Organizational Behavior and Human Decision Processes 97 (2005) 152160

www.elsevier.com/locate/obhdp

On the psychology of if only: Regret and the comparison between

factual and counterfactual outcomes
Eric van Dijk a,, Marcel Zeelenberg b
a

Department of Social and Organizational Psychology, Leiden University, P.O. Box 9555, 2300 RB Leiden, The Netherlands
b
Department of Economic and Social Psychology, Tilburg University, The Netherlands
Received 19 May 2004

Abstract
People experience regret when they realize that they would have been better oV had they decided diVerently. Hence, a central element in regret is the comparability of a decision outcome with the outcomes forgone. Up to now, however, the comparison process
that is so essential to the experience of regret has not been the subject of psychological research. In this article, we tune in on the comparison dependency of regret. We argue that factors that reduce the tendency to compare attenuate regret, and demonstrate that
uncertainty about counterfactual outcomes (Experiment 1), and incomparability of counterfactual and factual outcomes (Experiments 2 and 3) produce such eVects.
2005 Elsevier Inc. All rights reserved.
Keywords: Regret; Comparability; Uncertainty

Regret is a negative emotion that we experience when

we realize or imagine that our present situation would
have been better, if only we had decided diVerently. It is
a common experience that has serious behavioral implications for our day-to-day behavior. These may stem
from both the anticipation and experience of this emotion (for reviews, Connolly & Zeelenberg, 2002; Zeelenberg, 1999; Zeelenberg, Inman, & Pieters, 2001). Hence,
this emotion has attracted the attention of researchers in
diverse Welds, such as economics (Bell, 1982; Loomes &
Sugden, 1982), marketing (e.g., Inman, Dyer, & Jia,
1997), medicine (e.g., Brehaut et al., 2003), law (e.g.,
Guthrie, 1999), and in experimental (e.g., Mellers, Schwartz, & Ritov, 1999), social (e.g., Zeelenberg, van der
Pligt, & Manstead, 1998), and cross-cultural psychology
(e.g., Gilovich, Wang, Regan, & Nishina, 2003). To fully
understand regrets impact, it is important to develop

Corresponding author. Fax: +31 71 5273619.

E-mail address: Dijk@fsw.leidenuniv.nl (E. van Dijk).

0749-5978/$ - see front matter 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.obhdp.2005.04.001

our insights into the psychology of this emotion and the

processes that may moderate it.
To feel regret, one needs to run a mental simulation of
what happened and what could have happened instead,
and compare the two (Kahneman & Miller, 1986). Thus,
regret is related to counterfactual thoughts about what
could have been (Ritov, 1996; Roese, 1997), and hence
is the end result of a comparison process. Prior research,
however, has mostly neglected this comparison process,
and has paid little attention to how factual compare to
counterfactual outcomes. As a result, we argue, this
research has painted a rather incomplete picture of the
conditions that generate regret.

Fundamental diVerences between factual and

counterfactual outcomes
In the following, we draw attention to two fundamental diVerences between factual and counterfactual outcomes that we deem relevant with respect to the

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

comparison process that elicits regret. The Wrst issue pertaining to this process is that whereas people will often
be certain about the factual outcomes they obtained, the
outcomes they could have obtained will typically be surrounded with uncertainty. For example, we may know
the salary we currently earn, but we are often not certain
about the exact salary we would have obtained if only
we had studied Wnance rather than psychology. How
does this potential uncertainty about what might have
been aVect regret?
The second issue is that even if we are certain about
what we missed out on, these counterfactual outcomes
may be of a diVerent kind than the outcomes we actually obtained. What diVerence does it make whether
after buying a car, one Wnds out that a similar car was
available for $100 less, or whether a dissimilar car was
available for $100 less? How does reduced comparability between actual and factual outcomes aVect regret?
This is a basic question that taps directly into the comparison that is believed to be so central to the emotion
of regret, but to our knowledge it has not been
addressed before.
Using an experimental setup, we will argue and demonstrate that both uncertainty about counterfactual outcomes and comparability of counterfactual outcomes
with the factual outcome are factors that have a direct
impact on the emotion regret.

Experiment 1: Uncertainty about what could have been

In the Wrst experiment, we address the eVect of uncertainty about forgone outcomes. To examine this, we
investigated the eVect of uncertainty that people experience when they know the range of possible outcomes
they might have obtained had they decided diVerently,
but they do not know exactly which of these possible
outcomes it would have been. How do people deal with
such uncertainty? To the extreme, our basic proposition
is that people do not deal with it. That is, we suggest that
uncertainty about what could have been may keep people from experiencing regret. Our present reasoning is
based on research on the disjunction eVect (ShaWr &
Tversky, 1992; Tversky & ShaWr, 1992; see also Van Dijk
& Zeelenberg, 2003) that suggests that uncertainty and
ambiguity may induce people to engage in nonconsequential reasoning. That is, if people are uncertain of
which outcomes will be obtained, they are less likely to
think through the consequences of the possible outcomes.
The disjunction eVect and its relation to nonconsequential reasoning can best be illustrated by discussing
a study of Tversky and ShaWr (1992). In this scenario
study, participants had to imagine that they had taken
an exam: they either had to imagine failing the exam or
passing the exam, or that they did not know whether

153

they had failed or passed. The main dependent variable

was the willingness to purchase a vacation to Hawaii.
The results showed that both participants who had
learned that they passed the test and participants who
had learned that they failed the test were likely to purchase the vacation. Interestingly, those who were still
ignorant about their test result were unlikely to purchase the vacation. These Wndings suggest that people
may not think through the consequences of uncertain
outcomes. After all, should the participants have taken
the consequences into account they would have purchased the vacation. They would buy the vacation if
they failed, also if they passed the exam, so if they
would think through the two possible outcomes (either
you fail or you pass), they should also opt to purchase
the vacation when being ignorant about the test
outcome.
The research on the disjunction eVect thus suggests a
reluctance to base decisions on uncertain information. In
a similar vein, we suggest that people may not base their
emotions on the consequences of what could have been if
what could have been remains uncertain to them. If so,
this would imply that people are less likely to suVer from
aversive regret experiences. In our Wrst experimental
study, we put this reasoning to the test.
For this purpose, we designed a scenario study in
which participants learned that they participated in a
game in which several prizes could be won. All participants were informed that they had won a stress ball
(i.e., a little ball that you squeeze to reduce stress). In
addition, they received information about the prize
they would have won had they chosen diVerently.
Some participants read that the missed prize was a
CD of their choosing. Others read that they missed
out on a walkman, and some that they missed out on a
dinner for two. We also included a condition in which
participants were uncertain about their missed prize,
and all that they knew was that the missed prize was
either a CD of their choosing, a walkman, or a dinner
for two.
Method
Social Science students at Leiden University (47
males; 61 females; and Mage D 21.7 years) participated
voluntarily. They were randomly assigned to one condition of a 4-group design (27 participants per condition).
They read the following scenario:
Imagine that you participate in a game in which several
prizes can be won. To make the game interesting, the
prizes are hidden behind closed doors. Now it is your
turn to choose one out of two doors. Whatever will be
behind the door of your choosing, will be yours. You
pick a door, and behind this door you Wnd a stress ball.
After this, the organizers of the game show you that the

154

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

prize behind the other door was [a CD of your choosing;

walkman; dinner for two; either a CD of your choosing, or
a walkman, or a dinner for two].
After this, participants indicated to what extent they
would feel regret (1 D not at all; 9 D very much).
In pretesting the material of Experiment 1, we conducted a pilot to test whether the manipulation of
uncertainty had the intended eVects on uncertainty by
asking 52 participants (13 per condition) to what
extent they were uncertain about what prize would be
behind to other door (1 D absolutely uncertain;
9 D absolutely certain). The results of this pretest indicated that our manipulation had the intended eVect,
F (3, 48) D 12.79, p < .0001. LSD tests indicated that
participants in the Uncertain condition were signiWcantly less certain (M D 1.00) than participants in the
other three conditions, that did not diVer signiWcantly
from each other (Mdinner for two D 7.23; Mwalkman D 7.00;
and MCD D 6.15).
Results and discussion
The results of Experiment 1 are shown in Table 1. A
one-way ANOVA yielded a signiWcant eVect,
F (3, 104) D 15.35, p < .01. LSD-comparisons (p < .05)
showed that regret was signiWcantly lower for participants missing out on a CD than for participants missing
out on the more expensive prizes, a dinner for two and
walkman. Of course, the most interesting result was that,
as predicted, regret was signiWcantly lower in the Uncertain condition than in the other conditions. These Wndings corroborate the reasoning that uncertainty about
what could have been mitigates the experience of regret.
They also constitute the Wrst generalization of the disjunction eVect to emotions.
These Wndings suggest that uncertainty about what
could have been may reduce feelings of regret. If we are
uncertain about what could have been, we are less likely
to think through the consequences of what could have
been, and therefore less likely to compare what is with
what could have been. Does this mean that if we are
certain about what could have been, we will inevitably
be very susceptible to regret? In the search for a satisfactory answer to this question, it is again informative to
address the comparison process that is at the basis of
regret. This is what we did in Experiment 2.
Table 1
Mean regret ratings per condition, Experiment 1
Uncertain

Certain
CD

Dinner for two

Walkman

4.41a

5.96b

7.04c

6.89c

Note. Means with diVerent superscripts diVer signiWcantly (LSD,

p < .05). Ratings were made on a 9-point scale (1 D absolutely not;
9 D very much).

Experiment 2: Comparability of what could have been

with what is
Not all outcomes are equally easy to compare.
Sometimes comparisons are fairly easy. For example,
in the case of the investor who realizes that the shares
he sold suddenly increased in value, and that he would
have earned $10,000 more if he had not sold the shares.
But as we noted in our introduction, this example may
be more of an exception than a rule. Often, the counterfactual outcomes will be of a diVerent kind than the
factual outcomes. So, what if you know that you have
got an orange, and you also know with certainty that
you would have gotten two apples should you have
decided diVerently? Comparing apples to oranges is a
diYcult task. In more theoretical terms, this implies
that comparisons can diVer in complexity. As Medin,
Goldstone, and Markman (1995, p. 8) already
reasoned:
not all comparisons are equally easy to make; comparisons that involve substantially diVerent properties are
diYcult. It is easier, for example, to compare the merits of Mendelssohn and Schumann than to compare
the merits of Schumann and the Beatles. It may be
possible to convert both Schumann and the Beatles
into generic utilities, but this process seems to require
more work than comparing items that have similar
aspects.
If comparability lies at the heart of regret, the
(in)comparability of factual and counterfactual outcomes may be another feature explaining why we do not
constantly go about kicking ourselves over forgone outcomes. We investigated this hypothesized moderating
eVect of comparability on regret by presenting our participants a scenario in which they engaged in a lottery.
Half of the participants learned that they had won a D15
liquor store token. The others learned that they had won
a D15 book token. After this, participants were informed
that they would have won another prize had they
decided diVerently. We also manipulated the missed
prize. It was either a D50 book token or a D50 liquor
store token.
With this design, we were able to manipulate comparability (Johnson, 1984). The factual and counterfactual outcomes are comparable if the prize won and
the missed prize are in the same product category
(either both book tokens or both liquor tokens) and
relatively incomparable if they are in a diVerent
product category (i.e., you win a book token, but miss
out on a liquor token, or you win a liquor token and
miss out on a book token). Our basic hypothesis was
that regret would be more intense if the missed prize
and the prize won came from the same product category than if they came from a diVerent product
category.

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

155

Method

Experiment 3: Comparability and the need to compare

Social Science students at Leiden University (66

males; 114 females; and Mage D 21.08 years) participated
voluntarily. They were randomly assigned to one of the
four conditions of a 2 (Own prize: D15 book token vs.
D15 liquor store token) 2 (Missed prize: D50 book
token vs. D50 liquor store token) factorial design (45
participants per condition). They read the following scenario:

Experiment 2 provided support for the idea that

regret is related to outcome comparability. Another
issue to consider here is that individuals diVer in the
extent to which they compare outcomes, especially in the
situations that we have described here. Note that in
Experiment 2, participants learned that the counterfactual outcome accrued to another individual. In this case,
counterfactual thinking and social comparison go hand
in hand (Larrick, 1993; Olson, Buhrmann, & Roese,
2000).
In real life, people often learn about the outcomes forgone by comparing their own outcomes to those of others. As a result, the outcomes of others may serve as
counterfactual reference points when decision-makers
realize that it could have been them. People are very
sensitive to the outcomes of others, and decision-makers
can be especially dissatisWed when others receive a better
outcome. Previous research has shown that these social
comparison eVects can also contribute to the regret that
people may feel in response to a decision that goes awry
(Boles & Messick, 1995; Kumar, 2004; Zeelenberg &
Pieters, 2004). The observation that social comparisons
may play a role suggests an additional opportunity to
test our reasoning by studying individual diVerences.
Gibbons and Buunk (1999) recently argued that some
people are more than others likely to engage in social
comparison. They constructed a reliable and validated
scale (the IowaNetherlands comparison orientation
measure; INCOM) for measuring such individual diVerences in social comparison orientation. In Experiment 3,
we used this new insight to further explore the eVect of
comparability we revealed in Experiment 2.
The Wndings of Experiment 2 suggest that if the comparison of factual with counterfactual outcomes
becomes more diYcult (cf. Medin et al., 1995) people are
less likely to experience regret. But of course, more diYcult does not mean impossible. With a high motivation
to engage in comparison, people may be expected to
engage in comparisons, even if these turn out to be diYcult ones. Based on this reasoning, and using the same
paradigm as in Experiment 2, we expected that the eVect
of ease of comparability would be moderated by individual diVerences in the need to compare.
For our study, we administered the INCOM measure
(Gibbons & Buunk, 1999) to identify people with a high
vs. low need to compare. To investigate the eVect of comparison orientation, on regret ratings, we presented them
with the scratch card scenario. All participants read that
they had won a D15 book token. Half of them read that
they missed out on a D50 book token, whereas the other
half read that they missed out on a D50 liquor store
token. Our main hypothesis was that people with a high
need to compare would be less inXuenced by the diYculty to compare, and thus that they would suVer from

Imagine walking through your home town. You run into

a small fair. It appears that they have a lottery, with
instant scratch card lottery tickets. You doubt whether
you will buy a scratch card because they are expensive.
But since the lottery guarantees you to win a prize, you
decide to take a chance. You are just in time, there are
only two scratch cards left. You doubt which one to
take, but eventually you choose one. You open the
scratch card and as it turns out, you have won a liquor
store token [book token] for D15. After this, you notice
someone else buying the scratch card that was left, the
one you did not choose. This other person wins a book
token [liquor store token] of D50.
After this, participants were asked how much regret
they would feel (1 D not at all; 7 D very much).
Results and discussion
A 2 (Own prize) 2 (Missed prize) ANOVA on the
regret ratings only yielded a signiWcant interaction,
F (1, 176) D 11.07 p < .001. Table 2 shows that, in agreement with the notion that comparability is at the basis
of regret, and that incomparability may shield people
from the experience of regret, participants winning a
book token reported less regret when missing out on the
D50 liquor store token (M D 3.91) than when missing
out on the D50 book token (M D 4.71), F (1, 176) D 4.37,
p < .05. Participants winning a liquor store token
reported less regret when missing out on the D50 book
token (M D 3.64) than when missing out on the D50
liquor store token (M D 4.64), F (1, 176) D 6.83, p < .01.
Thus, more regret is anticipated when the obtained outcome and the missed outcome come from the same category as compared to when they come from diVerent
categories.

Table 2
Mean regret ratings per condition, Experiment 2
Own prize

Book token
Liquor store token

Missed prize
Book token

Liquor store token

4.71
3.64

3.91
4.64

Note. Ratings were made on a 7-point scale (1 D absolutely not;

7 D very much).

156

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

regret when Wnding out that someone else had obtained

higher outcomes, even if these outcomes were in a diVerent product category than the prize that they themselves
obtained.

Table 3
Mean regret ratings per condition, Experiment 3
Comparison
orientation

Missed prize
Same category
(book token)

DiVerent category
(liquor store token)

Method

Low
High

6.30
6.42

4.60
6.45

First-year Social Science students at Tilburg University (11 males, 63 females; and Mage D 19.42 years) participated voluntarily. They were randomly assigned to
one of the four conditions of (Missed prize: D50 book
token vs. D50 liquor store token) 2 (comparison orientation: low vs. high) factorial design.
In the beginning of the academic year, we administered the IowaNetherlands comparison orientation
measure (INCOM; Gibbons & Buunk, 1999) to a group
of Wrst-year students (
D 0.69). The INCOM consists of
11 items that measure the comparison orientation (e.g.,
I always pay a lot of attention to how I do things compared with how others do things; I always like to know
what others in a similar situation would do).
Eight months later, we contacted the students whose
score fell in the highest 30-percentile range of the social
comparison scores (i.e., those with a relatively high need to
compare) and the students whose score fell in the lowest
30-percentile range (i.e., those with a relatively low need to
compare). We asked them to participate in a study and we
presented them with the scratch card scenario.
Participants read the scenario of Experiment 2, in
which now all participants read that they had won a
book token for D15. As in Experiment 2, participants
read that the remaining scratch card was bought by
someone else, and that this other person won a book
token [liquor store token] of D50. After this, participants
were asked how much regret they would feel (1 D not at
all; 9 D very much).

Note. Ratings were made on a 9-point scale (1 D absolutely not;

9 D very much).

Results and discussion

Regret is rooted in a comparison of actual decision

outcomes with counterfactual outcomes. In this article,
we used this observation to gain more insight into the
comparison process underlying regret. In particular, we
were able to demonstrate that vulnerability to regret is
moderated by the uncertainty people may experience
regarding counterfactual outcomes, and the comparability of counterfactual outcomes with factual outcomes.
The more fundamental contribution to regret research is
that these Wndings bear directly on the comparison that
underlies regret.
We are not the Wrst to theorize on the role of potential
moderators regarding counterfactual outcomes. For
example, Seelau, Seelau, Wells, and Windschitl (1995)
argued that people do not consider all counterfactual
outcomes and maintained that some counterfactual outcomes may be less available in memory and appear less
lucid to people. The results of Experiment 1 add to this
literature by investigating the moderating eVect of uncer-

A 2 (Missed prize) 2 (comparison orientation)

ANOVA on the regret ratings yielded a signiWcant main
eVect of comparison orientation, F (1, 70) D 4.91, p < .05,
a marginal signiWcant main eVect of missed prize,
F (1, 70) D 3.77, p < .06, and a signiWcant interaction,
F (1, 70) D 4.00, p < .05.
The main eVect of comparison orientation indicated
that participants with a high need to compare (M D 6.44)
reported more regret than participants with a low need
to compare (M D 5.37). The marginally signiWcant main
eVect of missed prize indicated that participants tended
to report less regret when the missed prize was from a
diVerent product category as the obtained prize (i.e., a
liquor store token; M D 5.55) than when it was from the
same category (i.e., a book token; M D 6.38).
In agreement with our main hypothesis, these two
main eVects were qualiWed by the signiWcant interaction.

Participants with a relative low need for comparison

reported less regret when the missed prize was from a
diVerent product category as the obtained prize (i.e., a
liquor store token; M D 4.65) than when it was from the
same category (i.e., a book token; M D 6.30,
F (1, 70) D 7.35, p < .010). Consistent with our reasoning,
participants with a high need for comparison gave high
regret ratings regardless of the product category of the
missed prize (Mbook token D 6.42; Mliquor store token D 6.45),
F (1, 70) D .002, n.s., see Table 3.
These Wndings are not only important because they
corroborate the view that social comparison processes
may add to regret (cf. Boles & Messick, 1995; Kumar,
2004; Zeelenberg & Pieters, 2004). For the current purpose, these Wndings are of special importance because
they support our notion that to understand the psychological process leading up to regret, it is useful to explicitly investigate the comparison process that lies at the
basis of regret. In particular, the Wndings suggest a motivational dependency of the eVect of comparability we
documented in Experiment 2, in the sense that diYculty
to compare may be overcome if the need for comparison
is high.

General discussion

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

tainty, and introducing a novel connection between counterfactual thinking and nonconsequential reasoning.
Prior research on counterfactual reasoning implicitly
assumed that if people generate counterfactuals (either
because they generate these themselves or because they
are oVered some counterfactuals by the researcher), these
counterfactual outcomes are consequential in the sense
that they will be entered as an input in the comparison of
what is with what could have been. The literature on nonconsequential reasoning (ShaWr & Tversky, 1992; Van
Dijk & Zeelenberg, 2003) stresses that people do not
always engage in consequential reasoning, and that one
crucial aspect to consider is the (un)certainty people may
experience regarding outcomes. The current Wndings suggest that the fact that people may not think through the
consequences of uncertain counterfactual outcomes may
be a blessing because it may protect people from the aversive feeling of regret.
The second and third experiments on the comparability of counterfactual and factual outcomes illustrate a
similar process in the sense that people may be less likely
to engage in comparative reasoning if comparisons are
more diYcult to make. In this article, we show how the
diYculty to compare may aVect feelings of regret. In particular, the Wndings suggest that if outcomes are diYcult
to compare, people may be less likely to experience
regret. As Experiment 3 shows, however, this regretreducing aspect of the diYculty to compare can be overcome if the motivation to compare is high.
On a more general level, our insights can be related to
Tversky and GriYns (1991) contrast and endowment
model of well-being that states that current well-being is
not only dependent on current outcomes, but also on
past experiences. The eVect of endowment represents a
direct eVect of past outcomes: positive experiences make
us happy and negative experiences make us unhappy.
The contrast eVect is more indirect. Satisfaction with
current outcomes may be increased if the outcomes are
preceded by a negative experience, because people may
contrast their current outcome with their negative experience. Satisfaction with current outcomes may be
decreased if the outcomes are preceded by a positive
experience, because people may contrast their current
outcome with their positive experience. This contrast
eVect is fueled by comparison. In this respect, it is noteworthy that Tversky and GriYn also reasoned that comparability of the past with the present may moderate the
contrast eVect. Contrast diminishes as the comparability
of the past with the present diminishes: a bad meal at a
Chinese restaurant has less eVect on our reactions to a
subsequent meal if we enjoy that meal in a French restaurant than if we enjoy it in a Chinese restaurant. The
current Wndings are not only supportive of that reasoning, they also suggest that the basic reasoning extends
beyond the temporal context of past versus current outcomes and well-being.

157

A connection can also be made to Slovic, GriYn, and

Tverskys (1990) theorizing on the concept of compatibility. Slovic et al. investigated to what extent the compatibility between stimulus and response scale aVects the
weight that people assign to a stimulus attribute. Their
basic prediction was that the weight of a stimulus attribute is greater when it matches the response scale than
when it does not. To illustrate: in their Study 2, Slovic et
al. had participants predict the performance of 10 target
students in a history course on the basis of the students
performance in two other courses (English literature and
philosophy). For each of the target students, participants
were given a letter grade (from A+ to D) in one course
and a class rank (from 1 to 100) in the other course. Participants then had to predict how the students would
perform in a history class. Half of the participants had to
predict the students grade. The other half of the participants had to predict the students class rank. The results
indicated that participants who had to predict the grade
put more weight on grade information, and participants
who had to predict class rank put more weight on class
rank. The connection between this theorizing on compatibility and our theorizing on comparability is that in
both cases it is assumed that comparisons that are more
easy to make receive more weight in decision-making.
Slovic et al., however, restricted their analysis to the relation between stimulus and response scale, whereas we
concentrate on the comparison process of two stimuli
(the comparison process that aVects the regret
responses).
Another relation with comparability can be found in
Hsees theorizing on evaluability (e.g., Hsee, Loewenstein, Blount, & Bazerman, 1999; Hsee & Zhang, 2004).
A central theme in this theorizing is that some attributes
are diYcult to evaluate, and that attributes that are diYcult to evaluate have less impact on decision-making.
With its focus on the ease of evaluation, the evaluability
hypothesis shares some resemblance to our theorizing on
the ease of comparing factual to counterfactual outcomes. Of course, an important diVerence is that the
evaluability hypothesis mainly pertains to the evaluation
of outcomes people obtain. The central issue in our studies is not whether outcomes are easy or diYcult to evaluate, but whether factual and counterfactual outcomes
are easy or diYcult to compare.
Another issue worthy of discussion is that in our scenario studies, regret ratings primarily reXected participants predictions on how they would feel. Such
predictions of anticipated regret are essential for decision-making (Bell, 1982; Loomes & Sugden, 1982),
because decision-makers will try to minimize anticipated
regret. Nevertheless, it should be acknowledged that predictions regarding anticipated emotions need not be
accurate. For example, Gilbert and Ebert (2002) stated
that when people anticipate how they will feel, they
insuYciently anticipate the post decisional processes

158

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

such as dissonance reduction, self-deception, ego

defense, emotion-based coping, and so forth. In a recent
article, Gilbert, Morewedge, Risen, and Wilson (2004)
found that people may overestimate the regret in the
sense that experienced regret may be lower than anticipated regret. The main explanation for this Wnding was
that after a regretful experience, people may eYciently
avoid self-blame. Because people do not anticipate this
psychological process when they have to predict how
they will feel, people overestimate regret. As Gilbert et al.
put it: regret can be a bit of a boogeyman, looming
larger in prospect than it actually stands in experience
(p. 349). With these Wndings, Gilbert et al. did not mean
to downplay the importance of anticipated regret for
decision-making, and even noted that peoples decisions
are often based on their beliefs about how they would
feel (e.g., anticipated regret). Gilbert et al.s Wndings do
suggest that it is worthwhile in future research to complement the current Wndings on anticipated regret with
assessments of experienced regret. That anticipations of
regret need not necessarily be inaccurate was demonstrated by Mellers et al. (1999) in their Experiment 4. In
this experiment, they examined the resemblance between
anticipated and actual emotions, and found that participants were good at predicting their feelings. Their predictions were not perfect; they did not seem to anticipate
surprise, but they did anticipate the disappointment and
regret they later experienced (p. 341). Interestingly, they
next speculated that prediction may especially be accurate in simple situations, as found in our paradigm. (p.
341). Mellers et al. studied choices between two alternatives, which in terms of complexity seems similar to our
studies in which we informed participants about
obtained and missed outcomes.
It may be noted that in our studies we used a singleitem measurement of regret that was also used in previous research on regret and decision making (e.g., Arkes,
Kung, & Hutzel, 2002; Crawford, McConnell, Lewis, &
Sherman, 2002; Kray, 2000; Kumar, 2004; Ordez &
Connolly, 2000; Zeelenberg & Pieters, 2004; Zeelenberg,
van den Bos, van Dijk, & Pieters, 2002; Zeelenberg, van
Dijk, & Manstead, 2000). We thus stayed close to the
methodology used in other experimental research to
measure regret. One may wonder, however, whether we
would also have obtained these Wndings if we would
have used multiple items to measure regret. To address
this question, we decided to test whether the results of
Experiment 1 would be replicated if we would measure
regret with more items. In this replication of Experiment
1 on 100 additional participants, we asked our participants Wve questions: how much regret would you feel
(i.e., the item we included in our studies), how bad
would you consider your decision, to what extent
would you feel that you should have chosen the other
door, would you prefer to have chosen the other door,
and to what extent would you regret your decision, all

to be answered on 9-point scales (1 D not at all; 9 D very

much). These measures were combined to form a reliable
regret scale (
D 0.82). With this scale, the basic
Wndings of Experiment 1 were replicated. The regret
measure was signiWcantly aVected by our manipulations
(F (3, 96) D 4.79, p < .01). LSD-comparisons (p < .05)
again showed that regret was signiWcantly lower in the
Uncertain condition (M D 5.02) than in the other conditions
(Mdinner for two D 6.17;
Mwalkman D 6.54;
and
MCD D 6.10). This suggests that the Wndings we reported
are not restricted to cases of single-item measurement.
At this point, it is also appropriate to discuss the possible limitations of using scenario studies to measure
(anticipated) regret. Although scenario studies are often
used in research, some authors stress that one should be
careful and acknowledge the potential limitations of scenario studies. One of the most stringent positions in this
regard was advocated by Spencer (1978), who stated that
hypothetical role-playing should only be used to assess
demand characteristics. On the other hand, Connolly,
Ordez, and Coughlan (1997, p. 83) argued with regard
to anticipated regret ratings, that scenario studies adequately measure what they should measure: if ones
interest in the area is in the possible decisional impact of
anticipated regret. Of special interest are the arguments
provided by Greenberg and Eskew (1993). Their basic
message was not that scenario studies are generally inappropriate or appropriate. Rather, they made some interesting recommendations on how scenario studies might
look like, based on what should be considered to be the
basic purpose of the studies. Restricting their arguments
to organizational behavior, Greenberg and Eskew distinguished between two major purposes (p. 225): (a)
describing the attitudes and/or behaviors of people in an
organizational setting and (b) examining the basic
human processes of perception, judgment, or cognition.
Our current research Wts best with the latter purpose,
and it is for this purpose that the authors recommend
that participants should have a low level of involvement
and that responses should be limited. In this respect, they
gave an example of a study on decision-making by
Olshavsky (1979) in which participants had to imagine
that they were going on a ski trip to Aspen, or that they
were renting a stereo receiver. Information was limited,
and this setup was described by Greenberg and Eskew
not as a problem, but even as a virtue (p. 237). In our
opinion, this setup very much resembles our setup, and
deWnitely our main purpose, that is to reveal basic information about the fundamental comparison process that
underlies regret: the fact that counterfactuals regarding
what could have been are often surrounded with
uncertainty, and often of a diVerent kind than what is.
Taken together, our studies show the beneWts a process oriented approach for the understanding of the psychology of if only. In particular, they indicate that the
consequences of what could have been for how we feel

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

may be limited if what could have been is of a diVerent

kind than what is, either because the counterfactual
outcomes are uncertain, or because they are diYcult to
compare with our factual outcomes. We thus argue that
both aspectsuncertainty and comparabilitymay
each attenuate feelings of regret because they both aVect
the tendency to compare factual to counterfactual outcomes. It may also be noted, that in our Experiment 1 on
the eVect of uncertainty, we used products that in terms
of Experiments 2 and 3 could be described as less comparable (i.e., CDs, dinner for two, walkman, and stress
ball). This raises the question of whether uncertainty
eVects would also be obtained in situations of greater
comparability. For example, what if one would learn
that one obtained 1 CD, and the missed prizes were 20
CDs, 30 CDs, 40 CDs, or uncertain (i.e., either 20, 30, or
40)? One could argue that in such a setting uncertainty
may not have a strong dampening eVect on regret ratings
because people may reason that even though it is uncertain whether one misses out on 20, 30, or 40 CDs, it is
certain that one at least misses out on 20 CDs.
A recent study by Van Dijk and Zeelenberg (2003) on
eVects of uncertainty on economic decision-making suggests, however, that even under such conditions the disjunction eVect may operate. For example, in one of these
studies it was investigated to what extent uncertainty
would aVect the likelihood of people to fall prey to the
sunk cost eVect (Arkes & Blumer, 1985; i.e., the tendency
for increased risk-taking after having incurred sunk
costs). To study this, participants were presented a typical sunk cost scenario and either learned that the sunk
costs were low (500,000 Guilders), high (1.5 million Guilders), or that the size of the sunk costs were uncertain
(i.e., at the minimum 500,000 Guilders and 1.5 million
Guilders at the maximum). The results indicated that
participants fell prey to the sunk cost eVect in the low
sunk costs condition, in the high sunk costs condition,
but not in the uncertain sunk costs condition. That is, in
line with the basic tenet of the disjunction eVect, participants did not base their decisions on the uncertain information, even though in this case too, participants in the
uncertain conditions could have reasoned that the costs
would in any case be at least 500,000 Guilders. These
Wndings suggest that the disjunction eVect also operates
when the outcomes involved under uncertainty are easy
to compare.
In a way, our Wndings may set the record straight on
the experience of regret. The formal deWnition of regret
(i.e., regret results from an unfavorable comparison of
what is with what could have been) would lead one
to expect that the human kind is constantly haunted by
feelings of regret because there will always be inWnite
ways in which we might have obtained higher outcomes
if only we had decided diVerently. The fact that many of
these counterfactual outcomes are surrounded with
uncertainty and/or are of a diVerent kind, may explain

159

why we manage to be happy with what we have got

rather than dwell on what we missed. The above also
suggests that there may be a functional side to the matter: uncertainty and incomparability may function as a
protective shield, attenuating excessive feelings of regret.
In this respect, we believe that, in addition to cognitive
explanations that focus on the cognitive complexity of
comparing dissimilar outcomes (e.g., Medin et al., 1995)
and thinking through uncertain situations (e.g., ShaWr,
1994; ShaWr & Tversky, 1992), there may be a motivational component at work. By excluding the unknown
and the incomparable, we may be better able to experience life without the constant nagging feeling of regret.

References
Arkes, H. R., & Blumer, C. (1985). The psychology of sunk costs. Organizational Behavior and Human Decision Processes, 35, 124140.
Arkes, H. R., Kung, Y.-H., & Hutzel, L. (2002). Regret, valuation, and
inaction inertia. Organizational Behavior and Human Decision Processes, 87, 371385.
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
inXuence of multiple reference points on the evaluation of outcomes and decisions. Organizational Behavior and Human Decision
Processes, 61, 262275.
Brehaut, J. C., OConnor, A. M., Wood, T. J., Hack, T. F., SiminoV, L.,
Gordon, E., et al. (2003). Validation of a decision regret scale. Medical Decision Making, 23, 281292.
Crawford, M. T., McConnell, A. R., Lewis, A. C., & Sherman, S. J.
(2002). Reactance, compliance, and anticipated regret. Journal of
Experimental Social Psychology, 38, 5663.
Connolly, T., Ordez, L. D., & Coughlan, R. (1997). Regret and
responsibility in the evaluation of decision outcomes. Organizational Behavior and Human Decision Processes, 70, 7385.
Connolly, T., & Zeelenberg, M. (2002). Regret and decision making.
Current Directions in Psychological Science, 11, 212216.
Gibbons, F. X., & Buunk, B. P. (1999). Individual diVerences in social
comparison: Development and validation of a measure of social
comparison orientation. Journal of Personality and Social Psychology, 76, 129142.
Gilbert, D. T., & Ebert, J. E. J. (2002). Decisions and revisions: The
aVective forecasting of changeable outcomes. Journal of Personality
and Social Psychology, 75, 617638.
Gilbert, D. T., Morewedge, C. K., Risen, J. L., & Wilson, T. D. (2004).
Looking forward to looking backward. Psychological Science, 15,
346350.
Gilovich, T., Wang, R. X. F., Regan, D., & Nishina, S. (2003). Regrets
of action and inaction across cultures. Journal of Cross-Cultural
Psychology, 34, 6171.
Greenberg, J., & Eskew, D. E. (1993). The role of role playing in organizational research. Journal of Management, 19, 221241.
Guthrie, C. (1999). Better settle than sorry: The regret aversion theory
of litigation behavior. University of Illinois Law Review, 1999, 43
90.
Hsee, C. K., Loewenstein, G., Blount, S., & Bazerman, M. H. (1999).
Preference reversals between joint and separate evaluations of
options: A review and theoretical analysis. Psychological Bulletin,
125, 576591.
Hsee, C. K., & Zhang, J. (2004). Distinction bias: Misprediction and
mischoice due to joint evaluation. Journal of Personality and Social
Psychology, 86, 680695.

160

E. van Dijk, M. Zeelenberg / Organizational Behavior and Human Decision Processes 97 (2005) 152160

Inman, J. J., Dyer, J. S., & Jia, J. (1997). A generalized utility Model of
disappointment and regret eVects on post-choice valuation. Marketing Science, 16, 97111.
Johnson, M. D. (1984). Consumer choice strategies for comparing noncomparable alternatives. Journal of Consumer Research, 11, 741
753.
Kahneman, D., & Miller, D. T. (1986). Norm theory: Comparing reality to its alternatives. Psychological Review, 93, 136153.
Kray, L. J. (2000). Contingent weighting in self-other decision making.
Organizational Behavior and Human Decision Processes, 83, 82
106..
Kumar, P. (2004). The eVects of social comparison on inaction inertia.
Organizational Behavior and Human Decision Processes, 95, 175
185.
Larrick, R. P. (1993). Motivational factors in decision theories: The
role of self-protection. Psychological Bulletin, 113, 440450.
Loomes, G., & Sugden, R. (1982). Regret theory: An alternative theory
of rational choice under uncertainty. Economic Journal, 92, 805
824.
Medin, D. L., Goldstone, R. L., & Markman, A. B. (1995). Comparison
and choice: Relations between similarity processes and decision
processes. Psychonomic Bulletin & Review, 2, 119.
Mellers, B. A., Schwartz, A., & Ritov, I. (1999). Emotion-based choice.
Journal of Experimental Psychology: General, 128, 346361.
Olshavsky, R. W. (1979). Task complexity and contingent processing in
decision making: A replication and extension. Organizational
Behavior and Human Performance, 24, 300316.
Olson, J. M., Buhrmann, O., & Roese, N. J. (2000). Comparing comparisons: An integrative perspective on social comparison and counterfactual thinking. In J. Suls & L. Wheeler (Eds.), Handbook of
social comparison: Theory and research (pp. 379398). New York:
Plenum.
Ordez, L. D., & Connolly, T. (2000). Regret and responsibility: A
reply to Zeelenberg et al., 1998. Organizational Behavior and Human
Decision Processes, 81, 132142.
Ritov, I. (1996). Probability of regret: Anticipation of uncertainty resolution in choice. Organizational Behavior and Human Decision Processes, 66, 228236.
Roese, N. J. (1997). Counterfactual thinking. Psychological Bulletin,
121, 133148.
Seelau, E. P., Seelau, S. M., Wells, G. L., & Windschitl, P. D. (1995).
Counterfactual constraints. In N. J. Roese & J. M. Olson (Eds.),

What might have been: The social psychology of counterfactual

thinking (pp. 5779). Mahwah, NJ: Lawrence Erlbaum.
ShaWr, E. (1994). Uncertainty and the diYculty of thinking through
disjunctions. Cognition, 50, 403430.
ShaWr, E., & Tversky, A. (1992). Thinking through uncertainty: Nonconsequential reasoning and choice. Cognitive Psychology, 24, 449474.
Slovic, P., GriYn, D., & Tversky, A. (1990). Compatibility eVects in
judgment and choice. In R. M. Hogarth (Ed.), Insights in decision
making: Theory and applications (pp. 527). Chicago: University of
Chicago Press.
Spencer, C. D. (1978). Two types of role playing: Threats to internal
and external validity. American Psychologist, 33, 265268.
Tversky, A., & GriYn, D. (1991). Endowment and contrast in judgments of well-being. In R. J. Zeckhauser (Ed.), Strategy and Choice
(pp. 297318). Cambridge, MA: MIT Press.
Tversky, A., & ShaWr, E. (1992). The disjunction eVect in choice under
uncertainty. Psychological Science, 3, 305309.
Van Dijk, E., & Zeelenberg, M. (2003). The discounting of ambiguous
information in economic decision making. Journal of Behavioral
Decision Making, 16, 341352.
Zeelenberg, M. (1999). Anticipated regret, expected feedback and
behavioral decision-making. Journal of Behavioral Decision Making, 12, 93106.
Zeelenberg, M., Inman, J. J., & Pieters, R. G. M. (2001). What we do
when decisions go awry: Behavioral consequences of experienced
regret. In E. U. Weber, J. Baron, & G. Loomes (Eds.), ConXict and
tradeoVs in decision making (pp. 136155). Cambridge, MA: Cambridge University Press.
Zeelenberg, M., & Pieters, R. (2004). Consequences of regret aversion
in real life: The case of the Dutch postcode lottery. Organizational
Behavior and Human Decision Processes, 93, 155168.
Zeelenberg, M., van den Bos, K., van Dijk, E., & Pieters, R. (2002). The
inaction eVect in the psychology of regret. Journal of Personality
and Social Psychology, 82, 314327.
Zeelenberg, M., van der Pligt, J., & Manstead, A. S. R. (1998). Undoing
regret on Dutch television: Apologizing for interpersonal regrets
involving actions and inactions. Personality and Social Psychology
Bulletin, 24, 11131119.
Zeelenberg, M., van Dijk, W. W, & Manstead, A. S. R. (2000). Regret
and responsibility resolved? Evaluating Ordez and Connollys
(2000) conclusions. Organizational Behavior and Human Decision
Processes, 81, 143154.