Content uploaded by Christopher J. Anderson
Author content
All content in this area was uploaded by Christopher J. Anderson on Feb 21, 2016
Content may be subject to copyright.
Content uploaded by Christopher J. Anderson
Author content
All content in this area was uploaded by Christopher J. Anderson on Feb 21, 2016
Content may be subject to copyright.
The Psychology of Doing Nothing: Forms of Decision Avoidance Result
From Reason and Emotion
Christopher J. Anderson
University at Albany, State University of New York
Several independent lines of research bear on the question of why individuals avoid decisions by
postponing them, failing to act, or accepting the status quo. This review relates findings across several
different disciplines and uncovers 4 decision avoidance effects that offer insight into this common but
troubling behavior: choice deferral, status quo bias, omission bias, and inaction inertia. These findings are
related by common antecedents and consequences in a rational–emotional model of the factors that
predispose humans to do nothing. Prominent components of the model include cost–benefit calculations,
anticipated regret, and selection difficulty. Other factors affecting decision avoidance through these key
components, such as anticipatory negative emotions, decision strategies, counterfactual thinking, and
preference uncertainty, are also discussed.
The experience of postponing and avoiding certain choices is
universal, yet often appears to work against individuals’ goals.
Delays transform into lost opportunities, and adhering to the status
quo is frequently unjustified given advantageous alternatives. Still,
individuals persist in seeking default no-action, no-change options.
Decision avoidance deserves concentrated attention, yet it has
not been studied in an integrated manner because it does not fit
neatly into the current paradigms in clinical, cognitive, or social
psychology. Yet, it is a common phenomenon with high personal
and societal costs. Under conditions of high stress, this avoidance
can become extreme. Take, for example, the “old sergeant syn-
drome” described by Janis and Mann (1977b). Infantry on the front
lines of battle for long periods, witnessing the death of comrades
and having no hope of transfer, have been known to ignore
decisions required to protect themselves under fire or from routine
safety hazards. For them, decision avoidance costs lives. In more
common environments, the costs of decision avoidance are too
widespread and often ineffable to attempt to calculate here. Almost
everyone can cite examples of the high price of a failure to act, and
there is evidence that humans are only becoming more indecisive,
perhaps as a part of cultural evolution in the information age
(Dentsu Reports, 1999). Given the adverse outcomes they expose
themselves to by delaying and failing to act, why do humans so
frequently engage in decision avoidance?
Herein I consider a variety of choice behaviors as reflections of
an individual’s underlying decision avoidance, a pattern of behav-
ior in which individuals seek to avoid the responsibility of making
a decision by delaying or choosing options they perceive to be
nondecisions. This review reveals that in all such cases there is a
mixture of a few good, rational reasons for avoidance and a more
complex and rationally questionable role played by emotions such
as regret and fear. These issues form the basis of this article: (a) the
delineation of boundary conditions under which persons hesitate,
defer, or choose options that require no action on their part or no
change to the status quo and (b) the explanation for that behavior.
I proceed first by defining more precisely what is meant by the
concept of decision avoidance and how it relates to similar every-
day and technical terms. I then describe some initial principles
regarding this phenomenon and the methods used for arriving at
the inferences put forth in this article. Decision avoidance is then
explored through a model that postulates both rational and emo-
tional sources of avoidance. I conclude with an agenda for future
research, a recapitulation of the insights gleaned from this review,
and implied recommendations for indecisive persons.
Conceptual Analysis: Forms of Decision Avoidance
It is important at the outset to explain just what is meant in this
review by decision avoidance, as it includes several phenomena
not explicitly related in the research literature, yet excludes several
concepts that one might think of as synonymous with decision
avoidance.
Decision avoidance manifests itself as a tendency to avoid
making a choice by postponing it or by seeking an easy way out
that involves no action or no change. This concept is derived from
the earlier notion of a decision attitude: “the desire to make or
avoid decisions, independent of any consequence that they
achieve” (Beattie, Baron, Hershey, & Spranca, 1994, pp. 129–
130). In this regard, decision makers could be either decision
Christopher J. Anderson, Department of Psychology, University at Al-
bany, State University of New York.
This article was completed in partial fulfillment of the requirements for
Christopher J. Anderson’s doctoral degree.
I thank the members of my qualifying examination committee, Jim
Jaccard, Jim Neely, and Tram Neill. I am likewise indebted to Hal Arkes,
Jon Baron, Terry Connolly, Lijun Ji, Jason Riis, Neal Roese, Eldar Shafir,
Michael Tsiros, Frank Yates, and Marcel Zeelenberg, all of whom con-
tributed helpful information pertaining to their research. I would also like
to thank the members of the Department of Social and Economic Psychol-
ogy at Tilburg University in the Netherlands and the Decision and Policy
Sciences Group at Rockefeller College, State University of New York, for
their comments and questions regarding talks based on this article.
Correspondence concerning this article should be addressed to Christo-
pher J. Anderson, who is now at the Department of Social and Economic
Psychology, Tilburg University, P.O. Box 90153, 5000-LE Tilburg, the
Netherlands. E-mail: ca4809@albany.edu
Psychological Bulletin Copyright 2003 by the American Psychological Association, Inc.
2003, Vol. 129, No. 1, 139–167 0033-2909/03/$12.00 DOI: 10.1037/0033-2909.129.1.139
139
seeking or decision averse, depending on the context (as con-
ceived, it does not represent an individual difference).
Marked preferences for avoidant options have been discovered
in diverse areas of the literature; humans generally prefer no
change (status quo bias, Samuelson & Zeckhauser, 1988), no
action (omission bias, Ritov & Baron, 1992; inaction inertia,
Tykocinski, Pittman, & Tuttle, 1995), and delay (choice deferral,
Dhar, 1996) more than an initial analysis would indicate these
options warrant. It is with these four phenomena that I most closely
identify decision avoidance.
Decision avoidance is not the same as procrastination, although
there is relevant research on decisional procrastination, an individ-
ual difference in the tendency to defer decisions. For one, this
article is primarily concerned not with individual differences but
with general human tendencies and the conditions (environmental
and cognitive) that bring them about. An analysis complementary
to that presented here could focus on individual differences in
decision avoidance and contributors to those differences (there is
less research in this area, but it is a valuable line of inquiry).
Secondly, procrastination involves having an intention at some
level but then acting contrary to that intention by stalling (Sabini
& Silver, 1982). Decision avoidance, on the other hand, could be
consistent with a decision maker’s intentions. The literature in the
area known as decision avoidance does not strongly suggest oth-
erwise at this juncture.
Patterns of responding to difficult situations that are similar in
many respects to decision avoidance have been investigated within
the conflict model of decision making under the terms defensive
avoidance and unconflicted adherence (Janis & Mann, 1977a,
1977b). Unconflicted adherence occurs when there do not appear
to be significant risks if one maintains his or her current course of
action in a challenging situation. Instead of responding to a threat
by considering their alternatives in a systematic manner and mak-
ing a decision, individuals in this case tend to adhere to the status
quo in an unreflective manner. Defensive avoidance occurs when
there may be risks to maintaining the status quo but the prospects
for discovering better alternatives appears grim. The defensive
response takes several forms: evasive, in which reminders of the
decision are ignored and distractions are sought; buck passing, in
which responsibility for the decision is shifted to others; and
bolstering, in which the decision maker seeks reasons, in a biased
manner, to support an inferior course of action. Phenomena similar
to these concepts are revealed in the discussion of modern research
on decision avoidance (see Decision Avoidance Phenomena sec-
tion). However, the concept of decision avoidance differs in that it
applies beyond threatening conditions into more mundane deci-
sions. It also occurs under conditions beyond those specified in the
conflict model as the antecedents of defensive avoidance and
unconflicted adherence.
The studies reviewed here focus on decision-averse behavior,
and the understanding of them herein as expressions of a decision-
avoidant attitude is consistent with Beattie et al.’s (1994) sugges-
tion that omission bias may indicate a general decision aversion. I
have included three additional phenomena and suggest that deci-
sion avoidance is the dominant attitude of human decision makers
within certain boundary conditions, which I discuss in the context
of the model.
The forms of decision avoidance currently lack definition and
common explanation in the literature, and they are not part of an
integrated understanding of indecisive behavior. Unfortunately,
decision avoidance is precisely the type of behavioral phenomenon
of high social importance that is likely to be overlooked given the
paradigms and division of labor among the subdisciplines of
psychology. It is not attended to by clinical psychologists, for
although it is pervasive, it infrequently rises to the level of a
disorder, and if it does, it can be subsumed under some other
diagnostic category, such as anxiety disorders. It also has escaped
the attention of cognitive and social psychologists, as it is not the
type of behavior that fits easily into their programs of research.
Even the area of judgment and decision making had not addressed
the issue until recently, as the theoretical underpinnings of that
field tend to emphasize choice between several positive options
and did not explicitly represent the additional option people often
have—not to choose.
First Principles
Two important principles need to be recognized in any investi-
gation of decision avoidance. These are meant to be limiting and
clarifying principles rather than assumptions that ground conclu-
sions reached later.
The first important principle to recognize is conservation of
energy. Psychologists and biologists generally have focused, not
on the factors that keep an organism dormant, but rather on which
factors motivate or energize organisms to move and expend en-
ergy. There is validity to this approach; generally when one ob-
serves an organism doing nothing, the explanation is simple and
uninteresting: the organism is resting or conserving energy for
action when an appropriate opportunity or need presents itself.
Insofar as decision avoidance is concerned, it is thus important to
recognize at the outset that apparently avoidant behavior involving
inaction may occur simply because the decision maker does not
recognize that an opportunity has presented itself or that there is a
need to make a decision. One may simply be maintaining energy
for a later time rather than deliberately avoiding a decision.
The second principle concerns the relevant antecedents of de-
cision avoidance and is termed multiple causation. It is reasonable
to consider that a complex and common behavior, such as decision
avoidance, may have more than one cause. It also is likely that
many observed antecedents of decision avoidance have their in-
fluence through mediating variables. Finally, I consider it likely
that many of the antecedents of decision avoidance are sufficient,
but not necessary, causes. Thus, I have sought to make the model
of decision avoidance as simple as possible, but no simpler than
that.
Method of Investigation
The primary goal of this article is to establish the concept of
decision avoidance as a class of behavior that is revealed in several
literatures and to integrate what is known about the motivation for
decision avoidance. Having clarified some conceptual foundations,
I proceed by building a path model of decision avoidance. First, I
identify existing and unintegrated potential decision-avoidance
phenomena through literature review. Having identified the rele-
vant phenomena discovered to date, I examine the literature for
identification of the antecedents and consequences of these forms
of decision avoidance. Special attention is paid to shared causes,
140
ANDERSON
potential redundancy in terminology, potential influence of manip-
ulated variables through other mediating factors, and the identifi-
cation of potentially irrelevant antecedents considered in earlier
literature. The last step covered in this article is the derivation of
a path model of decision avoidance, its antecedents and conse-
quences. Finally, an important step beyond the current purview is
to conduct further research to test and/or correct this model and to
identify any other potential forms of decision avoidance currently
unknown to science. Some suggestions regarding directions for
these investigations are made in the concluding remarks.
A Rational–Emotional Model of Decision Avoidance
For most of its history, judgment and decision-making theories
in both psychology and economics have been characterized by the
consequentialist mode of thought; expected utility theory is the
prime example of a consequentialist theory. From this perspective,
decision making is a computational process that operates solely on
subjective probabilities and expected outcomes of different op-
tions. Feelings may be generated as by-products of this computa-
tional process but do not affect decisions in a substantial manner.
Recently, theorists have begun to disavow the limitations of this
model and propose nonconsequentialist theories in which emotions
do influence choice (e.g., Loewenstein, Weber, Hsee, & Welch,
2001). In fact, some of the research embedded in this perspective
is reviewed herein. It is worth noting, however, that the effects of
anticipated emotions such as regret could be incorporated into a
consequentialist model. It may be proposed that anticipated regret
is just another attribute on which options vary and that this at-
tribute is weighted and incorporated into evaluation just as any
other attribute, without the need for postulating other processes.
There are several problems with this alternative. First, it requires
the abandonment of basic axioms of most consequentialist models.
The most prominent examples of this are the violations of transi-
tivity that have been observed when anticipated regret is manipu-
lated. Because regret is a highly context-sensitive attribute, it
becomes difficult to order preferences reliably when it is taken into
account; preferences depend highly on the options presented in a
particular case (Loomes, Starmer, & Sugden, 1991, 1992; Redel-
meier & Shafir, 1995; see also Starmer & Sugden, 1996, 1998, for
an alternative interpretation). Although a consequentialist model
can incorporate emotion, it may not be able to do so and remain
normative in the classical sense.
1
Second, this alternative is insufficient to account for several
findings in the area of decision avoidance. As Loewenstein et al.
(2001) indicated, it is not just anticipated emotions but also antic-
ipatory emotions that can have an effect on risky choice. Antici-
patory emotions are experienced during the decision process and
include states such as fear, anxiety, and dread. Phenomenologi-
cally, these mental states refer to potential future outcomes, as
anticipated regret does, but the emotional experience itself occurs
in the present rather than in a mentally simulated future. The
nonconsequentialist model of emotions in decision making under
risk postulates that emotions interact with computational processes
and can also mediate choice directly. Although this model is
reasonable, a slightly different framework is more suitable to the
current understanding of decision avoidance, as per Figure 1. This
consequentialist model is almost identical to the model advanced
by nonconsequentialists, with two important differences. First,
note that emotions have no direct effect on behavior. This is
important because it identifies the model as consequentialist: Emo-
tions have their role in choice either because one is attempting to
reduce future negative emotion and is treating that goal as an
attribute of the options or because one is experiencing unpleasant
anticipatory emotions (e.g., fear, dread) and hopes to cope with
those emotions by selecting a particular behavioral option. Both
1
The other alternative for this sort of model is to recognize that alter
-
natives are more than their labels. Emotions must be recognized as real
properties of options as they are perceived by decision makers; the value of
an option cannot be seen as fixed and context independent because context
influences emotions, and fluctuating emotions must be allowed to be part
of utility to retain transitivity (normative status) in a consequentialist
model. Essentially, this view asserts that options with the same objective
description are “different options” in different contexts because they strike
the decision maker differently. In either case, something must be given up:
context independent utility or transitivity.
Figure 1. An alternative consequentialist model of emotional influence on decision making in which concur-
rently experienced emotion is not epiphenomenal (cf. Loewenstein et al., 2001). This constitutes a background
assumption for the more detailed rational–emotional model of decision avoidance. Incl. ⫽ includes.
141
DECISION AVOIDANCE
kinds of emotion, anticipated and anticipatory, could be factored
into choice in a fully computational, goal-based manner. In this
regard, another important difference between this model and more
traditional models such as those underlying subjective expected
utility (SEU) theory is that not acting or choosing is explicitly
defined in the research as a behavioral option, often used by
decision makers to reduce anticipatory emotion (e.g., Luce 1998;
Luce, Bettman, & Payne, 1997).
A model of variables that influence decision avoidance is pre-
sented in Figure 2; this is the specific rational–emotional model of
decision avoidance.
2
It is built on the consequentialist–emotion
assumptions previously discussed: Rational inferences based on
probability and outcome information are incorporated in a decision
along with anticipated and anticipatory emotion. Those emotional
influences have their impact because they bear on affective goals,
which a decision maker has for managing present and future
emotional states.
The rightmost column of Figure 2 represents specific emotional
outcomes assumed in many theoretical analyses to result from
decision avoidance, although these variables are rarely measured.
The previous column represents the three forms of decision avoid-
ance identified in this review (inaction inertia is not listed, as it
overlaps with these forms). The column prior to that represents the
variables postulated to have a direct impact on decision avoidance.
A proposed motivation for many acts of decision avoidance is to
regulate one’s emotional state. To this end, two emotional out-
comes are identified in the model. Although it is true that for any
given decision any number of outcome dimensions may be af-
fected by decision avoidance, these are the most general and
relevant effects. It is interesting to note that the vast majority of
studies support the conclusion that emotional goals influence de-
cision avoidance but that postdecisional emotions are infrequently
measured. This is probably due to the fact that most researchers
study anticipated emotion (a cognitive evaluation manipulated by
the researcher) or presently experienced anticipatory emotions and
frequently conclude that, by manipulating those variables and
altering patterns of decision avoidance, emotions can cause deci-
sion avoidance. It is reasonable to assume that people make
choices that reduce negative emotions. Nonetheless, it is important
for future research to explicitly measure the effect of decision
avoidance on later emotional outcomes.
Decision Avoidance Phenomena
I begin my presentation of the rational–emotional model by
discussing research on four decision avoidance phenomena iden-
tified in research literature. These phenomena, which have been
independently discovered and discussed by researchers, are then
revealed to be closely tied together by shared causes and effects.
2
A distilled version of this path model is presented in the current article;
a more detailed version reaching further back into earlier antecedents is
available from Christopher J. Anderson.
Figure 2. A rational–emotional model of decision avoidance, presented in the form of a path diagram. The
primary subjects of the investigation, the forms of decision avoidance, are shown in capital letters. Bold text
indicates emotional influences, which have thin arrows to indicate that other processes and situational factors in
turn influence these emotions.
142
ANDERSON
Status quo and omission biases. Human decision makers have
a tendency to prefer options that cause no change in the state of the
world (the status quo) and/or require no action on their part
(omissions; Ritov & Baron, 1990). The status quo bias is a deci-
sion maker’s inflated preference for the current state of affairs.
Two observations serve as criteria to establish the finding of a
status quo bias: first, that a large majority of persons repeat initial
choices in successive decision situations and, second, that this
occurs in spite of changing preferences. Repeated choice of one
option is unsurprising if that option continues to serve goals well.
Research provides abundant evidence for such a bias in both
hypothetical decision tasks and everyday life. As an example of the
former, a set of identical alternatives were distributed to college
students, except that some students had an option (that could be
any of the original options) labeled as the status quo (Samuelson &
Zeckhauser, 1988). For the vast majority of questions and com-
parisons, the status quo option dominated both the control neutral
options seen by a different group and the non-status quo options
within their own group. Field studies provide evidence of this
behavior in real-world decision making; for example, a greater
than expected number of persons opt to keep the same allocation
of retirement funds year after year even though they know there is
no cost for changing, and persons just beginning a retirement fund
invest in different ways that reflect the same preference changes,
even within the same age group (Samuelson & Zeckhauser, 1988;
see also Schweitzer, 1995). This research provides a challenge to
SEU theory, although status quo choices can be quite “rational”
when (a) preferences are unchanged, (b) there are costs for change,
or (c) there is uncertainty regarding the consequences of non-status
quo options. The aforementioned studies found evidence for a bias
in the absence of such concerns.
Closely related to this is the omission bias, an inflated prefer-
ence for options that do not require action (Spranca, Minsk, &
Baron, 1991). Ritov and Baron (1992) argued that status quo bias
is an extension of omission bias. In reviewing the literature, they
found that in all cases, the status quo option had been confounded
with the omission option. When questions were constructed so as
to separate these two dimensions of the status quo (i.e., unchanging
circumstances and inaction), individuals preferred the option that
required no action, regardless of whether it represented the status
quo or a change.
Some initial evidence (Ritov & Baron, 1992) favored the inter-
pretation that these two biases are a unitary phenomenon and that
observed status quo biases are dependent on one’s partiality for
choices that forgo action. However, Schweitzer (1994) produced
evidence for independent status quo and omission biases, a posi-
tion that Baron and Ritov (1994) supported. The biases were
established as independent by responses to hypothetical questions
that manipulated omission and status quo options independently.
Instead of choosing an option, participants assigned preference
ratings to all options. The degree of bias was established by
comparing proportions of preference for a given option on whether
it was labeled the status quo, omission,orneutral option
(Schweitzer, 1994). When this procedure is followed, exaggerated
preferences for options representing both status quo (without omis-
sion) and omission (without status quo) are observed. The biases
are thus independent of each other, and they appear to be relatively
equal in magnitude. Of course, in the majority of real-world cases,
the two biases work in concert, and the fact that they occur
independently does not rule out the possibility that they share
underlying causes. When preferences for omission and status quo
are correlated within subjects, the correlation is positive and sig-
nificant, indicating that the more one prefers the status quo, the
greater the preference for omission, and vice versa. This suggests
that the two independent effects may occur for the same reasons.
Why do individuals prefer inaction and the status quo when this
may interfere with their goals? After all, if a non-status quo–
omission option is preferable in a neutral context, the indication
should be that one’s goals would be better met by selecting that
option, regardless of whether some nonpreferred option is the
status quo or default position in another context.
First, I examine these situations from the vantage point of regret
theory. Norm theory (Kahneman & Miller, 1986) can be consid-
ered an adjunct to regret theory for the current purposes. This
theory states that emotional responses to the outcomes of decisions
are amplified by “abnormal causes,” and because actions are
abnormal causes, it thus seems likely that deviations from the
status quo are also, psychologically, abnormal causes. Emotional
responses are amplified to omissions and deviations from the
status quo because the alternatives to these options are easier to
mentally simulate. Thus, norm theory is a regret theory: Counter-
factual thinking about decisions is tied to an affective response
presumed to influence choice. Norm theory predicts omission bias
because individuals anticipate more potential regret as a result of
commissions, and they incorporate regret avoidance into their
valuation of the options.
3
However, norm theory’s explanation alone is inadequate. If
potential affective responses are enhanced by actions, decision
makers would also have to take into account the increased rejoic-
ing they would experience from a good outcome to an action
option. The enhanced rejoicing and regret should cancel out,
leaving the omission and commission options equal in expected
affective value. For this reason, it is necessary to supplement this
regret theory of omission bias with the assumption of loss aversion
described by prospect theory (Baron & Ritov, 1994). Loss aversion
describes people’s tendency to weight potential losses greater than
potential gains of the same amount. If this assumption holds true,
the utility of potential rejoicing experienced as a result of an action
would be less than that for potential increased regret, leading to an
increased preference for the omission option.
Riis and Schwarz (2000) advanced an alternative explanation
for status quo selection similar to the trade-off avoidance hypoth-
esis advanced for choice deferral below. As with deferral choices,
status quo options may be seen as less threatening and could thus
serve to reduce negative emotion that is experienced prior to
making the choice (termed anticipatory emotions). They attempted
to demonstrate the impact of this affective state on choice in a
series of studies requiring participants to make consecutive deci-
3
It might seem difficult to extend this explanation to omission bias in
judgments of others’ decisions. Two considerations come to mind: First, in
the context of this article, I am concerned with people’s decisions and what
motivates them, not with their judgments of others, which could be based
on different factors. Second, regret could be relevant to judgments of
others’ decisions if the person making the judgments assumes others are
concerned with regret and feel regret in the same situations as he or she
does and thus base judgments of some dimension partially on these
assumptions.
143
DECISION AVOIDANCE
sions. As predicted by the emotion hypothesis, individuals in-
creased their selection of the status quo option when they had to
make a second affective choice after making an initial difficult
choice that produced negative emotion. On average, selection of
the status quo option increased by 77% after an emotionally
aversive initial decision. The direction of the effect was consistent,
and it appears to be a robust finding.
Although these conflicting hypotheses might seem contradic-
tory, they are congruent with the rational–emotional model. Se-
lection difficulty, which is sometimes due to negative emotion, can
cause selection of avoidant options, as can anticipated regret. Both
are sufficient, but not necessary, causes of decision avoidance.
There is some evidence linking status quo choice to emotional
outcomes. The clearest example comes from a study (Luce, 1998)
that manipulated the difficulty of trade-offs to be made in a
decision as well as whether an avoidant option (in some cases, the
status quo) was available. Trade-offs on valued attributes increased
negative emotion, whereas the presence of avoidant options tended
to decrease overall negative emotion. The increased emotion pro-
voked by value trade-offs was more pronounced if no avoidant
options were present. When actually making a choice, an increased
preference for the status quo and other avoidant options was
present only if difficult trade-offs were also present. Participants
who made those choices decreased their composite negative
emotion.
A second source of evidence linking status quo choice to emo-
tional outcomes comes from a study (Inman & Zeelenberg, 2002)
that manipulated prior experience with choices with a repeat or
switching of that choice by a hypothetical individual. When the
individual chose to repeat the choice (i.e., selected the status quo),
and when the prior outcome of that decision was positive or no
information was available about its outcome, participants rated that
person as feeling less regret about a bad outcome. This indicates
that, by default, or when the status quo led to good outcomes,
choosing the status quo was believed by these participants to lead
to less regret. It is interesting, however, that status quo selection
led to more regret when prior outcomes of the same status quo
decision were undesirable. However, the general finding across
Inman and Zeelenberg’s (2002) studies demonstrates that, other
things being equal, selecting the status quo can reduce postdeci-
sional regret.
Omission, on the other hand, has been explicitly linked to
differences in regret in a growing body of research (e.g., Kahne-
man & Miller, 1986; Zeelenberg, van den Bos, van Dijk, & Pieters,
2002). Decision makers tend to associate action with more regret
than inaction; this has been referred to as the action effect, the
exaggeration effect, and emotional amplification (e.g., Kahneman
& Miller, 1986; Zeelenberg et al., 2002) in the research literature.
There are some boundary conditions for this effect and it merits
some review. However, individuals in these studies tend to make
judgments about the regrets of hypothetical persons in a vignette;
therefore, the results are most validly discussed under the heading
of anticipated regret, and cautiously generalized to this component
of the model.
Choice deferral. A situation in which an individual chooses
not to choose for the time being is a choice deferral. This includes
taking time to search for better alternatives, choosing not to pur-
chase any of a variety of options, or avoiding responsibility for the
decision altogether. The effects discussed in this category involve
decision avoidance characterized by postponing decision or refus-
ing to select an option.
Degree of conflict plays a role in choice deferral; this is signif-
icant because conflict is a context-dependent variable that has no
causal potency in SEU theory. One way researchers manipulate
conflict is by varying the relative levels of attractiveness of two
competing options. Higher degrees of conflict are associated with
a greater willingness to prolong search and with a refusal of any
option. This creates behavior that seems paradoxical from a sim-
plistic interpretation of SEU theory. Consider the following result
of Tversky and Shafir (1992): Decision makers did not prefer
choice deferral when a single option was presented. However,
when an additional option was made available, the share of the
choice deferral increased markedly, although very few participants
found the new option preferable to the original. This contradicts an
important assumption of SEU theory: Each option has a value, and
a decision maker chooses the option in a set with the highest value.
This principle, termed value maximization by the authors, implies
that what individuals choose in this situation should not be affected
by adding an option that is considered inferior.
Likewise, Redelmeier and Shafir (1995) found that individuals
often chose options they otherwise would have declined when
additional options were made available. For example, participants
who were doctors had to choose between treating two very differ-
ent patients, A and B. In this situation, the patients’ needs and the
context were such that most doctors chose to treat Patient B.
However, if Patient C, who was highly similar to B, was added to
the list of options, but only one of the three can be treated, the
doctors preferred to treat Patient A. In similar scenarios, individ-
uals preferred a status quo, default, or distinctive option (as in the
above example; Redelmeier & Shafir, 1995). The authors sug-
gested that this pattern obtains because the additional options
increase decisional conflict. Decision makers desire justifications
for choices they make, and these become more scarce as the
number of options increases. Conflict makes justification more
difficult and, as a result, leads people to seek options that reduce
their responsibility for the choice, such as deferral or the status
quo.
Redelmeier and Shafir (1995) did not have a theoretical reason
to predict deferral. Because the experience of conflict is excluded
from explanation within SEU theory, the focus is on finding an
effect incompatible with that theory to demonstrate that conflict
does have an effect on actual human decision making. The hy-
pothesis is that choice deferral occurs as a result of higher levels of
conflict; it is unclear whether conflict operates on selection by
making the justification of any particular option more difficult,
thus making deferral options more attractive, or whether lack of
justification increases choice conflict, which motivates escape or
avoidance of responsibility provided by a deferral option.
An alternative theory has been advanced that focuses on the
emotions experienced during the choice situation but shares with
the conflict hypothesis and the preference uncertainty hypothesis
(discussed next) a focus on broadening choice theories to include
factors grounded in the psychological experience of decision mak-
ing (Luce, Bettman, & Payne, 1997, 2001). This hypothesis is
grounded in the finding that conflicts in a decision, such as those
that put different values at odds, often lead to negative emotion
(Luce, Bettman, & Payne, 1997, 2001). Luce (1998) claimed that
the selection of avoidant options may serve to reduce decision-
144
ANDERSON
related negative emotion in addition to serving efficiency and
accuracy goals. This position is labeled a trade-off avoidance
hypothesis because the primary source of the negative emotion to
be reduced is the process of making compromises (trade-offs)
between options on attributes that are highly valued (i.e., are
weighted more heavily than others).
In work on trade-off avoidance, three types of decisions are
considered avoidant: status quo choices, prolonging search for
options, and choosing an alternative that dominates at least one
other option in the choice set. Consistent with the theory, increases
in trade-off difficulty led to increased negative emotion in Luce’s
(1998) Experiment 1. Furthermore, as trade-off difficulty in-
creased, choice shares of the three avoidant options increased. In
another study (Luce, 1998, Experiment 2), giving participants
imagery instructions served to increase negative emotions experi-
enced in a difficult choice situation and increased the percentage of
participants choosing an avoidant option. Avoidant choices were
generally preceded by more negative emotion and also were fol-
lowed by less negative emotion than other choices. This demon-
strates that emotion maintenance may be among the goals involved
in choice and that the emotion being maintained may be task
related. It is important to differentiate this influence from that of
emotions anticipated as a result of choosing a given option (either
in terms of its contribution to subjective utility or to anticipated
regret and rejoicing) or that of ambient emotions, which precede
the choice context and stem from outside of it (cf. Isen, 2001).
Dhar and colleagues (Dhar, 1996, 1997a, 1997b; Dhar & Now-
lis, 1999; Dhar, Nowlis, & Sherman, 1999; Dhar & Simonson,
1999) have produced an abundance of research advocating a
different position: that preference uncertainty produces the deci-
sional conflict (selection difficulty) that precedes choice deferral.
This hypothesis is based on the assumption that selection decisions
have primacy over deferral decisions; before deciding whether to
defer, an individual first attempts to find an acceptable or superior
option to select that will finalize the decision. Their studies re-
vealed that individuals given a choice between options with com-
mon good and unique bad features were more likely to defer
choosing than those faced with a pair of options with common bad
and unique good features (see, e.g., Dhar & Nowlis, 1999). Con-
trary to the conflict and trade-off avoidance hypotheses, this result
is interpreted as being due to differences in the attractiveness of the
options in the choice situation. Dhar and Nowlis (1999) claimed
that the individuals process the choice in such a way that common
features cancel out and that attention is focused on unique features.
Thus, the valence of the unique features can drive the attractive-
ness of the set of options. When the choice set has low attractive-
ness, individuals are expected to defer choice because none of the
options appears to satisfy their goals for the choice.
This interpretation led researchers to pursue a line of research
based on the idea that choice deferral often occurs because of
preference uncertainty, a state in which decision makers do not feel
they can determine with accuracy which option best meets their
goals. This also runs contrary to SEU theory because it focuses on
the idea that values may not be stable or fully determined before
entering the choice situation. The idea that elements of the choice
situation itself can alter preferences violates the assumptions of
SEU theory. Dhar (1997a) argued that preference uncertainty leads
to more hesitation and that uncertainty can be increased by options
with small differences in attractiveness. Supporting this notion,
Dhar (1997a, Experiments 1 and 2) found that individuals defer
choice more often when there are small differences in attractive-
ness but that there were no effects of the number of trade-offs
when attractiveness was held constant, contradicting conflict and
trade-off difficulty explanations. In a study using verbal protocols,
Dhar (1997a, Experiment 3) found that individuals produced a
greater number of thoughts when there were small differences in
attractiveness and that participants selecting deferral had more
balanced evaluations of the options. This supports the idea that
decision makers actively attempt to determine preferences in these
situations and defer when their preferences are more uncertain.
Further supporting this hypothesis are Dhar’s (1997a) findings that
asking participants to evaluate features of options led to more
deferral (Experiment 4), participants allowed to choose multiple
options from a set showed less choice deferral (Experiment 5), and
allowing individuals to practice a compensatory decision strategy
(learning how to manage trade-offs) decreased choice deferral
(Experiment 6).
Dhar’s (1997a) interpretation given to these findings is that
evaluating features forces decision makers to consider their pref-
erences more carefully and thus causes more uncertainty. Choos-
ing multiple options allows participants to make a decision while
maintaining preference uncertainty. Finally, using a compensatory
decision strategy aids individuals in choosing by assigning a
unidimensional value to the options, exposing their preferences.
However, it is worth noting that many of Dhar’s (1997a) findings
are not necessarily incongruent with the trade-off avoidance hy-
pothesis. Many of the manipulations could have their impact on
choice through task-related emotion; it is impossible to rule out
that possibility with the available data.
In a subsequent study, Dhar et al. (1999) found that the inci-
dence of deferral depends on which features are emphasized in an
initial comparison of the options. The comparison task is assumed
to influence preference judgments, making a choice between op-
tions with unique bad (and common good) features seem worse
after a dissimilarity judgment. Contrasting the alternatives in this
case leads to increased focus on the negative features, decreasing
the attractiveness of the choice set. Thus, deferral is increased by
unique bad pairs compared with unique good pairs after individ-
uals have made a dissimilarity judgment. A decision-tracking
study in which Dhar et al. (1999, Experiment 2) were able to
collect data on the features of options that participants viewed and
on how long they viewed them supported this interpretation.
Dhar and Nowlis (1999) focused on the roles of decision pro-
cesses and conflict in choice deferral rather than on their roles in
preference uncertainty. Recall that earlier research (Redelmeier &
Shafir, 1995; Tversky & Shafir, 1992) indicated that conflict might
be important in producing deferral choices and that experiments
(Dhar, 1997a) also had implied that the difference between com-
pensatory and noncompensatory decision processes might be im-
portant. Dhar and Nowlis found that increased conflict is associ-
ated with increased incidence of choice deferral. They also
discovered that choice deferral decreased under time pressure for
pairs of options designated as high conflict. This is important
because individuals tended to resort to noncompensatory decision
strategies under time pressure, effectively making the selection
decision easier. These researchers assumed that participants first
made the selection decision and then chose whether to defer
choice. Thus, they hypothesized that the difficulty of the selection
145
DECISION AVOIDANCE
decision could affect the deferral decision; if selection was easy,
they predicted, an individual would not defer. The effect of time
pressure seems to have occurred through the route of decision
strategies; individuals tended to pay more attention to unique
features, use noncompensatory strategies, and experience deci-
sions as easier and less conflicted in time-limited conditions.
Dhar and Nowlis (1999) also found that deferral is less likely in
approach–approach conflicts (decisions between two attractive
choices) than in avoidance–avoidance conflicts (choices between
two unattractive options). This supports their earlier notion that the
overall attractiveness of a choice set is important and implies that
conflict is important because of its influence on preferences.
Three different explanations have been proposed for various
findings involving choice deferral, each detailing processes and
situational variables considered unimportant from the vantage
point of SEU theory. Conflict theory states that individuals have
more difficulty justifying decisions under conflict and thus choose
deferral and status quo choices. Trade-off avoidance hypothesis
claims that difficult choices stimulate negative emotions, which
individuals seek to reduce by choosing avoidant options, including
the status quo and deferral of decision. Finally, preference uncer-
tainty hypothesis states that individuals do not always have clearly
defined preferences and that small differences in attractiveness can
increase this uncertainty. Preference uncertainty means that selec-
tion is difficult, and individuals tend to defer more often when
selection is problematic. This difficulty may be compounded by
the use of compensatory decision strategies. This theory has also
been extended to incorporate an understanding of how conflict
affects deferral; conflict seems to operate on deferral by increasing
selection difficulty. However, it is not the conflict per se that
produces deferral, as its effect disappears under time pressure.
These explanations are not mutually exclusive, and there is no
particular advantage to be gained by treating them as if they were.
In fact, there is considerable overlap between the positions. The
conflict, trade-off avoidance, and preference uncertainty effects
can be consolidated by incorporating a common mediating vari-
able, selection difficulty, into the analysis, which I discuss in the
Antecedents of Decision Avoidance section.
Inaction inertia. Although these phenomena have not been
considered related in previous research, the last finding I look at
also concerns a reluctance to take action. Inaction inertia refers
specifically to the tendency of a person to omit action when he or
she already has passed up a similar, more attractive opportunity to
act (Butler & Highhouse, 2000; Tykocinski & Pittman, 1998,
2001; Tykocinski et al., 1995). This occurs especially when the
subsequent opportunity is somehow less attractive, even if it still
represents a gain from one reference point. In one demonstration
(Tykocinski et al., 1995, Experiment 1), some participants missed
a hypothetical opportunity to buy a ski pass for $40, another group
missed an $80 price, and a third group had no initial opportunity.
All participants then had the opportunity to buy a pass for $90,
which still represents a 10% savings compared with the usual price
of $100. Participants in the large difference condition rated them-
selves as least likely to purchase the ticket. This effect has also
been demonstrated with nonfiscal costs and in an experimental
rather than a hypothetical setting (Tykocinski et al., 1995).
Tykocinski et al. (1995) have interpreted this finding as due to
a form of counterfactual thinking. Counterfactual thoughts are
alternative representations of events (i.e., outcomes that could have
occurred but did not; Roese, 1997, 1999), such as “If I had chosen
to quit smoking 5 years ago as I planned, perhaps I would not be
diagnosed with lung cancer today.” This particular example is one
of upward counterfactual thinking (imagining a better outcome),
whereas downward counterfactuals concern worse outcomes than
those experienced by the thinker. Upward counterfactuals are
associated with regret, which figures prominently in the standard
explanation of inaction inertia.
The anticipated regret explanation of inaction inertia assumes
that, in general, persons do not consider the merits of current
opportunities independently of past choices. Individuals in the high
discrepancy condition recalled the prior, much better opportunity
they had missed (Tykocinski et al., 1995). These people knew that
if they took the current offer, it would be easy to construct an
upward counterfactual for the situation and experience concomi-
tant regret. Anticipating this, they bypassed the subsequent oppor-
tunity—regardless of its merits—to avoid regret. Alternatively,
participants may experience regret over the initial decision and
seek to decrease or avoid amplifying that level of regret (M.
Zeelenberg, personal communication, September 21, 2001).
Although anticipated regret is a reasonable explanation, it is
valuable to examine how it was established. Tykocinski and col-
leagues (1995) used a disconfirmation strategy, investigating four
alternative hypotheses: perceptual–price contrast, cognitive disso-
nance, self-perception, and commitment.
The perceptual–price contrast explanation claims that inaction is
due simply to the knowledge of a lower cost, which correspond-
ingly devalues the current opportunity. Inconsistent with this hy-
pothesis, Tykocinski et al. (1995, Experiment 3) demonstrated that
inaction inertia is not evident when there is no initial action
opportunity but a cost contrast is otherwise achieved. Several
further studies also disconfirmed a price contrast explanation (see
Tykocinski & Pittman, 2001).
Cognitive dissonance and self-perception theories offer expla-
nations of inaction inertia that are similar. According to these
theories, decision makers desire to maintain a certain image of
themselves, and to do so, they must explain certain events and
occasionally revise evaluations of aspects of the world. In the case
of inaction inertia, it may be the case that people feel the need to
explain to themselves why they failed to act on that initial oppor-
tunity. To do so, individuals might devalue the goal or offer in
question and decide they must not like it, which leads them to
likewise devalue the similar present opportunity to obtain the same
goal or object. If this explanation is correct, one would expect that
inaction inertia would be mitigated if the individual was not to
blame for missing the initial opportunity. In that case, the inaction
can be quite readily attributed to external sources, and there is no
need to revise one’s assessment of the goal or object. However,
manipulations of initial responsibility did not reduce or eliminate
inaction inertia in Tykocinski et al.’s (1995) Experiment 4. This
generalizes to a nonhypothetical context; participants who were
forced to miss an early, winning bet were equally unresponsive to
a later opportunity to bet on the same horse, compared with
participants who also missed the bet, but because of their own
choice (Tykocinski et al., 1995, Experiment 5). These results are
also considered to be troublesome for an account based on behav-
ioral commitment because personal responsibility is typically a
prerequisite for obtaining commitment-based effects.
146
ANDERSON
The positive case for an explanation of inaction inertia was
established in a subsequent article by Tykocinski and Pittman
(1998) that focused on avoidance of anticipated counterfactual
regret. The researchers demonstrated that the future avoidability of
the object or some aspect of the opportunity plays a role in inaction
inertia (Experiments 1 and 2). For example, imagine that a con-
sumer had an opportunity to buy a used car at a very low price and
then postponed the purchase. Later, the buyer wants the car and
returns to the seller, but the seller wants a somewhat higher amount
this time (even though the price is still advantageous to the buyer).
According to the anticipated regret explanation, inaction inertia
would set in because the buyer knows that if the car is purchased,
regret would be experienced, even though the current price is still
a good deal. But suppose that if the consumer does not buy the car,
it will be displayed prominently in a lot that the consumer drives
by on the way to work. This person is going to have to see the car
tomorrow and every day after regardless, so that by avoiding
purchasing the car, the individual is not effectively avoiding re-
minders of it. According to the regret interpretation, inaction
inertia would now be eliminated, because the regret cannot be
avoided no matter what the individual does. Therefore, the buyer
ought to examine the offer on the basis of its merits, independently
of the earlier opportunity missed. Likewise, Tykocinski and Pitt-
man (1998) found that unavoidability does significantly reduce the
inaction inertia effect.
More convincing is the result of another study (Tykocinski &
Pittman, 1998, Experiment 4), which showed that if the initial
missed opportunity is somehow made to look less attractive than it
previously did, regret over taking a related offer dissipates and
inaction inertia is diminished. To support their interpretation, the
authors provided a postexperimental survey of participants’ rea-
sons for their choices. Reports making mention of regret were most
prevalent in the group that did not receive information degrading
the initial opportunity (i.e., the group displaying the most inaction
inertia).
Is there any information that further compels the anticipated
regret hypothesis? Further studies by Tykocinski and Pittman
(2001) pitted the contrast hypothesis against regret. Costs were
held equivalent across conditions, and availability of a bonus and
the timing of missing the bonus were varied. Regret should have
been more likely when participants missed an opportunity more
recently, because it is then easier to construct counterfactuals
(while the price contrast is held constant). Inaction inertia was
reduced when the opportunity was missed longer ago, consistent
with the regret interpretation. Their Experiment 2 contrasts a
situation in which an individual postpones acting on an opportu-
nity and is faced with a less attractive offer because the original
offer was sold out with a situation in which the original offer was
a typographical error. Indeed, inaction inertia was reduced by this
manipulation; the magnitude of the effect was much larger in the
sold out group. Another investigation (Arkes, Kung, & Hutzel,
2002) found that when the earlier offer was in another country,
participants did not show the inaction inertia effect in their deci-
sions about whether to make a current purchase. Analyses of regret
with a separate group of participants demonstrated that no regret
was associated with the distant earlier offer, but regret was asso-
ciated with offers that were made nearby.
There are potentially contradictory results pertaining to the
regret hypothesis. Tykocinski and colleagues (1995, Experi-
ments 4 and 5; Tykocinski & Pittman, 2001, Experiment 1) also
produced evidence discrepant with the anticipated regret explana-
tion. In both of these studies, inaction inertia was obtained even
when the individual was not responsible for missing the opportu-
nity. Further research has found mixed results in the relationship
among responsibility, regret, and inaction inertia (Anderson,
2002a; Zeelenberg, 2002).
There is some disagreement among theorists as to whether
responsibility is a necessary component for regret; many argue that
it is essential (e.g., Bell, 1985; Gilovich & Medvec, 1995a; Sug-
den, 1985). Others argue that responsibility is not an essential
component of regret (Connolly, Ordonez, & Coughlan, 1997),
although even among this group it is noted that the degree of regret
often correlates closely with the degree of perceived responsibility
(Simonson, 1992; this issue is more thoroughly discussed in the
Anticipated regret section below). Thus, the regret hypothesis as
applied to inaction inertia needs to explain why someone would
still fail to act in a situation in which he or she is not responsible
for missing the initial opportunity. In addition, some preliminary
studies indicate that although regret is correlated with inaction,
differences in ability to think counterfactually (and thus in ability
to anticipate regret) do not appear to be related to inaction inertia
(Anderson, 2001a, 2001b; current research is concerned with es-
tablishing validity of the measurements of individual differences
and with clinical populations in which individual differences in
counterfactual thinking are known to occur). Thus, some of the
results appear to be somewhat inconsistent with the regret hypoth-
esis’s predictions. Regret may be a sufficient but not necessary
cause of inaction inertia, as postulated by the rational–emotional
model. These potentially anomalous findings pose less of a prob-
lem when inaction inertia is viewed in light of other forms of
decision avoidance.
There are also alternative and complementary views emerging
with regard to inaction inertia. One possibility is that lower priced
items lead people to devalue those items so that later opportunities
with higher prices do not seem as valuable, even though those
higher prices may be objectively lower than normal (Arkes et al.,
2002). Another possibility is that individuals are exercising self-
control because the earlier price leads them to believe that they can
obtain a better price if they can exert enough temperance to wait.
Although both of these possibilities may seem similar to the
contrast hypothesis disconfirmed above, neither of them should be
affected by the manipulations that ruled out a contrast explanation.
For example, making the earlier opportunity a typo leaves contrast
intact but might also lead participants to revise valuations and
relax the exertion of self-control, because there is no potential
better opportunity to wait for. These possibilities could be viewed
as independent explanations, or self-control motivation and dimin-
ished valuations might both be reinforced by regret. It would seem
that the concepts are more complementary than contradictory;
taken together, this explanation also suggests a possible functional
role for regret in reinforcing (reasonable) diminished valuations
and facilitating the exercise of self-control. Some preliminary
research (Anderson, 2002a; Zeelenberg, 2002) is supportive of the
notion that inaction inertia can occur in the absence of regret, as
predicted by the rational–emotional model or complementary self-
control explanation of inaction inertia but not by the pure antici-
pated regret explanation. Future research has many possibilities to
explore.
147
DECISION AVOIDANCE
Antecedents of Decision Avoidance
Preference stability. Among the clear and rational influences
on decision avoidance is preference stability, the degree to which
people’s values remain the same over time and thus their consec-
utive decisions. When preferences (measured independent of
choice) change less, one expects an individual to select the status
quo option more often. Choice-independent measures of prefer-
ence can be made through multiattribute utility analyses or through
a simpler method with the same goal, such as ranking options,
providing a scale rating of each option, or assigning dollar values
to options (a special case of scale rating).
Costs. Many instances of status quo selection may also be
attributable to a different rational reason. Changing the status quo,
in the real world, often entails transaction costs that can be avoided
by maintaining the status quo. For example, to alter one’s holdings
in the stock market, one must usually pay an agent to first sell the
stock currently owned and then pay another commission to buy an
alternative stock. Individuals presumably trade off the potential
gains to be made by diverging from the status quo with the costs
of changing; when those costs are higher, status quo selection is
more likely, ceteris paribus.
Analogous to change costs, but less frequently discussed, are
action costs, which should influence the tendency to choose a route
of inaction. As the costs required to take an action increase,
selection of an omission option should increase, ceteris paribus.
These costs are as pervasive, if not more so, than change costs
associated with status quo selection. This is the case because, in
addition to culturally or situationally imposed costs, there is also a
cost component reflecting expenditure of energy that is not asso-
ciated with a pure status quo change cost, which can require either
omission or commission.
Analogous to costs for changing the status quo and for taking
action, there are also costs for delay. Often, a decision maker faces
declining outcomes as decisions are postponed and therefore must
make a decision about when to decide by trading off the gains that
could be produced by further thinking and consideration with the
losses that could be incurred by deferring the choice. This is the
last of the purely “rational” influences in the model.
Anticipated regret. Anticipated regret is hypothesized to in-
fluence decision avoidant behavior; Table 1 presents a summary of
research suggesting such a link. The general hypothesis is that
individuals seek to minimize regret resulting from decisions and
that choice of an avoidant option is a domain-general vehicle for
avoiding regret. In other words, a higher level of anticipated regret
resulting from a choice over all of the available alternatives mo-
tivates a search for the option that minimizes regret; status quo
choices, omission choices, and choice deferral are likely candi-
dates for selection when anticipated regret is an attribute under
consideration. Status quo selection presumably decreases regret
because it provides an additional justification for a choice; when
choices are seen as more reasonable or justified, one experiences
less regret over them, even if the outcome is poor (Inman &
Zeelenberg, 2002). Omission appears to function as a response to
anticipated regret; for example, researchers have found that the
bias toward omissions is exaggerated by worse outcomes (Baron &
Ritov, 1994, Experiment 1). Other conditions that tend to produce
increased anticipated regret also produce an increased preference
for omission options (Kordes-de Vaal, 1996; Ritov & Baron,
1995). Alternatively, it might be argued that omission can increase
regret, given the finding that individuals tend to regret bad out-
comes more when they expend less instrumental effort toward an
outcome (van Dijk, van der Pligt, & Zeelenberg, 1999). Effort and
omission are not identical variables, but they should be highly
correlated. Thus, these results do pose a challenge to the general
theory. Further research is needed to disentangle these conflicting
findings.
Sometimes individuals are less likely to defer a choice when
regret is a factor, but these cases are the exception that proves the
rule. What occurs in these situations is that decision makers
anticipate more regret on the basis of deferring the option because
the only information they are given in these studies concerns the
immediately present option; therefore, they anticipate substantial
costs for delaying (Simonson, 1992). When information is obtain-
able on outcomes available if they had made a decision either
sooner or later, individuals tend to be more influenced by out-
comes that are available later. The increased impact of later op-
tions indicates that to avoid regret, deferral is a preferable protec-
tive option. Decision makers appear to know this, and they set
more stringent criteria for options when they expect to receive
information on outcomes available after their choice, leading them
to defer choosing longer (Cooke, Meyvis, & Schwartz, 2001).
Complement: Anticipated blame. There is an alternative
source of decisional causation in these situations that rests on the
same cognitive structures that anticipated regret does. There is no
research evaluating decision avoidance from the perspective about
to be outlined, but it has a clear logic and is worth consideration.
In the status quo and omission bias literature, it has been found
(Ritov & Baron, 1995, 1999; Spranca et al., 1991) that individuals
associate less wrongdoing with omissions, and presumably this
might hold for nonomission status quo decisions as well, although
this is unknown. For example, in one study (Spranca et al., 1991),
students acting as hypothetical jurors awarded larger sums of
money for damages to individuals who were harmed by a com-
mission, compared with individuals who experienced something
equally harmful as a result of an omission.
Most individuals in the various studies show evidence of hold-
ing an omission bias, and for some of them it is so pervasive as to
outweigh more important concerns, such as saving lives (Baron &
Ritov, 1990). Assume for a moment that individuals have, at some
level, an awareness of their own omission bias and the fact that
others share this bias, as might be the case if omission bias results
from overgeneralization of a social norm (Baron, 1994; alternately,
the social norm could be unconscious). If individuals know that
other people attribute less responsibility and wrongdoing to omis-
sions, they could be biased toward omissions, not out of any
consideration for their own regret, but out of the will to avoid the
blame of others for a bad outcome. This blame could result in
further damaging outcomes for the decision maker. The question
then becomes one of how important it is for the decision maker to
avoid damaging outcomes resulting from blame. Because the
blame resulting from omission bias is out of the decision maker’s
control, the question of the rationality of incorporating this emo-
tion into choice is less compelling.
Anticipated blame should involve the same cognitive processes
and obey the same rules as anticipated regret. Even if decision
makers refuse to consider their own potential regret as an impor-
tant variable, the blame resulting when others evaluate their deci-
148
ANDERSON
Table 1
Research Reports Suggesting an Influence of Anticipated Regret on Decision Avoidance (DA) by DA Form
Study DA measure
Manipulation of
avoidance regret Major findings and/or interpretation
Omission
Ritov &
Baron
(1990)
Risk tolerance, written
reports
Participant-reported reasons for preferring omission primarily regarded
responsibility and fault.
Spranca et al.
(1991)
Morality, written reports Relative outcome, intention,
degree of risk
Harmful omissions were perceived as less immoral than equivalent
commissions. Participants’ justifications emphasize distinction of
action versus omission, responsibility (diffusion of and a general
decrement for omissions). It follows that what is perceived as more
moral would be more justified and less regrettable.
Baron (1992) Normative beliefs, emotions Debiasing Normative beliefs interact with anticipated emotion; omission bias (OB)
reversed after logical debiasing argument that affected most
participants’ anticipated regret.
Baron &
Ritov
(1994)
Preference ratings,
comparisons
Relative outcome, gain/loss Preference for omission increased by worse outcomes, irrespective of
gain/loss domain.
Ritov &
Baron
(1995)
Choice, satisfaction, written
reports
Outcome feedback,
outcome overlap, gain/
loss
Biased preference for omissions dependent on bad outcomes and
information about foregone outcomes, less likely under high overlap
(less potential for regret based on comparisons). Verbal justifications
were consistent with avoidance regret theory.
Kordes-de
Vaal
(1996)
Responsibility, morality Causality, intention Preference for omission stronger for intentional, causally potent choices.
Ritov &
Baron
(1999)
Decision threshold Sel.: protected values Holding protected values (i.e., being reluctant to trade off) was
correlated with increased preference for omission. OB is a reliable
individual difference. Trading off protected values is likely to lead to
regret, so such individuals may be more prone to anticipated regret,
thus the difference in OB.
Status quo
Tsiros &
Mittal
(2000)
Regret ratings,
counterfactuals
Outcome feedback,
reversibility, switch/stay
Most participants had counterfactual thoughts in the switch–irreversible
condition, also rated as lowest quality/safety decisions. Participants
had more counterfactual thoughts when switching from the status
quo. Participants avoided regret by choosing status quo when
feedback on foregone outcomes was available. Switching associated
with more regret regardless of feedback. Irreversible decisions also
produced high regret regardless of outcome feedback; reversible
decisions followed by unpleasant feedback produced a slight increase
in regret.
Inman &
Zeelenberg
(2002)
Choice, written reports Prior outcomes Positive prior outcomes increased tendency to select status quo option,
compared with negative prior outcomes. No prior information also
led to more status quo selection. More reasons were given for
switching than for repeating decisions, indicating participants need to
justify potentially regrettable choices.
Status quo and omission option confounded
Kahneman &
Tversky
(1982)
Preference/regret Mutability “People generally agree that the person who acted and experienced the
same [bad] outcome [as one who omitted action] feels more
regret.... Anticipation of regret is likely to favor inaction and
routine behavior” (pp. 170–171).
Redelmeier
& Shafir
(1995)
Choice Increased no. of options Compared with similar decisions with fewer alternatives, participants
chose the status quo–omit option more often and chose less
preferable but easier to justify options more often (transitivity
violation). Attributed to increased “conflict,” this could be due to
avoidance of regret.
Inaction inertia
a
Tykocinski et
al. (1995)
Choice Gain/loss, outcome
discrepancy
Inaction inertia occurred in neutral and loss frames only; there was no
effect in the gain frame. A number of manipulations predicted to
affect inaction inertia by nonregret hypotheses did not produce
results.
(table continues)
149
DECISION AVOIDANCE
sion and make the same cognitive processes may still be important
to avoid. Although individuals may not weight regret in their
choice, they can still make use of counterfactual thought structures
to anticipate the regret they would feel and the blame that might
thus come from others when they realize a better decision could
have been made. Omission options clearly offer a source for
diminishing perceptions of responsibility, causality, and wrongdo-
ing, and thus the degree of blame assigned to a decision maker.
Status quo choices might be able to offer this kind of protection as
well; therefore, these forms of decision avoidance might offer
protection from backlash to the decision maker. Deferral is less
clearly associated with decreased blame, but the heightened anx-
iety resulting from situations in which blame may be a factor could
lead an individual to seek escape of the decision by postponing it,
especially if no status quo or omission option is available. These
alternative hypotheses are worth considering, but they must be
seen as complementary sources of avoidance that could operate in
isolation of and in addition to other sources of avoidance.
Table 1 (continued)
Study DA measure
Manipulation of
avoidance regret Major findings and/or interpretation
Inaction inertia
a
(continued)
Tykocinski &
Pittman
(1998)
Choice, written reports Avoidability, outcome
discrepancy, avoidance
cost
Inaction inertia is only present when reminders of (regret-provoking)
option are avoidable. It is also reduced when costs are introduced for
avoiding the option. If participants receive information that the
earlier outcome was not as attractive as thought, inaction inertia is
eliminated. Mention of regret was more frequent in protocols of
participants with large discrepancy between past and present options
and for participants who did not receive information degrading prior
option.
Butler &
Highhouse
(2000)
Choice, regret ratings Outcome discrepancy, gain/
loss
Regret ratings correlated with differences in action tendencies.
Anderson
(2001a)
Choice, regret ratings Outcome discrepancy, time
pressure, Sel.:
counterfactual inference
ability
Speeded decisions produced more inaction inertia as well as higher
regret ratings. No differences in inaction inertia based on individual
differences in counterfactual thinking have been found to date.
Tykocinski &
Pittman
(2001)
Choice, written reports Outcome discrepancy,
mutability
Inaction inertia was only present for highly mutable foregone outcomes,
for which outcome contrast was controlled. Reported thoughts of
regret correlated closely with inaction behavior whereas other
emotions did not. When regret was partialed out, the inaction inertia
effect size was reduced, though still statistically significant.
Arkes et al.
(2002)
Choice, regret ratings Outcome discrepancy,
distance
Individuals felt no regret and display no inaction inertia when a prior
opportunity was highly distant compared with nearby offers.
Zeelenberg et
al. (2002)
Regret judgments, ratings Omission, commission;
prior outcome’s valence,
prior outcome’s
informativeness
Participants attributed more regret to individual who took action
(change) when prior outcomes were positive or not known. Inaction
produced more regret only when prior outcomes were negative. The
effect of omission was mediated by perceived responsibility. Prior
outcome only influences responsibility, and thus regret, when it is
informative about the wisdom of a prior choice.
Deferral
Simonson
(1992)
Timing of choice Outcome feedback, forced
regret evaluation
Participants chose not to defer only when they expected feedback about
future opportunities and explicitly considered regret. Deferral can act
to increase anticipated regret when future opportunities may be worse
(delay costs).
Beattie et al.
(1994)
Preference, regret Outcome feedback Participants were more regretful when bad outcomes ensued from a
personal choice compared with a choice made by others. In situations
in which regret is a factor, many prefer to have others make the
decision.
Cooke et al.
(2001)
Timing of choice; rated
satisfaction, regret
Outcome feedback, control
over decision
Regret ratings were affected more by outcomes that could have been
obtained if participants deferred longer, than by those they actually
deferred on. Upward comparisons have greater effect on satisfaction.
Comparisons to earlier obtainable outcomes had an effect when
purchase timing decision was uncontrollable. Postdecision outcome
information’s influence was not limited by control. Participants could
anticipate these results and set higher thresholds for decision when
they expected information on foregone postdecision outcomes,
causing them to defer deciding longer.
Note. Sel. ⫽ selected variable.
a
These studies could meet criteria for any of the three decision avoidance forms identified in the current article (omission, status quo, deferral).
150
ANDERSON
Contributors to Anticipated Regret
A number of factors can act to increase or decrease one’s
anticipated feelings of regret. These are worth brief consideration,
because manipulation of these variables should affect decision
avoidance according to the rational–emotional model. Figure 3
presents a model, derived from the literature on regret, detailing
how these contributors influence anticipated regret. This model is
used as a guide for the discussion, with more direct influences
discussed first.
Reversibility. A decision is reversible if its outcome can be
altered after the fact; irreversible decisions are permanent. For
example, consumers can often reverse their purchase decisions
within a certain time period by returning products to the store
where they purchased them. Decision makers might anticipate less
regret for decisions that are reversible compared with similar
irreversible decisions (Zeelenberg, Beattie, van der Pligt, & de
Vries, 1996).
Research with hypothetical scenarios demonstrates that individ-
uals do think that less regret is associated with reversible outcomes
(Tsiros & Mittal, 2000). Although these researchers asked about
postdecisional regret, this is best thought of as an anticipated regret
effect because participants were attempting to predict how some-
one would feel in those situations. It is interesting, though, that this
seems to be an inaccurate anticipation of how people feel about
changeable outcomes. Research that compared predicted satisfac-
tion with experienced satisfaction with reversible and irreversible
choices shows that, although individuals do predict more satisfac-
tion with reversible choices, this prediction is incorrect (Gilbert &
Ebert, 2002). In fact, people experience less satisfaction with
reversible decisions compared with irreversible decisions. Assum-
ing that satisfaction is inversely correlated with regret, this implies
that the Tsiros and Mittal (2000) model is valid only as a model of
anticipated regret. However, both studies affirm the anticipation
effect, which is of current concern in describing potential influ-
ences on decision avoidance. Irreversible decisions should produce
more anticipated regret and thus should increase the tendency
toward decision avoidance.
Expected outcome feedback. Initially, regret theory suggested
that counterfactual mutation of potential outcomes was a unilater-
ally applied process in decision making under uncertainty. Cur-
rently, consensus is building that counterfactual thinking and, thus,
anticipated regret occur within certain boundary conditions. One of
those conditions is the expectation of feedback regarding the
outcome of nonselected, foregone options. In other words, after
making a decision, outcome feedback entails learning what would
have occurred had one chosen a different option. Anticipated
regret is dependent on a decision maker’s expectation of encoun-
tering this information (Larrick & Boles, 1995; Ritov, 1996; Ritov
& Baron, 1995; Tsiros & Mittal, 2000; Zeelenberg & Beattie,
1997; Zeelenberg et al., 1996).
Degree of loss aversion. Although not explicitly experimented
with, this theory suggests that loss aversion is an important factor
in effects of anticipated regret on decision making. Loss aversion
refers to the human tendency to weight outcomes viewed as losses
from an arbitrary reference point more heavily than equivalent
gains. This is arguably important for effects of regret on overt
decision avoidance because otherwise downward counterfactuals
(comparisons between an outcome and worse outcomes) would
have an equal effect on anticipated emotion, and anticipated re-
joicing would cancel out effects of anticipated regret (Baron &
Figure 3. A model of the contributors to anticipated regret. Bold text indicates emotional contributors to
decision avoidance. Fbk. ⫽ feedback.
151
DECISION AVOIDANCE
Ritov, 1994; Ritov & Baron, 1995; note also that this is a modi-
fication to the original regret theory). To the extent that decision
makers differ in the degree of their loss aversion, the decision
weight disparity between regret and rejoicing varies.
Perceived responsibility. Decision makers also are more likely
to anticipate regret when they perceive themselves as personally
responsible for the outcome. For certain reasons, acts and changes
to the status quo lead decision makers to feel more responsible for
outcomes, a finding I take up in the discussion of abnormal causes
(see Abnormal options section). A lesser degree of regret results
when people make a choice that seems well justified or not
causally related to an experienced bad outcome (Inman & Zeelen-
berg, 2002; Kordes-de Vaal, 1996; Simonson, 1992; van Dijk et
al., 1999; Zeelenberg et al., 2002). This probably occurs because
regret involves aspects of self-blame in addition to the counterfac-
tual comparison considered central in regret theory.
There has been some disagreement over the role responsibility
plays in moderating the influence of contextual factors on regret.
Connolly et al. (1997) claimed that responsibility is neither a
necessary nor substantial component of regret. However, these
researchers used a nonspecific measure that may not primarily
reflect regret. Further studies with procedural and measurement
modifications supported the view that responsibility is one impor-
tant moderator of regret; although it is not clear whether it is a
necessary condition for regret, some evidence is consistent with
this hypothesis of responsibility having a necessary role (Ordonez
& Connolly, 2000; Zeelenberg, van Dijk, & Manstead, 1998,
2000). Furthermore, some experiments that did not measure re-
sponsibility as a moderating variable are likely to have affected it
with their manipulations, as in research on the effect of consistency
on regret (Anderson, 2002b; Seta, McElroy, & Seta, 2001).
Mutability. The mutability of an outcome refers to the ease of
constructing counterfactual alternatives to it. For example, if Ms.
E misses her flight by 5 min and Ms. L misses hers by 37 min, one
would expect Ms. E to feel more regret because it is far easier for
her to undo any number of prior actions that could have changed
the result. It would be more difficult for Ms. L to construct a
scenario in which she does not miss the plane, as her negative
outcome is what is referred to as less mutable.
Mutability of alternatives could influence anticipated regret
directly, in that the number and intensity of counterfactuals that
could be produced is usually associated with more regret and thus
is factored into anticipations of regret. Alternatively, in some
cases, a more mutable outcome might be associated with more
perceived responsibility for the outcome. In the example above,
Ms. E may imagine herself as more responsible for the bad
situation because it is easy to imagine personally having done
something different to alter the outcome. It is also the case that the
most immutable outcomes are those that one feels were out of
one’s control—further justifying a link between mutability and
responsibility. Whether mutability directly affects anticipated re-
gret, is moderated by perceived responsibility, or both is a fairly
speculative consideration; therefore, both of the links are modeled
to represent the current state of knowledge about this variable.
Several studies demonstrate an effect of mutability on regret
(Inman & Zeelenberg, 2002; Kahneman & Miller, 1986; Kahne-
man & Tversky, 1982; Landman, 1988; Tykocinski & Pittman,
2001); these are typically associated with the mutability of abnor-
mal causes, discussed below in the Abnormal options section.
There are two studies purporting to find no association between
counterfactual mutation and experienced regret (N’gbala &
Branscombe, 1997; Seta et al., 2001). However, there are several
unaddressed problems in the measurements of counterfactual
thinking and regret that are used in these studies, as discussed
earlier.
Anticipated future opportunities. One can account for some
apparently contradictory results by noting that anticipated regret in
decisions that can be made at multiple points in time is affected by
what the decision maker believes about the nature of future op-
portunities at any given decision point being investigated. It has
been suggested that the primary motivation for upward counter-
factuals is to improve future outcomes in similar situations, mak-
ing regret more likely when it is believed a similar decision must
be made in the future (Markman, Gavanski, Sherman, & Mc-
Mullen, 1993).
There is evidence to suggest that regret judgments made in
time-sensitive decisions are more influenced by future opportuni-
ties that will be foregone by accepting a current offer than by
opportunities that have already been foregone (Cooke et al., 2001).
If it is believed that future opportunities may dominate the pres-
ently available alternatives, more regret will be associated with
taking action than with deferring. However, if future opportunities
could be dominated by the currently available alternatives, more
regret should be associated with deferral (Simonson, 1992). Be-
cause deferral can be associated with more or less regret depending
on beliefs about future opportunities, it is important to consider
this variable in interpreting studies that show increased immediate
action resulting from a regret-producing situation. This result does
not undermine the association of regret with deferral but demon-
strates that it can reverse, depending on an individual’s beliefs
about the future.
Abnormal options. A normal cause of an outcome would be an
ordinary, expected decision or state of the world—that is, the
“norm.” Kahneman and Miller (1986) presented a detailed model
of how norms are generated and suggested that these norms
influence counterfactual thinking. When the cause of an outcome
departs from the norm (i.e., is an abnormal cause), generation of
counterfactuals is more likely than when the cause is normal
because the norm is a highly salient counterfactual alternative,
whereas abnormal causes are not. Thus, one expects that abnormal
causes are more mutable, lead to more anticipated regret when
negative outcomes are at stake, and are associated with more
perceived responsibility for the negative outcomes. In this treat-
ment, abnormal causes are referred to as abnormal options, for
when they are considered prior to a decision, they have not yet
caused an outcome.
Originally, it appeared that actions were always abnormal op-
tions and that omissions were salient, normal alternative causes
that produced less anticipated regret (Kahneman & Miller, 1986;
Kahneman & Tversky, 1982; Landman, 1988). When two people
experience the same negative outcome, observers rate the person
who reached that outcome by means of inaction as experiencing
less regret. However, results showing that regrets recalled from
long-term autobiographical memory are more frequently omis-
sions challenge the generality of this statement (Feldman, Miya-
moto, & Loftus, 1999; Gilovich & Medvec, 1995a). These results
can be reconciled by considering the subsequent factor.
152
ANDERSON
Outcome valence from prior choices. Earlier, it became ap-
parent that omission and status quo choice cannot be clearly and
consistently considered normal options. The variables considered
in this final stage of the model clarify this situation and also
interact with the difficulty path.
Decision makers rarely make choices in a vacuum, and espe-
cially in cases in which status quo or repeated decisions are
involved, they may rely on memory to guide their understanding of
a problem. A particularly salient dimension in such choices is the
outcome of prior related decisions. New findings indicate that
status quo and omission options do not always modulate regret in
real-world situations because decision makers draw on information
about prior outcomes to guide their judgments (Inman & Zeelen-
berg, 2002; Zeelenberg et al., 2002). If a decision in the past led to
negative results, taking no action to change (maintaining the status
quo) would be considered an abnormal cause, leading to more
perceived responsibility and thus more regret. The results of Inman
and Zeelenberg (2002) and Zeelenberg et al. (2002) support this
contention, having demonstrated that when prior outcomes were
positive or unknown, decision makers tended to associate less
responsibility and regret with the status quo–omission option. The
status quo and omission options were confounded in these studies;
therefore, it is unclear whether the effect applies to both status quo
and omission choices. Intuitively, it seems most natural for prior
outcomes to affect primarily status quo choices rather than omis-
sion choices, as the two are not the same. I also hypothesize that
the outcome of prior choices affects the decision maker’s predi-
lection for consistency, which is considered in the Consistency
with orientation section. When prior outcomes are negative, the
impact of a preference for consistency should be lowered.
Finally, research also suggests that prior outcomes can generate
ambient negative emotion if they involve a risky decision or a
decision that involves choosing among different degrees of losses
(i.e., avoidance–avoidance conflicts; Riis & Schwarz, 2000). This
negative emotion generates differences in selection difficulty, as
described in the sections detailing the difficulty path.
The consideration of prior outcome valence also has implica-
tions for work with inaction inertia. It has been predicted that if the
previous opportunity were worse than the current opportunity,
regret effects should motivate one to take action (Zeelenberg et al.,
2002). Initial research on this “inaction mobilization” effect has
found some cases in which this appears to happen (Anderson &
Zeelenberg, 2002).
Consistency with orientation. One last factor that is associated
with regret is the decision maker’s orientation with regard to basic
dimensions of the options (e.g., Is one in an active mood? Is one
generally an active person? Do one’s goals involve taking more
action or avoiding it?, etc.). Seta et al. (2001) claimed to have
discovered that orientation is a different cause of regret, but upon
reflection this component fits into the framework that emerges
from the other research on regret reviewed herein. Their research
demonstrates that when orientation is manipulated in hypothetical
others, regret tends to follow decisions that are inconsistent with
that orientation, whether it be active or inactive. Thus, omission
can be associated with more regret—if a person is described as a
risk seeker. This effect generalizes to past and present personally
made decisions in which orientation is manipulated by mood
induction.
4
Selection difficulty. Selection difficulty is difficult to opera-
tionalize or define independently of the variables that produce it.
Beattie and Barlas (2001) asked participants to self-rate the diffi-
culty of a decision with some success; other researchers have
implied that a form of psychological conflict mediates between
their manipulation and the participant response without specifying
it in their measures (e.g., Tversky & Shafir, 1992). Perhaps others
would equate it with the negative emotion that is experienced
alongside certain decision-task environments (e.g., those in Luce,
1998). It is also tempting to use avoidant choice selection as a
behavioral indicator of difficulty, but this is unwise if one wishes
to examine difficulty as an influence on avoidance. It is also
unwarranted given the current understanding of avoidant choice as
influenced by multiple factors; it is at best an impure measure of
selection difficulty.
It is important, but difficult to specify on the basis of present
evidence, what this component means. Perhaps there are neuro-
logical indicators of difficulty that allow for specification of the
component independently of manipulations assumed to produce it
and responses presumed to reflect it (see Greene, Somerville,
Nystrom, Darley, & Cohen, 2001; Gur, Skolnick, & Gur, 1994, for
potential starting points relevant to the responses examined in this
article). Otherwise, it would appear that measurement of difficulty
is limited to essentially introspective self-report that researchers
might wish to eschew (Dennett, 1991; Nisbett & Wilson, 1977).
Such self-reports may be most useful in conjunction with converg-
ing neurological and behavioral measures, when available.
I begin attempting to define selection difficulty by explaining
what it is not. It cannot be equated with preference uncertainty, as
there may be many options among which people may be relatively
indifferent but which would not pose especially difficult decisions
for most of them. Consider this example: A student forgets to bring
a pencil for a test and is offered two different brands of pencil from
a classmate. The student has no reason to see either pencil as
superior; the student’s primary goal is to obtain a pencil quickly
rather than spend time thinking or accept the status quo of not
having one. Likewise, it should not be equated with negative
emotion per se, because the individual may be situated with
ambient negative emotion or may be faced with an easy, but
unpleasant decision. Consider the example of a virologist who had
a plan to spare 99% of a population from an epidemic, but because
4
The researchers concluded that consistency of orientation has its effect
through its impact on the desirability of different options; however, their
data are more consistent with a view that a manipulation of orientation is
equivalent to a manipulation of desirability because the measures of each
are highly correlated and entering one of them into a regression equation
predicting regret has an effect equivalent to the other. It is more likely that
when a decision maker has an active orientation, actions become the norm
and thus a salient counterfactual alternative that produces more regret than
counterfactuals of omission, as described by Kahneman and Miller (1986).
Recent data from a number of studies (see Anderson, 2002b) support this
interpretation. Also, orientation is likely to be influenced by prior experi-
ences, regardless of whether one is satisfied with one’s decision making
record. Thus, the valence of prior outcomes is likely to affect whether
orientation is adhered to. When prior outcomes are unsatisfactory and stem
from an aspect of the decision maker’s orientation, orientation-inconsistent
behavior becomes a more salient alternative, and consistency could be
associated with more regret.
153
DECISION AVOIDANCE
of a mistake, must choose between two equally costly courses of
action, one that would lead to a loss of 1.5% of the population and
one that would lead to a loss of 4% of the population—an obvious
decision that probably produces a variety of negative emotions
nonetheless. Selection difficulty is thus experienced when individ-
uals find it difficult to choose a particular course of action, but it
may occur in the absence of uncertain preferences or negative
emotion, although these are strong correlates of difficulty. Al-
though it is a central experiential component of decision making
that is familiar to most humans, little is scientifically known about
choice difficulty (Hastie, 2001; some notable attempts to charac-
terize the related concept of conflict may be found in the work of
Janis & Mann, 1977a; Lewin, 1931; Miller, 1944, 1959). In addi-
tion to the potential remedies outlined above, it may be useful to
have individuals rank various decisions in terms of difficulty in
addition to timing them, because investigators cannot use decision
time as an indicator of difficulty without circularity and oversim-
plification as regards the rational–emotional model of decision
avoidance.
Although more precision is desirable regarding the nature of
selection difficulty, the general hypothesis being advanced is clear:
Increases in selection difficulty lead to increases in the shares of
decision-avoidant options. Evidence for this hypothesis is pre-
sented in Table 2.
Contributors to Selection Difficulty
As with regret, I examine some of the important factors that
produce selection difficulty because these manipulations should,
and in some cases have been shown to, affect decision avoidance.
Figure 4 presents a model, derived from the literature on selection
difficulty, detailing how these contributors influence difficulty.
This model is used as a guide for the discussion, with more direct
influences discussed first. (Outcome valence from prior choices is
excepted from this section, as its influence on both emotional
factors is discussed above.)
Decision strategy. Decision makers adopt different strategies
for choosing on the basis of a number of considerations; one
distinction within these strategies that is especially relevant is
between compensatory and noncompensatory decision rules
(Payne, Bettman, & Johnson, 1993). When adopting a compensa-
tory approach, the decision maker is willing to trade off the value
of one attribute on one option with a value of another attribute
within or across options. For example, a negative safety rating on
an automobile might be compensated for by a lower price on that
same automobile. In following this approach, one adopts the
general idea of a particular prescriptive framework for decision
makers faced with conflicts between attributes: multiattribute util-
ity theory.
Often, an individual is reluctant to make these trade-offs. In a
noncompensatory strategy, trade-offs are not considered. The au-
tomobile’s safety may be paramount, and no difference in price
can persuade some consumers to risk endangering the vehicle’s
passengers. Taking a noncompensatory approach to decision mak-
ing is not normative but does serve to make selection easier in
most cases. It does not require as much thinking as a compensatory
strategy because entire alternatives can be eliminated for display-
ing unsatisfactory values on important attributes, and one need not
compute myriad comparisons among the remaining alternatives.
Furthermore, one may also mitigate the negative emotion that is
frequently generated in the process of making trade-offs by adopt-
ing a noncompensatory strategy that eliminates the trade-off pro-
cess. This leads to an interesting quandary for normative theories
of decision making: Because the normative compensatory strategy
entails more difficulty when enacted by actual human decision
makers, decision makers using this strategy tend to produce more
potentially suboptimal deferral, status quo, and omission choices.
Via the selection of a strategy for deciding, decision makers may
unwittingly predispose themselves toward or against choosing
decision avoidance (Dhar, 1996, 1997a; Dhar & Nowlis, 1999;
Luce, 1998; Luce et al., 1997).
Reasons. A justification for a decision is equivalent to having
a reason for selecting a particular option. The meaning of a reason
for making a decision has been recently investigated and refined
by Schick (1997). Briefly, the classic formulation of a reason is a
mental state that includes components of belief and desire; it is a
cognitive evaluation that an option will produce certain results and
a motivational orientation toward those results. Schick’s reformu-
lated definition includes an understanding of an option in terms of
the attributes toward which one has particular beliefs and desires.
Most options could produce a variety of results and be interpreted
in the framework of several cognitive schemata, and a reason for
a decision is situated within a particular framework, an understand-
ing of the relevance of the decision to other knowledge.
Decisions in which there are few potential reasons for making a
particular choice are likely to be more difficult. The context in
which options are situated can add or remove context-dependent
reasons for choice, making the decision easier or more difficult
(Redelmeier & Shafir, 1995; Tversky & Shafir, 1992). The ratio of
reasons for selecting one option over another, the salience of
reasons, and the subtlety of differences among those reasons (i.e.,
on how many dimensions of a reason the options vary) are prob-
ably quite important, although current research does not permit
more precise statements on how reasons influence difficulty.
Preference uncertainty. Preference uncertainty is a state of
being unsure of which of two or more options best meets one’s
goals or criteria for choice. Preference uncertainty can be distin-
guished theoretically from decision difficulty, as outlined above,
because trivial choices in which one is uncertain may not be
difficult, but the two components generally are related in conse-
quential decisions. Although this caveat about trivial decisions is
worth bearing in mind, preference uncertainty is expected to pro-
duce more difficult decisions (Dhar, 1996, 1997a, 1997b; Dhar &
Nowlis, 1999; Dhar et al., 1999; Dhar & Sherman, 1996; Tversky
& Shafir, 1992). Even relatively insignificant choices may be
slightly more difficult in the condition of preference uncertainty.
Two of the factors discussed below (option attractiveness differ-
ence and option set size) have their effect on selection difficulty
through preference uncertainty.
Degree of structure. Unstructured, ill-defined decisions
should be more difficult than well-defined decisions. This structure
is usually revealed in the degree to which algorithmic methods can
be applied to produce a solution. In a well-defined problem, the
relevant variables are known, their values are retrievable, and the
process of combining this information is mechanical and guaran-
teed to produce a solution of a determined nature. Many decisions
lack these attributes and are, consequently, more difficult. This
difficulty is produced by doubt about the relevance, accuracy, or
154
ANDERSON
Table 2
Studies Suggesting a Link Between Selection Difficulty and Decision Avoidance (DA) by DA Form
Study DA measure
Manipulation of selection
difficulty Major findings and/or interpretation
Deferral
Tversky &
Shafir (1992)
Choice Attractiveness difference, no.
of options
Participants were more likely to ask for added option when present options were
similar in attractiveness; they were also more likely to prolong search when
both options were weak. Participants postponed choice more often with
increased number of options; thus, nonpreferred deferral option becomes
preferred when new (even inferior) option is added (transitivity violation).
Beattie et al.
(1994)
Preference Fairness, accountability Participants were more decision averse in situations that involved potentially
inequitable outcomes for others, particularly when they expected to be held
accountable for the decision.
Dhar (1996) Choice Alternative, attribute-based
presentation;
compensatory, nonrule
training
Deferral was increased when alternatives were presented simultaneously
(attribute-based processing) compared with sequentially (alternative-based).
Attribute-based processing increases uncertainty, whereas alternative-based
processing promotes satisficing. Effect of processing focus was qualified by
compensatoriness of decision rule. Alternative-based processing only decrease
deferral when participants used compensatory rule. A reverse effect of
compensatory strategy is obtained when processing is attribute-based.
Dhar & Sherman
(1996)
Choice Unique good and bad
features
Deferral increased with common good–unique bad features on options. Shared
features were underweighted because of attention. Also, trade-offs between
bad features may be more difficult.
Dhar (1997a,
Exp.1&2)
Choice,
protocols
Added attractive and inferior
options (few vs. many
trade-offs)
An additional attractive option increased choice deferral. An added inferior
option decreased choice deferral (transitivity violation). Increased deferral
correlated with more thoughts, and participants had roughly equal numbers of
favorable thoughts for each option.
Dhar (1997a,
Exp. 3)
Choice,
protocols
Attractiveness difference The author found more deferral with similar options and more thoughts from
deferring participants in this condition. There was an equal ratio of favorable
to unfavorable thoughts for deferring participants in this condition.
Dhar (1997a,
Exp. 4)
Choice Compensatory strategy Participants trained in this strategy were more likely to defer; thus, strategy
decreased perceived attractiveness differences, increasing difficulty.
Dhar (1997a,
Exp. 5)
Choice Nonexclusive choice Participants exhibited less deferral when allowed to select multiple options.
Deferral thus cannot be solely attributed to desiring more options—difficulty
and uncertainty in selection are important.
Dhar (1997a,
Exp. 6)
Choice Trade-off practice
(multiattribute utility
theory)
Deferral decreased when participants practiced normative compensatory strategy.
This demonstrates decreased difficulty when attractiveness difference is held
constant.
Dhar (1997a,
Exp. 7)
Choice Trade-off practice, dominated
option
Practice did not affect deferral when a dominated option was present. Trade-off
practice to reduce difficulty was instrumental only in situations of preference
uncertainty.
Luce et al.
(1997)
Strategy,
negative
emotion
Vividness, likelihood of
negative consequences;
attractiveness difference,
no. of trade-offs
Likelihood of negative outcomes increased negative task emotion. Higher
emotion groups had more information acquisitions, more time deciding, and
more attribute-based processing. This supports negative emotion minimization
as a third factor in processing strategy selection. Trade-off difficulty increased
negative emotion much more in high conflict choices. Increased number of
acquisitions and time spent is a function of conflict. Emotion mediates effect
of conflict manipulations on processing strategy.
Mann et al.
(1998)
Decisiveness Sel.: culture Asian participants rated selves as more likely to engage in “buck passing” and
decision procrastination, forms of defensive avoidance. They also rated selves
as having less confidence in decision-making ability.
Dhar & Nowlis
(1999)
Choice, process
tracing
Time pressure (TP),
attractiveness difference,
attractiveness of common
features, type of conflict,
primacy of selection
Deferral decreases under TP, but only under high conflict. TP favors
noncompensatory strategy, which decreases difficulty. Deferral decreases
when overall (common) attractiveness high; this effect decreases under TP,
which shifts attention away from common features. TP has no effect when
selection decision is not primary. TP also tends to decrease deferral in
approach–approach conflicts because attention is shifted to positive features.
Dhar et al.
(1999)
Choice, process
tracing
Unique good and bad
features, similarity and
dissimilarity judgments
Choice deferral increased for unique good option pairs following similarity
judgment, vice versa for unique bad pairs. Attentional focus mediates this
interaction.
Ji et al. (2000) Decisiveness,
decisiveness
values
Sel.: culture Japanese participants were more indecisive than American and Chinese
participants, which suggests avoidant choice patterns may be consistent within
individuals and influenced by local norms. Value ratings obtained from
participants in different cultures were mapped closely onto their personal
patterns of decisiveness.
Beattie & Barlas
(2001)
Time to decide,
rated
difficulty,
choice
Category of options requiring
trade-offs
Noncommodities (Ncs) are more difficult to trade off than currencies or
commodities (Cs). Nc–Nc trade-off decisions are slowest; C–C decisions are
faster. Ncs were traded off with either other category fastest, perhaps because
participants used difficulty-reducing noncompensatory strategy.
(table continues)
155
DECISION AVOIDANCE
use of information one has acquired that is assumed to be relevant
to the decision.
Attractiveness of option set. Another direct influence on se-
lection difficulty is the overall attractiveness of the options avail-
able to the decision maker. Attractiveness is a common term in this
literature and presumably means something similar or equivalent
to utility. It is typically manipulated by varying values of attributes
whose dichotomous poles can be safely assumed to have a partic-
ular valence for all participants in a study. In particular, overall
attractiveness is manipulated by altering the values of shared
features of options.
This relationship is inverse: If the overall attractiveness is low,
there is very little motivating the individual to select any option.
Likewise, if the attractiveness of the option set is greater, the
individual stands to do well by choosing any option, even if there
is little to distinguish the options from each other. If the attrac-
tiveness of common (unvarying) features is increased among op-
tions in a set while utility differences are held constant, choice
deferral decreases (Dhar & Nowlis, 1999). Presumably this occurs
because it is less difficult to choose when any option available
satisfies goals adequately than when the options generally fail to
meet goals.
Cultural values. There is some recent research that is sugges-
tive of an influence of culture on tendencies to delay decisions. Ji,
Yates, and Oka (2000) found that self-reports, anecdotes, and
assessments of values all indicated that Chinese and American
persons tend to be more decisive than Japanese persons. Although
it is too early to know how general and large these cultural
differences are, these authors suggested that culture has its influ-
ence on decisiveness through local norms and values. Cultures
vary in terms of who makes decisions of varying importance. Ji et
al. (2000) suggested that Japanese participants whose decisiveness
differed from that of American students had assimilated values that
promoted building group consensus at the expense of individual
initiative, relative to American values. Chinese culture also differs
from American culture, but Chinese people’s decisiveness does
not, perhaps because, although individuals are looked down on for
making unauthorized decisions, decision-making power is valued
and aspired to, even though it is not held by most people for a wide
sphere of decisions. American culture emphasizes individual de-
cision making through concepts such as freedom and rights.I
believe that these values do not directly produce differences in the
probability of deferral, but they make decisions more or less
difficult for individuals holding various values.
With regard to the cultures described above, it is worth noting
that there is some conflicting evidence regarding Chinese culture
and decisiveness (Mann et al., 1998; Tse, Lee, Vertinsky, &
Wehrung, 1988). In some studies (Tse et al., 1988), Chinese
officials were more indecisive than their Western counterparts,
which is unexpected given Ji et al.’s (2000) analysis above, in
which it was noted that authorities are vested with individual
decision-making power and that others aspire to be like them in
this regard. These studies were interpreted by Tse et al. (1988) in
the light of predilections for probabilistic or nonprobabilistic
thinking in different cultures (as opposed to differences in values).
According to this analysis, participants from the Chinese sample
were less decisive because they viewed outcomes of decisions as
either certain or uncertain rather than making distinctions in prob-
Table 2 (continued)
Study DA measure
Manipulation of selection
difficulty Major findings and/or interpretation
Inaction inertia
a
Luce (1998) Choice,
emotion,
latency
decision
Availability of avoidant
options, trade-offs on
valued attributes, imagery
Trade-off difficulty is positively related to negative emotion; presence of
avoidance options is negatively related. Trade-off difficulty has greater effect
when no avoidant option is present. Increased preference for avoidant option
only under high trade-off difficulty. This suggests that difficult trade-offs
increase selection difficulty and negative emotion and that choosing deferral,
status quo, or default options may serve to reduce negative emotion. Imagery
increases negative emotion. Increased latency follows more trade-offs and
imagery instructions; this result weakens after avoidant options are
introduced. More status quo choice was found in conditions associated with
negative emotion. Initial negative emotion and response time mediate the
effects of manipulations on avoidant choice.
Status quo and omission option confounded
Redelmeier &
Shafir (1995)
Choice No. of alternatives Added options similar to preferred option led to increased selection of options
(including status quo) considered inferior in dichotomous choice (transitivity
violation). This may be due to difficulty of selection between similar options;
thus, participants avoid difficult trade-offs by choosing a dissimilar option.
Riis & Schwarz
(2000)
Choice Order of decisions,
emotionality of decisions
Status quo selection increased when decision was preceded by negative emotion
eliciting risky choice, compared with same decision presented without a prior
choice or following a nonemotional choice. Fewer participants chose the
status quo option on a nonemotional decision preceded by an emotional
decision, compared with no preceding decision. Negative emotion apparently
does not affect avoidant option selection when selection is not difficult for
other reasons.
Note. Exp. ⫽ experiment; Sel. ⫽ selected variable.
a
This study could meet criteria for any of the three decision avoidance forms identified in the current article (omission, status quo, deferral).
156
ANDERSON
abilities and weighting them to the extent that a Western decision
maker might. When the outcomes are uncertain, as they frequently
are, it is considered more prudent to hesitate than to act as if one
had certainty, whereas someone who makes probability judgments
can estimate the best course of action and act on that. Mann et al.
(1998) also found that confidence in decision-making ability is
lower in Eastern cultures, occurring alongside an increased ten-
dency to pass the buck or procrastinate on a decision.
The basic effect of culture is reliable, but the mechanism of the
effect has remained unclear. It has recently been suggested that
culture may influence decision making through the reasons for
choice that individuals in a culture are likely to consider (Briley,
Morris, & Simonson, 2000). As predicted by this hypothesis, East
Asian–North American cultural differences were more evident in
environments that required decision makers to consider reasons for
their decisions (Briley et al., 2000).
Because it is the beliefs and values associated with a culture that
seem to produce these differences in selection difficulty, other
sources of differing values within a culture should also produce
differences in decision avoidance. Consistent with this suggestion,
subcultural differences in religious beliefs about predestination
and the “direct will of God” seem to be associated with differences
in decisiveness (Anderson, 2002c).
Effort–accuracy trade-off. Decision strategies are selected on
the basis of several considerations, the most basic of which are
trade-offs between the cognitive effort required by the strategy and
the accuracy of the results it produces. Of the two strategies
discussed earlier, compensatory strategies tend to require more
effort and result in more accuracy, accuracy being defined as
conformity to a normative solution preferred in a utility theory
analysis. When individuals prefer to expend less effort and are
willing to accept suboptimal decisions as a result, they use a
noncompensatory strategy. The issue of how these trade-offs are
made and influenced is a complex matter best treated elsewhere;
fortunately, several excellent reviews already exist (Payne, 1982;
Payne et al., 1993; see also discussion of Luce et al., 1997, below).
Negative emotion. Current research has been unspecific about
the nature of negative emotion generated prior to avoidant deci-
sions. I have tentatively suggested that negative emotions may be
associated with the anticipatory emotions of fear, anxiety, and
despair as discussed by Loewenstein et al. (2001). The available
evidence is consistent in indicating that some form of negative
emotion generated by aspects of the choice itself (Beattie & Barlas,
2001; Luce, 1998; Luce et al., 1997) or by prior circumstances
(Riis & Schwarz, 2000) precedes decision avoidance. According to
the detailed investigations of Luce (1998) and colleagues (Luce et
al., 1997), negative emotion is often generated when many trade-
offs are possible, especially when those pertain to valued at-
tributes. Coping with this negative emotion then becomes a third
goal in the selection of a decision strategy. Participants in these
experiments (Luce, 1998; Luce et al., 1997) who displayed more
intense negative emotion acquired more information and spent
more time analyzing information; their processing also became
more attribute based. Conflict, as manipulated by differences in
attractiveness of options, had similar effects, and these two factors
interacted such that the most attribute-based, information-
acquisitive behavior occurred under high conflict and high nega-
tive emotion (produced by trade-off difficulty). In other words,
Figure 4. A model of the contributors to selection difficulty. Thin arrows indicate linkages for which evidence
is scant or uncertain. Opt. ⫽ option; Attractiv.⫽ attractiveness; Attr. ⫽ attractive.
157
DECISION AVOIDANCE
processing became more vigilant. If any of the avoidant options
were present in the choice set, individuals were much more likely
to select those options when they were experiencing negative
emotion. The presence of those options served to decrease negative
emotion associated with the choice context, and actually choosing
one of the avoidant options decreased negative emotion even
further (Luce, 1998).
Other research has demonstrated that negative emotion gener-
ated by prior choices can increase decision avoidance even for
nonemotional decisions (Riis & Schwarz, 2000). Apparently, lin-
gering or ambient negative emotion can also lead decision makers
to adopt more vigilant processing strategies. Although this study
did not address intervening variables, it is consistent with the
notion that difficult decisions put individuals on alert, thereby
altering their processing strategies for subsequent decisions, even
if they are apparently unrelated.
Option attractiveness difference. The most straightforward
manipulation of preference uncertainty involves altering the attrac-
tiveness difference between the various choice options by mini-
mizing the differences on their feature values. It is not required that
a researcher know exactly what the participants value most in the
decision; so long as differences between the options on their
features are smaller, the attractiveness difference must be smaller
regardless of the direction and weight of individuals’ preferences.
Smaller differences in the attractiveness of options produces
more preference uncertainty. This uncertainty tends to cause
greater difficulty in selection, motivating decision avoidance.
These manipulations are frequently reported to be successful in
producing changes in the degree of preference for decision avoid-
ance, other things being equal (Dhar, 1997a; Dhar & Nowlis, 1999;
Luce et al., 1997; Tversky & Shafir, 1992). Smaller attractiveness
differences also produce more vigilant decision strategies, involv-
ing more acquisitions of information and time spent viewing
information, and smaller attribute differences also promote
attribute-based processing. All of these findings are compatible
with the idea of using a compensatory strategy for a more difficult
choice (Luce et al., 1997).
Option set size. Increasing the number of options available
results in a concomitant increase in preference uncertainty. As
more options are added, the tendency is for any dominant options
to diminish in superior status. Decision makers then need to
prioritize among the features of options that are most important to
them to determine which option best meets their goals. This
process is complicated as options are added, making it more
difficult to discriminate between the subjective utilities of op-
tions—that is, if the decision maker even applies such a normative
strategy to large option sets. Often, decision makers may respond
to such a situation by adopting a policy of satisficing or using an
elimination-by-aspects rule. I propose further that another outcome
of decision difficulty produced in this way is decision avoidance.
Several experiments support this notion, finding that when an
avoidant option, such as a deferral or default option, is available,
the preference for it is increased by adding more options (Dhar,
1997a; Tversky & Shafir, 1992; Redelmeier & Shafir, 1995). From
a normative perspective, this is puzzling because more options
only implies more potential for incrementally higher utility
achievements. Of course, this is completely unrepresentative of the
perception of such situations by a human decision maker; added
options often only complicate the task and motivate one to avoid
the decision. In this regard, increasing the size of option set can
also remove simple reasons for selecting an option, such as dom-
inance, from the potential justifications for a choice.
Time limitations. Time limitations impose constraints on the
strategies for selection a decision maker can use. This component
is manipulated by speeding the decision process, which leads to the
counterintuitive result of less choice deferral and inaction inertia
(Anderson, 2001a, 2001b; Dhar & Nowlis, 1999). One might
expect that decision makers with less time to decide would prefer
an option that extends the search for alternatives, having judged
their limited processing as inadequate. This might be the case, but
only if the decision maker is already committed to using a com-
pensatory strategy that requires difficult, time-consuming compar-
isons. Instead, it seems that expediting the decision process leads
most decision makers to adopt a noncompensatory strategy that
makes selection easier, thus decreasing the preference for a defer-
ral or other avoidant option. The strategic analysis of the effect is
supported by online process tracing measures of information
acquisition.
Attentional focus. Finally, it has been suggested that manipu-
lations of attention can influence choice deferral (e.g., Dhar &
Nowlis, 1999). Attentional focus is defined as the time spent
analyzing particular features of options and the direction of shifts
in attention (comparisons made within an option or across op-
tions). By focusing attention on particular features of options, a
choice set can be made to seem more or less attractive from the
perspective of the decision maker. Those who attend to negative
features of the choice set should see the overall set composition as
less attractive, increasing the difficulty of selecting any particular
option (Dhar & Nowlis, 1999; Dhar et al., 1999; Dhar & Sherman,
1996). For example, Dhar and Nowlis (1999, Experiment 5) found
that the ratio of unique to common features attended to increased
under time pressure. The unique features were positively valenced
in all choices. This decreased choice deferral, but only in cases in
which the selection decision was primary (i.e., choosing between
two brands). This allowed the attentional focus to indirectly influ-
ence selection difficulty. If the deferral decision was primary (i.e.,
“deciding whether to decide”), there was no influence of time
pressure on deferral, presumably because the selection difficulty
component was factored out of the choice. This also demonstrates
that time limitations have an effect on attention in these choice
tasks.
Dhar (1996) also demonstrated that the kind of attentional shifts
that are made after analyzing particular features (i.e., comparisons)
are also important in indirectly influencing the tendency to avoid
choosing. Two basic kinds of attentional shifts that can be identi-
fied are those to other features within the same option or those to
features of alternative options. Simultaneous presentation of two
options should encourage across-option shifts, whereas sequential
presentation should encourage within-option shifts (comparisons).
Across-option shifts of attention are consistent with the more
difficult compensatory decision strategy, and consistent with this
interpretation, simultaneous presentation of alternatives led to in-
creased choice deferral. A follow-up experiment (Dhar, 1996) that
also manipulated rules that participants were trained to use sup-
ports this interpretation; in that study, comparing across alterna-
tives increased choice deferral when participants used a compen-
satory rule.
158
ANDERSON
Conflict type. Studies of choice deferral have also indicated
that manipulations embedded in the typology of conflict proposed
by Kurt Lewin (1931) and explored by Neal Miller (1944, 1959)
may exert an influence on decision avoidance. In Miller’s (1944,
1959) drive theory of conflict, goals are of two basic types:
approach and avoidance, signifying whether the organism seeks to
obtain or to avoid goal objects. Conflicts can arise when an
organism has to choose between two mutually exclusive goal
objects it desires (approach–approach), between two that it wishes
to avoid (avoidance–avoidance), or how to act toward a goal it is
both attracted to and repulsed by (approach–avoidance).
Dhar and Nowlis (1999) manipulated the type of conflict human
decision makers were faced with through the valence of the com-
mon and unique features of choice options. If the two options had
unique bad features and common good features, the researchers
characterized the situation as avoidance–avoidance, assuming the
common features cancelled out and the decision maker would have
to choose which undesirable feature was less repulsive. When the
situation was reversed and the options had common bad and
unique good features, the conflict was termed approach–approach
(approach–avoidance conflicts were not included in this experi-
ment). When the researchers manipulated time pressure alongside
type of conflict, they found that the time pressure manipulation
decreased choice deferral as predicted, but only when the individ-
ual faced an approach–approach conflict. They interpreted this as
the effect of increased attention to the unique good features, which
thus increased the perceived attractiveness of the option set. How-
ever, this interpretation implies that avoidance–avoidance conflicts
would, under time pressure, lead to more attention to negative
features and lead to more choice deferral under time pressure
compared with unlimited time, a result that did not obtain.
In Miller’s (1944, 1959) theory and research on conflict,
approach–approach conflicts were supposed to produce only min-
imal vacillation. Vacillation is a state in which the organisms
tested would either falter in their intentional movements and halt
or would move toward one goal, slow their approach, and retreat,
only to cease retreat and begin approaching again. It may be
likened to a behavioral measure of indecision. According to Mill-
er’s (1944, 1959) analysis, only avoidance–avoidance and
approach–avoidance conflicts produced any important degree of
conflict, as measured by vacillation. To the degree that Miller’s
(1944, 1959) conflict can be mapped on to selection difficulty and
vacillation can be mapped on to choice deferral, these results
conform well to the original predictions. Because approach–
approach conflicts are not associated with a high degree of con-
flict, time pressure could easily sway an individual to cease inde-
cisive behavior and act. When there is no pressure to decide
quickly, though, the individual may prefer to prolong search be-
cause deferral may be rewarded by superior options that erase the
conflict altogether. Avoidance–avoidance and approach–
avoidance conflicts are less susceptible to this manipulation be-
cause the conflict is difficult and entails potential negative conse-
quences. An individual does not yield to time pressure and choose
because the aversive contingencies of the choice powerfully mo-
tivate organisms to escape these situations; so long as avoidant
options are available, they are preferred regardless of time limita-
tions. These competing interpretations of the effect of conflict
cannot be arbitrated with the insufficient evidence available:
Avoidance–avoidance claims that attentional focus and preference
uncertainty mediates the effect of conflict type, whereas approach–
avoidance holds a more important, direct influence for conflict
type on the difficulty of the decision. Both paths have been
tentatively sketched into the model.
Neuroticism. Although individual differences have not been
studied in mainstream decision avoidance research, they are of
primary interest to researchers in the field of procrastination, itself
an individual difference. There are several definitions of procras-
tination, but the studies most relevant to decision avoidance use a
definition based on an individual’s hesitation to make decisions,
termed decisional procrastination.
In a set of correlational studies, the personality factor of Neu-
roticism (emotional reactiveness) accounted for most of the ob-
served variance in decisional procrastination (Milgram & Tenne,
2000). The increased anxiety experienced by individuals who
score higher on indicators of neuroticism may account for their
predilection for avoiding choice. Thus, I have tentatively linked
neuroticism to negative emotions generated by a choice situation,
with all the concomitant influences on decision making that antic-
ipatory negative emotions have been shown to produce thus im-
plied. Although evidence is just beginning to accumulate regarding
this potential link, there are other studies in the area of procrasti-
nation that are suggestive of a link between personality factors
related to neuroticism or trait anxiety and decisional procrastina-
tion (Berzonsky & Ferrari, 1996; Effert & Ferrari, 1989; Ferrari,
1991; Ferrari & Dovidio, 2000; Haycock, McCarthy, & Skay,
1998; McCown & Johnson, 1991; Orange, 1997).
Trade-off difficulty. As discussed earlier, trade-offs between
valued attributes produce negative emotions before a choice is
made, thus differentiating this factor from regret or anticipated
regret. Research demonstrates that simply manipulating the
number of trade-offs is sufficient to produce this effect (Luce,
1998; Luce et al., 1997). A trade-off is said to be required if an
option is not clearly superior on all features compared with its
alternatives but is deemed both more and less desirable on the
basis of comparisons of particular features or attributes. The
more such comparisons are required to assess the differences
between the options, the more trade-offs are required to reach a
choice. (For a review of the effects of trade-off difficulty on
negative emotion and decision making, see Luce, Bettman, &
Payne, 2001.)
Trade-off category. The category of the attributes that must be
traded off against each other merits consideration when predicting
trade-off difficulty. Beattie and Barlas (2001) demonstrated that
noncommodities (attributes that are not typically traded in a mar-
ket) are difficult to trade off with other noncommodities, as mea-
sured by decision time. In that study, noncommodities were traded
with commodities or currencies fastest, apparently because partic-
ipants used a lexicographic choice rule that dictated preference of
the noncommodity over the other two categories. Consistent with
this notion, noncommodities were rated as more important and less
tradable than the other two categories. I hypothesize that the
observed differences in decision time in this study are consistent
with preferences for decision avoidance, although these reactions
have not been measured. The path from trade-off category to
decision avoidance is long and indirect but fundamentally simple:
The categories of attributes influence the difficulty of trade-offs,
and difficulty of trade-offs influences decision strategy. When
trading noncommodities off with other noncommodities, individ-
159
DECISION AVOIDANCE
uals feel more negative emotion because of the increased impor-
tance and difficulty associated with these decisions. Thus, they are
more vigilant in the processing and use more compensatory strat-
egies, which further increase decision difficulty. When the trade-
offs are between commodities and other commodities or noncom-
modities and other categories, they are less difficult; less negative
emotion is experienced, and noncompensatory rules may be used
(especially in the case of trade-offs between noncommodities and
other categories).
Consequences of Decision Avoidance
Experienced regret. The actual level of emotions that are
anticipated earlier in the decision-making process should be mea-
sured. Regret can be measured by direct self-report, although this
method has pitfalls, such as the slippage between the participant’s
definition of regret versus the experimenter’s theoretical definition
(Ordonez & Connolly, 2000; Zeelenberg, van Dijk, & Manstead,
2000). Thus, a more formal definition is appropriate: Regret is a
negatively valenced emotion that is marked by generation of
upward counterfactual thoughts. Individuals mentally simulate
prior event sequences, altering details (particularly their own ac-
tions and decisions) and observe imagined outcomes. Upward
counterfactuals define regret in that the comparison of the coun-
terfactual outcome to the achieved outcome alerts decision makers
to the fact that they could have obtained a better outcome, and
negative emotions are experienced concomitantly with this recog-
nition. Although two studies have challenged the association of
counterfactuals with regret (e.g., N’gbala & Branscombe, 1997;
Seta et al., 2001), these studies contradict themselves by including
counterfactual thinking as part