Special Issue

R875

Empathy and Empathy

compassion
Tania Singer1,*
and Olga M. Klimecki2,3,4 Compassion Empathic distress

As humans we are a highly social

species: in order to coordinate our Other-related emotion Self-related emotion
joint actions and assure successful Positive feelings: e.g., love Negative feelings: e.g., stress
communication, we use language
skills to explicitly convey information Good health Poor health, burnout
to each other, and social abilities
such as empathy or perspective Approach & prosocial motivation Withdrawal & non-social behavior
taking to infer another person’s
Current Biology
emotions and mental state. The
human cognitive capacity to draw Figure 1. Compassion and empathic distress.
Schematic model that differentiates between two empathic reactions to the suffering of others.
inferences about other peoples’
beliefs, intentions and thoughts has
been termed mentalizing, theory of the potentially very different outcomes distress, which is also referred to as
mind or cognitive perspective taking. that empathic or compassionate personal distress; and compassion,
This capacity makes it possible, for responses to others’ distress may which is also referred to as empathic
instance, to understand that people have, it is of great importance to concern or sympathy (Figure 1). For
may have views that differ from our understand which factors determine simplicity, we will refer to empathic
own. Conversely, the capacity to the emergence of these different distress and compassion when
share the feelings of others is called social emotions and to know more speaking about these two different
empathy. Empathy makes it possible about whether and how such families of emotions. While empathy
to resonate with others’ positive and emotional responses can be trained refers to our general capacity to
negative feelings alike — we can thus and changed. resonate with others’ emotional
feel happy when we vicariously share states irrespective of their valence —
the joy of others and we can share Psychological perspective positive or negative — empathic
the experience of suffering when we Although the concepts of empathy distress refers to a strong aversive
empathize with someone in pain. and compassion have existed for and self-oriented response to the
Importantly, in empathy one feels with many centuries, their scientific suffering of others, accompanied
someone, but one does not confuse study is relatively young. The term by the desire to withdraw from a
oneself with the other; that is, one still empathy has its origins in the Greek situation in order to protect oneself
knows that the emotion one resonates word ‘empatheia’ (passion), which is from excessive negative feelings.
with is the emotion of another. If this composed of ‘en’ (in) and ‘pathos’ Compassion, on the other hand, is
self–other distinction is not present, (feeling). The term empathy was conceived as a feeling of concern
we speak of emotion contagion, a introduced into the English language for another person’s suffering
precursor of empathy that is already following the German notion of which is accompanied by the
present in babies. ‘Einfühlung’ (feeling into), which motivation to help. By consequence,
While shared happiness certainly originally described resonance with it is associated with approach and
is a very pleasant state, the sharing works of art and only later was used prosocial motivation.
of suffering can at times be difficult, to describe the resonance between Research by Daniel Batson and
especially when the self–other human beings. The term compassion Nancy Eisenberg in the fields of
distinction becomes blurred. Such is derived from the Latin origins social and developmental psychology
a form of shared distress can be ‘com’ (with/together) and ‘pati’ (to confirmed that people who feel
especially challenging for persons suffer); it was introduced into the compassion in a given situation help
working in helping professions, such English language through the French more often than people who suffer
as doctors, therapists, and nurses. word compassion. In spite of the from empathic distress. Furthermore,
In order to prevent an excessive philosophical interest for empathy and Daniel Batsons’ work showed that
sharing of suffering that may turn the fundamental role that compassion the extent to which people feel
into distress, one may respond to the plays in most religions and secular compassion can, for instance, be
suffering of others with compassion. ethics, it was not until the late 20th increased by explicitly instructing
In contrast to empathy, compassion century that researchers from social participants to feel with the target
does not mean sharing the suffering and developmental psychology person. Interestingly, the capacity to
of the other: rather, it is characterized started to study these phenomena feel for another person is not only a
by feelings of warmth, concern and scientifically. property of a person or a situation,
care for the other, as well as a strong According to this line of but can also be influenced by
motivation to improve the other’s psychological research, an empathic training.
wellbeing. Compassion is feeling for response to suffering can result in In order to train social emotions like
and not feeling with the other. Given two kinds of reactions: empathic compassion, recent psychological
Current Biology Vol 24 No 18
R876

repeatedly found evidence for the

existence of such shared neuronal
networks (Figure 2). For example,
meta-analyses on empathy for pain
studies have revealed that a portion
of the anterior insula and a specific
part of the anterior cingulate cortex
were consistently activated, both
during the experience of pain as well
as when vicariously feeling with the
suffering of others.
Importantly, the magnitude of these
empathy-related activations was
modulated by individual differences
in the degree to which participants
reported having experienced negative
feelings while empathizing with the
Figure 2. Neural network underlying empathy for pain. other. Although empathy has been
Depicted functional neural activations on the right are the result of a meta-analysis based on studied most extensively in the
nine fMRI studies investigating empathy for pain. AI, anterior insula; aMCC, anterior middle domain of pain, similar paradigms
cingulate cortex; IFG, inferior frontal cortex. Right side of figure reproduced with permission have also been used for the study
from Lamm et al. (2011).
of touch, disgust, taste or social
rewards. Depending on the emotion in
research has increasingly made use pure altruistic helping as opposed question, such shared networks were
of meditation-related techniques to reciprocity-based helping. This observed in somatosensory cortex
that foster feelings of benevolence indicates that compassion training for vicarious neutral touch, medial
and kindness. The most widely used especially increases prosocial orbitofrontal cortex for vicarious
technique is called ‘loving kindness motivation rather than just norm- pleasant touch, ventral striatum for
training’. This form of mental practice adherence. shared social rewards and parts of the
is carried out in silence and relies on anterior insula when empathizing with
the cultivation of friendliness towards A neuroscientific perspective on taste and disgust.
a series of imagined persons. One empathy After having established this basic
would usually start the practice by This purely behavioral psychological neural mechanism underlying our
visualizing a person one feels very research is more and more supported ability to share feelings with others,
close to and then gradually extend and extended by recent findings from a second generation of empathy
the feeling of kindness towards social neuroscience. Some years studies — again mostly focusing on
others, including strangers and, at ago, this relatively new discipline vicarious pain — has investigated the
a later stage, also people one has embarked on the investigation of modulation of such empathic brain
difficulties with. Ultimately, this social emotions such as empathy responses by various factors. Indeed,
practice aims at cultivating feelings and compassion and their plasticity. the results reveal that empathic brain
of benevolence towards all human A multitude of neuroimaging responses are modulated by factors
beings. studies using functional magnetic that range from person-specific
Using this kind of training, resonance imaging (fMRI) has, for characteristics, such as gender,
researchers around Barbara example, shown that empathizing to context-specific factors. For
Fredrickson have shown that several with another person’s feelings relies example, in several fMRI studies in
weeks of regular compassion training on the activation of neural networks our lab we could show that perceived
can have a beneficial impact on self- that also support the first-person group membership or fairness of
reported feelings of positive affect, experience of these feelings. another person matters for how much
personal resources, and well-being A very prominent way to study empathy one will actually experience
during everyday life. Interestingly, such ‘shared neuronal networks’ for the other. Thus, witnessing the
the beneficial effects of compassion underlying empathic experiences suffering of a perceived in-group
training are not limited to the person is the domain of pain. In such member (same football team) or
who is training, but can also benefit ‘empathy for pain paradigms’, of someone who played fairly in
others. More recent research in our scanned participants typically either economic games beforehand evoked
lab has shown that participants receive painful stimulation to body more pronounced empathy-related
who undergo loving kindness and parts themselves or are presented anterior insula activations than when
compassion training increased their with pictures or cues that indicate witnessing an unfair person or an
helping rates towards strangers in that another person is currently out-group member (rival football
a computer game when compared experiencing pain. By then comparing team) suffering pain. Importantly, the
to an active memory control group. the brain activations that are elicited magnitude of the empathy-related
Interestingly, the amount of time by the first-hand experience of pain signal in the anterior insula predicted
participants practiced compassion with those purely elicited by the the extent to which participants
predicted how much a certain type of vicarious observation of another later engaged in altruistic helping
helping behavior increased, namely person in pain, researchers have behavior.
Special Issue
R877

Plasticity of the socio-emotional A

brain
Despite existing psychological
findings suggesting the possibility of aMCC

transforming social emotions through

GP
training, it was only very recently that Put
AI VS
neuroscience began to investigate Compassion network NAcc
sgACC mOFC
the neural plasticity underlying our VTA
Empathy for pain network SN
capacity for empathy and compassion
(Figure 3A). As usual in plasticity
research, one begins with cross- B VTA, SN mOFC GP, Put
sectional studies which compare
Compassion
experts in a given field to novices. In group Compassion
the case of studying the malleability
of the compassionate brain, the Memory
group Memory
experts were long-term meditators
that had trained compassion over Time x = 10 x=8 x=2 x=2
many years. The results of a study Pre Training 1 Post 1
conducted by Antoine Lutz and Measurement Measurement
Richard Davidson revealed that
when exposed to distressing sounds, C AI aMCC sgACC mOFC AI VS, NAcc
expert meditators reveal increased
activations in middle insula as
compared to novice meditators.
These studies were then followed
by longitudinal designs in which
meditation-naïve subjects underwent -43 -32 -10 8 12 32 44 52 18
short-term training of affective Empathy Compassion Overlaps with empathy meta-analysis
capacities.
In a series of studies performed
Affect
in our lab, for example, the brains group Empathy Compassion
of meditation-naïve participants
were scanned before and after Memory
Memory Memory
they underwent either empathy or group
compassion training. During the Time
scanning, participants were watching Pre Post 1 Post 2
Training 1 Training 2
short film excerpts depicting others’ Measurement Measurement Measurement
suffering. Throughout the experiment, Current Biology
participants provided self-reports
on their feelings in response to each Figure 3. Differential neural networks for empathy and compassion.
of these film clips. These studies (A) Training compassion or empathy leads to differential plasticity in neural networks. (B) Com-
revealed that, in comparison to a passion training compared to memory training augments activations in ventral tegmental area/
memory control group, short-term substantia nigra (VTA/SN), medial orbitofrontal cortex (mOFC), and striatum, the latter span-
ning globus pallidus (GP) and putamen (Put). (C) Empathy training (in blue) leads to increased
compassion training of several days activations in anterior insula (AI) and anterior middle cingulate cortex (aMCC), while subse-
was able to increase positive affect quent compassion training (in red) augments activations in medial orbitofrontal cortex (mOFC),
and activations in a neural network subgenual anterior cingulate cortex (sgACC) and the ventral striatum/nucleus accumbens (VS,
usually related to positive emotions NAcc). Original brain data in (B) and (C) adapted with permission from Klimecki et al. (2013).
(spanning medial orbitofrontal cortex
and striatum; Figure 3B). This finding differs from plasticity involved in affect. In line with previous results,
underlines the malleability of social compassion training, we conducted compassion training again led to
emotions as it shows that a short- another longitudinal study in which an increase in a non-overlapping
term compassion training of several participants first engaged in empathy brain network, including medial
days can foster positive feelings and training before receiving compassion orbitofrontal cortex and ventral
related brain activations, even when training in a second step (Figure 3C). striatum (Figure 3C). The comparison
persons are exposed to the distress This study revealed that several of the effects of both training regimes
of others. days of empathy training led to an on observed functional brain plasticity
Interestingly, this compassion- activation increase in insula and thus indicates that empathy and
related brain network differed from anterior middle cingulate cortex, compassion training indeed elicited
the above-mentioned networks as well as to an increase in self- changes in differential brain networks
implicated in empathy for pain reported negative affect. In contrast, associated with opposed patterns in
(encompassing anterior insula and subsequent compassion training in experienced affect.
anterior middle cingulate cortex). In the same participants could reverse Taken together, these results
order to formally compare whether this effect by decreasing negative underline the important distinction
plasticity involved in empathy training affect and increasing positive between empathy and compassion,
Current Biology Vol 24 No 18
R878

both on a psychological and

neurological level. Accordingly,
neuroscience research may be able to
determine critical periods throughout
Cochlear implants
exposure to the distress and suffering ontogeny which indicate when it is
of others can lead to two different best to teach these socially relevant Olivier Macherey1,*
emotional reactions. Empathic skills during development. Such and Robert P. Carlyon2
distress, on the one hand, results in knowledge could help to assure
negative feelings and is associated an effective education fostering Cochlear implants are the first
with withdrawal. When experienced subjective wellbeing, adaptive example of a neural prosthesis that
chronically, empathic distress most emotion-regulation, meaningful can substitute a sensory organ: they
likely gives rise to negative health relationships and human prosociality. bypass the malfunctioning auditory
outcomes. On the other hand, periphery of profoundly-deaf people
compassionate responses are based Further reading to electrically stimulate their auditory
on positive, other-oriented feelings Batson, C.D. (2009). These things called empathy: nerve. The history of cochlear
eight related but distinct phenomena. In The
and the activation of prosocial Social Neuroscience of Empathy, J. Decety
implants dates back to 1957, when
motivation and behavior. Given the and W. Ickes, eds. (Cambridge: MIT Press), Djourno and Eyriès managed, for the
potentially detrimental effects of pp. 3–15. first time, to elicit sound sensations
de Vignemont, F., and Singer, T. (2006). The
empathic distress, the finding of empathic brain: how, when and why? Trends in a deaf listener using an electrode
existing plasticity of adaptive social Cogn. Sci. 10, 435–441. implanted in his inner ear. Since
Eisenberg, N. (2000). Emotion, regulation, and
emotions is encouraging, especially moral development. Annu. Rev. Psychol. 51,
then, considerable technological
as compassion training not only 665–697. and scientific advances have been
promotes prosocial behavior, but Fredrickson, B.L., Cohn, M.A., Coffey, K.A., Pek, J., made. Worldwide, more than 300,000
and Finkel, S.M. (2008). Open hearts build lives:
also augments positive affect and positive emotions, induced through loving- deaf people have been fitted with a
resilience, which in turn fosters better kindness meditation, build consequential cochlear implant; it has become a
personal resources. J. Pers. Soc. Psychol. 95,
coping with stressful situations. 1045–1062.
standard clinical procedure for born-
This opens up many opportunities Frith, C.D., and Frith, U. (2006). The neural basis of deaf children and its success has
for the targeted development of mentalizing. Neuron 50, 531–534. led over the years to relaxed patient
Hein, G., Silani, G., Preuschoff, K., Batson, C.D.,
adaptive social emotions and and Singer, T. (2010). Neural responses to selection criteria; for example, it is
motivation, which can be particularly ingroup and outgroup members’ suffering now not uncommon to see people
predict individual differences in costly helping.
beneficial for persons working in Neuron 68, 149–160.
with significant residual hearing
helping professions or in stressful Klimecki, O.M., Leiberg, S., Lamm, C., and undergoing implantation. Although
environments in general. Singer, T. (2013). Functional neural plasticity the ability to make sense of sounds
and associated changes in positive affect
after compassion training. Cereb. Cortex 23, varies widely among the implanted
Future outlook 1552–1561. population, many cochlear implant
Klimecki, O.M., Leiberg, S., Ricard, M., and
Despite these recent advances in Singer, T. (2014). Differential pattern of
listeners can use the telephone and
the neuroscientific study of social functional brain plasticity after compassion and follow auditory-only conversations in
phenomena such as empathy and empathy training. Soc. Cogn. Affect. Neurosci. quiet environments.
9, 873–879.
compassion and their plasticity, Lamm, C., Decety, J., and Singer, T. (2011). Meta- The core functions of a cochlear
many questions remain to be analytic evidence for common and distinct implant are to convert the input
neural networks associated with directly
answered. Currently, researchers are experienced pain and empathy for pain.
sounds into meaningful electrical
investigating the longer-term effects Neuroimage 54, 2492–2502. stimulation patterns, and then to
of different types of such socio- Leiberg, S., Klimecki, O., and Singer, T. (2011). deliver these patterns to the auditory
Short-term compassion training increases
affective training techniques, focusing prosocial behavior in a newly developed nerve fibers. In this primer, we shall
not only on their effect on functional prosocial game. PLoS One 6, e17798. describe how these two steps are
Lutz, A., Brefczynski-Lewis, J., Johnstone, T., and
brain plasticity but also on changes Davidson, R.J. (2008). Regulation of the neural
performed, show how the original
in brain structure, health-related circuitry of emotion by compassion meditation: information present in the sounds is
variables (stress hormones, immune Effects of meditative expertise. PLoS One 3, degraded as a result of both device
e1897.
parameters, neurogenetic markers) Singer, T. (2012). The past, present and future of and sensory limitations, and discuss
as well as ecologically valid everyday social neuroscience: a European perspective. current research trends aiming
Neuroimage 61, 437–449.
behavior and cognition (thoughts, Singer, T., Seymour, B., O’Doherty, J., Kaube, H.,
to improve speech perception,
prosocial actions, relationships to Dolan, R.J., and Frith, C.D. (2004). Empathy particularly in challenging listening
others). for pain involves the affective but not sensory conditions.
components of pain. Science 303, 1157–1162.
Longitudinal follow-up studies will
also have to determine how long Normal and impaired hearing
such beneficial changes can be ¹Max Planck Institute for Human Cognitive In normal hearing, sound pressure
and Brain Sciences, Department of Social
maintained and how these changes Neuroscience, Leipzig, Germany. ²Swiss
waves travel down the ear canal and
can be sustained. In addition, future Center for Affective Sciences, University of cause the eardrum to vibrate. These
research is needed to delineate Geneva, Switzerland. ³Laboratory for the vibrations are directly transmitted
in more detail the neurobiological Study of Emotion Elicitation and Expression, to the entrance of the cochlea by
mechanisms underlying the Department of Psychology, University the small bones of the middle ear
differential changes observed after of Geneva, Switzerland. 4Laboratory for (Figure 1). The cochlea is responsible
Behavioral Neurology and Imaging of
empathy and compassion training. Cognition, Department of Neuroscience,
for transducing these mechanical
One such question relates to the Medical School, University of Geneva, vibrations into action potentials that
neurotransmitters that are involved. Switzerland. will further propagate towards the
And finally, future developmental *E-mail: singer@[Link] brain and eventually elicit a sound
 RESEARCH ARTICLE

The Impact of Emotions and Empathy-

Related Traits on Punishment Behavior:
Introduction and Validation of the Inequality
Game
Olga M. Klimecki1,2*, Patrik Vuilleumier1,3, David Sander1,2
1 Swiss Center for Affective Sciences, University of Geneva, Geneva, Switzerland, 2 Laboratory for the
Study of Emotion Elicitation and Expression, Department of Psychology, University of Geneva, Geneva,
Switzerland, 3 Laboratory for Behavioral Neurology and Imaging of Cognition, Department of Neuroscience,
Medical School, University of Geneva, Geneva, Switzerland

* [Link]@[Link]

OPEN ACCESS
Abstract
Citation: Klimecki OM, Vuilleumier P, Sander D In the prevention and resolution of conflicts in social contexts, an important step is to under-
(2016) The Impact of Emotions and Empathy-Related stand how different emotions and empathic traits are linked to punishment behaviors. Unfor-
Traits on Punishment Behavior: Introduction and
tunately, few paradigms exist to study these phenomena. Here, we developed the
Validation of the Inequality Game. PLoS ONE 11(3):
e0151028. doi:10.1371/[Link].0151028 Inequality Game (IG) as an economic and verbal interaction paradigm in which participants
are faced with an “unfair other” as opposed to a “fair other” and subsequently have the
Editor: Elisabetta Palagi, University of Pisa, ITALY
opportunity to engage in a range of social behaviors. These social behaviors include coop-
Received: July 21, 2015
erative or competitive economic choices and nice or derogatory verbal behavior toward the
Accepted: February 23, 2016 unfair and fair other. Participants could thus engage in punishment or forgiveness behavior
Published: March 15, 2016 toward the unfair other as well as in cooperative or aggressive behavior toward the fair
Copyright: © 2016 Klimecki et al. This is an open other. We validated the IG through multimodal measures comprising the assessment of per-
access article distributed under the terms of the sonality traits, emotions (by means of facial expressions and self-reports), arousal (by
Creative Commons Attribution License, which permits means of skin conductance responses), physical effort (force exertion), and behavioral
unrestricted use, distribution, and reproduction in any
reactions. Second, we examined the influence of emotions and empathy-related traits on
medium, provided the original author and source are
credited. punishment behavior. With regard to emotions, we observed a positive relation between
malicious joy and punishment behavior. This result highlights the role of reward-related
Data Availability Statement: As data contain
confidential information of study participants, data are mechanisms in favoring punishment behavior. In addition, different empathic traits had
available from the Swiss Center of Affective Sciences opposing effects on antisocial behavior. Whereas personal distress predicted aggressive
at the University of Geneva, Switzerland. Please verbal behavior, perspective taking and empathic concern predicted a reduction in punish-
contact Dr. Olga Klimecki at [Link]@[Link]
for further information.
ment behavior. Empathic traits also modulated emotional experience and person evalua-
tions, such that perspective taking was related to more positive affect (less frowning and
Funding: This work was supported by a Marie Curie
fellowship to Olga M. Klimecki and it received funding
more smiling) and a more favorable evaluation of the unfair other. The current data validate
from the European Union Seventh Framework the IG, reveal that malicious joy is positively related to punishment behavior, and show that
Programme (FP7/2007-2013) under grant agreement different types of empathic traits can have opposing effects on antisocial behavior as well
n° 267171. The research was also supported by the
as on related emotions and person evaluations.
National Center of Competence in Research (NCCR)
for the Affective Sciences, financed by a grant from
the Swiss National Science Foundation (51NF40-

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 1 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

104897), hosted by the University of Geneva. The Introduction

funders had no role in study design, data collection
and analysis, decision to publish, or preparation of Although the role of emotions and empathic processes in social interactions is increasingly
the manuscript. being studied (e.g., [1–4]), many questions remain open regarding their relation to antisocial
Competing Interests: The authors have declared
behavior. This might partly be because extant paradigms for the study of emotional and behav-
that no competing interests exist. ioral responses to aggression (e.g., [5, 6]) do not allow for a clear delineation of reactive as
opposed to instrumental (or proactive) aggression [7]. To address these issues, we developed
and validated a novel task, the Inequality Game (IG). This task was used to investigate emo-
tional and behavioral reactions to anger-provoking situations and to examine how emotional
experiences and empathic personality traits influence these behaviors.
A few experimental paradigms allow for the assessment of behavioral reactions to anger elic-
itation [5, 6, 8, 9], but they have a number of shortcomings. Anger and aggression can be elic-
ited in the competitive reaction time task [5] and the point subtraction task [6], both of which
rely on a temporal structure in which a provocation and the possibility to retaliate are inter-
spersed. In such paradigms, punishment behavior might be driven by reactive aggression,
which denotes an affect-driven type of aggression [7], or by instrumental aggression [7], which
is less emotional but more oriented towards a goal, such as to change the other’s behavior.
Other paradigms relying on humiliation, unjustified reprimands, and insults (e.g., [9]) consti-
tute singular events that are not suitable for concurrently assessing physiological reactions,
since the latter necessitates recording data over several trials (such as skin conductance
responses [SCRs] or functional magnetic resonance imaging). Still other paradigms from
behavioral economics, such as mixed-motive games [8], rely on interdependent payoff matrices
in which the choice of one player affects his own and another person's outcome. This structure,
however, confounds other-regarding behavior with self-interests. There is thus a shortage of
paradigms that at the same time allow for i) the separate assessment of emotional and behav-
ioral reactions to anger elicitation, ii) the concurrent recording of physiological data across
repeated trials, and iii) a distinction between self- and other-regarding preferences.
Emotions play a crucial role in guiding judgments and behavior (e.g., [10, 11]). Positive and
negative emotions may influence interpersonal exchanges in opposing ways (for review, see
[12, 13]). Although different emotions of the same valence can also have opposing effects on
decision making (e.g., [14]), positive emotions in general tend to be associated with more coop-
eration [1], while negative emotions in general and anger in particular have been implicated in
intensifying conflict and aggression (e.g., [2, 15]). Although some research has shown that
induced positive emotions can promote cooperative behavior (e.g., [1]), there is a shortage of
studies investigating the link between emotional experiences, such as anger, with social behav-
ior, such as aggression. Anger and aggression are conceptually linked, although anger does not
always lead to aggression [15]. Another emotion that might be of interest here is malicious joy,
as it has been linked to punishment [16,17]. Observing an unfair person being punished was
thus associated with malicious joy [16] and punishing an unfair person was accompanied by
neural activations in reward-related areas [17]. Despite these findings, the relation between
emotional experiences, such as malicious joy, and active punishment behavior remains largely
unresolved.
Smooth social interactions are greatly facilitated by perspective taking and empathy. Per-
spective taking is defined as the ability to infer the thoughts and intentions of others through
cognitive processes that allow one to simulate the perceptions and beliefs of others [18]. Empa-
thy is commonly defined as the affective sharing and understanding of both pleasant and
unpleasant emotions [19]. When it comes to sharing of unpleasant emotions, two components
of empathy can be differentiated–empathic distress (also called personal distress, e.g., in [20])
and empathic concern (also called compassion, e.g., in [21] or sympathy, e.g., in [22]). Whereas

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 2 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

empathic distress denotes the sharing of others’ suffering to a strong degree, leading to a conta-
gion with negative affect, empathic concern occurs when the suffering of others is met with a
more positive feeling of caring and the motivation to help (for details, see [23]). A meta-analy-
sis revealed that in situations that involve no conflict, the capacity for perspective taking was
related to prosocial behavior [24]. However, it is unknown whether a similar relation also
holds for conditions eliciting anger and bearing the possibility for punishment and aggression.
Another meta-analysis revealed that empathy in general was associated with reduced aggres-
sion [25]. In situations involving needy others, however, empathic skills may have opposing
effects on helping behavior [3]. More specifically, other-oriented empathic feelings, called
empathic concern or compassion, are associated with prosocial behavior and positive affect [3,
26], whereas feelings of empathic distress lead to reduced helping behavior and increased nega-
tive feelings [3]. In light of these opposing effects of different empathy-related components, an
open question is whether empathic distress and empathic concern may also have an opposing
impact on emotional and behavioral reactions to anger provocation.
What are potential factors reducing punishment behavior and aggressive response following
anger elicitation? Previous research has shown that the experience of positive affect is usually
associated with more cooperation [1, 12, 13]. However, it remains an open question as to
whether strengthening other-related positive affect can also reduce punishment behavior.
Other results suggest that the opportunity to “settle the score,” i.e., to engage in revenge behav-
ior, can reduce subsequent aggression [27]. This contrasts with the finding that engaging in
aggressive actions can prompt more aggression (e.g., [28]). In light of these opposing findings,
more research on the effects of "settling the score" is needed.
The present study addressed these issues by designing and validating the Inequality Game
(IG) as a novel paradigm with several complementary measures. Our first aim was to develop a
task that can dissociate emotional and behavioral responses to anger elicitation, and allows for
the concurrent recording of physiological data. To validate this task, we expected to find evi-
dence for a reliable engagement of participants during the game, as reflected not only by self-
reports and behavioral ratings (prediction 1a), but also by physiological measures related to
arousal (quantified by SCRs [29, 30]) and effort (quantified by handgrip force [31, 32]) (predic-
tion 1b). We also expected that the IG would induce anger (correlating with individual trait
anger), punishment behavior in economic exchanges, and unfavorable person evaluation for
the unfair other (prediction 1c). Finally, we expected that the IG would allow for the classifica-
tion of participants in different types (prosocial, sanctioning, or competitive) based on their
behavior (prediction 1d).
A second aim of our study was to more specifically examine the relationship between emo-
tions and punishment behavior. We tested whether emotional expressions (e.g., facial expres-
sions or self-reports) could predict subsequent behavioral reactions towards fair and unfair
others (prediction 2a). Spontaneous facial expressions were recorded to probe for displays of
happiness (as indexed by smiles) and negative affect (as indexed by frowns). Previous work
with the Facial Action Coding System [33] established that the activity of the zygomaticus mus-
cle (smile) is indicative of happiness, while the activity of the corrugator muscle (frown) is asso-
ciated with unpleasant emotions such as anger, sadness, or fear [34]. In addition, we explored
how more specific emotional experiences, such as anger, malicious joy, and regret, were linked
to punishment behavior. These indices allowed us to test whether aggression is primarily
related to negative emotions, such as anger [2], and to explore whether reward-related emo-
tions, such as malicious joy, may also play a role in driving punishment behavior (question 2b).
The third aim of our study was to examine whether–in analogy to findings from the domain
of helping behavior [3, 24]–different empathy-related skills can impact punishment behavior
in opposing ways. More specifically, we predicted that empathic concern and perspective

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 3 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

taking would be related to a reduction in punishment behavior, while the tendency to experi-
ence personal distress would be related to more frequent punishment behavior (prediction 3a).
In addition, we explored how emotions, empathy-related traits, person evaluation, and social
behavior are linked to each other in the context of anger provocation (question 3b).
A final aim was to investigate factors that may promote a reduction in punishment behav-
ior. To this end, we examined two possibilities. First, we hypothesized that strengthening posi-
tive other-related emotions prior to playing the IG would reduce subsequent punishment
behavior (prediction 4a). Second, we wanted to test two competing hypotheses: would "settling
the score" in one situation (the IG) indeed reduce subsequent punishment behavior in another
task, or would there still be significant carry-over effects to a different situation (question 4b)?
To this end, we included a seemingly unrelated task subsequent to the IG in which participants
could assign desirable and undesirable food items to their former interaction partners.
In summary, the current work allowed us to establish the validity of the IG and its usefulness
for obtaining a range of different measures (including self-reports, physiological data, and
behavioral data) in relation to anger and social conflict. In addition, our results shed new light
on the relation between emotions, empathy-related traits, and punishment behavior.

Materials and Methods

Participants
The final sample consisted of 40 men who were mainly students (age: M = 24.78 years;
SD = 5.34; range: 18–37). Six of the initial 46 participants were excluded, as they saw through
the nature of the manipulation. We restricted our sample to male participants because of previ-
ously reported sex differences in aggression [35, 36] and emotional responses to unfairness
[16], with the aim to ensure that we observed a range of aggressive responses in this initial
validation study. Prior to the experiment, volunteers completed online questionnaires concern-
ing demographic information, alexithymia (Toronto Alexithymia Scale [TAS-20] [37], French
version [38]), and depression (Beck’s Depression Inventory [BDI-II] [39]). Volunteers were
included if they met the following criteria: male, aged 18–40, TAS score  60, and BDI
score  28. Participants provided informed written consent, received 25 CHF for their partici-
pation, and were fully debriefed after the experiment. The study was approved by the Research
Ethics Committee of the Faculty of Psychology and Educational Sciences at the University of
Geneva.

Measures
Inequality Game. The IG was developed to separately assess emotional and behavioral
reactions to anger provocation in a socio-economic interaction. It was programmed by using
Cogent 2000 (developed by the Cogent 2000 team at the FIL and the ICN and Cogent Graphics
developed by John Romaya at the LON at the Wellcome Department of Imaging Neuroscience,
London, UK). Participants were told that they played the game with two other participants.
Unbeknownst to them, the behavior of the two other players was preprogrammed to be fair or
unfair. To increase credibility, participants could freely interact with these two ostensible other
participants prior to the testing. At this point, the two confederates did not know whether they
would be assigned the fair or unfair role during the IG.
The paradigm had a 2 x 2 x 2 within-subject design with the following three factors: the
power of the participant (low power in the first phase of the game, followed by high power in
the second phase of the game), the behavior of the other (fair or unfair), and the event type dur-
ing the interaction (economic interaction or verbal feedback).

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 4 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

The first phase of the game served to assess the participant’s emotional reactions to anger
elicitation. Therefore, the participant was in low power and the two others (the fair and unfair
other) were in high power. In each trial of the game (Fig 1), the participant was paired with
either the fair or with the unfair other. During economic events, the participant interacted with
one of the two others to make a common decision on a 2 x 2 payoff matrix that determined the
payoff of both players (the participant and the other). The player in high power controlled the
lines and chose first. The player in the low power condition controlled the columns and chose
second. The player in high power affected up to 90% of the other’s gain in a given trial. He
could, for instance, choose between a cooperative distribution (e.g., both interaction partners
receive 9 or 10 CHF) and a competitive distribution (e.g., 9 or 10 CHF for oneself and 1 or 0
CHF for the other). Conversely, the player in the low power condition could affect only up to

Fig 1. Trial timeline of the Inequality Game, depicted for the high power phase. After player selection, the photographs of both paired players were
presented, along with their assignment to control the lines (high power) or the columns (low power). The payoff structure of a new trial was shown for 4 s. The
participant (here Nicolas, in green) then chose one of the two lines. After displaying the participant’s choice, the ostensible other (here Jean, in blue) made a
decision. The common choice (Final Result) determined the payoff (here: 8 CHF for Nicolas and 1 CHF for Jean). Half of the economic interactions were
followed by feedback trials. Here, the player with high power (green) had the opportunity to select between one of four feedback choices for the other, which
were subsequently displayed. In the low power phase, the participant was in control of the columns (blue).
doi:10.1371/[Link].0151028.g001

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 5 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

10% of the gain difference in a given trial (corresponding to differences of 1 CHF). The game
was designed in such a way that the fair other selected cooperative outcomes, thereby maximiz-
ing both his own and the participant’s benefits, whereas the unfair other chose competitive out-
comes that maximized his own benefits while minimizing the participant’s benefits.
We implemented another more social interaction by ensuring that half of the economic
choices were followed by a trial in which the player in high power could select one of four feed-
back messages for the other. In the low power phase, the fair other always sent nice feedback
messages to the participant (“You are very nice” three times and “You are nice” three times).
Conversely, the unfair other always sent derogatory feedback messages to the participant
(“You are not nice” three times and “You are annoying” three times).
In the subsequent phase of the IG, the participant was in turn given high power, allowing us
to probe for the participants’ behavioral reactions to anger provocation. In this phase, partici-
pants controlled the economic distributions for the fair and unfair other and selected one of
the four feedback messages for the fair and unfair other. Participants could thus choose
between cooperative or competitive economic distributions for the others and between nice or
derogatory feedback sentences for the others.
In total, each phase (low and high power) consisted of 24 economic choices and 12 feedback
choices, which were equally split into interactions with the fair and unfair other. Twelve of the
24 payoff matrices reflected situations in which the high power other chose between a coopera-
tive and a competitive allocation of money for the other (e.g., 8/9, 9/8 or 8/1, 9/0 CHF). In
order to control for the monetary reward, we ensured that the remaining 12 trials had essen-
tially predetermined economic outcomes (win or no win). These control conditions consisted
of six win trials (e.g., 10/9, 10/10, 9/10, 9/10 CHF) and six no win trials (e.g., 0/1, 1/1, 0/0, 1/1
CHF). When the fair and unfair other were in low power or during the control conditions
(win, no win for both players), their choices always served to maximize their own gain.
In contrast to mixed-motive games [8], the high power player’s payoff in the IG was largely
independent of the low power player’s payoff and thus reflects other-regarding choices. Follow-
ing the classification of persons as prosocial when they maximize joint outcomes and competi-
tive when they maximize their relative advantage over other’s outcomes [40, 41], the IG allows
for the classification of participants as prosocial, sanctioning, and competitive. Participants
were identified as prosocial if they chose more cooperative than competitive economic out-
comes for both the fair and unfair other. They were classified as sanctioning if they recipro-
cated by choosing more cooperative than competitive economic outcomes for the fair other
and more competitive than cooperative outcomes for the unfair other. Finally, they were classi-
fied as competitive if they chose competitive outcomes more often than cooperative outcomes
for both the fair and unfair other, thereby maximizing the relative difference between their
own and the other’s economic gain.
Participants’ economic and feedback behavior in the high power phase of the IG was opera-
tionalized as follows. Economic behavior was quantified by subtracting the number of competi-
tive economic choices from the number of cooperative economic choices, yielding an indicator
of cooperation (positive values) as opposed to competition (negative values). In order to test
the effect of participants’ choices on economic outcomes, the sum of hypothetical gains from
the high power phase was computed for all three players. Feedback behavior was quantified by
computing the sum of weighted frequencies with which participants chose each of the four
feedback sentences (“You are nice” was assigned a weight of +2, “You are nice” +1, “You are
not nice” -1, and “You are annoying” -2). The feedback index, which was separately computed
for the fair and unfair other, thus reflects the degree to which participants chose nice (positive
value) as opposed to derogatory messages (negative value). Overall, participants’ behavior in
the high power phase towards the unfair other can be interpreted as punishment behavior

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 6 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

(competitive economic choices and derogatory behavior) or forgiveness behavior (cooperative

economic choices and nice feedback), whereas social behavior towards the fair other is indica-
tive of aggressive behavior (competitive economic choices and derogatory feedback) or proso-
cial behavior (cooperative economic choices and nice feedback).
Skin Conductance Responses (SCRs). To obtain an objective measure of emotional
arousal, skin conductance responses (SCRs) were continuously recorded during the IG by
means of two electrodes attached to the index and middle finger of participants’ non-dominant
hand (the handgrip device was held with their dominant hand). Data were collected with the
MP150 Biopac System (Santa Barbara, CA, USA) and analyzed offline with Acqknowledge 4.1
(Goleta, CA, USA). Because of technical problems, we had no SCR data from five participants,
resulting in a final sample size of n = 35 for SCR data. SCRs were low-pass filtered at 0.25 Hz
and high-pass filtered at 0.05 Hz. SCRs were extracted for each frame of the IG (see Fig 1) and
each experimental condition. Our analysis focused on SCRs at the time of the high-power play-
er’s economic or feedback choice. Amplitudes of SCRs that began between 1 and 3 s after stim-
ulus onset were taken into account if they exceeded a threshold of 0.01 μS. Amplitudes were
log-transformed by using the formula Aln = ln(A + 1).
Effort. To obtain an objective measure of motivational involvement during the IG, partici-
pants were required to make their behavioral choices (with their dominant hand) by using an
isometric handgrip device (TDS121C), connected to the MP150 Biopac System (Santa Barbara,
CA, USA). Prior to the experiment, we calibrated the baseline (participants were merely hold-
ing the device) and the maximal force with which participants could press the device. For data
analysis, the data points were converted to the percentage of each participant’s maximal grip
force (% of maximum).
Facial Expressions. To assess spontaneous emotional reactions, participants’ facial expres-
sions were recorded throughout the IG by using a Logitech USB HD Pro web camera C920
(Apples, Switzerland) mounted on the computer screen. Video recordings were synchronized
with the physiological recordings through the MP150 Biopac System (Santa Barbara, CA,
USA). Because of technical problems, video data from two participants were not properly
recorded. This resulted in a sample size of n = 38 for facial expression data. The Computer
Expression Recognition Toolbox [42] was used to quantify the intensity of smiles and frowns,
indexing positive and negative affect, respectively. Based on the Facial Action Coding System
[33], this program extracts the intensity of several facial action units (AUs) from video data.
Past research [33] suggests that activity of the zygomaticus muscle (smile) is indicative of hap-
piness, while activity of the corrugator muscle (frown) is associated with negative emotions
such as anger, sadness, or fear [34]. Therefore, we used data from AU 12 (lip corner puller,
zygomaticus major, smile) and AU 4 (brow lowerer, corrugator, depressor glabellae, depressor
supercilii, frown) for our analyses. We smoothed and thresholded the data obtained from the
Computer Expression Recognition Toolbox with a custom made toolbox from our laboratory.
Food Allocation Task. To assess the propensity to punishment or retaliation following
anger provocation, subsequent to the IG, participants were invited to take part in another
experiment which was presented as unrelated but done with the same volunteers for conve-
nience. This experiment was described as a food evaluation task in which all three participants
had to allocate, taste, and evaluate different food items in separate rooms. The food allocation
task (for a detailed description, see see Supporting Materials and Methods in Text A in S1 File)
was used to examine the extent of prosocial or aggressive behavior (amount of allocated choco-
late sauce and wasabi, respectively) in a novel situation. With the addition of prosocial behav-
ior, the present paradigm can thus be seen as an extension of the hot-sauce paradigm [43].
After this task, the aim of the entire experiment was revealed and neither the participant nor
the two confederates actually tasted any food items.

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 7 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

Questionnaires. To assess the validity of the IG in relation to well-known individual dif-

ferences, we asked participants to fill in the following online questionnaires prior to the experi-
ment: the State-Trait Anger Expression Inventory (STAXI-II [44], French version [45]), the
Aggression Questionnaire (AQ [46], French version [47]), and the Levenson Self-Report Psy-
chopathy Scale [48], French version [49]). As previous studies found that questionnaire mea-
sures of physical and verbal aggression were positively related to behavioral activation
(behavioral approach) and negatively related to behavioral inhibition (aversive motivation)
[50], we also included the Behavioral Inhibition System/Behavioral Activation System (BIS/
BAS) scales [51] (French version [52]). To test for dimensions of empathy, we included the
Interpersonal Reactivity Index (IRI [20], French version [53]). Questionnaires were collected
prior to the experiment by using the online program Questionnaire Machine (developed by
Christoph Hofstetter, University of Geneva). Table A in S1 File summarizes the psychometric
characteristics of the sample. Information on the subscales of the questionnaires can be found
in the Supporting Materials and Methods in Text A in the S1 File.
During the experiment itself, participants also filled in additional questionnaires (all using
11-point response scales ranging from 0, not at all, to 10, extremely). In order to inform partici-
pants that the two others disliked spicy food, we administered a food preference questionnaire
at the beginning of the experiment. It asked “How much do you like the following flavors?” in
relation to four items: sweet, salty, mild, and spicy. When filling in this questionnaire, the two
alleged other participants explicitly stated that they dislike spicy food and fear eating it in the
food tasting part of the experiment. Emotions in relation to the IG (joy, anger, disappointment,
and sadness) were assessed with a self-report questionnaire depicting instances of the fair or
unfair others' economic choices and feedback messages. The self-report questionnaire also
assessed self-reported emotions (anger, joy, disappointment, regret, culpability, malicious joy,
generosity, and sadness) related to the participants’ own choices (cooperative or competitive
economic decisions, as well as nice or derogatory feedback messages to the fair and unfair
other, respectively). Self-report questionnaires on emotions were administered after the IG
because previous studies observed that explicit emotion assessments can reduce subsequent
aggressive behavior [54].
Finally, after the food allocation task, participants were given a debriefing questionnaire
that probed for their involvement in the economic interaction game, the extent to which they
felt the presence of the others during the game, and their impression of the others as fair, agree-
able, good-looking, and reliable. This questionnaire also asked participants about their strategy
during the economic interaction game, whether they took the food preferences of the others
into account when allocating the food items, and what they thought about the experiment’s
aim. These data served as additional information on whether participants believed that the
interactions were real.
Analyses. Statistical analyses were performed by using SPSS Statistics 21 software. We
conducted repeated measures multivariate analyses of variance (MANOVAs), as well as analy-
ses of variance (ANOVAs) and t-tests. Correlations between normally distributed variables
were tested with Pearson correlations (denoted by r) and correlations with variables that were
not parametrically distributed were tested with Spearman correlations (denoted by rs) on rank
transformed data.

Results
Aim 1: Validation of the IG
Validity of the IG. Detailed results related to the validity of the IG can be found in Text A
in the S1 File. In line with prediction 1a, the validity and efficacy of our paradigm was

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 8 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

supported by participants’ self-reports of high involvement (M = 7.45, SD = 1.71) and reports

of feeling the others’ presence during the game (M = 6.23, SD = 2.48; both measured on a scale
from 0, not at all, to 10, extremely). This was further supported by higher arousal responses
(SCRs) and stronger physical effort (grip force) in reaction to others’ deliberate economic
choices, in comparison to control conditions lacking social relevance (Fig A, Table B and Text
A in S1 File), in line with prediction 1b.
Emotions Elicited by the IG. Similarly, self-reports of emotions also supported our pre-
diction 1c, with higher ratings of anger in response to the unfair other´s behavior in the low-
power phase (Fig 2). In addition, the unfair other´s behavior induced stronger feelings of

Fig 2. Self-reported feelings in response to the fair other’s and the unfair other’s economic choices (A) and feedback choices (B) during the low
power phase of the Inequality Game. Charts depict means and 2 standard errors of the mean. ***p < .001; **p < .01.
doi:10.1371/[Link].0151028.g002

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 9 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

disappointment and sadness, whereas the fair other´s behavioral choice elicited more joy (for
details, see Text A in S1 File).
The external validity of the IG as a paradigm to elicit anger and aggression was confirmed
by correlations with questionnaire data pertaining to personality traits related to anger and
aggression (details in Text A in S1 File). Anger in response to competitive economic choices
correlated with trait aggression (AQ; rs = .41, p < .01) and angry temperament (STAXI-T;
rs = .33, p < .05). Persons who had a tendency to show angry reactions to provocation in their
everyday life (STAXI-R) made more competitive economic choices for the fair and unfair other
(both rs  .33, both p < .05) and chose more derogatory feedback for the fair and unfair other
(both r  .36, both p < .05). Details on participants´ emotional experiences related to the high-
power phase and the difference of emotions in the high- and low-power phase can be found in
Text A and in Tables C-E in S1 File). Finally, in line with prediction 1c, the behavior of the fair
and unfair other during the IG led to a more favorable evaluation of the fair other (Text A and
Table F in S1 File)–despite counterbalancing confederates to the fair and unfair condition.
Reciprocation of Fair and Unfair Behavior. To test for the behavioral impact of the fair-
ness manipulation in the low-power phase, we assessed whether participants displayed differ-
ential behavior towards the fair and unfair other during the subsequent high-power phase. A
repeated-measures MANOVA was performed with the within-subject factor other (two levels:
fair and unfair), the between-subject factor focus (positive qualities or physical description),
and the two dependent variables, economic choice and feedback choice. This revealed a large
effect of other (F(2,37) = 10.98, p < .001, ƞ2 = .37), while the effect of focus was not significant
(F(2,37) = 0.16, p = .86). Confirming prediction 1c on the behavioral level, univariate ANOVAs
showed that both economic choices (F(1,38) = 15.69, p < .001, ƞ2 = .29) and feedback choices
(F(1,38) = 13.17, p < .01, ƞ2 = .26) were more favorable for the fair other than for the unfair
other (Fig 3).
To further test whether participants’ economic and feedback behavior towards the fair and
unfair other differed from the respective behavior of the alleged others, we used one-sample t-
tests. These analyses revealed that, for both economic choices and feedback messages, the par-
ticipants’ behavior was less extreme than the behavior of the respective other (i.e., less prosocial
then the fair partner and less antisocial than the unfair; all t(39)  4.22; all p < .001, Fig 3).
Finally, we found that economic choices and feedback choices were positively correlated.
This was the case for behavior towards both the fair and the unfair other (both rs  .38, both p
< .05), indicating that participants’ behavior was consistent across event types.
Classification of Participants Reveals Preference for Conciliatory Behavior. On the
basis of data from the high-power phase, participants could be categorized according to their
preferences for cooperative, sanctioning, or competitive economic behavior (Fig 4A; see meth-
ods and prediction 1d). In our sample, 21 of 40 participants were prosocial, 11 sanctioning, 7
competitive, and 1 ambiguous (no systematic preference). As detailed in the results section of
Text A in the S1 File, the independent payoff structure of the IG was corroborated, as partici-
pants’ behavioral preferences for prosocial, sanctioning, or competitive behavior affected only
the others’ economic outcomes, while participants' own payoffs remained unaffected by these
preferences (Fig 4B). Furthermore, participants with distinct classifications also differed in
terms of anger-related personality traits (detailed in Text A in S1 File).
Using binomial tests, we determined whether participants’ actual behavior deviated from
behavior predicted by chance distribution. These tests revealed that despite exposure to unfair-
ness and insults in the low power phase, more participants than expected by chance preferred
to engage in prosocial behaviors towards both the fair and unfair other in the subsequent high
power phase. This was the case in terms of cooperative economic choices (p < .01) and nice
feedback messages (p < .001).

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 10 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

Fig 3. Participants’ behavior in the high power phase of the Inequality Game. Participants’ economic
choices (A) and feedback choices (B). For completeness, the corresponding behavior of the fair and unfair
other in the low power phase is displayed on the right-hand side of each plot. Charts depict means and 2
standard errors of the mean. ***p < .001; **p < .01.
doi:10.1371/[Link].0151028.g003

Aim 2: Relationship between Emotions and Behavior

To examine the relation between emotions and behavioral reactions to anger provocation, we
used several measures, including spontaneous displays of positive and negative affect during

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 11 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

Fig 4. Participants’ behavior in the high power phase of the Inequality Game. (A) Classification of participants based on their economic choices in the
high power phase and (B) hypothetical economic gains in the high power phase for each player split by inter-individual classification of participants. Chart A
depicts frequencies; Chart B depicts means and 2 standard errors of the mean. ***p < .001; **p < .01.
doi:10.1371/[Link].0151028.g004

the provocation phase and self-reports of specific emotional experiences in particular situations
of the IG. This allowed us to obtain both implicit and explicit measures, probing general
valence and more specific emotions, respectively.
Facial Expressions Related to Pleasant and Unpleasant Emotions. To test whether a
spontaneous display of happiness (with a smiling facial expression) as opposed to unpleasant
emotions (with a frowning facial expression) during the provocation phase may predict

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 12 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

subsequent behavioral reactions, we correlated the intensity of participants’ smiles and frowns
during the low power phase with their economic and feedback choices in the subsequent high
power phase. In line with our hypothesis (prediction 2a) that positive emotions are linked to
prosocial behavior, the average smile intensity during the low power phase predicted fair eco-
nomic choices towards the unfair other in the high power phase (rs = .28, p < .05, one-tailed).
No such effect was found for frowns (rs = -.2, p = .22). There was no relation between the dis-
play of frowns or smiles in the low power phase and feedback behavior in the high power phase
(all r  .1, all p  .56).
Relation between Self-Reported Emotions and Behavior. We also assessed whether spe-
cific emotional responses during the IG were related to different behaviors (prediction 2b). In
the low power phase, self-reported feelings (joy, anger, disappointment, and sadness) evoked
by others’ economic and feedback behavior, were not correlated with subsequent economic or
feedback behavior in the high power phase (all rs  .25, all p  .13). In contrast, behavioral
choices in the high power phase were correlated with some of the associated self-reported feel-
ings. When participants chose cooperative monetary outcomes for the fair other, stronger feel-
ings of regret were related to less cooperative economic behavior (rs = -.34, p < .05). Similarly,
stronger feelings of disappointment were related to less frequent nice feedback choices for the
fair other (rs = -.4, p < .05).
When participants interacted with the unfair other, self-reports of malicious joy during
punishment behavior correlated with the number of competitive economic choices (rs = .4,
p < .05). Similarly, malicious joy experienced during the delivery of derogatory feedback corre-
lated with the degree to which such feedback was given (rs = .44, p < .05). Self-reports of joy
pointed in the same direction (rs = .37, p < .05), while the experience of regret was negatively
related to the amount of derogatory feedback given (rs = -.4, p < .05). All other correlations were
not significant (all rs  .29, all p  .07), including the relation between state anger and punish-
ment behavior. A formal comparison of correlation coefficients with Williams’ t-test revealed
that malicious joy was more strongly related to competitive economic behavior (t(27) = 2.53,
p < .05) and derogatory feedback choices (t(26) = 2.83, p < .01) towards the unfair other than
anger was.

Aim 3: Relationship between Empathy-related Traits and Behavioral

Reactions to Anger Provocation
To examine the role of empathy, we tested whether empathic traits measured by the IRI sub-
scales perspective taking, empathic concern, and personal distress were differentially related to
participants’ behavior during the high power phase of the IG and during food allocation (pre-
diction 3a). As expected, the capacity for perspective taking (e.g., "When I’m upset at someone,
I usually try to ‘put myself in his shoes’ for a while") predicted nice feedback choices for the
unfair other (r = .39, p < .05) and cooperative economic choices for the unfair other (rs = .32,
p = .05). Perspective taking had no relation to behavior towards the fair other. In contrast, per-
sonal distress (e.g., "Being in a tense emotional situation scares me") predicted derogatory feed-
back towards both the fair and unfair other (r  .37, p < .05). Empathic concern had no
relationship with choices in the IG.
With regard to food allocation, empathic concern (e.g., "I often have tender, concerned feel-
ings for people less fortunate than me") predicted how much desirable food (chocolate sauce)
participants allocated to the unfair other (rs = .46, p < .01). Perspective taking and empathic
distress showed no significant correlation with behavior in this task.
These findings show that when anger is induced, persons high in perspective taking and
empathic concern displayed less punishment behaviors. Conversely, people prone to

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 13 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

experience distress displayed more verbal aggression, even towards fair others. The link
between the experience of distress and antisocial behavior is further corroborated by the obser-
vation that the higher participants scored on behavioral inhibition (BIS; e.g., “Criticism or
scolding hurts me quite a bit”), the more likely they were to send derogatory feedback to the
fair and unfair other (both r  .45, both p < .01). With regard to prediction 3b, perspective tak-
ing correlated with more positive affect (less frowning, more smiling) and more favorable eval-
uation of the unfair other (all p < .05; see details in the results of Text A in S1 File).

Aim 4: Factors Reducing Punishment Behavior

A last issue addressed in our paradigm concerns two potential ways of reducing punishment
and retaliation behavior after anger elicitation. First, we tested whether the instruction that
asked participants to focus on the positive qualities as opposed to the physical appearance of
the others prior to the task would influence subsequent punishment behavior in the IG (predic-
tion 4a). Contrary to our expectations, we observed no significant effect (see above). Second,
we examined carry-over effects between the IG and behavior in a new context–i.e., the subse-
quent food allocation task (question 4b). To this end, we conducted two analyses. A repeated
measures within-subject MANOVA was performed with the factor other (fair and unfair) and
the dependent variable allocated food (chocolate sauce as desirable food and wasabi as undesir-
able food). This analysis revealed an effect of the factor "other" with a medium effect size (F
(2,38) = 3.33, p < .05, ƞ2 = .15; Fig 5). A follow-up univariate ANOVA showed that participants
allocated more wasabi to the unfair other than to the fair other (F(1,39) = 6.37, p < .05). More-
over, paired t-tests revealed that participants allocated more chocolate sauce than wasabi to the
fair other (t(39) = 3.59, p < .01). We also tested whether punishment behavior in the food allo-
cation task was correlated with punishment behavior in the high-power phase of the IG. This
showed a positive trend for a relation between competitive economic choices and subsequent
wasabi allocation to the unfair other (rs = .28, p = .08), while feedback choices were not related
to wasabi allocation (rs = -.19, p = .25).

Fig 5. Amount of wasabi and chocolate sauce allocated to the fair and unfair other. Graph depicts means and 2 standard errors of the mean. **p < .01;
*p < .05.
doi:10.1371/[Link].0151028.g005

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 14 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

Discussion
Despite the high prevalence of conflict in our society, there has so far been a shortage of tools
to empirically study the impact of inter-individual differences on behavioral responses to anger
elicitation in controlled laboratory settings. We therefore developed the IG as a paradigm that
allows for the assessment of emotional responses to anger elicitation and subsequent other-
regarding behavior. First, we validated the IG using a multi-method approach with compre-
hensive questionnaires, behavioral measures, and physiological measures. In addition, we
introduced an extension of a previous measure of aggression–the hot sauce paradigm [43]–that
also included a measure of prosocial behavior. Converging evidence from these various mea-
sures demonstrates that our paradigm was effective and powerful to modulate emotional and
social responses in our participants.
Secondly, we took advantage of this new paradigm to examine several open questions con-
cerning the links between emotions, empathic traits, and social behavior. Our findings reveal
that active punishment behavior in the IG was primarily related to the experience of malicious
joy, rather than just to negative emotions (such as anger). This extends the notion that mali-
cious joy may be experienced when observing unfair others suffer [16], and that neural activa-
tions are increased in reward-related areas during the punishment of unfair others [17]. We
also found that empathic traits were related to forgiveness behavior and aggressive behavior in
opposing ways. Empathic concern and perspective taking predicted a reduction in punishment
behavior (which can be seen as an increase in forgiveness behavior), whereas personal distress
predicted aggressive behavior, even towards the fair other. These novel insights may not only
help in the understanding of emotional processes underlying punishment behavior, but may
also inform interventions that aim at promoting forgiveness behavior.

Validation of the IG
With regard to the validation of the IG and associated measures (Predictions 1a-1c), we found
the following. The efficacy and validity of the IG were confirmed by self-reports of high
involvement, while external validity was established by convergent correlations between self-
reports of anger and relevant trait questionnaires, such as the STAXI. Among self-reported
emotions, the unfair other’s provocation elicited anger, disappointment, and sadness. This
result confirms the notion that injustice [55] and insults [56, 57] elicit anger. It also shows that
rather than eliciting only anger, competitive economic choices and derogatory feedback mes-
sages can induce a wider range of negative emotions, as found in other conditions of injustice
[58]. This highlights the importance of studying several emotions in conflict-prone interac-
tions. Consistent with previous findings [59], fair behavior in the IG elicited joy. However,
whereas cooperative economic choices were related to joy irrespective of who chose them (the
other or the participant), competitive choices elicited a different pattern of emotions, depend-
ing on whether they were passively received or actively chosen. Receiving a competitive out-
come elicited more anger, more disappointment, and more sadness than making competitive
choices for the other. Conversely, actively choosing competitive outcomes for the other was
associated with more joy. This finding shows that emotions differ markedly depending on situ-
ational variations of power and agency [60] and underlines that a particular kind of joy may
contribute to the active punishment of unfair others (for the discussion of malicious joy, see
below).
Likewise, measures of arousal revealed that participants reacted to socially relevant situa-
tions (cooperative and competitive choices of others) with stronger physiological responses
(SCRs) and stronger grip force than they did to control conditions of no social relevance (win
and no win conditions). The elevated SCRs for both cooperative and competitive choices of

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 15 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

others confirm the sensitivity of arousal to both positive and negative events [30]. In contrast,
handgrip force increased when participants received cooperative as compared with competitive
distributions. These data are in accord with previous observations that physical effort is sensi-
tive to monetary reward motivation [31, 32], and they extend these findings to the domain of
social reward.
On the behavioral level, participants reciprocated the fair and unfair choices of others in the
IG, albeit to an attenuated extent compared to the fair and unfair other. Participants could be
categorized on the basis of these choices (prediction 1d), in line with the social value orienta-
tion framework [61], corresponding to prosocial, sanctioning, or competitive behaviors. Inter-
estingly, participants’ economic and feedback choices in the IG overall were more prosocial
than expected by chance distribution. This is in line with previous observations that aggression
is not a necessary consequence of provocation, frustration, or anger [15, 62], and raises the
question of how inter-individual differences in behavioral responses to anger elicitation can be
explained.

Role of emotions and empathy-related traits

We were able to relate social behavior in the IG with both implicit (prediction 2a) and explicit
(question 2b) measures of experienced emotions. By recording spontaneous facial expressions,
we found that more frequent smiles during the low power phase of the IG predicted subsequent
forgiveness behavior. This result extends the notion that positive emotions foster cooperation
while negative emotions increase competition [12, 13] by showing that such effects can be
unveiled through spontaneous facial expressions preceding social behavior. On the other hand,
although participants reported more anger when receiving unfair outcomes and when actively
engaging in punishment behavior, state anger in both cases was not related to participants' pro-
pensity for punishment behavior. This result may seem surprising in light of the observed cor-
relation between trait anger and aggression. However, it is in line with previous reports of a
weak relationship between self-reported state anger and aggression [63].
Whereas state anger did not correlate with aggression in the current experiment, active pun-
ishment behavior in the IG was related to the experience of malicious joy. This finding empha-
sizes that malicious joy is not only experienced when passively witnessing an unfair other
person experiencing pain [16], but that it may also be linked to active punishment behavior.
This observation broadens Berkowitz’s hypothesis by underlining that in addition to negative
emotions such as anger, reward-related emotions such as malicious joy can play a role in active
punishment. In contrast, we observed that the negative emotion of regret was related to a
reduction in both prosocial and punishment behavior, further arguing against the simple
notion that unpleasant emotions are linked to aggression [2]. Instead, it seems that the regret
could rather lead to a disengagement from the associated behavior.
With regard to the role of empathy-related traits (prediction 3a), we found that forgiveness
behavior was predicted by empathic concern and perspective taking. In contrast, individual
proneness to empathic distress was related to increased aggressive behavior, even towards the
fair and innocent other. This result limits and extends several previous findings. First, our data
underline that empathic traits are not necessarily linked to reduced aggressive behavior [25],
but that different empathic traits can be related to aggressive behavior in opposing ways. Previ-
ous studies on prosocial behavior reported that empathic distress was associated with reduced
helping [3], while empathic concern increased helping [3, 4]. We show that a similar distinc-
tion applies to antisocial behavior following anger induction.
Furthermore, the observed relation between perspective taking and conciliatory behavior
extends previous findings on the beneficial impact of perspective taking skills onto negotiation

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 16 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

outcomes [64]. In addition to the beneficial effects of perspective taking on altruistic behavior
and helping behavior [24], perspective taking in our experiment was positively related to the
engagement in forgiveness behavior. Overall, these results underline the importance of consid-
ering empathy as a multidimensional construct, especially when studying social interactions.
Interestingly, derogatory feedback messages to the fair and unfair other were not only pre-
dicted by personal distress, but also by the tendency for behavioral inhibition, a construct related
to anxiety and submissiveness [51]. Although this is in contrast to the previous observation that
behavioral inhibition is negatively related to aggression [50], the current data suggest that the
proneness to experience aversive social feelings is related to augmented antisocial behavior.
Given observations that positive affect is associated with more cooperation [1,12,13], we
hypothesized that focusing on positive qualities of the others might reduce subsequent punish-
ment behavior (prediction 4a). This hypothesis was not confirmed. This might be due to the
temporal delay between this description exercise (prior to the IG) and the opportunity for pun-
ishment behavior (in the second part of the IG). Future studies should test whether focusing on
the others' positive qualities at a different time or using a different exercise can affect social
behavior.
Finally, to test the hypothesis that “settling the score” may decrease subsequent aggression
[27] (question 4b), we also examined whether punishment behavior in the Inequality Task pre-
dicted the allocation of undesirable spicy food (wasabi) to the unfair other in a subsequent task.
Our data revealed a trend for economic punishment to predict the amount of wasabi allocated
to the unfair other. This finding indicates that an opportunity for "settling the score" in the
same situation did not diminish subsequent punishment. Rather, punishment behavior in the
first situation (IG) was related to augmented antisocial behavior in the subsequent food alloca-
tion situation. Taken together, in the present experiment, neither focusing on other´s positive
qualities nor “settling the score” reduced subsequent punishment.

Limitations and future directions

This paper introduces the IG as a novel tool to study behavioral reactions linked to anger
induction in controlled laboratory settings. We validate this task in a multi-method study and
reported predictors of social behavior at the level of emotions and empathy-related personality
traits. Despite these assets, there are several limitations to the results, which might be addressed
in future studies. A central limitation is that–apart from the manipulation of the others’ fair-
ness in the IG–most reported findings are correlational. Future studies are needed to test for
causal relationships and between personality traits, emotions, and social behavior. Moreover,
because of previously reported sex differences in aggression [35, 36], as well as emotional
responses to unfairness [16], the present study recruited male participants only. Future studies
are needed to test whether similar relations hold true in females.
Another limitation is linked to the recording of subjective feelings after the IG, which was
done to avoid interference of self-reports on behavior [54]. The temporal delay between provo-
cation and self-reports may have distorted participants’ recollection of their feelings during the
IG. In addition, self-reports of participants’ feelings were collected only in response to recipro-
cal situations (i.e., cooperative economic choices and nice feedback for the fair other and com-
petitive economic choices and derogatory feedback for the unfair other). To obtain finer-
grained insights, future studies should assess a broader range of situations. Although the pres-
ent study took into account physiological markers related to arousal, it would be interesting to
elucidate the neural substrates of anger and to dissociate them from aggressive behavior. In this
context, it will also be important to test which neural functions control social behavior and fos-
ter forgiveness in response to provocations.

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 17 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

Supporting Information
S1 File. Text A (including Supporting Material and Methods, Supporting Results, Support-
ing Discussion), Supporting Figure and Tables.
(PDF)

Acknowledgments
We thank Sylvain Delplanque for advice on the analysis of psychophysiology data and the
intern students Thibaut Cot, Lydia Lombardi, and Yoav Haccoun for their help.

Author Contributions
Conceived and designed the experiments: OK PV DS. Performed the experiments: OK. Ana-
lyzed the data: OK. Wrote the paper: OK PV DS.

References
1. Carnevale PJD, Isen AM. The influence of positive affect and visual access on the discovery of integra-
tive solutions in bilateral negotiation. Organ Behav Hum Decis Process. 1986; 37(1): 1–13.
2. Berkowitz L. On the formation and regulation of anger and aggression: A cognitive-neoassociationistic
analysis. Am Psychol. 1990; 45(4): 494. PMID: 2186678
3. Batson CD, O'Quin K, Fultz J, Vanderplas M, Isen AM. Influence of self-reported distress and empathy
on egoistic versus altruistic motivation to help. J Pers Soc Psychol. 1983; 45(3): 706–718.
4. Leiberg S, Klimecki O, Singer T. Short-term compassion training increases prosocial behavior in a
newly developed prosocial game. PLoS One, 2011; 6(3): e17798. doi: 10.1371/[Link].0017798
PMID: 21408020
5. Taylor SP. Aggressive behavior and physiological arousal as a function of provocation and tendency to
inhibit aggression. J Pers. 1967; 35(2): 297–310. PMID: 6059850
6. Cherek D, Roache JD, Egli M, Davis C, Spiga R, Cowan K. Acute effects of marijuana smoking on
aggressive, escape and point-maintained responding of male drug users. Psychopharmacology, 1993;
111(2): 163–168. PMID: 7870947
7. Dodge KA, Coie JD. Social-information-processing factors in reactive and proactive aggression in chil-
dren’s peer groups. J Pers Soc Psychol. 1987; 53(6): 1146–1158. PMID: 3694454
8. Komorita SS, Parks CD. Interpersonal relations: Mixed-motive interaction. Annu Rev Psychol., 1995;
46: 183–207.
9. Doob AN. Catharsis and aggression: The effect of hurting one's enemy. Journal of Experimental
Research in Personality, 1970; 4, 291–296.
10. Isen AM, Some ways in which positive affect influences decision making and problem solving. In: Lewis
M, Haviland-Jones JM, Feldman-Barrett L, editors. Handbook of emotions, 3, New York, NY, US: Guil-
ford Press; 2008. pp. 548–573.
11. Schwarz N, Clore GL. Mood, misattribution, and judgments of well-being: Informative and directive
functions of affective states. J Pers Soc Psychol. 1983; 45(3): 513.
12. Lyubomirsky S, King L, Diener E. The benefits of frequent positive affect: does happiness lead to suc-
cess? Psychol Bull, 2005; 131(6), 803. PMID: 16351326
13. Van Kleef GA, Sinaceur M. The demise of the 'rational' negotiator: Emotional forces in conflict and
negotiation. In: Olekalns MA, Adair W, editors. Handbook of research on negotiation. Cheltenham, UK:
Edward Elgar; 2013. pp. 103–131.
14. Lerner JS, Keltner D. Beyond valence: Toward a model of emotion-specific influences on judgement
and choice. Cogn Emot. 2000; 14(4): 473–493.
15. Averill JR. Studies on anger and aggression: Implications for theories of emotion. Am Psychol. 1983;
38(11): 1145–1160. PMID: 6650969
16. Singer T, Seymour B, O'Doherty JP, Stephan KE, Dolan RJ, Frith CD. Empathic neural responses are
modulated by the perceived fairness of others. Nature. 2006; 439(7075): 466–469. PMID: 16421576
17. de Quervain DJ, Fischbacher U, Treyer V, Schellhammer M, Schnyder U, Buck A, et al. The neural
basis of altruistic punishment. Science. 2004; 305(5688): 1254–1258. PMID: 15333831
18. Frith CD, Frith U. The neural basis of mentalizing. Neuron. 2006; 50(4): 531–534. PMID: 16701204

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 18 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

19. Klimecki O, Singer T. Empathy from the perspective of social neuroscience. In: Armony J, Vuilleumier
P, editors. Handbook of human affective neuroscience. New York: Cambridge University Press;
2013. pp. 533–549.
20. Davis MH. Measuring individual differences in empathy: Evidence for a multidimensional approach. J
Pers Soc Psychol. 1983; 44(1): 113–126.
21. Keltner D, Goetz JL. Compassion. In: Baumeister RF, Vohs KD, editors. Encyclopedia of social psy-
chology. Thousand Oaks, CA: Sage Publications; 2007. pp. 159–161.
22. Eisenberg N, Fabes RA, Miller PA, Fultz J, Schell R, Mathy RM, et al. Relation of sympathy and per-
sonal distress to prosocial behavior: A multimethod study. J Pers Soc Psychol. 1989; 57(1): 55–66.
PMID: 2754604
23. Klimecki OM. The plasticity of social emotions. Soc Neurosci. 2015. 10 (5): 466–473. doi: 10.1080/
17470919.2015.1087427 PMID: 26369728
24. Underwood B, Moore B. Perspective-taking and altruism. Psychol Bull. 1982; 91(1): 143–173.
25. Miller PA, Eisenberg N. The Relation of empathy to aggressive and externalizing/antisocial behavior.
Psychol Bull. 1988; 103(3): 324–344. PMID: 3289071
26. Klimecki OM, Leiberg S, Ricard M, Singer T. Differential pattern of functional brain plasticity after com-
passion and empathy training. Soc Cogn Affect Neurosci. 2014; 9(6): 873–879. doi: 10.1093/scan/
nst060 PMID: 23576808
27. Doob AN, Wood LE. Catharsis and aggression: Effects of annoyance and retaliation on aggressive
behavior. J Pers Soc Psychol. 1972; 22(2): 156. PMID: 5024076
28. Bushman BJ. Does venting anger feed or extinguish the flame? Catharsis, rumination, distraction,
anger, and aggressive responding. Pers Soc Psychol Bull. 2002; 28(6): 724–731.
29. Critchley HD. Electrodermal responses: What happens in the brain. Neuroscientist. 2002; 8(2): 132–
142. PMID: 11954558
30. Greenwald MK, Cook EW, Lang PJ. Affective judgment and psychophysiological response: Dimen-
sional covariation in the evaluation of pictorial stimuli. J Psychophysiol. 1989; 3(1): 51–64.
31. Talmi D, Seymour B, Dayan P, Dolan RJ. Human pavlovian-instrumental transfer. J Neurosci. 2008; 28
(2): 360–368. doi: 10.1523/JNEUROSCI.4028-07.2008 PMID: 18184778
32. Pessiglione M, Schmidt L, Draganski B, Kalisch R, Lau H, Dolan RJ, et al. How the brain translates
money into force: A neuroimaging study of subliminal motivation. Science. 2007; 316(5826): 904–906.
PMID: 17431137
33. Ekman P, Friesen WV, Hager JC. Facial Action Coding System investigator’s guide. Salt Lake City:
Research Nexus; 2002.
34. Ekman P. About brows: Emotional and conversational signals. In: von Cranach M, Foppa K, Lepenies
W, Ploog D, editors. Human ethology. Cambridge: Cambridge University Press; 1979. pp. 169–249.
35. Archer J. Sex differences in aggression in real-world settings: A meta-analytic review. Rev Gen Psy-
chol. 2004; 8(4): 291.
36. Björkqvist K. Sex differences in physical, verbal, and indirect aggression: A review of recent research.
Sex Roles. 1994; 30(3–4): 177–188.
37. Bagby RM, Parker JD, Taylor GJ. The twenty-item Toronto Alexithymia Scale: I. Item selection and
cross-validation of the factor structure. J Psychosom Res. 1994; 38(1): 23–32. PMID: 8126686
38. Loas G, Corcos M, Stephan P, Pellet J, Bizouard P, Venisse JL, et al. Factor analysis of the French ver-
sion of the 20-item Toronto Alexithymia Scale (TAS-20). Psychopathology. 1996; 29(2): 139–144.
PMID: 8861519
39. Beck A, Steer R, Brown G. (1996). The Beck Depression Inventory–Second Edition Manual. San Anto-
nio, TX: The Psychological Corporation.
40. Deutsch M. The effect of motivational orientation upon trust and suspicion. Hum Relat. 1960; 13(2):
123–139.
41. De Dreu CKW, Weingart LR, Kwon S. Influence of social motives on integrative negotiation: A meta-
analytic review and test of two theories. J Pers Soc Psychol. 2000; 78(5): 889–905. PMID: 10821196
42. Littlewort G, Whitehill J, Wu T, Fasel I, Frank M, Movellan J, Bartlett M. The Computer Expression Rec-
ognition Toolbox (CERT). In: IEEE International Conference on Automatic Face and Gesture Recogni-
tion. 2011: 298–305.
43. Lieberman JD, Solomon S, Greenberg J, McGregor HA. A hot new way to measure aggression: Hot
sauce allocation. Aggress Behav. 1999; 25(5): 331–348.
44. Spielberger CD. Manual for the State-Trait Anger Expression Inventory (STAXI). Odessa, FL: Psycho-
logical Assessment Resources; 1988.

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 19 / 20

 Impact of Emotions and Empathy-Related Traits on Punishment Behavior in the Inequality Game

45. Borteyrou X, Bruchon-Schweitzer M, Spielberger CD. The French adaptation of the STAXI-2, C.D.
Spielberger's State-trait anger expression inventory. Encephale. 2008; 34(3): 249–255. doi: 10.1016/j.
encep.2007.06.001 PMID: 18558145
46. Buss AH, Perry M. The Aggression Questionnaire. J Pers Soc Psychol. 1992; 63(3): 452–459. PMID:
1403624
47. Bouchard J. Validation de la version française du AQ auprès de deux échantillons: étudiants universi-
taires (étude1) et adultes non-recrutés en milieu universitaire (étude 2). Chicoutimi, Canada: 2007.
48. Levenson MR, Kiehl KA, Fitzpatrick CM. Assessing psychopathic attributes in a noninstitutionalized
population. J Pers Soc Psychol. 1995; 68(1): 151–158. PMID: 7861311
49. Savard C, Lussier Y, Sabourin S. French-Canadian validation of the Levenson self-report psychopathy
scale. In: Annual Meeting of the Canadian Psychological Association. Montréal: Canada; 2005.
50. Smits DJM, Kuppens P. The relations between anger, coping with anger, and aggression, and the BIS/
BAS system. Pers Individ Dif. 2005; 39(4): 783–793.
51. Carver CS, White TL. Behavioral-inhibition, behavioral activation, and affective responses to impending
reward and punishment: The BIS BAS scales. J Pers Soc Psychol. 1994; 67(2): 319–333.
52. Caci H, Deschaux O, Bayle FJ. Psychometric properties of the French versions of the BIS/BAS scales
and the SPSRQ. Pers Individ Dif. 2007; 42(6): 987–998.
53. Gilet A-L, Mella N, Studer J, Gruehn D, Labouvie-Vief G. Assessing dispositional empathy in adults: A
French validation of the Interpersonal Reactivity Index (IRI). Can J Behav Sci. 2013; 45(1): 42–48.
54. Berkowitz L. Aggression: Its causes, consequences, and control. Philadelphia: Temple University
Press; 1993.
55. Pillutla MM, Murnighan JK. Unfairness, anger, and spite: Emotional rejections of ultimatum offers.
Organ Behav Hum Decis Process. 1996; 68(3): 208–224.
56. Harmon-Jones E, Sigelman J. State anger and prefrontal brain activity: Evidence that insult-related rel-
ative left-prefrontal activation is associated with experienced anger and aggression. J Pers Soc Psy-
chol. 2001; 80(5): 797–803. PMID: 11374750
57. Denson TF, Pedersen W, Ronquillo J, Nandy AS The angry brain: Neural correlates of anger, angry
rumination, and aggressive personality. J Cogn Neurosci. 2009; 21(4): 734–744. doi: 10.1162/jocn.
2009.21051 PMID: 18578600
58. Mikula G, Scherer KR, Athenstaedt U. The role of injustice in the elicitation of differential emotional
reactions. Pers Soc Psychol Bull. 1998; 24(7): 769–783.
59. Hegtvedt KA, Killian C. Fairness and emotions: Reactions to the process and outcomes of negotiations.
Soc Forces. 1999; 78(1): 269–302.
60. Scherer KR. Appraisal considered as a process of multilevel sequential checking. In: Scherer KR,
Schorr A, Johnstone T, editors. Appraisal processes in emotion: Theory, methods, research. New
York: Oxford University Press; 2001. pp. 92–120.
61. Van Lange PAM. The pursuit of joint outcomes and equality in outcomes: An integrative model of social
value orientation. J Pers Soc Psychol. 1999; 77(2): 337–349.
62. Berkowitz L. Frustration aggression hypothesis: Examination and reformulation. Psychol Bull. 1989;
106(1): 59–73. PMID: 2667009
63. Kassinove H, Roth D, Gregory Owens S, Fuller JR. Effects of trait anger and anger expression style on
competitive attack responses in a wartime prisoner’s dilemma game. Aggressive Behavior. 2002. 28
(2): 117–125.
64. Galinsky AD, Ku G. The effects of perspective-taking on prejudice: The moderating role of self-evalua-
tion. Pers Soc Psychol Bull. 2004; 30(5): 594–604. PMID: 15107159

PLOS ONE | DOI:10.1371/[Link].0151028 March 15, 2016 20 / 20

1967] Reviews

Confronting Injustice: The Edmond Cahn Reader. EDITED BY LENORE

CAHN. Little, Brown & Co., Boston, 1966. Pp. xxiv, 428. $8.95.

In his tribute to the late Professor Edmond Cahn, which opened the
memorial edition of the New York University Law Review, Justice
Hugo Black took a text from Cahn's first book, The Sense of Injustice.
Wrote Black:
"Here was a man-a lawyer-who was not ashamed to endorse the
idea that 'justice or righteousness is the source, the substance, and the
ultimate end of law,' and that a legal concept is most worthwhile when
it 'becomes relevant to the homely experiences of individual human
beings.' This was to me a pleasing contrast to the idea expressed by
many that 'courts do not sit to administer justice, but to administer
the law,' as though there is a kind of hostility between the two." 1
Edmond Cahn's precept concerning justice and injustice was car-
ried forward in his second book, The Moral Decision, and in his last
published work, The Predicament of Democratic Man, the concept
was broadened to take in the entire society. It is by no means surprising
that at the moment of his untimely death, in 1964, he was writing a
book with the assigned title, The Meaning of Justice. It is consequently
wholly fitting that "The Edmond Cahn Reader," drawn together from
many sources and posthumously published in 1966, should be entitled
Confronting Injustice.
Thanks to this concentration of thought upon the meaning of
justice, during the twenty years after he ceased to be a tax lawyer to
become legal philosopher, the close-to-forty articles here assembled
from legal and other magazines, public addresses and correspondence,
have a truly remarkable consistency and cohesion. It is almost as if he
had completed the book whose two written chapters tragically termi-
nate the present volume. The style is clear, lucid and charming, giving
no hint of the heavy labors Cahn is said to have bestowed upon his
writing.
Credit for selection of the contents of this book belongs to Lenore
Cahn, who played an active role in the development of her husband's
ideas and ideals. The introduction and illuminating headnotes are
from the pen of Norman Redlich, former student and faculty associate
of Professor Cahn, and at present executive assistant corporation coun-
sel of the City of New York. His contribution combines deserved
eulogy with exposition of Cahn's ideas and purposes and has a clarify-
ing and unifying effect.
The basic unity, however, is in Cahn's thought. It is significant that
1 Black, About Edmond Cahn, 40 N.Y.U.L. REv. 207 (1965).
 The University of Chicago Law Review [Vol. 34:963
the opening and near-closing articles of Confronting Injustice have
closely-related titles, "The Consumers of Injustice" and "The Shift to
a Consumer Perspective." The first is a reprint of Cahn's Horace M.
Kallen Lecture of 1959 at the New School for Social Research; the
last presents the fundamental idea of the book whose production was
precluded by his death. In both he was dealing with a new and truly
American factor in law and politics-the shift of perspective from legal
concepts held in monarchies, oligarchies and aristocracies to the power
and responsibility of the citizen in a representative democracy. The
old perspective was that of the ruler or the government official; the
new perspective is that of "the democratic citizen in the role of con-
sumer of the law." 2 Not that this shift is fully achieved in American
jurisprudencel
Injustice persists, in Cahn's estimation, because the old imperial
perspective continues to dominate legislatures and courts through
tradition, ignorance and the self-interest of its beneficiaries. "The
ugliest sign of our thralldom to the old outlook is that it tends to
desensitize men of fine intellect and good will."3 They learn somehow
not to notice what happens to individuals, and suppress their own in-
evitable involvement in the ills of the victims. Such men beguile them-
selves with a rule of averages. The "third degree" and such horrors are
not used (or at least exposed) frequently, so on balance all is well. Even
Justice Cardozo, rightly "regarded as a paragon of moral insight,"4
wrote in his most celebrated book that "the eccentricities of judges
balance one another," causing a little confusion until, happily, "The
future takes care of such things." 5
"Revere Cardozo as we may," comments Edmond Cahn, "we cannot
help retorting that averages in the administration of justice do not
avail the person who is wronged grievously in his own, particular
case." 6 Official misconduct, seen, heard or read about, arouses the
sense of injustice, stirring the sympathetic reaction of outrage, resent-
ment, and anger, "for our physiology has equipped us to regard an act
'
of injustice to another as a personal aggression against ourselves. T
Is that all? Clearly not, adds Cahn. "While the sense of injustice
uses empathy, projection, and emotion, it simultaneously summons
perception, reasoning, intelligence, and judgment-all the capacities
2 E. CAHN, CONFRONTING INJUSTICE 6 (L. Cahn ed. 1966).
3 Id. at 7.
4 Id. at 9.
5 Id. at 9-10.
6 Id. at 10.
7 Id. at 12.
1967] Reviews

that make for understanding and the application of sense. ' 8 The
author, using the method he employed so strikingly in The Moral De-
cision, goes to the courts to trace the transition that has taken place in
the last forty years, in reaction, for example, against the excesses of the
prosecution and the callousness of judges in the case of Tom Mooney.
Cahn's keen perceptions govern the several chapters devoted to the
Bill of Rights, especially to freedom of religion and the press. The
title of one of them, "The Firstness of the First Amendment," fur-
nishes a sufficient indication of this emphasis. The touchstone is James
Madison's prophecy in 1789 that if these rights are implanted in the
Constitution, independent tribunals of justice "will be an impene-
trable bulwark against every assumption of power in the legislative or
executive." 9 Actually, the rights thus implanted have been frustrated,
followed by a belated movement of fulfillment in the judicial opinions
of Justices Holmes, Brandeis, Stone, Cardozo, Hughes, Black, Douglas,
Murphy, Rutledge, Brennan, Warren and others who have made up
occasional and precarious majorities.
Like an interlude from the Middle Ages is a chapter comprised of
1962 correspondence on freedom of the press between Professor Cahn
and David Ben-Gurion, then Prime Minister of Israel. Cahn wrote to
protest against a proposed Israeli law that would enable the govern-
ment to shut down any newspaper found guilty of publishing two
libels in the course of two years. Almost any reflection on a public
person, the nation, or the people in general was to be considered libel-
ous. Ben-Gurion stuck to his guns, defending the doctrine of "prior
restraint" just as it was defended by Archbishop Whitgift and Star
Chamber judges in Elizabethan England. He then drew the same
distinction between "freedom" and "licentiousness" of the press that
was drawn by Blackstone in 1769 and by the framers of the American
Alien and Sedition Acts of 1798. In one of his replies Cahn assailed
the concept of "group libel" as something that precluded the defense
of truth or reasonable basis of belief-an argument that he could have
directed equally against the United States Supreme Court for its de-
cision in Chicago's Beauharnaiscase.
In his exchange with Ben-Gurion, Cahn carefully limited the issue
to restraint previous to publication. He was not talking about punish-
ment of published libels. That same tactic was followed by English
libertarians in the eighteenth century, when they concentrated on re-
form of the libel law instead of advocating complete freedom of opin-
8 Id. at 13.
9 Id. at 90.
966 The University of Chicago Law Review

ion. That has been taken by modern defenders of seditious libel laws
as an indication that Englishmen, and inferentially the sponsors of
the first amendment, accepted the punishment of published utterances.
But Cahn elsewhere rejects that utterly-just as English libertarians
would have done if they had discerned the remotest chance of success.
Cahn's commitment to absolute freedom of expression put him on
the side of Justices Black and Douglas and against the "balancing
doctrine" of Justice Felix Frankfurter. The latter construed the words
of the first amendment, "Congress shall make no law" restricting the
freedom of religion, speech, press, or assembly, to mean that Congress
shall make no such law unless there is sufficient reason for doing so. But
Cahn took note of overlapping inconsistencies on both sides. Black and
Douglas grade the first eight amendments, making the first pre-eminent.
But they put them all on a par in judging the weight of the fourteenth
amendment upon them. Frankfurter grades them with respect to in-
clusion under the fourteenth amendment, excluding some from its
force, but he puts them on a par in relation to their intrinsic force and
importance. Cahn clarifies these inconsistencies in terms of constitu-
tional realities. Judged by their fruits, "the composite doctrines of
Justices Black and Douglas are designed consistently to promote the
constitutional guarantees, and the composite doctrines of Justice
Frankfurter are designed as consistently to subordinate, demote, or
delete them.""'
The Cahn Reader heartily approves Judge Jerome Frank's doctrine
of "fact-skepticism": a questioning of the reliability of evidence based
on the senses of sight, sound and memory. This goes to the heart of the
question of justice in the criminal courts, and is climactic on the subject
of capital punishment. On that theme the Reader reproduces Cahn's
introduction to Arthur Koestler's Reflections on Hanging. In this
preface he vividly transfers the impact of English legal atrocities to the
American scene.
Running all through Confronting Injustice is the appeal for a new
outlook in American legal education. It is written for law students,
faculties, lawyers and judges, and has in it the seeds of growth of a
modem miracle in the law.
Irving Bran
10 Id. at 303.
t Irving Brant is a noted journalist and historian.
Journal of Experimental Psychology: General © 2017 American Psychological Association
2017, Vol. 146, No. 10, 1448 –1459 0096-3445/17/$12.00 [Link]

The Secret to Happiness: Feeling Good or Feeling Right?

Maya Tamir Shalom H. Schwartz

The Hebrew University The Hebrew University and National Research
University—Higher School of Economics

Shige Oishi Min Y. Kim

The University of Virginia Keimyung University

Which emotional experiences should people pursue to optimize happiness? According to traditional subjective
well-being research, the more pleasant emotions we experience, the happier we are. According to Aristotle,
the more we experience the emotions we want to experience, the happier we are. We tested both predictions
in a cross-cultural sample of 2,324 participants from 8 countries around the world. We assessed experienced
emotions, desired emotions, and indices of well-being and depressive symptoms. Across cultures, happier
people were those who more often experienced emotions they wanted to experience, whether these were
pleasant (e.g., love) or unpleasant (e.g., hatred). This pattern applied even to people who wanted to feel less
pleasant or more unpleasant emotions than they actually felt. Controlling for differences in experienced and
desired emotions left the pattern unchanged. These findings suggest that happiness involves experiencing
emotions that feel right, whether they feel good or not.

Keywords: happiness, well-being, emotion, culture, motivation

Supplemental materials: [Link]

Happiness is often defined as “a state of well-being and con- Such emotions are not necessarily pleasant emotions and may even be
tentment” (Merriam-Webster, n.d.). It is perhaps one of the most unpleasant, like anger or fear (Thomson, 1955). Indeed, Aristotle held
salient of human pursuits (Diener, Sapyta, & Suh, 1998). How can that the absence of unpleasant emotions is not an indicator of happi-
one attain this state of well-being? One answer is by increasing ness. Instead, happiness is linked to feeling unpleasant emotions when
pleasure and decreasing pain (Kahneman, 1999). Indeed, some they are appropriate and goal-conducive. The present research is a
psychologists argue that happiness involves maximizing pleasant first attempt to test whether feeling the right emotions may be critical
emotions and minimizing unpleasant emotions (e.g., Diener, 1984; in attaining happiness.
Kahneman, Diener, & Schwarz, 1999; Kuppens, Realo, & Diener,
2008; Lucas, Diener, & Suh, 1996). This approach has dominated What Are the Right Emotions?
the field of subjective well-being (SWB) over the last 30 years (see
For Aristotle, happiness entails experiencing the right emotions
Diener, 2013, for a recent review). The present article explores
(Nicomachean Ethics, 1105b25– 6). In Book 2 of Nicomachean
another possible answer based on the Aristotelian approach. Aris-
Ethics, Aristotle states that “to have these feelings at the right
totle suggested that happiness involves feeling the right emotions.
times on the right grounds towards the right people for the right
motive and in the right way is (. . .) the mark of virtue” (1106b9 –
1107a1; Thomson, 1955, p. 101). Anger, fear, as well as pleasure
This article was published Online First August 14, 2017.
are right, for some people, for some reasons. For example, for a
Maya Tamir, Department of Psychology, The Hebrew University; Sha- minority group member who seeks justice because people in the
lom H. Schwartz, Department of Psychology, The Hebrew University; and majority mistreat him, feeling anger may just be the right emotion.
Department of Psychology, National Research University—Higher School Whether an emotion is right, therefore, depends on the goals and
of Economics, Moscow; Shige Oishi, Department of Psychology, The needs of each individual. Whereas anger may feel right to some, it
University of Virginia; and Min Y. Kim, Department of Psychology, may feel wrong to others. Happiness, according to Aristotle,
Keimyung University. should involve feeling emotions that people deem to be appropri-
This work was supported by Grant 794/11 from the Israel Science ate given their needs and motives. Building on Aristotle’s account,
Foundation. We thank Jan Cieciuch, Michaela Riediger, Claudio Torres, therefore, we define “feeling right” as feeling emotions that one
Christie Scollon, Vivian Dzokoto, and Xiaolu Zhou for their help with data
considers to be desirable.
collection. The ideas and data presented in this article were previously
presented at colloquium talks at the Hebrew University and University of
Which emotions people consider desirable differs systematically
California, Berkeley. across situations, individuals, and cultures. Individuals differ in the
Correspondence concerning this article should be addressed to Maya extent to which they desire pleasant states. For instance, whereas
Tamir, Department of Psychology, The Hebrew University, Mount Scopus, some individuals desire high arousal pleasant emotions, such as
Jerusalem, 91905, Israel. E-mail: tamirm@[Link] excitement, others desire low activation pleasant emotions, such as

1448
 FEELING GOOD OR FEELING RIGHT 1449

calmness (Rusting & Larsen, 1995). This pattern also differs by viduals and cultures depending on their values. For instance,
culture, such that Americans, on average, desire high arousal satisfaction with one’s daily achievements predicted the overall
pleasant emotions more than East Asians do (Tsai, Knutson, & daily life satisfaction of individuals more strongly if they were
Fung, 2006). Individuals also differ in the extent to which they high (vs. low) in achievement values (Oishi, Diener, Lucas, & Suh,
desire unpleasant states. For instance, European Americans seem 1999). Likewise, satisfaction with one’s finances predicted life
to be more motivated to minimize unpleasant states, compared satisfaction more in poor countries than in wealthy ones (Oishi,
with Germans (Koopmann-Holm & Tsai, 2014). In addition, Diener, Lucas, & Suh, 1999). This pattern also extends to hedonic
whereas European Americans seek to maximize pleasant experi- experiences. For instance, daily physical pleasure predicted overall
ences and minimize unpleasant experiences, members of collec- daily satisfaction more among people high (vs. low) in sensation
tivistic cultures seek more balanced emotional experiences, and are seeking (Oishi, Schimmack, & Diener, 2001). Similarly, experi-
less motivated to minimize unpleasant experiences (e.g., Miy- encing excitement increased life satisfaction more among individ-
amoto, Ma, & Petermann, 2014; Sims et al., 2015). uals high (vs. low) in sensation seeking (Oishi, Schimmack, &
We have recently proposed that people’s values are one factor Colcombe, 2003, Study 5). At the national level, positive emotions
that might determine which emotions they desire (Tamir et al., related more strongly to life satisfaction in countries that stress
2016). In a cross-cultural study, we found that people desired self-expression values than in countries that stress survival values
emotions that were consistent with their core values. For instance, (Kuppens et al., 2008).
individuals who valued self-transcendence (benevolence, univer- To the extent that value priorities differ across cultures
salism) desired more love, trust, and compassion (i.e., self- (Schwartz, 2011), it is reasonable to assume that different emo-
transcending emotions) than others did, whereas those who valued tions may be associated with happiness across cultures (Mesquita,
self-enhancement (power, achievement) desired more pride, but de Leersnyder, & Albert, 2014). Kitayama, Markus, and Kurokawa
also more anger, hatred and contempt (i.e., self-enhancing emo- (2000), for instance, found that interpersonally disengaging pleas-
tions) than others did. ant emotions (e.g., pride) were strongly associated with happiness
Aristotle claimed that feeling emotions that are consistent with among Americans, whereas interpersonally engaging pleasant
one’s values feels right, and feeling right relates to greater happi- emotions (e.g., fureai) were strongly associated with happiness
ness. Is this indeed the case? Furthermore, is this the case even among Japanese (see also Kitayama, Ishii, Imada, Takemura, &
when the right emotions are unpleasant to experience? To address Ramaswamy, 2006). de Leersnyder, Kim, and Mesquita (2015)
these questions, the present investigation tested whether individu- reported that people are happier the more they experience emo-
als are happier if they experience the emotions they desire than if tions that are characteristic of their culture. They found that the
they do not, whether the emotions are pleasant or unpleasant. higher the correlation between a person’s emotion profile and the
We assessed the degree to which a person feels the right emo- average profile of their culture the greater that person’s well-being
tion by computing the discrepancy between the amount of the among Koreans, Belgians, and European Americans. In general,
emotion the person desires to feel and the amount of that emotion there is evidence that people tend to be happier when their personal
the person actually feels. Previous research on pleasant emotions values are consistent with the dominant cultural values (e.g.,
found that the smaller the discrepancy between experienced and Fulmer et al., 2010).
desired pleasant states, the more satisfied people are (e.g., Kämpfe Both values and desired emotions differ across cultures (e.g.,
& Mitte, 2009; Larsen & McKibban, 2008; Rice, McFarlin, & Eid & Diener, 2001; Schwartz, 2011). However, according to the
Bennett, 1989). When individuals desire pleasant emotions, the Aristotelian account, regardless of their culture, individuals should
Aristotelian prediction is the same as the prediction of traditional be happier the more they feel emotions that they personally deem
SWB researchers: people are happier if they experience as much of desirable. Thus, although experienced emotions and desired emo-
a pleasant emotion as they desire. However, when people desire tions may differ across cultures, the degree to which individuals
unpleasant emotions, the Aristotelian prediction and the prediction experience discrepancies between them (i.e., the degree to which
of traditional SWB researchers are in stark contrast. The Aristo- people feel ‘right’) should relate consistently to happiness across
telian prediction is that such people would be happier the more cultures. An alternative prediction is that feeling right relates
they feel the emotion they desire, even though that emotion is differently to happiness across cultures. For instance, feeling right
unpleasant. The traditional SWB prediction is that people would be might relate more strongly to happiness in wealthier countries than
happier the less they feel that unpleasant emotion, whether they in poorer countries, where happiness may depend more on satis-
desire it or not. That is because SWB researchers typically treat faction of basic needs (see Tay & Diener, 2011). To test these
pleasant emotions as good and unpleasant emotions as bad (at least hypotheses, the present investigation examined the links between
for SWB; Diener, 1984; Kahneman, 1999). feeling right and happiness across a set of countries that vary on an
index of general development (United Nations Development Pro-
Right Emotions and Happiness Across Individuals gramme, 2014).
and Cultures
The Current Research
To our knowledge, the Aristotelian claim regarding feeling right
has not been explicitly tested in well-being research. However, this Previous research on the link between well-being and feeling
claim is consistent with several existing theoretical approaches. “right” focused exclusively on pleasant affective experiences (e.g.,
For example, according to the value-as-a-moderator model of excitement in Oishi et al., 2003) and a small set of cultures (e.g.,
SWB (Oishi, Diener, Suh, & Lucas, 1999), sources of SWB (e.g., Japan vs. U.S.). None of the previous research tested whether
achievement, self-esteem, social relationships) differ across indi- feeling “right” predicts life satisfaction and depressive symptoms
1450 TAMIR, SCHWARTZ, OISHI, AND KIM

above and beyond feeling “good.” The current research assessed (i.e., creating words from letters in longer words; e.g., ‘go’ from
the desirability of distinct emotions, including both pleasant (e.g., ‘geography’). Finally, participants rated their experienced emo-
love and compassion) and unpleasant (e.g., anger and hatred) tions, completed indices of depressive symptoms and well-being,
emotions, and tested the Aristotelian prediction across cultures. and provided demographic information.
The inclusion of distinct emotions enabled us to test whether
happiness is related to feeling right across distinct emotions or
Materials
whether happiness is related to experiencing certain right emotions
more than others. We assessed relations of the absolute discrep- Desired and experienced emotions. To assess desired emo-
ancies between experienced and desired emotions with greater tions, participants rated how often they wanted to experience
well-being and depressive symptoms in eight countries around the various emotions in their daily life. To assess the actual experience
globe. of emotions, participants rated how often they experienced the
We recently identified four emotion categories that map onto same emotions in their daily life. All ratings were made on 1
key dimensions of core values (i.e., self-transcendence, self- (never) to 5 (most of the time) scales. We assessed self-
enhancement, openness to experience, and conservation; Tamir et transcending emotions (i.e., love, affection, trust, empathy, com-
al., 2016). We demonstrated that, across cultures, people desire passion; ␣ ⫽ .59 and .66 for desired and experienced emotions,
emotions that are more consistent with their values. People who respectively), negative self-enhancing emotions (i.e., anger, con-
value self-transcendence (e.g., universalism), for instance, desire tempt, hostility, hatred; ␣ ⫽ .66 and .70 for desired and experi-
more empathy than others do. Similarly, people who value self- enced emotions, respectively),2 opening emotions (i.e., interest,
enhancement (e.g., power) desire more anger than others do. curiosity, excitement, enthusiasm, passion; ␣ ⫽ .64 and .70 for
Would people who desire more anger be happier if they experi- desired and experienced emotions, respectively), and conserving
enced more anger? Or would experiencing more anger make them emotions (i.e., calmness, relaxation, relief, contentment; ␣ ⫽ .53
less happy? To address these questions, we assessed discrepancies and .67 for desired and experienced emotions, respectively). Emo-
between experienced and desired self-transcending emotions (e.g., tion terms were presented in a predetermined and fixed order.3
love, empathy), negative self-enhancing emotions (e.g., anger, To confirm that our emotion indices were empirically distinct,
hatred), opening emotions (e.g., excitement, interest) and conserv- we ran principal axis factor analyses with varimax rotation, sepa-
ing emotions (e.g., calmness, relief). We tested whether smaller rately on desired and experienced emotions, imposing four-factor
absolute discrepancies are linked to higher life satisfaction and less solutions. The scree plots and eigenvalues supported the choice of
depressive symptoms. four factor solutions. Table 2 presents the varimax rotated factor
matrices for experienced and desired emotions, respectively. The
Method table presents only loadings greater than .20. In each matrix, the
items assigned a priori to each emotion composite loaded posi-
tively and most highly on one factor. For desired emotions, neg-
Participants
ative self-enhancing emotions loaded on Factor 1, conserving
Participants in this study came from eight countries (i.e., United emotions on factor 2, opening emotions on factor 3, and self-
States, Brazil, China, Germany, Ghana, Israel, Poland, and Singa- transcending emotions on factor 4. For experienced emotions,
pore). We chose these countries to represent seven of the eight conserving emotions loaded on factor 1, negative self-enhancing
world cultural regions (i.e., Anglo, Latin American, Confucian, emotions on factor 2, self-transcending emotions on factor 3, and
West European, Sub-Saharan African, East Central European, and opening emotions on factor 4. These findings lend support to the
South Asian) distinguished by both Inglehart and Baker (2000) and discriminant validity of the four desired and experienced emotions
Schwartz (2006). We recruited only native speakers in each coun- scales.
try. The entire sample included 2,324 university students (57.5%
female, Mage ⫽ 22.47), recruited either through ads or through 1
In Tamir et al. (2016), we examined links between values and desired
local participant pools in their respective universities. Participants emotions, controlling for experienced emotions. No analyses were reported
received monetary compensation (equivalent to $3–$5) or course in that paper involving well-being or depressive symptoms. In contrast, the
credit. Table 1 presents the characteristics of each sample. current study investigates relations of well-being and depressive symptoms
to discrepancies between the two types of emotions. This study does not
concern values at all and includes the emotions themselves as controls
Procedure rather than the predictors of interest.
2
The original composite included ‘pride’ as an additional item. How-
Participants completed the study in their native language or in ever, to compare feeling ‘right’ to ‘feeling good’ we created a composite
their formal language of instruction. They responded either online that is exclusively negative in valence and dropped this item from the
(i.e., the study was administered on Qualtrics) or in writing (see composite. We refer to the new composite as capturing ‘negative self-
Table 1). For non-English versions, we carried out iterations of enhancing emotions’.
3
For desired emotions, items were presented in the following order:
translation and back-translation by independent bilinguals until we contentment, affection, enthusiasm, interest, compassion, contempt, curi-
obtained satisfactory versions. Separate gender-matched versions osity, trust, excitement, hostility, pride, empathy, passion, hatred, calm-
of the survey were used in those languages that distinguish gender. ness, relief, relaxation, love, and anger. For experienced emotions, items
After giving consent, participants first completed a scale of basic were presented in the following order: empathy, compassion, excitement,
hostility, relief, calmness, passion, pride, enthusiasm, love, contempt, cu-
values (PVQ-R; Schwartz et al., 2012).1 Participants then rated riosity, anger, relaxation, contentment, affection, interest, trust, and ha-
their desired emotions. To minimize carryover effects, they next tred. The order of the desired emotion terms differed from the order of
completed an unrelated, affectively neutral task for five minutes experienced emotion terms.
 FEELING GOOD OR FEELING RIGHT 1451

Table 1
Sample Characteristics

Country N % Female Language Age M (SD) % Married Mode % Psychology majors

Brazil 653 64 Portuguese 24.35 (5.78) 8 Online 17

China 213 53 Chinese 20.82 (1.97) 1 Paper and pencil 53
Germany 200 50 German 25.03 (4.12) 10 Paper and pencil and online 30
Ghana 207 59 English 22.90 (2.81) 3 Paper and pencil 46
Israel 248 53 Hebrew 24.21 (2.91) 12 Online 24
Poland 299 52 Polish 21.72 (1.74) 3 Paper and pencil 7
Singapore 201 69 English 21.23 (1.83) 0 Online 12
United States 303 54 English 19.51 (1.80) 0 Online 34

Well-being and depressive symptoms. We assessed well- their latent factors were equal across cultures (i.e., metric invari-
being with the Satisfaction With Life Scale (SWLS; Diener, Em- ance). All our measures demonstrated either partial or full metric
mons, Larsen, & Griffin, 1985) that includes 5 items (␣ ⫽ .83) invariance. This indicates that each of the cultural groups calibrate
rated on a scale ranging from 1 (strongly disagree) to 7 (strongly their measures the same way, so the manifest scales have the same
agree) scale. We measured depressive symptoms with the 10-item meaning across groups. This justifies comparing patterns of asso-
(␣ ⫽ .83) version of the Center of Epidemiological Studies De- ciations between measures across cultures. Finally, we tested
pression Scale (CES–D; Andresen, Malmgren, Carter, & Patrick, whether item intercepts were equal across cultures (i.e., scalar
1994). Participants rated how frequency they experienced each invariance). As is often the case in cross-cultural studies, our
symptom on a scale from 1 (rarely or none of the time) to 4 (most measures did not demonstrate scalar invariance. This means that
or all of the time). identical observed scores do not necessarily map on to the same
Cross-cultural measurement equivalence. We applied mul- latent scores across cultures. Hence, it is not appropriate to com-
tigroup confirmatory factor analyses (MGCFA) to test the mea- pare means across cultures.
surement equivalence of our measures (e.g., Byrne, Shavelson, & The final metric invariance model for the life satisfaction scale
Muthen, 1989; Vandenberg & Lance, 2000). These analyses are had good fit (confirmatory fit index [CFI] ⫽ .965, root-mean-
important to ensure that individuals from different cultures re- square error of approximation [RMSEA] ⫽ .033, standardized
sponded to our measures in comparable ways. We evaluated the root-mean-square residual [SRMR] ⫽ .048), as did the final metric
models with multiple fit indices (Hu & Bentler, 1999; Marsh, Hau, invariance model for the depressive symptoms scale (CFI ⫽ .967,
& Wen, 2004). First, for each measure, we tested whether items RMSEA ⫽ .021, SRMR ⫽ .065). The metric equivalence of
loaded on the same latent factor across cultures (i.e., configural experienced and desired self-transcending, opening, and conserv-
invariance). Second, we tested whether the loadings of items on ing emotions was established and the relevant fit coefficients are

Table 2
Varimax Rotated Matrices From Principal Axis Factor Analyses of Experienced (Left) and Desired (Right) Emotions

Experienced emotions Desired emotions

Factor Factor
2 1
1 (negative self- 3 4 (negative self- 2 3 4
Item (conserving) enhancing) (self-transcending) (opening) Item enhancing) (conserving) (opening) (self-transcending)

Relaxation .698 Hatred .644 ⫺.205

Calmness .620 Hostility .619
Contentment .604 Anger .502 ⫺.202
Relief .451 Contempt .500
Hatred ⫺.201 .701 Relaxation ⫺.265 .563
Hostility .626 Relief .488
Contempt .528 Calmness ⫺.231 .389
Anger ⫺.261 .494 Contentment ⫺.227 .388
Compassion .725 Curiosity .592
Empathy .565 Interest .267 .532
Love .279 ⫺.266 .437 Enthusiasm .264 .509
Trust .266 ⫺.287 .363 Passion .239 .434
Affection .224 ⫺.230 .361 Excitement ⫺.284 .341
Interest .301 .201 .582 Compassion .780
Curiosity .556 Empathy .239 .401
Enthusiasm .292 .215 .498 Affection .281 .348
Excitement ⫺.249 .244 .415 Love ⫺.263 .235 .341
Passion .275 .292 .389 Trust ⫺.345 .265 .327
1452 TAMIR, SCHWARTZ, OISHI, AND KIM

reported in Tamir et al. (2016). Because we dropped ‘pride’ from be associated with greater well-being and less depressive symp-
our original self-enhancement composite, we reran the measure- toms, even when controlling for experienced and desired emo-
ment equivalence models for desired and experienced self- tions. We further predicted that such patterns would hold across
enhancement, excluding pride. The partial metric invariance model cultural samples. We tested these hypotheses with multilevel
for the negative self-enhancing emotions had good fit for experi- modeling analyses using the Hierarchical Linear Modeling pro-
enced emotions (CFI ⫽ .989, RMSEA ⫽ .049, SRMR ⫽ .030) and gram HLM 7.0 (Raudenbush, Bryk, Cheong, Congdon, & Du
adequate fit for desired emotions (CFI ⫽ .913, RMSEA ⫽ .097, Toit, 2011). For each of our four target emotion categories, we
SRMR ⫽ .069). tested one model that predicted well-being and another that
Human Development Scores. The Human Development In- predicted depressive symptoms. The predictor variable was the
dex (HDI) provides scores that rank countries on their level of absolute discrepancy between desired and experienced emotion,
development. HDI combines indicators of life expectancy, educa- controlling for experienced emotion, desired emotion, gender,
tion, and per capita income. We used HDI scores for 2013 (United and age. The effects of a discrepancy score may be due to either
Nations Development Programme, 2014), the year we gathered our or both of the variables from which it is derived and with which
data. it is correlated. We, therefore, included both experienced and
desired emotions as predictors to control for their effects. Any
Results effect of the absolute discrepancy scores over and above the
effects of the experienced and desired emotions can then be
attributed to the discrepancy between them (see Dyrenforth,
Discrepancies Between Experienced and
Kashy, Donnellan, & Lucas, 2010).5 In our analyses, gender
Desired Emotions
was grand-mean centered and other predictors were group-mean
To compute discrepancy scores, we subtracted experienced centered. By running random-coefficient regression models, we
emotions from desired emotions, separately for each target emo- also tested whether effects of the absolute discrepancy and
tion category. We use difference scores because they provide an control variables varied across cultures. Below is an example of
intuitively clear, direct representation of the conceptual discrep- a Level 1 equation predicting life satisfaction:
ancy between desired and experienced emotions. Below, we ex- SWLSij ⫽ ␤0j ⫹ ␤1jGender ⫹ ␤2jAge ⫹ ␤3jExperienced Emotion
plain how we dealt with the limitations of difference scores (Grif-
fin, Murray, & Gonzalez, 1999; Zuckerman, Gagne, Nafshi, Knee, ⫹ ␤4jDesired Emotion ⫹ ␤5jABS(Desired
& Kieffer, 2002). Figure 1 presents the histograms, means, and ⫺ Experienced Emotion) ⫹ rij
standard deviations for the discrepancies in each emotion category.
On average, people desired more pleasant (i.e., self-transcending, ␤0j is the mean level of life satisfaction across groups. ␤ij are the
opening, and conserving emotions) and less unpleasant (i.e., neg- average regression coefficients of the predictor variables across
ative self-enhancing) emotions than they experienced. However, groups. rij is the individual level variance in life satisfaction
there was substantial variation in discrepancy scores. Many indi- that the predictor variables do not explain.
viduals desired more pleasant emotions than they experienced, but Table 4 presents the results of the HLM analyses predicting life
some desired less. For instance, 11% of our sample wanted to feel satisfaction. As expected, experiencing more frequent positive
less self-transcending emotions, such as love and empathy, than emotions (i.e., self-transcending, opening, and conserving emo-
they actually felt. Many individuals desired less unpleasant emo- tions) and less frequent negative emotions (i.e., negative self-
tions than they experienced, but some desired more. For instance, enhancing emotions) predicted greater life satisfaction (Row 4).
10% of our sample wanted to feel more negative self-enhancing Similarly, desiring more frequent positive emotions and less fre-
emotions, such as anger and hatred, than they actually felt. quent negative emotions also predicted greater life satisfaction
(Row 5). Critically, supporting our hypothesis, smaller absolute
discrepancies between experienced and desired emotions predicted
Happiness and Absolute Discrepancies Between
greater life satisfaction for all four emotion categories (Row 6).
Experienced and Desired Emotions The random effects analyses assessed variation across cultures.
Correlations. We examined correlations of absolute emo- Life satisfaction varied across cultures (Row 7) as did the effects
tional discrepancy scores with well-being and depressive symp- of desired negative self-enhancing and opening emotions (Row
toms within each emotion category.4 We examined the correlations 11). In contrast, effects of experienced emotions (Row 10) and of
in the entire sample and separately for people who desired more desired self-transcending and conservation emotions did not vary
frequent emotions than they experienced and for people who across cultures (Row 11). Critically for our hypothesis, the effects
desires less frequent emotions than they experienced. As shown in of the absolute discrepancies between experienced and desired
Table 3, greater absolute discrepancies between experienced and
desired emotions correlated with lower life satisfaction and more 4
In these analyses, we weighted samples equally to control for unequal
depressive symptoms. This was true for people who felt less sample sizes.
pleasant and more unpleasant emotions than they desired, but, 5
A reviewer noted that discrepancy scores may have different effects
critically for the Aristotelian hypotheses, it was also true for people depending on experienced emotions. To test this possibility requires adding
who felt more pleasant and less unpleasant emotions than they an interaction term between the discrepancy scores and experienced emo-
tions to the multilevel models for each emotion. We conducted these
desired. analyses. They did not affect the conclusions reported below. Part 1 of the
Multilevel modeling. We predicted that smaller absolute online supplementary materials describes the results of these additional
discrepancies between desired and experienced emotions would analyses.
 FEELING GOOD OR FEELING RIGHT 1453

Figure 1. Distributions of the discrepancies between experienced and desired emotions for the four types of
emotions. Means and standard deviations are noted on the upper right of each graph. The dotted line represents
zero discrepancy. Number of people in the sample who were below the dotted line (i.e., experienced more
frequent emotions than they desired) or above it (i.e., desired more frequent emotions than they experienced) are
listed to the left and right of it, respectively. The normal curve is depicted in black on each histogram.
1454 TAMIR, SCHWARTZ, OISHI, AND KIM

Table 3 understand the source of this variation, we examined whether

Zero-Order Correlations Between Indices of Well-Being and socioeconomic development, one of the most important character-
Depressive Symptoms and Absolute Discrepancies of Desired istics on which countries differ, moderated this effect. Specifically,
and Experienced Emotions, Separately for Each Target we tested the possible moderation effect of the country level of
Emotion Category development as indexed by HDI. We did this by running the
multilevel analyses for self-transcending emotions with HDI as a
Emotion and sample Life satisfaction Depressive symptom Level 2 predictor and as a moderator of the effect of the absolute
Self-transcending emotion discrepancy score. The cross-level interaction between
Total ⫺.24 .30 HDI and the absolute discrepancy tested the moderation effect.
Experienced ⬎ Desired ⫺.18 .18 Table 6 presents the results of these analyses for both life satisfac-
Desired ⬎ Experienced ⫺.27 .32 tion (left) and depressive symptoms (right). The interactions between
Negative self-enhancing
Total ⫺.22 .36 HDI and the absolute discrepancy were significant for both (Row 8).
Experienced ⬎ Desired ⫺.22 .38 This indicates that the country level of development moderated the
Desired ⬎ Experienced ⫺.20 .17 effects of the absolute discrepancy between desired and experienced
Opening emotions on life satisfaction and on depressive symptoms. Moreover,
Total ⫺.20 .26
once this moderation effect was included, no significant cross-cultural
Experienced ⬎ Desired ⫺.00a .10
Desired ⬎ Experienced ⫺.23 .27 variation in the effects of the absolute discrepancy remained (Row
Conserving 16). Figure 2 portrays the significant interactions. Figure 2 (left)
Total ⫺.21 .34 portrays the slopes for the interaction predicting life satisfaction and
Experienced ⬎ Desired ⫺.02a .11 Figure 2 (right) portrays the slopes for the interaction predicting
Desired ⬎ Experienced ⫺.24 .36
M (SD) 4.21 (1.16) 1.96 (0.53) depressive symptoms.
Figure 2 (left) shows that life satisfaction dropped more sharply
Note. The table presents correlations in the entire sample (Total) and in more developed than in less developed countries as the discrep-
separately for individuals who experienced the emotion more frequently
than they wanted to experience it (Experienced ⬎ desired) and individuals ancy people felt between the self-transcending emotions (e.g., love
who wanted to experience the emotion less frequently than they actually and empathy) they desired and those they experienced increased.
did (Desired ⬎ Experienced). Figure 2 (right) shows that depressive symptoms rose more sharply
a
p ⬎ .05, for all other entries p ⬍ .05. in more developed than in less developed countries as the discrep-
ancy people felt between the self-transcending emotions (e.g., love
and empathy) they desired and those they experienced increased.
emotions on life satisfaction did not vary by culture, except for
self-transcending emotions (Row 12). Smaller absolute discrepan- Discussion
cies between experienced and desired self-transcending emotions
For Aristotle, happiness entailed feeling the right feelings (Nico-
predicted greater life satisfaction in every country, but this effect
machean Ethics, 1105b25– 6). What feels right, in turn, differs
was not significant in Ghana (slope ⫽ ⫺.15, p ⫽ .34) and China
across people and situations. Happier people, he argued, are those
(slope ⫽ ⫺.42, p ⫽ .071).
who feel what they consider the right feelings, given their unique
Table 5 presents the results of the HLM analyses predicting
circumstances. Our findings support Aristotle’s claims empiri-
depressive symptoms. As expected, experiencing less frequent
cally. We found that happier people are those who more frequently
positive emotions (i.e., self-transcending, opening, and conserving
experience the emotions they want to experience, whether those
emotions) and more frequent negative emotions (i.e., negative
emotions are pleasant (e.g., love or excitement) or unpleasant (e.g.,
self-enhancing) predicted depressive symptoms (Row 4). Simi-
anger and hatred).
larly, desiring more frequent negative emotions and less frequent
What feels right often feels good. However, the two types of
positive emotions (Row 5) predicted depressive symptoms. Sup-
feelings are conceptually distinct and do not necessarily overlap.
porting our hypothesis, larger absolute discrepancies between ex-
Our design enabled us to distinguish empirically between these
perienced and desired emotions predicted depressive symptoms for
two types of feelings and to test whether happiness is linked to
all emotion categories (Row 6).
maximizing ‘right’ feelings even if they are not ‘good’ feelings.
The random effects analyses revealed that depressive symptoms
Consistent with the predictions of the traditional SWB approach,
varied across cultures (Row 7). Effects of experienced emotions (Row
we found that people were happier the more they experienced
10) and desired emotions (Row 11) on depressive symptoms did not
vary across cultures, except for desired opening emotions. Critically
6
for our hypothesis, the effects of the absolute discrepancies between In addition to testing the effects of absolute discrepancy scores, we tested
experienced and desired emotions on depressive symptoms did not the quadratic effect of the signed difference scores on life satisfaction and
vary by culture, except for self-transcending emotions (Row 12). depression in another series of multilevel models. Predictors in these models
included gender, age, experienced emotions, desired emotions, the signed
Larger absolute discrepancies between experienced and desired self- discrepancy scores, and the squared discrepancy scores. All these models
transcending emotions predicted more depressive symptoms in every yielded significant fixed effects for the squared discrepancy scores, as hypoth-
country, but this effect was not significant in Ghana (slope ⫽ .08, p ⫽ esized. Indeed, in 7 of the 8 models the squared discrepancy score was the only
.224) or China (slope ⫽ .02, p ⫽ .856).6 significant predictor. However, these models did not produce reliable estimates
of random effects due to multicollinearity between the experienced or desired
Variation across cultures. As noted, for self-transcending emotions and the signed discrepancy scores. We, therefore, report the findings
emotions, the effects of emotion discrepancies on life satisfaction with absolute differences in the text and provide the fixed effects of the
and on depressive symptoms varied across cultures. In order to quadratic models in Part 2 of the online supplementary materials.
 FEELING GOOD OR FEELING RIGHT 1455

Table 4
Multilevel Models Predicting Life Satisfaction From the Absolute Discrepancy of Desired and Experienced Emotion, Controlling for
Gender, Age, Experienced Emotion, and Desired Emotion

Emotion
Negative self-
Self-transcending enhancing Opening Conserving
Effects Coeff. SE Coeff. SE Coeff. SE Coeff. SE

Fixed effects
Overall mean 4.33ⴱⴱ .11 4.33ⴱⴱⴱ .11 4.33ⴱⴱⴱ .11 4.34ⴱⴱⴱ .11
Gender ⫺.001 .06 .10 .06 .13ⴱ .06 .18ⴱ .07
Age ⫺.05ⴱ .02 ⫺.04 .02 ⫺.04 .02 ⫺.03 .01
Experienced emotion .26ⴱ .08 ⫺.39ⴱⴱ .11 .38ⴱⴱ .09 .55ⴱⴱⴱ .08
Desired emotion .24ⴱ .07 ⫺.11 .10 .08 .10 .12 .07
Absolute discrepancy ⫺.61ⴱⴱ .12 ⫺.26ⴱ .09 ⫺.41ⴱⴱ .08 ⫺.27ⴱⴱ .08

Variance ␹2 Variance ␹2 Variance ␹2 Variance ␹2

Random effects
Culture sample mean .09 132.67ⴱⴱⴱ .09 133.60ⴱⴱⴱ .09 133.77ⴱⴱⴱ .10 145.56ⴱⴱⴱ
Gender slope .006 6.18 .007 5.10 .007 6.52 .02 10.08
Age slope .002 50.82ⴱⴱⴱ .002 41.92ⴱⴱⴱ .002 40.44ⴱⴱⴱ .001 25.56ⴱⴱⴱ
Experienced emotion slope .02 14.25 .054 15.67ⴱ .29 10.68 .02 11.68
Desired emotion slope .01 9.61 .05 15.19ⴱ .05 15.32ⴱ .02 7.55
Absolute discrepancy slope .09 22.89ⴱⴱ .03 7.92 .02 7.72 .02 4.95
% variance explained .15 .12 .13 .17
ⴱ ⴱⴱ ⴱⴱⴱ
p ⬍ .05. p ⬍ .01. p ⬍ .001.

pleasant emotions and the less they experienced unpleasant emo- Implications for Studying Emotions and Well-Being
tions, on average. However, over and above this effect, we also
found that people were happier when they experienced smaller In this investigation, we adopted Aristotle’s ideas regarding the
discrepancies between the emotions they experienced and the potential links between emotions and happiness. This Aristotelian
emotions they desired. The secret to happiness, then, may involve model differs from hedonic models of well-being because it sug-
not only feeling good but also feeling right. gests that both pleasant and unpleasant emotional experiences may

Table 5
Multilevel Models Predicting Depressive Symptoms From the Absolute Discrepancy of Desired and Experienced Emotion, Controlling
for Gender, Age, Experienced Emotion, and Desired Emotion

Emotion
Negative self-
Self-transcending enhancing Opening Conserving
Effects Coeff. SE Coeff. SE Coeff. SE Coeff. SE

Fixed effects
Overall mean 1.95ⴱⴱⴱ .04 1.95ⴱⴱⴱ .04 1.95ⴱⴱⴱ .04 1.95ⴱⴱⴱ .04
Gender .05 .04 .03 .03 .01 .04 ⫺.01 .04
Age .002 .004 ⫺.0008 .003 ⫺.002 .003 ⫺.003 .003
Experienced emotion ⫺.11ⴱ .03 .26ⴱⴱⴱ .04 ⫺.12ⴱ .04 ⫺.26ⴱⴱⴱ .04
Desired emotion ⫺.02 .03 .06 .03 ⫺.07 .04 ⫺.02 .03
Absolute discrepancy .26ⴱⴱ .06 .10ⴱ .03 .23ⴱⴱ .05 .14ⴱⴱ .04

Variance ␹2 Variance ␹2 Variance ␹2 Variance ␹2

Random effects
Culture sample mean .01 91.89ⴱⴱⴱ .01 103.27ⴱⴱⴱ .01 93.99ⴱⴱⴱ .01 102.43ⴱⴱⴱ
Gender slope .01 27.10ⴱⴱⴱ .003 14.22ⴱ .009 23.05ⴱⴱ .006 18.67ⴱⴱ
Age slope .0001 8.65 .00001 8.80 .00001 7.07 .00002 5.17
Experienced emotion slope .002 12.05 .006 14.53ⴱ .003 6.01 .009 12.06
Desired emotion slope .002 10.71 .002 6.54 .008 14.13ⴱ .003 8.44
Absolute discrepancy slope .02 21.12ⴱⴱ .0003 7.32 .009 7.75 .004 8.34
% variance explained .11 .18 .12 .18
ⴱ ⴱⴱ ⴱⴱⴱ
p ⬍ .05. p ⬍ .01. p ⬍ .001.
1456 TAMIR, SCHWARTZ, OISHI, AND KIM

Figure 2. Life satisfaction (left) and depressive symptoms (right) as a function of the absolute discrepancy
between experienced and desired self-transcending emotions (⫾1 SD from the mean) in countries above and
below the mean of Human Development Index (HDI) in our sample.

be linked to well-being if they feel ‘right.’ Emotional experiences monic models (e.g., Ryan & Deci, 2001; Ryff, 1989) assume the
feel right when people perceive them as desirable, given their universal importance of specific life domains or psychological
motives and the situational context. The Aristotelian model also needs for all people and cultures. For instance, Ryff’s (1989)
differs from extant eudaimonic models of well-being. First, eudai- eudaimonic model assumes that eudaimonia consists in function-
ing well in six life domains: autonomy, self-acceptance, environ-
mental mastery, personal growth, purpose in life, and positive
Table 6
relations with others. Similarly, Ryan and Deci’s (2001) well-
Multilevel Models Testing Whether Socioeconomic Development
being model assumes that the satisfaction of basic human needs
(HDI) Moderates the Effects of the Absolute Discrepancy
such as autonomy, competence, and relatedness is key. In contrast,
Between Desired and Experienced Self-Transcending Emotions
our current research concerns feeling “right” emotions rather than
on Life Satisfaction (Left) and Depressive Symptoms (Right)
functioning well in specific life domains or satisfying basic psy-
Life Depressive chological needs. Second, like SWB models (e.g., Diener, 2013),
satisfaction symptoms eudaimonic models focus on positive attributes such as autonomy,
Effects Coeff. SE Coeff. SE successful goal pursuit, and positive relations (e.g., Emmons,
1986; Sheldon & Kasser, 1995; Waterman, 1984). None of these
Fixed effects models posits that well-being may derive from feeling anger or
Overall mean 4.33ⴱⴱⴱ .11 1.95ⴱⴱⴱ .04
Individual level hatred.
Gender .002 .05 .05 .04 The hedonic, eudaimonic, and Aristotelian models of happiness
Age ⫺.05 .02 .0002 .005 are likely to overlap to some extent and to complement each other.
Experienced emotion .26ⴱ .09 ⫺.10 .04 For instance, it is plausible that emotions feel right when basic
Desired emotion .24ⴱ .08 ⫺.03 .04
Absolute discrepancy ⫺.61ⴱⴱⴱ .08 .27ⴱⴱⴱ .04 human needs are satisfied and that satisfying such needs induces
Country level positive affect. In such cases, happiness reflects feeling right,
HDI 1.74 .76 ⫺.23 .33 satisfying basic needs, and maximizing pleasure. However, to
Interaction: HDI ⫻ Absolute attain the best understanding of happiness and its variation across
discrepancy ⫺2.1ⴱⴱ .48 .88ⴱ .29
people, it is important to identify the unique cases in which the
models of happiness make opposing predictions. The current in-
Variance ␹2 Variance ␹2
vestigation identified cases in which people are happier when they
Random effects feel right even if they do not feel good. Of course, such cases are
Culture sample mean .09 113.61ⴱⴱⴱ .01 90.38ⴱⴱⴱ less common, but studying them enabled us to compare the pre-
Gender slope .006 6.25 .01 27.22ⴱⴱⴱ dictions of different models. It demonstrated the unique impor-
Age slope .002 48.51ⴱⴱⴱ .00009 9.52
Experienced emotion slope .03 13.37 .008 12.09
tance of feeling right for happiness, independent of pleasure and
Desired emotion slope .03 9.63 .005 10.73 pain.
Absolute discrepancy slope .02 5.09 .005 6.84 Our findings show that happiness depends, in part, on the match
Residual variance 1.12 .25 between what people feel and what they want to feel. Conse-
Note. HDI ⫽ Human Development Index. quently, understanding the link between emotions and happiness
ⴱ
p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001. requires studying not only experienced emotions but also desired
 FEELING GOOD OR FEELING RIGHT 1457

emotions. Our findings reinforce the evidence for the potential role potential accounts. Testing each account is an important task for
of desired emotions in well-being (e.g., Bastian, Kuppens, De future research.
Roover, & Diener, 2014) and depression (e.g., Millgram, Joor- Second, our findings do not allow us to infer causality.
mann, Huppert, & Tamir, 2015). They highlight the importance of Smaller discrepancies between experienced and desired emo-
understanding what people strive for emotionally. Consistent with tions may lead to greater happiness, but it is also plausible that
the importance of discrepancies between actual and desired states people who experience greater happiness report smaller dis-
(e.g., Higgins, 1987), our findings further highlight the importance crepancies between experienced and desired emotions. Future
of understanding how people compare their experienced emotional research should employ experimental designs to test these pos-
state to the emotions they strive for. sibilities. It would also be beneficial to use longitudinal designs
to examine experienced emotions, desired emotions, and hap-
piness as they evolve and interact over time.
Implications for Studying Emotions and Well-Being
Finally, we investigated a limited set of emotion categories. Our
Across Cultures
selection was based on the idea, which was tested elsewhere
The emotions people desire differ substantially across individ- (Tamir et al., 2016), that certain categories of emotions map on to
uals, cultures, and situations (for a review, see Tamir, 2016). If certain values. Whereas our previous investigation was focused on
well-being is linked to discrepancies between experienced and differences between emotion categories, in this investigation, we
desired emotions, researchers and practitioners must be sensitive to highlight a principle that applies equally across them. Regardless
such differences in desired emotions and understand what under- of whether emotions differed as a function of values, valence or
lies them. Some people may be happier if they experience more other dimensions, smaller discrepancies between desired and ex-
empathy, and some may be happier if they experience more anger, perienced emotion was linked to greater happiness. The emotion
at least to some extent. Such patterns are likely to depend on categories we examined allowed us to compare emotions that
cultural contexts (e.g., de Leersnyder, Kim, & Mesquita, 2015; differ in valence but not arousal (i.e., negative self-enhancing and
Ford et al., 2015). To enhance happiness around the world, re- opening emotions), in arousal but not valence (i.e., opening and
searchers should acknowledge and respect the differences in the conserving emotions), and in other ways (e.g., self-transcending
emotions people desire and understand how they vary across emotions and opening emotions). We did, however, examine a
cultures. limited set of emotions. For instance, we assessed only one cate-
Although we found that people across cultures differ in the gory of unpleasant emotions (i.e., negative self-enhancing emo-
emotions they experience and desire, people were generally hap- tions). Future research could test whether the current effects gen-
pier the more their emotional experiences matched the emotions eralize to other negative emotions such as fear, guilt, sadness and
they desired. This held for negative self-enhancing (e.g., anger, shame, and assess potential cultural differences.
contempt), opening (e.g., interest, excitement), and conserving
(e.g., calmness, relief) emotions. However, for self-transcending Summary
emotions (e.g., empathy, love), the effects of the match between
experienced and desired emotions on happiness varied somewhat What emotions should people strive for to be happy? Consistent
across countries. Specifically, the match between experienced and with Aristotle’s claims, our investigation suggests that people are
desired self-transcendence values enhanced life satisfaction and happier when they experience emotions they desire, whether such
reduced reported symptoms of depression more in highly devel- emotions are pleasant or unpleasant. To the extent that people
oped countries than in less developed countries. Unlike the other desire emotions that are consistent with their values, this suggests
emotions we examined, self-transcending emotions (e.g., love, that happiness entails feeling emotions that are valued, as deter-
trust) are linked to social connectedness. Social connectedness, in mined by the unique personal, social, and cultural context of each
turn, is presumably a basic human need and a key determinant of individual.
well-being (e.g., Myers & Diener, 1995; Ryan & Deci, 2001).
Perhaps, therefore, for people who struggle to meet their basic References
needs, the amount of love they actually feel matters more for their
Andresen, E. M., Malmgren, J. A., Carter, W. B., & Patrick, D. L. (1994).
happiness than whether this amount feels right or not. Future Screening for depression in well older adults: Evaluation of a short form
research should explore further when, why, and how the links of the CES–D (Center for Epidemiologic Studies Depression Scale).
between emotion discrepancies and well-being varies across coun- American Journal of Preventive Medicine, 10, 77– 84.
tries. Bastian, B., Kuppens, P., De Roover, K., & Diener, E. (2014). Is valuing
positive emotion associated with life satisfaction? Emotion, 14, 639 –
645. [Link]
Limitations and Future Directions Byrne, B. M., Shavelson, R. J., & Muthén, B. (1989). Testing for the
Our correlational design has several limitations. First, our find- equivalence of factor covariance and mean structures: The issue of
ings are consistent with two interpretations (see Hardin & Larsen, partial measurement invariance. Psychological Bulletin, 105, 456 – 466.
[Link]
2014; Larsen & McKibban, 2008)—namely, that happiness is
de Leersnyder, J., Kim, H., & Mesquita, B. (2015). Feeling right is feeling
related to greater ideal self-actualization (i.e., experiencing emo- good: Psychological well-being and emotional fit with culture in
tions one desires) or to greater actual self-regard (i.e., desiring autonomy-versus relatedness-promoting situations. Frontiers in Psy-
emotions one experiences). By providing empirical evidence for chology, 6, 630. [Link]
links of happiness to discrepancies between desired and experi- Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95, 542–
enced emotions, our findings highlight the need to consider these 575. [Link]
1458 TAMIR, SCHWARTZ, OISHI, AND KIM

Diener, E. (2013). The remarkable changes in the science of subjective Kitayama, S., Markus, H. R., & Kurokawa, M. (2000). Culture, emotion,
well-being. Perspectives on Psychological Science, 8, 663– 666. http:// and well-being: Good feelings in Japan and the United States. Cognition
[Link]/10.1177/1745691613507583 and Emotion, 14, 93–124. [Link]
Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The Koopmann-Holm, B., & Tsai, J. L. (2014). Focusing on the negative:
satisfaction with life scale. Journal of Personality Assessment, 49, 71– Cultural differences in expressions of sympathy. Journal of Personality
75. [Link] and Social Psychology, 107, 1092–1115. [Link]
Diener, E., Sapyta, J. J., & Suh, E. (1998). Subjective well-being is a0037684
essential to well-being. Psychological Inquiry, 9, 33–37. [Link] Kuppens, P., Realo, A., & Diener, E. (2008). The role of positive and
.org/10.1207/s15327965pli0901_3 negative emotions in life satisfaction judgment across nations. Journal of
Dyrenforth, P. S., Kashy, D. A., Donnellan, M. B., & Lucas, R. E. (2010). Personality and Social Psychology, 95, 66 –75. [Link]
Predicting relationship and life satisfaction from personality in nation- .1037/0022-3514.95.1.66
ally representative samples from three countries: The relative impor- Larsen, J. T., & McKibban, A. R. (2008). Is happiness having what you
tance of actor, partner, and similarity effects. Journal of Personality and want, wanting what you have, or both? Psychological Science, 19,
Social Psychology, 99, 690 –702. [Link] 371–377. [Link]
Eid, M., & Diener, E. (2001). Norms for experiencing emotions in different Lucas, R. E., Diener, E., & Suh, E. (1996). Discriminant validity of
cultures: Inter- and intranational differences. Journal of Personality and well-being measures. Journal of Personality and Social Psychology, 71,
Social Psychology, 81, 869 – 885. [Link] 616 – 628. [Link]
.81.5.869 Marsh, H. W., Hau, K. T., & Wen, Z. (2004). In search of golden rules:
Emmons, R. A. (1986). Personal strivings: An approach to personality and Comment on hypothesis-testing approaches to setting cutoff values for
subjective well-being. Journal of Personality and Social Psychology, 51, fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999).
1058 –1068. [Link] Structural Equation Modeling, 11, 320 –341. [Link]
Ford, B. Q., Dmitrieva, J. O., Heller, D., Chentsova-Dutton, Y., Gross- s15328007sem1103_2
mann, I., Tamir, M., . . . Mauss, I. B. (2015). Culture shapes whether the Mesquita, B., de Leersnyder, J., & Albert, D. (2014). The cultural regula-
pursuit of happiness predicts higher or lower well-being. Journal of tion of emotions. In J. J. Gross (Ed.), Handbook of emotion regulation
Experimental Psychology: General, 144, 1053–1062. [Link] (2nd ed., pp. 284 –301). New York, NY: Guilford Press.
10.1037/xge0000108 Millgram, Y., Joormann, J., Huppert, J. D., & Tamir, M. (2015). Sad as a
Fulmer, C. A., Gelfand, M. J., Kruglanski, A. W., Kim-Prieto, C., Diener, matter of choice? Emotion-regulation goals in depression. Psychological
E., Pierro, A., & Higgins, E. T. (2010). On “feeling right” in cultural Science, 26, 1216 –1228. [Link]
contexts: How person-culture match affects self-esteem and subjective Miyamoto, Y., Ma, X., & Petermann, A. G. (2014). Cultural differences in
well-being. Psychological Science, 21, 1563–1569. [Link] hedonic emotion regulation after a negative event. Emotion, 14, 804 –
.1177/0956797610384742 815. [Link]
Griffin, D., Murray, S., & Gonzalez, R. (1999). Difference score correlations Myers, D. G., & Diener, E. (1995). Who is happy? Psychological Science,
in relationship research: A conceptual primer. Personal Relationships, 6, 6, 10 –19. [Link]
505–518. [Link] Oishi, S., Diener, E. F., Lucas, R. E., & Suh, E. M. (1999). Cross-cultural
Happiness. (n.d.). In Merriam-Webster’s online dictionary. Retrieved from variations in predictors of life satisfaction: Perspectives from needs and
[Link] values. Personality and Social Psychology Bulletin, 25, 980 –990. http://
Hardin, E. E., & Larsen, J. T. (2014). Distinct sources of self- [Link]/10.1177/01461672992511006
discrepancies: Effects of being who you want to be and wanting to be Oishi, S., Diener, E. F., Suh, E. M., & Lucas, R. E. (1999). Value as a
who you are on well-being. Emotion, 14, 214 –226. [Link] moderator in subjective well-being. Journal of Personality, 67, 157–184.
.1037/a0033893 [Link]
Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Oishi, S., Schimmack, U., & Colcombe, S. (2003). The Contextual and
Psychological Review, 94, 319 –340. [Link] systematic nature of life satisfaction judgments. Journal of Experimental
295X.94.3.319 Social Psychology, 39, 232–247. [Link]
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance 1031(03)00016-7
structure analysis: Conventional criteria versus new alternatives. Struc- Oishi, S., Schimmack, U., & Diener, E. (2001). Pleasures and subjective
tural Equation Modeling, 6, 1–55. [Link] well-being. European Journal of Personality, 15, 153–167. [Link]
519909540118 .org/10.1002/per.406
Inglehart, R., & Baker, W. E. (2000). Modernization, cultural Change, and Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., Congdon, R. T., & Du
the persistence of traditional values. American Sociological Review, 65, Toit, M. (2011). HLM 7. Lincolnwood, IL: Scientific Software Interna-
19 –51. [Link] tional Inc.
Kahneman, D. (1999). Objective happiness. In D. Kahneman, E. Diener, & Rice, R. W., McFarlin, D. B., & Bennett, D. E. (1989). Standards of
N. Schwarz (Eds.), Well-being: Foundations of hedonic psychology (pp. comparison and job satisfaction. Journal of Applied Psychology, 74,
3–25). New York, NY: Sage. 591–598. [Link]
Kahneman, D., Diener, E., & Schwarz, N. (1999). Well-being. The Foun- Rusting, C. L., & Larsen, R. J. (1995). Moods as sources of stimulation:
dations of Hedonic Psychology. New York, NY: Sage. Relationships between personality and desired mood states. Personality
Kämpfe, N., & Mitte, K. (2009). What you wish is what you get? The and Individual Differences, 18, 321–329. [Link]
meaning of individual variability in desired affect and affective discrep- 0191-8869(94)00157-N
ancy. Journal of Research in Personality, 43, 409 – 418. [Link] Ryan, R. M., & Deci, E. L. (2001). On happiness and human potentials: A
.org/10.1016/[Link].2009.01.007 review of research on hedonic and eudaimonic well-being. Annual
Kitayama, S., Ishii, K., Imada, T., Takemura, K., & Ramaswamy, J. (2006). Review of Psychology, 52, 141–166. [Link]
Voluntary settlement and the spirit of independence: Evidence from .psych.52.1.141
Japan’s “Northern frontier”. Journal of Personality and Social Psychol- Ryff, C. D. (1989). Happiness is everything, or is it? Explorations on the
ogy, 91, 369 –384. [Link] meaning of psychological well-being. Journal of Personality and Social
 FEELING GOOD OR FEELING RIGHT 1459

Psychology, 57, 1069 –1081. [Link] based account. Journal of Personality and Social Psychology, 111,
.1069 67– 82. Advance online publication. [Link]
Schwartz, S. H. (2006). A theory of cultural value orientations: Explication pspp0000072
and applications. Comparative Sociology, 5, 137–182. [Link] Tay, L., & Diener, E. (2011). Needs and subjective well-being around the
10.1163/156913306778667357 world. Journal of Personality and Social Psychology, 101, 354 –365.
Schwartz, S. H. (2011). Values: Individual and cultural. In S. M. Breugel- [Link]
mans, A. Chasiotis, & F. J. R. van de Vijver (Eds.), Fundamental Thomson, J. A. K. (1955). The ethics of Aristotle—The Nicomachean
questions in cross-cultural psychology (pp. 463– 493). New York, NY: ethics. Baltimore, MD: Penguin.
Cambridge University Press. [Link] Tsai, J. L., Knutson, B., & Fung, H. H. (2006). Cultural variation in affect
11974090.019 valuation. Journal of Personality and Social Psychology, 90, 288 –307.
Schwartz, S. H., Cieciuch, J., Vecchione, M., Davidov, E., Fischer, R., [Link]
Beierlein, C., . . . Konty, M. (2012). Refining the theory of basic United Nations Development Programme. (2014). Human development
individual values. Journal of Personality and Social Psychology, 103, report. Retrieved from [Link]
663– 688. [Link] Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the
Sheldon, K. M., & Kasser, T. (1995). Coherence and congruence: Two measurement invariance literature: Suggestions, practices, and recom-
aspects of personality integration. Journal of Personality and Social mendations for organizational research. Organizational Research Meth-
Psychology, 68, 531–543. [Link] ods, 2, 4 –70.
Sims, T., Tsai, J. L., Jiang, D., Wang, Y., Fung, H. H., & Zhang, X. (2015). Waterman, A. S. (1984). The psychology of individualism. New York, NY:
Wanting to maximize the positive and minimize the negative: Implica- Praeger.
tions for mixed affective experience in American and Chinese contexts. Zuckerman, M., Gagné, M., Nafshi, I., Knee, C. R., & Kieffer, S. C.
Journal of Personality and Social Psychology, 109, 292–315. [Link] (2002). Testing discrepancy effects: A critique, a suggestion, and an
.[Link]/10.1037/a0039276 illustration. Behavior Research Methods, Instruments, & Computers, 34,
Tamir, M. (2016). Why do people regulate their emotions? A taxonomy of 291–303. [Link]
motives in emotion regulation. Personality and Social Psychology Re-
view, 20, 199 –222. [Link] Received May 2, 2016
Tamir, M., Schwartz, S. H., Cieciuch, J., Riediger, M., Torres, C., Scollon, Revision received February 17, 2017
C., . . . Vishkin, A. (2016). Desired emotions across cultures: A value- Accepted February 20, 2017 䡲

E-Mail Notification of Your Latest Issue Online!

Would you like to know when the next issue of your favorite APA journal will be available
online? This service is now available to you. Sign up at [Link] and you will
be notified by e-mail when issues of interest to you become available!