Content uploaded by Tatiana Sokolova
Author content
All content in this area was uploaded by Tatiana Sokolova on Sep 22, 2017
Content may be subject to copyright.
Take It or Leave It: How Choosing versus
Rejecting Alternatives Affects Information
Processing
TATIANA SOKOLOVA
ARADHNA KRISHNA
People can make decisions by choosing or by rejecting alternatives. This research
shows that changing a task from choice to rejection makes people more likely to
rely on deliberative processing, what we label the task-type effect. To demonstrate
this effect, we use a set of established decision biases that can be attenuated
under deliberative processing. We show that changing a task from choice to rejec-
tion makes people express more consistent preferences between safe and risky
options in the Asian disease problem (study 1A) and in financial decision making
(study 1B), even with real monetary consequences (study 1C). Further, switching
a task from choice to rejection increases the quality of consideration sets in the
context of hotel reviews (study 2) and leads to more rational decisions in the con-
text of cell phone plan selection (study 3). Studies 4 and 5 tap into the process
underlying the effect of task type. We demonstrate that a rejection task produces
decisions similar to those observed in a choice task when decision makers are
cognitively depleted (study 4) or encouraged to rely on their feelings (study 5).
The findings provide insight into the effect of task type on deliberation and decision
outcomes.
Keywords: choice and rejection, information processing, framing effects
Consumers make their decisions in different ways. In
some situations, they choose alternatives from a set of
available options to form their consideration sets; in other
situations they reject the less attractive alternatives from a
list of available options. For instance, when consumers use
Pinterest, they “pin,” or choose, the most interesting items.
In contrast, when they review jobs on LinkedIn, they can
only hide, or reject, the less attractive jobs from their sug-
gestion lists. By the same token, takers of the Graduate
Management Admission Test sometimes have to choose
the correct logical argument in support of the focal claim;
but sometimes they have to eliminate, or reject, the logic-
ally flawed argument from the set of available alternatives.
Finally, irrespective of the context, people can adopt differ-
ent decision strategies on their own. For instance, when
shopping online, consumers can place only their most pre-
ferred items into their shopping carts, or they can put mul-
tiple items into their carts and then reject the products that
are relatively less attractive before proceeding to the
checkout.
While the two decision strategies (choice vs. rejection)
should normatively lead to the same outcomes, research in
consumer psychology and judgment and decision making
has shown that oftentimes they produce different decisions
(Dhar and Wertenbroch 2000;Laran and Wilcox 2011;
Shafir 1993). People adopt different selection criteria
(Yaniv and Schul 2000) and allocate different weights and
amounts of attention to disparate types of information in
Tatiana Sokolova (sokolova@umich.edu) is a postdoctoral scholar at
Professor Krishna’s Sensory Marketing Lab at the University of
Michigan. Aradhna Krishna (aradhna@umich.edu) is the Dwight F.
Benton Professor of Marketing at the Ross School of Business, University
of Michigan, Ann Arbor, MI 48109. Direct correspondence to Tatiana
Sokolova. The authors acknowledge the helpful input of the editor, associ-
ate editor, and reviewers. The authors thank members of the Sensory
Marketing Lab at the University of Michigan, Andrea Bonezzi, and
Panikos Georgallis for comments on earlier versions of the article.
Darren Dahl served as editor, and Zeynep Gu¨rhan-Canli served as associ-
ate editor for this article.
Advance Access publication August 22, 2016
V
CThe Author 2016. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc.
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com Vol. 43 2016
DOI: 10.1093/jcr/ucw049
614
choice versus rejection decisions (Laran and Wilcox 2011;
Meloy and Russo 2004;Shafir 1993). Rejection produces
larger consideration sets (Huber, Neale, and Northcraft,
1987;Yaniv and Schul 1997,2000) and causes preference
reversals by increasing the impact of negative (Shafir
1993) and preference-inconsistent attributes in decision
making (Laran and Wilcox 2011). In this article, we add to
the previous line of work on choice versus rejection by
showing that people are more likely to rely on deliberative
processing in rejection (vs. choice) tasks.
We test our conceptualization across seven studies. To
demonstrate that rejection relies on more deliberative pro-
cessing compared to choice, we use a set of established de-
cision biases that can be attenuated under deliberative
processing. We then compare the magnitude of these biases
across choice and rejection decisions. In studies 1A, 1B,
and 1C, we first replicate the robust finding that framing
options as gains versus losses affects decisions in choice
tasks (De Martino et al. 2006;Simon, Fagley, and Halleran
2004;Tversky and Kahneman 1981). We then show that
such gain versus loss framing changes decisions to a sig-
nificantly lesser extent in rejection tasks, consistent with
more deliberative processing. In studies 2 and 3, we exam-
ine the effect of choice versus rejection tasks in the con-
texts of online reviews (study 2) and complex product
purchases (study 3). In both studies, we show that people
make more rational and objectively superior decisions in
rejection. In studies 4 and 5, we find direct evidence of the
mechanism underlying the effect of task type on decision
making. In line with our theorizing, we find that rejection
decisions become similar to choice decisions when people
are cognitively depleted (study 4) or are encouraged to rely
on their feelings (study 5).
Our work contributes to the literature on choice ver-
sus rejection effects by demonstrating that task type
(choice vs. rejection) not only affects the importance
and evaluations of specific attributes pertaining to the
alternatives (Laran and Wilcox 2011;Meloy and Russo
2004;Shafir 1993) but also changes the way in which
the information about the alternatives is processed. We
demonstrate that changing the task from choice to rejec-
tion makes people more likely to use deliberative pro-
cessing. At the same time, we identify task type as a
novel boundary condition reducing the impact of gain
versus loss framing and increasing the impact of aggre-
gate versus anecdotal evidence on individual decisions.
In terms of practical implications, we show that Web
site interface decisions, such as opting-in (choosing) or
hiding (rejecting) buttons, can affect how consumers
process information about the available alternatives and
impact their preferences.
The remainder of the article is organized as follows. We
first discuss the literature pertinent to this research and
build up our main prediction. Next, we describe our seven
studies and their findings. Finally, we discuss the
theoretical and practical implications of our research, rule
out possible alternative accounts, and outline the directions
for future research.
LITERATURE REVIEW AND
CONCEPTUAL FRAMEWORK
Several prior studies have directly focused on the effect
of choice versus rejection, uncovering the effects of task
type on the size of the consideration sets, attention to spe-
cific attributes, attribute weights, as well as specific attri-
bute evaluations. In terms of the consideration set size,
research indicates that—with “not choosing” being the sta-
tus quo in choice tasks and “not rejecting” being the status
quo in rejection tasks—people tend to include fewer op-
tions into their consideration sets in choice versus rejec-
tion. For example, people shortlist significantly fewer job
applicants when their task is to select the applicants they
would interview, as opposed to eliminate the applicants
they would not interview for a job (Huber et al. 1987).
Similar findings have been obtained in the contexts of car-
eer counseling and general knowledge testing
(Krishnamurthy and Nagpal 2008;Yaniv and Schul 1997,
2000).
Furthermore, choice and rejection direct decision mak-
ers’ attention to different attributes of the alternatives and
lead to preference reversals. For example, Shafir (1993)
proposes that choice makes people focus on positive attri-
butes, whereas rejection makes them focus on negative at-
tributes. Consistent with this task-compatibility
framework, options with “enriched” positive and negative
attributes (e.g., high quality and high price) are preferred to
the “impoverished” options (e.g., average quality and aver-
age price) in a choice task, but are no longer preferred to
them in a rejection task (Shafir 1993). In a similar vein,
Laran and Wilcox (2011) suggest that rejection prompts
elaboration on preference-inconsistent attributes or attri-
butes that are considered less important given the currently
active goals. For example, people seeking “indulgence”
focus on proximity to nightlife when choosing apartments,
and they focus on price when rejecting them.
Finally, Meloy and Russo (2004) show that task type not
only affects the attention allocated to different attributes,
but it also influences the evaluations of those attributes. In
their studies, participants’ evaluations of positively
valenced attributes became more extreme when they had to
choose (vs. reject) one of the two alternatives. The opposite
was true for negatively valenced attributes, a result attrib-
uted to greater information distortion in task-compatible
decisions (DeKay, Pati~
no-Echeverri, and Fischbeck 2009;
Meloy and Russo 2004).
In this article, we focus on another aspect of the
decision-making process that is affected by changing the
task from choice to rejection. Specifically, we examine
SOKOLOVA AND KRISHNA 615
how rejection affects the extent to which people rely on de-
liberative processing in their decisions. We discuss this
next.
Rejection and Deliberative Processing
We propose that a rejection task will induce more delib-
erative processing, compared to a choice task. Here we
build on research in consumer behavior and cognitive
psychology to develop this prediction.
Losses and Attention to Negative Attributes. First, re-
jection decisions trigger consideration of loss of one or
several forgone options (Dhar and Wertenbroch 2000;
Park, Jun, and MacInnis 2000). Consideration of losses has
been linked to greater visual attention (Hochman and
Yechiam 2011) and more rational decisions in risky
choices (Yechiam and Hochman 2013). By the same token,
consideration of losses (e.g., price increases), compared to
potential gains (e.g., price discounts), has been shown to
attenuate price framing effects, a result consistent with the
idea that loss considerations should enhance deliberation
(Chatterjee et al. 2000). With loss considerations more
prominent in rejection, consumers should be more prone to
deliberation in rejection compared to choice.
Aside from the focus on the losses, a greater focus on
negative attributes in rejection (Shafir 1993) can also trig-
ger more deliberative processing. Studies show that nega-
tive information can prompt greater deliberation (Kuvaas
and Selart 2004; Malkoc, Hedgcock, and Hoeffler 2012).
For example, decision makers exhibit better information
recall after receiving negatively framed information, com-
pared to positively framed information (Kuvaas and Selart
2004). Similarly, people exhibit more vigilant processing
when deciding among unattractive (vs. attractive) alterna-
tives: they are less affected by irrelevant decoy options
when they see their choice alternatives as relatively un-
attractive (vs. attractive) (Malkoc et al. 2012).
Negation and Deliberation. Next, studies looking at
the difference between acceptance and negation, two deci-
sions that mirror choice and rejection, are also pertinent to
our research. For example, early correlational studies on
acceptance and negation suggest that “yea-sayers”—people
who tend to reply “yes” to “yes/no” questions—are more
impulsive and emotional, and they are less prone to exhibit
inhibition and control. In contrast, “nay-sayers”—people
who exhibit an overall disagreement tendency—are more
reflective and analytical, are less likely to behave impul-
sively, and more likely to deliberate on their responses
(Couch and Keniston 1960).
Furthermore, when put under cognitive load, people be-
come less likely to deny facts (Gilbert, Tafarodi, and
Malone 1993;Knowles and Condon 1999), a result consist-
ent with the basic proposition that negation entails the cog-
nitively demanding deliberative processing to a greater
extent than acceptance. It should be noted that these
findings merely imply an association between negation and
deliberation, and they do not allow us make inferences re-
garding causal links between negation and deliberative
processing. However, these findings are generally consist-
ent with our main proposition that rejection, a decision that
parallels negation, should have a stronger association with
deliberative processing compared to choice.
Other Research. Finally, several prior studies, although
not designed to investigate the role of deliberative process-
ing in rejection, give further support to our prediction. For
example, a study by Heller, Levin and Goransson (2002)
showed that people are significantly more likely to use the
exclusion (i.e., rejection) versus the inclusion (i.e., choice)
decision strategy when answering questions that have a
correct answer (e.g., “Which <city>hosted the Olympic
Summer Games of 1976?’’), compared to personal judg-
ment questions (e.g., ‘‘Which <city>would make the best
site for future Olympic Summer Games?”). In other words,
people are more prone to use rejection for questions requir-
ing conscious deliberation and to use choice for questions
that could be answered by relying on their feelings, a result
implicating an association between rejection and delibera-
tive processing.
Further, several studies imply that rejection draws on
cognitive resources more than choice does, meaning that
rejection might involve more deliberative processing com-
pared to choice (Krishnamurthy and Nagpal 2008;Laran
and Wilcox 2011). As discussed earlier, Laran and Wilcox
(2011) find that people prefer the indulgent alternative in a
choice task and the relatively less indulgent (but cheaper)
alternative in a rejection task. More importantly, their re-
sults indicate (in their study 4) that cognitive load affects
rejection more than choice. Krishnamurthy and Nagpal
(2008) find a similar pattern of results for the effect of cog-
nitive depletion on rejection versus choice. Taken together,
these results indicate that rejection relies on a limited pool
of cognitive resources to a greater extent than choice does.
Based on the prior research discussed here, we propose
that decisions entail more deliberative processing in rejec-
tion tasks (vs. choice tasks). For easier explication, we call
this the task-type information processing effect, or the task-
type effect. We test this proposition using a set of established
decision biases that are attenuated under greater deliber-
ation. Across seven studies we replicate previously observed
decision biases using a choice task, and then we show that
these effects are reduced if we use a rejection task.
STUDY 1A: REJECTION AND FRAMING
EFFECTS IN THE ASIAN DISEASE
PROBLEM
The goal of study 1A was to test the effect of task type
on gain versus loss framing effects using the Asian disease
616 JOURNAL OF CONSUMER RESEARCH
problem scenario. In the standard formulation of the Asian
disease problem, people choose between two programs of
combatting an unusual Asian disease (Tversky and
Kahneman 1981): a riskless program A and a risky pro-
gram B. Depending on how the options are described (as
gains vs. losses), people change their preferences between
the two programs: they tend to select the riskless option in
the domain of gains (“200 people <out of 600>will be
saved”) and tend to avoid this option in the domain of
losses (“400 people <out of 600>will die”). Importantly,
the gain-loss framing effect is reduced under deliberative
processing. For example, the effect is weaker among peo-
ple high in need for cognition and math skills (Simon et al.
2004;Smith and Levin 1996). By the same token, recent
neuroimaging data suggest that cognitive effort reduces
gain-loss framing effects by making people less likely to
choose the safe option in the domain of gains (Gonzalez
et al. 2005).
Per our theorizing, when working on a rejection task ver-
sus a choice task, people should rely more on deliberative
processing. Thus we expected to replicate the standard
framing effect when people have to choose one of the two
programs: A or B. We expected that this framing effect
would be attenuated when people have to reject one of the
two programs.
Design and Procedure
Mechanical Turk (MTurk) panelists were invited to fill
out a short computer-based survey for a small monetary
compensation. The study adopted a 2 (framing: gain vs.
loss) 2 (task: choice vs. rejection) between-subjects de-
sign. The participants read the hypothetical Asian disease
problem scenario proposed by Tversky and Kahneman
(1981). Depending on their experimental condition, they
read the problem either in the gain or the loss frame (frame
condition). Their task was to choose or to reject (task
condition) one of two programs, the riskless program A or
the risky program B.
As discussed earlier, people tend to select the riskless
program A in the domain of gains; however, they tend to
select the risky program B and to avoid the riskless pro-
gram A in the domain of losses. For the sake of exposition,
we refer to the shifts in preferences across gain and loss
frames as framing effects and to the shifts in preferences
across choice and rejection tasks as task-type effects. We
refer to our dependent variable as decision outcome. Table 1
lists the decision options in the four conditions.
After making their decision in the Asian disease prob-
lem, the participants in the gain (loss) frame answered a
question regarding the number of lives that would be saved
(lost) with program A, which served as an attention check.
Data and Results
A total of 203 MTurk panelists took part in the study.
Four participants were from outside the United States and
were removed. Three participants were removed due to re-
peat participation. Twelve participants failed the attention
check. The final sample included 184 participants (108
male).
To analyze the effect of problem framing on decisions
across the two task types, we ran a binary logistic regres-
sion. We used decision outcome, the selection of the risk-
less versus risky program, as the dependent variable. The
dependent variable was set to 0 when a participant selected
(i.e., chose or did not reject) the riskless program and to 1
when she selected the risky program. Frame (gain vs. loss),
task type (choice vs. rejection), and their interaction were
the independent variables—frame was set to 0 (1) when the
problem was framed in terms of losses (gains); task type
was set to 0 (1) when the task was to choose (reject) one of
the two programs.
The analysis revealed a significant effect of frame on de-
cision outcome (b ¼1.49, p¼.001) and no significant
TABLE 1
DECISION OPTIONS IN THE FOUR CONDITIONS OF STUDY 1A
FRAME
GAIN LOSS
TASK TYPE
CHOICE
(A) If program A is adopted, 200 people will be saved.
(B) If program B is adopted, there is a one-third prob-
ability that 600 people will be saved and a two-
thirds probability that no people will be saved.
Which Program would you CHOOSE?
(A) If program A is adopted, 400 people will die.
(B) If program B is adopted, there is a one-third prob-
ability that nobody will die and a two-thirds probabil-
ity that 600 people will die.
Which Program would you CHOOSE?
REJECTION
(A) If program A is adopted, 200 people will be saved.
(B) If program B is adopted, there is a one-third prob-
ability that 600 people will be saved and a two-
thirds probability that no people will be saved.
Which Program would you REJECT?
(A) If program A is adopted, 400 people will die.
(B) If program B is adopted, there is a one-third prob-
ability that nobody will die and a two-thirds probabil-
ity that 600 people will die.
Which Program would you REJECT?
SOKOLOVA AND KRISHNA 617
effect of task type (b ¼0.06, p¼.88). There was a sig-
nificant interaction between frame and task type (b ¼1.31,
p¼.035). In the choice task, where participants had to
choose one of the two programs, 76% selected the riskless
option in the gain frame, compared to only 42% in the loss
frame (Wald v
2
¼10.88, p¼.001), replicating prior find-
ings. However, in the rejection task, where participants had
to reject one of the two programs, 48% selected (i.e., did
not reject) the riskless option in the gain frame, compared
to 43% in the loss frame (Wald v
2
¼0.19, p¼.66). Thus
the effect of gain-loss framing was reduced in the rejection
task. These results support our main hypothesis (figure 1).
Discussion
Study 1A showed that rejection reduces the framing ef-
fect in the Asian disease problem. Taken together with
prior findings on the moderating role of deliberative pro-
cessing in framing effects (Gonzalez et al. 2005;Simon
et al. 2004), this result provides preliminary evidence that
participants’ decisions are more deliberative in rejection
compared to choice.
STUDY 1B: REJECTION AND FRAMING
EFFECTS IN GAMBLING DECISIONS
Study 1B was designed to replicate conceptually the re-
sults observed in study 1A using a different gain-loss fram-
ing scenario. We used a modified version of the scenario
used by Tversky and Kahneman (1981), in which people
had to choose between two monetary gains or two monet-
ary losses. We expected to replicate the standard framing
effect in the choice task, that is, to find that people are
more (less) likely to choose the riskless option A when it
offers a sure gain (sure loss). We expected that this framing
effect would be attenuated in the rejection task.
Design and Procedure
MTurk panelists were invited to fill out a short
computer-based survey for a small monetary compensa-
tion. The study adopted a 2 (framing: gain vs. loss) 2
(task: choice vs. rejection) between-subjects design. The
expected utility of each option was a monetary gain in
the gain frame condition and a monetary loss in the loss
frame condition. The participants had to choose or to
reject one of the two options. Table 2 lists the decision
options in the four conditions.
Note that in the original design, Tversky and
Kahneman (1981) set the expected values of the sure
gain ($240) and the gamble ($250) to be slightly differ-
ent. This allowed for a stronger test of their theory in the
domain of gains where a perfectly rational decision
maker would actually prefer the gamble (and not simply
be indifferent between the two options). The expected
values of the sure option and the gamble were identical
in the loss frame in the original scenario by Tversky and
TABLE 2
DECISION OPTIONS IN THE FOUR CONDITIONS OF STUDY 1B
FRAME
GAIN LOSS
TASK TYPE
CHOICE
(A) A sure gain of $240
(B) 25% chance to gain $1000, and 75% chance to
gain nothing
Which option would you CHOOSE?
(A) A sure loss of $740;
(B) 75% chance to lose $1000, and 25% chance to
lose nothing
Which option would you CHOOSE?
REJECTION
(A) A sure gain of $240
(B) 25% chance to gain $1000, and 75% chance to
gain nothing
Which option would you REJECT?
(A) A sure loss of $740;
(B) 75% chance to lose $1000, and 25% chance to
lose nothing
Which option would you REJECT?
FIGURE 1
STUDY 1A: REJECTION TASK REDUCES FRAMING EFFECT
76%
48%
42% 43%
0%
20%
40%
60%
80%
100%
CHOICE REJECTION
SHARE OF THE RISKLESS OPTION
GAIN LOSS
618 JOURNAL OF CONSUMER RESEARCH
Kahneman (1981) ($750). To allow for a strong test of
their theory even in the domain of losses (and for sym-
metry), we modified the original task and set the value
of the safe option at $740, such that the perfectly
rational decision maker would prefer the guaranteed loss
in this task. Following the decision regarding the two
options the participants answered an attention check
question regarding the amount of money gained (lost)
with option A.
In addition to participants’ decisions regarding the two
options, we also recorded their response times in the main
task in this study and in subsequent studies. It is assumed
that longer response times are indicative of more delibera-
tive processing. Although several researchers have ques-
tioned the validity of response time data analysis in
detecting more (vs. less) deliberative processing (Evans,
Dillon, and Rand 2015;Krajbich et al. 2015), we collected
response time data in our studies. Assuming that longer re-
sponse times are associated with more deliberative process-
ing, we expected that the participants would take longer to
make their decisions in the rejection (vs. choice) task.
Data and Results
A total of 203 MTurk panelists took part in the study.
Eight participants were from outside the United States and
were removed. An additional 22 participants failed the at-
tention check. The final sample included 173 participants
(99 male).
To analyze the effect of problem framing on decisions
across the two task types, we ran a binary logistic regres-
sion similar to the one used in study 1A, with decision out-
come as the dependent variable and frame (gain vs. loss),
task type (choice vs. rejection), and their interaction as the
independent variables.
The analysis revealed a significant effect of frame on de-
cision outcome (b ¼3.56, p<.001) and no significant
effect of task type (b ¼0.40, p¼.39). There was a sig-
nificant interaction between frame and task type (b ¼2.87,
p<.001). In the choice task, where participants had to
choose one of the two options, 96% selected the riskless
option in the gain frame, compared to only 29% in the loss
frame (Wald v
2
¼26.68, p¼.001), replicating prior find-
ings. However, in the rejection task, where participants had
to reject one of the two options, 54% selected (i.e., did not
reject) the riskless option in the gain frame, compared to
38% in the loss frame (Wald v
2
¼2.42, p¼.12). Thus the
effect of problem framing was reduced in the rejection
task. These results support our main hypothesis (figure 2).
Response Time Data. We analyzed participants’ re-
sponse times in the main task. In this and subsequent stud-
ies participants’ response times were log-transformed and
trimmed at 2 SDs from their respective group means (Fazio
1990). A two-way analysis of variance (ANOVA) with
frame and task type as between-subjects factors revealed a
significant effect of frame (M
gain
¼2.65 vs. M
loss
¼2.87,
F(1, 162) ¼7.38, p¼.007), meaning that decisions took
longer in the domain of losses than in the domain of gains.
The analysis also revealed a marginally significant effect
of task type (M
choice
¼2.69 vs. M
rejection
¼2.83, F(1, 162)
¼2.99, p¼.086). Thus, in this study, participants took
longer to make their decisions in the rejection (vs. choice)
task, consistent with expectations. The interaction between
frame and task type was not significant.
Discussion
The results of study 1B conceptually replicate the find-
ings of study 1A. In this study we used the domain of mon-
etary gains and losses to show that rejection reduces the
effect of gain-loss framing.
STUDY 1C: REJECTION AND FRAMING
EFFECTS IN AN INCENTIVE-
COMPATIBLE TASK
The goal of study 1C was to replicate conceptually the
results of the first two studies using an incentive-compat-
ible procedure. As in study 1B we used the context of mon-
etary gains and losses, adopting a modified version of the
scenario used by De Martino and colleagues (2006).We
expected to replicate the standard gain-loss framing effect
in the choice task and that the framing effect would be atte-
nuated in the rejection task.
Design and Procedure
Students at a large North American university partici-
pated in a computer-based study in exchange for course
FIGURE 2
STUDY 1B: REJECTION TASK REDUCES FRAMING EFFECT
96%
54%
29%
38%
0%
20%
40%
60%
80%
100%
CHOICE REJECTION
SHARE OF THE RISKLESS OPTION
GAIN LOSS
SOKOLOVA AND KRISHNA 619
credit. The study adopted a 2 (framing: gain vs. loss) 2
(task: choice vs. rejection) between-subjects design.
At the beginning of the study, participants were told that
we would randomly select four study participants who
would be paid per their decisions in the study. Next, they
were presented with the hypothetical scenario proposed by
De Martino and colleagues (2006). The scenario said that
they would receive $50 but would not be able to retain this
initial amount of $50 with certainty. Instead, they had to
select one of two options: a sure option or a gamble.
Participants’ response times were recorded.
Table 3 summarizes the decision options in the four
conditions. Note that unlike in study 1B, the expected
values of the riskless and risky options are identical
within and across the gain frames ($20 of $50 for both
options) and the loss frames ($30, leaving participants
with $20 of $50).
Data and Results
A total of 158 undergraduate students participated in the
study (65 male).
To analyze the effect of problem framing on decision
outcomes across the two task types, we ran a binary
logistic regression similar to the one used in studies 1A
and 1B—with decision outcome as the dependent variable
and frame (gain vs. loss), task type (choice vs. rejection),
and their interaction as the independent variables.
The analysis revealed a significant effect of frame on de-
cision outcome (b ¼1.65, p¼.001) and no significant
effect of task type (b ¼0.32, p¼.49). There was a sig-
nificant interaction between frame and task type (b ¼2.01,
p¼.003). When the task was to choose one of the two op-
tions, 79% of participants selected the riskless option in the
gain frame, compared to 42% in the loss frame (Wald v
2
¼
10.92, p¼.001), replicating the framing effect. However,
when the task was to reject one of the two options, 41% of
participants selected (i.e., did not reject) the riskless option
in the gain frame, compared to 50% in the loss frame
(Wald v
2
¼0.63, p¼.43)—that is, the effect of problem
framing was reduced in the rejection task. These results
support our main hypothesis.
Response Time Data. A two-way ANOVA with frame
and task type as between-subjects factors revealed a sig-
nificant effect of frame (M
gain
¼3.14 vs. M
loss
¼3.27, F(1,
148) ¼4.77, p¼.031), meaning that decisions took longer
in the domain of losses than in the domain of gains. The ef-
fect of task type was not significant (M
choice
¼2.20 vs.
M
rejection
¼2.22, F<1). The interaction between frame
and task type was not significant.
Discussion
Study 1C replicates the results of the first two studies
and demonstrates that rejection reduces the effect of prob-
lem framing in an incentive-compatible framework.
STUDY 2: REJECTION AND THE ROLE
OF AGGREGATE VERSUS ANECDOTAL
EVIDENCE
Study 2 tested the effect of task type in the context of
online reviews where people typically receive two types of
information: an aggregate rating based on ratings from
multiple users and a sample of individual ratings and ver-
bal reviews. Prior research has shown that people often dis-
count aggregate numerical information in the presence of
anecdotal evidence (Alter et al. 2007;Tversky and
Kahneman 1973;Yang, Saini, and Freling 2015). For ex-
ample, when asked to evaluate the probability that Jack is
an engineer, people may ignore the proportion of engineers
in the sample and rely on Jack’s verbal description instead,
a decision bias often referred to as base rate neglect
(Tversky and Kahneman 1973). Similarly, when choosing
between two insurance plans offered by companies A and
B, people can disregard the aggregate satisfaction ratings
of the two companies and instead rely on the personal ex-
perience of a specific consumer when making their deci-
sions (Yang et al. 2015). Importantly, the tendency to
TABLE 3
DECISION OPTIONS IN THE FOUR CONDITIONS OF STUDY 1C
FRAME
GAIN LOSS
TASK TYPE
CHOICE
(A) keep $20
(B) 40% chance to keep all ($50), and 60% chance to
keep nothing
Which option would you CHOOSE?
(A) lose $30;
(B) 40% chance to lose nothing, and 60% chance to
lose all ($50)
Which option would you CHOOSE?
REJECTION
(A) keep $20
(B) 40% chance to keep all ($50), and 60% chance to
keep nothing
Which option would you REJECT?
(A) lose $30;
(B) 40% chance to lose nothing, and 60% chance to
lose all ($50)
Which option would you REJECT?
620 JOURNAL OF CONSUMER RESEARCH
discount aggregate numerical information in the face of an-
ecdotal evidence is reduced under more deliberative pro-
cessing (Alter et al. 2007;Yang et al. 2015). Since, per our
theorizing, rejection tasks entail more deliberation com-
pared to choice tasks, we expected that users would give
more weight to aggregate numerical ratings versus anec-
dotal evidence in rejection tasks compared to choice tasks.
Design and Procedure
MTurk panelists were invited to fill out a short
computer-based survey for a small monetary compensa-
tion. The study adopted a 2 (task type: choice vs. rejection)
2 (information type: “bad rating–good reviews” vs.
“good rating–bad reviews”) mixed factorial design. Task
type was manipulated between subjects. Information type
was manipulated within subjects.
Participants had to narrow down a list of 12 hotels to a
smaller set by either saving attractive hotels to their list
(choice) or by removing unattractive hotels from their list
(rejection). For each hotel, participants received informa-
tion about its aggregate numerical rating based on at least
120 user ratings (i.e., aggregate information). In addition,
they saw two sample ratings and verbal reviews from indi-
vidual users (i.e., anecdotal information). We acknowledge
that processing large amounts of anecdotal information (vs.
aggregate ratings) requires substantial effort, making reli-
ance on anecdotal information as deliberative as reliance
on aggregate ratings. Thus to ensure that greater reliance
on aggregate versus anecdotal information was indicative
of deliberative processing, we limited the number of indi-
vidual ratings and reviews to two. At the beginning of the
study, the participants saw the following instructions:
You will get information about 12 hotels, one hotel at a
time. On each screen you will see the ratings (on a scale
from 1 to 10) and randomly selected reviews of a given
hotel. Your task is to narrow down the selection of 12 hotels
to a few attractive options. You will do that by saving hotels
to your list (removing hotels from your list). The hotels you
save (do not remove) would be the hotels you consider to be
attractive. Read the information carefully and decide for
each hotel whether you would like to save it to your list
(remove it from your list).
Out of the 12 hotels, 4 hotels had bad aggregate ratings
but good individual ratings and reviews, and 4 hotels had
good aggregate ratings but bad individual ratings and re-
views; the remaining 4 hotels served as fillers and had
moderate aggregate numerical ratings and moderate indi-
vidual ratings and reviews. For all 12 hotels, participants
got aggregate “overall” numerical ratings and also ratings
for value, staff, facilities, location, comfort and cleanliness;
these six specific dimension ratings were highly correlated
(a¼.99) and were averaged to form the “overall” ratings.
Additionally, for each hotel, participants saw two sample
ratings and reviews. Appendix A provides examples of
each information type (“bad rating—good reviews,” “good
rating—bad reviews,” “filler”). Examples of screenshots
are presented in appendix B.
Hotels were presented one at a time, in random order.
The task was self-paced. After participants had seen the in-
formation for each hotel and decided which hotels to
choose (reject) to form their final lists, we asked them sev-
eral follow-up questions. First, we asked them to rate the
importance of five different pieces of information—overall
numerical ratings, numerical ratings on specific dimen-
sions, number of reviews, individual reviews, and individ-
ual ratings—in their decisions about the hotels. They rated
how important each of the five pieces of information was
on a 5 point scale (1 ¼Not at all; 5 ¼Very much). Next,
as an attention check, participants answered whether they
were saving hotels to their list or removing hotels from the
list. Finally, they filled out their demographic information.
We expected that participants in the rejection task would
assign greater importance to the aggregate (aggregate rat-
ings) versus anecdotal information (two individual ratings
and reviews), and be less likely to select “bad rating—
good reviews,” compared to participants in the choice task.
By the same token, we expected that participants in the re-
jection task would be more likely to select “good rating—
bad reviews” hotels, compared to participants in the choice
task. We did not expect to find significant differences
across the two tasks for “filler” hotels. Fillers were
included to prevent the participants from guessing the hy-
pothesis and to avoid perfect multicollinearity.
Pretest
We pretested our stimuli to ensure that the “mismatch”
between aggregate ratings and individual ratings and re-
views did not raise any suspicions regarding the quality of
the information about the 12 hotels. The participants read
information about 12 hotels provided by an online booking
Web site. Next, they indicated to what extent they per-
ceived the Web site to be typical, normal, and odd (re-
verse-coded) on a 7 point scale anchored on “Not at all” on
the left and “Very” on the right. After that, the participants
rated the online booking Web site on its overall quality, the
quality of its rating system, and the quality of its review
system, using a 7 point scale anchored on “very low” on
the left and “very high” on the right.
Fifty MTurk panelists completed the pretest. Two par-
ticipants were from outside the United States and were
excluded. The final sample included 48 participants (31
male). We computed the means of typicality (a
typicality
¼
.92) and quality ratings (a
quality
¼.89) to create aggregate
typicality and aggregate quality ratings, respectively. A
one-sample ttest indicated that participants’ ratings of
Web site typicality were significantly above the midpoint
of the typicality scale (M
typicality
¼4.88, t¼3.90, p<.001).
SOKOLOVA AND KRISHNA 621
Similarly, a one-sample ttest indicated that participants’
ratings of Web site quality were significantly above
the midpoint of the quality scale (M
quality
¼4.60, t¼3.12,
p¼.003). Thus the pretest indicated that the information
about the 12 hotels was perceived as typical and of rela-
tively high quality.
Data and Results
A total of 121 MTurk panelists took part in the study.
One participant was from outside the United States and was
removed. Two participants were removed due to repeat par-
ticipation. One participant failed the attention check. The
final sample included 117 participants (57 male).
We analyzed the shares of hotels with bad ratings and
good reviews and the shares of hotels with good ratings
and bad reviews across conditions in a repeated-measures
ANOVA. For each participant, we computed the share of
“bad rating—good reviews” hotels on the final list and the
share of “good rating—bad reviews” hotels on the final
list.
Note that the number of hotels in the final set can be
lower in the choice versus the rejection task due to the sta-
tus quo bias (the status quo was “not to choose” in the
choice task vs. “not to reject” in the rejection task) (Huber
et al. 1987;Yaniv and Schul 1997,2000). This was indeed
the case in our data (M
choice
¼5.76 vs. M
rejection
¼6.67,
F(1, 115) ¼8.71, p¼.004). Consequently, using the abso-
lute number of hotels on the list as the dependent measure
was going to reduce the difference between choice and re-
jection tasks for the “bad rating—good review” hotels and
inflate the difference for the “good rating–bad review”
hotels. Thus we used percentages instead of absolute meas-
ures in our analysis.
The model included task type (choice vs. rejection) as a
between-subjects factor, information type (bad rating—
good review vs. good rating—bad review) as a within-
subjects factor, and their interaction. The share of fillers
was not included in the model to avoid perfect multicolli-
nearity (P
filler
¼1–(P
good
þP
bad
)). There were no
significant differences in the shares of fillers across choice
and rejection tasks (P
choice
¼39% vs. P
rejection
¼41%,
F<1).
Figure 3 presents the shares of “bad rating—good re-
views” hotels, and of “good rating—bad reviews” hotels
across the two task-type conditions. There was no main ef-
fect of review type (F(1, 115) ¼2.43, p¼.122) or condi-
tion (F<1). The analysis revealed a significant interaction
between task and information type (F(1, 115) ¼4.15, p¼
.044). Simple contrasts revealed that, in line with our pre-
dictions, the share of “bad rating—good reviews” hotels
was lower in the rejection (vs. choice) task condition (P
choice
¼39% vs. P
rejection
¼28%, F(1, 115) ¼4.96, p¼
.028). Further, simple contrasts showed that the share of
“good rating—bad reviews” hotels was marginally higher
in the rejection (vs. choice) condition (P
choice
¼22% vs.
P
rejection
¼31%, F(1, 115) ¼2.78, p¼.098).
Response Time Data. In this study, we expected re-
sponse time data to follow a different pattern. We expected
the participants to focus more on aggregate (i.e., aggregate
ratings) versus anecdotal (i.e., two individual ratings and
reviews) information in rejection compared to choice.
Since processing aggregate ratings can be less time con-
suming than processing individual ratings and reviews, we
expected respondents to take less (and not more) time with
rejection versus choice. The response time pattern is gener-
ally aligned with these predictions. The regression analysis
with task type and information type dummies as independ-
ent variables indicated that effect of task type was not sig-
nificant (b
rejection
¼0.10, p¼.428). However, decisions
took directionally less time in the rejection (vs. choice)
task. In addition, decisions regarding “good rating—bad re-
views hotels” took longer compared to decisions regarding
“bad rating—good reviews hotels” (b
bad rating—good reviews
¼0.23, p<.001) and “filler hotels” (b
filler
¼0.24, p<
.001).
Mediation Analysis. We expected that participants
would assign greater importance to aggregate ratings ver-
sus anecdotal information in rejection (vs. choice). This
difference in the importance assigned to the two types of
information was expected to mediate the effect of task type
on the shares of “bad rating—good reviews” and “good rat-
ing—bad reviews” hotels.
To test our predictions, we ran a mediation analysis with
the INDIRECT macro by Preacher and Hayes (2008). The
first mediation model used task type as the independent
variable (task ¼1 in rejection condition; task¼0 in choice
condition), the importance of aggregate (i.e., the overall
FIGURE 3
STUDY 2: SHARES OF HOTELS IN CHOICE VERSUS
REJECTION
39%
22%
28% 31%
0%
20%
40%
60%
80%
100%
"BAD RATING - GOOD
REVIEWS" HOTELS
"GOOD RATING - BAD
REVIEWS" HOTELS
SHARE IN THE CONSIDERATION SET
CHOICE REJECTION
622 JOURNAL OF CONSUMER RESEARCH
rating) versus anecdotal information (computed as the
average of individual ratings and individual reviews; r¼
.86, p<.001) as the mediator, and the share of “bad rat-
ing—good reviews” hotels as the dependent variable. The
second mediation model used the share of “good rating—
bad reviews” hotels as the dependent variable and was
otherwise identical to the first model.
The first equation in models 1 and 2 is the same, and the
results indicated that task type had a marginally significant
positive effect on the importance assigned to aggregate
versus anecdotal information (b ¼0.57, t¼1.66, p¼
.099). This suggests that in line with our predictions, when
the task changed from choice to rejection, people assigned
greater importance to aggregate versus anecdotal informa-
tion. The second equation in model 1 (model 2) indicates
that the importance assigned to aggregate versus anecdotal
information had a significant negative effect on the share
of “bad rating—good reviews” hotels (b ¼0.10, t¼
12.49, p<.001), a significant positive effect on the share
of “good rating—bad reviews” hotels (b ¼0.11, t¼12.30,
p<.001). The results of equation 2 in the two models sug-
gest that when the importance of aggregate versus anec-
dotal information increased, the share of “bad rating—
good reviews” hotels in participants’ consideration sets
decreased and the share of “good rating—bad reviews”
hotels increased. Finally, the third equation in model 1,
which focused on the mean indirect effect of task type on
the share of “bad rating—good review” hotels through im-
portance of aggregate versus anecdotal information (based
on 1000 bootstrap samples), was marginally significant,
with a point estimate of 0.06 and a 90% confidence inter-
val (CI) excluding zero (0.11 to 0.00). Similarly, the
third equation in model 2 focusing on the mean indirect ef-
fect of task type on the share of “good rating—bad review”
hotels through importance of aggregate versus anecdotal
information (based on 1000 bootstrap samples) was also
marginally significant, with a point estimate of 0.06 and a
90% CI excluding zero (0.00–0.12). Together, results from
the three equations making up mediation models 1 and 2
indicate that the effect of task type on the shares of “bad
rating—good reviews” and “good rating—bad reviews”
hotels was mediated by the importance of aggregate versus
anecdotal information. Figures 4a and 4b present the re-
sults of the mediation analysis.
Discussion
Study 2 showed that hotels with bad ratings and good re-
views were less likely to be shortlisted when participants
were rejecting (vs. choosing) hotels. In contrast, hotels
with good ratings and bad reviews were more likely to be
shortlisted when participants were rejecting (vs. choosing)
hotels. Consistent with the idea that deliberative processing
plays a greater role in rejection versus choice, participants
assigned greater importance to aggregate versus anecdotal
information when they were removing hotels from their
lists (i.e., rejecting), as opposed to when they were adding
hotels to their lists (i.e., choosing).
STUDY 3: REJECTION AND DECISION
MAKING IN A COMPLEX TASK
Study 3 focused on differences between choice and re-
jection in the context of complex and cognitively
FIGURE 4a
MODEL 1: REJECTION DECREASES THE SHARE OF BAD RATING—GOOD REVIEWS HOTELS
SOKOLOVA AND KRISHNA 623
demanding decisions. We used a complex phone plan se-
lection decision proposed by Mishra, Mishra, and
Nayakankuppam (2007). In the scenario people had to
choose between two cell phone plans: one with fewer “any-
time” minutes and a lenient penalty for exceeding the
monthly limit of “anytime” minutes; and another, object-
ively superior, plan with more “anytime” minutes and a
strict penalty structure (Cheema and Patrick 2012;Mishra
et al. 2007). Prior research has shown that people tend to
focus on the visually salient penalty aspect of the plan de-
scription, resulting in suboptimal decisions. However,
when prompted to rely on deliberative processing, they
consider both the number of anytime minutes and the pen-
alty structure, resulting in better decisions (Cheema and
Patrick 2012). Thus in this study we examine phone plan
selection decisions across choice and rejection tasks, ex-
pecting to replicate prior results with the choice task and to
observe more optimal decisions in the rejection task.
Design and Procedure
MTurk panelists were invited to fill out a short
computer-based survey for a small monetary compensa-
tion. The study adopted a single-factor (task type: choice
vs. rejection) between-subjects design. The participants
made a decision by choosing (rejecting) one of the follow-
ing cell phone plans:
a. Plan from Firm A that offers 160 anytime minutes
for $19.99 a month;
b. Plan from Firm B that offers 200 anytime minutes
for $19.99 a month.
In making their decisions, participants also had to con-
sider the penalties for exceeding the respective monthly
limits of firms A and B. The charges for exceeding the
monthly limits are presented in table 4 (appendix C pro-
vides the screenshots).
FIGURE 4b
MODEL 2: REJECTION INCREASES THE SHARE OF GOOD RATING—BAD REVIEWS HOTELS
NOTE.—Standard errors are presented in parentheses.
*Effect significant at 10% level; **Significant at 5% level; ***Significant at 0.1% level.
TABLE 4
PENALTY STRUCTURE FOR EXCEEDING THE MONTHLY LIMIT
IN FIRM A AND FIRM B
Penalty for exceeding
anytime minutes by
Firm A imposes
(per minute) ($)
Firm B imposes
(per minute ) ($)
10 minutes 0.00 2.00
20 minutes 0.25 2.00
30 minutes 0.50 2.00
40 minutes 1.00 2.00
More than 40 minutes 2.00 2.00
624 JOURNAL OF CONSUMER RESEARCH
If we focus on the relatively salient table of penalties for
exceeding the monthly limit, we feel that plan A is more
frugal. However, a more careful evaluation of the two
plans reveals that plan B is actually the frugal one—it
offers more minutes for the same amount of money and im-
poses the same penalty as plan A for exceeding the 200-mi-
nute time limit. While choosing plan B is the correct
decision, people tend to select the relatively more expen-
sive plan A (Cheema and Patrick 2012;Mishra et al.
2007). We predicted that presenting the decision regarding
the two plans as a rejection task would enhance delibera-
tive processing and reduce the share of plan A. After sub-
mitting their decision regarding the two options,
participants were transferred to the next screen where, as
an attention check, they had to report the number of mi-
nutes offered by plan A. Participants’ response times in the
main task were recorded.
Data and Results
A total of 142 MTurk panelist took part in the study.
One participant was from outside the United States and
was removed. One participant was removed due to repeat
participation. Twenty-nine participants failed the attention
check. The final sample included 111 participants (68
male).
To analyze the effect of task type on the selection of the
expensive cell phone plan, we used the one-tailed Fisher
exact test. The analysis revealed that people were less
likely to select (i.e., not reject) the expensive plan A in the
rejection task: 70% of participants selected the more ex-
pensive phone plan A in the choice task condition, but only
51% of participants selected it in the rejection task condi-
tion (p¼.034).
Response Time Data. A one-way ANOVA with task
type as a between-subjects factor revealed a significant ef-
fect of task type (M
choice
¼3.62 vs. M
rejection
¼3.87, F(1,
107) ¼5.58, p¼.020) on response times in the phone plan
selection task. As expected, participants took significantly
longer to make their decisions in the rejection (vs. choice)
task in this study.
Discussion
In this study we find further support for our main prop-
osition by showing that rejection affects performance in
the context of complex product purchases. Prior studies
have shown that enhancing the role of deliberative process-
ing (Cheema and Patrick 2012) makes people more likely
to make the correct decision in the cell phone plan scen-
ario. Supporting the proposition regarding the enhanced
role of deliberative processing in rejection, our results
show that rejection also increases the quality of decisions
in the cell phone plan scenario.
STUDY 4: ROLE OF COGNITIVE
DEPLETION
Study 4 tested the underlying mechanism of the task-
type effect. In this study, we introduced a cognitive
depletion manipulation for some of the participants. We
expected that cognitive depletion would make people less
likely to engage in deliberative processing. As such, we ex-
pected cognitive depletion to make rejection decisions
more similar to choice decisions, thus diminishing the
task-type effect.
To induce cognitive depletion we used the Stroop color
identification task (1935). The Stroop (1935) task is con-
sidered to be cognitively depleting because it requires that
people override their initial tendency to read the word (an
automatic response) and name its color instead
(Pocheptsova et al. 2009; Webb and Sheeran 2003). We
expected that the cognitive depletion manipulation would
affect participants’ subsequent decisions concerning two
monetary gains and make the participants in the rejection
task condition behave more similarly to the participants in
the choice task condition.
Design and Procedure
Students at a large North American university partici-
pated in a computer-based study in exchange for course
credit. The study adopted a 2 (task: choice vs. rejection)
2 (cognitive depletion: control vs. depleted) between-
subjects design.
The study was composed of two parts. The first part of
the study was adapted from Webb and Sheeran (2003)
and was based on the Stroop task (1935). In this part par-
ticipants were presented with 48 words, one word at a
time. Each word represented the name of a color
(“BLUE,” “GREEN,” “RED,” “YELLOW”) and was dis-
played using a font of one of four colors (“BLUE,”
“GREEN,” “RED,” “YELLOW”). Importantly, 75% of
the words did not match the color of the ink they were
typed in. For example, the word “YELLOW” could have
been colored green. At the bottom of the computer screen
were four buttons with the names of the four ink colors
used in the task. The names of the colors on the response
buttons were colored black.
Participants assigned to the control condition were in-
structed to merely read the color words. Participants as-
signed to the cognitive depletion condition were instructed
to click on the button matching the color of the font and to
ignore what any given word said. Thus when the word
“YELLOW” was typed in green ink, participants in the
cognitive depletion condition were supposed to respond
“GREEN.” Participants completed 48 such screens.
In the second part of the study, participants read a hypo-
thetical scenario about two monetary gains used in study
SOKOLOVA AND KRISHNA 625
1B. They had to make a decision by choosing or rejecting
one of the following gains:
A. sure gain of $240;
B. 25%chance to gain $1000 and a 75%chance to
win nothing.
Note that we used the gain frame and not the loss frame,
since we had previously found bigger differences between
choice and rejection tasks in the gain frame (see studies
1A–1C). Finally, as an attention check, participants had to
recall the amount of money they would gain with the risk-
less option A. Participants’ response times in the main task
were recorded.
Data and Results
A total of 179 undergraduate students participated in the
study. Eight participants did not follow the instructions in
the cognitive depletion task (color identification rate less
than 50%) and were excluded. Four participants failed to
correctly recall the amount of money they would gain with
the riskless option and were excluded. The final sample
included 167 participants (68 male).
To analyze the effect of task type (rejection coded as 0,
choice coded as 1) on the share of the riskless option (i.e.,
the sure gain) across two cognitive depletion conditions
(control coded as 0, cognitive depletion coded as 1), we
ran a binary logistic regression. Decision outcome served
as the dependent variable; task type, cognitive depletion,
and their interaction served as the independent variables.
The analysis revealed a significant effect of task type
(b ¼1.31, p¼.009) and of cognitive depletion
(b ¼0.96, p¼.044). The interaction between task type
and cognitive depletion did not reach statistical signifi-
cance (b ¼0.80, p¼.284). Replicating the results of study
1B in the control condition, 81% of participants opted for
the riskless gain in the choice task, compared to 54% in the
rejection task (Wald v
2
¼6.89, p¼.009). But when par-
ticipants were cognitively depleted, the difference between
the choice and rejection tasks was directionally reduced.
Now, 83% of participants opted for the riskless option in
the choice task condition, compared to 75% in the rejection
task condition (Wald v
2
¼0.86, p¼.36), meaning that the
task-type effect was eliminated in the cognitive depletion
condition.
Importantly, in line with our conceptualization, we
observed that cognitive depletion had no effect on the share
of the riskless gain in the choice condition (P
control
¼81%
vs. P
cognitive depletion
¼83%, Wald v
2
¼0.08, p¼.78). In
contrast, the cognitive depletion manipulation significantly
increased the share of the riskless option in the rejection
condition (P
control
¼52% vs. P
cognitive depletion
¼73%,
Wald v
2
¼4.05, p¼.044). Figure 5 presents the shares of
the riskless option across the four conditions.
Response Time Data. A two-way ANOVA with task
type and cognitive depletion as between-subjects factors
revealed no significant main effects of task type (M
choice
¼
2.84 vs. M
rejection
¼2.78, F(1, 156)¼1.33, p¼.250) or
cognitive depletion (M
control
¼2.80 vs. M
cognitive depletion
¼
2.82, F(1, 156)¼0.05, p¼.819) on response times in the
gain selection task. The interaction between task type and
cognitive depletion was marginally significant (F(1,
156)¼2.91, p¼.090). Contrary to the pattern observed in
studies 1B, 1C, and study 3, simple contrasts indicated that
in the control condition people look less time to make their
decisions in the rejection (vs. choice) task (M
choice
¼2.88
vs. M
rejection
¼2.72, F(1, 156) ¼4.09, p¼.045). The
effect was directionally reversed and no longer significant
in the cognitive depletion condition (M
choice
¼2.80 vs.
M
rejection
¼2.83, F(1, 156) ¼0.15, p¼.697).
Discussion
The results of study 4 show that cognitive depletion
increased the share of the riskless option in the rejection
task and made people act more similarly to people in the
choice task. Thus in line with our reasoning, the results of
study 4 indicate that deliberative processing is the driver of
the proposed task-type effect.
STUDY 5: ROLE OF FEELING-BASED
EVALUATION
Study 5 further examined the underlying mechanism of
the task-type effect. In this study, we manipulated the ex-
tent to which participants rely on their feelings when mak-
ing their decisions, expecting them to engage in less
FIGURE 5
COGNITIVE DEPLETION INCREASES RISK AVERSION IN THE
REJECTION TASK
81% 83%
53%
75%
0%
20%
40%
60%
80%
100%
CONTROL DEPLETION
SHARE OF THE RISKLESS OPTION
CHOICE REJECTION
626 JOURNAL OF CONSUMER RESEARCH
deliberative processing in the feeling-based evaluation con-
dition. As such, we expected the feeling-based evaluation
condition to make rejection decisions more similar to
choice decisions, thus diminishing the task-type effect.
Design and Procedure
MTurk panelists were invited to fill out a short
computer-based survey for a small monetary compensa-
tion. The study adopted a 2 (task: choice vs. rejection) 2
(evaluation: control vs. feeling-based) between-subjects
design. Similar to study 4, we used the gain frame scenario
from study 1B. Participants had to make a decision by
choosing or rejecting one of two monetary gains—a risk-
less versus a risky option. We also manipulated the evalu-
ation basis by asking participants in the feeling-based
condition to think about “how they felt about each of the
two options” before making their decision regarding which
option to choose (reject). Participants in the control condi-
tion did not receive such instructions and merely had to
choose (reject) one of the two options. Additionally, in this
study we randomized the order of the two options to rule
out the possibility that the data pattern is driven by partici-
pants’ tendency to choose or reject the first option they see
(Mantonakis et al. 2009). Next, as an attention check, par-
ticipants had to recall the amount of money they would
gain with the riskless option. Participants’ response times
in the main task were recorded.
Data and Results
A total of 264 MTurk panelists took part in the study. Six
participants were from outside the United States and were
removed. Nine participants were removed due to repeat par-
ticipation. Ten participants failed the attention check. The
final sample included 239 participants (135 male).
Order had no significant main or interaction effects (p>
.50), and the data were collapsed across the two orders. To
analyze the effect of task type (rejection coded as 0, choice
coded as 1) on the share of the riskless option (i.e., the sure
gain) across the two evaluation conditions (control coded
as 0, feeling-based coded as 1), we ran a binary logistic re-
gression. Decision outcome was the dependent variable;
task type, evaluation mode, and their interaction were the
independent variables. The analysis revealed a significant
effect of task type on decision outcome (b ¼2.89, p<
.001) and a significant effect of evaluation type
(b ¼0.91, p¼.018). There was also a significant inter-
action between task type and evaluation condition
(b ¼1.75, p¼.035).
Replicating the results of study 1B in the control condi-
tion, 95% of participants opted for the riskless gain in the
choice task, compared to 52% in the rejection task (Wald
v
2
¼20.08, p<.001). When the instructions induced a
feeling-based evaluation of the two alternatives, the
difference between the choice and rejection tasks was
reduced. Now, 89% of participants opted for the riskless
option in the choice task, compared to 73% in the rejection
task (Wald v
2
¼1.15, p¼.03). Feeling-based evaluation
instructions had no effect on the share of the riskless gain
in the choice condition (P
control
¼95% vs. P
feeling based
¼
89%, Wald v
2
¼1.32, p¼.25). In contrast, the feeling-
based evaluation manipulation significantly increased the
share of the riskless option in the rejection condition
(P
control
¼52% vs. P
feeling based
¼73%, Wald v
2
¼5. 56,
p¼.02). Figure 6 presents the shares of the riskless option
across the four conditions.
Response Time Data. Neither order nor its interactions
with other factors had a significant effect on response times
(all F’s <1), so we collapsed the data across the two order
conditions. A two-way ANOVA with task type and evalu-
ation type as between-subjects factors revealed a signifi-
cant effect of task type (M
choice
¼2.51 vs. M
rejection
¼
2.67, F(1, 222) ¼8.26, p¼.004) on response times in the
gain selection task. In this study, again, participants took
significantly longer to make their decisions in the rejection
(vs. choice) task. The analysis also revealed a main effect
of evaluation type (M
control
¼2.48 vs. M
feeling based
¼2.70,
F(1, 222) ¼16.19, p<.001): people took longer to make
their decisions in the feeling-based (vs. control) condition.
The latter result can be attributed to the word count differ-
ence across the two conditions (n
control
¼32 vs. n
feeling based
¼54). The interaction between task type and evaluation
type was not significant.
Discussion
The results of study 5 show that inducing feeling-based
processing in the rejection task made participants act more
FIGURE 6
FEELING-BASED PROCESSING MANIPULATION INCREASES
RISK AVERSION IN THE REJECTION TASK
95%
89%
52%
73%
0%
20%
40%
60%
80%
100%
CONTROL FEELING-BASED
SHARE OF THE RISKLESS OPTION
CHOICE REJECTION
SOKOLOVA AND KRISHNA 627
similarly to participants in the choice task. Thus the results
indicate that deliberative processing serves as the driver of
the proposed task-type effect.
GENERAL DISCUSSION
We show across a set of decision contexts that changing
a task from choice to rejection makes people more likely to
use deliberative processing, what we label the task-type ef-
fect. We replicate prior results pertaining to various deci-
sion biases in choice tasks (e.g., gain-loss framing in
studies 1A, 1B, and 1C; discounting of aggregate informa-
tion in study 2; use of salient cues in study 3); and show
that the results change in rejection tasks, becoming more
consistent with deliberative processing. Thus we contribute
to the literature on choice and rejection by showing that
task type not only changes the weights allocated to option
attributes (Laran and Wilcox 2011;Shafir 1993), but also
determines how the information about the options is
processed.
In study 1A, we use the Asian disease problem to dem-
onstrate that a rejection task makes people less susceptible
to framing effects, compared to a choice task. In study 1B
and in an incentive-compatible, study 1C, we replicate the
moderating effect of task type on framing effects in the
context of monetary gains and losses. It is worth noting
that in studies 1A. 1B, and 1C, the effect of task type was
only present in the domain of gains (all p’s <.01), but not
in the domain of losses (all p’s >.30). As discussed earlier,
decisions in the domain of losses are, by default, more
likely to be driven by deliberative processing (Chatterjee
et al. 2000;Yechiam and Hochman 2013). Thus the fact
that the rejection task, expected to enhance deliberative
processing, did not affect participants in the loss frame is
consistent with our theorizing.
Next, in study 2, we demonstrate the effect of task type
on information processing in the context of online reviews.
We find that a rejection task reduces the share of hotels
with bad aggregate ratings but good individual reviews,
and it increases the share of hotels with good aggregate rat-
ings but bad individual reviews in users’ consideration
sets. In study 3, we test the effect of task type in a complex
product purchase scenario, where people have been shown
to make suboptimal decisions (Cheema and Patrick 2013;
Mishra et al. 2007). We replicate prior research findings
with a choice task. We then show that people are less likely
to select the objectively more expensive but seemingly
cheap cell phone plan A, versus the objectively superior
plan B, in a rejection task, an outcome consistent with
greater deliberation in rejection tasks.
Studies 4 and 5 provide evidence that deliberative pro-
cessing underlies the task-type effect. Study 4 shows that
cognitive depletion reduces the task-type effect such that
people in a rejection task behave more similarly to those in
a choice task. This suggests that rejection involves greater
use of cognitive resources and thus entails more delibera-
tive processing. Study 5, analogously, shows that when
people are encouraged to rely on their feelings, their deci-
sions in a rejection task are closer to the decisions in a
choice task—again, indicating that rejection involves more
deliberative processing compared to choice.
In sum, in seven studies we demonstrate that decision
making becomes more consistent with deliberative pro-
cessing when a rejection (vs. choice) task is used. We also
show that when people are less likely to rely on delibera-
tive processing (e.g., when cognitively depleted or encour-
aged to rely on their feelings), the reported effect of task
type is attenuated. It should be noted that the results for the
effect of task type on participants’ response times are not
unequivocal. Studies 1B, 3, and 5 indicate that decisions
took significantly longer in rejection (vs. choice) tasks, and
study 1C shows no significant difference across conditions
(study 2 has a predicted reverse pattern that is supported).
However, study 4 shows a contrary pattern. Nonetheless, in
an unreported meta-analysis of the response times across
the studies (i.e., studies 1B, 1C, 3, 4, and 5; we exclude
study 2 where we did not expect rejection to take longer
than choice), we find an overall positive effect of rejection
on response times (d ¼.20, z¼2.79, p¼.005), consistent
with the idea that rejection entails more deliberative
processing.
Alternative Accounts
Prior research has provided substantial insight into the
differences between choice and rejection in terms of the
underlying evaluation processes and resulting preferences.
In this section, we elaborate on the link between our find-
ings and those reported in prior research, and we rule out
some alternative explanations of our results.
Elaboration on Preference-Inconsistent Attributes. Laran
and Wilcox (2011) propose that choice versus rejection can
shift preferences toward options consistent versus incon-
sistent with the currently activated preferences and goals.
Their results are aligned with our conceptualization. The
main premise of this article is that rejection enhances de-
liberative processing. Greater deliberation, in turn, should
increase decision makers’ propensity to override their auto-
matic responses (Evans and Stanovich 2013). Priming ma-
nipulations used by Laran and Wilcox could have created
such automatic responses (e.g., priming savings would
make the cheap option the intuitive response). According
to our theory, in a choice task participants would be more
likely to select the primed, or the intuitively appealing op-
tion. In contrast, in a rejection task, participants would be-
come more likely to override their intuition and less likely
to select (i.e., not reject) the primed alternative, a
628 JOURNAL OF CONSUMER RESEARCH
prediction consistent with the data pattern observed in the
experiments by Laran and Wilcox.
At the same time, we do not think that the framework
proposed by Laran and Wilcox can fully account for the
empirical evidence reported in this article. The notion of
active preferences and goals (e.g., indulge vs. save money
on an apartment) is central to their theory. However, the
presence of baseline preferences and goals is somewhat un-
likely in several of the scenarios used in this article, such
as the Asian disease problem scenario (study 1) or the on-
line hotel reviews scenario (study 2). In sum, while our
framework is consistent with the empirical evidence re-
ported by Laran and Wilcox, their theoretical framework
cannot account for all our findings.
Task-Compatibility Framework. One may also suggest
that the pattern of results in some of our studies (e.g., the
Asian disease problem in study 1A and the gambling
scenario is study 1C) could be explained by the task-
compatibility framework proposed by Shafir (1993).
According to the task-compatibility framework, choice
makes people focus more on the positively valenced (vs.
negatively valenced) attributes of each option—for in-
stance, the number of lives saved in the Asian disease
scenario. In contrast, rejection makes them focus more on
the negative (vs. positive) attributes—the number of lives
lost. A focus on the 200 lives saved with certainty in a
choice task should increase the share of the riskless option
in the gains frame. A focus on the number of lives lost in
a rejection task should make people more likely to infer
that 400 people will die if the “200 lives saved” option is
taken in the gains frame. Consequently, per the task-com-
patibility account, the share of the riskless option in the
gain frame would go down in a rejection (vs. a choice)
task. However, the data pattern observed in the loss frame
in our studies is not as easily explained by this alternative
account. We find that task-type does not affect decisions
in the loss frame, yet Shafir’s framework would predict a
difference between choice and rejection in the loss frame
as well. We run a simulation with 5 million observations
(details in online appendix A) to create the data pattern
following from Shafir’s framework. The simulation does
not support the task-compatibility framework as an ex-
planation of our results.
Additionally, the results of study 3 run counter to what
Shafir’s framework would predict. According to the task-
compatibility framework people should have been focusing
more on the minutes offered (the benefit, or the positively
valenced attribute) versus the penalties charged (the cost,
or the negatively valenced attribute) in the choice task; and
focusing more on the penalties changed versus the minutes
offered in the rejection task. In the choice task, an
increased focus on the number of minutes offered
(160 minutes in plan A and 200 minutes in plan B)
combined with little attention to the penalties charged
should have led the participants to prefer plan B to plan A.
In contrast, in the rejection task, an increased focus on the
penalties (lenient in plan a and strict in plan B) combined
with little attention to the number of minutes offered
should have led the participants to prefer plan A to plan B.
However, the opposite was true: people were less likely to
choose plan a in the rejection task than they were in the
choice task (P
choose
¼70% vs. P
reject
¼51%). At the
same time, this pattern is consistent with our hypothesis
that rejection entails more deliberative processing,
whereby people are more likely to consider both the num-
ber of minutes offered and the penalties charged in the cell
phone plan scenario.
Implications for Practice
Our research offers several practical implications for
consumers and managers. Our findings imply that people
could benefit from reformulating their decisions into rejec-
tion tasks when they want to behave more rationally. For
example, when looking at Web sites that provide overall
ratings of different options (e.g. hotels, restaurants, mov-
ies), as well as sample reviews for these options, they may
try rejecting alternatives to form their consideration sets, if
they want to be more rational. Similarly, when working on
a difficult test, test takers may be better off by first con-
sidering the alternatives that should be filtered out, instead
of focusing on which option to choose, since the latter
strategy can make the test takers choose the option that im-
mediately “feels” right and less likely to carefully elaborate
on their decision.
From the perspective of firms, we suggest that Web site
designers pay greater attention to Web site interfaces—
whether they involve opting-in (choosing) or hiding (re-
jecting) buttons. Our findings imply that Pinterest, a Web
site mostly used for recreational purposes and thus less
congruent with deliberative processing, should adopt the
opt-in interface, which Pinterest does. Similarly, LinkedIn
and Research Gate, Web sites supporting job search, a pro-
cess that requires considerable deliberation, seem to follow
the right strategy when providing the option to “hide” the
unattractive job offers.
Future Research
An important question that remains to be addressed in
future research pertains to the moderating effect of individ-
ual differences on the effects of task type. We would
speculate that the motivation to do well in a task, and cog-
nitive ability, would increase deliberation in choice tasks
more than in rejection tasks and reduce the task-type
effects reported in the current article.
Future research could also examine the effect of delib-
erative processing manipulations across choice versus
SOKOLOVA AND KRISHNA 629
rejection conditions. We find that rejection decisions be-
come similar to choice decisions when people are less
prone to rely on deliberative processing (study 4 and study
5). Yet we may wonder whether choice decisions would
become more similar to rejection decisions when people
are prompted to rely on deliberation or are encouraged to
ignore their feelings. While we believe this is possible, we
also think that there could be potential asymmetries in de-
cision makers’ propensity to shift from versus to delibera-
tive processing. Specifically, we expect that people are
more easily swayed away from deliberative processing in
rejection than toward it in choice. Further research could
address this question.
At the same time, while we do believe that rejection can
prompt deliberative processing and increase decision qual-
ity in contexts beyond those explored in the current article,
it would be important to explore potential boundary condi-
tions and reversals of the observed effects. For example,
increased deliberation associated with rejection could po-
tentially increase the amount of motivated reasoning and
information distortion, making people more likely to arrive
at their “desired,” but not necessarily accurate, conclusions
(Kunda 1990;Meloy and Russo 2004;Mishra, Shiv, and
Nayankankuppam 2008;Risen 2016). Similarly, rejection
may produce less accurate decisions in contexts where reli-
ance on feelings and intuitions is actually beneficial.
Recent research suggests that feeling-based (vs. delibera-
tive) processing may sometimes lead to superior decision
outcomes (Dijksterhuis 2004;Pham, Lee, and Stephen
2012). For example, feeling-based processing can outper-
form deliberation in contexts where access to uncon-
sciously acquired information (e.g., weather patterns) is
necessary to make more accurate decisions (Pham et al.
2012). An exploration of the effects of rejection in similar
contexts could provide insights into the boundary condi-
tions of the task-type effect.
In conclusion, whereas the boundary conditions of the
task-type effect merit further exploration, this research
shows that formulating a task as a choice or as a rejection
has important implications for how consumers evaluate al-
ternatives, and for what decisions they make.
DATA COLLECTION INFORMATION
The first author managed the collection of data for stud-
ies 1A, 1B, 2, 3, and 5 using the MTurk online panel be-
tween fall 2014 and fall 2015. The first author managed
the data collection for studies 1C and 4 at the University of
Michigan Ross Behavioral Research Lab in fall 2015 using
the help of research assistants. The first and second authors
jointly analyzed these data.
630 JOURNAL OF CONSUMER RESEARCH
Rating Randomly selected hotel reviews
BAD RATING – GOOD REVIEW
Rating from 138 users
6.7
Value for money: 6.9
Staff: 6.8
Facilities: 6.5
Location: 6.8
Comfort: 6.6
Cleanliness: 6.7
John, USA
9.6
The staff was amazing! Joshua at the front desk was great! I loved the
room style, we overlooked the city on the 13th floor. This hotel smells like
the beach! It is a wonderful soothing scent..:) Overall it was a nice stay
and the location was great.
Terry, USA:
8.9
I have never been more impressed with the value of a hotel. The hotel staff
was also very polite and accommodating. My two business colleagues
traveling with me were just as impressed. I would stay here every trip to
the city.
GOOD RATING – BAD REVIEW
Rating from 137 users
9.1
Value for money: 8.8
Staff: 9.1
Facilities: 9.2
Location: 9.4
Comfort: 9.1
Cleanliness: 8.9
Marc, USA:
7.1
Nothing was replaced in my room. When I drank my 2 small bottles of water
they were never replaced. When I used the 1 chamomile tea bag it was
never replaced. I don’t think I should have to ask to have water in my
room daily or tea bags. Very disappointing.
Bill, USA:
7.3
The front desk staff were good. Parking was the worst. Shower leaking all
over bath floor. Elevators appeared to be having some serious issues
including very long waits all the time for one to come.
FILLER
Rating from 134 users
7.6
Value for money: 7.7
Staff: 7.6
Facilities: 7.4
Location: 7.8
Comfort: 7.8
Cleanliness: 7.5
Deanna, USA:
7.7
The hotel is nicely designed and is clean. They’re having some “break-in”
issues such as bad light bulbs and misaligned glass shower doors. The lo-
cation is convenient for accessing public transport.
Sean, USA:
7.4
The hotel is within walking distance of many attractions as well as train and
bus stops which will take you to farther-off places of interest. Good places
to eat nearby as well. The staff was helpful.
Appendix A
Examples of Ratings and Reviews Used in Study 2
SOKOLOVA AND KRISHNA 631
Appendix D
Stimuli Used in Study 5
REFERENCES
Alter, Adam L., Daniel M. Oppenheimer, Nicholas Epley, and
Rebecca N. Eyre (2007), “Overcoming Intuition:
Metacognitive Difficulty Activates Analytic Reasoning,”
Journal of Experimental Psychology: General, 136 (4),
569–76.
Chatterjee, Subimal, Timothy B. Heath, Sandra J. Milberg, and
Karen R. France (2000), “The Differential Processing of
Price in Gains and Losses: The Effects of Frame and Need
for Cognition,” Journal of Behavioral Decision Making,13
(1), 61–75.
Cheema, Amar and Vanessa M. Patrick (2012), “Influence of
Warm versus Cool Temperatures on Consumer Choice: A
Resource Depletion Account,” Journal of Marketing
Research, 49 (6), 984–95.
Couch, Arthur and Kenneth Keniston (1960), “Yeasayers and
Naysayers: Agreeing Response Set as a Personality
Variable,” Journal of Abnormal and Social Psychology,60
(2), 151–71.
De Martino, Benedetto, Dharshan Kumaran, Ben Seymour, and
Raymond J. Dolan (2006), “Frames, Biases, and Rational
Decision-Making in the Human Brain,” Science, 313 (5787),
684–87.
DeKay, Michael L., Dalia Pati~
no-Echeverri, and Paul S.
Fischbeck (2009), “Distortion of Probability and Outcome
Information in Risky Decisions,” Organizational Behavior
and Human Decision Processes, 109 (1), 79–92.
Dhar, Ravi and Klaus Wertenbroch (2000), “Consumer Choice
Between Hedonic and Utilitarian Goods,” Journal of
Marketing Research, 37 (1), 60–71.
Dijksterhuis, Ap (2004), “Think Different: The Merits of
Unconscious Thought in Preference Development and
Decision Making,” Journal of Personality and Social
Psychology, 87 (5), 586–98.
Evans, Anthony M., Kyle D. Dillon, and David G. Rand (2015),
“Fast But Not Intuitive, Slow But Not Reflective: Decision
Conflict Drives Reaction Times in Social Dilemmas,”
Journal of Experimental Psychology: General, 144, 951–66.
Evans, Jonathan St. BT and Keith E. Stanovich (2013), “Dual-
Process Theories of Higher Cognition Advancing the
Debate,” Perspectives on Psychological Science, 8 (3),
223–41.
Fazio, Russell H. (1990), “A Practical Guide to the Use of Response
Latency in Social Psychological Research,” Research Methods
in Personality and Social Psychology, 11, 74–97.
Gilbert, Daniel T., Romin W. Tafarodi, and Patrick S. Malone
(1993), “You Can’t Not Believe Everything You Read,”
Journal of Personality and Social Psychology, 65 (2),
221–33.
Gonzalez, Cleotilde, Jason Dana, Hideya Koshino, and Marcel
Just (2005), “The Framing Effect and Risky Decisions:
Examining Cognitive Functions with fMRI,” Journal of
Economic Psychology, 26 (1), 1–20.
Heller, Daniel, Irwin P. Levin, and Martin Goransson (2002),
“Selection of Strategies for Narrowing Choice Options:
Antecedents and Consequences,” Organizational Behavior
and Human Decision Processes, 89 (2), 1194–213.
Hochman, Guy and Eldad Yechiam (2011), “Loss Aversion in the
Eye and in the Heart: The Autonomic Nervous System’s
Responses to Losses,” Journal of Behavioral Decision
Making, 24 (2), 140–56.
Huber, Vandra L., Margaret A. Neale, and Gregory B. Northcraft
(1987), “Decision Bias and Personnel Selection Strategies,”
Organizational Behavior and Human Decision Processes,40
(1), 136–47.
Knowles, Eric S. and Christopher A. Condon (1999), “Why
People Say “Yes”: A Dual-Process Theory of Acquiescence,”
Journal of Personality and Social Psychology, 77 (2),
379–86.
EVALUATION
CONTROL FEELING-BASED
TASK TYPE CHOICE You need to choose one of the
two options:
A. a sure gain of $240
B. 25% chance to gain $1000, and 75% chance
togain nothing
Which option would you CHOOSE? (95%*)
You need to choose one of the two options:
A. a sure gain of $240
B. 25% chance to gain $1000, and 75% chance to gain nothing
Think about how you feel about each of the two options
and decide which option you LIKE MORE.
Based on that decide – which option would you CHOOSE?
(89%*)
REJECTION You need to reject one of the two options:
A. a sure gain of $240
B. 25% chance to gain $1000, and 75% chance
to gain nothing
Which option would you REJECT? (52%*)
You need to reject one of the two options:
A. a sure gain of $240
B. 25% chance to gain $1000, and 75% chance to gain nothing
Think about how you feel about each of the two options and
decide which option you LIKE LESS.
Based on that decide – which option would you REJECT?
(73%*)
*Share of the riskless option
634 JOURNAL OF CONSUMER RESEARCH
Krajbich, Ian, Bjo¨rn Bartling, Todd Hare, and Ernst Fehr (2015),
“Rethinking Fast and Slow Based on a Critique of Reaction-
Time Reverse Inference,” Nature Communications, 6 (7455),
2–9.
Krishnamurthy, Parthasarathy and Anish Nagpal (2008), “The
Impact of Goal Framing on the Choose–Reject Discrepancy,”
Advances in Consumer Research—North American
Conference Proceedings, 35, 784–85.
Kunda, Ziva (1990), “The Case for Motivated
Reasoning,” Psychological Bulletin, 108 (3), 480–98.
Kuvaas, Ba˚rd and Marcus Selart (2004), “Effects of Attribute Framing
on Cognitive Processing and Evaluation,” Organizational
Behavior and Human Decision Processes, 95 (2), 198–207.
Laran, Juliano and Keith Wilcox (2011), “Choice, Rejection, and
Elaboration on Preference-Inconsistent Alternatives,”
Journal of Consumer Research, 38 (2), 229–41.
Malkoc, Selin A., William Hedgcock, and Steve Hoeffler (2013),
“Between a Rock and a Hard Place: The Failure of the
Attraction Effect among Unattractive Alternatives,” Journal
of Consumer Psychology, 23 (3), 317–29.
Mantonakis, Antonia, Pauline Rodero, Isabelle Lesschaeve, and
Reid Hastie (2009), “Order in Choice Effects of Serial Position
on Preferences.” Psychological Science, 20 (11), 1309–12.
Meloy, Margaret G. and J. Edward Russo (2004), “Binary Choice
Under Instructions to Select versus Reject,” Organizational
Behavior and Human Decision Processes, 93 (2), 114–28.
Mishra, Himanshu, Arul Mishra, and Dhananjay Nayakankuppam
(2007), “Seeing Through the Heart’s Eye: The Interference of
System 1 in System 2,” Marketing Science, 26 (5), 666–78.
Mishra, Himanshu, Baba Shiv, and Dhananjay Nayankankuppam
(2008), “The Blissful Ignorance Effect: Pre- versus Post-
Action Effects on Outcome Expectancies Arising from
Precise and Vague Information,” Journal of Consumer
Research, 35 (4), 573–85.
Park, C. Whan, Sung Youl Jun, and Deborah J. MacInnis (2000),
“Choosing What I Want versus Rejecting What I Do Not
Want: An Application of Decision Framing to Product
Option Choice Decisions,” Journal of Marketing
Research, 37 (2), 187–202.
Pham, Michel Tuan, Leonard Lee, and Andrew T. Stephen (2012),
“Feeling the Future: The Emotional Oracle Effect,” Journal
of Consumer Research, 39 (3), 461–77.
Pocheptsova, Anastasiya, On Amir, Ravi Dhar, and Roy F.
Baumeister (2008), “Deciding Without Resources:
Psychological Depletion and Choice in Context,” Journal of
Marketing Research, 46(3), 344–55.
Preacher, Kristopher J. and Andrew F. Hayes (2008), “Asymptotic
and Resampling Strategies for Assessing and Comparing
Indirect Effects in Multiple Mediator Models,” Behavior
Research Methods, 40 (3), 879–91.
Risen, Jane L. (2016), ‘Believing What We Do Not Believe:
Acquiescence to Superstitious Beliefs and Other Powerful
Intuitions,” Psychological Review, 123 (2), 182–207.
Shafir, Eldar (1993), “Choosing versus Rejecting: Why Some
Options Are Both Better and Worse than Others.” Memory &
Cognition, 21 (4), 546–56.
Simon, Andrew F., Nancy S. Fagley, and Jennifer G. Halleran
(2004), “Decision Framing: Moderating Effects of Individual
Differences and Cognitive Processing,” Journal of
Behavioral Decision Making, 17 (2), 77–93.
Smith, Stephen M. and Irwin P. Levin (1996), “Need for
Cognition and Choice Framing Effects,” Journal of
Behavioral Decision Making, 9, 283–90.
Stroop, J. Ridley (1935), “Studies of Interference in Serial Verbal
Reactions,” Journal of Experimental Psychology, 18 (6),
643–62.
Tversky, Amos and Daniel Kahneman (1973), “Availability: A
Heuristic for Judging Frequency and Probability,” Cognitive
Psychology, 5 (2), 207–32.
——— (1981), “The Framing of Decisions and the Psychology of
Choice,” Science, 211 (4481), 453–58.
Webb, Thomas L., and Paschal Sheeran (2003), “Can
Implementation Intentions Help to Overcome Ego-
Depletion?” Journal of Experimental Social Psychology, 39,
279–86.
Yang, Zhiyong, Ritesh Saini, and Traci Freling (2015), “How
Anxiety Leads to Suboptimal Decisions Under Risky Choice
Situations,” Risk Analysis, 35 (10), 1789–800.
Yaniv, Ilan and Yaacov Schul (1997), “Elimination and Inclusion
Procedures in Judgment,” Journal of Behavioral Decision
Making, 10 (3), 211–20.
——— (2000), “Acceptance and Elimination Procedures in Choice:
Noncomplementarity and the Role of Implied Status Quo,”
Organizational Behavior and Human Decision Processes,82
(2), 293–313.
Yechiam, Eldad and Guy Hochman (2013), “Loss-Aversion or
Loss-Attention: The Impact of Losses on Cognitive
Performance,” Cognitive Psychology, 66 (2), 212–31.
SOKOLOVA AND KRISHNA 635
CopyrightofJournalofConsumerResearchisthepropertyofOxfordUniversityPress/USA
anditscontentmaynotbecopiedoremailedtomultiplesitesorpostedtoalistservwithout
thecopyrightholder'sexpresswrittenpermission.However,usersmayprint,download,or
emailarticlesforindividualuse.
