INTERPERSONAL RELATIONS AND GROUP PROCESSES
The Paradoxical Consequences of Revenge
Kevin M. Carlsmith
Timothy D. Wilson
University of Virginia
Daniel T. Gilbert
People expect to reap hedonic rewards when they punish an offender, but in at least some instances,
revenge has hedonic consequences that are precisely the opposite of what people expect. Three studies
showed that (a) one reason for this is that people who punish continue to ruminate about the offender,
whereas those who do not punish “move on” and think less about the offender, and (b) people fail to
appreciate the different affective consequences of witnessing and instigating punishment.
Keywords: affective forecasting, punishment, revenge, rumination
On April 16, 2007, Seung Hui Cho massacred 32 people at
Virginia Tech University before turning the gun on himself. In the
aftermath, some lamented that his death had robbed the survivors
of the emotional satisfaction of exacting their own revenge. As one
woman from the Midwest wrote on an Internet blog, “I don’t think
there would be anything temporary about the satisfaction I would
feel in being permitted to execute the person who killed my child”
“Sweet is revenge,” wrote Lord Byron (Don Juan, Canto I,
Stanza 124). Indeed, this belief may be both venerable and wide-
spread, but is it right? Research in psychology and behavioral
economics has shown that people often mispredict their hedonic
reactions to future events because they mispredict how often they
will think about those events once the events have ended (Schkade
& Kahneman, 1998; Wilson & Gilbert, 2003; Wilson, Wheatley,
Meyers, Gilbert, & Axsom, 2000). We hypothesized that acts of
revenge would cause people to continue thinking about the trans-
gressor whom they have punished, which would prolong their
hedonic reactions to the transgression rather than shorten them.
Thus we predict that people punish others, in part, to repair their
negative mood and to provide psychological closure to the precip-
itating event, but that the act of punishment yields precisely the
opposite outcome. As the poet John Milton wrote, “Revenge, at
first though sweet, Bitter ere long back on itself recoils” (Paradise
Lost, Book IX, Line 171).
Our research, then, focuses on three questions: (a) What is the
affective consequence of exacting revenge? (b) Do people accu-
rately predict this consequence, and, if not, (c) Why do people
make this error of prediction?
Although there is an enormous amount of literature on punish-
ment, it has generally focused on the justification for imposing
punishment (Bentham, 1843/1962; Ezorsky, 1972; Kant, 1790/
1952) or on the consequences of receiving punishment (Skinner,
1938, 1974; Watson, 1924). More recently, psychologists have
begun to explore the motives behind the desire to punish (see
Carlsmith & Darley, 2008, for a review), but there has been little
empirical research performed on the question of whether instigat-
ing punishment against another person has affective consequences
for the one who punishes. This absence is both surprising and
unfortunate given the ubiquity of punishment across societies
(Gintis, 2008) and the self-evident importance of this punishment
to both instigator and target. Although this discussion is potentially
relevant to the myriad forms of punishment (e.g., societal, legal,
organizational, etc.), we focus specifically on interpersonal aggres-
sion in response to individual norm-violating behaviors. This sort
of aggression is typically referred to as revenge or retaliation by
laypeople (McKean, 2005), psychologists (Frijda, 1994), and phi-
losophers (Baier, 1955; Hobbes, 1950; Quinton, 1954).
People believe in the utility of aggression. There is widespread
acceptance of the notion that aggression, either directed against the
target of one’s ire or displaced to some other object, relieves the
tension, and thus the anger, that had been pent up inside. Evidence
for this belief abounds in our language (“get it out of your sys-
Kevin M. Carlsmith, Department of Psychology, Colgate University;
Timothy D. Wilson, Department of Psychology, University of Virginia;
Daniel T. Gilbert, Department of Psychology, Harvard University.
We acknowledge the assistance provided by James Elmers, Alison
Potocki, Emily Drummond, Sue LeBarron, and Rachel Smith in conducting
Correspondence concerning this article should be addressed to Kevin M.
Carlsmith, Department of Psychology, Colgate University, 13 Oak Drive,
Hamilton, NY 13346. E-mail: firstname.lastname@example.org
Journal of Personality and Social Psychology, 2008, Vol. 95, No. 6, 1316–1324
Copyright 2008 by the American Psychological Association 0022-3514/08/$12.00 DOI: 10.1037/a0012165
tem”), in our public service advertisements (“punch a wall, not
your child”), in plays (The Spanish Tragedy,Hamlet), novels (The
Count of Monte Cristo, Moby Dick, The Cask of Amontillado), and
films (Death Wish I–V, Revenge, Kill Bill). Bushman (2002) cited
numerous self-help books (e.g., Lee, 1993) that extol the virtues of
the cathartic release that comes from punching pillows or other
objects that stand in for the true target.
This belief in catharsis is not surprising, perhaps, given psy-
chology’s own deep roots in the theory. Early psychological re-
search on aggression adopted the hydraulic model, in which ag-
gression was perceived to be a cathartic release to the pressure
brought about by frustration or another stimulus (Berkowitz, 1989;
Dollard, Doob, Miller, Mowrer, & Sears, 1939; Freud, 1922/1961).
In this model, the eliciting stimulus creates an internal pressure
that must be released, and the most likely form of release is
aggression. Once cathartic aggression takes place, subsequent ag-
gression is less likely. Thus, venting one’s aggression on an
appropriate object (e.g., a pillow) was seen as an effective way to
reduce interpersonal aggression and spirals of conflict.
Although the early research on catharsis was mixed, more recent
investigations have come to a clear consensus against the theory.
Bushman (2002) demonstrated that cathartic aggression not only
failed to reduce subsequent aggression, but in fact increased it.
Participants in his experiment were initially angered by a fellow
student who provided insulting feedback on an essay and were
then given an opportunity to vent their anger on a punching bag.
Shortly thereafter, participants played a competitive game with the
object of their anger and were given the opportunity to deliver
aversive noise blasts. Bushman found that use of the punching bag
actually increased subsequent aggression rather than reducing it.
Moreover, he found that participants who imagined that the punch-
ing bag was the other student (a rumination manipulation) ag-
gressed even more than did those participants in the control con-
dition and those focused on the health benefits of the punching bag
(a distraction manipulation). Critical for our purposes, participants
who vented their anger and subsequently aggressed against the
other student also experienced more anger at the end of the
Thus, Bushman (2002) found that physical venting did not
reduce the desire to aggress and that the combination of venting
and aggressing led to increased anger. The focus of Bushman’s
experiment, though, was on the effects of cathartic aggression, and
thus the aggression against the offending student (e.g., the noise
blasts) served as a dependent measure rather than as an indepen-
dent variable. It is thus unclear whether it was the venting or the
aggression that increased the anger and whether the aggression
caused the anger or whether the anger caused the aggression.
Cathartic Aggression as Mood Repair
Other research has shown that participants aggress against of-
fending others in the belief that it will improve their mood (Bush-
man, Baumeister, & Phillips, 2001). Participants were angered and
then given an opportunity to aggress against the offender, and
those who believed in the cathartic value of aggression— either
through preexisting attitudes or through manipulation—aggressed
more severely than did those who did not believe in catharsis. This
effect disappeared, however, when students believed they had
ingested a “mood-freezing” pill. Thus, participants were strategi-
cally using aggression to repair their own mood.
Bushman et al. (2001) primarily examined people’s beliefs
about the hedonic utility of aggression. In one study, however, they
examined the actual impact of aggression on subsequent mood.
Those who punished most severely (because they believed in the
utility of punishment) in fact experienced increased negative af-
fect. Once again, however, the aggression variable was measured
rather than manipulated. The authors concluded that these results
“may be relevant to the question of whether aggression actually
does accomplish mood repair, although our study was not designed
to provide definitive evidence on that question” (Bushman et al.,
2001, p. 26).
In sum, research shows that cathartic venting does not reduce
subsequent aggression against the offending object. However, peo-
ple have a clear belief in the value of catharsis and regulate
aggression in service to mood. But it is unclear whether aggression
improves mood or if, indeed, it actually makes the aggressor feel
worse. In order to examine this question, we used punishment as
an independent variable (people did or did not have the opportu-
nity to punish an offender), with the hypothesis that people would
take advantage of the opportunity to punish but that doing so
would prolong rumination and negative affect.
The Accuracy of Affective Forecasting
There is ample evidence that people often make mistakes when
forecasting their future affective reactions (Loewenstein,
O’Donoghue, & Rabin, 2003; Mellers & McGraw, 2001; Wilson
& Gilbert, 2003), and the mistaken beliefs about the utility of
aggression described above may well be a case in point.
The most commonly observed error in affective forecasting is
the impact bias, whereby people overestimate the intensity and
duration of their future affective reactions. People have been found
to overestimate the hedonic impact of a wide range of events, from
romantic breakups and political elections to the loss of a job or
receipt of unwanted results of a pregnancy test (for a review, see
Wilson & Gilbert, 2003). One cause of the impact bias is focalism,
the tendency to overestimate how much an event that one is
thinking about will occupy one’s thoughts as time goes by. Wilson
et al. (2000), for example, found that fans overestimated how
happy or sad they would be after their favorite team won or lost a
football game, in part because they overestimated how much they
would think about the game after it occurred (see also Schkade &
Kahneman, 1998). Put differently, people underestimated the ex-
tent to which the everyday routine of their lives would occupy their
thoughts, crowding out thoughts about the game.
Thoughts about some events, however, get crowded out more
easily than others. There is evidence that anger, in particular, is an
attention-focusing emotion, making it difficult to think about other
things (Fredrickson & Branigan, 2005). Anger thoughts can thus
be a vicious cycle; the more people think about them the angrier
they get, and the angrier they get, the harder it is to think about
anything else (Bushman, 2002; Konecni, 1974; Rusting & Nolen-
Hoeksema, 1998). Anger may thus be an exception to previous
research on affective forecasting and focalism. Instead of overes-
timating the intensity and duration of their anger following pun-
ishment, people may actually underestimate it. People’s lay theory
seems to be that revenge is a cathartic act that will bring them
CONSEQUENCES OF REVENGE
closure, allowing them to stop thinking about the precipitating
event, when in fact that course of action might keep their attention
on the event and prevent them from coping in other ways.
Suppose, for example, that people did not have the opportunity
to punish the object of their ire. They would be forced to deal with
their negative feelings in other ways, perhaps by deciding that in
the grand scheme of things, the person’s actions were not very
important and not worth worrying about. By trivializing the be-
havior, people could move on and not think about it further
(Simon, Greenberg, & Brehm, 1995). Ironically, the act of pun-
ishing might increase ruminations about the incident and prevent
people from dealing with their negative emotions in other ways,
such as thinking about unrelated issues, which has been shown to
reduce anger (e.g., Bushman, 2002).
We hypothesized that people believe that punishing an offender
will improve their mood and bring about psychological closure,
but in fact punishment will increase rumination about the offender
and lead to a continuation of negative affect. Indeed, there might
be a recursive relationship between affect and rumination such that
increased negative affect also leads to increased rumination. To
test these hypotheses, we used the “free rider paradigm” in which
participants played multiple trials of an interactive game with three
other people. Players could earn money if they cooperated, but a
player who did not cooperate (a “free rider”) could earn more and
cause others to earn less. Unbeknownst to the participants, the
game was staged so that on several trials one of the players
encouraged others to cooperate and then did not cooperate himself
or herself. Punishers were given the opportunity to financially
penalize this free rider at a small cost, whereas nonpunishers
played the same game but were given no opportunity to penalize
the free rider. Both groups reported their hedonic reactions, and
punishers also estimated their hedonic reactions had they had no
opportunity to penalize the free rider. An additional group of
forecasters played the game and predicted what their hedonic
reactions would be if they penalized the free rider. There were thus
three conditions: punishers, nonpunishers, and forecasters. The
latter group made prospective predictions of affect, but we also
asked the punishers to make retrospective predictions of affect had
they been in a different condition.
Participants and procedure. Forty-eight university students
were allotted a stake of $1.00 and told that they could invest in the
group or save it. After each round of play the total invested amount
from all 4 players—plus a 40% dividend—was distributed to all
players equally regardless of individual contribution. This arrange-
ment created a classic prisoner’s dilemma: The optimal group
outcome was achieved when each individual cooperated, but each
individual maximized his or her personal outcome by defecting.
Participants generally convened in groups of 4. When fewer
than 4 participants were available, participants were led to believe
that a 4th individual showed up after the original members entered
their cubicles. Participants played a few verbal practice rounds
with the experimenters to ensure they understood the instructions.
Once the game began, all of the interactions occurred in private
rooms via individual computer terminals.
Computer software simulated the play of the other participants.
Two were programmed to cooperate initially and to use a tit-for-tat
strategy on subsequent rounds. The 3rd player, always a female,
initially bid $0.50 and sent everyone an instant message urging
them to cooperate. She then defected on all subsequent rounds.
After each of eight investment decisions, participants received
feedback about each group member’s investment, a record of each
person’s total earnings, and a prompt to invest for the following
round. The program was convincing; few participants doubted that
the other players were fellow students.
Participants were randomly assigned to a punishment, no-
punishment, or forecaster condition. At the end of the game, partici-
pants in the punishment condition were told that, if they wanted, they
could anonymously penalize the other players by subtracting points
from these players’ total earnings. Each point cost the punisher $.05
but reduced the punished member’s profit by three times that amount.
After indicating whether they wanted to punish any of the other
players and by how much, participants in the punishment condition
completed the dependent measures. Participants in the no-punishment
condition completed the dependent measures right after the conclu-
sion of the game with no mention of the opportunity to penalize other
players. Participants in the forecaster condition received the same
instructions participants in the punishment condition received about
penalizing the other players, except that instead of actually being
given the opportunity to punish, they were asked to imagine that they
could do so. They then completed the dependent measures according
to how they thought they would feel if they had had the opportunity
Dependent measures. Participants in the punishment and no-
punishment conditions rated their hedonic reactions at the conclu-
sion of the game and 10 min later after completing a filler task
(writing a detailed description of their typical day). At each time
point they rated the extent to which they were experiencing pos-
itive affect (pleased, positive, satisfied) and negative affect (neg-
ative, vengeful, irritated) on 7-point scales anchored by not at all
and extremely. Participants also reported the extent to which they
were thinking about each of the other players, again on 7-point
scales (1 ⫽not at all,7⫽very much). Participants in the
punishment condition were also asked immediately postpunish-
ment to estimate what they thought their hedonic reactions would
be (on the same scales) if they had not had the opportunity to
penalize the free rider. After the game all participants were probed
for suspicion, debriefed, and paid slightly more than they expected
to earn in the game.
Preliminary analyses. Three participants failed to play coop-
eratively and therefore earned more money than each of the other
players. Thus, each of these participants was the free rider in the
group and had an entirely different psychological experience than
did the other participants. We excluded these participants (2 in the
punishment condition, 1 in the no-punishment condition) from all
analyses reported below, leaving a total of 15 participants in each
condition. Participants earned an average of $2.51 (excluding the
cost of punishing), whereas the free rider earned $5.59, t(44) ⫽
10.41, p⬍.01. Fourteen of 15 people in the punishment condition
chose to spend their money to punish the free rider, with an
average outlay of $0.51. This resulted in an average punishment of
1318 CARLSMITH, WILSON, AND GILBERT
$1.53. No participant penalized either of the other (cooperative)
players in the game.
Data reduction. Initial analyses revealed that the positive and
negative affect items were highly correlated and that the manipula-
tions had similar effects on each type of item. We thus reverse scored
the negative items and averaged them with the positive items. Initial
analyses also revealed that although there was a main effect of time on
this index, F(1, 41) ⫽28.79, p⬍.001, reflecting the fact that
participants’ mood improved during the 10-min distraction test, time
did not interact with the experimental manipulations, F(2, 42) ⫽0.12,
p⫽.89. In order to simplify the analyses, we thus averaged the affect
ratings over time, resulting in a highly reliable index (␣⫽.93). We
also report the results of individual affect items.
Hedonic reactions. Did punishing the free rider change peo-
ple’s affect? The answer is yes: People in the punishment condi-
tion actually felt worse than people in the no-punishment condi-
tion. Did people anticipate this effect? The answer is no; those in
the forecaster condition, who rated how they would feel if they
could punish, predicted that they would feel better than people who
actually were allowed to punish. The main effect of condition on
affect was significant, F(2, 42) ⫽6.83, p⬍.01, and planned
contrasts revealed that people in the punishment condition (M⫽
3.52, SD ⫽.95) reported a significantly worse mood than did
people in both the no-punishment (M⫽4.79, SD ⫽1.22) and
forecaster (M⫽4.53, SD ⫽.77) conditions, ps⬍.01. The
no-punishment and forecaster conditions were not significantly
different from each other ( p⫽.47).
A detailed analysis of each affect item confirms the findings
described above. Each negatively valenced item was higher for the
punishers and lower for the nonpunishers. Similarly, each posi-
tively valenced item was lower for the punishers and higher for the
nonpunishers. Each of these differences was significant by
Tukey’s least significant difference (LSD) at p⫽.05. The largest
mean differences between punishers and nonpunishers were for the
items vengeful (2.07), irritated (1.67), negative (1.27), and positive
(⫺1.20). Thus it is primarily (but not exclusively) an increase in
negative affect that results from administering punishment.
Mean differences in rumination. There were no differences
between the punishment and no-punishment conditions at Time 1
in how much people reported thinking about the free rider (M⫽
6.07 vs. 5.73, F⬍1.0). However, people in the punishment
condition reported thinking about the free rider significantly more
at Time 2 than did people in the no-punishment condition (M⫽
4.33 vs. 2.67), F(1, 28) ⫽4.46, p⫽.04. The Time ⫻Condition
(punishers vs. nonpunishers) interaction was significant, F(1,
28) ⫽4.25, p⬍.05, indicating that although all participants
initially focused attention on the free rider, the act of punishment
kept the free rider at the forefront of the participants’ thoughts.
Nonpunishers, by contrast, were able to focus attention elsewhere.
Rumination as mediator between punishment and affect. Fig-
ure 1 reveals that rumination about the free rider at Time 2 was
related to both punishment (b⫽.43) and affect (b⫽⫺.42) as
required for a mediated effect, and that the direct effect of pun-
ishment on affect was reduced when rumination was included in
the equation. The beta coefficient dropped from b⫽⫺.49 to ⫺.31,
which was marginally significant by the Sobel test( z⫽1.77, p⫽
.08). The same analysis does not hold for rumination at Time 1,
suggesting that it is the continued rumination that leads to negative
affect rather than initial rumination.
Affect as mediator between punishment and rumination. We
anticipated, and found, a recursive relationship between affect and
rumination such that each variable served as a partial mediator for
the effect of punishment on the other. Thus, affect at Time 2 was
related to both punishment (b⫽⫺.49) and rumination (b⫽⫺.40),
and the direct effect of punishment on rumination was reduced
when affect was included in the equation. The beta coefficient
dropped from b⫽.43 to .22, which was also marginally significant
by the Sobel test (z⫽1.77, p⫽.08).
Affective forecasts. As noted above, forecasters predicted
that if they could punish the free rider they would feel signif-
icantly better than punishers reported feeling. That is, they
failed to anticipate the negative hedonic effects of punishment.
This was also true at the individual level: The correlation
between forecasters’ predicted affect and predicted level of
punishment was r(15) ⫽.48, p⫽.07. This is in stark contrast
to the negative correlation in the actual punishment condition,
r(15) ⫽⫺.50, p⫽.06. The difference between these correla-
tions is significant (z⫽2.63, p⬍.01), showing that actual and
predicted experience ran in opposite directions. People ex-
pected that higher punishment would lead to a better mood, but
in fact higher punishment was associated with a worse mood.
However, forecasters did not predict that they would feel better
than people in the no-punishment condition reported feeling,
which on the face of it is inconsistent with our hypothesis that
people believe punishment improves their affective states. In the
absence of a no-punishment–forecaster condition, however, in
which people predicted how they would feel without the opportu-
nity to punish, this finding is difficult to interpret. That is, our
forecasters were predicting how they would feel in a situation that
the nonpunishers did not experience, making the comparison prob-
lematic (we return to this issue in Study 2). More informative are
punishers’ beliefs about how they would feel if they had not been
given the opportunity to punish. When asked, all participants in the
punishment condition indicated that they would have been less
happy than they actually were had they not been allowed to punish,
⫽4.00 vs. M
⫽5.02, t(14) ⫽⫺5.35, p⬍.001.
Rumination forecasts. Finally, forecasters also failed to per-
ceive the relationship between punishment and rumination. Al-
though the actual punishers revealed a strong positive correlation,
r(15) ⫽.40, between time spent ruminating about the free rider
and the severity of the punishment, forecasters thought the rela-
tionship ran in the other direction, r(15) ⫽⫺.19. A test of the
difference of these two correlations yielded z⫽1.48, p⫽.07.
Thus, people thought that punishing the free rider made them feel
better, and perhaps even allowed them to “move on” and forget
about the free rider. In reality, the opposite occurred: They con-
* p< .05 # p< .10
Figure 1. Study 1: The negative relationship between punishment of the
perpetrator and personal affect is mediated by rumination about the per-
CONSEQUENCES OF REVENGE
tinued to think about the free rider and continued to experience
We created a real situation in which people were angry at a free
rider and then manipulated whether or not they could retaliate
against the free rider by means of financial penalties. The result of
the punishment was that both the participants and the free rider lost
money, and the punishers were less happy than were those who did
not have the opportunity to punish. Ten minutes after the game was
over, punishers were thinking about the free rider significantly
more, and this increased rumination partially mediated the persis-
tent negative affect. Although the actual affective consequences of
punishment were clear, people in the punishment condition did not
seem to understand these consequences. In fact, each and every
participant in the punishment condition reported that their affect
would have been lower had they not been allowed to punish. This
is a remarkable finding, given that these participants were at the
time experiencing more negative affect as a result of the punish-
ment. Further, forecasters failed to predict the negative affective
consequences of revenge, in part because (a) the act of punishment
led people to ruminate about the perpetrator more, (b) this rumi-
nation prolonged the negative affect, (c) the negative affect pro-
longed rumination, and (d) people failed to predict the effects of
punishment on rumination and affect.
Another reason that people mispredicted the effects of punish-
ment may be that they thought only about the pleasure they
anticipated from the free rider’s imminent misfortune and failed to
anticipate the negative affect that would accrue as a result of
instigating the punishment. That is, the negative impact of pun-
ishment may stem from the fact that participants had, so to speak,
to thrust the knife themselves, which may have violated their sense
that they were good people who do not harm others. Further,
people had to spend some of their own money in order to punish
the free rider. Perhaps people could avoid the unpleasant effects of
revenge if they were not the agent responsible for it. If they learned
that someone else punished the free rider, they would see her get
her comeuppance while avoiding the psychological and financial
costs of delivering the punishment themselves.
In Study 2 we tested the hypotheses that people fail to appreciate
the distinction between instigating and witnessing a negative event
befalling the free rider, namely, that people would expect to feel
better after the free rider received a financial punishment but that
they would be indifferent to the source of that punishment. An-
other purpose of Study 2 was to compare directly people’s fore-
casts about how they would feel if they could versus if they could
not exact revenge. As mentioned earlier, in Study 1 it was difficult
to compare people’s forecasts about how they would feel if they
could punish with the experiences of people who could not punish
in the absence of a condition in which people forecasted how they
would feel if they could not punish. In Study 2 we asked people to
provide both forecasts to better test the prediction that people
believe punishing would improve their affective states.
Participants. Ninety-eight undergraduates participated in par-
tial fulfillment of a course requirement. Seventy-three of the par-
ticipants were from the same university the participants were from
in Study 1; 25 were from a different, though academically com-
parable, university. Seventy-seven participants were women, and
21 were men.
Materials and procedure. Participants read a detailed descrip-
tion of the basic paradigm used in Study 1 and were asked to imagine
that they were one of the participants. They were told that they and the
2 other cooperative players had invested $3.87 and received $1.85
profit at the end of the game. The free rider, however, had only
invested $1.00 yet received $3.43 profit at the end. (Earnings and
investments were based on the actual average results from a pilot
version of the free rider game.) Participants then forecasted how they
would feel at that point in the game on the same affect measures used
in the previous study. These measures are the equivalent of the
no-punish control condition of Study 1. Participants then learned
about the punishment options and were asked to report how they
would feel if they had decided to punish the free rider, and also how
they would feel if one of the other players had punished the free rider
(yielding a within-subjects design). In one condition they were told,
“In the game that you just finished, you decide to spend $0.25 on
penalties, a move that will reduce the winnings of [the free rider] by
$0.75.” In the other condition they were told that “one of the other
players” spent the $0.25. Thus there were three within-subject con-
ditions: a no-punish control condition; a punisher condition, in which
the participant imagines instigating the punishment; and a witness
condition, in which the participant imagines witnessing another player
punishing the free rider. The order of presentation of the punisher and
witness conditions was counterbalanced across participants.
In addition to affect measures, participants reported the extent to
which they would be thinking about the free rider (i.e., the rumi-
nation question from Study 1) after each set of affect measures. At
the conclusion of the survey, participants completed manipulation
checks regarding details of the game. Participants completed this
survey in one of two formats: as a stand-alone web survey or as a
paper-and-pencil survey appended to the end of an existing survey.
Preliminary analyses. There were no main effects or interac-
tions involving the order of question presentation, the format of the
survey, or university affiliation and so these variables are not dis-
cussed further. As before, the positive and negative affect items were
highly correlated and produced reliable affect scales. We thus col-
lapsed across valence for each of the three conditions and created
unitary affect scales as in Study 1 (␣⫽.78, .81, and .77, respectively).
Predictions of affect. Overall, people expected that they would
feel quite negative after playing the game in which the free rider
reaped profits at their expense. The mean affect score in the control
condition (prior to any mention of punishment) was 2.36 (on a
scale of 1 to 7) with a standard deviation of .84. However, there
was also a strong belief that the opportunity to punish would have
salubrious effects on mood, F(2, 190) ⫽98.71, p⬍.001. Partic-
ipants predicted a mean affect of 4.33 (SD ⫽1.05) if they insti-
gated the punishment and a mean affect of 4.76 (SD ⫽.99) if they
witnessed someone else instigating the punishment. The no-
punishment condition was significantly different from the two
punishment conditions ( ps⬍.001), and the two punishment
conditions did not differ from each other ( p⫽.16).
1320 CARLSMITH, WILSON, AND GILBERT
A detailed analysis of each affect item confirms the findings
described above. Each negatively valenced item was predicted to
be higher in the no-punishment condition and lower in the pun-
ishment conditions. Similarly, each positively valenced item was
predicted to be lower in the no-punishment condition and higher in
the punishment conditions. Each of these differences was signifi-
cant by Tukey’s LSD at p⫽.05. The largest mean differences
between the punish and no-punish conditions were for the items
pleased (⫺2.80), irritated (2.49), and negative (2.33).
Predictions of rumination. In Study 1 we found that punishers
ruminated more about the free rider and that this increased rumi-
nation partially explained subsequent negative affect. Participants
in the present study, like the forecasters in Study 1, did not predict
that punishers would ruminate more; in fact, they predicted the
opposite. In the control condition, their agreement with the state-
ment “I am thinking about” the free rider was 4.57 on a 7-point
scale. When they imagined instigating the punishment themselves,
that number dropped to 4.29, and when they imagined another
player punishing the free rider, the number dropped to 3.53. All
differences were significant by Tukey’s LSD ( p⬍.001).
As hypothesized, participants predicted that they would feel
much better after being given the opportunity to punish than
before. This clarifies the results of Study 1, in which we did not
include both a no-punish and punish forecaster condition. Second,
participants believed that punishment would reduce rumination
rather than increase it, replicating Study 1. Third, participants
failed to differentiate between witnessing and instigating the pun-
ishment. In Study 3 we tested the hypothesis that this distinction is
critical for the affect people actually experience, namely, that
people feel worse when they deliver the punishment themselves.
That is, Study 3 tested the hypothesis that participants’ forecasts in
Study 2 were wrong, in the sense that having a hand in the
punishment (instigating it) has more negative hedonic conse-
quences than witnessing someone else doing the job.
We performed a replication of Study 1 with the addition of
conditions in which the punishment (or anticipated punishment)
was carried out by one of the other players and the participant was
merely a witness to the punishment. This design permitted a direct
replication of Study 1, a replication of Study 2 using actual game
players rather than vignette-based predictions, and a test of
whether the affective forecasting errors found previously were
attributable, in part, to a failure to recognize the affective distinc-
tion between instigating and witnessing punishment.
Participants. Seventy-five university students from the same
population as in Study 1 participated in the experiment for partial
course credit. There were 45 women and 30 men.
Procedure. The procedure was identical to that in Study 1 with
the following exceptions. In the punishment conditions, partici-
pants were informed that 1 player had been randomly selected to
be allowed to financially penalize any or all of the other players.
In the punisher condition, the participant learned that he or she had
been the one selected. This condition mimics the punishment
condition of Study 1. In the witness condition, the participant
learned that 1 of the other cooperative players had been selected
and had chosen to levy a fine on the free rider of $1.65. Thus, in
the first condition the participant instigated the punishment, and in
the second condition the participant witnessed the punishment.
There were two corresponding forecaster conditions in which
participants played the entire game but were asked to predict how
they would feel if they had either instigated or witnessed a $1.65
punishment levied against the free rider. Finally, there was a
no-punishment control condition in which participants were never
informed about any punishment options.
Materials. The materials were identical to those used in Study
1 except where noted. New to this study was a single item at the
end of the experiment asking participants to predict how they
would have felt had they been in one of the other conditions. Thus,
the instigators (both actual and forecasters) were asked, “Imagine
that it had been another player who did the punishing rather than
you. Do you think you would be feeling better, worse, or about the
same right now?” Participants in the witness conditions and the
no-punish control condition, by contrast, were asked how they
would be feeling had they punished the free rider. The question
used a 7-point scale ranging from 1 (worse than now)to7(better
than now), with a neutral midpoint at 4 (about the same).
Participants earned an average of $1.89, and the free rider
earned $4.41. Eleven of 15 people in the instigator condition
penalized the free rider by spending an average of $0.43, resulting
in an average punishment of $1.28. Nobody penalized either of the
cooperative confederates in the game. As before, we checked to
see whether we could collapse the affect ratings across time and
valence. The 12-item affect scale had acceptable reliability (␣⫽
.81), and subsequent tests revealed no interactions with time or
valence relevant to the hypotheses. There was a significant main
effect for time such that people’s mood improved during the
10-min distracter task, F(1, 27) ⫽7.10, p⬍.05, and there was a
marginal Valence ⫻Condition interaction, revealing that the rep-
lication of Study 1 (see below) was somewhat more pronounced
for negative affect, F(1, 27) ⫽3.61, p⫽.05.
Replicating Study 1, participants in the punisher condition (M⫽
3.74, SD ⫽.86) experienced significantly worse mood than did the
no-punish controls (M⫽4.58, SD ⫽.58), t(70) ⫽2.98, p⬍.01.
This replicates the finding that the act of punishing the free rider
lowered participants’ mood. Also as predicted, there was no affective
cost to seeing someone else punish the free rider: Participants who
merely witnessed the punishment of the free rider (M⫽4.33, SD ⫽
.75) were in no worse a mood than were the no-punish controls (M⫽
4.58, SD ⫽.58), t(70) ⫽0.86, p⫽.39, and were in a significantly
better mood than were the punishers, t(70) ⫽2.10, p⬍.04.
As seen in Table 1, forecasters did not predict this affective
difference between punishing and witnessing. Those asked how
they would feel if they instigated the punishment predicted that
they would feel better than the punishers actually felt, t(70) ⫽
All predictions in Study 3 were tested with planned comparisons.
CONSEQUENCES OF REVENGE
2.58, p⫽.02, replicating the finding in Study 1 that people do not
anticipate the negative affective consequences of punishment. Also
as predicted, there was no significant difference between the
forecasts of people in the punisher and witness conditions, t(70) ⫽
0.14, p⫽.89, suggesting that forecasters did not appreciate the
different effects of exacting revenge oneself versus witnessing
someone else doing it. A contrast testing the prediction that the
mean in the actual punishers condition was lower than the means
in the actual witness, forecaster punisher, and forecaster witness
conditions was significant, t(70) ⫽2.87, p⬍.01, and the residual
from this contrast was not significant.
As in Study 1, the means in the forecaster conditions were not
significantly higher than the mean in the no-punish control condition,
t(70) ⫽0.68, p⫽.50. However, when we asked people directly at the
end of the study how they would have felt had they been in the other
condition, a different picture emerged. The punishers reported that
they would feel slightly, though not significantly, worse if they had
not been allowed to punish. Their average response was 3.76 (SD ⫽
1.26), slightly below the midpoint of the scale (4 ⫽about the same).
Both the witnesses (M⫽5.0, SD ⫽1.57) and the nonpunishers (M⫽
5.29, SD ⫽1.38), however, reported that they would feel significantly
better had they been permitted to punish the free rider. Both groups
were higher than the punishers ( p⬍.001), as well as the neutral
midpoint of the scale, t(27) ⫽3.43, p⬍.01; t(13) ⫽3.48, p⬍.01.
Thus, not only did these nonpunishers fail to recognize the hedonic
cost associated with punishment, they actually predicted hedonic
gains from the act of punishment.
Rumination about the free rider. As in Study 1, there were no
differences between conditions at Time 1 in how much people
reported thinking about the free rider, F(4, 70) ⫽0.67, p⫽.61.
Also replicating Study 1, punishers thought about the free rider at
Time 2 more than the nonpunishers did (M⫽4.2 vs. 2.5), t(70) ⫽
2.29, p⫽.03, and more than punisher forecasters predicted they
would (M⫽2.8), t(70) ⫽1.95, p⫽.06. These differences
disappear when the amount of punishment is statistically con-
trolled, F(2, 39) ⫽1.59, p⫽.22, suggesting that punishment
mediates this relationship. And, as before, rumination about the
free rider was related to both punishment (b⫽.42) and mood (b⫽
⫺.42) at Time 2, and the direct effect of punishment on mood was
reduced from b⫽⫺.40 ( p⬍.05) to ⫺.22 (ns) when rumination
was included in the equation. This reduction was marginally sig-
nificant by the Sobel test (z⫽1.65, p⫽.10). As predicted,
punishers reported thinking about the free rider more than did
witnesses, but this difference was not significant, t(70) ⫽0.85,
p⫽.40 (see means in Table 2).
Not only did rumination partially mediate the effect of punish-
ment on affect, but once again affect appeared to partially mediate
the effect of punishment on rumination. Affect was related to
punishment (b⫽⫺.40) and rumination (b⫽⫺.42) at Time 2, and
the direct effect of punishment on rumination was reduced from
b⫽.42 ( p⬍.05) to .26 (ns) when mood was included in the
equation. This reduction was marginally significant by the Sobel
test (z⫽1.61, p⫽.11).
Further, forecasters failed to predict the negative affective con-
sequences of revenge, in part because (a) the act of punishment led
people to ruminate about the perpetrator more, (b) this rumination
prolonged the negative affect, (c) the negative affect prolonged
rumination, and (d) people failed to predict the effects of punish-
ment on rumination and affect.
Study 3 replicated Study 1: People expected the act of punish-
ment to make them feel better, but they were wrong, it actually
made them feel worse. As in Study 1, the results suggest that one
mechanism by which this occurs is that punishment causes people
to think about the offender more, and this in turn maintains their
negative affect. We also have evidence that negative affect pro-
motes rumination. Study 3 extended these findings by showing that
punishing the free rider themselves made people feel worse than
witnessing someone else delivering the punishment, a distinction
that forecasters did not anticipate. Thus, the present studies impli-
cate two reasons why people mispredict the affective costs of
punishment: They underestimate the extent to which punishment
will make them ruminate about the free rider, and they fail to
realize that this is especially true if they instigate the punishment,
as opposed to seeing someone else do it.
One interesting finding in Study 3 was that people in the witness
condition did not report feeling better than people in the no-
punishment control condition. That is, seeing someone else punish
the free rider did not have the affective cost of punishing her
oneself (as we predicted), but neither did it confer an affective
benefit. On the face of it this finding is inconsistent with the
literature on schadenfreude, the idea that people take pleasure in
the misfortune of others (Hareli & Weiner, 2002; Leach, Spears,
The results on the individual affect items were consistent with those
reported here. Each negatively valenced item was higher for the punishers
than for people in the other conditions. Similarly, each positively valenced
item was lower for the punishers. The effect was more pronounced for the
negative affect items negative (1.52), vengeful (1.45), and irritated (0.59).
Thus, as in Study 1, it is primarily an increase in negative affect that results
from administering punishment.
Study 3: Mean Participant Affect (With Standard Deviation in
Parentheses) by Condition (7-Point Scale)
Condition Experiencer Forecaster
No-punish control 4.58
Note. The higher the number, the more positive the reported affect.
Means with different superscripts are significantly different, p⬍.05.
Study 3: Mean Participant Rumination (With Standard
Deviation in Parentheses) by Condition (7-Point Scale)
Condition Experiencer Forecaster
Witness 3.57 (2.21) 3.33 (2.35)
No-punish control 2.50
Note. The higher the number, the more people reported thinking about
the free rider. Means with different superscripts are significantly different,
p⬍.06. Means without subscripts do not differ from any other means.
1322 CARLSMITH, WILSON, AND GILBERT
Branscombe, & Doosje, 2003; Smith et al., 1996). One difference
between the present experiments and work on schadenfreude is
that our participants had been directly harmed by the free rider, as
opposed to being passive bystanders to an event. In the Smith et al.
(1996) studies, for example, participants read a news account of a
target individual who failed to be accepted to medical school, and
in the Leach et al. (2003) studies, participants learned that a rival
soccer team lost to the home team. In both cases the misfortune
was quite distant from the participant. It may be that when people
are directly harmed by a transgressor and experience anger, the
pleasure of seeing that person suffer is mitigated by the rumination
effects we found in the current studies. That is, learning that
someone else punished the free rider may have some affective
benefit, but it may also keep one’s attention on the transgression,
making it difficult to move on and think about something else. The
important point for our purposes is that people who delivered the
punishment themselves ruminated the most and experienced the
most negative affect, contrary to people’s forecasts.
The present studies addressed three questions: (a) What is the
affective consequence of exacting revenge? (b) Do people accurately
predict this consequence, and, if not, (c) Why do people make this
error of prediction? In answer to the first question, Studies 1 and 3
found that most people given the opportunity to punish a free rider did
so, but they ended up feeling worse than did people who did not have
the opportunity to punish. In answer to the second question, Studies
1–3 found that people did not accurately predict the negative affective
consequences of exacting revenge; in fact, they predicted that it would
make them feel better. In answer to the third question, Studies 1–3
found that people believed that exacting revenge would bring closure,
in the sense that they would think less about the free rider, when in
fact it had the opposite effect—punishing the free rider made people
think about her more, which in turn made them feel worse. Further,
Study 3 found that the punishment 3rumination 3negative affect
sequence occurred only when people punished the free rider them-
selves and not when they witnessed someone else delivering the
punishment. And yet, forecasters did not anticipate that delivering the
punishment themselves would make them feel worse than seeing
someone else do it.
The present results add to the growing body of research showing
that forecasters do not simply exaggerate the impact of an emo-
tional event, they sometimes misunderstand which of two sets of
conditions will make them the happiest (Gilbert & Ebert, 2002;
Wilson, Centerbar, Kermer, & Gilbert, 2005). This finding is
particularly striking in the domain of anger, because across human
history there have been countless acts of revenge, many motivated
by the belief that “to witness suffering does one good, to inflict it
even more so” (Nietzsche, 1887/1996, p. 48)—an affective fore-
cast that the present studies show is wrong, at least under some
The results also suggest an exception to the focalism hypothesis
(Wilson et al., 2000), which argues that people typically overestimate
how much they will think about a future emotional event. We found
the opposite effect in the present studies, in that people in the pun-
ishment conditions underestimated how much they would think about
the free rider. Although it will take further research to identify the
precise conditions under which people under- or overestimate how
much they will think about an event, we suspect that the present
results are specific to forecasts about anger, which is an emotion that
makes it difficult to think about anything else (Fredrickson & Brani-
gan, 2005). Anger led to punishment which led to increased rumina-
tion about the free rider, and this rumination may have prevented
people from engaging other strategies (e.g., trivialization) that would
have allowed them to move on and think about something else.
Participants failed to anticipate these effects, possibly due to their lay
theories that punishment would improve their mood (e.g., Bushman et
It is beyond the scope of this article to identify why these
theories are wrong, especially when most people have experience
in situations like the ones created in our experiments. Many people
have retaliated against a driver who was rude them, for example,
or in some other way punished someone who angered them. Why
have they not learned that such retaliation can make them feel
worse? One reason may be that seeing a transgressor receive his or
her comeuppance is pleasurable under some conditions, as found
in the literature on schadenfreude. People may overgeneralize from
such situations, failing to appreciate the conditions (such as those
identified in the present studies) that turn punishment into an
affectively costly action.
The Neural Basis of Punishment
Our findings are particularly interesting in light of recent reports
of the neural bases of punishment. De Quervain and colleagues
(2004) found that the dorsal striatum, the brain region that is
closely related to pleasure, was activated when participants de-
cided to punish a fellow participant who had violated a social
norm. In other words, that punishment was associated with posi-
tive affect, not negative affect. These measurements, however,
were taken during the 1-min period prior to making the punish-
ment decision, precisely the same time frame represented by the
forecasters in the present experiment. In both studies, then, people
showed clear expectations that the punishment would lead to
pleasure. The present studies, however, also measured hedonic
reactions after the punishment was conducted and found that
punishers were substantially less happy than the nonpunishers after
both a 1- and 10-min delay. Thus, our results corroborate de
Quervain et al.’s findings but add the crucial finding that predicted
and experienced emotion are substantially different.
Motives for Punishment
Our findings support a functional account of punishment—
people use punishment to strategically repair their negative mood.
Clearly, though, this is not the only function of punishment, and it
is left to future research to integrate the different explanations of
punishment and to delineate the circumstances under which one
particular motive or function becomes dominant.
Carlsmith and his colleagues (Carlsmith, 2006; Carlsmith &
Darley, 2008; Carlsmith, Darley, & Robinson, 2002; Carlsmith,
Monahan, & Evans, 2007) have shown that citizens’ intuitive
theories of punishment are consistent with Immanuel Kant’s the-
ory of retributive punishment. When people punish, they seek to
CONSEQUENCES OF REVENGE
give a perpetrator what he or she deserves and to make the
punishment proportional to the offense. The function of the pun-
ishment in these experiments appears to be one of satisfying the
“justice motive” rather than a more instrumental goal. Specifically,
Carlsmith found little support for the hypothesis that punishers
seek to use punishment for the utilitarian goals of deterrence
(Carlsmith, 2006; Carlsmith et al., 2002) or incapacitation (Darley,
Carlsmith, & Robinson, 2000).
The critical difference between studies that have found support
for retributive punishment and the current hedonic regulation
accounts appears to be the participants’ level of involvement.
Specifically, the justice motive is dominant in third-party obser-
vations of injustice—such as one would experience as a juror. By
contrast, when the respondent has been directly harmed by the
perpetrator (as in the present experiments), additional motives may
be activated, including the desire to repair one’s mood.
In sum, our results suggest that, in some instances at least, revenge
has hedonic consequences that are the opposite of what people expect.
Revenge can prolong peoples’ hedonic reactions to a transgression
because punishing others can cause people to continue to think about
(rather than to forget) those whom they have punished. As Sir Frances
Bacon noted more than three centuries ago, “A man that studieth
revenge, keeps his own wounds green, which otherwise would heal,
and do well” (Spedding, Ellis, & Heath, 1858).
Baier, K. (1955). Is punishment retributive? Analysis, 16, 25–32.
Barnstead-Klos, A. (2007, May 6). Re: Virginia Tech Massacre thoughts.
Message posted to http://tinyurl.com/5fy2n2
Bentham, J. (1962). Principles of penal law. In J. Bowring (Ed.), The works
of Jeremy Bentham (p. 396). Edinburgh, Scotland: W. Tait. (Original
work published 1843)
Berkowitz, L. (1989). Frustration-aggression hypothesis: Examination and
reformulation. Psychological Bulletin, 106, 59 –73.
Bushman, B. J. (2002). Does venting anger feed or extinguish the flame?
Catharsis, rumination, distraction, anger, and aggressive responding.
Personality and Social Psychology Bulletin, 28, 724 –731.
Bushman, B. J., Baumeister, R. F., & Phillips, C. M. (2001). Do people
aggress to improve their mood? Catharsis beliefs, affect regulation
opportunity, and aggressive responding. Journal of Personality and
Social Psychology, 81, 17–32.
Carlsmith, K. M. (2006). The roles of retribution and utility in determining
punishment. Journal of Experimental Social Psychology, 42, 437– 451.
Carlsmith, K. M., & Darley, J. M. (2008). Psychological aspects of retrib-
utive justice. In M. P. Zanna (Ed.), Advances in experimental social
psychology, (Vol. 40, pp. 193–236). San Diego, CA: Elsevier.
Carlsmith, K. M., Darley, J. M., & Robinson, P. H. (2002). Why do we
punish? Deterrence and just deserts as motives for punishment. Journal
of Personality and Social Psychology, 83, 284 –299.
Carlsmith, K. M., Monahan, J., & Evans, A. (2007). The function of
punishment in the “civil” commitment of sexually violent predators.
Behavioral Sciences & the Law, 25, 437– 448.
Darley, J. M., Carlsmith, K. M., & Robinson, P. H. (2000). Incapacitation
and just deserts as motives for punishment. Law and Human Behavior,
24, 659 – 683.
De Quervain, D. J.-F., Fischbacher, U., Treyer, V., Schellhammer, M.,
Schnyder, U., Buck, A., & Fehr, E. (2004). The neural basis of altruistic
punishment. Science, 305, 1254 –1258.
Dollard, J., Doob, L., Miller, N., Mowrer, O. H., & Sears, R. (1939).
Frustration and aggression. New Haven, CT: Yale University Press.
Ezorsky, G. (1972). Philosophical perspectives on punishment. Albany,
NY: SUNY Press.
Fredrickson, B. L., & Branigan, C. (2005). Positive emotions broaden the
scope of attention and thought-action repertoires. Cognition and Emo-
tion, 19, 313–332.
Freud, S. (1961). Civilization and its discontents. New York: Norton.
Frijda, N. H. (1994). The lex talionis: On vengeance. In S. H. M. van
Goozen, N. E. Van De Poll, & J. A. Sergeant (Ed.), Emotions: Essays on
emotion theory (pp. 263–289). Hillside, NJ: Erlbaum.
Gilbert, D. T., & Ebert, J. E. J. (2002). Decisions and revisions: The
affective forecasting of changeable outcomes. Journal of Personality
and Social Psychology, 82, 503–514.
Gintis, H. (2008). Punishment and cooperation. Science, 319, 1345–1346.
Hareli, S., & Weiner, B. (2002). Dislike and envy as antecedents of
pleasure at another’s misfortune. Motivation and Emotion, 26, 257–277.
Hobbes, T. (1950). Leviathan. New York: E. P. Dutton.
Kant, I. (1952). The science of right (W. Hastie, Trans.). In R. Hutchins
(Ed.), Great books of the Western world (pp. 397– 446). Edinburgh,
Scotland: T. & T. Clark. (Original work published 1790)
Konecni, V. (1974). Self-arousal, dissipation of anger, and aggression.
Personality and Social Psychology Bulletin, 1, 192–194.
Leach, C. W., Spears, R., Branscombe, N. R., & Doosje, B. (2003).
Malicious pleasure: Schadenfreude at the suffering of another group.
Journal of Personality and Social Psychology, 84, 932–943.
Lee, J. (1993). Facing the fire: Experiencing and expressing anger appro-
priately. New York: Bantam Books.
Loewenstein, G., O’Donoghue, T., & Rabin, M. (2003). Projection bias in
predicting future utility. Quarterly Journal of Economics, 118, 1209 –1248.
McKean, E. (Ed.). (2005). The new Oxford American dictionary (2nd. ed.).
Oxford, England: Oxford University Press.
Mellers, B. A., & McGraw, A. P. (2001). Anticipated emotions as guides
to choice. Current Directions in Psychological Science, 10, 210 –214.
Nietzsche, F. (1996). On the genealogy of morality (D. Smith, Trans.). Oxford,
England: Oxford University Press. (Original work published 1887)
Quinton, A. M. (1954). On punishment. Analysis, 14, 512–517.
Rusting, C. L., & Nolen-Hoeksema, S. (1998). Regulating responses to
anger: Effects of rumination and distraction on angry mood. Journal of
Personality and Social Psychology, 74, 790 – 803.
Schkade, D. A., & Kahneman, D. (1998). Does living in California make
people happy? A focusing illusion in judgments of life satisfaction.
Psychological Science, 9, 340 –346.
Simon, L., Greenberg, J., & Brehm, J. (1995). Trivialization: The forgotten
mode of dissonance reduction. Journal of Personality and Social Psy-
chology, 68, 247–260.
Skinner, B. F. (1938). The behavior of organisms. New York: Appleton-
Skinner, B. F. (1974). About behaviorism. New York: Knopf.
Smith, R. H., Turner, T. J., Garonzik, R., Leach, C. W., Urch-Druskat, V.,
& Weston, C. M. (1996). Envy and Schadenfreude’. Personality and
Social Psychology Bulletin, 22, 158 –168.
Spedding, J., Ellis, R. L., & Heath, D. D. (Eds.). (1858). The works of
Francis Bacon (Vol. 6). London: Longman.
Watson, J. B. (1924). Behaviorism. New York: Norton.
Wilson, T. D., Centerbar, D. B., Kermer, D. A., & Gilbert, D. T. (2005). The
pleasures of uncertainty: Prolonging positive moods in ways people do not
anticipate. Journal of Personality and Social Psychology, 88, 5–21.
Wilson, T. D., & Gilbert, D. T. (2003). Affective forecasting. In M. P.
Zanna (Ed.), Advances in experimental social psychology (Vol. 35, pp.
345– 411). San Diego, CA: Elsevier.
Wilson, T. D., Wheatley, T. P., Meyers, J., Gilbert, D. T., & Axsom, D.
(2000). Focalism: A source of durability bias in affective forecasting.
Journal of Personality and Social Psychology, 78, 821– 836.
Received July 10, 2006
Revision received March 10, 2008
Accepted March 12, 2008 䡲
1324 CARLSMITH, WILSON, AND GILBERT