Content uploaded by Carolyn Baer
Author content
All content in this area was uploaded by Carolyn Baer on Oct 28, 2021
Content may be subject to copyright.
PREPRINT Not the version of record, currently in press at Child Development
1
Mini Managers: Children strategically divide cognitive labor among col-
laborators, but with a self-serving bias
Carolyn Baer
University of British Columbia
University of California, Berkeley
Darko Odic
University of British Columbia
Strategic collaboration according to the law of comparative advantage involves dividing tasks based
on the relative capabilities of group members. Three experiments (N = 405, primarily White and Asian,
45% female, collected 2016-2019 in Canada) examined how this strategy develops in children when
dividing cognitive labor. Children divided questions about numbers between two partners. By 7 years,
children allocated difficult questions to the skilled partner (Experiment 1, d = 1.42; Experiment 2, d =
0.87). However, younger children demonstrated a self-serving bias, choosing the easiest questions for
themselves. Only when engaging in a third-party collaborative task did 5-year-olds assign harder ques-
tions to the more skilled individual (Experiment 3, d = 0.55). These findings demonstrate early under-
standing of strategic collaboration subject to a self-serving bias.
Keywords: collaboration, cognitive labor, division of labor, skill, difficulty, law of comparative ad-
vantage
Supplemental Materials: https://osf.io/xeb6z/
Collaboration is a foundational feature of human
culture and cognition, illustrated in part by its early
emergence in childhood (Tomasello, 1999; Warneken,
2018). One of the challenging tasks when we engage in
these collaborations is to strategically divide labor to
achieve the team’s goals rather than only our own. For
instance, imagine yourself as a child in math class.
Your teacher has just described today’s activity: you
and a partner must answer 100 math questions with a
range of difficulties as fast as possible. But, while you
are generally good at math, your partner is not. How do
you best divide the questions to maximize the team’s
chance of success?
There are many possible solutions to this challenge.
To maximize accuracy, you could answer all the ques-
tions yourself. This would cost you considerable time
and might hurt the feelings of your partner. To instead
minimize time, you could divide the questions ran-
domly into two equal piles. This might come at the cost
of accuracy when your partner attempts the harder
questions in their pile.
There is also a solution that optimizes both accuracy
and time based on the economic law of comparative ad-
vantage (Ricardo, 1891). You could divide the ques-
tions in half by difficulty, assigning the easier questions
to your partner and taking on the harder questions your-
self. To illustrate, if you can answer 96% of easy ques-
tions correctly and your partner can answer 87%, there
is a 9% collective benefit if you answer the easy ques-
tions. But if you can answer 94% of hard questions cor-
rectly against 72% for your partner, the collective ben-
efit is a much larger 24% if your partner answers the
easy questions. (Estimates based on actual performance
from Baer & Odic (2019) Study 1 for 5- and 7-year-
olds in a numerical comparison task like what we use
in the present studies.) This means that the team gains
9% if you answer easy questions, but misses out on the
much larger 24% benefit if you had answered the
harder ones. Formally, you would have a ‘comparative
advantage’ on the difficult questions because the bene-
fit to the team is comparatively larger when you answer
difficult instead of easy questions.
In everyday life, we employ economists to perform
these computations on the scale of international trade
Carolyn Baer, University of British Columbia and University
of California, Berkeley; Darko Odic, University of British Co-
lumbia.
This work was funded by the Social Sciences and Humanities
Research Council of Canada through an Insight Development
Grant to DO and a Canada Graduate Scholarship to CB. Thanks
to Natasha Au, Bana Ashour, Nicaela Weigel, Nicole Gertz,
Emilie Kniefel, Sally Poon, and Inderpreet Gill for their assis-
tance with data collection, to CB’s writing groups for feedback
on portions of the manuscript, and to all the schools and families
for their support.
Correspondence concerning this paper should be addressed to
Carolyn Baer, Department of Psychology, University of Califor-
nia, Berkeley, 2121 Berkeley Way West, Berkeley, California,
USA, 94720. Email: carolynbaer@berkeley.edu
Baer and Odic 2
and hire managers to make these decisions within or-
ganizations. But these decisions are not only made by
highly-trained people; they also permeate everyday
life. Groupwork must be divided among classmates,
roommates must divide chores, and co-authors must
decide which sections of the manuscript to write. These
far-reaching uses of strategic thinking about skill and
difficulty open a question of whether such strategic
thinking might be fundamental to collaborative reason-
ing in humans. To examine this possibility, we investi-
gate how strategic divisions of labor emerge in child-
hood.
The Development of Strategic Division of Labor
To understand when and how children reason about
division of labor, we break down the formal computa-
tions needed to determine comparative advantage into
several broad steps. We intend these to serve as an ab-
stract set of minimal steps required for dividing labour
rather than an exact model for how we carry them out.
Strategists must estimate the chances of success on the
given tasks for both parties. Those estimates need to be
compared for their relative advantages. Finally, the re-
sulting solution needs to be enacted. Current evidence
suggests that each of these steps should be possible for
even preschool children.
Estimating the Chances of Success of Self and
Other
The literature on metacognitive reasoning demon-
strates that preschool children can sensibly reason
about their chances of success, or what is more com-
monly referred to as confidence (Pouget et al., 2016).
Preschoolers report higher confidence on items they an-
swer correctly than on items answered incorrectly
(Hembacher & Ghetti, 2014; Lyons & Ghetti, 2011),
demonstrating that can differentiate their likely success
from likely failure. This reasoning also influences chil-
dren’s strategies, like decisions to ask for help, opt out
of a task, or seek additional information (Balcomb &
Gerken, 2008; Call & Carpenter, 2001; Coughlin et al.,
2015; Goupil et al., 2016; Hembacher & Ghetti, 2014).
Particularly relevant to our question here are recent
findings that children strategically choose to answer
items with higher chances of success. When given the
option to select from a pair of perceptual quantity com-
parisons (e.g., which set has more dots, or which shape
is bigger), 5-9-year-old children generally chose those
featuring larger, easier ratios that they were more likely
to answer correctly (Baer et al., 2021; Baer & Odic,
2019). Children therefore seem able to estimate their
individual relative chances of success and enact a sim-
ple strategy to maximize that success.
Young children can also reason about others’
chances of success. Three-year-olds recognize and stra-
tegically seek help from individuals who have been
previously accurate or have relevant expertise, infer-
ring that they will continue to be informative in the fu-
ture (Birch et al., 2008; Harris et al., 2018; Koenig &
Harris, 2005; Lutz & Keil, 2002; Mills, 2013; Pasquini
et al., 2007). By 4 years, children tailor their communi-
cation based on the linguistic skill and world
knowledge of others (Baer & Friedman, 2018; Shatz &
Gelman, 1973), using fewer words and talking about
general rather than specific properties to ensure suc-
cessful understanding. Further, preschoolers opt to help
someone faced with a difficult task rather than an easy
one, reflecting an understanding that the chances of
success were lower on the difficult task (Bennett-Pierre
et al., 2018; Bridgers et al., 2020; Ronfard & Harris,
2017). Children therefore demonstrate their under-
standing of others’ chances of success for achieving
their own goals and when assisting others achieve
goals.
Formulating and Enacting a Solution
Collaboration involves working towards a group
goal rather than individual goals. While the optimal
strategy for an individual may be to prioritize easier
work and seek help on difficult work, the law of com-
parative advantage shows that a skilled team member
always pursuing the easiest option can negatively im-
pact the team’s collective chance of success. Collabo-
ration therefore requires thinking about and balancing
both one’s own and another’s chances of success.
Two studies provide initial evidence that children
consider both parties’ capabilities in collaborative set-
tings. Warneken and colleagues (2014) had pairs of 3-
5-year-old children tasked with retrieving a reward us-
ing two complementary tools. After viewing their part-
ner choose one of the tools, 5-year-olds consistently
chose the other tool to ensure that the team had access
to both. Three-year-olds also made this strategic
choice, but only if their partner’s choice was initially
constrained to only one of the tools (Warneken et al.,
2014). This behavior reflects an emerging understand-
ing that success as a group depends on the combined
and complementary capabilities of the team members.
Another study by Magid and colleagues (2018) sug-
gests that children can further apply this collaborative
strategizing to the skill of each partner in dealing with
Baer and Odic 3
physical constraints. In their study, 4-year-olds needed
to win two one-shot carnival games that required
throwing objects at targets either close or far away. The
games needed to be won simultaneously, so children
were given a partner and asked to assign each person
one game to play. If their partner was older than them
(with presumably better chances of success), children
assigned the harder game to the partner and the easier
game to themselves. In contrast, children assigned the
harder game to themselves if their partner was younger
(Magid et al., 2018). This strategy, consistent with the
law of comparative advantage, suggests that even pre-
school children can consider the relative chances of
each partner successfully overcoming the physical con-
straints of the tasks.
Strategic Division of Cognitive Labor
When these findings are put together, there is good
reason to believe that children as young as 4 years old
can follow the law of comparative advantage when
making strategic collaborative decisions. In this work,
we test whether this strategy extends to a non-physical
domain, as would be expected if it was of prime im-
portance for human collaboration.
Human collaboration spans many different domains
from physical (e.g., lifting a couch together) to emo-
tional labor (e.g., caring for elders). In particular, many
of the everyday tasks that require dividing labor are not
subject to physical constraints like the distance of tar-
gets, but by cognitive constraints like knowledge,
memory, or mental effort. In our math class example,
some group members may not have the knowledge re-
quired to answer harder questions at all – a constraint
on their performance that is determined by the effort
and ability required entirely of the ¬mind, not the body.
With the evidence that children can estimate their own
and others’ success on similar knowledge-based tasks
(e.g., Baer & Friedman, 2018; Baer & Odic, 2019), it
seems highly likely that children would similarly use
the law of comparative advantage when strategizing
about tasks from this cognitive domain. Given this pos-
sibility, along with the importance of cognitive labor
for engaging in modern society, we therefore focus on
collaboration in cognitive labor here.
At the same time, strategically dividing cognitive la-
bor may be more difficult than dividing physical labor
for young children because of potential differences at
each of the three theorized steps outlined above. First,
children could have more difficulty reasoning about an-
other’s chances of success at tasks requiring cognitive
skill than physical skill (Niebaum & Munakata, 2020).
Cognitive skills like knowledge or mental capacity
have few, if any, concrete correlates available prior to
observing their success or failure. In contrast, many
physical skills can be inferred from appearances only.
For example, we don’t expect infants to be capable of
running, but we do expect people with large muscles to
be able to lift heavy weights. It could therefore be that
children can learn about physical skills more readily
because there are concrete correlates in the world,
while cognitive skill understanding may lag behind
given its more abstract nature. If so, children may need
more time and experience to learn about these abstract
relations, or because they conceptualize cognitive skill
as linked to a concrete property (like age or wearing
glasses). The current literature has little to say about
this, as there is currently a lack of evidence about
whether children reason differently about cognitive and
physical skills.
A second possibility is that the law of comparative
advantage strategy must be re-learned for each applica-
ble domain. Children do not always apply strategies
learned in one domain to others (Bellon et al., 2020;
Geurten et al., 2018; Vo et al., 2014). For instance, chil-
dren who strategically placed high bets on easier num-
ber discriminations did not necessarily bet high on eas-
ier emotion discriminations (Vo et al., 2014; see also
Bellon et al., 2020; Geurten et al., 2018). Therefore,
even though Magid and colleagues (2018) found per-
formance consistent with the law of comparative ad-
vantage in young children within the physical domain,
children might need to re-learn this strategy separately
for other domains.
A final possibility considered here is that children
may be subject to stronger task biases that overshadow
the appropriate strategy when executing their strategy
to divide cognitive labor. One such bias is well-docu-
mented in the literature on children’s understanding of
fairness (Blake, 2018; Blake et al., 2014). Children up
to 6-8 years show a self-serving bias, prioritizing their
own self-interest over others’ by giving themselves
more stickers than a partner (Blake et al., 2015; Blake
& McAuliffe, 2011; Sheskin et al., 2016). At the same
time, children this age and even preschoolers will iden-
tify these biased distributions as unfair when made be-
tween two other parties (Chernyak & Sobel, 2015; Ro-
chat et al., 2009; Sheskin et al., 2016). Children thus
possess knowledge of the ‘correct’ option but it is over-
shadowed by other influences on their behavior, like a
desire to give themselves an advantage (Blake, 2018).
A similar phenomenon may exist here, where children
can formulate a strategy according to the law of com-
parative advantage, but their behavior is dominated by
Baer and Odic 4
a self-serving bias. This bias could differentially affect
the division of cognitive labor if intellectual compe-
tence is more highly prized by children than physical
competence. There is again little evidence in the current
literature to speak to this (though see Asaba & Gweon,
2019; Zhao et al., 2018 for evidence that children are
motivated to appear competent).
The Current Studies
In our studies, we ask whether children apply the
law of comparative advantage to their strategic division
of cognitive labor. We presented children with a task
relying on cognitive skill: intuitive number compari-
sons (“which one has more dots”; Halberda & Feigen-
son, 2008). Children divided the questions between
themselves and a peer described as being ‘better’ or
‘not as good’ at the task. Children strategically assigned
themselves the easier questions when their partner was
‘better’ and likely more capable of succeeding on the
difficult items. However, to anticipate the findings of
our first experiment, we did not find that children made
strategic divisions of cognitive labor until much older,
at age 9.
In two follow-up studies, we investigate two of the
three possible difficulties as explanations for why chil-
dren did not appear to follow the law of comparative
advantage when dividing cognitive labor: difficulty es-
timating the chances of success and difficulty in enact-
ing a solution. If children fail to use this strategy across
all studies, even with modifications to account for these
two difficulties, we would instead have initial evidence
that children struggle to formulate the strategy when
reasoning about cognitive labor.
Experiment 1
Method
Children were made to believe that they would play
a collaborative “Number Game” (described below)
with another child, who was either better or worse at
that game than the child. We then offered children the
opportunity to divide questions from the Number Game
between the two partners to maximize their chances of
success. Of critical interest was whether children would
assign the easier of the two questions to the relatively
less skilled group member, consistent with the law of
comparative advantage. That is, would children assign
easier questions to themselves if they were less skilled,
but assign those same easy questions to their partner if
the partner was less skilled?
Participants
One-hundred and fifty-nine children participated in
the study (81 girls) between March 2016 and March
2017. Children were between the ages of 6 and 10 years
old (M = 7;11 [years; months], range = 6;0 - 9;11) as
this age group is known to attend to and strategically
compare cognitive difficulty in the number discrimina-
tion task we used (Baer et al., 2018; Baer & Odic,
2019). Our target sample size, set prior to testing based
on the recommendations of Simmons et al., (2011)
though not formally preregistered, was 20 children in
each condition at each age (160 total). Most children
participated in a quiet area of their school and some
participated in an on-campus lab, all in Vancouver,
British Columbia, Canada. We did not formally collect
demographic information, but children largely matched
the demographics of the region (predominantly middle-
class, and White or East or South-East Asian), and all
children spoke English. Two additional children were
excluded for having known or suspected developmen-
tal disabilities (reported by the child’s classroom
teacher). Informed consent was given by children’s
parents prior to beginning the study.
Materials and Procedures
Tasks were presented on a laptop computer using
Psychtoolbox-3 in MATLAB (Brainard, 1997; The
MathWorks Inc., 2015). Children responded verbally
or by pointing to their answer on the screen. The exper-
imenter entered all responses to reduce the influence of
memory and motor development on the results.
Number Game. In each trial, children selected
which of two groups of dots is more numerous without
counting (e.g., 10 yellow dots is more than 5 blue dots,
see Figure 1). As mentioned earlier, smaller ratios, like
8 yellow to 9 blue (a ratio of 1.13), result in longer de-
cision times and lower accuracy than large ratios, like
15 yellow to 5 blue (ratio 3.0; Baer & Odic, 2019; Hal-
berda & Feigenson, 2008). These reaction time and ac-
curacy differences signal that smaller ratios are more
difficult and require more cognitive skill than larger ra-
tios. By 5 years of age, children are sensitive to these
differences in difficulty and can strategically identify
the easier of two ratios (Baer & Odic, 2019, 2020).
Therefore, we manipulated the difficulty of each trial
through the ratio of dots, from a hard 1.13 to an easy
3.0.
Baer and Odic 5
Figure 1. Stimuli Used in the Experiments. (a) shows sample
trials from the Number Game at different ratios. Children in-
dicate which color has more dots. Larger ratios (on the left)
are easier and smaller ratios (on the right) are more difficult.
(b) shows a sample trial from the collaborative task. Children
assign each question to a group member. In this example, the
easier question is on the left.
Children were first introduced to the Number Game
through 10 warm-up trials. On each trial, the dots ap-
peared on the screen for 1.2 sec, and the child indicated
whether the yellow or blue set was more numerous. The
computer played audio feedback about the accuracy of
each trial (e.g., “Oh, that’s not right.” or “Great!”). Be-
cause our collaborative task (detailed below) some-
times required that children believe they were less
skilled than their partner, two of the ten warm-up trials
were impossible (e.g., 10 yellow and 10 blue dots) and
always led to negative feedback. These impossible tri-
als served to signal to children that they were not per-
fectly skilled at the game, making it plausible that an-
other child could be more skilled than them. The re-
maining 8 trials were relatively easy ratios (2.0 and 3.0)
to get children comfortable with the task.
Collaborative Task. Children were given a simple
collaborative goal: they and a partner both needed to
answer a lot of Number Game questions correctly to
win the game. Each partner would answer half of the
total questions independently, and the two independent
scores would be combined to form a team score. The
exact number of correct answers needed was not pro-
vided as we did not want children to begin counting the
number of trials, or feel too anxious about getting an
answer wrong. If children asked for a specific number
(which happened very rarely), the experimenter told
them that she didn’t know the exact number, but em-
phasized that it was ‘a lot’ to make clear that both the
child and their partner would need to do well in order
to win. Teams consisted of two children: the child par-
ticipant and a virtual gender-matched ‘partner,’ intro-
duced through a photo of a child (approximately 7 years
old and White) on the computer screen. Partners were
said to be in another room or at another school, but con-
nected virtually. In the rare event a child expressed
doubt about the partner’s existence, the experimenter
explained that the computers were connected (through
FaceTime or Skype), and pretended to message ‘the
other experimenter’ through an iPad.
We then provided children with information about
the relative skill levels of each partner. Half of children
were told that their partner answered more of the warm-
up questions correctly and was therefore better at the
Number Game than the child (“Partner Better” condi-
tion, exact numbers were not provided). The other half
heard that the partner didn’t answer as many questions
correctly and was therefore not as good at the Number
Game (“Partner Worse” condition, randomly as-
signed). The experimenter also provided affective cues
to differentiate the conditions (e.g., excited expression
and tone if the partner did better, as though children
were lucky to have a skilled partner, or a worried ex-
pression and tone if the partner did worse). The exper-
imenter asked children to indicate which person was
better at the game and corrected them before moving
on if they answered incorrectly or did not want to an-
swer.
As the focal task, we asked children to split pairs of
Number Game questions between the partners. Our pri-
mary dependent variable was how often children would
assign the easier question to themselves, expecting this
to occur more frequently in the Partner Better condition
than in the Partner Worse condition if the children are
following the law of comparative advantage (e.g.,
Magid et al., 2018). In 14 test trials, children saw a pair
of Number Game questions that differed in difficulty
and were asked to choose one for their partner to an-
swer and one for themselves to answer (see Figure 1
and see Baer & Odic, 2019 for a similar design in a non-
social task). If children only selected one question, the
experimenter asked children to select which person
would answer that question, and then indicated that the
remaining question would go to the other partner, to
make sure that children understood that they needed to
assign both questions. Number Game questions varied
in ratio from 1.13 (hard) to 3.0 (easy) and were paired
Baer and Odic 6
to make test trials with ‘metaratios’ from 1.33 (small
difference in difficulty, e.g., ratios 1.5 and 1.13) to 2.65
(large difference in difficulty; see Baer & Odic, 2019).
Children saw two trials at the easiest metaratio first,
then cycled through the full range of metaratios in a
random order that was the same for all children. We
expected children’s performance to be best on the larg-
est metaratios, but additional metaratios were included
to match the stimuli from Baer & Odic, 2019 and to
control for non-numeric cues on these trials (e.g., the
cumulative area of the dots). We made no further pre-
dictions about this variable. The task was untimed, but
children were prevented from counting the dots. Ques-
tions were never labelled as ‘easy’ or ‘hard;’ children
had to infer these difficulties on their own. No feedback
was provided to children about their choices at any
point during the task. While splitting the trials, children
were not asked to answer the questions, and therefore
did not get any feedback about their performance on
those trials.
Following the division of questions, children an-
swered 14 Number Game questions. All children an-
swered the same easy 14 questions, but were told that
they were answering the questions they chose earlier.
At the end of the session, children were told that their
team had answered enough questions correctly to win
and were given a small prize.
Results
All analyses below collapse across gender, as there
was no impact of gender on these results.
First, we confirmed that children understood the
Number Game using the 14 questions presented follow-
ing the division of labor. As shown in Table 1, children
at all ages clearly understood this task, as they selected
the more numerous set well above chance of 50%.
Given that children understood the Number Game,
we next examined their division of labor to see whether
they strategically allocated questions according to the
law of comparative advantage. We conducted a 2 (Con-
dition, between subjects: Partner Better, Partner
Worse) by 4 (Age Group, between subjects: 6, 7, 8, 9)
by 6 (Metaratio, within subjects: 1.33, 1.5, 1.77, 2.0,
2.26, 2.65) ANOVA on children’s choice to assign the
easier question to themselves, averaged across the tri-
als. The analysis of Condition was meant to confirm
prior work (Magid et al., 2018), while Age Group and
Metaratio were exploratory variables.
Table 1
Means and Tests Against Chance in Experiment 1
Age
(Years)
Mean
(%)
SD
df
t
p
d
Number Game (chose more numerous)
6
85.26
12.60
38
17.47
< .001
2.80
7
86.67
11.75
39
19.73
< .001
3.12
8
86.99
10.38
40
22.81
< .001
3.56
9
89.32
9.36
38
26.24
< .001
4.20
Division (gave easier to self)
Partner Better Condition
6
61.56
16.21
20
3.27
.004
0.71
7
62.50
20.71
19
2.70
.014
0.60
8
61.07
24.90
19
1.99
.061
0.44
9
73.02
18.26
17
2.64
.017
1.26
Partner Worse Condition
6
60.71
17.19
17
2.64
.017
0.62
7
57.86
26.00
19
1.35
.192
0.30
8
45.92
26.78
20
-0.70
.493
0.15
9
42.52
23.20
20
-1.48
.155
0.32
Overall, children’s strategies were consistent with
the law of comparative advantage: they were more
likely to take the easier question when they were less
skilled than their partner than when they were better,
F(1, 151) = 10.65, p = .001, ηp2 = .07. However, this
effect was modulated by age, F(3, 151) = 3.39, p = .020,
ηp2 = .06. As shown in Figure 2, a Tukey posthoc anal-
ysis revealed that only 9-year-olds matched skill and
difficulty, keeping many more difficult trials for them-
selves in the Partner Worse condition than in the Part-
ner Better condition, t(37) = 30.50, p < .001, d = 1.46.
The conditions did not differ from one another in any
other age group, though, as can be seen in Table 1 and
Figure 2, 8-year-olds showed the same differentiation
as the 9-year-olds but did not reach traditional levels of
significance.
We also found a main effect of Metaratio (the ratio
of difficulties being compared on each division trial),
F(4.4, 663.75) = 4.65, p = .001, ηp2 = .03 (Greenhouse-
Geisser corrected for sphericity). Similar to when chil-
dren compare difficulties to make strategic choices
only for themselves, children were more likely to
choose the easier question when the difference in diffi-
culty was large than when is was small (Baer et al.,
2018; Baer & Odic, 2019). There were no other signif-
icant main effects or interactions. We also repeated
these analyses in the Supplemental Materials treating
age as a continuous variable rather than categorial, and
Baer and Odic 7
Figure 2. Percentage of Trials Assigning the Easier Ques-
tion to Self in Experiment 1. Error bars represent 1 standard
error.
looking only at performance on the first trial. Both
analyses replicate the patterns reported here.
Finally, children generally allocated the easier ques-
tion to themselves across both conditions. As shown in
Table 1 and Figure 2, children at all ages took the easier
question above chance of 50% when their partner was
better, as expected given a strategy of matching skill
with difficulty. However, when their partner was
worse, 6-year-olds still took the easier question above
chance – opposite the matching strategy – and 7-9-
year-olds’ selections did not differ from chance.
Discussion
Only 9-year-olds in the current study strategically
divided cognitive labor according to the law of compar-
ative advantage. This is surprising given evidence that
4-year-olds enacted this strategy for a physical task
(Magid et al., 2018) and that 5-year-olds can reason
about the relative difficulties of questions in the identi-
cal Number Game (Baer & Odic, 2019). Moreover,
children in all age groups appeared generally biased to
allocate the easier question to themselves, even when
their partner was worse at the game, suggesting that
they did recognize their own chances of success were
higher on the easier questions than the harder ones.
If children could detect the differences in difficul-
ties, then perhaps the challenge lies in their understand-
ing of the skill manipulation. In a follow-up experiment
in the Supplemental Material, we rule out other low-
level explanations for not understanding the skill ma-
nipulation through a post-test comprehension check.
We found that 91% of children passed the check yet
still took easier questions for themselves. Therefore,
properties of the design like the small number of warm-
up trials from which to form ability beliefs, the reliance
on verbal testimony about partner performance, or for-
getting over the course of the study are unlikely to ac-
count for the results.
Our initial intuition was that the term “better”
should have directly cued skill without requiring any
inference, and was therefore the best choice of skill ma-
nipulation. However, 4-year-olds succeeded at dividing
physical labor in the studies by Magid et al., (2018)
when cued to skill using the relative age of the partner
(“younger” or ”older”), which we learned about after
completing this experiment. This highlights an intri-
guing possibility: perhaps as children learn about oth-
ers around them, they may find it easier to link success
with concrete, observable traits than these abstract
ones. Therefore, children may have encoded the term
“better”, but not fully understood its implications for
the underlying cognitive skills. In our next study, there-
fore, we used age as a more easily observable trait that
correlates strongly with many cognitive skills like those
required here (Halberda & Feigenson, 2008; and see
Magid et al., 2018). We told children that their partner
was either older or younger than they were, accompa-
nied by a picture of an older or younger child, respec-
tively. We hypothesized that age alone might be a more
salient and relevant cue to cognitive ability for children
than direct testimony about skill, and that children
might use this information to strategically divide cog-
nitive labor younger than age 9.
Experiment 2
Method
Participants. One hundred and sixty-six children
participated in the study (74 girls) between February
and July 2019 in the same manner and geographic
location as Experiment 1. Our target sample size, as in
Experiment 1, was 20 children per condition per age
group (160 total). We recruited children between the
ages of 4 to 8 years (M = 5;11, range = 4;0 - 7;11)
because our goal was to detect success in children
younger than age 9 and because the manipulation of age
was successfully used by Magid and colleagues (2018)
in 4-year-old children. Four additional children were
excluded from the analyses for not competing the study
(2), not understanding English (1), and because they
participated in Experiment 1 (1).
Materials and Procedures. We used the same de-
sign as Experiment 1 with a few key changes. First, we
did not tell children anything about the relative skill or
Baer and Odic 8
past success of their partner, but instead told children
that their partner was older or younger than them. To
help reinforce this, the picture of their virtual partner
was either about 2-3 years old if younger, or around 7-
8 years old if older, ensuring that every age tested
would believe the manipulation. Second, we provided
a longer warm-up phase, including 30 trials ranging in
ratio from 1.13 to 3.0. Like in Experiment 1, we in-
cluded 6 impossible trials so that children did not get
all questions correct. Third, we used more distinct me-
taratios (2.0, 2.88, and 3.66) during the collaboration
phase of the study to make sure that children could con-
sistently tell apart the two difficulties. Fourth, we asked
children two post-test comprehension questions to con-
firm that they remembered the key age manipulation
(“Who is older?”) and to see if they explicitly mapped
age on to skill (“Who is better at the [Number]
Game?”).
Results
Children’s gender is collapsed in these analyses, as
its inclusion did not influence any of the results below.
Gender interacted with metaratio, F(1.84, 275.98) =
3.77, p = .027, ηp2 = 0.03 (Greenhouse-Geisser cor-
rected), but as neither are key variables of interest and
do not interact with the condition manipulation, we did
not look at this further.
First, we examined children’s post-test comprehen-
sion answers to confirm that the skill manipulation
worked. When asked which partner was older, 97.6%
of children gave the correct answer. When instead
asked which partner was better at the Number Game,
109 children (65.7%) responded with the older partner.
Interestingly, an exploratory logistic regression re-
vealed that this was influenced by condition (Odds Ra-
tio (OR) = 4.73, χ2(1) = 17.40, p < .001), age (OR =
2.19, χ2(1) = 10.10, p = .001), and their interaction (OR
= 6.6, χ2(1) = 6.60, p = .010). Simple slopes revealed
that there was no age effect for linking age to skill if
their partner was younger (81.7% of children linked the
two), β = -0.20, SE = 0.29, t(162) = -0.68, p = .494, but
a strong age effect if their partner was older, β = 0.79,
SE = 0.25, t(162) = 3.17, p = .002 (from 36.3% at age
4 to 85.0% at age 7). We make two conclusions from
this. First, children remembered the age manipulation
and generally linked it to skill. Second, children, par-
ticularly younger children, were reluctant to label
themselves as ‘worse’ at the task. No results below
change when removing children who failed the com-
prehension check.
We next checked that children understood the Num-
ber Game. As shown in Table 2, children selected the
more numerous color more often than 50% (chance) at
all ages tested.
Table 2
Means and Tests Against Chance in Experiment 2
Age
Mean
SD
df
t
p
d
Number Game (chose more numerous)
4
71.87
12.98
40
10.79
< .001
1.69
5
80.38
8.04
43
25.05
< .001
3.78
6
84.39
11.72
43
19.47
< .001
2.93
7
89.28
6.34
36
37.68
< .001
6.19
Division (gave easier to self)
Partner Older Condition
4
45.13
14.69
21
-1.56
.135
0.33
5
49.03
16.83
21
-0.27
.789
0.06
6
56.07
21.53
19
1.26
.222
0.28
7
65.00
24.84
19
2.70
.014
0.60
Partner Younger Condition
4
46.24
6.89
18
-2.38
.029
0.55
5
39.94
17.45
21
-2.71
.013
0.58
6
51.19
25.33
23
0.23
.820
0.05
7
44.96
20.94
16
-0.99
.336
0.24
Did children strategically assign the difficult ques-
tions to the older partner? We conducted an ANOVA
with Condition (Partner Older, Partner Younger), Age
(4, 5, 6, 7), and Metaratio (2.0, 2.8, 3.66) on children’s
choice to assign themselves the easier question. Chil-
dren were generally more likely to take the easy ques-
tion when their partner was older than when their part-
ner was younger, F(1, 158) = 6.42, p = .012, ηp2 = .04.
Children’s age also affected their behavior: older chil-
dren in the sample were also more likely to take the
easy question than younger children, F(3, 158) = 2.92,
p = .036, ηp2 = .05. While not reaching conventional
levels of significance, the analyses and Figure 3 hint at
an interaction between the Condition and Age, F(3,
158) = 2.14, p = .097, ηp2 = .04. Post-hoc Tukey tests
revealed that only 7-year-olds changed their division of
labor strategy based on the relative age of their partner,
t(35) = -20.04, p = .044, d = 0.87, shown in Figure 3.
This appears to be a much clearer difference between
conditions compared to 7-year-olds in Experiment 1,
but also a more tenuous success than that of older chil-
dren given the non-significant interaction term here,
and so should be interpreted with caution. There was
no main effect nor any interactions with Metaratio.
Baer and Odic 9
These results also emerge when treating age as a con-
tinuous variable and when only examining the first trial
(see Supplemental Materials).
As shown in Table 2, children were not biased to
assign themselves the easier question, and instead gen-
erally divided question difficulty at chance levels.
Figure 3. Percentage of Trials Assigning the Easier Question
to Self in Experiment 2. Error bars represent 1 standard error.
Discussion
In contrast to Experiment 1, we have preliminary
evidence that 7-year-olds strategically assigned easier
questions to the relatively less skilled (i.e., younger)
partner when age was used as a proxy for skill. As an
easily observable trait in childhood, and one that often
covaries with cognitive ability given the dramatic cog-
nitive development in this age range, children may
more readily link age with cognitive skill than the com-
parative testimony used in Experiment 1. These results
suggest that at least some of the apparent failure to stra-
tegically divide cognitive labor in 5-8-year-old children
might stem from the labelling of skill, especially for the
youngest children.
The results of Experiment 2 also highlight another,
non-mutually exclusive possibility: children in Experi-
ment 1 may have had a self-serving bias, choosing to
act in their own self-interest rather than in the best in-
terests of the team. Recall that in Experiment 1, chil-
dren on average assigned themselves the easier ques-
tion in both conditions. This is the same behavior seen
when children try to maximize their own success,
shown when children divided labor for a competition
(Magid et al., 2018), and in an asocial setting (Baer et
al., 2018; Baer & Odic, 2019). In Experiment 2, how-
ever, children did not show this general preference for
easy trials, instead assigning themselves the easier
question about half of the time, independent of their
partner’s skill or age. We suspect that the key reason
for this change is in the overt skill comparison used in
Experiment 1 (“Sam was even better than you”), which
tends to lead to reward-maximizing behavior in school-
aged children. For instance, preschool children who
were told they had a reputation for being smart were
more likely to cheat (Zhao et al., 2018), and children
who were outperformed by peers persisted longer than
when peers performed worse (Magid & Schulz, 2015).
In contrast, children in Experiment 2 only learned about
the relative age of the partner, which potentially spared
them from these competitive feelings. In fact, we only
observed similar self-preservation behavior when we
asked children to compare which partner was better as
part of our comprehension check at the end of the study.
Children who were younger than their partner were
more likely to respond that they were better or simply
refuse to answer. Even if children knew they were
younger, they seem to have felt uncomfortable saying
they were worse at the game. If true, children’s self-
serving biases might mask their underlying understand-
ing of strategic division of cognitive labor.
A similar kind of self-serving bias can be seen in the
literature on children’s fairness (e.g., Blake et al.,
2015), where children are tasked with dividing re-
sources like stickers or candies between partners. When
presented with a split that gives them an advantage
(e.g., they get 4 candies when their partner gets 1), chil-
dren up to middle childhood (around age 8) generally
accept the split (Blake et al., 2015; Blake & McAuliffe,
2011; Sheskin et al., 2016). When instead presented
with a split that unfairly advantages their partner (e.g.,
gives the partner 4 candies when they only get 1), chil-
dren generally reject the split, leaving both partners
with no resources. However, even preschool children
say that it is only fair when the two partners receive
equal resources, resulting in a gap between their
knowledge of fair distributions and their actual distri-
bution behavior (Blake, 2018; Blake et al., 2014). To
combat this discrepancy and capture children’s under-
standing of fairness using a behavioral measure, re-
searchers rely on a third-party task, where children di-
vide resources between two other parties rather than be-
tween themselves and a partner. When dividing re-
sources between others, even preschool children divide
resources equally (Chernyak & Sobel, 2015; Rochat et
al., 2009; Sheskin et al., 2016).
To understand the contribution of a self-serving bias
in children’s collaborative behavior, we used a third-
party design in the next study. Rather than asking chil-
dren to divide questions between themself and a part-
ner, we asked children to divide questions between two
other children that differ in skill (but are of same age).
Baer and Odic 10
In doing so, we eliminate any potential self-serving bi-
ases, allowing us to directly test whether children
younger than 7 will strategically divide cognitive labor
according to the law of comparative advantage.
Experiment 3
Method
Participants. Eighty children between the ages of 4
to 8 years (M = 6;0, range = 4;0 - 7;11, 29 girls) partic-
ipated in the study between November 2016 and March
2017 in the same manner and geographic location as
Experiments 1 and 2. This study was planned and con-
ducted prior to learning about the findings of Magid et
al., (2018), and prior to Experiment 2. We present it
third for ease of interpretation given the stronger meth-
odological overlap between Experiments 1 and 2. Our
target sample size was again 20 children per condition
per age group (which given the within-subject design
meant 80 children). Two additional children were ex-
cluded from the analyses for not competing the study.
Materials and Procedures. As in Experiment 2,
children began the study with 30 trials of the Number
Game to orient them to the task (the impossible trials
were not needed to convince children that they weren’t
perfect, and so were not used here). Then, children were
asked to divide the same 14 pairs of questions between
two partners as in Experiment 2.
Rather than asking children to divide questions be-
tween themselves and a virtual partner, children di-
vided questions between two other children who were
playing the Number Game together on a team later that
day. The two other children were gender-matched to
the participant and presented using small laminated
photos. Photos were always placed one above the other
rather than beside each other to avoid children associ-
ating the partners with sides of the screen.
To introduce the collaborative task, the experi-
menter said that both partners needed to get all the
questions correct to win. One partner was said to be
‘good’ at the game, while the other partner was said to
be ‘not so good’ at the game. Children were asked a
comprehension question about whether each partner
was good at the game or not and were corrected if nec-
essary. We asked children to help the partners win the
game by splitting up the questions so they could both
get them all correct. As in Experiments 1 and 2, chil-
dren saw pairs of Number Game questions on the
screen and selected one question to give to each part-
ner. Following the 14 test trials, the experimenter re-
peated the comprehension questions to check if chil-
dren remembered the skill difference.
Results
There was once again no impact of children’s gen-
der or metaratio on the results, so these variables were
removed from further analyses. Twelve children failed
the post-test comprehension check (i.e., incorrectly
identified the skill of both partners), but the results be-
low do not change when removing these children. We
also confirmed that children at all ages understood the
Number Game and selected the more numerous color
during the 30 warm-up trials (see Table 3).
Table 3
Means and Tests Against Chance in Experiment 3
Age
(Years)
Mean
SD
df
t
p
d
Number Game (chose more numerous)
4
62.17
17.07
19
3.19
.005
0.71
5
79.00
11.75
19
11.03
< .001
2.47
6
82.67
7.84
19
18.63
< .001
4.16
7
88.17
4.39
19
38.88
< .001
8.69
Division (gave easier to unskilled)
4
52.14
13.73
19
0.70
.494
0.16
5
63.93
25.54
19
2.44
.025
0.55
6
67.50
18.02
19
4.34
< .001
0.97
7
72.50
21.40
19
4.70
< .001
1.05
As this study was within-subjects, we were most in-
terested to know if children were assigning harder
questions to the ‘good’ partner and therefore easier
questions to the ‘not so good’ partner. As shown in Ta-
ble 3 and Figure 4, children aged 5 and older followed
this pattern significantly above chance of 50%. A one-
way ANOVA found that children were significantly
more likely to match difficulty with skill at older age
groups, F(3, 76) = 3.70, p = .015, ηp2 = .13. Specifically,
7-year-olds were significantly more likely to match dif-
ficulty and skill than were 4-year-olds, Tukey posthoc
t(38) = 20.36, p = .012, d = 1.13. Thus, when children’s
self-interest was removed by using a third-party task
rather than a first-person task, children as young as 5
years old demonstrated strategic behavior consistent
with the law of comparative advantage – assigning
harder questions to the most competent partner. We re-
peat this analysis treating age continuously and only us-
ing performance on the first trial, and find that these
Baer and Odic 11
results are robust to these modifications (see Supple-
mental Material).
Figure 4. Percentage of Trials Assigning Easier Questions to
Worse Partner in Experiment 3. Error bars represent 1 stand-
ard error.
General Discussion
It can be challenging to optimally allocate tasks to
group members with differing skills. In three studies,
we examined whether children from ages 4-9 apply a
strategy based on the law of comparative advantage to
their collaborations in a numerical discrimination task.
In Experiment 1, children aged 6-9 assigned them-
selves easier questions when paired with a higher-
skilled partner, consistent with the law of comparative
advantage. But surprisingly, only 9-year-olds consist-
ently assigned themselves the harder questions if their
partner was less skilled. In Experiment 2, 7-year-olds
began to demonstrate this strategic reasoning when
their partner was described as older or younger, sug-
gesting that at least some of the apparent failure in Ex-
periment 1 might stem from children’s developing abil-
ity to estimate the probability of their partner’s success.
In Experiment 3, 5-7-year-olds assigned harder ques-
tions to a child ‘good’ at the task and easier questions
to a child ‘not so good’ at the task in a third-party par-
adigm where children were not direct participants in the
collaboration. This suggests that another influence on
performance in Experiment 1 was likely a self-serving
bias, in which children knew the best split of questions,
but did not enact it as a strategy out of a desire to priv-
ilege themselves.
These findings paint a nuanced picture of develop-
ing labor division. First, they demonstrate that by 5
years, children can enact a collaborative strategy in the
domain of cognitive labor consistent with the law of
comparative advantage, in which a higher-skilled part-
ner completes difficult tasks given the greater ad-
vantage to the team. This extends prior work of strate-
gic collaboration on a physical task (throwing objects;
Magid et al., 2018) and demonstrates breadth of this
reasoning to cognitive tasks (reasoning about num-
bers). Second, they show that enacting this strategy can
be constrained by a self-serving bias. Despite under-
standing the collaborative strategy at age 5, children
didn’t enact this strategy in their own collaborations
until 7-9 years. Like findings in the literature on chil-
dren’s fairness, developing collaborations appear to
also be influenced by goals to benefit the self above
others. Third, at least some of the effects can be ex-
plained by how ability was communicated to the chil-
dren. Seven-year-old children did not enact a strategy
consistent with the law of comparative advantage when
their partner was “not as good” as them, but did when
their partner was described as “younger”. This high-
lights a caution for researchers interested in measuring
or manipulating beliefs about ability or competence: we
may underestimate children’s reasoning by using ab-
stract terms like “better”.
Strategizing About Skill on Cognitive Tasks
Our goal was to examine the breadth of children’s
collaborative strategies, and to test what components of
their strategy use might differ when children apply the
same strategy in different domains. To that end, our re-
sults provide several new insights and highlight new di-
rections for understanding how children strategize
about cognitive skill.
With respect to the way children estimate the
chances of success on cognitive tasks, these findings
demonstrate that certain cues are more tightly linked to
judgments of skill than others. A simple change in the
partner’s skill manipulation from an abstract compari-
son (‘better’) to an age comparison (‘older’) resulted in
children applying the law of comparative advantage at
age 7 rather than 9. This could signal that children
begin to learn about skills by tracking the co-occur-
rence between performance and observable traits, later
mapping that understanding onto abstract comparisons.
At the same time, children as young as 5 years inter-
preted the abstract term ‘good’ to mean skilled in Ex-
periment 3. A fruitful avenue for future research will be
to uncover how children learn which cues are relevant
to their judgments of skill, and what other cues children
use (e.g., stereotypes about group membership; Bian et
al., 2017, or observed performance; Ronfard & Cor-
riveau, 2016).
Baer and Odic 12
With respect to formulating a strategy consistent
with the law of comparative advantage, we replicate
and extend the findings of Magid and colleagues (2018)
that this strategy emerges early in childhood. We found
success at 5 years, but not at 4 years as in their study.
However, we do not want to make strong claims that
the strategy must be learned independently for the
physical and cognitive domains based on this small dif-
ference. The number task we used has recently been
used in studies about children’s solo strategies, but with
limited success in children under age 5 (e.g., Baer &
Odic, 2019). In contrast, 4-year-olds applied the same
strategies on area discrimination (identifying the larger
of two shapes), a dimension that shows earlier and
more rapid development than number (Baer et al.,
2021; Odic, 2018). Therefore, the apparent failure of 4-
year-olds in Experiment 3 may have more to do with
children’s difficulty discriminating the difficulties ra-
ther than not understanding the strategy. Future work
with different tasks or more disparate difficulties that
4-year-olds can reliably detect will be necessary to pin
down when this strategy emerges.
A second avenue for future exploration in strategy
formation is what computations children perform when
devising this strategy. The formal computations for the
law of comparative advantage involve estimating four
separate probabilities of success and then comparing
their subtractions. Could children accomplish the same
outcome without these formal calculations? A simple
strategy could be to pair easier questions with the less
skilled group member, without thinking about the alter-
native of giving those easier questions to the skilled
partner. With this approach, children could theoreti-
cally formulate a strategy prior to estimating each
party’s chances of success on each individual trial (e.g.,
by deciding to give all easy questions to the partner
upon hearing they are not skilled, before seeing any
questions to be divided). This could explain children’s
success in the third-party task, where matching skill to
difficulty results in the optimal strategy. However, we
would have expected to see similar success in Experi-
ments 1 and 2 if this were true, yet we did not. One
means of testing this alternative is to present children
with a situation in which the law of comparative ad-
vantage favors the opposite pattern: assigning the
harder items to the unskilled partner. This would hap-
pen when both items are very difficult, making the
small increased chances of success by the skilled part-
ner on the easier question comparatively more advan-
tageous for the group.
Our results also prompt future exploration into the
use of this strategy in other cognitive skills and other
domains of labor. We focused here on a well-studied
perceptual magnitude task, but it would also be inter-
esting to explore how children divide labor on other
cognitive tasks. For memory, do children spend more
time trying to remember items that they have unique
access to over items that a partner can also access? Or
for language, do bilingual children consider each
party’s linguistic skill when negotiating which lan-
guage to communicate in? Outside of physical and cog-
nitive labor, future work could also explore how chil-
dren collaborate on emotional labor, such as dividing
care for others. The law of comparative advantage the-
oretically applies to all these tasks, provided there is a
way to quantify ‘success’ in each domain.
With respect to enacting the chosen solution, chil-
dren in our studies were influenced by a self-serving
bias, which could stem from several sources. One pos-
sibility is that children were motivated to seek rewards
(i.e. positive feedback from answering correctly) after
being compared to a peer (Magid & Schulz, 2015). An-
other possibility is that children were driven to reduce
their cognitive effort, choosing questions that could be
answered very quickly and without much thought (see
Halberda & Feigenson, 2008; Niebaum & Munakata,
2020). Yet another possibility is that children were
driven to repair their reputation following an unfavora-
ble comparison (see Shaw et al., 2014 for an example
within the literature on fairness). A growing body of
work suggests that children are very concerned with
maintaining their reputation (Silver & Shaw, 2018),
particularly about competence (Asaba & Gweon, 2019;
Zhao et al., 2018). Preliminary evidence of this possi-
bility comes from Experiment 2, where we found that
younger children claimed to be more skilled regardless
of how old their partner was, whereas older children
matched skill to relative age in both conditions. As no
such bias was seen when children collaborated on a
physical task (Magid et al., 2018), this may reveal dif-
ferences in how children value their reputation for com-
petency in different domains. Cognitive competency
might be especially tied to one’s identity and desired
reputation given its reliance on properties of the mind,
which is more closely linked to the self than other parts
of the body (Starmans & Bloom, 2012). Proving one’s
cognitive skill may be particularly important for chil-
dren as they develop independent identities.
Particularly relevant when considering self-serving
bias are the specific properties of our studies which
may limit generalization. For instance, children had to
take the experimenter’s word about a virtual partner,
which may have led children to doubt the true collabo-
Baer and Odic 13
rative nature of the task. Without full belief in the col-
laborative goal, children may have fallen back on a ra-
tional solo strategy of always answering the easiest
questions (Baer et al., 2018; Baer & Odic, 2019). Fur-
ther, we only recruited children from Vancouver, Brit-
ish Columbia, Canada and tested them in school set-
tings, which reflect a very small portion of children in
the world (Henrich et al., 2010; Nielsen et al., 2017).
Children in other societies or testing locations have
shown even stronger self-serving tendencies. One
cross-cultural study of fairness across seven societies,
for instance, found that only in some societies world-
wide (including Canada where testing took place)
would children avoid advantageous situations (Blake et
al., 2015). We might expect similar cultural differences
in collaborative strategies, as well.
Implications and Conclusions
In conclusion, 5-year-olds enact a collaborative
strategy consistent with the law of comparative ad-
vantage. Our results suggest that children’s perfor-
mance was somewhat limited by challenges in estimat-
ing another’s chances of success, and more dramati-
cally influenced by self-serving bias when enacting the
strategy.
These findings showcase a basic understanding
early in life about how to strategically divide cognitive
labor, something we often associate with highly-trained
economists and managers yet remains a fundamental
part of human collaboration. This contributes to ongo-
ing investigations into the cognition enabling humans
to engage in sophisticated social behaviors. For exam-
ple, our work provides initial evidence that children use
their own metacognitive confidence to assess what is
likely for both their own success and for others’ suc-
cess. This hypothesis can then be applied to many other
social behaviors including teaching, helping, and com-
municating that all rely on understanding others’ capa-
bilities.
Further, knowing that this basic strategic under-
standing is influenced by self-serving biases can be ex-
tremely informative for those interested in developing
better strategic thinking. For instance, decision-making
might be best when developing a strategy for others ra-
ther than for the self (see Kross & Grossmann, 2012).
Our work also highlights how some cues to compe-
tency might lead to different attitudes about capability.
This could be informative for work on modifying atti-
tudes around people with disabilities, where accommo-
dations for ‘visible’ disabilities are more positively
viewed than for ‘invisible’ disabilities (e.g., a broken
leg vs. dyslexia; Deckoff-Jones & Duell, 2018). Future
work into the factors impacting strategic decisions and
their development will help refine training programs
for these purposes.
References
Asaba, M., & Gweon, H. (2019). Young children can rationally re-
vise and maintain what others think of them. PsyArXiv.
https://doi.org/10.31234/osf.io/yxhv5
Baer, C., & Friedman, O. (2018). Fitting the message to the listener:
Children selectively mention general and specific facts. Child
Development, 89, 461–475. https://doi.org/10.1111/cdev.12751
Baer, C., Gill, I. K., & Odic, D. (2018). A domain-general sense of
confidence in children. Open Mind: Discoveries in Cognitive
Science, 2, 86–96. https://doi.org/10.1162/opmi_a_00020
Baer, C., Malik, P., & Odic, D. (2021). Are children’s judgments of
another’s accuracy linked to their metacognitive confidence
judgments? Metacognition and Learning, 16(2), 485–516.
https://doi.org/10.1007/s11409-021-09263-x
Baer, C., & Odic, D. (2019). Certainty in numerical judgments de-
velops independently of the Approximate Number System.
Cognitive Development, 52, 100817.
https://doi.org/10.1016/j.cogdev.2019.100817
Baer, C., & Odic, D. (2020). The relationship between children’s
approximate number certainty and symbolic mathematics.
Journal of Numerical Cognition, 6(1), 50–65.
https://doi.org/10.5964/jnc.v6i1.220
Balcomb, F. K., & Gerken, L. (2008). Three-year-old children can
access their own memory to guide responses on a visual match-
ing task. Developmental Science, 11, 750–760.
https://doi.org/10.1111/j.1467-7687.2008.00725.x
Bellon, E., Fias, W., & Smedt, B. D. (2020). Metacognition across
domains: Is the association between arithmetic and metacogni-
tive monitoring domain-specific? PLOS ONE, 15, e0229932.
https://doi.org/10.1371/journal.pone.0229932
Bennett-Pierre, G., Asaba, M., & Gweon, H. (2018). Preschoolers
consider expected task difficulty to decide what to do and whom
to help. In C. W. Kalish, M. Rau, J. Zhu, & T. T. Rogers (Eds.),
Proceedings of the 40th Annual Meeting of the Cognitive Sci-
ence Society.
Bian, L., Leslie, S.-J., & Cimpian, A. (2017). Gender stereotypes
about intellectual ability emerge early and influence children’s
interests. Science, 355, 389–391. https://doi.org/10.1126/sci-
ence.aah6524
Birch, S. A. J., Vauthier, S. A., & Bloom, P. (2008). Three-and four-
year-olds spontaneously use others’ past performance to guide
their learning. Cognition, 107, 1018–1034.
https://doi.org/10.1016/j.cognition.2007.12.008
Blake, P. R. (2018). Giving what one should: Explanations for the
knowledge-behavior gap for altruistic giving. Current Opinion
in Psychology, 20, 1–5.
https://doi.org/10.1016/j.copsyc.2017.07.041
Blake, P. R., & McAuliffe, K. (2011). “I had so much it didn’t seem
fair”: Eight-year-olds reject two forms of inequity. Cognition,
120(2), 215–224. https://doi.org/10.1016/j.cogni-
tion.2011.04.006
Blake, P. R., McAuliffe, K., Corbit, J., Callaghan, T. C., Barry, O.,
Bowie, A., Kleutsch, L., Kramer, K. L., Ross, E., Vongsachang,
H., Wrangham, R., & Warneken, F. (2015). The ontogeny of
fairness in seven societies. Nature, 528, 258–261.
https://doi.org/10.1038/nature15703
Baer and Odic 14
Blake, P. R., McAuliffe, K., & Warneken, F. (2014). The develop-
mental origins of fairness: The knowledge–behavior gap.
Trends in Cognitive Sciences, 18(11), 559–561.
https://doi.org/10.1016/j.tics.2014.08.003
Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vi-
sion, 10, 433–436. https://doi.org/10.1163/156856897X00357
Bridgers, S., Jara-Ettinger, J., & Gweon, H. (2020). Young children
consider the expected utility of others’ learning to decide what
to teach. Nature Human Behaviour, 4(2), 144–152.
https://doi.org/10.1038/s41562-019-0748-6
Call, J., & Carpenter, M. (2001). Do apes and children know what
they have seen? Animal Cognition, 3, 207–220.
https://doi.org/10.1007/s100710100078
Chernyak, N., & Sobel, D. M. (2015). Equal but not always fair:
Value-laden sharing in preschool-aged children. Social Devel-
opment. http://onlinelibrary.wiley.com.ezproxy.li-
brary.ubc.ca/doi/10.1111/sode.12136/pdf
Coughlin, C., Hembacher, E., Lyons, K. E., & Ghetti, S. (2015).
Introspection on uncertainty and judicious help-seeking during
the preschool years. Developmental Science, 18, 957–971.
https://doi.org/10.1111/desc.12271
Deckoff-Jones, A., & Duell, M. N. (2018). Perceptions of appropri-
ateness of accommodations for university students: Does disa-
bility type matter? Rehabilitation Psychology, 63(1), 68–76.
https://doi.org/10.1037/rep0000213
Geurten, M., Meulemans, T., & Lemaire, P. (2018). From domain-
specific to domain-general? The developmental path of meta-
cognition for strategy selection. Cognitive Development, 48,
62–81. https://doi.org/10.1016/j.cogdev.2018.08.002
Goupil, L., Romand-Monnier, M., & Kouider, S. (2016). Infants ask
for help when they know they don’t know. Proceedings of the
National Academy of Sciences, 113, 3492–3496.
https://doi.org/10.1073/pnas.1515129113
Halberda, J., & Feigenson, L. (2008). Developmental change in the
acuity of the “Number Sense”: The Approximate Number Sys-
tem in 3-, 4-, 5-, and 6-year-olds and adults. Developmental
Psychology, 44, 1457–1465. https://doi.org/10.1037/a0012682
Harris, P. L., Koenig, M. A., Corriveau, K. H., & Jaswal, V. K.
(2018). Cognitive foundations of learning from testimony. An-
nual Review of Psychology, 69, 251–273.
https://doi.org/10.1146/annurev-psych-122216-011710
Hembacher, E., & Ghetti, S. (2014). Don’t look at my answer: Sub-
jective uncertainty underlies preschoolers’ exclusion of their
least accurate memories. Psychological Science, 25, 1768–
1776. https://doi.org/10.1177/0956797614542273
Henrich, J., Heine, S. J., & Norenzayan, A. (2010). The weirdest
people in the world? Behavioral and Brain Sciences, 33(2–3),
61–83. https://doi.org/10.1017/S0140525X0999152X
Koenig, M. A., & Harris, P. L. (2005). Preschoolers mistrust igno-
rant and inaccurate speakers. Child Development, 76, 1261–
1277. https://doi.org/10.1111/j.1467-8624.2005.00849.x
Kross, E., & Grossmann, I. (2012). Boosting wisdom: Distance
from the self enhances wise reasoning, attitudes, and behavior.
Journal of Experimental Psychology: General, 141(1), 43–48.
https://doi.org/10.1037/a0024158
Lutz, D. J., & Keil, F. C. (2002). Early Understanding of the Divi-
sion of Cognitive Labor. Child Development, 73(4), 1073–
1084. https://doi.org/10.1111/1467-8624.00458
Lyons, K. E., & Ghetti, S. (2011). The development of uncertainty
monitoring in early childhood. Child Development, 82, 1778–
1787. https://doi.org/10.1111/j.1467-8624.2011.01649.x
Magid, R. W., DePascale, M., & Schulz, L. E. (2018). Four- and 5-
year-olds infer differences in relative ability and appropriately
allocate roles to achieve cooperative, competitive, and prosocial
goals. Open Mind, 2, 72–85.
https://doi.org/10.1162/opmi_a_00019
Magid, R. W., & Schulz, L. E. (2015). Quit while you’re ahead:
Preschoolers’ persistence and willingness to accept challenges
are affected by social comparison. Proceedings of the Annual
Meeting of the Cognitive Science Society.
Mills, C. M. (2013). Knowing when to doubt: Developing a critical
stance when learning from others. Developmental Psychology,
49, 404–418. https://doi.org/10.1037/a0029500
Niebaum, J., & Munakata, Y. (2020). Deciding What to Do: Devel-
opments in Children’s Spontaneous Monitoring of Cognitive
Demands. Child Development Perspectives, 14(4), 202–207.
https://doi.org/10.1111/cdep.12383
Nielsen, M., Haun, D., Kärtner, J., & Legare, C. H. (2017). The
persistent sampling bias in developmental psychology: A call
to action. Journal of Experimental Child Psychology, 162, 31–
38. https://doi.org/10.1016/j.jecp.2017.04.017
Odic, D. (2018). Children’s intuitive sense of number develops in-
dependently of their perception of area, density, length, and
time. Developmental Science, 21, e12533.
https://doi.org/10.1111/desc.12533
Pasquini, E. S., Corriveau, K. H., Koenig, M., & Harris, P. L.
(2007). Preschoolers monitor the relative accuracy of inform-
ants. Developmental Psychology, 43(5), 1216.
Pouget, A., Drugowitsch, J., & Kepecs, A. (2016). Confidence and
certainty: Distinct probabilistic quantities for different goals.
Nature Neuroscience, 19, 366–374.
https://doi.org/10.1038/nn.4240
Ricardo, D. (1891). Principles of Political Economy and Taxation.
G. Bell and sons.
Rochat, P., Dias, M. D. G., Liping, G., Broesch, T., Passos-Ferreira,
C., Winning, A., & Berg, B. (2009). Fairness in Distributive
Justice by 3- and 5-Year-Olds Across Seven Cultures. Journal
of Cross-Cultural Psychology, 40(3), 416–442.
https://doi.org/10.1177/0022022109332844
Ronfard, S., & Corriveau, K. H. (2016). Teaching and preschoolers’
ability to infer knowledge from mistakes. Journal of Experi-
mental Child Psychology, 150, 87–98.
https://doi.org/10.1007/s13164-017-0343-6
Ronfard, S., & Harris, P. L. (2017). Children’s decision to transmit
information is guided by their evaluation of the nature of that
information. Review of Philosophy and Psychology, 1–13.
https://doi.org/10.1007/s13164-017-0344-5
Shatz, M., & Gelman, R. (1973). The development of communica-
tion skills: Modifications in the speech of young children as a
function of listener. Monographs of the Society for Research in
Child Development, 38, 1–38. https://doi.org/10.2307/1165783
Shaw, A., Montinari, N., Piovesan, M., Olson, K. R., Gino, F., &
Norton, M. I. (2014). Children develop a veil of fairness. Jour-
nal of Experimental Psychology: General, 143, 363–375.
https://doi.org/10.1037/a0031247
Sheskin, M., Nadal, A., Croom, A., Mayer, T., Nissel, J., & Bloom,
P. (2016). Some Equalities Are More Equal Than Others: Qual-
ity Equality Emerges Later Than Numerical Equality. Child De-
velopment, 87(5), 1520–1528.
https://doi.org/10.1111/cdev.12544
Silver, I. M., & Shaw, A. (2018). Pint-Sized Public Relations: The
Development of Reputation Management. Trends in Cognitive
Sciences, 22(4), 277–279.
https://doi.org/10.1016/j.tics.2018.01.006
Baer and Odic 15
Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-Pos-
itive Psychology: Undisclosed Flexibility in Data Collection
and Analysis Allows Presenting Anything as Significant. Psy-
chological Science, 22, 1359–1366.
https://doi.org/10.1177/0956797611417632
Starmans, C., & Bloom, P. (2012). Windows to the soul: Children
and adults see the eyes as the location of the self. Cognition,
123(2), 313–318. https://doi.org/10.1016/j.cogni-
tion.2012.02.002
The Math Works Inc. (2015). MATLAB (9.0 (R2015b)) [Computer
software]. The Math Works Inc.
Tomasello, M. (1999). The cultural origins of human cognition (pp.
vi, 248). Harvard University Press.
Vo, V. A., Li, R., Kornell, N., Pouget, A., & Cantlon, J. F. (2014).
Young children bet on their numerical skills metacognition in
the numerical domain. Psychological Science, 25, 1712–1721.
https://doi.org/10.1177/0956797614538458
Warneken, F. (2018). How Children Solve the Two Challenges of
Cooperation. Annual Review of Psychology, 69(1), 205–229.
https://doi.org/10.1146/annurev-psych-122216-011813
Warneken, F., Steinwender, J., Hamann, K., & Tomasello, M.
(2014). Young children’s planning in a collaborative problem-
solving task. Cognitive Development, 31, 48–58.
Zhao, L., Heyman, G. D., Chen, L., & Lee, K. (2018). Telling young
children they have a reputation for being smart promotes cheat-
ing. Developmental Science, 21(3), e12585.
https://doi.org/10.1111/desc.12585