (1996), "Why Do Bad Moods Increase Self-
Defeating Behavior? Emotion, Risk-Taking, and Self-Regulation," Journal of
Personality and Social Psychology, 71 (December), 1250-1267.
Levin, Paula R. and Alice M. Isen (1975), "Further Studies on the Effect of Feeling Good
on Helping," Sociometry, 38 (March) 141-147.
Loewenstein, George F., Elke U. Weber, Christopher K. Hsee, and Ned Welch (2001),
"Risk as Feelings," Psychological Bulletin, 127 (March), 267-286.
Luce, Mary F. (1998), "Choosing to Avoid: Coping with Negatively Emotion-laden
Consumer Decisions," Journal of Consumer Research, 24 (March), 409-433.
Luomala, Harri T. and Martti Laaksonen (1997), "Mood-Regulatory Self-Gifts:
Development of a Conceptual Framework," Journal of Economic Psychology, 18,
Macdiarmid, Jennie I. and Marion M. Hetherington (1995), "Mood Modulation by Food:
An Exploration of Affect and Cravings in 'Chocolate Addicts'," British Journal of
Clinical Psychology, 34 (February), 129-138.
Macht, Michael (1999), "Characteristics of eating in anger, fear, sadness and joy,"
Appetite, 33 (August), 129-139.
Macht, Michael, S. Roth, and H. Ellgring (2002), "Chocolate Eating in Healthy Men
During Experimentally Induced Sadness and Joy," Appetite, 39 (October), 147-158.
Mano, Haim (1992), "Judgment under Distress: Assessing the Role of Unpleasantness
and Arousal in Judgment Formation," Organizational Behavior and Human
Decision Processes, 52 (July) 216-245.
Mano, Haim (1994). "Risk-taking, Framing Effects, and Affect," Organizational
Behavior and Human Decision Processes, 57 (January), 38-57.
Manucia, Gloria K., Donald J. Baumann, and Robert B. Cialdini (1984), "Mood
Influences on Helping: Direct Effects or Side Effects?," Journal of Personality and
Social Psychology, 46 (February) 357-364.
Martin, Leonard L. and Gerald L. Clore (2001), Theories of Mood and Cognition,
Mahwah, NJ: Erlbaum.
Martin, Leonard. L. and Peggy Stoner (1996), "Mood as input: what we think about how
we feel determines how we think," in Striving and Feeling: Interactions among
Goals, Affect, and Self-regulation, ed. Leonard L. Martin and Abraham Tesser,
Mahwah, NJ: Erlbaum, 279-302.
Martin, Leonard L., David W. Ward, John W. Achee, and Robert S. Wyer (1993), "Mood
as input: People have to interpret the motivational implications of their moods,"
Journal of Personality and Social Psychology, 64 (March), 317-326.
Mayer, John D., Yvonne N. Gaschke, Debra L. Braverman, and Temperance W. Evans
(1992), "Mood-Congruent Judgment is a General Effect," Journal of Personality
and Social Psychology, 63 (July), 119-132.
Mick, David G. and Michelle DeMoss (1990a), "To Me from Me: A Descriptive
Phenomenology of Self-Gifts," in Advances in Consumer Research, Vol.17, ed.
Richard Pollay et al., Provo, UT: Association for Consumer Research, 677-682.
(1990b), "Self-Gifts: Phenomenological Insights from Four Contexts," Journal of
Consumer Research, 17 (December), 322-332.
Miniard, Paul, Sunil Bhatla, and Deepak Sirdeshmukh (1992) "Mood as a determinant of
post-consumption product evaluations: Mood effects and their dependency on the
affective intensity of the consumption experience," Journal of Consumer
Psychology 1 (2), 173-195.
Mittal, Vikas and William T. Ross (1998), "The Impact of Positive and Negative Affect
and Issue Framing of Issue Interpretation and Risk-taking," Organization Behavior
and Human Decision Processes, 76 (December), 298-324.
Moore, Bill S., Bert Underwood, and David L. Rosenhan (1973), "Affect and Altruism,"
Developmental Psychology, 8 (January), 99-104.
Morris, William N. and Nora P. Reilly (1987), "Toward the Self-regulation of Mood:
Theory and Research," Motivation and Emotion, 11 (September), 215-249.
Nygren, Thomas E., Alice M. Isen, Pamela J. Taylor, and Jessica Dulin (1996), "The
Influence of Positive Affect on the Decision Rule in Risky Situations: Focus on
Outcome (and Especially Avoidance of Loss) Rather than Probability,"
Organizational Behavior and Human Decision Processes, 66 (April), 59-72.
Oliver, Georgiana and Jane Wardle (1999), "Perceived Effects of Stress on Food
Choice," Physiology & Behavior, 66 (May), 511-515.
Ortony, Andrew, Gerald L. Clore, and Allan Collins (1988), The Cognitive Structure of
Emotions, New York: Cambridge University Press.
Parrott, W. Gerrod and John Sabini (1990), "Mood and Memory Under Natural
Conditions: Evidence for Mood Incongruent Recall," Journal ofPersonality and
Social Psychology, 59 (February), 321-336.
Patel, Kushal A. and D.G. Schlundt (2001), "Impact of Moods and Social Context on
Eating Behavior," Appetite, 36 (April), 111-118.
Pham, Michel T. (1998), "Representativeness, Relevance, and the Use of Feelings in
Decision Making," Journal of Consumer Research, 25 (September), 144-159.
Pham, Michel T., Joel B. Cohen, John W. Pracejus, and G. David Hughes (2001), "Affect
Monitoring and the Primacy of Feelings in Judgment," Journal of Consumer
Research, 28 (September), 167-188.
Pietromonaco, Paula R. and Karen S. Rook (1987), "Decision Style in Depression: the
Contribution of Perceived Risks versus Benefits," Journal ofPersonality and
Social Psychology, 52 (February), 399-408.
Piliavin, John A., Jane F. Dovidio, Samuel L. Gaertner, and Russell D. Clark III (1981),
Emergency Intervention, New York, NY: Academic.
(1982), "Responsive Bystanders: the Process of Intervention," in Cooperation and
Helping Behavior, ed. Valerian J. Derlega and Janusz Grzelak, New York, NY:
Pine, Charles J. (1985), "Anxiety and Eating Behavior in Obese and Nonobese American
Indians and White Americans," Journal ofPersonality and Social Psychology, 49
Polivy, Janet and Peter C. Herman (1976), "Clinical Depression and Weight Change: A
Complex Relation," Journal ofAbnormalPsychology, 85 (June), 338-340.
Polivy, Janet, C. Peter Herman, and Shelley Warsh (1978), "Internal and External
Components of Emotionality in Restrained and Unrestrained Eaters," Journal of
Abnormal Psychology, 87 (October), 497-504.
Raghunathan, Rajagopal and Michel Tuan Pham (1999), "All Negative Moods Are Not
Equal: Motivational Influences of Anxiety and Sadness on Decision Making,"
Organizational Behavior and Human Decision Processes, 79, 1 (July), 56-77.
Regan, Dennis T., Margo Williams, and Sondra Sparling (1972), "Voluntary Expiation of
Guilt: a Field Experiment," Journal ofPersonality and Social Psychology, 24
Reznick, Harrell and Philip Balch (1977), "The Effects of Anxiety and Response Cost
Manipulations on the Eating Behavior of Obese and Normal Weight Subjects,"
Addictive Behaviors, 2(4), 219-225.
Salovey, Peter, John D. Mayer, David L. Rosenhan (1991), "Mood and Helping: Mood as
a Motivator of Helping and Helping as a Regulator of Mood," in Prosocial
Behavior, Vol. 12, ed. Margaret S. Clark, Thousand Oaks, CA: Sage, 215-237.
Schaller, Mark and Robert B. Cialdini (1990), "Happiness, Sadness, and Helping: a
Motivational Integration," in Handbook of Motivation and Cognition: Foundations
of Social Behavior, Vol. 2, ed. E. Tory Higgins and Richard M. Sorrentino, New
York: Guilford, 265-296.
Schmitz, Beth Anne (1996), The Relationship between Affect and Binge Dating,
Dissertation Abstracts International: Section B: The Sciences and Engineering, 56
(12-B, June), 7055.
Schwarz, Norbert (1990), "Feelings as Information: Informational and Motivational
Functions of Affective States," in Handbook of Motivation and Cognition.
Foundations of Social Behavior, ed. Richard M. Sorrentino and E. Tory Higgins,
New York, NY: Guilford, 527-561.
Schwarz, Norbert (2001), "Feelings as Information: Implications for Affective Influences
on Information Processing," in Theories of Mood and Cognition, ed. Leonard L.
Martin and Gerald L. Clore, Mahwah, NJ: Erlbaum, 159-176.
Schwarz, Norbert and Gerald L. Clore (1983), "Mood, Misattribution, and Judgment of
Well-Being: Informative and Directive Functions of Affective States," Journal of
Personality and Social Psychology, 45, (September), 513-523.
(1996), "Feelings and Phenomenal Experiences," in Social Psychology: Handbook
ofBasic Principles, ed. E. Tory Higgins and Arie W. Kruglanski, New York:
Siemer, Matthias and Rainer Reisenzein (1998), "Effects of Mood on Evaluative
Judgments: Influence of Reduced Processing Capacity and Mood Salience,"
Cognition and Emotion, 12 (November), 783-805.
Smith, Andrew P., Rachel Clark, and John Gallagher (1999), "Breakfast Cereal and
Caffeinated Coffee: Effects on Working Memory, Attention, Mood, and
Cardiovascular Function," Physiology & Behavior, 67 (1), 9-17.
Steptoe, Andrew, Tessa M. Pollard, and Jane Wardle, J. (1995), "Development of a
Measure of the Motives Underlying the Selection of Food: the Food Choice
Questionnaire," Appetite, 25 (December), 267-284.
Thayer, Robert E. (1989), The Biopsychology of Mood andArousal, New York: Oxford.
Thayer, Robert E. (1996), The Origins ofEveryday Moods: Managing Energy, Tension,
and Stress, New York, NY: Oxford University Press.
Tice, Dianne M., Ellen Bratslavsky, and Roy F. Baumeister (2001), "Emotional Distress
Regulation Takes Precedence Over Impulse Control: If You Feel Bad, Do It!,"
Journal of Personality and Social Psychology, 80 (January), 53-67.
Wegener, Duane T. and Richard E. Petty (1994), "Mood Management across Affective
States: The Hedonic Contingency Hypothesis," Journal ofPersonality and Social
Psychology, 66 (June), 1034-1048.
Weinstein, Suzanne, Shide, David J., and Barbara J. Rolls (1997), "Changes in Food
Intake in Response to Stress in Men and Women: Psychological Factors," Appetite,
28 (February), 7-18.
Willner, Paul, David Benton, Emma Brown, Survjit Cheeta, Gareth Davies, Janine
Morgan, and Michael Morgan (1998), 'Depression' Increases 'Craving' for Sweet
Rewards in Animal and Human Models of Depression and Craving,"
Psychopharmacology, 136 (April), 272-283.
Willner, Paul and Sarah Healy (1994), "Decreased Hedonic Responsiveness during a
Brief Depressive Mood Swing," Journal of Affective Disorders, 32 (September),
Westermann, Rainer, Kordelia Spies, Gunter Stahl, and Friedrich W. Hesse (1996),
"Relative Effectiveness and Validity of Mood Induction Procedures: A Meta-
Analysis," European Journal of Social Psychology, 26 (July-August) 557-580.
Wright, William F. and Gordon H. Bower (1992), "Mood Effects on Subjective
Probability Assessment," Organizational Behavior and Human Decision
Processes, 52 (July), 276-291.
Wyer, Robert S., Gerald L. Clore, and Linda M. Isbell (1999), "Affect and Information
Processing," In Advances in Experimental Social Psychology, Vol. 31, ed. Mark P.
Zanna, San Diego, CA: Academic Press, 1-77.
Zillmann, Dolf (1988), "Mood Management: Using Entertainment to Full Advantages,"
in Communication, Social Cognition, andAffect, ed. Lewis Donohew and Howard
E. Sypher, Hillsdale, NJ: Erlbaum, 147-171.
I graduated in Business Administration at the Universidade Federal de Santa
Catarina in Florianopolis (SC) Brazil, in December 1993. A few years later, I received
my master's degree in business (marketing option) from the Universite de Montreal -
HEC School. My interests in marketing and, particularly, in consumer behavior led me to
move to the University of Florida to pursue a Ph.D. in the marketing department. I will be
joining the Haas School of Business at the University of California-Berkeley as an
Assistant Professor of Marketing in July 2004.