ORGANIZATIONAL BEHAVIOR AND HUMAN DECISION PROCESSES

Vol. 72, No. 1, October, pp. 6378, 1997

ARTICLE NO. OB972730

Consequences of Regret Aversion 2: Additional

Evidence for Effects of Feedback on
Decision Making
Marcel Zeelenberg
Eindhoven University of Technology, Eindhoven, The Netherlands

and

Jane Beattie
University of Sussex, Brighton, United Kingdom

We discuss the effects of anticipated and experienced regret on

decision making under uncertainty. In previous research, using
the standard, context-free, gamble paradigm, we found that deci-
sion makers anticipate the regret they can experience as a result
of post-decisional feedback on forgone outcomes (Zeelen-
berg, Beattie, van der Pligt, & de Vries, 1996). In the present
research we move away from the gamble paradigm, on to richer
contexts. In Experiments 1 and 2, involving investment decision
making and decision making in the ultimatum game, it is shown
that the expectation of feedback on forgone outcomes influences
decision making and can promote more risk seeking behavior.
Experiment 3 focused on effects of retrospective regret and shows
that actual feedback on foregone outcomes influences the experi-
ence of regret and subsequent decision making. The results of
these studies support our earlier work on regret aversion. q 1997
Academic Press

Jane Beattie died from cancer on March 25, 1997, at the age of 36. The research presented in
this article was a real joint effort. Because of her illness and her untimely death Jane was never
able to read the final version of this paper. This research was supported by a European Community
Marie Curie Fellowship (ERBFMBICT961052) awarded to Marcel Zeelenberg. Preliminary data
of this paper were presented at IAREP-21, held from 11 to 14 September 1996, in Paris, France,
and at SPUDM-16, held from 17 to 21 August 1997, in Leeds, United Kingdom. Thanks are due
to Ian Todd for his help in collecting the data of Experiment 1, and to Henk Aarts, Rosie Murray,
and two anonymous reviewers for their helpful comments on an earlier draft of this article.
Address correspondence and reprint requests to Marcel Zeelenberg, who is now at Tilburg
University, Department of Business Administration, PO Box 90153, 5000 LE Tilburg, The Nether-
lands. E-mail: [email protected].
63 0749-5978/97 $25.00
Copyright q 1997 by Academic Press
All rights of reproduction in any form reserved.
64 ZEELENBERG AND BEATTIE

Negative emotional states often evolve in response to unfavorable outcomes

of decisions. Regret is experienced when it turns out, in retrospect, that you
should have chosen something different. For example, when you invested your
uncles inheritance in a particular option and find out that another option
would have provided you with a much larger profit, or when you, while negotiat-
ing, decided to make the other party a too generous offer and learn later that
they would also have accepted a much lower offer.
Regret is an aversive state, and it is therefore not surprising to see that we
do not sit and wait for this emotion to occur. To the contrary, we are apt to
choose in such a way that we minimize our future regrets. In the present paper
we focus on how this regret aversion influences our decisions under uncertainty.
Although ideas about the role of regret were long present in decision research
(e.g., Savage, 1951), it was not until the early 80s before the anticipatory
aspects of regret were explicitly incorporated in Bells (1982) and Loomes and
Sugdens (1982) regret theory. In regret theory it is assumed that people
compare the actual outcome with what the outcome would have been, had a
different choice been made, and that they experience regret when the foregone
outcome would have been better, and rejoicing when the foregone outcome
would have been worse. It is also assumed that these emotional consequences
of decisions are anticipated and taken into account when making decisions.
Obviously, not all our decisions are influenced by the anticipation of regret,
and an important question in regret research is therefore concerned with the
boundary conditions under which regret exerts its influence. Recent research
showed that the expectation of feedback on the rejected options is one of the
prime determinants of whether people anticipate regret or not (Josephs Larrick,
Steele, & Nisbett, 1992; Larrick & Boles, 1995; Ritov & Baron, 1995; Ritov,
1996; Zeelenberg, Beattie, van der Pligt, & de Vries, 1996). We discuss here
primarily our own work on the effects of expected feedback, because the present
research is a follow up on these studies. In our former work we showed that
the anticipation of regret can promote both risk avoiding and risk seeking
tendencies, whereas earlier research suggested that it would only promote risk
aversion (see for details, Zeelenberg et al., 1996).
The design of the experiments reported in Zeelenberg et al. (1996) was as
follows. Participants were given a choice between two gambles, one being
relatively risky and the other being relatively safe. Participants always knew
the outcome of the riskier gamble and the probability of winning it. For example
in our Experiment 1 the riskier gamble would result in a gain of 130 Dutch
Guilders with a probability of 35%, or in no gain with a probability of 65%. In
the safer gamble they could gain an unknown amount X with a probability of
65%, or no gain with a probability of 35%. It was the participants task to write
down the value of X for which they found the gambles equally attractive. Next,
feedback on one of the gambles was manipulated orthogonally to the riskiness
of the gambles. In all three experiments we had a Safe Feedback condition, in
which the safer gamble would always be resolved, and a Risky Feedback condi-
tion, in which the riskier gamble would always be resolved. In addition, all
participants expected to learn the outcome of the chosen gamble. Participants
 CONSEQUENCES OF REGRET AVERSION 65

in Safe Feedback condition were predicted to choose the safer gamble. The
safer gamble would provide them with feedback on the chosen gamble only,
and protect them from threatening feedback on the riskier gamble. Likewise,
participants in the Risky Feedback condition, who would always learn the out
come of the riskier gamble, were predicted to opt for the riskier gamble. This
predicted pattern was found in all three studies, in both high and low variance
gambles, and in gambles involving both gains and losses. These studies thus
show that the anticipation of regret influences behavioral choice and can pro-
mote risk averse and risk seeking tendencies.
In this paper we present three extensions of Zeelenberg et al. (1996). The
first extension has to do with the fact that we used a choice between matched
gambles paradigm in our former studies. Because participants made the gam-
bles equally attractive, they would be indifferent when choosing between the
two. The expectation of feedback then pushed participants preferences either
in the direction of the riskier gamble or in the direction of the safer gamble.
One might argue that the influence of regret on decision making is exaggerated
in such a paradigm, because of the initial indifference between the two gambles,
and that our experiments only show that anticipated regret can be a tie-
breaking mechanism. Experiment 1 of the present paper goes one step further
and shows that expected feedback can also influence preferences when there
is no initial indifference between the options.
The second extension is that we now focus on decisions that are less context-
free. In Zeelenberg et al. (1996) we used the standard, context-free, gamble
paradigm. Some researchers might argue that this approach has limited gener-
ality, and hence propose a move away from the gamble as research methodology,
on to richer contexts. The present research attempts to do so. Experiment 1
focuses on a financial decision between two uncertain investment options.
Experiments 2a and 2b focus on decision making in an interpersonal context;
namely, the ultimatum game (Guth, Schmittberger & Schwarze, 1982).
The third extension is that we now also study experienced regret. We argue
that not only anticipated future regret affects decision making, but that also
experienced retrospective regret has an influence. Anticipated regret is evoked
by the expectation of feedback, whereas experienced regret stems from actual
feedback. In Experiment 3 we therefore manipulate the actual feedback follow-
ing a decision and measure the intensity of experienced retrospective regret.
By asking participants to make a second decision, after they received feedback
and indicated their regret, we assess the influence of this retrospective regret
on subsequent decisions.
In sum, the research presented in the present paper attempts to speak to
and extend earlier work on regret, feedback, and decision making. The present
research tries to show that expected, and actual, feedback does also affect
decision making in less context free situations where the choice options are
not matched and thus not equally preferable. In these experiments we manipu-
late the expected or actual feedback on the outcomes of the unchosen options.
The prediction is that participants will make regret minimizing decisions.
66 ZEELENBERG AND BEATTIE

EXPERIMENT 1: ANTICIPATED REGRET IN FINANCIAL DECISION

MAKING

Experiment 1 tested whether the manipulation of feedback could also influ-

ence decision making, and cause preference reversals, when the two options
are initially not equally attractive. Participants were asked to evaluate two
financial investment options that differed in riskiness: The High Interest Ac-
count (HIA) and the Government Bond (GB). The HIA was the safer option
since it would always provide them with a modest profit. The GB was riskier,
it could result in a large profit, or no profit at all.
We expected that in a situation where there is only feedback on the chosen
option (Choice Only Feedback condition), people would prefer to invest in the
safer option (HIA). This prediction is based on the finding that people are
generally risk averse when they choose between possible gains (cf., Kahne-
man & Tversky, 1979). In the case where there always is feedback on the riskier
option (Risky Feedback condition), choosing for the safer option can result in
regret when one finds out that the riskier option would have been a better
choice. This feedback was manipulated by informing the participants that their
sisters money was already invested in the GB.1 In this case we expected that
people would prefer to invest in this riskier option.

Method
Forty-one English undergraduate students volunteered to participate in this
experiment and were randomly assigned to one of the two Feedback conditions.
There were 20 participants in the Risky Feedback condition and 21 in the
Choice Only Feedback condition. Participants in the Risky Feedback condition
read the following scenario:
Your uncle has just died and left you 1000. You now have to decide how to invest the money
for five years. Your uncle has also left your sister 1000, but her money is already invested
for the same five years period in a Government Bond, which is guaranteed to pay back a total
sum between 1000 and 1800 at the end of the five years. You can choose to invest your
money in this type of investment too. A friend has just told you about another type of investment
which you could choose, a High Interest Account, which is guaranteed to pay back a total
sum between 1250 and 1350 at the end of the five years. You know that at the end of the
five years you will find out how much money you would have made if you had chosen the
Government Bond because your sister will tell you.

Participants in the Choice Only Feedback condition read a similar scenario,

in which there was no sister and their friend told them about the two options.
There was therefore no feedback on the HIA (the risky option), as there was
in the Risky Feedback condition. After reading the scenario all participants
were asked to indicate the strength of their preference for both investment
options. They could do this on two 7-point scales with endpoints labeled Defi-
nitely would not invest (1) to Definitely would invest (7).

1
Information about the outcome of an unchosen alternative provided by a social comparison
with a person who does obtain this outcome results in amplified regret, as was shown by Boles
and Messick (1995).
 CONSEQUENCES OF REGRET AVERSION 67

Results and Discussion

Figure 1 shows participants mean willingness to invest in both options
for the two feedback conditions. Participants answers were submitted to a 2
(Feedback: Risky vs Choice Only) 3 2 (Investment: HIA vs GB) analysis of
variance with Investment as a within-subjects factor. This analysis yielded a
Feedback 3 Investment interaction, F(1,39) 5 8.80, p , .005. As expected,
participants in the Choice Only Feedback condition showed risk aversion; they
were more willing to invest in the safer HIA (M 5 5.10) than in the riskier
GB (M 5 4.05), t(19) 5 3.05, p , .05. Participants in the Risky Feedback
condition showed the predicted opposite pattern; they were more willing to
invest in the riskier GB (M 5 5.19) than in the safer HIA (M 5 3.33), t(20) 5
2.23, p , .05.
These results clearly indicate that the expectation of feedback on the outcome
of a riskier option can promote risk seeking choices, even if the initial preference
was for the safer option. In a choice between two investment options partici-
pants preferred the safer investment over the risky investment if there was
only feedback on the chosen option. However, when participants expected that
there would always be feedback on the riskier option they preferred riskier
investment over the safer investment.
These findings corroborate and extend those reported in Zeelenberg et al.
(1996). They also show that expected feedback can cause a preference for risk
seeking in situations where the general preference is risk averse. These data
were obtained using a somewhat more ecologically valid paradigm, that is,
having participants make choices they could encounter in their own lives,
instead of using the traditional gamble paradigm.

EXPERIMENTS 2a AND 2b: ANTICIPATED REGRET IN THE

ULTIMATUM GAME

The present experiments explored another extension of our earlier findings.

Experiment 1 and our earlier experiments studied purely individual decision

FIG. 1. Willingness to invest as a function of feedback, measured on 7-point scales with

endpoints labeled Definitely would not invest (1) to Definitely would invest (7).
68 ZEELENBERG AND BEATTIE

making. The decisions were made by individuals and the outcomes only affected
the decision maker him or herself. In real life, however, our decisions often not
only affect ourselves, but also others. Moreover, our own outcomes might be
influenced by decisions of a number of other individuals. Hence, when making
decisions we often take into account the decisions and outcomes of others.
In Experiments 2a and 2b we tested whether anticipated regret could also
influence these interpersonal decisions. It has been argued that emotions play
a large role in negotiations (Barry & Oliver, 1996; Pillutla & Murnighan, 1996).
Larrick and Boles (1995) were the first to show that expected feedback can
influence negotiation decisions. Participants negotiated about a signing bonus
they could earn when deciding to work for a certain company ALPHA. They
either expected to learn or expected not to learn the offer of a competing
company BETA after they reached an agreement with ALPHA. Participants
who expected to learn the offer of BETA (Feedback condition), could regret or
rejoice about their decision to accept the bonus offered by ALPHA. These
participants wanted to have a higher bonus and were consequently less likely
to reach agreement, and were thus more risk-seeking, than those who did not
expect to learn the offer of BETA (No Feedback condition).
Although our Experiments 2a and 2b are similar to Larrick and Boles (1995)
experiment, they also differ in at least two ways. First, we used a different
feedback manipulation. In their research the feedback was information about
what could have been the result of negotiating with a third party, whereas in
our experiments the feedback was provided by one of the parties within the
negotiation. Second, and more importantly, the present experiments used differ-
ent negotiation task, namely the ultimatum game (Guth et al., 1982).
The ultimatum game is very simple. Two players are allotted a sum of money,
e.g. $10. Player 1 (often called the Proposer) offers some portion of the money
(e.g., $4) to player 2 (the Responder). If the Responder accepts, she gets the
$4, and the Proposer gets the rest ($6). If the Responder rejects the offer,
both players get nothing. The predictions from rational economic theory are
straightforward. Responders should accept the smallest amount of money, that
is, one cent, since this is more than they would get by rejecting the offer.
Proposers know this and should thus offer Responders only the one cent. Earlier
research using this game, however, showed that people hardly ever offer the
other player only one cent. Moreover, if they do so, the other players refuse
the offer. Commonly the average offers are in the regions of 3040%, with a
5050 split often as the mode. Offers of less then 20% are frequently rejected
(for an overview, see Camerer & Thaler, 1995).
How might anticipated regret influence the behavior of proposers? Proposers
can regret two things, offering too little money when the offer is rejected, and
offering too much when the offer is accepted. If we consider the fact that the
modal offer is 50%, and that offers over 20% are almost always accepted, there
is more chance of offering too much. However, there are two reasons why in a
normal ultimatum game the regret about offering too much money is generally
less severe than regret about offering too little money. First, when proposers
regret offering too much, they still have money, whereas when proposers regret
 CONSEQUENCES OF REGRET AVERSION 69

offering too little, they do not get any money at all. Second, although proposers
whos offers are accepted might infer that their offer was probably too high,
they do not know to what extent, and therefore the possible regret over offering
too much will not be that painful. Whereas for proposers whos offers are
rejected it is crystal clear that they offered too little.
Thus, in a normal ultimatum game the regret minimizing option is offering
too much. However, everything changes when feedback is introduced. If the
responders minimal acceptable offer is communicated to the proposers, they
might learn that a much lower offer would also have been accepted. Feedback
of this kind can make regret about an offer that is too high more severe because
it points out exactly how much less the proposer could have offered.2 Proposers
who expect this feedback might anticipate the possible regret, and move away
from the 50/50 split to less egalitarian offers.
In the present experiments we manipulated whether or not proposers knew
in advance that they would learn the responders minimal acceptable offer
after they had made their offer. We expected that proposers who expected
feedback on the minimal acceptable offer would be inclined to make lower
offers than proposers who did not expect this feedback. This was because
lowering their offers would lower the amount of possible regret because the
offer would be closer to the minimal acceptable offers. At the same time lowering
their offers would not need to result in a higher likelihood of the offer being
rejected, because offers are hardly ever rejected if they remain higher than 20%.

Experiment 2a
Method. Seventy Dutch undergraduate students volunteered to participate
in this experiment. They were randomly assigned to one of the two conditions
(No Feedback vs Feedback). There were 35 participants in each condition.
Upon arrival in the laboratory participants were seated behind a computer
screen via which the rules of the game were explained to them. Participants
were told that half of them were randomly selected to be proposer, and the
other half to be responder. In reality all of them were proposers. They were
told that they had to divide an amount of 100 Dutch Guilders between them-
selves and the Responder. Participants in the Feedback condition were told
that they would always learn the respondents minimal acceptable offer, and
that they would thus learn the exact amount of money that they should have
offered more to get their offer accepted or could have offered less and have
their offer still accepted. After all participants made their offer, the experiment
ended and the participants were debriefed.
Results. Because of the skewed distributions of the offers we used a Wil-
coxon-Mann-Whitney test to test for differences between the two conditions.

2
The regret stemming from an offer that turns out to be too high can be characterized as a type
of winners curse. The winners curse refers to a situation in which one has paid more than the
opponents reservation price. See Thaler (1992), and Bazerman and Neale (1992) for excellent
descriptions of this phenomenon.
70 ZEELENBERG AND BEATTIE

This test showed that, as predicted, offers in the Feedback condition (M 5

36.46 Dutch Guilders) were lower than in the No Feedback condition (M 5
42.80 Dutch Guilders), Z (corrected for ties) 5 2.18, p , .05.
The results were thus in accordance with our predictions. However, the data
do not conclusively rule out the possibility that other factors might have caused
the difference between the conditions. For example, one might argue that
decisions in the ultimatum game are very much based on strategic considera-
tions, and that our manipulation induced more strategic offers. This might
have happened because our manipulation focuses participants attention on
the minimal acceptable offer, whereas normally participants might not have
considered this (of course it might still be the case that this strategic thinking
is instigated by the anticipation of regret resulting from too high offers). We
ran a second study in attempt to investigate the relation between possible
considerations and the offers made.

Experiment 2b

This experiment had the same design as Experiment 2a. Dutch undergradu-
ate students volunteered to participate in this experiment. They were randomly
assigned to one of the 2 conditions (No Feedback vs. Feedback). There were
35 participants in each condition.
There were two differences between the present experiment and the former.
Whereas Experiment 2a involved real playing of the ultimatum game, the
present experiment was part of a one-hour series of different judgmental tasks,
and therefore participants were asked to indicate what they would decide when
playing the game (not an uncommon procedure in ultimatum game research,
see, e.g., Pillutla & Murnighan, 1996).
The second difference is that we now asked participants some questions
about their offer, after they stated it. First we asked them to indicate the extent
to which they thought that the other person would accept their offer. They
could do this on a 9-point scale with endpoints labeled will definitively not
accept (1) and will definitively accept (9). Next we asked them to indicate on
9-point scales, with endpoints does not apply (1) and does apply (9), to what
extent four different considerations played a role in their decision. These consid-
erations were: I wanted my offer to be as strategic as possible, I did not
want to feel regret over a too high offer, I did not want to feel regret over a
too low offer, I was afraid that my offer would not be accepted.
The results of the present experiment were strikingly similar to the former.
Offers in the Feedback condition (M 5 37.34 Dutch Guilders) were again lower
than in the No Feedback condition (M 5 43.57 Dutch Guilders), Z (corrected
for ties) 5 1.96, p , .05. Nevertheless, participants in both conditions found
it equally likely that their offers would be accepted by the responder (for both
conditions, M 5 7.69). Figure 2 depicts participants offers to the responders
in both feedback conditions, collapsed over Experiments 2a and 2b.
Also interesting are the relations between participants considerations and
 CONSEQUENCES OF REGRET AVERSION 71

FIG. 2. Amount of money (of a fixed pie of 100 Dutch Guilders) offered by the proposers to
responders as a function of feedback (combined data from Experiments 2a and 2b).

their offers.3 Overall we found that only ratings on the consideration I did
not want to feel regret over a too high offer correlated significantly with the
amount offered to the responder, r 5 2.35, p , .005. Thus, the more participants
indicated that they anticipated the regret over an offer that could be too high,
the lower their offers were. The correlations between the other considerations
and the offer were all non-significant and smaller than 0.1. If we look at the
correlations between the considerations and the offers in the different feedback
conditions we get an even clearer picture. In the No Feedback condition there
were no significant correlations between the considerations and the offer. In the
Feedback condition, however, only the correlation between the consideration I
did not want to feel regret over a too high offer and the amount offered to the
responder was significant, r 5 2.48, p , .005.
These correlations confirm our reasoning. We hypothesized that when people
anticipate regret while playing the ultimatum game, they would anticipate the
regret arising from offers that would be too high. This would then promote
lower offers. We also hypothesized that the anticipation of regret were likely
to have its influence when people expect feedback on the minimal acceptable
offer of the responder. In the present experiment we did not only found that
the offers were lower in the Feedback condition, but we also found that the
offers were lower as more regret was anticipated.

Discussion
The results of these experiments clearly indicate that the expectation of
feedback, a possible cause of regret, influences peoples choices in the ultimatum

3
No between condition differences were found for the ratings of the considerations. The mean
rating for each consideration was: I wanted my offer to be as strategic as possible (M 5 5.9), I
did not want to feel regret over a too high offer (M 5 3.7), I did not want to feel regret over a
too low offer (M 5 5.3),I was afraid that my offer would not be accepted (M 5 5.2).
72 ZEELENBERG AND BEATTIE

game. When deciding how much to offer the responder, participants who expec-
ted feedback on the responders minimal acceptable offer made lower offers
than participants who did not expect to receive this feedback. This behavior
reflects regret aversion because lower offers result in less regret if accepted.
The offers were, in general, not that low that they would be rejected, which
suggests that participants minimized both the regret that could arise from
offering too much and the regret that could arise from offering too little.
This finding thus extends our previous findings, because it shows that the
motive of minimizing regret can also influence decision making in interpersonal
contexts. It thereby replicates the findings of Larrick and Boles (1995) in a
very different negotiation context. As in Experiment 1, the results also show
that regret effects can be obtained in richer situations, in which different
motives may influence the final decision.

EXPERIMENT 3: EXPERIENCED REGRET IN THE ULTIMATUM GAME

This experiment addressed the effects of experienced regret on future behav-

ior. Until now regret research has predominantly focused on the effects of the
anticipation of this emotional state on decision making. This is, however, only
half of the story. As emotion researchers know, the emotion a person brings
into the situation has a considerable impact on how that person behaves (e.g.,
Zeelenberg, van der Pligt, & Manstead, in press-a; Zeelenberg, van Dijk,
Manstead, & van der Pligt, in press-b). Hence, decision makers who have just
experienced regret will probably behave differently than decision makers who
did not experience regret. Since regret is a common emotion, that is experienced
quite regularly (Landman, 1993; Shimanoff, 1984), it seems not only worth-
while, but also a necessary next step, to investigate how this experience shapes
our future behavior (cf. Zeelenberg, 1996). We suspect that the behavior of
people who experience regret can be characterized as a sort of emotion manage-
ment. They will behave in such a way that their regret will disappear. They
will probably choose options that shield them from further regret, and so restore
their original, more or less neutral, emotional state.
In the present experiment the focus was again of decision making in the
ultimatum game. In Experiments 2a and 2b we saw that participants who
expect feedback on the responders minimal acceptable offer anticipate the
regret from offering too much and decide in such a way to prevent this regret
from happening. What happens if they do not succeed in this? How will this
regret influence their decisions when playing a second ultimatum game? These
are the questions addressed by Experiment 3.
In this experiment participants played the ultimatum game and expected
to receive feedback on the responders minimal acceptable offer. They learned
that the minimal acceptable offer is either 2 Guilders less than they offered
(2 Guilders Too Much condition) or 10 Guilders less than they offered (10
Guilders Too Much condition). The prediction was that participants in the latter
condition would experience most regret. Next participants played a second
ultimatum game against another responder, and we predicted that participants
 CONSEQUENCES OF REGRET AVERSION 73

in the 10 Guilders Too Much condition would lower their offers more than
participants in the 2 Guilders Too Much condition. Moreover, we predicted that
they would do so because of the regret they experienced over the first offer.

Method
Forty-six Dutch undergraduate students volunteered to participate in this
experiment. They were randomly assigned to one of the 2 Feedback conditions
(2 Guilders Too Much vs 10 Guilders Too Much). There were 23 participants
in each condition.
The procedure of this experiment was for the first part similar to that for
the participants in the feedback condition of Experiment 2a. However, after
participants made their offer the experiment did not end. Participants were
then informed about the decision of the responder. Depending on the condition
they were in participants either learned that their offer was accepted and they
could have offered 2 Guilders less, or that offer was accepted and they could
have offered 10 Guilders less. Following this feedback, the regret over their
offer was assessed. This was done by presenting the participants on the com-
puter screen with the feedback and asking them to indicate, on 7-point scales,
how much regret they experienced and how good they thought their offer was
in retrospect.4 Both ratings, after reverse coding for the second question, were
combined in a regret measure (Cronbachs a 5 .76).
Next, participants were informed that they were going to play a second
round. We wanted the difference in regret to be the main difference between
the two conditions. Therefore we told all participants that they would now play
with a different responder. They were also told that the average minimal
acceptable offer in the first round was 22 Guilders. After all participants had
made their second offer, the experiment ended and the participants were de-
briefed.

Results
Table 1 depicts the results of this experiment. These support our predictions.
The first offer was very much similar to the offers in the feedback conditions

TABLE 1
Mean First Offer, Regret, and Second Offer for Both Feedback Conditions

Feedback condition

2 Guilders too much 10 Guilders too much F(1,44) p,

First offer 35.69 30.43 2.79 ns.

Regret 1.30 2.09 11.02 .003
Second offer 34.69 26.34 8.41 .006

4
According to Bell (1982, p. 961) regret stems from realizing after the fact that one has made
the wrong decision.
74 ZEELENBERG AND BEATTIE

of Experiments 2a and 2b, and there was no significant difference between the
two conditions on this offer. Next we analyzed the regret ratings. Note that
the overall regret was very low because all participants got their offers accepted.
There was, however, a significant difference in the regret experienced after
receiving the feedback. Participants in the 10 Guilders Too Much condition
reported more regret than participants in the 2 Guilders Too Much condition.
We hypothesized that the experienced regret would influence the subsequent
offer made. As can be seen from Table 1, the second offer was lower in the 10
Guilders Too Much condition than in the 2 Guilders Too Much condition. In a
analysis of covariance the effect of Feedback remained significant after control-
ling for differences in the first offer, F(1,43) 5 7.41, p , .01, even though the
first offer appeared to be a highly significant covariate, F(1,43) 5 139.45, p ,
.001. More importantly, when we also included the experienced regret as a
covariate (F(1,42) 5 9.31, p , .005), the effect of Feedback was no longer
significant, F(1,42) 5 2.73, ns. This supports our prediction that the effects of
feedback on the second offer are mediated by the regret experienced over the
first offer.

Discussion
This experiment shows that the experience of retrospective regret influences
subsequent decision making. Participants played an ultimatum game and re-
ceived feedback on how much less they could have offered and still had their
offer accepted. Participants who could have offered 10 Guilders less experienced
more regret than participants who could have offered only 2 Guilders less.
When participants were asked to play a second round of the ultimatum game
their offers were influenced by the amount of regret experienced. It thus seems
that participants engaged in a sort of emotion, or regret, management; they
behaved in such a way that their regret will disappear, or future regret will
be minimized. This finding extends our previous findings from Experiments 1
and 2, and from Zeelenberg et al. (1996), because it shows that not only antici-
pated future regret, but also experienced retrospective regret influences behav-
ioral decision making. To our knowledge the present data are the first demon-
stration of such an influence.
Pillutla and Murnighan (1996) recently also studied on the role of emotions
in decision making in the ultimatum game. They focused on how experienced
anger might influence the decisions of responders whether or not to accept the
offer made by the proposer. We believe that their experiment, our experiments,
and Larrick and Boles (1995) experiment, provide initial support for Barry
and Olivers (1996) recent approach to the role of emotion in negotiation, in
which emotion plays a large role at several stages in the negotiation process.

GENERAL DISCUSSION

Taken together, the results of these experiments show that both the antici-
pation of regret caused by the manipulation of expected feedback, and the
 CONSEQUENCES OF REGRET AVERSION 75

experience of regret caused by actual feedback, have a profound influence on

decisions in several contexts. Evidence for effects of anticipated regret was
found in financial investment decisions (Experiment 1) and in decisions in
the ultimatum game (Experiments 2a and 2b). Experiment 3 showed that
experienced regret in an ultimatum bargaining situation influences subsequent
decisions. The research presented thus extends our earlier research presented
in Zeelenberg et al. (1996) by showing regret effects in decisions representing
more real life decision making, and by showing effects of experienced regret
on subsequent decisions.
In addition, Experiment 1 showed that regret effects can be strong.
Zeelenberg et al. (1996) showed regret effects for choices between two equally
preferred options. Experiment 1 showed that even when there is a clear prefer-
ence for a certain course of action, anticipated regret might push people toward
another course of action.
Experiments 2a, 2b, and 3 extend our earlier work by showing regret effects
in interpersonal decision making. This suggests that also other interpersonal
decisions might be influenced by regret. For example peoples decision whether
or not to contribute to a public good might be influenced by the regret they
may anticipate if they did contribute and the public good was not realized, or
if they did not contribute and the public good was not realized, because of their
decision. In contrast, decision makers might also anticipate the regret arising
from contributing when the public good would also have been realized without
their contribution. A careful study of feedback structures in such social dilemma
situations is needed to predict which behavior is regret minimizing. If it is
shown that in these situations people will be motivated to minimize regret,
then external provisions of feedback might influence choice behavior, and help
to solve social dilemmas.
The present demonstrations thus show that peoples tendency to minimize
regret can be powerful motivators while making decisions. Analogies of these
results found in real life decision making indicate that regret effects are not
just laboratory findings. How regret may influence real life decisions is illus-
trated by the success of the Dutch postal code lottery. In this lottery the winning
number is a randomly selected postal code. If your postal code is selected and
you bought a lottery ticket you can win as much as 15,700,000 Dutch Guilders
(grand prize in May, 1997; approximately $8,000,000). The lottery ticket only
costs 10 Guilders. The decision whether or not to buy a ticket might be based
on regret aversion.
Consider the following situation: your postal code has been selected, but you
did not buy a ticket. Your neighbor, however, did buy a ticket and won the
grand prize. How would you feel? Probably, feelings of regret will be present,
and extremely painful. If you consider this possibility of regret before deciding
to buy a ticket or not, it might prompt you to buy a ticket and be protected
from severe regret. This does not apply to normal lotteries, because in these
you will never learn whether you would have won if you do not play.
The organizers of the postal code lottery also recognize the power of regret.
In their advertisements they state the following: Dont you have any tickets?
76 ZEELENBERG AND BEATTIE

Then your neighbors will win everything. So make sure that you buy some
now. Moreover, since the lottery is connected to a nation wide broadcast TV
game show, the feedback is hard to avoid. This, and the intensity of the possible
regret, might explain why so many people play this lottery.
Our present and earlier research showed effects of regret in decision making.
It is important to note that the effects of regret are not easily extended to other
emotions, such as disappointment. Regret and disappointment have in common
the fact that they are experienced when the outcome of a decision is unfavorable.
They both arise from thoughts about what would have been, had things
been different. There is virtually no empirical research focusing on effects of
anticipated disappointment on decision making. However, recent research on
the experience of disappointment (van Dijk & van der Pligt, 1997) suggests
that the findings for regret cannot easily be generalized to disappointment.
Other recent research focused on differences between regret and disappoint-
ment and showed that two emotions are different in many respects. They have
different antecedents, that is regret arises from comparing an obtained outcome
with a better outcome that might have occurred had a different choice been
made, whereas disappointment arises from comparing an obtained outcome
with a better outcome that might have resulted from the same choice being
made (Zeelenberg et al., 1997). Regret and disappointment feel quite different
(cf., Zeelenberg et al., in press-b), and they have different effects on post deci-
sional evaluations (Inman, Dyer, & Jia, 1997). We thus argue, following van
der Pligt, Zeelenberg, van Dijk, de Vries, & Richard (1997), that it is important
to be specific about the emotion under investigation because each emotion has
its own effects on choice behavior. Therefore we think that research on the
effects of anticipated disappointment is needed, in order to come to a better
understanding of the different effects of the anticipation of different emotions
in the decision making process.

Conclusion

People are motivated to avoid or minimize post-decisional regret. This moti-

vation exerts impact on their decisions, because the possibility of future regret
is anticipated and taken into account when making decisions, and because
experienced retrospective regret promotes decisions that make this regret dis-
appear (the so-called regret management). As a result people can become risk
averse or risk seeking, depending on which of the possible choice options is
the regret minimizing option. In principle this can be considered rational,
because it protects the decision maker from some of the aversive consequences
of decision making. There might be cases, however, in which one can argue
that regret results in bad decisions or in reduced learning from experience
(because feedback is avoided), and might be considered irrational. Now that
we have established that regret does influence decision making and that feed-
back is one of its prime determinants, future theorizing could focus on the
normative status of these effects. Future research should focus more on effects
 CONSEQUENCES OF REGRET AVERSION 77

of experienced regret on choice behavior. These enterprises are required to

come to an complete understanding of the role of regret in decision making.

REFERENCES

Barry, B., & Oliver, L. (1996). Affect in dyadic negotiation: A model and propositions. Organizational
Behavior and Human Decision Processes, 67, 127143.
Bazerman, M. H., & Neale, M. A. (1992). Negotiating rationally. New York: The Free Press.
Bell, D. E. (1982). Regret in decision making under uncertainty. Operations Research, 30, 961981.
Boles, T. L., & Messick, D. M. (1995). A reverse outcome bias: The influence of multiple reference
points on the evaluation of outcomes and decisions. Organizational Behavior and Human Decision
Processes, 61, 262275.
Camerer, C., & Thaler, R. H. (1995). Ultimatums, dictators and manners. Journal of Economic
Perspectives, 2, 209219.
Guth, W., Schmittberger, R., & Schwarze, B. (1982). An experimental analysis of ultimatum
bargaining. Journal of Economic Behavior and Organization, 3, 367388.
Inman, J. J., Dyer, J. S., & Jia, J. (1997). A Generalized Utility Model of Disappointment and
Regret Effects on Post-Choice Valuation. Marketing Science, 16, 97111.
Josephs, R. A., Larrick, R. P., Steele, C. M., & Nisbett, R. E. (1992). Protecting the self from the
negative consequences of risky decisions. Journal of Personality and Social Psychology, 62, 2637.
Kahneman, D., & Tversky, A. (1997). Prospect theory: An analysis of decision under risk. Economet-
rica, 47, 263291.
Landman, J. (1993). Regret: The persistence of the possible. New York: Oxford University Press.
Larrick, R. P., & Boles, T. L. (1995). Avoiding regret in decisions with feedback: a negotiation
example. Organizational Behavior and Human Decision Processes, 63, 8797.
Loomes, G. & Sugden, R. (1982). Regret Theory: An alternative theory of rational choice under
uncertainty. Economic Journal, 92, 805824.
Pillutla, M. M., & Murnighan, J. K. (1996). Unfairness, anger, and spite: Emotional rejections of
ultimatum offers. Organizational Behavior and Human Decision Processes, 68, 208224.
Ritov, I. (1996). Probability of regret: Anticipation of uncertainty resolution in choice. Organiza-
tional Behavior and Human Decision Processes, 66, 228236.
Ritov, I., & Baron, J. (1995). Outcome knowledge, regret, and omission bias. Organizational Behav-
ior and Human Decision Processes, 64, 119127.
Savage, L. J. (1951). The theory of statistical decision. Journal of the American Statistical Associa-
tion, 46, 5567.
Shimanoff, S. B. (1984). Commonly named emotions in everyday conversations. Perceptual and
Motor Skills, 58, 514.
Thaler, R. H. (1992). The winners curse: Paradoxes and anomalies of economic life. New York:
The Free Press.
van der Pligt, J., Zeelenberg, M., van Dijk, W. W., de Vries, N. K., & Richard, R. (1997). Affect,
attitudes, and decisions: Lets be more specific. European Review of Social Psychology, 8, in press.
van Dijk, W. W., & van der Pligt, J. (1997). The impact of probability and magnitude of outcome
on disappointment and elation. Organizational Behavior and Human Decision Processes, 69, 277
284.
Zeelenberg, M. (1996). On the importance of what might have been: Psychological perspectives on
regret and decision making. Ph.D. dissertation, University of Amsterdam.
Zeelenberg, M., Beattie, J., van der Pligt, J., & de Vries, N. K. (1996). Consequences of regret
aversion: effects of expected feedback on risky decision making. Organizational Behavior and
Human Decision Processes, 65, 148158.
78 ZEELENBERG AND BEATTIE

Zeelenberg, M., van der Pligt, J., & Manstead, A. S. R. (in press-a). Undoing Regret on Dutch
Television: Apologizing for interpersonal regrets involving actions and inactions. Personality
and Social Psychology Bulletin.
Zeelenberg, M., van Dijk, W. W., Manstead, A. S. R., & van der Pligt, J. (in press-b). The experience
of regret and disappointment. Cognition and Emotion, 12.
Zeelenberg, M., van Dijk, W. W., van der Pligt, J., Manstead, A. S. R., van Empelen, P., & Reinder-
man, D. (1997). Emotional reactions to the outcomes of decisions: Counterfactual thought and
the experience of regret and disappointment. Manuscript under review.

Received: October 20, 1997