Running Head: DECISION JUSTIFIABILITY AND ANTICIPATED REGRET
Decision Justifiability and Anticipated Regret
Jochen Reb
Terry Connolly
University of Arizona
April 18, 2005
Please address correspondence to:
Jochen Reb
Eller College of Management
University of Arizona
P.O. Box 210108
Tucson, AZ 85721
(520) 621-1911
(520) 621-4171 (fax)
jmreb@email.arizona.edu
Acknowledgment: This work was supported in part by Grant # F496200031037 from the Air
Force Office of Scientific Research.
Decision Justifiability and Regret
ABSTRACT
We studied the role of decision justifiability in determining the intensity of anticipated
regret (cf. Connolly & Zeelenberg, 2002). In three experiments we examined whether perceived
justifiability can account for the effects of normality, a factor previously thought to affect
anticipated regret through ease of counterfactual mutability (cf. Kahneman & Miller, 1986), and
decision process quality, a previously unstudied factor. Study 1 found that perceived justifiability
could account for the effects on anticipated regret of action (whether the decision was to act or
not) and social normality (whether or not the decision matched those of friends and
acquaintances). Study 2 showed that following a careful decision process reduced anticipated
regret over a bad outcome and that perceived justifiability mediated this effect. Study 3
replicated the effect of decision process quality on anticipated regret through perceived
justifiability in the presence of both action and social normality information.
KEYWORDS: Decision Making; Decision Process Quality; Justifiability; Anticipated Regret
2
Decision Justifiability and Regret
DECISION JUSTIFIABILITY AND ANTICIPATED REGRET
Important general theories of human choice and behavior, such as utility theory or the
theory of planned behavior, have typically conceptualized people’s decisions as the result of
cognitions and calculations, such as probability and value estimates, at the expense of emotions.
While philosophers have long speculated about the role of emotions in decision making -- and
often come to the conclusion that emotions interfere with rational choice -- careful investigation
of the effects of emotions on decision and of decisions on emotions is quite recent. A variety of
emotions has been examined in this context, and decision-related regret has attracted particular
attention (e.g., Savage, 1951; Bell, 1982; Landman, 1993). Numerous studies have shown that
anticipated regret affects choice. Specifically, individuals often choose the option that reduces
expected regret. Such regret avoidance has been found in numerous domains including health
decisions (e.g., Connolly & Reb, 2003; Wroe, Turner, & Salkovskis, 2004), sexual behavior
(Richard, van der Pligt, & de Vries, 1996), consumer behavior (Simonson, 1992), and
negotiation (Larrick & Boles, 1995). For example, in a study conducted in New Zealand, Wroe
et al. asked women in the third trimester of their pregnancy to predict (a) the regret they would
experience if they did not immunize their child against six major diseases and the child
experienced the disease; and (b) the regret they would experience if they did immunize their
child and it experienced serious immunization side-effects. They found that these anticipated
regret ratings strongly predicted both intention to vaccinate and actual vaccination at the time the
babies were 8-10 weeks old.
Given the role that anticipated regret plays in predicting choices we need to understand
better why decision makers anticipate more regret for some options than for others. The present
paper examines one potentially important determinant of the intensity of anticipated regret: the
3
Decision Justifiability and Regret
justifiability of the decision. According to decision justification theory regret after a bad outcome
is reduced if the decision maker considers the decision to have been justifiable (Connolly &
Zeelenberg, 2002; see also Connolly & Reb, in press). The theory further predicts that decision
makers will, at the predecisional stage, anticipate less regret for a decision to the extent that they
consider the decision justifiable. While many past studies are consistent with this hypothesis (for
a review see Connolly & Zeelenberg, 2002), hardly any research has explicitly been designed to
examine the relation between anticipated regret and perceived justifiability. We examine here
two important aspects of the role of decision justifiability in determining anticipated regret. First,
we examine whether perceived justifiability can account for the previously established effect of
normality on regret (Kahneman & Miller, 1986). Second, we predict that decision makers
anticipate lower regret for a careful decision process and that is because they perceive such a
process as more justifiable.
Determinants of Anticipated Regret Intensity
Previous research has identified three major factors that affect the intensity of
experienced regret, anticipated regret, or both. We focus here on anticipated regret though the
common finding (see Zeelenberg, 1996) is that anticipated and experienced regret match each
other relatively closely and are affected by the same factors (though perhaps not with equal
intensity, cf. Gilbert, Morewedge, Risen, & Wilson, 2004).
Relative Outcome Severity
First, the intensity of anticipated (and experienced) regret is thought to be influenced by
relative outcome severity. Regret theory (Bell, 1982; Loomes & Sugden, 1982) argues that
anticipated (and experienced) regret is a function of how bad (severe) the outcome of the option
chosen is relative to some comparison standard, typically the outcome that would have been
4
Decision Justifiability and Regret
received had one chosen a different option. The more severe the expected outcome for an option
relative to a foregone outcome, the higher the anticipated regret (cf. Savage, 1951; Mellers,
Schwartz, & Ritov, 1999). The idea makes immediate intuitive sense: Consider your anticipated
regret for missing out on $100 versus $1 million. Clearly, the latter would be higher because of
the higher relative outcome severity.
Mutability
Second, the intensity of experienced and, therefore, anticipated regret is thought to
affected by mutability (Kahneman & Miller, 1986). In a landmark paper, Kahneman and Tversky
(1982) showed that regret intensity depends on whether the option chosen involves action or
inaction. Holding outcome severity constant, they found in a scenario study that people thought
an investor who switched stocks (and lost money) would experience more regret than an investor
who did not switch stocks (and lost the same amount of money), the so-called action effect.
Kahneman and Tversky argued that it is easier to counterfactually imagine that one had not
switched stocks than it is that one had switched stocks. Further, they argued that the ease of
availability of such upward counterfactual thoughts leads to more intense regret. (For an
introduction to the counterfactual thinking literature see, e.g., Roese, 1997; Roese & Olsen,
1995.) Findings such as this led to the development of norm theory (Kahneman & Miller, 1986),
which argues that a bad decision outcome leads to more regret (more generally, emotional
amplification) when the decision was abnormal, because abnormality stimulates counterfactual
thoughts of what could have been. The counterfactual “I could have done the usual thing” comes
immediately to mind.
Consistent with the mutability account of regret intensity, regret has been found to be
judged more intense after less normal behaviors such as actions, as compared to inactions
5
Decision Justifiability and Regret
(Landman, 1987), and personally unusual behaviors as compared to personally normal behaviors
(taking a different driving route to work than usual: Kahneman & Miller, 1986; making a
decision inconsistent with one’s personality: Seta, McElroy, & Seta, 2001). However, some
subsequent research showed results inconsistent with the mutability account. For example,
N’gbala and Branscombe (1997) found an action effect only in a within-subjects design, not in a
between-subjects design. They argued plausibly that norm theory (Kahneman & Miller, 1986)
would have predicted just the opposite: a stronger effect in the between-subjects design. In the
within-subjects case, the argument that counterfactuals are more easily available following
abnormal decisions is moot because both action (abnormal) and inaction (normal) alternatives
are explicitly given to the participants as part of the experimental materials.
Decision Justifiability
Findings such as these together with theoretical arguments (e.g., Janis & Mann, 1977;
Sugden, 1985) have led to the identification of a third potential influence on the intensity of
anticipated regret: the justifiability of the decision (Connolly & Zeelenberg, 2002; Connolly &
Reb, in press). The presumed role of justifiability is central to decision justification theory
(Connolly & Zeelenberg, 2002). According to this theory, there are two major components of
regret: regret associated with the outcome being (comparatively) poor, and regret associated with
the decision itself or the process that led up to it being poor. Self-blame is an essential ingredient
of this second component of regret: the less justifiable the decision is perceived to be, the more
intense will be regret after a bad outcome.
Connolly and Zeelenberg (2002) have shown that differences in justifiability perceptions
can – post hoc – account for a variety of past findings. For example, Zeelenberg, van den Bos,
van Dijk, and Pieters (2001) showed that a soccer coach who acts and changes the team after
6
Decision Justifiability and Regret
winning in the past is thought to experience more regret following a subsequent loss than a coach
who does not act (does not change the team). However, the same choice is thought to lead to less
intense regret when it takes place after a previous loss. The finding that a win-stay lose-switch
strategy can protect from regret is clearly consistent with the argument that making a more
justifiable decision leads to less regret (It may well, of course, be consistent also with a normality
argument). Crawford, McConnell, Lewis, and Sherman (2002) found that experimental
participants following the (bad) advice of an unknown stranger as to which of two football teams
to bet on experienced more intense regret when their team lost than did those who ignored the
advice. Apparently accepting advice from strangers, though plausible at the time, appeared
unjustified (and thus blame-worthy) in retrospect. In a consumer preference study (Simonson,
1992), participants primed to think about regret chose the safer option (a SONY, a highly
reputed, name-brand product) over the riskier option (a cheaper, no-name product) more often
than did participants in a control condition. The SONY reputation, even if costly, appears to
justify the choice and inoculate against self-blame. In a student-oriented version of the twoinvestors problem, Connolly, Ordóñez, & Coughlan (1997) showed that students displaced by
computer reassignment from a good course section into a poorer one were expected to feel
significant regret, though not as much as other students who brought this misfortune on
themselves. The first group appears to suffer only outcome regret; the second group experiences
self-blame regret as well.
In addition to these post-hoc accounts of earlier findings in terms of decision justification
theory at least one study explicitly tested if the intensity of regret depends on whether the
decision makers made a choice based on good or bad reasons. In this study Inman and
Zeelenberg (2002), following closely the win-stay lose-change idea (Zeelenberg et al., 2001),
7
Decision Justifiability and Regret
presented participants with scenarios in which consumers either re-purchased a product or
service or switched after previous positive or negative experience with the product or service.
Consistent with decision justification theory and the earlier findings of Zeelenberg et al. (2001)
respondents judged that a re-purchasing a product after past bad experience with it led to more
regret than switching to another product. Similarly, switching to another product from a product
after past good experience with it also led to more regret than re-purchasing the same product. In
a study of related anticipated emotions such as guilt and sadness, Baron (1992) found a similar
influence of reasons. Specifically, respondents who first rated their anticipated emotions in
response to a scenario involving two options and were then presented with a good argument
favoring one option over another tended to change their anticipated emotions in the direction
favored by the argument.
While these findings and post-hoc accounts are suggestive, more research is needed that
explicitly tests the justifiability account and decision justification theory. Such research should
(1) examine whether the effects of variables previously found to influence the intensity of
anticipated regret can be accounted for by justifiability perceptions; and (2) identify new
decision-related variables that could be expected to affect decision justifiability and test whether
these variables also affect anticipated regret. The first study is designed to address the former
issue.
STUDY 1: NORMALITY, JUSTIFIABILITY, AND ANTICIPATED REGRET
As we have shown past research found that the normality of a choice can affect regret:
abnormal choices and actions appear to lead to more intense anticipated regret than do normal
ones (Kahneman & Miller, 1986; Kahneman & Tversky, 1982). This has been explained in norm
8
Decision Justifiability and Regret
theory terms (Kahneman & Miller, 1986) as due to the easier mutability of abnormal choices and
behaviors leading to easier availability of counterfactuals and, thus, more intense regret. Study 1
examines an alternative explanation of the effect of normality on regret intensity. This
explanation proposes that abnormal choices lead to more intense regret when they are perceived
as less justifiable. In other words, we argue that justifiability perceptions mediate any effects on
anticipated regret of factors that manipulate normality.
In the present research we use a decision context in which anticipated regret has been
found to predict real and important decisions and that has been the topic of past research: a
parent’s decision whether or not to vaccinate her or his child. In the study of Wroe et al. (2004)
described above, the authors compared different potential predictors of actual immunization
decisions and found that “anticipated regret … was the strongest predictor of likelihood of
immunizing the child” (p. 38), predicting 57% of the variance (demographic variables, in
contrast, predicted only 1% of the variance). They also reported that most parents decided to
vaccinate their children.
The vaccination decision has also been the context of a number of studies examining a
possible omission bias, a tendency to prefer inaction over action, in health-related decisions, a
bias possibly resulting from increased anticipated regret for action (e.g., Ritov & Baron, 1990;
Ritov & Baron, 1995). We are not concerned here with the debate about the existence of this bias
(the interested reader can refer to the paper cited above as well as Connolly & Reb, 2003; Baron
& Ritov, 2004). We use this context because anticipated regret has been shown to be a highly
significant predictor of actual choices in this context and because the context allowed us to
construct different (and, to us, plausible) versions of the scenario in which the parent behaves
9
Decision Justifiability and Regret
either normally or abnormally, using either a careful or a careless decision process, and decides
either to take action or not.
We implemented two normality manipulations. First, we varied whether the decision
maker decides to do what her friends and family do (social normality). Second, we manipulated
whether the decision maker decides to vaccinate or not (action). Past research suggests that in
this context the active option (i.e., vaccinating) is very common and strongly preferred, at least in
the US population (Connolly & Reb, 2003). Thus, if anything, we expect a reversal of the action
effect such that action will be regretted less than inaction if bad outcomes eventuate.
As part of our research strategy, we manipulated social normality both within- and
between-subjects. We expected that the respondents’ perceptions of the justifiability of making a
given choice will be amplified when respondents are given the opportunity to contrast it with
another decision. Whether a decision conforms to what one’s friends and family do (“socially
normal”) may not be seen as more justifiable (and, therefore, leading to less anticipated regret)
when evaluated in isolation. However, when explicitly contrasted with a socially abnormal
decision leading to the same outcome, decision makers may perceive the socially normal choice
as more justifiable (and, therefore, less regrettable). We manipulated action only betweensubjects because, based on previous findings showing the strong preference and general
acceptance of vaccination as the normal choice in the population (Connolly & Reb, 2003; Wroe
et al., 2004) we expected vaccinating (i.e., acting) to provide a strong enough justification to
show in a between-subjects manipulation.
Importantly, we predicted that the effects of action and social normality on anticipated
regret would only occur when the manipulations also affect justifiability perceptions in a pattern
consistent with the regret effects (i.e., lower justifiability associated with higher anticipated
10
Decision Justifiability and Regret
regret). In the absence of experimental effects on justifiability perceptions, we did not expect any
experimental effects on anticipated regret. Further, we expected that perceived justifiability
would predict anticipated regret.
Method
Design and Procedure
Participants read a scenario in which two protagonists each face the decision whether or
not to vaccinate their small child. Both protagonists decide either to act or not. One chooses to do
what most of her family and friends did when faced with a similar decision, the other chooses the
opposite. The decision is always followed by the same bad outcome. Thus, the experiment
manipulated two factors each with 2 levels: action (vaccinate vs. don’t vaccinate) and social
normality (socially normal decision vs. socially abnormal decision). Action was manipulated
between-subjects only, whereas social normality was manipulated both between-subjects and
within-subjects, as described below.
Participants read on the first page of the experimental materials a scenario involving two
mothers who each had to decide whether or not to vaccinate her child against a flu. The scenario
explained that the risk of an unvaccinated child getting the flu was about equal to the risk of a
vaccinated child getting equally serious vaccine side-effects. On the next page, respondents read
about the decision of the first mother, in which action and social normality information was
manipulated. In all conditions, the outcome of this decision was equally bad: the child “was very
sick and unhappy for almost three weeks”, as a result either of the flu or of vaccine side-effects.
After reading about the decision and the outcome, participants rated how much regret they would
feel if they were in the protagonist’s place.
11
Decision Justifiability and Regret
On the next page, the decision of the second mother was described. She differed from the
first only in that the social normality of her choice was opposite that of the first mother; action or
inaction stayed the same. Thus, social normality was manipulated within-subjects at this point.
Order was counterbalanced so that for about half the participants the first mother was described
as having made the socially normal decision, while for the other participants she made the
socially abnormal decision. The outcomes of both decisions were equally bad. Participants again
rated the intensity of regret they would feel. On the next and final page, respondents rated several
statements concerning the normality and justifiability of the decisions, as well as what they
would have done in the protagonists’ situation.
To summarize, action was manipulated between-subjects. The effect of social normality
could be analyzed both between-subjects, using only the responses to the first decision and
outcome (on the second page of the materials) and within-subjects, using the responses to the
first and second decisions and outcomes (pages two and three of the materials).
Participants
Sixty-six undergraduate students at a large southwestern university participated for extra
course credit. They took about ten minutes to complete the questionnaire. (Ten additional
respondents were excluded for giving severely inconsistent responses, suggesting that they did
not take the task seriously or had not understood the instructions).
Materials and Measures
Scenario. The first page contained the following scenario adapted from one used by
Connolly and Reb (2003):
In Arizona there have been several outbreaks of a certain kind of flu, which can
cause severe illness in children under three. Most children just get normal flu symptoms
12
Decision Justifiability and Regret
but, in a small number of cases, the children get very sick with high fevers, a lot of pain,
and generally feel horrible for several weeks.
A vaccine for this kind of flu has been developed and tested. The vaccine
eliminates any possibility of the child getting the flu. The vaccine, however, can
sometimes cause side-effects that are very similar to the severe flu symptoms: high fever,
pain, feeling horrible for several weeks. Fortunately, these unpleasant effects are rare. In
fact, the risk of a vaccinated child getting the unpleasant side-effects is about as low as
the chance of a non-vaccinated child getting severe flu symptoms.
Both Cindy and Linda are married and each of them has a one-year old child.
Each of them has to decide whether or not to vaccinate the child against this flu.
The second and third pages each started with a description of one of the mothers’
decision and its outcome, giving information on the experimental factors. It read (manipulations
in parentheses):
[Cindy decides [not] to vaccinate her child, [like most of the people she knows /
although most of the people she knows decide [not] to vaccinate their children].
Unfortunately, her child [gets a severe case of the flu / experiences the severe side effects
of the vaccine] and is very sick and unhappy for almost three weeks.
Emotion ratings. After reading about the decision and its outcome, respondents were
asked to imagine themselves in the protagonist’s situation and to indicate how they would feel.
Regret was measured on an 11-point scale (0: “I wouldn’t experience this at all”; 10: “I would
experience this a lot”). The regret measure was embedded in a list of eleven other decisionrelated emotions (taken from Reisenzein and Hofmann, 1993) in order to avoid demand effects.
13
Decision Justifiability and Regret
After completing these scales, respondents turned the page, read about the second protagonist,
and completed the second set of emotion measures.
Decision ratings. After completing the emotion ratings for both scenarios, participants
turned to the last page where they rated several items concerning the decisions of the two
protagonists as well as their own vaccination preferences on 5-point Likert-scales (anchored at
“strongly agree” and “strongly disagree”, coded as +2 and –2, respectively). Decision ratings
were put on a separate page to reduce possible demand effects. A pair of items intended as
manipulation checks asked about the social normality of the decision of the two protagonists
(“[Linda / Cindy] chose the same alternative as most people she knows”). A second pair
measured perceptions of decision justifiability (“[Linda/Cindy] made a good choice”).
Results
Manipulation Check
The manipulation of social normality was successful. A between-subjects ANOVA on
social normality ratings found the expected main effect of social normality (socially normal M =
1.45, socially abnormal M = -1.29, F(1, 62) = 75.14, p < .001). The same main effect was also
revealed in a repeated-measures ANOVA (socially normal M = 1.41, socially abnormal M = 1.39, F(1, 62) = 163.19, p < .001). No other effects (action, order, and interactions) in the
between-subjects and repeated-measures ANOVAs were significant, all ps > .1.
Anticipated Regret
To examine whether any of the experimental manipulations affected anticipated regret we
first analyzed only the responses to the first decision with a between-subjects ANOVA including
action and social normality as between-subjects factors. We found a significant main effect for
action, F(1, 62) = 6.76, p < .05. Vaccination was considered less regrettable (M = 6.41, SD =
14
Decision Justifiability and Regret
3.43) than non-vaccination (M = 8.18, SD = 1.75). Whether the decision was socially normal (M
= 7.42, SD = 2.41) or socially abnormal (M = 7.23, SD = 3.17) had no significant effect on
anticipated regret intensity, F(1, 62) = .09, ns. The interaction was also non-significant, F(1, 62)
= .19, ns.
We next conducted a repeated-measures ANOVA on the full data set (i.e., responses to
the decisions of both protagonists) with social normality as a within-, and action and order as a
between-subjects factors. The main effect for action remained significant in the same direction,
F(1, 61) = 7.97, p < .01 (action M = 6.19, inaction M = 8.15). Social normality now also affected
regret intensity, F(1, 61) = 5.76, p < .05. A socially normal choice was considered less
regrettable (M = 6.97) than a socially abnormal choice (M = 7.46). The effect of social normality
became significant only when the manipulation was highlighted in a within-subjects design. The
main effect for order and all interactions were non-significant, all ps > .1.
Perceived Decision Justifiability
We examined the relation between respondents’ judgments of the justifiability of the
decision and their anticipated regret in several ways. First, analyzing only the responses to the
first decision, we found in a linear regression analysis that perceived justifiability predicted
anticipated regret, β = -.38, t(64) = 3.30, p < .01. The more a decision was perceived as
justifiable, the less regrettable it was anticipated to be in the event of a bad outcome. Social
normality perceptions, on the other hand, did not significantly predict anticipated regret, β = -.06,
t(64) = .49, ns. The same significance test results were obtained in analyses of the whole data set
(i.e., responses to both decisions).
Next, we tested whether the experimental manipulations that affected anticipated regret
also affected perceived justifiability. A between-subjects ANOVA with social normality and
15
Decision Justifiability and Regret
action as between-subjects factors showed a significant main effect for action such that a
decision to vaccinate was perceived as more justifiable (M = .47, SD = 1.16) than a decision not
to vaccinate (M = -.21, SD = 1.15), F(1, 62) = 6.22, p < .05. In addition, ANOVA revealed a
significant main effect for social normality, F(1, 62) = 6.29, p < .05. A decision to do what most
of the people one knows do was considered more justifiable (M = .48, SD = 1.15) than a socially
abnormal decision (M = -.20, SD = 1.16). The interaction with action was non-significant, p > .1.
The same results were obtained in a repeated-measures ANOVA with social normality as
within-subjects factor, and action and order as between-subjects factors: for action, F(1, 6) =
8.92, p < .01 (action M = .51, inaction M = -.25); for social normality, F(1, 62) = 4.63, p < .05
(socially normal M = .23, socially abnormal M = .02). The interaction between action and social
normality was non-significant, p > .1. In addition, the main effect for order was marginally
significant, F(1, 62) = 3.96, p = .09, such that ratings were somewhat higher when the socially
normal decisoin was rated first (M = .35) than when the abnormal decision was rated first (M = .09). However, none of the interactions involving order were significant, all ps > .1.
Because perceived justifiability predicted anticipated regret and was affected by the same
experimental manipulations as anticipated regret it is possible that perceived justifiability acted
as a mediator for the effects on regret. We tested whether the effect of action (independent
variable) on anticipated regret (dependent variable) in the between-subjects analysis was
mediated through perceived justifiability (presumed mediator). In addition to demonstrating a
significant relation between the independent variable and the dependent variable, the
independent variable and the presumed mediator, and the presumed mediator and the dependent
variable as we have done already, one needs to test whether inclusion of the mediator reduces the
impact of the independent variable on the dependent variable (cf. Baron & Kenny, 1986). When
16
Decision Justifiability and Regret
we regressed anticipated regret on both action and perceived justifiability, the effect of action
remained significant (β = .23, t(64) = 1.92, p < .05, one-tailed), and perceived justifiability
remained a significant predictor (β = -.32, t(64) = 2.68, p < .01). A test for mediation according
to Sobel (1982) was significant (z = 1.82, p < .05, one-tailed). Thus, decision justifiability
perceptions partially mediated the effect of action on anticipated regret intensity.
Discussion
Study 1 was intended to examine the role of perceived justifiability in the effect of
normality of choice on anticipated regret. We manipulated whether a mother decides to vaccinate
her child or not (action, manipulated between-subjects) and whether she decides to do what most
of the people she knows do (social normality, manipulated both between- and within-subjects).
We measured both anticipated regret and perceived justifiability. We found that action affected
anticipated regret in both between- and within-subjects analyses. Social normality, however,
affected anticipated regret only when manipulated within-subjects.
The results showed the hypothesized role of justifiability in explaining regret effects.
First, perceived justifiability predicted anticipated regret. The more a decision was perceived as
justifiable, the less regrettable it was considered in the event of a bad outcome. Second, a
mediation analysis found that perceived justifiability mediated the effect of action on anticipated
regret in the between-subject analysis. Perceived justifiability was rated at the end of the study
after both decisions were presented, introducing a potential confound on the measure through a
contrast effect of the within-subjects manipulation. However, action was held constant for each
respondent so post hoc contrast effects are irrelevant. It is difficult to see how the within-subjects
manipulation of social normality could have been confounded with the effect of action on
17
Decision Justifiability and Regret
perceived justifiability. Thus, we are confident that the results of the mediation analysis were not
affected in this case by the within-subjects manipulation of social normality.
Third, highlighting the difference in social normality between the two protagonists
through a within-subjects manipulation led to a significant effect on anticipated regret that was
not obtained for the between-subjects manipulation. We argue that this is because withinsubjects, justifiability issues were made salient by direct comparison of normal and abnormal
behavior. Without this additional salience the (between-subjects) normality manipulation may
have been too weak to affect regret. As N’gbala & Branscombe (1997) argue, the mutability
hypothesis predicts a stronger effect of normality in the between-subjects design because it rests
on the argument that abnormal behaviors evoke counterfactuals more easily. However, in a
within-subjects design, even the counterfactual for normal behavior is essentially obvious
because both decision alternatives (normal and abnormal) are described to the respondent. This
would imply that the normality effect would be found in the between-subjects rather than the
within-subjects design, the opposite of our finding. Justifiability appears to have trumped
mutability in determining regret in this context. Regression analysis also showed that social
normality perceptions did not predict anticipated regret, further weakening the case that
normality as such determines anticipated regret.
Study 1 examined the effects of factors associated with the normality of a decision that
have been previously shown to affect (anticipated) regret. While in the past effects of normality
have been explained through the ease of counterfactual generation (Kahneman & Miller, 1986)
we predicted that differences in perceived justifiability could account for the effects of action and
social normality. The results of Study 1 support this interpretation. However, to further probe the
importance of justifiability in the anticipation of regret it is important to identify new variables
18
Decision Justifiability and Regret
that can be expected to affect perceived justifiability, and thus anticipated regret, in addition to
reexamining previously studied variables such as normality. Study 2 does this by investigating
the effect of decision process quality on anticipated regret and perceived justifiability.
STUDY 2: DECISION PROCESS QUALITY AND REGRET
Existing studies of justification and anticipated regret have examined justifications that
are highly context-specific: previous experience with a particular product (Inman & Zeelenberg,
2002), the reputation enjoyed by a specific brand (Simonson, 1992), or an individual’s
entrepreneurial personality (Seta et al., 2001). Study what examined the effect of another
consideration, social conformity, that might serve as a regret-reducing justification in a variety of
situations, though perhaps not a very compelling one. In the present experiment we examine a
justification that is potentially powerful over a wide range of choices: the care and thoroughness
of the decision process itself. Janis and Mann (1977) suggest that the anticipation of regret can
motivate what they call a “vigilant decision process”. Our question here is whether the use of
such a process does, in fact, lead to reduced expectations of decision-related regret.
We study the effect of decision process quality on anticipated regret for the same
vaccination scenario used before. The protagonist in the scenario was described either as
following a careful decision process (collecting a lot of relevant information, surveying all
available options and carefully considering the advantages and disadvantages of each) or a
careless decision process (doing none of these). We also manipulated action such that one group
read about a decision to vaccinate, the other group about a decision not to vaccinate. Although
we were mainly interested in the decision process quality main effect, this additional
manipulation allowed us to examine possible interactions between the two variables (that is, to
19
Decision Justifiability and Regret
see whether the effect of decision process quality on anticipated regret depended on whether the
decision was to vaccinate or not to vaccinate).
Method
Design and Procedure
The design and procedure were identical to Study 1 except that decision process quality
was manipulated instead of social normality. As with the manipulation of social normality in
Study 1, decision process quality was manipulated both within-subjects and between-subjects. In
contrast to Study 1, however, we expected decision process quality to be a strong enough
manipulation of justifiability to affect anticipated regret even in both within- and betweensubjects designs. This prediction was borne out in the data and we will therefore focus largely on
the between-subjects analyses and only briefly describe the parallel within-subjects results.
These between-subjects analyses use only the responses to the first choice, which was either
made using a careful or a careless decision process. As in Study 1, action was manipulated fully
between-subjects, so that a respondent read either about two protagonists who both decided to
vaccinate or two who decided not to vaccinate.
Participants
148 students at a large southwestern university participated for extra-credit and took
about 10 minutes to fill out the questionnaire.
Materials and Measures
The scenario was slightly changed from the previous one. Specifically, at the beginning
of the scenario, second sentence, where it read “in a small number of cases” in Study 1, the
scenario now read, less ambiguously, “in about 1% of cases”. The first page of the scenario also
20
Decision Justifiability and Regret
contained the action manipulation and information about the decision outcome, information
presented on the following page in Study 1. It ended as follows.
… Both Cindy and Linda are married and each has a one-year old child.
Both mothers [decide to / decide not to] vaccinate their child. Unfortunately,
both children did [experience the severe side effects of the vaccine / get a severe
case of the flu] and were very sick and unhappy for almost three weeks.
Participants were then presented with two emotion rating sheets, one of which was
headed by a description of a careful decision process, the other of a careless decision process.1
The careful decision process was described as:
[Protagonist] talked to several doctors, read medical journals, consulted with family and
friends, and thought seriously about what to do.
The careless decision process stated:
[Protagonist] did not collect any information, did not talk with friends or doctors before
she made the decision, and didn’t think very seriously about the decision.
Anticipated regret was measured on the same 11-point (0-10) scale as in Study 1.
Perceived justifiability was measured through a rating of anticipated self-blame made on similar
scale, reverse-coded such that higher values indicate higher perceived justifiability. Measures of
anticipated regret and perceived justifiability of the first decision were taken immediately after
respondents read the first part of the scenario, before they read about the second decision. The
between-subjects analyses of perceived justifiability of Decision 1 are thus uncontaminated by
contrast effects involving Decision 2.
Results
Anticipated Regret
21
Decision Justifiability and Regret
We conducted an ANOVA on anticipated regret with decision process quality and action
as between-subjects factors. This analysis revealed a significant main effect for decision process
quality, F(1, 140) = 18.51, p < .001). As expected, anticipated regret was lower when the bad
outcome followed a careful decision process (M = 5.64, SD = 3.01) than when it followed a
careless decision process. (M = 7.67, SD = 2.57). This effect did not depend on whether the
decision was to act or not (interaction F(1, 140) = .14, ns). Also, anticipated regret did not
depend on whether the decision was to vaccinate or not, F(1, 140) = .05, ns. The same
significance test results emerged from a repeated-measures ANOVA. Decision process quality,
the within-subjects factor, again affected anticipated regret, F(1,139) = 62.44, p < .001). The
effects of action, order, and the interactions were all non-significant, all ps > .1.
Perceived Decision Justifiability
Again, we examined the relation between perceived justifiability and anticipated regret in
several ways. First, we correlated anticipated regret with perceived justifiability, analyzing only
the responses to the first decision. As expected, perceived justifiability strongly predicted
anticipated regret, β = -.66, t(146) = 10.64, p < .001. The same significance test result was
obtained in an analysis of the full data set (i.e., responses to both decisions).
Next, we examined whether the effect of decision process quality on anticipated regret
was matched by a consistent effect on perceived justifiability. A between-subjects ANOVA on
perceived justifiability showed a significant main effect for decision process quality, F(1, 140) =
14.15, p < .001), with perceived justifiability higher after a careful process (M = 4.08, SD = 2.94)
than after a careless process (M = 2.33, SD = 2.60). The effect of action and the interaction effect
were non-significant, both p > .1. The same main effect for decision process quality emerged
from a repeated-measures ANOVA, F(1, 140) = 80.16, p < .01. In addition, decision process
22
Decision Justifiability and Regret
quality interacted significantly with order, F(1, 140) = 19.75, p < .001. Analysis of means
showed that the effect of decision process quality was stronger when the careless protagonist was
presented first (M = 4.37 vs. M = 7.67) than when the careful protagonist was presented first (M
= 5.92 vs. M = 7.07). No other main or interaction effects were significant, all ps > .1.
Finally, we conducted a mediation analysis of the between-subjects effect of decision
process quality on anticipated regret through perceived justifiability (cf. Baron & Kenny, 1986).
We have already shown that decision process quality (independent variable) affected anticipated
regret (dependent variable); that decision process quality affected perceived justifiability
(presumed mediator); and that perceived justifiability predicted anticipated regret. A regression
analysis with decision process quality and perceived justifiability as simultaneous predictors of
anticipated regret showed that both decision process quality, β = .16, t(146) = 2.45, p < .05, and
perceived justifiability, β = -.61, t(146) = 9.57, p < .001, remained significant. A Sobel (1986)
test for mediation was significant, z = 3.56, p < .001. Thus, we find that perceptions of decision
justifiability partially mediated the effect of decision process quality on anticipated regret.
Discussion
Study 2 was designed to test the prediction that decision process quality affects
anticipated regret. We argued that if single good reasons for a decision, such as good past
performance of a product (Inman & Zeelenberg, 2002) or a soccer team (Zeelenberg et al.,
2001), can lead to increased justifiability and lower anticipated regret, a careful decision process
should achieve the same result as well or better. As expected, we found that decision process
quality affected anticipated regret intensity such that a bad outcome following a careful decision
process was expected to lead to less intense regret than a bad outcome following a careless
decision process.
23
Decision Justifiability and Regret
As expected, perceived justifiability predicted anticipated regret. Moreover, a mediation
analysis showed that perceived justifiability mediated the effect of decision process quality on
anticipated regret. These results are consistent with decision justification theory and lend support
to the idea that decision justifiability has a significant influence on the intensity of anticipated
regret.
In contrast to Study 1, action did not affect anticipated regret. While we cannot be sure
about the cause of this difference in results, one explanation consistent with the justifiability idea
is that different justifications are not simply additive in their effects. In the presence of a stronger
justification, the influence of weaker justifications may be reduced or eliminated. As we argued
earlier, the quality of the decision process is likely a very strong justification of a choice. When a
person makes a choice after collecting a lot of information and deliberating carefully, either
choice (i.e., vaccinating and not vaccinating) seems justified, and the action effect disappears.
However, in the absence of this process information, vaccinating might appear more justified, as
it reflects the majority choice.
STUDY 3: DECISION PROCESS QUALITY, NORMALITY AND ANTICIPATED REGRET
Study 3 combines the variables of Studies 1 and 2 and manipulates decision process
quality, action and social normality within one experiment. In this study we wanted to further
examine the idea that the effects of different justifications are not necessarily additive. It is
possible that that a characteristic that served as a justification loses its influence on anticipated
regret when another, stronger justification is added to the picture, as in our tentative explanation
(above) for the disappearance of the action-regret link in Study 2. The present study will examine
this possible non-additivity in the joint effect of three potential justifications: decision process
24
Decision Justifiability and Regret
quality, action, and social normality. One can also imagine other forms of complex interactions
and we will examine these on a more exploratory basis.
In Study 3 we manipulated decision process quality between- and within-subjects as in
Study 2. However, returning to the procedure of Study 1, we measured justifiability perceptions,
together with several manipulation check variables, at the end of the experiment. This was done
to rule out a potential influence of rating decision justifiability (as well as the manipulation
checks) on the subsequent anticipated regret ratings, the primary focus of the study. The cost, in
this case, was potential contamination of the justification measure with the contrast effect of a
within-subject manipulation of decision process quality.
Method
Participants read about a protagonist who either vaccinated her child or not, did or did not
do what most people she knows do, and engaged in a careful or careless decision process. Thus,
we manipulated three factors each with 2 levels: action (vaccinate vs. don’t vaccinate), social
normality, and decision process quality. Action and social normality were manipulated betweensubjects. Decision process quality was manipulated both between-subjects and within-subjects as
in Study 2.
One hundred and fifty-four undergraduate students at a large southwestern university
participated for extra course credit. They took about ten minutes to complete the questionnaire.
The materials and measures were identical to those for Study 1 except that the scenario
contained three manipulations. The relevant paragraph of the scenario read as follows.
[Cindy gives the matter a lot of thought, consults several doctors and medical
journals, and talks to her family and friends. Finally, she / Linda doesn’t give the matter
much thought, nor does she discuss it with her doctor, her family or her friends. She just]
25
Decision Justifiability and Regret
decides [not] to vaccinate her child, [like most of the people she knows / although most
of the people she knows decide [not] to vaccinate their children]. Unfortunately, her child
[gets a severe case of the flu / experiences the severe side effects of the vaccine] and is
very sick and unhappy for almost three weeks.
The emotion ratings were the same as before and were followed by decision ratings. A
pair of items measured decision justifiability: “[Linda / Cindy] made a justifiable decision”. The
manipulation checks asked about the carefulness of the decision process (“[Linda / Cindy]
followed a careful decision process”) and the social normality of the decision (“[Linda / Cindy]
chose the same alternative as most people she knows”). All decision ratings were made on the
same 5-point Likert-scale (coded -2 to +2) as in Study 1.
Results
Manipulation Checks
The manipulations of decision process quality and social normality were successful.
Between-subjects ANOVA on perceptions of the carefulness of the decision process showed the
expected main effect for decision process quality, F(1, 146) = 224.25, p < .001 (careful M =
1.48, careless M = -1.32); no other main effect or interaction was significant, all p > .1. Betweensubjects ANOVA on perceived social normality showed the expected main effect of social
normality, F(1, 146) = 116.81, p < .001 (socially normal M = .99, socially abnormal M = -1.24);
no other main effect or interaction was significant, all p > .1.
These effects were replicated in repeated-measures ANOVAs (using ratings of both
protagonists): a main effect of decision process quality on perceived carefulness (careful M =
1.54, careless M = -1.42, F(1, 145) = 404.88, p < .001), and a main effect of social normality on
perceived social normality (socially normal M = .99, socially abnormal M = -1.17, F(1, 146) =
26
Decision Justifiability and Regret
172.13, p < .001). In addition, within-subjects analyses showed a main effect of order on
perceived carefulness, F(1, 145) = 4.42, p < .05, such that carefulness ratings were lower when
the careful process was rated first (M = -.03) rather than second (M = .14). Importantly, order did
not significantly interact with any of the other variables, all ps > .1. Also no other main effects
and interactions significantly affected perceived decision process carefulness, all ps > .1. For
perceived social normality, we found a main effect for order, F(1, 146) = 3.83, p = .05, such that
normality ratings were lower when the careful process was rated first (M = -.25) rather than
second (M = .07). However, order did not interact with any of the experimental variables, all ps >
.1. In addition, there was a main effect for decision process quality, F(1, 145) = 7.27, p < .01,
such that a careful decision process was perceived as more socially normal (M = .07) than a
careless process (M = -.28). However, this effect was much weaker than the effect of social
normality on the same variable. No other main effects or interactions significantly affected
perceived social normality, all ps > .1.
Anticipated Regret
We first performed a between-subjects ANOVA on the first responses with decision
process quality, social normality, and action as between-subjects factors. As expected, we found
a main effect for decision process quality, F(1, 146) = 4.28, p < .05, such that anticipated regret
was lower when a careful decision process was used (M = 6.43, SD = 3.43) than when a careless
process was used (M = 7.43, SD = 2.88). The main effect for action was non-significant, F(1,
146) = 1.65, ns, (inaction M = 7.23, SD = 3.23, action M = 6.64, SD = 3.14). The main effect for
social normality was also non-significant, F(1, 146) = .68, ns (socially normal M = 7.20, SD =
3.16, socially abnormal M = 6.69, SD = 3.21), as were all of the interaction terms, all ps > .1.
27
Decision Justifiability and Regret
We next ran a repeated-measures ANOVA with decision process quality as within- and
social normality, action, and order as between-subjects factors. Replicating the between-subjects
results, the main effect of decision process quality was again significant, F(1, 146) = 21.48, p <
.001 (careful M = 6.23, careless M = 7.22), but the other main effects were not, all ps > .1. This
significant main effect was, however, qualified by a two-way interaction between decision
process quality and social normality, F(1, 146) = 7.32, p < .01. The only significant simple effect
was between a careless and a careful decision process with a socially normal choice. A careless
decision process leading to a decision to do what most people one knows do was expected to be
more regrettable (M = 7.63) than a careful decision process leading to the same socially normal
choice (M = 6.04), F(1, 75) = 23.44, p < .01. Abnormal choices were expected to be equally
regrettable whether the process that produced them was careless (M = 6.82) or careful (M =
6.41). While speculative interpretation is easy (e.g., a careless, socially conforming choice might
be seen as parental abdication) we note that the effect (a) was not predicted, (b) is modest in size,
(c) appeared only in within-subjects, not in between-subjects analysis, and (d) was not replicated
in the analysis of perceived justification, reported below. In the absence of subsequent
replication we therefore treat this finding with considerable caution.
Perceived Decision Justifiability
Again, we used several different analyses to examine the relation between perceived
justifiability and anticipated regret. First, we found that perceived justifiability significantly
predicted anticipated regret, analyzing only the responses to the first decision, β = -.30, t(152) =
3.89, p < .001. The more justifiable a decision was perceived, the less regrettable it was
considered. Social normality perceptions did not significantly predict anticipated regret, β = .13,
t(152) = 1.60, ns (note that the direction of the association was such that higher perceived
28
Decision Justifiability and Regret
normality went along with higher anticipated regret, if anything). The same significance test
results held when analyzing responses to both decisions.
Next, we examined whether decision process quality affected perceived justifiability in a
manner that matches its effect on anticipated regret. A between-subjects ANOVA with decision
process quality, social normality, and action as between-subjects factors showed a significant
main effect for decision process quality, F(1, 146) = 101.47, p < .001, such that a carefully made
decision was perceived as more justifiable (M = 1.13, SD = 1.04) than a carelessly made decision
(M = -.73, SD = 1.25). The main effect for action was not significant, F(1, 146) = .90, ns
(Minaction = .13, Maction = .23). Also, a socially normal decision was not considered more
justifiable (M = .14) than a socially abnormal decision (M = .21), F(1, 146) = .01, ns. None of the
interactions were significant, all ps > .1. This pattern of results matches the pattern of results for
anticipated regret.
This result was replicated in a repeated-measures ANOVA with decision process quality
as within- and social normality, action, and order as between-subjects factors. We found a main
effect of decision process quality (careful M = 1.14, careless M = -.74, F(1, 145) = 194.38, p <
.001). The interaction between decision process quality and social normality observed for
anticipated regret was not significant for perceived justifiability, F(1, 145) = .54, ns. Thus, while
the main effect of decision process quality on anticipated regret was mirrored by the justifiability
perceptions, the interaction between process quality and social normality was not. No other main
or interaction effects were significant, all ps > .1.
Finally, we performed a mediation analysis to test whether perceived justifiability
mediated the effect of decision process quality on anticipated regret found in the betweensubjects analysis. We already reported that decision process quality affected anticipated regret;
29
Decision Justifiability and Regret
that decision process quality affected perceived justifiability; and that perceived justifiability
predicted anticipated regret. When regressing anticipated regret on both the decision process
quality and decision justifiability perceptions, decision process quality was not a significant
predictor of regret, β = -.06, t(152) = .57, ns, but perceived justifiability remained significant, β =
-.34, t(152) = 3.35, p < .01). A Sobel (1982) test for mediation was significant (z = 3.18, p < .01).
We conclude that decision justifiability perceptions mediated the between-subjects effect of
decision process quality on anticipated regret intensity.
Discussion
In Study 3 we manipulated decision process quality, action, and social normality. We
found that a bad outcome following a careful, rather than careless, decision process was expected
to be less regrettable even when information about action and social normality was present.
Further, supporting a justifiability explanation of regret intensity, perceived justifiability was
higher for a careful decision process and the effect of process quality on regret was fully
mediated through justifiability perceptions. In the presence of process quality and social
normality information, whether the decision was to act or not did not affect regret or justifiability
perceptions.
GENERAL DISCUSSION
In this paper, we studied the effects of decision process quality, social normality, and
action (i.e., whether the decision was to act or not) on anticipated regret. Our major concern has
been with the role of perceived decision justifiability in explaining the intensity of anticipated
regret. The three experiments reported yielded several important results.
30
Decision Justifiability and Regret
First, past research and theorizing has suggested that the normality of a decision affects
(anticipated) regret through the easier counterfactual mutability of abnormal choices (Kahneman
& Tversky, 1982; Kahneman & Miller, 1986). Study 1 showed that the effects of action and
social normality on anticipated regret can be to a substantial degree explained through perceived
decision justifiability. We did not replicate the so-called “action effect” (e.g., Kahneman &
Tversky, 1982). Our results suggest instead that whether a decision to act or not leads to more or
less regret depends on the specific justifiability of the option rather than action or inaction per se
(cf. Connolly & Reb, 2003; Inman & Zeelenberg, 2002). We chose a context in which “acting” is
endorsed by the majority of the U.S. population: the decision whether to vaccinate one’s child
against a serious disease. In this context, we found that the effect of action on anticipated regret
was mediated through perceptions of decision justifiability. In addition, we found that social
normality affected anticipated regret only when the manipulation was made salient in a withinsubjects manipulation. As predicted, in a within-subjects design, the differential justifiability of
normal and abnormal behavior becomes more salient leading to a stronger effect on anticipated
regret.
The results are consistent with decision justification theory (Connolly & Zeelenberg,
2002). We now have two plausible explanations of these effects, one based on justifiability and
the other based on mutability and counterfactual thinking (Kahneman & Miller, 1986). These
explanations are not necessarily mutually exclusive. Future research should examine how
justifiability and mutability relate to each other in explaining these effects, whether one
explanation is more powerful than the other, and which explanation is more parsimonious.
Second, we found that a careful decision process is expected to protect decision makers
to some degree from regret following a bad outcome. The effect of decision process quality on
31
Decision Justifiability and Regret
anticipated regret was significant in both Study 2 and Study 3, where social normality
information was also given, in both between-subjects and within-subjects analyses, and
independent of whether the decision was to vaccinate or not to vaccinate. Mediation analysis
showed the effect of decision process quality on anticipated regret to be significantly mediated
through perceived justifiability, supporting decision justification theory.2
All three studies showed the important role of decision justifiability in determining the
intensity of anticipated regret. Table 1 summarizes the main effects of the experimental variables
on anticipated regret and on perceived decision justifiability. With only one exception the
experimental variables showed either significant main effects on both variables or on neither.
The exception is Study 1, where the between-subjects analysis of social normality showed a
significant effect on perceived justifiability but not on anticipated regret. It should be noted that,
even when anticipated regret was reduced by strong justifiability, the mean score remained
around the mid-point of the scale. Consistent with decision justification theory’s two-component
model of regret, substantial outcome regret (over the child’s sickness) remains even if the selfblame component of regret is entirely eliminated.
-----------------------------------Insert Table 1 around here
-----------------------------------Two additional findings hint at the complexities of combining multiple sources of
justifiability. First, we found no significant main effects on anticipated regret of either action or
of social normality when decision process quality information was given (Study 3). These results
suggest that the effects of weaker justifications, such as whether a decision is socially
conforming, may be driven out by stronger justifications, such as the carefulness of the decision
32
Decision Justifiability and Regret
process. Future research needs to examine in more detail the conditions under which different
justifications crowd each other out. We also found a significant interaction between process
quality and social normality on anticipated regret in Study 3, but place little confidence in this
result absent of future replication.
Limitations and Future Research
Several limitations of our studies point to directions for future research. First, Janis and
Mann (1977) predicted that regret aversion leads to “vigilant” decision making. Consistent with
this prediction we have shown that people anticipate more regret when a poor outcome follows a
careless decision process than a careful one. However, we did not demonstrate that decision
makers will actually improve their decision processes to avoid this anticipated regret. Future
research should examine whether regret aversion, either experimentally induced or measured as
an individual difference variable, leads to more vigilant, careful decision making. Such a finding
would have important implications for the debate about the functionality of regret (e.g.,
Zeelenberg, 1999) and would suggest that anticipated regret can have a beneficial effect on
decision making. Further, we did not show that a careful decision process (or deciding to
vaccinate, for that matter) actually reduces experienced regret after a bad decision outcome.
Future research should examine the intensity of experienced regret as a function of the quality of
the decision process.
Decision justification theory (Connolly & Zeelenberg, 2002) assigns a key role to how
justifications affect (anticipated) regret. Our results went a substantial way in examining and
comparing different potential justifications. We found that the relation between different
justifications can be complex. In our three studies different justifications acted independently on
regret, dominated each other (driving each other out), and produced interactions with each other.
33
Decision Justifiability and Regret
Clearly, the effects of multiple justifications need not be additive. Future research should
examine in more detail the ways in which justifications combine and interact.
A further limitation of our study was the use of a single context (in which anticipated
regret has been shown to strongly predict actual choice, however; see Wroe et al., 2004). It is
quite possible that justifications also interact with the decision context. In the context studied
here, medical decisions for a child, a careful, comprehensive analytical decision process was
seen as best justifying the choice the parent made. Such a process might not be seen as
appropriate for a choice in which intuition and spontaneity is called for, as in the choice of a
spouse. Similarly, the seriousness of the stakes involved might change the kind of justification
called for. A modest flutter at a gaming table might be justified by “I simply felt like it”. A
moderately expensive purchase of some new electronic gadget or a meal at a trendy restaurant
might be adequately justified by noting that lots of other people were doing the same. Perhaps
the “good decision process” justification used here is, in fact, seen as appropriate only for a
subset of life’s big decisions.
Finally, in all three studies, all decisions led to negative outcomes. One implication of
decision justification theory’s distinction between self-blame regret and outcome regret is that
regret might be experienced even in the absence of a bad outcome when the decision is perceived
unjustifiable. Sugden (1985) suggested that one might experience extreme regret over a decision
to drive home drunk from a party, even if one made the trip safely. Such self-blame regret in the
absence of a bad outcome has not, to our knowledge, been demonstrated, and would in fact be
somewhat surprising in light of the extensive literature (e.g. Pratto & John, 1991; Taylor, 1991)
showing that negative events stimulate much more cognitive activity than do good ones. Future
34
Decision Justifiability and Regret
research should examine whether or not a poor decision outcome is a prerequisite to the
experience or anticipation of self-blame regret.
Conclusion
This paper examined the role of decision justifiability in determining the intensity of
anticipated regret. We found support for the central prediction of decision justification theory
(Connolly & Zeelenberg, 2002) that anticipated regret is reduced when a decision is perceived as
justifiable. We also showed that engaging in a careful decision process is expected to protect the
decision maker to some degree from regret following a bad decision outcome. Finally, when
multiple sources of possible justifications are offered simultaneously, their combined effect is
not, in general, simply the sum of the effects of each. Complex interactions are to be expected.
Overall, these results support the ability of decision justification theory to provide an account of
decision makers’ regret anticipations. They also suggest a positive view of the role of regret in
decision making. Earlier research on the so-called action effect suggested that anticipated regret
might be a biasing factor, leading to a preference for inaction even when the active option was
objectively superior. The current research suggests, instead, that anticipated regret can have a
beneficial effect on decision making, by stimulating a search for more justifiable decisions, and
thus encouraging more careful decision processes.
35
Decision Justifiability and Regret
FOOTNOTES
1
The header also included what was intended as a manipulation of “personal normality”,
whether or not the protagonist made this choice as she normally made medical decisions for her
child. In retrospect it became clear that the wording of this manipulation was unclear. Since the
data showed no significant relationship between the manipulation and anticipated regret, we
collapsed on this variable and excluded it from further analysis.
2
A reviewer points out that cognitive dissonance theory might provide an alternative
account of the apparent ameliorating effect of good decision process on anticipated regret. Given
that good decision process implies greater decision effort, the presumed dissonance would
involve the elements “I made a large effort” and “I achieved (or expect to achieve) a poor (i.e.
regrettable) result”. This dissonance could be reduced by reassessing downwards the
regrettability of the outcome (the sick child), generating the negative correlation observed in
Studies 2 and 3 between decision process quality and anticipated regret. This alternative account
seems plausible to us (and, indeed, reminds us of the important similarities between decision
justification and cognitive dissonance theories), and it is not obvious how we might distinguish
the two accounts on the present data. On grounds of parsimony alone we prefer decision
justification theory (which accounts for all three main effects via a justification-regret
mechanism) over dissonance theory (which would account for the process quality findings of
Studies 2 and 3 but does not offer an obvious account of the action and normality effects found
in Study 1 or of their suppression in Studies 2 and 3). However, a serious effort to distinguish
between the two accounts would require both a more complete development of decision
justification theory and new data collection beyond the scope of the present report.
36
Decision Justifiability and Regret
REFERENCES
Baron, J. (1992). The effect of normative beliefs on anticipated emotions. Journal of Personality
and Social Psychology, 63, 320-330.
Baron, J., & Ritov, I. (2004). Omission bias, individual differences, and normality.
Organizational Behavior and Human Decision Processes, 94, 74-85.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social
psychological research: Conceptual, strategic and statistical considerations. Journal of
Personality and Social Psychology, 51, 1173-1182.
Bell, D. E. (1982). Regret in decision making under uncertainty. Operations Research, 30, 961981.
Connolly, T., Ordóñez, L., & Coughlan, R. (1997). Regret and responsibility in the evaluation of
decision outcomes. Organizational Behavior and Human Decision Processes, 70, 73-85.
Connolly, T., & Reb, J. (in press). Regret in cancer-related decisions. Health Psychology.
Connolly, T., & Reb, J. (2003). Omission bias in vaccination decisions: Where’s the “omission”?
Where’s the “bias”? Organizational Behavior and Human Decision Processes, 91, 186-202.
Connolly, T. & Zeelenberg, M. (2002). Regret in decision making. Current Directions in
Psychological Science, 11, 212-220.
Crawford, M. T., McConnell, A. R., Lewis, A. C., & Sherman, S. J. (2002). Reactance,
compliance, and anticipated regret. Journal of Experimental Social Psychology 38, 56–63.
Gilbert, D. T., Morewedge, C. K., Risen, J. L., & Wilson, T. D. (2004). Looking forward to look
backward: The misprediction of regret. Psychological Science, 15, 346-350.
Inman, J. J., & Zeelenberg, M. (2002). Regret in repeat purchase versus switching decisions: The
attenuating role of decision justifiability. Journal of Consumer Research, 29, 116-128.
37
Decision Justifiability and Regret
Janis, I. L., & Mann, L. (1977). Decision making. New York: Free Press.
Kahneman, D., & Miller, D. T. (1986). Norm theory: Comparing reality to its alternatives.
Psychological Review, 93, 136-153.
Kahneman, D., & Tversky, A. (1982). The psychology of preferences. Scientific American, 246,
160-173.
Landman, J. (1987). Regret and elation following action and inaction: Affective responses to
positive versus negative outcomes. Personality and Social Psychology Bulletin, 13, 524-536.
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, 87-97.
Loomes, G., & Sugden, R. (1982). Regret theory: an alternative theory of rational choice under
uncertainty. Economic Journal, 92, 805-824.
Mellers, B. A., Schwartz, A., & Ritov, I. (1999). Emotion-based choice. Journal of Experimental
Psychology: General, 128, 332-245.
N’gbala, A., & Branscombe, N. R. (1997). When does action elicit more regret than inaction and
is the counterfactual mutation the mediator of this effect? Journal of Experimental Social
Psychology, 33, 324-343.
Pratto, F., & John, O. P. (1991). Automatic vigilance: The attention-grabbing power of negative
social information. Journal of Personality and Social Psychology, 61, 380-391.
Reisenzein, R., & Hofmann, T. (1993). Discriminating emotions from appraisal-relevant
situational information: Baseline data for structural models of cognitive appraisal. In N. H.
38
Decision Justifiability and Regret
Frijda (Ed.), Appraisal and beyond: The issue of cognitive determinants of emotion (pp. 271293). Hillsdale: Lawrence Erlbaum Associates.
Richard, R., van der Pligt, J., & de Vries, N. K. (1996). Anticipated regret and time perspective:
Changing sexual risk-taking behavior. Journal of Behavioral Decision Making, 9, 185-199.
Ritov, I., & Baron, J. (1990). Reluctance to vaccinate: Omission bias and ambiguity. Journal of
Behavioral Decision Making, 3, 263-277.
Ritov, R., & Baron, J. (1995). Outcome knowledge, regret, and omission bias. Organizational
Behavior and Human Decision Processes, 64, 119-127.
Roese, N. J. (1997). Counterfactual thinking. Psychological Bulletin, 121, 133-148.
Roese, N. J., & Olsen, J. M. (Eds.). (1995). What might have been: The social psychology of
counterfactual thinking. Mahwah, NJ: Lawrence Erlbaum Associates.
Savage, L. J. (1951). The theory of statistical decision. Journal of the American Statistical
Association, 46, 55-67.
Seta, J. J., McElroy, T., & Seta, C. E. (2001). To do or not to do: Desirability and consistency
mediate judgments of regret. Journal of Personality and Social Psychology, 80, 861-870.
Simonson, I. (1992). The influence of anticipating regret and responsibility on purchase
decisions. Journal of Consumer Research, 19, 105-118.
Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation
models. In S. Leinhardt (Ed.), Sociological Methodology 1982 (pp. 290-312). Washington
DC: American Sociological Association.
Sugden, R. (1985). Regret, recrimination and rationality. Theory and Decision, 19, 77-99.
Taylor, S.E. (1991). Asymmetric effects of positive and negative events: The mobilizationminimization hypothesis. Psychological Bulletin, 110, 67-85.
39
Decision Justifiability and Regret
Wroe, A. L., Turner, N., & Salkovskis, P. M. (2004) Understanding and predicting parental
decisions about early childhood immunizations. Health Psychology, 23, 33-41.
Zeelenberg, M. (1996). On the importance of what might have been: Psychological perspectives
on regret and decision making. Doctoral dissertation: University of Amsterdam.
Zeelenberg, M. (1999). The use of crying over spilled milk: A note on the rationality and
functionality of regret. Philosophical Psychology, 12, 325-340.
Zeelenberg, M., van den Bos, K., van Dijk, E., & Pieters, R. (2001). The inaction effect in the
psychology of regret. Journal of Personality and Social Psychology, 82, 314-327.
40
Decision Justifiability and Regret
Table 1: Summary of Significant Main Effects of Experimental Variables on Anticipated Regret
and Perceived Decision Justifiability
Dependent Variable*
Study 1
Study 2
Study 3
Experimental
Between-Subjects or
Anticipated
Decision
Variable
Within-Subjects Analysis
Regret
Justifiability
Action
Between
Yes
Yes
Action
Within
Yes
Yes
Social normality
Between
No
Yes
Social normality
Within
Yes
Yes
Decision process quality
Between
Yes
Yes
Decision process quality
Within
Yes
Yes
Action
Between
No
No
Action
Within
No
No
Decision process quality
Between
Yes
Yes
Decision process quality
Within
Yes
Yes
Action
Between
No
No
Action
Within
No
No
Social normality
Between
No
No
Social normality
Within
No
No
* In all cases in which the experimental variable was shown to affect both dependent variables,
greater decision justifiability is associated with lower levels of anticipated regret.
41