XXX10.1177/0956797618783747
corrigendum2018
Corrigendum
Corrigendum: Group Influences on
Engaging Self-Control: Children Delay
Gratification and Value It More When
Their In-Group Delays and Their
Out-Group Doesn’t
Psychological Science
1
© The Author(s) 2018
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797618783747
https://doi.org/10.1177/0956797618783747
www.psychologicalscience.org/PS
Original article: Doebel, S., & Munakata, Y. (2018). Group influences on engaging self-control: Children delay
gratification and value it more when their in-group delays and their out-group doesn’t. Psychological Science, 29,
738–748. doi:10.1177/0956797617747367
In our original article, we discussed a hazard ratio as
though it were an odds ratio in three places. This Corrigendum corrects these errors. Note that the statistics associated with these sentences will not change, nor do these
changes affect the study’s conclusions whatsoever.
First, the second sentence of the second paragraph
of the Results of Experiment 1 (p. 742) will be changed
to read as follows: “Children whose group did not wait
for two marshmallows (wait time Mdn = 9.48 min) gave
into the temptation to eat the marshmallow at nearly
twice the rate of children whose group waited (Mdn =
15.00 min), χ2(1) = 3.39, p = .07, hazard ratio = 1.95,
95% confidence interval (CI) = [0.94, 4.01].” The second
sentence of the fourth paragraph in the same section
(p. 742) will be changed to read as follows: “Children
who did not receive any information about their group’s
delay behavior (wait time Mdn = 2.95 min) gave into
the temptation to eat the marshmallow at over twice
the rate of children whose in-group waited, χ2(1) =
4.57, p = .033, hazard ratio = 2.18, 95% CI = [1.05, 4.50].”
Finally, the first sentence of the second paragraph
of the Results of Experiment 2 (p. 745) will be changed
to read as follows: “Consistent with Experiment 2’s
first preregistered hypothesis and replicating the findings of Experiment 1, children whose group did not
wait for two marshmallows (Mdn = 5.77 min) gave into
the temptation to eat the marshmallow at twice the rate
of children whose group waited (wait time Mdn = 14.95
min), χ2(1) = 5.96, p = .015, hazard ratio = 2.05, 95%
CI = [1.14, 3.70].”
747367
research-article2018
PSSXXX10.1177/0956797617747367Group Influences on Self-ControlDoebel, Munakata
Research Article
Group Influences on Engaging
Self-Control: Children Delay Gratification
and Value It More When Their In-Group
Delays and Their Out-Group Doesn’t
Psychological Science
2018, Vol. 29(5) 738–748
© The Author(s) 2018
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797617747367
https://doi.org/10.1177/0956797617747367
www.psychologicalscience.org/PS
TC
Sabine Doebel and Yuko Munakata
Department of Psychology and Neuroscience, University of Colorado Boulder
Abstract
Self-control emerges in a rich sociocultural context. Do group norms around self-control influence the degree to which
children use it? We tested this possibility by assigning 3- to 5-year-old children to a group and manipulating their
beliefs about in-group and out-group behavior on the classic marshmallow task. Across two experiments, children
waited longer for two marshmallows when they believed that their in-group waited and their out-group did not,
compared with children who believed that their in-group did not wait and their out-group did. Group behavior
influenced children to wait more, not less, as indicated by comparisons with children in a control condition who were
assigned to a group but received no information about either groups’ delay behavior (Experiment 1). Children also
subsequently valued delaying gratification more if their in-group waited and their out-group did not (Experiment 2).
Childhood self-control behavior and related developmental outcomes may be shaped by group norms around selfcontrol, which may be an optimal target for interventions.
Keywords
self-control, executive function, social influences, cognitive development, open data, open materials, preregistered
Received 5/3/17; Revision accepted 10/30/17
What shapes our ability to exercise self-control—to
inhibit the urge to order chocolate cake when on a diet
or to save money now for a bigger payoff later? People
vary in how successful they are at engaging self-control.
Children are notorious for struggling with it. Childhood
self-control predicts important life outcomes (e.g.,
employment, academic success, health, wealth, criminality; Mischel, Shoda, & Rodriguez, 1989; Moffitt et al.,
2011), suggesting it is an early-emerging and stable individual difference that supports adaptive behavior throughout life. Identifying factors that predict self-control in
childhood and beyond is thus of considerable interest.
Self-control is often explained in terms of earlydeveloping cognitive skills, such as executive functions
(e.g., goal maintenance, working memory, and inhibition), prefrontal circuitry associated with executive
functions (Bunge & Zelazo, 2006; Casey et al., 2011),
and psychological traits (e.g., Duckworth, Tsukayama,
& Kirby, 2013; Moffitt et al., 2011). However, self-control
emerges not in isolation but in a rich sociocultural
context that may influence how it is exercised and
develops. Self-control is linked to parenting (e.g.,
Bernier, Carlson, & Whipple, 2010; Lamm et al., 2017)
and socioeconomic status (SES; e.g., Hackman, Gallop,
Evans, & Farah, 2015) and may be moderated by beliefs
about the reliability of the environment and social trust.
Children and adults wait less for rewards when the
person providing the reward is unreliable (e.g., Kidd,
Palmeri, & Aslin, 2013) or untrustworthy (Michaelson,
de la Vega, Chatham, & Munakata, 2013; Michaelson &
Munakata, 2016). Children also adjust whether they
exercise self-control in response to an adult model’s
Corresponding Author:
Sabine Doebel, Department of Psychology and Neuroscience,
University of Colorado Boulder, 1805 Colorado Ave., Muenzinger 345
UCB, Boulder, CO 80309
E-mail: sabine.doebel@colorado.edu
Group Influences on Self-Control
delay behavior (Bandura & Mischel, 1965) or undesirable outcomes (Lee & Carlson, 2015).
We propose and test another possibility—that selfcontrol is influenced by group membership and norms.
Adults and children can be strongly influenced by the
behavior and norms of others in their attitudes, judgments, and behaviors (Kim, Chen, Smetana, & Greenberger,
2016; Myers & Lamm, 1976). While self-control might
seem like a universally valued behavior, the importance
placed on self-control in modern society is relatively new
(Pinker, 2011). And while self-control is prominent in
both popular press coverage and scientific journal articles
today, cultures, communities, peer groups, and family
members may vary in how much they emphasize selfcontrol and the importance of using it to achieve goals
(Carlson & Zelazo, 2011; Lamm et al., 2017). Thus, variations in group behavior and norms around self-control
might influence whether self-control is used and how it
develops. Cultures differ in delaying gratification (e.g.,
Lamm et al., 2017); however, such patterns could be due
to either cultural norms around self-control or other differences between cultures.
One way to assess influences of group norms is by
manipulating children’s identification with a group who
shows or does not show self-control. Group behavior
motivates children from infancy (e.g., Xiao et al., 2017).
Preschoolers show in-group biases based on factors
such as eye color and accent (Bigler, Jones, & Lobliner,
1997; Kinzler, Corriveau, & Harris, 2011) even when
there is no substantive basis for group membership
(Dunham, Baron, & Carey, 2011), and such biases influence their actions ( Jordan, McAuliffe, & Warneken,
2014). Thus, if group norms play a role in self-control,
then learning that one’s in-group engages in self-control
(and one’s out-group does not) should influence children to increase their self-control behavior.
Gaining insight into the influence of group behavior
on self-control is timely given interest in interventions
to support the development of self-control and the mixed
findings thus far (Diamond & Lee, 2011). Common issues
are the absence of long-term effects and limited improvements that are generally confined to tasks similar to
those used to train control (so-called narrow transfer).
If group behavior influences children’s tendency to use
self-control, then interventions that capitalize on this
(e.g., by working to influence norms and values) may
achieve broader transfer and longer lasting gains.
We examined the influence of group membership on
self-control behavior by testing whether in-group identity influences children’s tendency to delay gratification
on the classic marshmallow task, in which children are
given the option of having a small reward right away
(one marshmallow) or waiting for a larger reward (two
marshmallows; Mischel et al., 1989). In two experiments,
739
we assigned children to an in-group using a minimalgroups induction procedure that creates in-group bias
in young children (Dunham et al., 2011) and manipulated the degree to which self-control was presented as
a group norm. Children were randomly assigned to conditions varying in what children were told about the
behavior of their group and of another group on the
marshmallow task. We predicted that children would
delay gratification longer on the task when their group
delayed while the other group did not, compared with
children whose group did not delay gratification while
the other group did.
Experiment 1
Method
Participants. A total of 98 children 3 to 4 years old (age:
M = 4.22 years, SD = 0.33, range = 3.55–5.00; males = 46)
were recruited from a database of families who had previously indicated interest in participating in child development research. Of these children, 8 were excluded because
they became upset while waiting for two marshmallows
(n = 1), required their parent to stay in the room throughout the procedure (n = 2), left the room (n = 2), or ate the
marshmallow before being exposed to the manipulation
(n = 1), or because of experimenter error that affected the
manipulation (n = 1) or equipment failure that affected the
manipulation (n = 1). We selected our sample size with a
view to the feasibility of collecting the data within an academic semester and to roughly double that of previous
studies (Kidd et al., 2013; Michaelson & Munakata, 2016).
A power analysis informed by previous effect sizes was
not possible because of a lack of precedent for this specific experimental manipulation in the literature. Data
were collected between February and May 2016.
For 97% of our participants, at least one parent had
a 4-year college degree or higher, 2% completed high
school and some college, and 1% completed high
school only. The racial makeup of the sample was 97%
Caucasian, 1% African American, 1% Native Hawaiian,
and 1% American Indian. Ninety-six percent of the
sample was non-Hispanic/non-Latino.
Design. We employed a between-subjects experimental
design, randomly assigning each child to one of three conditions in which (a) in-group members waited for a larger
reward and out-group members did not (group-waited
condition), (b) in-group members did not wait for a larger
reward and out-group members did (group-did-not-wait
condition), and (c) no information was provided about ingroup or out-group members’ delay behavior (control condition). All children then completed the marshmallow test
(Mischel et al., 1989).
740
Procedure. Children were brought into a small room and
seated at a white table. A video camera was set up on a
tripod to record the session, along with a webcam, which
fed into an adjacent room where the child’s parent was
observing. A lidded opaque rectangular container contained
the plated marshmallow and green T-shirt. The room was
otherwise empty and without decoration. The experiment
began with the in-group induction, followed by an introduction to the marshmallow test, the experimental manipulation, and then the marshmallow test proper. Children
were also asked questions at different points in the procedure, which served as manipulation checks.
Following Dunham et al. (2011), we assigned each
child to a group and did not use any procedure to
specifically encourage in-group affiliation. The experimenter knelt beside the seated child and said, “Guess
what? You get to be in the green group today. Alright.
Just like these kids here.” The experimenter placed in
front of the child a laminated 8- × 12-in. card depicting
headshots of four children wearing green shirts presented in a 2 × 2 format with a boy on the left and a
girl on the right in the top row and the reverse pattern
on the bottom row. These pictures have been used and
validated in previous research on young children’s ingroup preferences (Dunham et al., 2011). The children
in the pictures appear to be 4 to 5 years of age. All of
the children were Caucasian.
The experimenter continued, “The green group is
great. Here’s a special green shirt for you to wear
because you’re in the green group.” The experimenter
then helped the child put on the shirt. Next, the experimenter placed on the table (to the right or left of the
in-group, randomized between participants) a picture
of the out-group and said, “Now look here, this is the
orange group. This isn’t your group; it’s a different
group.” The format and gender makeup were identical
across the in- and out-groups. The children also looked
similar across the groups, but to ensure no biases
toward particular children, we randomized across participants which group of children was shown with
green shirts and presented as the in-group.
The experimenter then asked, “So which group are
you in?” If the child hesitated, the experimenter repeated
the question. Once the child responded, the experimenter said, “Yes, you’re in the green group.” If the
child responded incorrectly, the experimenter said,
“Actually, you’re in the green group.” Next, the experimenter presented children with an envelope, opened
it, pulled out a small sticker, and said,
So, I have this present, a sticker, that I want to
give away. I will put it in this envelope and give
it to the kids you tell me to give it to. Who should
I give it to? Can you point or tell me?
Doebel, Munakata
The experimenter recorded children’s responses and
did not provide any feedback. The experimenter then
began the marshmallow test (adapted from Kidd et al.,
2013, and Michaelson & Munakata, 2016). The experimenter placed a single marshmallow on a black plate
centered in front of the child, 4 in. from the edge of
the table, and announced,
Okay, it’s snack time! You have a choice for your
snack. You can eat this one marshmallow right
now or—if you wait for me to go get more
marshmallows from the other room—you can have
two marshmallows to eat instead. How does that
sound?
If the child declared that he or she just wanted one
marshmallow, the experimenter simply repeated the
choice, saying, “Okay, well, you can have this one marshmallow now or you can wait and get two later!” If the
child attempted to grab the marshmallow, the experimenter said, “Oh, let me tell you something else first!”
The experimenter placed a laminated 8- × 12-in.
picture depicting the in-group (in green shirts) and
out-group (in orange shirts) on a small bookstand 4 in.
behind the plated marshmallow. In the group-waited
condition, the experimenter pointed to the images on
the laminated sheet and said,
Look! These kids are in the green group, too, just
like you! And guess what? They waited until they
could have two marshmallows. And these kids are
in the orange group, and they didn’t wait until
they could have two marshmallows.
The left half of the laminated picture depicted a row
of headshots of the four in-group members, below
which were images of marshmallows: On the left was
an image of one marshmallow (representing the immediate reward), and on the right was an image of two
marshmallows (representing the delayed reward). Consistent with the in-group waiting for the larger reward,
an arrow pointed from the in-group to the two marshmallows. The right half of the laminated picture
depicted the four out-group members above the same
marshmallow images but with an arrow pointing to the
single marshmallow instead of the two marshmallows,
consistent with the out-group not waiting for the larger
reward.
In the group-did-not-wait condition, the laminated
picture was identical except that the arrows pointing
to the marshmallows were consistent with the ingroup not waiting for the larger reward and the outgroup waiting for the larger reward. The experimenter
said,
Group Influences on Self-Control
Look! These kids are in the green group, too, just
like you! And guess what? They didn’t wait until
they could have two marshmallows. And these
kids are in the orange group, and they waited
until they could have two marshmallows.
In the control condition, the laminated picture
depicted the in-group and the out-group in the same
manner, but did not depict the marshmallows and thus
provided no information about the groups’ delay behavior. The experimenter said, “Look! These kids are in the
green group, too, just like you! And these kids are in
the orange group.” We opted not to include a statement
about the in-group receiving a choice between waiting
and having two marshmallows because we anticipated
doing so could induce children to speculate about their
group’s behavior, potentially influencing children’s wait
times.
The experimenter then asked the child, “So what did
the green group do?” If the child said he or she did not
know or provided an otherwise uninformative answer,
the experimenter asked, “Did they wait or not wait?” If
the child responded correctly, the experimenter said,
“Yes that’s right they waited until they could have two
marshmallows.” If the child responded incorrectly, the
experimenter said, “Actually, they waited until they
could have two marshmallows.” In the control condition, this question was omitted.
The experimenter then said, “Okay, I’m going to go
get more marshmallows from the other room. I’ll leave
this marshmallow here for you to eat if you want to
while I’m gone, and if you haven’t eaten any of it when
I come back, you can have two marshmallows instead.”
At this point, the experimenter left the room and went
to the adjacent room to watch the child via webcam.
The session was stopped after the child took a bite or
lick of the marshmallow or waited the full 15 min. At
this point, the experimenter returned to the room and
said, “Okay, all done with snack time!” and gave children the second marshmallow if they waited the full
15 min. The experimenter then proceeded with the
posttest questions. The session was discontinued if children became distressed or left the room.
After the test, the experimenter then asked each child
which group he or she was in. This memory check was
identical to the first, except that no feedback was provided. Finally, the experimenter laid out the pictures of
the two groups and asked the child, “Which kids are
nicer?” A verbal or point response was recorded.
A coder who was blind to the experimental condition
and study hypotheses recorded when each child’s first
lick or bite of the marshmallow occurred (as in Kidd
et al., 2013; Michaelson & Munakata, 2016). To ensure
reliability, we checked the coding against that of a
741
second naive coder, r(86) = .98, p < .001. Coders differed by less than 3 s on 94% of judgments. The five
cases in which the coding discrepancy was larger than
3 s were discussed and resolved by the coders. A
research assistant who was blind to the experimental
condition and study hypotheses then transcribed all
self-talk in the videos. A naive coder recorded whether
children made any reference to the in- and out-groups
(green and orange, respectively) while waiting. Coding
was checked against that of a second naive coder, and
there were no discrepancies.
Analytic approach. The study design and analytic plan
were preregistered on the Open Science Framework
(https://osf.io/wrqtu/). Our preregistered hypothesis was
that children whose group waited for two marshmallows
would wait longer for two marshmallows than children
whose group did not wait. All analyses beyond the confirmatory test comparing the group-waited and group-didnot-wait conditions were exploratory. Our preregistered
analytic plan was to conduct a confirmatory test of this
hypothesis using linear regression; however, the dependent variable was heavily right-censored, rendering linear
regression inappropriate as it would produce biased estimates and inflate the Type II error rate (Mills, 2011). Thus
we abandoned the planned linear analysis in favor of the
more appropriate survival analysis using Cox proportional hazards regression models, developed specifically
to handle censored data (Cox, 1972). All analyses were
conducted using the R statistical package (R Core Team,
2006) and the lme4 add-on package to implement mixedeffects models (Bates, Mächler, Bolker, & Walker, 2015).
No data were excluded. One session was terminated
early during the delay period (at 9 min 51 s) because of
experimenter error (the experimenter mistakenly thought
the child had tasted the marshmallow); however, because
the data point was valid (i.e., the only difference from
other censored data points is that censoring occurred
earlier), it was retained, and the delay time was coded as
censored. Because of an equipment malfunction, video
recordings of sessions were not available for 2 participants in the control group. One of these participants
unambiguously waited the entire 15-min period, and the
other participant ate the marshmallow 15 s after the
experimenter left the room, as coded by the experimenter
using the webcam and a timer. Excluding these participants from the sample did not affect the results.
Results
All children remembered which group they were in,
both before and after the marshmallow test. Children
also identified with their group: 80% of children chose
to give a sticker to their group instead of the out-group
Doebel, Munakata
742
prior to learning anything about their group’s behavior,
and, after the marshmallow test, 81% indicated that their
group was nicer. These rates were well above chance,
ts > 7, ps < .001. Almost all children remembered, both
prior to and following the marshmallow test, whether
their group waited or did not wait—pretest: M = 94% (51
of 54; 6 children were not asked this question because
of experimenter error), SD = 23%, t(53) = 14.13, p < .001;
posttest: M = 90% (54 of 60), SD = 30%, t(59) = 10.24,
p < .001.
The results of our confirmatory test were consistent with
our preregistered hypothesis, although they did not meet
the traditional .05 alpha threshold. Children whose group
did not wait for two marshmallows (wait time Mdn =
9.48 min) gave into the temptation to eat the marshmallow at nearly twice the rate of children whose group
waited (Mdn = 15.00 min), χ2(1) = 3.39, p = .07, hazard
ratio = 1.95, 95% confidence interval (CI) = [0.94, 4.01].
We also conducted an exploratory analysis with age
included as a covariate because age tends to correlate
with self-control ability; in this model, the greater likelihood of resisting the marshmallow for children whose
group waited was significant, χ2(1) = 4.19, p = .041,
hazard ratio = 2.13, 95% CI = [1.02, 4.43]. The effect of
age was not statistically significant (p = .1277).
Exploratory comparisons with the control condition
suggested that group behavior influenced children to
wait longer but did not influence children to wait less.
Children who did not receive any information about their
group’s delay behavior (wait time Mdn = 2.95 min) gave
into the temptation to eat the marshmallow at over twice
the rate of children whose in-group waited, χ2(1) = 4.57,
p = .033, hazard ratio = 2.18, 95% CI = [1.05, 4.50]. On
the other hand, there was no significant difference in the
wait times of children whose group members did not
wait for two marshmallows and children who did not
receive information about their group’s delay behavior,
χ2(1) = 0.16, p > .250, hazard ratio = 0.88, 95% CI = [0.46,
1.66]. An exploratory analysis in which age and gender
were added to the model did not change these results,
χ2(1) = 0.16, p > .250, hazard ratio = 0.87, 95% CI = [0.45,
1.7].
This pattern was further confirmed by children’s verbalized reasons for waiting. Among those in the groupwaited condition, 24% of children (7 of 29; 1 additional
child did not answer the question) cited the in- or
out-group’s behavior in their reason for waiting (e.g.,
“Because the green group waited,” “Because I’m in the
green group,” “Orange didn’t wait”). By contrast, 3%
(1 of 29; 1 additional child did not answer the question) in the group-did-not-wait condition cited in- or
out-group behavior in their reason for not waiting,
χ2(1) < 3.62, p = .057. Similarly, among children who
did not receive information about their group’s
behavior, 0% (0 of 29; 1 additional child was not asked
the question because of experimenter error) cited the
groups’ behavior as their reason for waiting or not
waiting, significantly less than in the group-waited condition, χ2(1) = 5.84, p = .016. The number of children
who referred to the groups’ behavior in the group-didnot-wait and control conditions did not differ, χ 2(1) <
1, p > .250.
Children’s self-talk during the delay period was also
consistent with our primary findings. Seventy-three percent of children engaged in some form of self-talk during the delay period, and this did not vary by condition,
p > .250. Children in all conditions talked about the
green group (25%) and talked much less about the
orange group (10%), χ2(1) = 4.65, p = .031, consistent
with in-group identification. The number of children
who talked about the green group in the group-waited
condition (40%) was higher than in the group-did-notwait condition (17%), χ2(1) = 4.02, p = .045, and marginally higher than in the control condition (18%), χ2(1) =
3.43, p = .064. There was no significant difference
between the group-did-not-wait and control conditions,
p > .250.
We also explored the possibility that children
changed how they felt about their group after learning
about their group’s delay behavior, that is, that they
were more biased in favor of their group after learning
that their group did wait, and they were less biased in
favor of their group after learning that their group did
not wait. This would be consistent with children valuing
waiting over not waiting at baseline. A mixed logistic
regression (with random intercepts for participants to
account for dependency among observations) indicated
that the odds of children identifying with their group
did not vary depending on whether they were asked
before or after the marshmallow test, p > .250, whether
they were in the group-waited or group-did-not-wait
condition, p = .105, or the interaction of these factors,
p = .105. Numerically, children in the group-waited
condition showed more of an in-group bias after learning about their group’s behavior (pretest: M = 73%,
posttest: M = 90%), whereas children in the group-didnot-wait condition showed the opposite pattern (pretest: M = 73%, posttest: M = 67%).
Discussion
Experiment 1 provided the first evidence that group
behavior influences self-control in children: Children
tended to wait longer when their in-group delayed
gratification and their out-group did not, compared
with the reverse case. Group norms may have influenced children’s engagement of self-control because
they wanted to do what was normative in their group,
Group Influences on Self-Control
avoid doing what was normative in the out-group, or
both.
Alternatively, children may have been motivated to
behave in accord with their group’s behavior, without
changing their thinking about whether or not delaying
gratification is generally a good thing. In Experiment
2, we tested the possibility that group behavior influenced how children evaluate self-control.
While our group-waited manipulation induced children to wait longer, the group-did-not-wait condition
did not have a comparable negative effect on wait times.
In addition, although not statistically significant, children’s identification with their group numerically
increased after the marshmallow test in the group-waited
condition, and numerically decreased in the group-didnot-wait condition. These patterns suggest that children
may value delaying but cannot always act accordingly.
Learning that the in-group delayed and the out-group
did not may have supported delaying and increased ingroup identification, whereas learning that the in-group
did not delay and the out-group did may have led to
dissonance between the valuing of delay behavior and
in-group identity, resulting in a reduction in in-group
preference and no reduction in delaying. In Experiment
2, we added more questions to better test whether children’s identification with their group changed after learning about their group’s delay behavior.
Finally, in Experiment 2, we used shared preference
as a basis for children’s group membership (as in Billig
& Tajfel, 1973) and to increase group affiliation, which
was expected to strengthen our manipulation by making children more likely to behave in accordance with
their group.
Experiment 2
Experiment 2 tested three hypotheses. If group norms
influence self-control, then children in the group-waited
condition should (a) wait longer than children in the
group-did-not-wait condition (replicating Experiment
1), (b) evaluate new individuals who delay gratification
more positively than children in the group-did-not-wait
condition, and (c) identify more with their group than
children in the group-did-not-wait condition, if delaying
gratification is a preexisting norm.
Method
Participants. A total of 100 children 3 to 5 years old
(age: M = 4.39 years, SD = 0.33, range = 3.58–5.58, males =
40) were recruited to participate in this experiment, as
specified in our preregistered plan, using the same method
as in Experiment 1. Of these children, 13 were not included
in the final sample because they refused to complete the
743
tasks (n = 1), became upset while waiting for the two
marshmallows (n = 4), left the room (n = 2), or ate the
marshmallow before being exposed to the manipulation
(n = 5), or because of experimenter error that affected the
manipulation (n = 1), yielding a final sample of 87. Data
were collected between July 2016 and April 2017. For 95%
of our participants, at least one parent had a 4-year college
degree or higher, 4% completed high school and some college, and 1% completed high school only. The racial
makeup of the sample was 90% Caucasian and 10% mixed
race. The ethnic makeup of the sample was primarily nonHispanic/non-Latino (96%).
Design. This experiment had the same between-subjects
design as Experiment 1, in which we manipulated ingroup and out-group delay behavior. Each child was randomly assigned to either the group-waited condition or
the group-did-not-wait condition. We did not run the control condition in this experiment, as it was not relevant to
our hypotheses.1
Procedure. Experiment 2 followed the same general
procedure as Experiment 1 but with some adjustments to
increase power, control for potential confounds, and test
new hypotheses. We highlight these changes below.
Children were assigned to the in-group in much the
same way as in Experiment 1, but with two changes.
First, we used shared preferences as a basis for children’s
group membership to increase group affiliation (Billig
& Tajfel, 1973). Second, we made adjustments in wording
to ensure that children did not perceive the experimenter
as preferring one group over the other (avoiding the
potential confound of children following their group
because they wanted to please the experimenter).
The experimenter said, “I’m going to show you some
pictures of things and ask you about what you like!
Look at these animals. Can you point to the one you
like more?” The experimenter asked three more questions such as these about foods, toys, and treats and
recorded children’s choices. Next, the experimenter
said,
Okay, now look! This is the green group! The
green group likes the same things you like. They
like , , , and . Now
look here, this is the orange group! The orange
group likes different things than you like. They
like , , , and . So you
are going to be in the green group! Here’s a
special green shirt for you to wear because you’re
in the green group!
If a child refused to make a choice for one of the
questions, the experimenter noted this and omitted it
744
Doebel, Munakata
from the remainder of the procedure. As in Experiment
1, children were asked to indicate which group was
their group. They were also asked to indicate which
group was not their group. Correct responses were
affirmed and incorrect responses were corrected. As in
Experiment 1, children were also asked questions to
assess in-group identity, but in the current experiment,
four such questions were asked in each phase (i.e.,
pre–marshmallow-test and post–marshmallow-test) in
an effort to reduce measurement error and increase
power to detect effects involving this variable. Children
were asked questions such as “Who should I give a
sticker to?” and “Who helped their friend at school?”
The marshmallow test was introduced in the same
way as in Experiment 1. The wording of the experimental manipulation was adjusted to equate for length in
descriptions of each group’s behavior. In the groupwaited condition, the experimenter said,
An evaluation of delay choice (adapted from Shutts,
Banaji, & Spelke, 2010) was added in Experiment 2 to
test the hypothesis that group behavior influenced children via a change to how they evaluated delaying gratification. Children completed four trials in which they
were presented with scenarios involving a pair of children who differed in their delay behavior. Each trial
involved different rewards that were increasingly
abstract from what they had experienced in the experiment (marshmallows, cookies, stickers, and money).
They were shown a page with small pictures of two
children of the same gender, one on the right side of
the page, and one on the left. These images were from
the Child Affective Facial Expression (CAFE) stimuli set
hosted on Databrary (LoBue, 2014; LoBue & Thrasher,
2015). Each depicted child had small pictures of rewards
directly beneath them. A trial was introduced by the
experimenter saying, for example,
Look! These kids are in the green group, just like
you! And guess what? They didn’t have one
marshmallow; they waited until they could have
two marshmallows. And these kids are in the
orange group, not your group! They had one
marshmallow; they didn’t wait until they could
have two marshmallows.
Jenny and Kate love marshmallows! Their mom
said they could have one marshmallow right away,
or if they waited until she found more marshmallows,
they could have two instead. Jenny ate one marsh
mallow right away. Kate waited until she could
have two marshmallows.
In the group-did-not-wait condition, the experimenter said,
Look! These kids are in the green group, just like
you! And guess what? They had one marshmallow;
they didn’t wait until they could have two
marshmallows. And these kids are in the orange
group, not your group! They didn’t have one
marshmallow; they waited until they could have
two marshmallows.
Children were then asked to indicate what each
group did. If they failed to provide an informative
response, the experimenter said, “Did they wait or not
wait?” Correct responses were affirmed, and incorrect
responses were corrected. The marshmallow test was
then administered as in Experiment 1. Once the child
waited the full 15 min or tasted the marshmallow, the
experimenter returned to the room and said, “Snack
time is over! I just have a few more questions for you.”
Children were then asked several of the same questions
they were asked in the pre–marshmallow-test phase:
memory for group, memory for group behavior, and
group identity. The only difference was that no feedback was provided following children’s responses to
any of the questions.
The experimenter then asked three questions
designed to tap children’s implicit preference for one
of the children. For example, children were asked “Who
do you like more?” “Who is nicer?” and “Jenny loves
playing Kazoop. It’s her favorite game to play. Kate loves
playing Babber. It’s her favorite game to play. Now it’s
your turn. Would you rather play Kazoop like Jenny or
Babber like Kate?” Novel words were used in the last
question to ensure children were not influenced by their
own preferences that were unrelated to the scenarios.
Coding was conducted in the same manner as in
Experiment 1. Similar to Experiment 1, the two coders’
ratings were highly correlated, r(86) = .99, p < .001. Coders differed by less than 3 s on 92% of judgments. The
seven cases in which the coding discrepancy was larger
than 3 s were discussed and resolved by the coders.
Analytic approach. As in Experiment 1, the study design
and analytic plan were preregistered with the Open
Science Framework (https://osf.io/7gszx). In our analytic
plan, we specified that we would use Cox regression
models, as in Experiment 1, and linear regression to test
our preregistered hypotheses and that alternatives to
regression would be used if any of the assumptions
underlying the test were violated. Specifically, to test
whether condition influenced children’s wait times, we
planned to use Cox regression to compare models with
Group Influences on Self-Control
Results
As in Experiment 1, all children remembered which
group they were in with 100% accuracy. Children also
identified with their group—pretest: M = 73%, SD =
22%, t(86) = 9.6, p < .001; posttest: M = 66%, SD = 26%,
t(86) = 5.9, p < .001, and remembered both prior to and
following the marshmallow task whether their group
waited or did not wait—pretest: M = 92% (78 of 85; two
children were not asked this question because of experimenter error), SD = 28%, t(84) = 13.92, p < .001; posttest: M = 92% (79 of 86; one child was not asked this
question because of experimenter error), SD = 28%,
t(85) = 14.11, p < .001.
Consistent with Experiment 2’s first preregistered
hypothesis and replicating the findings of Experiment
1, children whose group did not wait for two marshmallows (Mdn = 5.77 min) gave into the temptation to eat
the marshmallow at twice the rate of children whose
group waited (wait time Mdn = 14.95 min), χ2(1) = 5.96,
p = .015, hazard ratio = 2.05, 95% CI = [1.14, 3.70]. These
results held when controlling for age, χ2(1) = 5.16, p =
.023, hazard ratio = 2.00, 95% CI = [1.1, 3.63].
To test whether the influence of group behavior varied
depending on children’s tendency to identify with their
group, we conducted an exploratory test in which ingroup identity, condition, and their interaction were
included in the model. The interaction was significant,
χ2(1) = 4.61, p < .03, hazard ratio = 8.02, 95% CI = [1.2,
53.67]. Among children who tended to identify with their
group (preferring the in-group on > 50% of the in-groupidentity questions, n = 63), the effect of condition was
significant (group-waited condition: Mdn = 15 min, groupdid-not-wait condition: Mdn = 3.3 min), χ2(1) = 11.8, p <
.001, hazard ratio = 3.28, 95% CI = [1.6, 6.7]. Among
children who did not show evidence of identifying with
their group (preferring the in-group on ≤ 50% of the ingroup-identity questions, n = 24), the condition effect was
not significant (group-waited condition: Mdn = 5.15 min,
0.8
Preference for Individuals Who Delay
Gratification
condition (group waited vs. group did not wait) as the
only predictor and compare it with a model without any
predictors. To test whether condition influenced children’s evaluation of delaying gratification, we planned to
use linear regression to compare a model with condition
(group waited vs. group did not wait) as the only predictor to a model without any predictors. To test whether
in-group bias varied depending on condition, we planned
to use multiple linear regression to compare a model
including condition (group waited vs. group did not
wait), time (pre– vs. post–marshmallow test), and their
interaction to a model excluding the interaction term,
with the expectation that the model including the interaction term would be a better fit to the data.
745
0.6
0.4
0.2
0.0
Group-Did-NotWait Condition
Group-Waited
Condition
Fig. 1. Results from Experiment 2: children whose in-group waited
and whose out-group did not were more likely to prefer individuals who delayed gratification. Error bars represent 95% confidence
intervals.
group-did-not-wait condition: Mdn = 15 min), χ2(1) = 0.74,
p > .250, hazard ratio = 0.60, 95% CI = [0.19, 1.91].
Consistent with our second preregistered hypothesis,
children in the group-waited condition were more likely
to prefer new individuals who delayed gratification
(M = 74%, SD = 19%), compared with children in the
group-did-not-wait condition (M = 53%, SD = 22%),
t(85) = 4.66, p < .001 (Fig. 1). An exploratory analysis
indicated that this effect held when controlling for children’s delay performance, t(84) = 4.13, p < .001. Children’s delay performance also independently predicted
their preference for new individuals who delayed gratification, t(84) = 3.21, p = .002.
We did not find evidence supporting our third preregistered hypothesis. Children in both conditions
tended to prefer their in-group—group-waited condition: M = 72%, SD = 26%, t(87) = 7.78, p < .001; groupdid-not-wait condition: M = 67%, SD = 22%, t(85) = 7.4,
p < .001. In-group bias did not significantly vary
by condition, t < 1, p > .250, but it did vary by time,
χ2(1) = 3.98, p = .046, such that children preferred their
group more at pretest (M = 73%, SD = 22%) than at
posttest (M = 66%, SD = 26%). There was no interaction
between condition and time, χ2(1) > 1, p > .250.
Discussion
Experiment 2 replicated and extended the key finding
of Experiment 1 by showing that children who believed
that their in-group delayed gratification and their outgroup did not delay waited longer and subsequently
Doebel, Munakata
746
preferred new individuals who delayed gratification,
compared with children who believed that their ingroup did not delay and their out-group did. Critically,
the new individuals were not presented as group members, so if children were simply trying to behave in
accord with their group’s behavior, they would have
had no reason to later prefer others who delayed gratification. Children led to believe that their group delayed
gratification may think this is a group value and internalize this value.
Identification with the in-group was not affected by
group behavior (i.e., whether or not the group delayed).
Perhaps delaying gratification was not a preexisting
norm, or it was but children’s in-group identity was
robust to negative information about their group. The
latter possibility may be more likely, given that in neither condition did children prefer new individuals who
did not delay over new individuals who did delay. Thus,
children may have adjusted their valuation of delaying
gratification on the basis of group behavior, but never
to the point of preferring individuals who did not delay,
given their preexisting norms.
General Discussion
We provide the first evidence that group membership
and norms influence self-control behavior and evaluations in young children. Young children were more
likely to delay gratification and value it when their
group delayed and another group did not, compared
with the reverse. These findings demonstrate an important way in which self-control behavior does not simply
reflect self-control ability but is also influenced by
social contextual factors (Gardner & Steinberg, 2005;
Kidd et al., 2013; Lee & Carlson, 2015; Michaelson &
Munakata, 2016; Pepper & Nettle, 2017). More broadly,
these findings show that the influence of social groups
on children extends beyond the realm of in-group
biases, social expectations, and moral behavior ( Jordan
et al., 2014; Liberman, Woodward, & Kinzler, 2017) to
the domain of self-control.
This work also supports a new perspective on individual differences in self-control behavior in childhood,
which predict concurrent and later life outcomes. In
addition to variation in relevant cognitive and neural
systems, variation in how much self-control is emphasized—in the family, community, or peer group—may
influence how much individuals use self-control both
in childhood and later in life (see also Lamm et al.,
2017). Norms could influence children to work harder
to exert control, could support their self-control by
allowing them to imagine themselves or their group
using self-control, or could lead them to utilize the
self-control they have. Moreover, opportunities to practice self-control early in life may shape the neural substrates that support it (Diamond & Lee, 2011; Doebel,
Michaelson, & Munakata, 2017; Zelazo, 2015), resulting
in reciprocal, cascading effects (Karmiloff-Smith, 1998;
Smith & Thelen, 2003) that make it easier for children
to exercise self-control and to improve across the life
span.
Our findings have implications for interventions to
improve self-control in those who struggle with it. Supporting values and norms around self-control may be
a fruitful approach to improving self-control in individuals for whom self-control is not a salient or wellestablished norm or value. For example, interventions
to improve self-control could be conducted in a group
format, and self-control could be promoted as a group
value.
Future research can examine such possibilities, in
addition to further investigating the ways in which
group behavior and norms influence self-control. For
example, are children driven to engage in self-control
by the actions of their in-group, those of the outgroup, or both (given that in-group bias and out-group
prejudice are distinct phenomena; Aboud, 2003)? Do
such influences of group norms vary by age (given
that in-group bias may increase with age; Yee &
Brown, 1992) or by culture (given that cultures vary
in how relevant others are to self-construal; e.g.,
Markus & Kitayama, 1991)? Does group behavior exert
its influence via cognitive mechanisms posited to support self-control, for example by strengthening relevant goal representations (Munakata, Snyder, &
Chatham, 2012)? Addressing such questions will
advance our understanding of how self-control develops and is shaped by sociocultural factors to support
adaptive behavior across the life span and will inspire
new approaches to improving it.
Action Editor
Charles Hulme served as action editor for this article.
Author Contributions
S. Doebel conducted all analyses and drafted the manuscript.
S. Doebel and Y. Munakata developed the study concept and
design, provided critical revisions, and approved the final
manuscript for submission.
Acknowledgments
The authors thank Hannah Wickenheiser, Hailey Buckingham, Samantha Stone, Cleo Andersen-Green, Margaret
McCandless, Marina Blum, and Taylor Kotary for assistance
with data collection, and members of the Cognitive Development Center for helpful discussion.
Group Influences on Self-Control
Declaration of Conflicting Interests
The author(s) declared that there were no conflicts of interest
with respect to the authorship or the publication of this
article.
Funding
This research was supported by the Eunice Kennedy Shriver
National Institute of Child Health & Human Development of
the National Institutes of Health under Award Numbers
F32HD079191 and R01 HD078532.
Open Practices
TC
All data and materials have been made publicly available via
the Open Science Framework (Experiment 1: https://osf.io/
fwt65/; Experiment 2: https://osf.io/da56v/). The design and
analysis plans for the experiments were preregistered at the
Open Science Framework (Experiment 1: https://osf.io/wrqtu/;
Experiment 2: https://osf.io/7gszx/). The complete Open
Practices Disclosure for this article can be found at http://journals
.sagepub.com/doi/suppl/10.1177/0956797617747367. This article has received badges for Open Data, Open Materials, and
Preregistration. More information about the Open Practices
badges can be found at http://www.psychologicalscience.org/
publications/badges.
Note
1. It is possible that our between-subjects design was not sensitive enough to detect differences between the control and groupdid-not-wait conditions. A pre-post design in which baseline
delay is compared with posttest delay might be more sensitive
and could be explored in future research, but researchers would
need to address the challenges inherent in having children complete the marshmallow test or a similar paradigm twice.
References
Aboud, F. E. (2003). The formation of in-group favoritism
and out-group prejudice in young children: Are they
distinct attitudes? Developmental Psychology, 39, 48–60.
doi:10.1037/0012-1649.39.1.48
Bandura, A., & Mischel, W. (1965). Modifications of selfimposed delay of reward through exposure to live and
symbolic models. Journal of Personality and Social
Psychology, 2, 698–705. doi:10.1037/h0022655
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015).
Fitting linear mixed-effects models using lme4. Journal
of Statistical Software, 67(1). doi:10.18637/jss.v067.i01
Bernier, A., Carlson, S. M., & Whipple, N. (2010). From external regulation to self-regulation: Early parenting precursors of young children’s executive functioning. Child
Development, 81, 326–339. doi:10.1111/j.1467-8624.2009
.01397.x
Bigler, R. S., Jones, L. C., & Lobliner, D. B. (1997). Social
categorization and the formation of intergroup attitudes
in children. Child Development, 68, 530–543.
747
Billig, M., & Tajfel, H. (1973). Social categorization and similarity in intergroup behaviour. European Journal of Social
Psychology, 3, 27–52.
Bunge, S. A., & Zelazo, P. D. (2006). A brain-based account
of the development of rule use in childhood. Current
Directions in Psychological Science, 15, 118–121.
Carlson, S. M., & Zelazo, P. D. (2011). The value of control
and the influence of values. Proceedings of the National
Academy of Sciences, USA, 108, 16861–16862. doi:10.1073/
pnas.1113235108
Casey, B. J., Somerville, L. H., Gotlib, I. H., Ayduk, O., Franklin,
N. T., Askren, M. K., . . . Glover, G. (2011). Behavioral and
neural correlates of delay of gratification 40 years later.
Proceedings of the National Academy of Sciences, USA,
108, 14998–15003. doi:10.1073/pnas.1108561108
Cox, D. R. (1972). Regression models and life-tables. Journal
of the Royal Statistical Society B: Statistical Methodology,
34, 187–220.
Diamond, A., & Lee, K. (2011). Interventions shown to aid
executive function development in children 4 to 12 years
old. Science, 333, 959–964.
Doebel, S., Michaelson, L. E., & Munakata, Y. (2017). Beyond
personal control: The role of developing self-control
abilities in the behavioral constellation of deprivation. Behavioral and Brain Sciences, 40, Article e324.
doi:10.1017/S0140525X17000942
Duckworth, A. L., Tsukayama, E., & Kirby, T. A. (2013). Is it
really self-control? Examining the predictive power of the
delay of gratification task. Personality and Social Psychology
Bulletin, 39, 843–855. doi:10.1177/0146167213482589
Dunham, Y., Baron, A. S., & Carey, S. (2011). Consequences of
“minimal” group affiliations in children. Child Development,
82, 793–811. doi:10.1111/j.1467-8624.2011.01577.x
Gardner, M., & Steinberg, L. (2005). Peer influence on risk
taking, risk preference, and risky decision making in
adolescence and adulthood: An experimental study.
Developmental Psychology, 41, 625–635.
Hackman, D. A., Gallop, R., Evans, G. W., & Farah, M. J.
(2015). Socioeconomic status and executive function:
Developmental trajectories and mediation. Developmental
Science, 18, 686–702. doi:10.1111/desc.12246
Jordan, J. J., McAuliffe, K., & Warneken, F. (2014). Development
of in-group favoritism in children’s third-party punishment of selfishness. Proceedings of the National Academy
of Sciences, USA, 111, 12710–12715.
Karmiloff-Smith, A. (1998). Development itself is the key
to understanding developmental disorders. Trends in
Cognitive Sciences, 2, 389–398. doi:10.1016/S13646613(98)01230-3
Kidd, C., Palmeri, H., & Aslin, R. N. (2013). Rational snacking: Young children’s decision-making on the marshmallow task is moderated by beliefs about environmental
reliability. Cognition, 126, 109–114. doi:10.1016/j.cognition
.2012.08.004
Kim, E. B., Chen, C., Smetana, J. G., & Greenberger, E. (2016).
Does children’s moral compass waver under social pressure? Using the conformity paradigm to test preschoolers’ moral and social-conventional judgments. Journal of
Experimental Child Psychology, 150, 241–251.
748
Kinzler, K. D., Corriveau, K. H., & Harris, P. L. (2011).
Children’s selective trust in native-accented speakers.
Developmental Science, 14, 106–111.
Lamm, B., Keller, H., Teiser, J., Gudi, H., Yovsi, R. D., Freitag,
C., . . . Lohaus, A. (2017). Waiting for the second treat:
Developing culture-specific modes of self-regulation.
Child Development. Advance online publication. doi:10
.1111/cdev.12847
Lee, W. S., & Carlson, S. M. (2015). Knowing when to be
“rational”: Flexible economic decision making and executive function in preschool children. Child Development,
86, 1434–1448. doi:10.1111/cdev.12401
Liberman, Z., Woodward, A. L., & Kinzler, K. D. (2017).
The origins of social categorization. Trends in Cognitive
Sciences, 21, 556–568.
LoBue, V. (2014). The Child Affective Facial Expression
(CAFE) set. Databrary. doi:10.17910/B7301K
LoBue, V., & Thrasher, C. (2015). The Child Affective Facial
Expression (CAFE) set: Validity and reliability from
untrained adults. Frontiers in Psychology, 5, Article 1532.
doi:10.3389/fpsyg.2014.01532
Markus, H. R., & Kitayama, S. (1991). Culture and the self:
Implications for cognition, emotion, and motivation.
Psychological Review, 98, 224–253.
Michaelson, L. E., de la Vega, A., Chatham, C. H., & Munakata, Y.
(2013). Delaying gratification depends on social trust.
Frontiers in Psychology, 4, Article 355. doi:10.3389/
fpsyg.2013.00355
Michaelson, L. E., & Munakata, Y. (2016). Trust matters: Seeing
how an adult treats another person influences preschoolers’ willingness to delay gratification. Developmental
Science, 19, 1011–1019. doi:10.1111/desc.12388
Mills, M. (2011). Introducing survival and event history analysis. Thousand Oaks, CA: SAGE.
Mischel, W., Shoda, Y., & Rodriguez, M. I. (1989). Delay
of gratification in children. Science, 244, 933–938.
doi:10.1126/science.2658056
Moffitt, T. E., Arseneault, L., Belsky, D., Dickson, N., Hancox,
R. J., Harrington, H., . . . Sears, M. R. (2011). A gradient
Doebel, Munakata
of childhood self-control predicts health, wealth, and
public safety. Proceedings of the National Academy
of Sciences, USA, 108, 2693–2698. doi:10.1073/pnas
.1010076108
Munakata, Y., Snyder, H. R., & Chatham, C. H. (2012).
Developing cognitive control: Three key transitions.
Current Directions in Psychological Science, 21, 71–77.
doi:10.1177/0963721412436807
Myers, D. G., & Lamm, H. (1976). The group polarization
phenomenon. Psychological Bulletin, 83, 602–627.
Pepper, G. V., & Nettle, D. (2017). The behavioural constellation of deprivation: Causes and consequences. Behavioral
and Brain Sciences, 40, Article e314. doi:10.1017/S0140
525X1600234X
Pinker, S. (2011). The better angels of our nature: Why violence
has declined. New York, NY: Viking.
R Core Team. (2008). R: A language and environment for statistical computing. Retrieved from http://www.R-project
.org
Shutts, K., Banaji, M. R., & Spelke, E. S. (2010). Social categories guide young children’s preferences for novel objects.
Developmental Science, 13, 599–610. doi:10.1111/j.14677687.2009.00913.x
Smith, L. B., & Thelen, E. (2003). Development as a dynamic
system. Trends in Cognitive Sciences, 7, 343–348. doi:
10.1016/S1364-6613(03)00156-6
Xiao, N. G., Wu, R., Quinn, P. C., Liu, S., Tummeltshammer, K. S.,
Kirkham, N. Z., . . . Lee, K. (2017). Infants rely more
on gaze cues from own-race than other-race adults for
learning under uncertainty. Child Development. Advance
online publication. doi:10.1111/cdev.12798
Yee, M. D., & Brown, R. (1992). Self-evaluations and intergroup attitudes in children aged three to nine. Child
Development, 63, 619–629. doi:10.1111/j.1467-8624.1992
.tb01650.x
Zelazo, P. D. (2015). Executive function: Reflection, iterative reprocessing, complexity, and the developing brain.
Developmental Review, 38, 55–68. doi:10.1016/j.dr.2015
.07.001
Reading Response for Doebel & Munakata (2018)
20 points total
Only typed Canvas assignments will be accepted.
1. In Experiment 1, the authors hypothesize that group behavior may influence
children’s delay of gratification. Why do the authors think this? What did they do
to test this idea? (Write 1-2 sentences in response to each of these questions, in
your own words). 4 points.
2. What is a preregistered hypothesis? (HINT: you may have to do some outside
work on this!) 2 points.
3. Why might preregistration make scientific findings more trustworthy? 2 points.
4. What were the main results of Experiment 1? How do the authors interpret these
results? 4 points.
5. In Experiment 2, the researchers hypothesized that children wait more when their
group waits because their group’s behavior affects how children value selfcontrol. How did the authors test this and what did they find? 4 points.
6. What do the findings of the experiments suggest about how self-control develops?
How could this be tested? 4 points.
Reading Response for Doebel & Munakata (2018)
20 points total
Only typed Canvas assignments will be accepted.
1. In Experiment 1, the authors hypothesize that group behavior may influence
children’s delay of gratification. Why do the authors think this? What did they do
to test this idea? (Write 1-2 sentences in response to each of these questions, in
your own words). 4 points.
2. What is a preregistered hypothesis? (HINT: you may have to do some outside
work on this!) 2 points.
3. Why might preregistration make scientific findings more trustworthy? 2 points.
4. What were the main results of Experiment 1? How do the authors interpret these
results? 4 points.
5. In Experiment 2, the researchers hypothesized that children wait more when their
group waits because their group’s behavior affects how children value selfcontrol. How did the authors test this and what did they find? 4 points.
6. What do the findings of the experiments suggest about how self-control develops?
How could this be tested? 4 points.
XXX10.1177/0956797618783747
corrigendum2018
Corrigendum
Corrigendum: Group Influences on
Engaging Self-Control: Children Delay
Gratification and Value It More When
Their In-Group Delays and Their
Out-Group Doesn’t
Psychological Science
1
© The Author(s) 2018
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797618783747
https://doi.org/10.1177/0956797618783747
www.psychologicalscience.org/PS
Original article: Doebel, S., & Munakata, Y. (2018). Group influences on engaging self-control: Children delay
gratification and value it more when their in-group delays and their out-group doesn’t. Psychological Science, 29,
738–748. doi:10.1177/0956797617747367
In our original article, we discussed a hazard ratio as
though it were an odds ratio in three places. This Corrigendum corrects these errors. Note that the statistics associated with these sentences will not change, nor do these
changes affect the study’s conclusions whatsoever.
First, the second sentence of the second paragraph
of the Results of Experiment 1 (p. 742) will be changed
to read as follows: “Children whose group did not wait
for two marshmallows (wait time Mdn = 9.48 min) gave
into the temptation to eat the marshmallow at nearly
twice the rate of children whose group waited (Mdn =
15.00 min), χ2(1) = 3.39, p = .07, hazard ratio = 1.95,
95% confidence interval (CI) = [0.94, 4.01].” The second
sentence of the fourth paragraph in the same section
(p. 742) will be changed to read as follows: “Children
who did not receive any information about their group’s
delay behavior (wait time Mdn = 2.95 min) gave into
the temptation to eat the marshmallow at over twice
the rate of children whose in-group waited, χ2(1) =
4.57, p = .033, hazard ratio = 2.18, 95% CI = [1.05, 4.50].”
Finally, the first sentence of the second paragraph
of the Results of Experiment 2 (p. 745) will be changed
to read as follows: “Consistent with Experiment 2’s
first preregistered hypothesis and replicating the findings of Experiment 1, children whose group did not
wait for two marshmallows (Mdn = 5.77 min) gave into
the temptation to eat the marshmallow at twice the rate
of children whose group waited (wait time Mdn = 14.95
min), χ2(1) = 5.96, p = .015, hazard ratio = 2.05, 95%
CI = [1.14, 3.70].”
747367
research-article2018
PSSXXX10.1177/0956797617747367Group Influences on Self-ControlDoebel, Munakata
Research Article
Group Influences on Engaging
Self-Control: Children Delay Gratification
and Value It More When Their In-Group
Delays and Their Out-Group Doesn’t
Psychological Science
2018, Vol. 29(5) 738–748
© The Author(s) 2018
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797617747367
https://doi.org/10.1177/0956797617747367
www.psychologicalscience.org/PS
TC
Sabine Doebel and Yuko Munakata
Department of Psychology and Neuroscience, University of Colorado Boulder
Abstract
Self-control emerges in a rich sociocultural context. Do group norms around self-control influence the degree to which
children use it? We tested this possibility by assigning 3- to 5-year-old children to a group and manipulating their
beliefs about in-group and out-group behavior on the classic marshmallow task. Across two experiments, children
waited longer for two marshmallows when they believed that their in-group waited and their out-group did not,
compared with children who believed that their in-group did not wait and their out-group did. Group behavior
influenced children to wait more, not less, as indicated by comparisons with children in a control condition who were
assigned to a group but received no information about either groups’ delay behavior (Experiment 1). Children also
subsequently valued delaying gratification more if their in-group waited and their out-group did not (Experiment 2).
Childhood self-control behavior and related developmental outcomes may be shaped by group norms around selfcontrol, which may be an optimal target for interventions.
Keywords
self-control, executive function, social influences, cognitive development, open data, open materials, preregistered
Received 5/3/17; Revision accepted 10/30/17
What shapes our ability to exercise self-control—to
inhibit the urge to order chocolate cake when on a diet
or to save money now for a bigger payoff later? People
vary in how successful they are at engaging self-control.
Children are notorious for struggling with it. Childhood
self-control predicts important life outcomes (e.g.,
employment, academic success, health, wealth, criminality; Mischel, Shoda, & Rodriguez, 1989; Moffitt et al.,
2011), suggesting it is an early-emerging and stable individual difference that supports adaptive behavior throughout life. Identifying factors that predict self-control in
childhood and beyond is thus of considerable interest.
Self-control is often explained in terms of earlydeveloping cognitive skills, such as executive functions
(e.g., goal maintenance, working memory, and inhibition), prefrontal circuitry associated with executive
functions (Bunge & Zelazo, 2006; Casey et al., 2011),
and psychological traits (e.g., Duckworth, Tsukayama,
& Kirby, 2013; Moffitt et al., 2011). However, self-control
emerges not in isolation but in a rich sociocultural
context that may influence how it is exercised and
develops. Self-control is linked to parenting (e.g.,
Bernier, Carlson, & Whipple, 2010; Lamm et al., 2017)
and socioeconomic status (SES; e.g., Hackman, Gallop,
Evans, & Farah, 2015) and may be moderated by beliefs
about the reliability of the environment and social trust.
Children and adults wait less for rewards when the
person providing the reward is unreliable (e.g., Kidd,
Palmeri, & Aslin, 2013) or untrustworthy (Michaelson,
de la Vega, Chatham, & Munakata, 2013; Michaelson &
Munakata, 2016). Children also adjust whether they
exercise self-control in response to an adult model’s
Corresponding Author:
Sabine Doebel, Department of Psychology and Neuroscience,
University of Colorado Boulder, 1805 Colorado Ave., Muenzinger 345
UCB, Boulder, CO 80309
E-mail: sabine.doebel@colorado.edu
Group Influences on Self-Control
delay behavior (Bandura & Mischel, 1965) or undesirable outcomes (Lee & Carlson, 2015).
We propose and test another possibility—that selfcontrol is influenced by group membership and norms.
Adults and children can be strongly influenced by the
behavior and norms of others in their attitudes, judgments, and behaviors (Kim, Chen, Smetana, & Greenberger,
2016; Myers & Lamm, 1976). While self-control might
seem like a universally valued behavior, the importance
placed on self-control in modern society is relatively new
(Pinker, 2011). And while self-control is prominent in
both popular press coverage and scientific journal articles
today, cultures, communities, peer groups, and family
members may vary in how much they emphasize selfcontrol and the importance of using it to achieve goals
(Carlson & Zelazo, 2011; Lamm et al., 2017). Thus, variations in group behavior and norms around self-control
might influence whether self-control is used and how it
develops. Cultures differ in delaying gratification (e.g.,
Lamm et al., 2017); however, such patterns could be due
to either cultural norms around self-control or other differences between cultures.
One way to assess influences of group norms is by
manipulating children’s identification with a group who
shows or does not show self-control. Group behavior
motivates children from infancy (e.g., Xiao et al., 2017).
Preschoolers show in-group biases based on factors
such as eye color and accent (Bigler, Jones, & Lobliner,
1997; Kinzler, Corriveau, & Harris, 2011) even when
there is no substantive basis for group membership
(Dunham, Baron, & Carey, 2011), and such biases influence their actions ( Jordan, McAuliffe, & Warneken,
2014). Thus, if group norms play a role in self-control,
then learning that one’s in-group engages in self-control
(and one’s out-group does not) should influence children to increase their self-control behavior.
Gaining insight into the influence of group behavior
on self-control is timely given interest in interventions
to support the development of self-control and the mixed
findings thus far (Diamond & Lee, 2011). Common issues
are the absence of long-term effects and limited improvements that are generally confined to tasks similar to
those used to train control (so-called narrow transfer).
If group behavior influences children’s tendency to use
self-control, then interventions that capitalize on this
(e.g., by working to influence norms and values) may
achieve broader transfer and longer lasting gains.
We examined the influence of group membership on
self-control behavior by testing whether in-group identity influences children’s tendency to delay gratification
on the classic marshmallow task, in which children are
given the option of having a small reward right away
(one marshmallow) or waiting for a larger reward (two
marshmallows; Mischel et al., 1989). In two experiments,
739
we assigned children to an in-group using a minimalgroups induction procedure that creates in-group bias
in young children (Dunham et al., 2011) and manipulated the degree to which self-control was presented as
a group norm. Children were randomly assigned to conditions varying in what children were told about the
behavior of their group and of another group on the
marshmallow task. We predicted that children would
delay gratification longer on the task when their group
delayed while the other group did not, compared with
children whose group did not delay gratification while
the other group did.
Experiment 1
Method
Participants. A total of 98 children 3 to 4 years old (age:
M = 4.22 years, SD = 0.33, range = 3.55–5.00; males = 46)
were recruited from a database of families who had previously indicated interest in participating in child development research. Of these children, 8 were excluded because
they became upset while waiting for two marshmallows
(n = 1), required their parent to stay in the room throughout the procedure (n = 2), left the room (n = 2), or ate the
marshmallow before being exposed to the manipulation
(n = 1), or because of experimenter error that affected the
manipulation (n = 1) or equipment failure that affected the
manipulation (n = 1). We selected our sample size with a
view to the feasibility of collecting the data within an academic semester and to roughly double that of previous
studies (Kidd et al., 2013; Michaelson & Munakata, 2016).
A power analysis informed by previous effect sizes was
not possible because of a lack of precedent for this specific experimental manipulation in the literature. Data
were collected between February and May 2016.
For 97% of our participants, at least one parent had
a 4-year college degree or higher, 2% completed high
school and some college, and 1% completed high
school only. The racial makeup of the sample was 97%
Caucasian, 1% African American, 1% Native Hawaiian,
and 1% American Indian. Ninety-six percent of the
sample was non-Hispanic/non-Latino.
Design. We employed a between-subjects experimental
design, randomly assigning each child to one of three conditions in which (a) in-group members waited for a larger
reward and out-group members did not (group-waited
condition), (b) in-group members did not wait for a larger
reward and out-group members did (group-did-not-wait
condition), and (c) no information was provided about ingroup or out-group members’ delay behavior (control condition). All children then completed the marshmallow test
(Mischel et al., 1989).
740
Procedure. Children were brought into a small room and
seated at a white table. A video camera was set up on a
tripod to record the session, along with a webcam, which
fed into an adjacent room where the child’s parent was
observing. A lidded opaque rectangular container contained
the plated marshmallow and green T-shirt. The room was
otherwise empty and without decoration. The experiment
began with the in-group induction, followed by an introduction to the marshmallow test, the experimental manipulation, and then the marshmallow test proper. Children
were also asked questions at different points in the procedure, which served as manipulation checks.
Following Dunham et al. (2011), we assigned each
child to a group and did not use any procedure to
specifically encourage in-group affiliation. The experimenter knelt beside the seated child and said, “Guess
what? You get to be in the green group today. Alright.
Just like these kids here.” The experimenter placed in
front of the child a laminated 8- × 12-in. card depicting
headshots of four children wearing green shirts presented in a 2 × 2 format with a boy on the left and a
girl on the right in the top row and the reverse pattern
on the bottom row. These pictures have been used and
validated in previous research on young children’s ingroup preferences (Dunham et al., 2011). The children
in the pictures appear to be 4 to 5 years of age. All of
the children were Caucasian.
The experimenter continued, “The green group is
great. Here’s a special green shirt for you to wear
because you’re in the green group.” The experimenter
then helped the child put on the shirt. Next, the experimenter placed on the table (to the right or left of the
in-group, randomized between participants) a picture
of the out-group and said, “Now look here, this is the
orange group. This isn’t your group; it’s a different
group.” The format and gender makeup were identical
across the in- and out-groups. The children also looked
similar across the groups, but to ensure no biases
toward particular children, we randomized across participants which group of children was shown with
green shirts and presented as the in-group.
The experimenter then asked, “So which group are
you in?” If the child hesitated, the experimenter repeated
the question. Once the child responded, the experimenter said, “Yes, you’re in the green group.” If the
child responded incorrectly, the experimenter said,
“Actually, you’re in the green group.” Next, the experimenter presented children with an envelope, opened
it, pulled out a small sticker, and said,
So, I have this present, a sticker, that I want to
give away. I will put it in this envelope and give
it to the kids you tell me to give it to. Who should
I give it to? Can you point or tell me?
Doebel, Munakata
The experimenter recorded children’s responses and
did not provide any feedback. The experimenter then
began the marshmallow test (adapted from Kidd et al.,
2013, and Michaelson & Munakata, 2016). The experimenter placed a single marshmallow on a black plate
centered in front of the child, 4 in. from the edge of
the table, and announced,
Okay, it’s snack time! You have a choice for your
snack. You can eat this one marshmallow right
now or—if you wait for me to go get more
marshmallows from the other room—you can have
two marshmallows to eat instead. How does that
sound?
If the child declared that he or she just wanted one
marshmallow, the experimenter simply repeated the
choice, saying, “Okay, well, you can have this one marshmallow now or you can wait and get two later!” If the
child attempted to grab the marshmallow, the experimenter said, “Oh, let me tell you something else first!”
The experimenter placed a laminated 8- × 12-in.
picture depicting the in-group (in green shirts) and
out-group (in orange shirts) on a small bookstand 4 in.
behind the plated marshmallow. In the group-waited
condition, the experimenter pointed to the images on
the laminated sheet and said,
Look! These kids are in the green group, too, just
like you! And guess what? They waited until they
could have two marshmallows. And these kids are
in the orange group, and they didn’t wait until
they could have two marshmallows.
The left half of the laminated picture depicted a row
of headshots of the four in-group members, below
which were images of marshmallows: On the left was
an image of one marshmallow (representing the immediate reward), and on the right was an image of two
marshmallows (representing the delayed reward). Consistent with the in-group waiting for the larger reward,
an arrow pointed from the in-group to the two marshmallows. The right half of the laminated picture
depicted the four out-group members above the same
marshmallow images but with an arrow pointing to the
single marshmallow instead of the two marshmallows,
consistent with the out-group not waiting for the larger
reward.
In the group-did-not-wait condition, the laminated
picture was identical except that the arrows pointing
to the marshmallows were consistent with the ingroup not waiting for the larger reward and the outgroup waiting for the larger reward. The experimenter
said,
Group Influences on Self-Control
Look! These kids are in the green group, too, just
like you! And guess what? They didn’t wait until
they could have two marshmallows. And these
kids are in the orange group, and they waited
until they could have two marshmallows.
In the control condition, the laminated picture
depicted the in-group and the out-group in the same
manner, but did not depict the marshmallows and thus
provided no information about the groups’ delay behavior. The experimenter said, “Look! These kids are in the
green group, too, just like you! And these kids are in
the orange group.” We opted not to include a statement
about the in-group receiving a choice between waiting
and having two marshmallows because we anticipated
doing so could induce children to speculate about their
group’s behavior, potentially influencing children’s wait
times.
The experimenter then asked the child, “So what did
the green group do?” If the child said he or she did not
know or provided an otherwise uninformative answer,
the experimenter asked, “Did they wait or not wait?” If
the child responded correctly, the experimenter said,
“Yes that’s right they waited until they could have two
marshmallows.” If the child responded incorrectly, the
experimenter said, “Actually, they waited until they
could have two marshmallows.” In the control condition, this question was omitted.
The experimenter then said, “Okay, I’m going to go
get more marshmallows from the other room. I’ll leave
this marshmallow here for you to eat if you want to
while I’m gone, and if you haven’t eaten any of it when
I come back, you can have two marshmallows instead.”
At this point, the experimenter left the room and went
to the adjacent room to watch the child via webcam.
The session was stopped after the child took a bite or
lick of the marshmallow or waited the full 15 min. At
this point, the experimenter returned to the room and
said, “Okay, all done with snack time!” and gave children the second marshmallow if they waited the full
15 min. The experimenter then proceeded with the
posttest questions. The session was discontinued if children became distressed or left the room.
After the test, the experimenter then asked each child
which group he or she was in. This memory check was
identical to the first, except that no feedback was provided. Finally, the experimenter laid out the pictures of
the two groups and asked the child, “Which kids are
nicer?” A verbal or point response was recorded.
A coder who was blind to the experimental condition
and study hypotheses recorded when each child’s first
lick or bite of the marshmallow occurred (as in Kidd
et al., 2013; Michaelson & Munakata, 2016). To ensure
reliability, we checked the coding against that of a
741
second naive coder, r(86) = .98, p < .001. Coders differed by less than 3 s on 94% of judgments. The five
cases in which the coding discrepancy was larger than
3 s were discussed and resolved by the coders. A
research assistant who was blind to the experimental
condition and study hypotheses then transcribed all
self-talk in the videos. A naive coder recorded whether
children made any reference to the in- and out-groups
(green and orange, respectively) while waiting. Coding
was checked against that of a second naive coder, and
there were no discrepancies.
Analytic approach. The study design and analytic plan
were preregistered on the Open Science Framework
(https://osf.io/wrqtu/). Our preregistered hypothesis was
that children whose group waited for two marshmallows
would wait longer for two marshmallows than children
whose group did not wait. All analyses beyond the confirmatory test comparing the group-waited and group-didnot-wait conditions were exploratory. Our preregistered
analytic plan was to conduct a confirmatory test of this
hypothesis using linear regression; however, the dependent variable was heavily right-censored, rendering linear
regression inappropriate as it would produce biased estimates and inflate the Type II error rate (Mills, 2011). Thus
we abandoned the planned linear analysis in favor of the
more appropriate survival analysis using Cox proportional hazards regression models, developed specifically
to handle censored data (Cox, 1972). All analyses were
conducted using the R statistical package (R Core Team,
2006) and the lme4 add-on package to implement mixedeffects models (Bates, Mächler, Bolker, & Walker, 2015).
No data were excluded. One session was terminated
early during the delay period (at 9 min 51 s) because of
experimenter error (the experimenter mistakenly thought
the child had tasted the marshmallow); however, because
the data point was valid (i.e., the only difference from
other censored data points is that censoring occurred
earlier), it was retained, and the delay time was coded as
censored. Because of an equipment malfunction, video
recordings of sessions were not available for 2 participants in the control group. One of these participants
unambiguously waited the entire 15-min period, and the
other participant ate the marshmallow 15 s after the
experimenter left the room, as coded by the experimenter
using the webcam and a timer. Excluding these participants from the sample did not affect the results.
Results
All children remembered which group they were in,
both before and after the marshmallow test. Children
also identified with their group: 80% of children chose
to give a sticker to their group instead of the out-group
Doebel, Munakata
742
prior to learning anything about their group’s behavior,
and, after the marshmallow test, 81% indicated that their
group was nicer. These rates were well above chance,
ts > 7, ps < .001. Almost all children remembered, both
prior to and following the marshmallow test, whether
their group waited or did not wait—pretest: M = 94% (51
of 54; 6 children were not asked this question because
of experimenter error), SD = 23%, t(53) = 14.13, p < .001;
posttest: M = 90% (54 of 60), SD = 30%, t(59) = 10.24,
p < .001.
The results of our confirmatory test were consistent with
our preregistered hypothesis, although they did not meet
the traditional .05 alpha threshold. Children whose group
did not wait for two marshmallows (wait time Mdn =
9.48 min) gave into the temptation to eat the marshmallow at nearly twice the rate of children whose group
waited (Mdn = 15.00 min), χ2(1) = 3.39, p = .07, hazard
ratio = 1.95, 95% confidence interval (CI) = [0.94, 4.01].
We also conducted an exploratory analysis with age
included as a covariate because age tends to correlate
with self-control ability; in this model, the greater likelihood of resisting the marshmallow for children whose
group waited was significant, χ2(1) = 4.19, p = .041,
hazard ratio = 2.13, 95% CI = [1.02, 4.43]. The effect of
age was not statistically significant (p = .1277).
Exploratory comparisons with the control condition
suggested that group behavior influenced children to
wait longer but did not influence children to wait less.
Children who did not receive any information about their
group’s delay behavior (wait time Mdn = 2.95 min) gave
into the temptation to eat the marshmallow at over twice
the rate of children whose in-group waited, χ2(1) = 4.57,
p = .033, hazard ratio = 2.18, 95% CI = [1.05, 4.50]. On
the other hand, there was no significant difference in the
wait times of children whose group members did not
wait for two marshmallows and children who did not
receive information about their group’s delay behavior,
χ2(1) = 0.16, p > .250, hazard ratio = 0.88, 95% CI = [0.46,
1.66]. An exploratory analysis in which age and gender
were added to the model did not change these results,
χ2(1) = 0.16, p > .250, hazard ratio = 0.87, 95% CI = [0.45,
1.7].
This pattern was further confirmed by children’s verbalized reasons for waiting. Among those in the groupwaited condition, 24% of children (7 of 29; 1 additional
child did not answer the question) cited the in- or
out-group’s behavior in their reason for waiting (e.g.,
“Because the green group waited,” “Because I’m in the
green group,” “Orange didn’t wait”). By contrast, 3%
(1 of 29; 1 additional child did not answer the question) in the group-did-not-wait condition cited in- or
out-group behavior in their reason for not waiting,
χ2(1) < 3.62, p = .057. Similarly, among children who
did not receive information about their group’s
behavior, 0% (0 of 29; 1 additional child was not asked
the question because of experimenter error) cited the
groups’ behavior as their reason for waiting or not
waiting, significantly less than in the group-waited condition, χ2(1) = 5.84, p = .016. The number of children
who referred to the groups’ behavior in the group-didnot-wait and control conditions did not differ, χ 2(1) <
1, p > .250.
Children’s self-talk during the delay period was also
consistent with our primary findings. Seventy-three percent of children engaged in some form of self-talk during the delay period, and this did not vary by condition,
p > .250. Children in all conditions talked about the
green group (25%) and talked much less about the
orange group (10%), χ2(1) = 4.65, p = .031, consistent
with in-group identification. The number of children
who talked about the green group in the group-waited
condition (40%) was higher than in the group-did-notwait condition (17%), χ2(1) = 4.02, p = .045, and marginally higher than in the control condition (18%), χ2(1) =
3.43, p = .064. There was no significant difference
between the group-did-not-wait and control conditions,
p > .250.
We also explored the possibility that children
changed how they felt about their group after learning
about their group’s delay behavior, that is, that they
were more biased in favor of their group after learning
that their group did wait, and they were less biased in
favor of their group after learning that their group did
not wait. This would be consistent with children valuing
waiting over not waiting at baseline. A mixed logistic
regression (with random intercepts for participants to
account for dependency among observations) indicated
that the odds of children identifying with their group
did not vary depending on whether they were asked
before or after the marshmallow test, p > .250, whether
they were in the group-waited or group-did-not-wait
condition, p = .105, or the interaction of these factors,
p = .105. Numerically, children in the group-waited
condition showed more of an in-group bias after learning about their group’s behavior (pretest: M = 73%,
posttest: M = 90%), whereas children in the group-didnot-wait condition showed the opposite pattern (pretest: M = 73%, posttest: M = 67%).
Discussion
Experiment 1 provided the first evidence that group
behavior influences self-control in children: Children
tended to wait longer when their in-group delayed
gratification and their out-group did not, compared
with the reverse case. Group norms may have influenced children’s engagement of self-control because
they wanted to do what was normative in their group,
Group Influences on Self-Control
avoid doing what was normative in the out-group, or
both.
Alternatively, children may have been motivated to
behave in accord with their group’s behavior, without
changing their thinking about whether or not delaying
gratification is generally a good thing. In Experiment
2, we tested the possibility that group behavior influenced how children evaluate self-control.
While our group-waited manipulation induced children to wait longer, the group-did-not-wait condition
did not have a comparable negative effect on wait times.
In addition, although not statistically significant, children’s identification with their group numerically
increased after the marshmallow test in the group-waited
condition, and numerically decreased in the group-didnot-wait condition. These patterns suggest that children
may value delaying but cannot always act accordingly.
Learning that the in-group delayed and the out-group
did not may have supported delaying and increased ingroup identification, whereas learning that the in-group
did not delay and the out-group did may have led to
dissonance between the valuing of delay behavior and
in-group identity, resulting in a reduction in in-group
preference and no reduction in delaying. In Experiment
2, we added more questions to better test whether children’s identification with their group changed after learning about their group’s delay behavior.
Finally, in Experiment 2, we used shared preference
as a basis for children’s group membership (as in Billig
& Tajfel, 1973) and to increase group affiliation, which
was expected to strengthen our manipulation by making children more likely to behave in accordance with
their group.
Experiment 2
Experiment 2 tested three hypotheses. If group norms
influence self-control, then children in the group-waited
condition should (a) wait longer than children in the
group-did-not-wait condition (replicating Experiment
1), (b) evaluate new individuals who delay gratification
more positively than children in the group-did-not-wait
condition, and (c) identify more with their group than
children in the group-did-not-wait condition, if delaying
gratification is a preexisting norm.
Method
Participants. A total of 100 children 3 to 5 years old
(age: M = 4.39 years, SD = 0.33, range = 3.58–5.58, males =
40) were recruited to participate in this experiment, as
specified in our preregistered plan, using the same method
as in Experiment 1. Of these children, 13 were not included
in the final sample because they refused to complete the
743
tasks (n = 1), became upset while waiting for the two
marshmallows (n = 4), left the room (n = 2), or ate the
marshmallow before being exposed to the manipulation
(n = 5), or because of experimenter error that affected the
manipulation (n = 1), yielding a final sample of 87. Data
were collected between July 2016 and April 2017. For 95%
of our participants, at least one parent had a 4-year college
degree or higher, 4% completed high school and some college, and 1% completed high school only. The racial
makeup of the sample was 90% Caucasian and 10% mixed
race. The ethnic makeup of the sample was primarily nonHispanic/non-Latino (96%).
Design. This experiment had the same between-subjects
design as Experiment 1, in which we manipulated ingroup and out-group delay behavior. Each child was randomly assigned to either the group-waited condition or
the group-did-not-wait condition. We did not run the control condition in this experiment, as it was not relevant to
our hypotheses.1
Procedure. Experiment 2 followed the same general
procedure as Experiment 1 but with some adjustments to
increase power, control for potential confounds, and test
new hypotheses. We highlight these changes below.
Children were assigned to the in-group in much the
same way as in Experiment 1, but with two changes.
First, we used shared preferences as a basis for children’s
group membership to increase group affiliation (Billig
& Tajfel, 1973). Second, we made adjustments in wording
to ensure that children did not perceive the experimenter
as preferring one group over the other (avoiding the
potential confound of children following their group
because they wanted to please the experimenter).
The experimenter said, “I’m going to show you some
pictures of things and ask you about what you like!
Look at these animals. Can you point to the one you
like more?” The experimenter asked three more questions such as these about foods, toys, and treats and
recorded children’s choices. Next, the experimenter
said,
Okay, now look! This is the green group! The
green group likes the same things you like. They
like , , , and . Now
look here, this is the orange group! The orange
group likes different things than you like. They
like , , , and . So you
are going to be in the green group! Here’s a
special green shirt for you to wear because you’re
in the green group!
If a child refused to make a choice for one of the
questions, the experimenter noted this and omitted it
744
Doebel, Munakata
from the remainder of the procedure. As in Experiment
1, children were asked to indicate which group was
their group. They were also asked to indicate which
group was not their group. Correct responses were
affirmed and incorrect responses were corrected. As in
Experiment 1, children were also asked questions to
assess in-group identity, but in the current experiment,
four such questions were asked in each phase (i.e.,
pre–marshmallow-test and post–marshmallow-test) in
an effort to reduce measurement error and increase
power to detect effects involving this variable. Children
were asked questions such as “Who should I give a
sticker to?” and “Who helped their friend at school?”
The marshmallow test was introduced in the same
way as in Experiment 1. The wording of the experimental manipulation was adjusted to equate for length in
descriptions of each group’s behavior. In the groupwaited condition, the experimenter said,
An evaluation of delay choice (adapted from Shutts,
Banaji, & Spelke, 2010) was added in Experiment 2 to
test the hypothesis that group behavior influenced children via a change to how they evaluated delaying gratification. Children completed four trials in which they
were presented with scenarios involving a pair of children who differed in their delay behavior. Each trial
involved different rewards that were increasingly
abstract from what they had experienced in the experiment (marshmallows, cookies, stickers, and money).
They were shown a page with small pictures of two
children of the same gender, one on the right side of
the page, and one on the left. These images were from
the Child Affective Facial Expression (CAFE) stimuli set
hosted on Databrary (LoBue, 2014; LoBue & Thrasher,
2015). Each depicted child had small pictures of rewards
directly beneath them. A trial was introduced by the
experimenter saying, for example,
Look! These kids are in the green group, just like
you! And guess what? They didn’t have one
marshmallow; they waited until they could have
two marshmallows. And these kids are in the
orange group, not your group! They had one
marshmallow; they didn’t wait until they could
have two marshmallows.
Jenny and Kate love marshmallows! Their mom
said they could have one marshmallow right away,
or if they waited until she found more marshmallows,
they could have two instead. Jenny ate one marsh
mallow right away. Kate waited until she could
have two marshmallows.
In the group-did-not-wait condition, the experimenter said,
Look! These kids are in the green group, just like
you! And guess what? They had one marshmallow;
they didn’t wait until they could have two
marshmallows. And these kids are in the orange
group, not your group! They didn’t have one
marshmallow; they waited until they could have
two marshmallows.
Children were then asked to indicate what each
group did. If they failed to provide an informative
response, the experimenter said, “Did they wait or not
wait?” Correct responses were affirmed, and incorrect
responses were corrected. The marshmallow test was
then administered as in Experiment 1. Once the child
waited the full 15 min or tasted the marshmallow, the
experimenter returned to the room and said, “Snack
time is over! I just have a few more questions for you.”
Children were then asked several of the same questions
they were asked in the pre–marshmallow-test phase:
memory for group, memory for group behavior, and
group identity. The only difference was that no feedback was provided following children’s responses to
any of the questions.
The experimenter then asked three questions
designed to tap children’s implicit preference for one
of the children. For example, children were asked “Who
do you like more?” “Who is nicer?” and “Jenny loves
playing Kazoop. It’s her favorite game to play. Kate loves
playing Babber. It’s her favorite game to play. Now it’s
your turn. Would you rather play Kazoop like Jenny or
Babber like Kate?” Novel words were used in the last
question to ensure children were not influenced by their
own preferences that were unrelated to the scenarios.
Coding was conducted in the same manner as in
Experiment 1. Similar to Experiment 1, the two coders’
ratings were highly correlated, r(86) = .99, p < .001. Coders differed by less than 3 s on 92% of judgments. The
seven cases in which the coding discrepancy was larger
than 3 s were discussed and resolved by the coders.
Analytic approach. As in Experiment 1, the study design
and analytic plan were preregistered with the Open
Science Framework (https://osf.io/7gszx). In our analytic
plan, we specified that we would use Cox regression
models, as in Experiment 1, and linear regression to test
our preregistered hypotheses and that alternatives to
regression would be used if any of the assumptions
underlying the test were violated. Specifically, to test
whether condition influenced children’s wait times, we
planned to use Cox regression to compare models with
Group Influences on Self-Control
Results
As in Experiment 1, all children remembered which
group they were in with 100% accuracy. Children also
identified with their group—pretest: M = 73%, SD =
22%, t(86) = 9.6, p < .001; posttest: M = 66%, SD = 26%,
t(86) = 5.9, p < .001, and remembered both prior to and
following the marshmallow task whether their group
waited or did not wait—pretest: M = 92% (78 of 85; two
children were not asked this question because of experimenter error), SD = 28%, t(84) = 13.92, p < .001; posttest: M = 92% (79 of 86; one child was not asked this
question because of experimenter error), SD = 28%,
t(85) = 14.11, p < .001.
Consistent with Experiment 2’s first preregistered
hypothesis and replicating the findings of Experiment
1, children whose group did not wait for two marshmallows (Mdn = 5.77 min) gave into the temptation to eat
the marshmallow at twice the rate of children whose
group waited (wait time Mdn = 14.95 min), χ2(1) = 5.96,
p = .015, hazard ratio = 2.05, 95% CI = [1.14, 3.70]. These
results held when controlling for age, χ2(1) = 5.16, p =
.023, hazard ratio = 2.00, 95% CI = [1.1, 3.63].
To test whether the influence of group behavior varied
depending on children’s tendency to identify with their
group, we conducted an exploratory test in which ingroup identity, condition, and their interaction were
included in the model. The interaction was significant,
χ2(1) = 4.61, p < .03, hazard ratio = 8.02, 95% CI = [1.2,
53.67]. Among children who tended to identify with their
group (preferring the in-group on > 50% of the in-groupidentity questions, n = 63), the effect of condition was
significant (group-waited condition: Mdn = 15 min, groupdid-not-wait condition: Mdn = 3.3 min), χ2(1) = 11.8, p <
.001, hazard ratio = 3.28, 95% CI = [1.6, 6.7]. Among
children who did not show evidence of identifying with
their group (preferring the in-group on ≤ 50% of the ingroup-identity questions, n = 24), the condition effect was
not significant (group-waited condition: Mdn = 5.15 min,
0.8
Preference for Individuals Who Delay
Gratification
condition (group waited vs. group did not wait) as the
only predictor and compare it with a model without any
predictors. To test whether condition influenced child...
Purchase answer to see full
attachment