ArticlePDF Available

Abstract and Figures

Researchers, educators, and parents have long believed that children learn cause and effect relationships through exploratory play. However, previous research suggests that children are poor at designing informative experiments; children fail to control relevant variables and tend to alter multiple variables simultaneously. Thus, little is known about how children's spontaneous exploration might support accurate causal inferences. Here the authors suggest that children's exploratory play is affected by the quality of the evidence they observe. Using a novel free-play paradigm, the authors show that preschoolers (mean age: 57 months) distinguish confounded and unconfounded evidence, preferentially explore causally confounded (but not matched unconfounded) toys rather than novel toys, and spontaneously disambiguate confounded variables in the course of free play.
Content may be subject to copyright.
Serious Fun: Preschoolers Engage in More Exploratory Play When
Evidence Is Confounded
Laura E. Schulz and Elizabeth Baraff Bonawitz
Massachusetts Institute of Technology
Researchers, educators, and parents have long believed that children learn cause and effect relationships
through exploratory play. However, previous research suggests that children are poor at designing
informative experiments; children fail to control relevant variables and tend to alter multiple variables
simultaneously. Thus, little is known about how children’s spontaneous exploration might support
accurate causal inferences. Here the authors suggest that children’s exploratory play is affected by the
quality of the evidence they observe. Using a novel free-play paradigm, the authors show that preschool-
ers (mean age: 57 months) distinguish confounded and unconfounded evidence, preferentially explore
causally confounded (but not matched unconfounded) toys rather than novel toys, and spontaneously
disambiguate confounded variables in the course of free play.
Keywords: causal learning, exploratory play, preschoolers’ scientific reasoning, ambiguous evidence,
confounded variables
Causal knowledge is fundamental to individuals’ understanding
of the world. It informs their moral judgments, their explanations
of the past, and their plans for the future. Little wonder that Hume
(1739/2000) called causal knowledge the “cement of the universe.”
However, relatively little is known about how children learn the
causal structure of events.
Piaget (1930) believed that young children came to understand
causal relationships through active exploration of their environ-
ment. Recent research suggests that very young children know
much more about the causal structure of the world than Piaget
believed (Baillargeon, Kotovsky, & Needham, 1995; Flavell,
Green, & Flavell, 1995; Gelman & Wellman, 1991; Kalish, 1996;
Leslie & Keeble, 1987; Saxe, Tenenbaum, & Carey, 2005; Spelke,
Breinlinger, Macomber, & Jacobson, 1992). However, the idea at
the heart of the Piageti an account—that children “construct”
knowledge (and particularly causal knowledge) by active explora-
tion—remains widely accepted.
Yet despite substantial agreement that children learn through
play (e.g., Bruner, Jolly, & Sylva, 1976; Singer, Golinkoff, &
Hirsh-Pasek, 2006), little is known about how exploratory play
might support accurate causal inferences. Descriptive studies of
play are more common than experimental research or theoretical
accounts, and much of the seminal work on exploratory play
predates recent research on children’s causal reasoning (Berlyne,
1954; Piaget, 1951). Except for the well-established finding that
children (and many nonhuman animals) selectively explore novel
stimuli (Berlyne, 1960; Dember & Earl, 1957; Henderson &
Moore, 1980; Hutt & Bhavnani, 1972; Pavlov, 1927), there is little
evidence for systematic patterns in children’s exploratory behav-
ior. Moreover, considerable research suggests that even older
children and naive adults are poor at designing causally informa-
tive experiments and have difficulty anticipating the type of evi-
dence that would support or undermine causal hypotheses (Chen &
Klahr, 1999; Inhelder & Piaget, 1958; Kuhn, 1989; Kuhn, Amsel,
& O’Laughlin, 1988; Koslowski, 1996; Masnick & Klahr, 2003).
Such findings pose a challenge for the constructivist account. The
number of events children might explore in principle is vastly
greater than the number of events they can explore in practice. If
children’s exploratory play is largely unsystematic, how might
they generate the type of evidence that could support efficient
causal learning?
We hypothesized that children’s exploratory play might be
affected not just by the novelty or perceptual complexity of stimuli
but also by the quality of the evidence they observe. Although
young children do not design controlled experiments, in simple
contexts, they seem to recognize the difference between informa-
tive and uninformative evidence (Masnick & Klahr, 2003; Sodian,
Zaitchik, & Carey, 1991), and they can use patterns of evidence to
make predictions, interventions, and even counterfactual claims
(Gopnik & Schulz, 2004, 2007; Gopnik, Sobel, Schulz, & Gly-
mour, 2001; Kushnir & Gopnik, 2005, 2007; Schulz & Gopnik,
2004, 2007; Shultz & Mendelson, 1975; Siegler & Liebert, 1975;
Sobel, 2004; Sobel & Kirkham, 2006). We predicted that pre-
schoolers would distinguish confounded and unconfounded evi-
dence and would engage in more exploratory play when evidence
failed to disambiguate the causal structure of events. If children
systematically engage in more exploratory play when causal evi-
dence is confounded, then even if children do not generate con-
trolled experiments, they might isolate the relevant variables in the
Laura E. Schulz and Elizabeth Baraff Bonawitz, Department of Brain
and Cognitive Sciences, Massachusetts Institute of Technology.
This research was supported by a James H. Ferry and a McDonnell
Foundation Collaborative Initiative Causal Learning grant to Laura E.
Schulz. We thank Darlene Fe r r a n t i , Noah Goodman, Kat e Hooppell,
Adrianna Jenkins, Rebecca Saxe, Andrew Shtulman, and Josh Tenenbaum
for helpful comments and suggestions; the Discovery Center at the Mu-
seum of Science, Boston; and participating parents and children.
Correspondence concerning this article should be addressed to Laura E.
Schulz, Department of Brain and Cognitive Sciences, Massachusetts Insti-
tute of Technology, 46-4011, 77 Massachusetts Avenue, Cambridge, MA
02139. E-mail:
Developmental Psychology Copyright 2007 by the American Psychological Association
2007, Vol. 43, No. 4, 1045–1050 0012-1649/07/$12.00 DOI: 10.1037/0012-1649.43.4.1045
course of free play and generate the type of evidence that could
support accurate causal learning.
In this study, we introduced children to a toy and showed them
either confounded or unconfounded evidence about the causal
structure of the toy. We removed the toy and then returned it along
with a novel toy. We allowed the children to play freely for 60 s.
We predicted that children who observed confounded evidence
would preferentially play with the familiar toy but that children
who observed unconfounded evidence would show the standard
novelty preference and play primarily with the novel toy.
Participants. We recruited 64 preschoolers (mean age: 57
months; range: 48–70 months) from the Discovery Center of a
metropolitan science museum and from urban area preschools.
Sixteen children were tested in each of four conditions: a con-
founded evidence condition and three unconfounded conditions,
described below. Approximately equal number of boys and girls
participated in each condition (45% girls overall). Although most
of the children were White and middle class, a range of ethnicities
and socioeconomic backgrounds reflecting the diversity of the
local population was represented.
Materials. Two boxes were constructed from 15 cm ! 15 cm
balsa boards. One box had a single lever and was covered in
yellow felt. The other box had two levers and was covered in red
felt. On the yellow box a small (5 cm high) fuzzy duck toy was
attached to a dowel 20 cm in length and 1 cm in diameter that
passed through a small hole in the side of the box. The dowel acted
as a lever. When the dowel was depressed on the outside of the
box, the inside end moved upwards, causing the duck to pop up
through a slit in the felt on the top of the box. The construction of
the double-lever box was identical, except there was a second lever
on the side of the box adjacent to the first lever. On this second
lever, a small L-shaped bracket was attached to a (7 cm high)
puppet made of drinking straws so that when the second lever was
depressed, the straw puppet could pop up without affecting the
movement of the first lever. The ends of the two levers were less
than 40 cm apart and were easily manipulated, both separately and
simultaneously, by preschool children.
Procedure. Children were tested individually in a quiet corner
of their preschools or in the Discovery Center. The experimenter
sat next to the child at a table. Both boxes were on a far corner of
the table and were covered with a cloth so the child could not see
them. The experimenter said, “We’re going to play a game today.”
The experimenter brought the red two-lever box out from under
the cloth and introduced it to the child.
In the confounded condition, the experimenter said, “You push
down your lever and I’ll push down my lever at the same time.
Ready: one, two, three, down!” When both levers were depressed,
a duck and a straw puppet popped out of the middle of the box. The
spatial locations of the duck and the puppet were uninformative
about their causal relationships with the lever; that is, the objects
appeared in the middle of the box so it was not possible, just by
looking, to determine which lever controlled which object. After
approximately 2 s, the experimenter said, “One, two, three, up!”
The experimenter and the child simultaneously released the levers
and the duck and puppet disappeared from view. Counting aloud
was an effective means of coordinating the child’s actions with the
experimenter’s so that the onset and offset of events appeared
simultaneous; pilot work established that even adult observers
failed to perceive temporal cues that would disambiguate the
causal structure of the toy. The procedure was then repeated twice
more so that in total both levers were pushed three times and both
effects (the duck and the puppet) occurred three times. Because the
two candidate causes were always manipulated simultaneously, the
evidence failed to disambiguate the many possible causal struc-
tures that might underlie the event (either lever might activate the
duck or the puppet, one or both levers might activate both, or the
levers might interact).
The unconfounded/matched-for-effect condition was designed
to replicate the effects of the confounded condition but with an
unambiguous causal structure. The unconfounded/matched-for-
effect condition was identical to the confounded condition except
that the child and the experimenter pressed and released their
levers simultaneously only twice. On the next trial, the experi-
menter said, “Let’s take turns.” (The order of turn taking and the
particular effect was counterbalanced among participants.) “You
go ahead. One, two, three, your turn!” The child pushed his or her
lever and just the duck popped up. The experimenter then said,
“Now it’s my turn.” After the child released his or her lever, the
experimenter counted “One, two, three,” pushed her lever, and just
the puppet popped up. Thus, as in the confounded condition, each
effect (the duck and the puppet) occurred three times; however,
this evidence fully disambiguated the causal structure of the toy:
The child could see that one lever activated the duck and the other
activated the puppet.
Conceivably, however, the exposure to the additional trial (as
indicated by the “one, two, three” counting ritual) might decrease
children’s interest in the familiar toy. The unconfounded/matched-
for-trials condition was designed to control for the possibility that
the additional trial bored the children. The condition was identical
to the unconfounded/matched-for-effect condition except that in
this condition, the child and the experimenter pressed and released
their levers simultaneously only once. On the second trial, the
child pressed his or her lever by him- or herself; on the third trial,
the experimenter pressed her lever by herself. Thus, each effect
occurred twice; however, as in the confounded condition, there
were three distinct trials. Again, the evidence fully disambiguated
the causal structure of the toy.
Alternatively, children might play more with the familiar toy in
the confounded condition than in the unconfounded conditions
because they were allowed to play independently with the toy in
the unconfounded conditions but not in the confounded condition.
To control for this possibility, we tested children in an
unconfounded/no-independent-play condition.
In this condition,
the experimenter and the child pressed and released their levers
simultaneously once. Then the experimenter pushed one lever and
just the duck popped up. She released that lever and pushed the
other lever and just the puppet popped up. The experimenter and
the child then pressed and released their levers simultaneously a
second time. As in the confounded condition, the child never had
a chance to manipulate the toy without the experimenter; however,
We are indebted to an anonymous reviewer for suggesting this control
this evidence fully disambiguated the causal structure of the toy.
There was no significant difference in the length of time children
were exposed to the effects of the familiar toy in the confounded
condition and any of the unconfounded conditions (confounded:
M " 12.1 s; unconfounded/matched for effects: M " 13.5 s;
unconfounded/matched for trials: M " 11.6 s; unconfounded/no
independent play: M " 13 s; for each comparison t[30]).
After the child observed the evidence, the experimenter returned
the red box to the far end of the table and uncovered the novel
yellow box. The experimenter then rotated the table so that the
boxes were just out of arms’ reach of the child (so the child had to
stretch to reach either box). The boxes were located approximately
2 feet (60.96 cm) apart from each other (left/right position of the
boxes counterbalanced among children). The experimenter said,
“I’ll be back in just a minute. Go ahead and play” and walked out
of the child’s line of sight. After 60 s, the experimenter returned,
thanked the child for participating, and ended the experiment.
Results and Discussion
Children were counted as playing with a box as long as they
were touching the box, and we coded the total amount of time that
each child played with each box. We analyzed children’s explor-
atory play in three ways: we looked at whether, on average,
children played longer with the familiar box or the novel box; we
looked at how many individual children preferentially played with
each box; and we looked at whether children’s first reach was to
the familiar box or to the novel box. Additionally, we coded
children’s actions in the confounded condition to see whether
children who played with the familiar toy spontaneously disam-
biguated the evidence. Children were counted as fully disambig-
uating the evidence if, in the course of their free exploratory play,
they depressed and released each lever separately at least once. If
children only isolated one of the two levers, if they only ever
moved both levers together, or if they engaged only in unrelated
exploratory play (e.g., reaching in the box; shaking the box), they
were counted as failing to fully disambiguate the evidence.
All data were coded by Elizabeth Baraff Bonawitz and were
recoded by a blind coder.
Intercoder agreement on children’s play
time was high across all conditions (r " .949); coders agreed
perfectly on children’s first reach and whether children fully
disambiguated the evidence (100% agreement). If children played
for less than 15 s overall, they were dropped from the study and
replaced. One child was replaced in the confounded condition; no
children were replaced in any other condition.
By all three measures, children were more likely to explore the
familiar toy in the confounded condition than in the unconfounded
conditions; there were no differences among the three uncon-
founded conditions on any analysis (see Figure 1). We compared
how long children played with each toy in each condition by doing
a2! 4 mixed analysis of variance, with play time on each toy as
the within-subjects variable and condition as the between-subjects
variable. There was an interaction between condition and toy
preference, F(3, 60) " 11.64, p # .0001. There was also a main
effect of toy type, F(3, 60) " 6.53, p # .05, suggesting that,
collapsing across the four conditions, children preferred the novel
toy. There was no main effect of condition, suggesting that chil-
dren in each group played for the same amount of time overall
(mean playtime across conditions was 27.6 s per toy).
To follow-up on the omnibus analysis of variance, we did
pairwise analyses of the four conditions. Each analysis was a 2 !
2 mixed analysis of variance, with play time on each toy as the
within-subjects variable and condition as the between-subjects
variable. Comparisons between the confounded condition and each
unconfounded condition revealed no main effect of play time
(averaging across the two conditions, children did not prefer one
toy to the other) and no main effect of condition (overall, children
played for the same amount of time in each condition) but did
reveal a significant interaction: Children spent more time playing
with the familiar toy in the confounded condition than in the
matched-for-effects condition, F(1, 32) " 17.32, p # .001, in the
matched-for-trials condition, F(1, 32) " 13.86, p # .001, and in
the no-independent-play condition, F(1, 32) " 10.83, p # .01.
Additionally, more children spent the majority of their time
playing with the familiar toy in the confounded condition than in
the matched-for-effects condition, $
(1, N " 32) " 8.13, p # .01,
in the matched-for-trials condition, $
(1, N " 32) " 6.15, p #
.025, and in the no-independent-play condition, $
(1, N " 32) "
4.5, p # .05. Finally, children were more likely to reach first for
the familiar toy in the confounded condition than in the matched-
for-effects condition, $
(1, N " 32) " 5.24, p # .025, and there
was a similar trend for children in the matched-for-trials condition,
(1, N " 32) " 3.46, p " .06, and in the no-independent-play
condition, $
(1, N " 32) " 3.46, p " .06.
Within the confounded condition, children played significantly
longer with the familiar toy than the novel toy, t(15) " 2.79, p #
.01. These results reversed for children in the matched-for-effects,
t(15) " 3.1, p # .01, and matched-for-trials, t(15) " 2.48, p " .01,
conditions. Children played equally long with both toys in the
no-independent-play condition, t(15) " 1.06. In the confounded
condition, a nonsignificant majority of children played most with
the familiar toy (nonsignificant by binomial test; one-tailed
throughout) but more children played most with the novel toy in
the matched-for-effects ( p " .01 by binomial test) and matched-
for-trials ( p # .05 by binomial test) conditions and were margin-
ally more likely to play with the novel toy in the no-independent-
play condition ( p " .07 by binomial test). Finally, in the
confounded condition, children’s first reach was just as likely to be
for the familiar toy as for the novel toy (nonsignificant by binomial
test), whereas children were significantly more likely to reach first
for the novel toy than for the familiar toy in all unconfounded
conditions (matched for effects: p # .01 by binomial test; matched
for trials: p " .01 by binomial test; no independent play: p # .01
by binomial test).
It is possible that the children were simply more interested in the
simultaneous effects than in the separate effects. However, if
children were more interested in the three simultaneous effects of
the confounded condition than, for instance, in the two simulta-
neous effects of the unconfounded/matched-for-effects and no-
independent-play conditions, one might also expect children to
play more with the familiar toy in those conditions than in the
unconfounded/matched-for-trials condition in which the effects
occurred simultaneously only once. In fact, there were no signif-
icant differences among the unconfounded conditions. This finding
suggests that it is the absence of disambiguating evidence rather
The clips for 2 children in the unconfounded/no-independent-play
condition were lost because of technical error after the original coding.
These 2 children were not recoded.
than the presence of simultaneous effects that encourages chil-
dren’s exploration.
We also looked at the actions children performed on the familiar
box in the confounded condition. In the course of their free play
with the familiar box, children often manipulated the levers simul-
taneously. Critically, however, 12 of the 16 children (75%) also
manipulated each lever separately, fully disambiguating the evi-
This finding suggests that children’s free exploratory play
could, in principle, generate the type of evidence that would
support accurate causal learning.
General Discussion
Our findings suggest that preschoolers’ spontaneous exploratory
play is sensitive not just to stimulus features such as novelty and
perceptual salience but also to formal properties of evidence, like
confounding. Note that in all four conditions, children were famil-
iarized with the same toy and children’s exposure to the toy’s
effects and affordances was closely matched across conditions. A
single manipulation seemed to drive the effect: If the two levers
were moved separately, on just a single trial, the children spent
most of their free time playing with the novel toy; if the two levers
were always moved simultaneously, children spent most of their
free time playing with the familiar toy. Children appear to recog-
We subsequently coded children’s actions on the familiar box in the
unconfounded conditions. In the matched-for-effects condition, 31% of the
children manipulated each lever separately; in both the matched-for-trials
and no-independent-play conditions, 50% of the children manipulated each
lever separately (this percentage does not include the 2 children in the
no-independent-play condition whose clips were lost; see Footnote 2). Of
course, the children played longer with the familiar box in the confounded
condition. However, it is interesting that the children were, if anything,
more likely to manipulate the levers separately in the confounded condition
than in the unconfounded conditions, where they had actually observed the
separate manipulations.
Figure 1. Experimental design and results
nize confounded evidence and are motivated to explore stimuli
whose causal structure is ambiguous.
In this study, we relied on an implicit measure of children’s
understanding of confounding—spontaneous exploratory play. Be-
cause previous research suggests that children have a poor meta-
cognitive understanding of confounding and experimental design
(Chen & Klahr, 1999; Inhelder & Piaget, 1958; Koslowski, 1996;
Kuhn, 1989; Kuhn et al., 1988; Masnick & Klahr, 2003), we
expected that children might not be aware of their own motivation
for exploration. However, it is possible that at least some of the
children might have been able to say that they were more curious
about one toy than another or articulate why they wanted to play
with one toy more than another. Future researchers might inves-
tigate the extent to which preschoolers explicitly recognize con-
founded evidence. Additionally, the majority of children in this
study were visitors to a science museum, an environment partic-
ularly likely to encourage exploratory play. Future researchers
might look at the extent to which these findings generalize to other
settings and populations.
Do children actually learn causal relationships from the evi-
dence of their own interventions? Our experiment does not address
this question directly, although in simple cases (i.e., when each
lever either does or does not cause an effect), it seems probable
that children would. However, we do not want to suggest that in all
cases children’s free exploratory play reliably leads to accurate
causal learning. There is every reason to believe that in many
contexts children’s spontaneous exploratory play might not suffice
for correct causal inferences. Children might be inaccurate for
many reasons: because they are unable to disambiguate the rele-
vant variables, because they fail to disambiguate the relevant
variables, or because they fail to attend sufficiently to the evidence
they generate in exploratory play. Nonetheless, children’s ten-
dency to selectively explore confounded events could be advanta-
geous for causal learning; regardless of whether children learn
from their explorations in any particular instance, overall, they
would be more likely to explore where there is something to be
It is important to note, however, that recent research suggests
that children do indeed learn from exploratory play in some
contexts. Preschoolers, for instance, were able to use disambigu-
ating evidence generated by their spontaneous play with a gear toy
to distinguish causal chain and common cause structures (Schulz,
Gopnik, & Glymour, 2007). Such findings are consistent with the
possibility that children’s selective exploration of confounded ev-
idence might support causal learning. Future researchers might
investigate the generality of this hypothesis by (a) looking at the
range of contexts and ages in which children are sensitive to and
selectively explore confounded evidence (e.g., whether such find-
ings hold for toddlers and infants), (b) looking at the extent to
which children’s free exploratory play generates informative evi-
dence, and (c) looking at the extent to which children learn from
the evidence of their own interventions.
Children’s exploratory play is a complex, dynamic phenome-
non, indubitably affected by many factors (e.g., the child’s tem-
perament, the child’s comfort and energy level, and the perceived
cost or benefit of various actions in terms of effort expended,
knowledge gained, and external reinforcement). However, these
results suggest that children’s normative understanding of evi-
dence and their curiosity about the causal structure underlying
observed evidence play a significant role in their decision to
explore. At least in simple cases, preschool children distinguish
confounded and unconfounded evidence and selectively engage in
more exploration when the causal structure of events is ambiguous.
The exploratory play of even very young children appears to
reflect some of the logic of scientific inquiry.
Baillargeon, R., Kotovsky, L., & Needham, A. (1995). The acquisition of
physical knowledge in infancy. In D. P. D. Sperber (Ed.), Causal
cognition: A multidisciplinary debate (pp. 79 –116). New York: Claren-
don Press/Oxford University Press.
Berlyne, D. E. (1954). An experimental study of human curiosity. British
Journal of Psychology, 45, 256 –265.
Berlyne, D. E. (1 960). Co nflict, arousal and curiosity. New York:
Bruner, J., Jolly, A., & Sylva, K. (1976). Play—Its role in development and
evolution. New York: Basic Books.
Chen, Z., & Klahr, D. (1999). All other things being equal: Acquisition and
transfer of the control of variables strategy. Child Development, 70,
1098 –1120.
Dember, W. N., & Earl, R. W. (1957). Analysis of exploratory, manipu-
latory, and curiosity behaviors. Psychological Review, 64, 91–96.
Flavell, J. H., Green, F. L., & Flavell, E. R. (1995). Young children’s
knowledge about thinking. Monographs of the Society for Research in
Child Development, 60(1, Serial No. 24).
Gelman, S. A., & Wellman, H. M. (1991). Insides and essence: Early
understandings of the non-obvious. Cognition, 38, 213–244.
Gopnik, A., & Schulz, L. E. (2004). Mechanisms of theory-formation in
young children. Trends in Cognitive Science, 8, 371–377.
Gopnik, A., & Schulz, L. E. (2007). Causal learning: Psychology, philos-
ophy, and computation. New York: Oxford University Press.
Gopnik, A., Sobel, D. M., Schulz, L. E., & Glymour, C. (2001). Causal
learning mechanisms in very young children: Two-, three-, and four-
year-olds infer causal relations from patterns of variation and covaria-
tion. Developmental Psychology, 37, 620 629.
Henderson, B., & Moore, S. G. (1980). Children’s responses to objects
differing in novelty in relation to level of curiosity and adult behavior.
New York: Dover.
Hume, D. (2000). A treatise on human nature (D. F. Norton & M. J.
Norton, Eds.). New York: Oxford University Press. (Original work
published 1739)
Hutt, C., & Bhavnani, R. (1972, May 19). Predictions from play. Nature,
237, 171–172.
Inhelder, B., & Piaget, J. (1958). The growth of logical thinking from
childhood to adolescence. New York: Basic Books.
Kalish, C. (1996). Causes and symptoms in preschoolers’ conceptions of
illness. Child Development, 67, 1647–1670.
Koslowski, B. (1996). Theory and evidence: The development of scientific
reasoning. Cambridge, MA: MIT Press.
Kuhn, D. (1989). Children and adults as intuitive scientists. Psychological
Review, 96, 674 689.
Kuhn, D., Amsel, E., & O’Laughlin, M. (1988). The development of
scientific thinking skills. Orlando, FL: Academic Press.
Kushnir, T., & Gopnik, A. (2005). Children infer causal strength from
probabilities and interventions. Psychological Science, 16, 678 683.
Kushnir, T., & Gopnik, A. (2007). Conditional probability versus spatial
contiguity in causal learning: Preschoolers use new contingency evi-
dence to overcome prior spatial assumptions. Developmental Psychol-
ogy, 43, 186 –196.
Leslie, A. M., & Keeble, S. (1987). Do six-month-old infants perceive
causality? Cognition, 25, 265–288.
Masnick, A. M., & Klahr, D. (2003). Error matters: An initial exploration
of elementary school children’s understanding of experimental error.
Journal of Cognition and Development, 4, 67–98.
Pavlov, I. P. (1927). Conditioned reflexes: An investigation of the physi-
ological activity of the cerebral cortex. New York: Dover.
Piaget, J. (1930). The child’s conception of physical causality. New York:
Harcourt, Brace.
Piaget, J. (1951). Plays, dreams, and imitation in childhood. London:
Routledge and Kegan Paul.
Saxe, R., Tenenbaum, J. B., & Carey, S. (2005). Secret agents: Inferences
about hidden causes by 10- and 12-month-old infants. Psychological
Science, 16, 995–1001.
Schulz, L. E., & Gopnik, A. (2004). Causal learning across domains.
Developmental Psychology, 40, 162–176.
Schulz, L. E., Gopnik, A., & Glymour, C. (2007). Preschool children learn
about causal structure from conditional interventions. Developmental
Science, 10, 322–332.
Shultz, T. R., & Mendelson, R. (1975). The use of covariation as a
principle of causal analysis. Child Development, 46, 394 –399.
Siegler, R. S., & Liebert, R. M. (1975). Acquisition of formal scientific
reasoning by 10- and 13-year-olds: Designing a factorial experiment.
Developmental Psychology, 11, 401– 412.
Singer, D. G., Golinkoff, M. R., & Hirsh-Pasek, K. (2006). Play "
learning: How play motivates and enhances children’s cognitive and
social– emotional growth. New York: Oxford University Press.
Sobel, D. M. (2004). Exploring the coherence of young children’s explan-
atory abilities: Evidence from generating counterfactuals. British Jour-
nal of Developmental Psychology, 22, 37–58.
Sobel, D. M., & Kirkham, N. Z. (2006). Blickets and babies: The devel-
opment of causal reasoning in toddlers and infants. Developmental
Psychology, 42, 1103–1115.
Sodian, B., Zaitchik, D., & Carey, S. (1991). Young children’s differenti-
ation of hypothetical beliefs from evidence. Child Development, 62,
Spelke, E. S., Breinlinger, K., Macomber, J., & Jacobson, K. (1992).
Origins of knowledge. Psychological Review, 99, 605– 632.
Received June 16, 2006
Revision received February 8, 2007
Accepted February 19, 2007 !
Call for Papers:
Special Section on New Perspectives on the Effects of Action on
Perceptual and Cognitive Development
Developmental Psychology invites manuscripts for a special section on new perspectives on the
effects of action on perceptual and cognitive development, to be compiled by guest editor David H.
Rakison and Associate Editor Amanda Woodward. The goal of the special section is to bring
together diverse new approaches to understanding the effects of action on perceptual and cognitive
development. It is hoped that submitted papers will report work from across a broad set of
perspectives, including perceptual, cognitive, social, biological, or cultural, and with a wide variety
of methodologies. Papers must present new empirical findings as well as theoretical integration
relevant to the topic of the special section. The main text of each manuscript, exclusive of figures,
tables, references, and/or appendixes, should not exceed 20 double-spaced pages (approximately
5,000 words).
The submission deadline is July 30, 2007. Initial inquiries regarding the special section may be
sent to David H. Rakison at or to Amanda Woodward at Manuscripts must be submitted electronically through the Manuscript
Submission Portal of Developmental Psychology at
submission.html and a hard copy sent to Cynthia Garcı´a Coll, Editor, Developmental Psychology,
Center for the Study of Human Development, Brown University, Box 1831, Providence RI 02912.
Please be sure to specify in the cover letter that your submission is intended for the special section.
For instructions to authors and other detailed submission information, see the journal Web site at
... (i.e., "Is it this one?"). Although the link between verbal skills and performance on the 20-questions game has never been investigated directly, efficient search emerges much earlier in tasks that do not involve verbal question asking (e.g., Bonawitz et al., 2012;Cook et al., 2011;Domberg et al., 2020;Kushnir & Gopnik, 2005;E. Schulz et al., 2019;L. E. Schulz & Bonawitz, 2007), which suggests a strong involvement of language skills in the development of informative question asking. ...
... 14 SWABODA, MEDER, AND RUGGERI Ruggeri et al., 2017;E. Schulz et al., 2019; L. E. Schulz & Bonawitz, 2007) and supports the view that children are skilled active learners from a very young age (e.g., Gopnik et al., 1997;Gopnik & Wellman, 2012). ...
We investigate whether a spatial representation of a search task supports 4- to 7-year-old children's information-search strategies, relative to their performance in a question-asking game. Children played two computationally and structurally analogous search games: a spatial search task, the maze-exploration game, in which they had to discover the path through a maze by removing masks covering its passages; and a verbal search task, the 20-questions game, where they had to identify a target monster from a set of eight monsters by asking yes-no questions. Across four experiments, we found that children searched more efficiently when they could make queries nonverbally (Experiments 1 and 2a). We also found that merely providing children with a visual conceptual aid that supports their representation of the hypothesis space (Experiment 2b), or familiarizing them with the hypothesis-space structure (Experiment 3) was not sufficient to improve their search strategies. Together, our results suggest that young children's difficulties in the 20-questions game are mainly driven by the verbal requirements of the task. However, they also demonstrate that efficient search strategies emerge much earlier than previously assumed in tasks that do not rely on verbal question generation. These findings highlight the importance of developing age-appropriate paradigms that capture children's early competence, in order to gain a more comprehensive picture of their emerging information-search abilities. (PsycInfo Database Record (c) 2022 APA, all rights reserved).
... When given a choice between a novel toy and toy young children are familiar with, children usually choose to interact with the recognized toy (L. E. Schulz & Bonawitz, 2007). As stated before, many young children probably do not interact with animal-based loose parts such as antlers, turtle shells, snakeskin, and clam shells on a regular basis. ...
... When a novel toy and familiar toy are present in the same space young children usually choose to interact with the familiar toy (L. E. Schulz & Bonawitz, 2007). Educators who wish to utilize loose part play to provide divergent play opportunities need to make those items a part of their daily routine in indoor and outdoor play spaces. ...
Outdoor play is an important aspect of young children’s health social-cognitive development. However, play in natural environments is declining due to urbanization and various safety concerns. Many urban preschools have outdoor play spaces that lack natural elements that stimulate children’s autonomy, creativity, and imaginative play. Furthermore, parents who find outdoor environments intimidating and fraught with danger limit young children’s outdoor experiences that inhibit their motor fitness, socialization with peers, and ecological awareness. Two qualitative case studies examined preschool children’s outdoor play. Study one focused on preschool children’s loose parts play in urban settings while study two examined parent’s attitudes towards outdoor play with young children. Key findings included children engaged in dramatic play more with natural loose parts than manufactured loose parts. And playgrounds with age/developmentally appropriate equipment, barriers/fences, and open/centralized play spaces with clear views would make parents more comfortable in providing outdoor play. Advisor: Lisa Pennisi
... Causal intervention involves learners experimenting to understand how things work (Liquin & Lombrozo, 2020b). Preschool learners have been found to choose actions to isolate variables to determine the cause and effect of processes, and develop hypotheses of how things work, which are then refined based on more information (Gopnick et al., 2001;Kidd & Hayden, 2015a;Schulz & Bonawitz, 2007). ...
... Similarly, object-oriented play was strongly associated with selecting the correct tool to retrieve a toy during an experimental task (Gredlein and Bjorklund, 2005). Recent studies also suggests that children solve instrumental problems during play by generating evidence that supports accurate causal learning (Schulz and Bonawitz, 2007;, discovering new object affordances (Bonawitz et al., 2009), and innovating objects using novel materials (Lew-Levy et al., 2021). Together, these findings lend support to the theoretical link between object play and some forms of creative problem solving. ...
Full-text available
Learning to use, make, and modify tools is key to our species’ success. Researchers have hypothesized that play with objects may have a foundational role in the ontogeny of tool use and, over evolutionary timescales, in cumulative technological innovation. Yet, there are few systematic studies investigating children’s interactions with objects outside the post-industrialized West. Here, we survey the ethnohistorical record to uncover cross-cultural trends regarding hunter-gatherer children’s use of objects during play and instrumental activities. Our dataset, consisting of 434 observations of children’s toys and tools from 54 hunter-gatherer societies, reveals several salient trends: Most objects in our dataset are used in play. Children readily manufacture their own toys, such as dolls and shelters. Most of the objects that children interact with are constructed from multiple materials. Most of the objects in our dataset are full-sized or miniature versions of adult tools, reflecting learning for adult roles. Children also engage with objects related to child culture, primarily during play. Taken together, our findings show that hunter-gatherer children grow up playing, making, and learning with objects.
This introductory chapter provides an overview of the entire content of the book, Play and STEM Education in the Early Years: International Policies and Practices, providing various viewpoints representing 47 STEM experts from 16 countries defining developmental milestones during the early years and the importance of play during this critical developmental period. As the book is divided into four sections: (1) Play as the foundation of STEM experiences during a child’s learning journey; (2) Policies and training for formal education environments; (3) Early years experiences in kindergarten and formal schools for 5–8-year-olds; and (4) Informal settings and family involvement in play, this chapter provides preliminary information and definitions as related to educational development in the early years. The chapter maintains that with a greater understanding of what constitutes play and why it is important, an appreciation and acceptance of the beneficial roles for the child, as well as the adult’s role within the development of play, will emerge. It is posited that sharing international policies and practices promoting STEM subjects in the early years, through unstructured and structured play, will provide exemplary models that parents, citizen scientists, daycare practitioners, primary school teachers and preservice teachers, as well as researchers and policy makers, can employ to design the best learning experiences for the children in their care.
In this paper, we argue that, as HCI becomes more multimodal with the integration of gesture, gaze, posture, and other nonverbal behavior, it is important to understand the role played by affordances and their associated actions in human-object interactions (HOI), so as to facilitate reasoning in HCI and HRI environments. We outline the requirements and challenges involved in developing a multimodal semantics for human-computer and human-robot interactions. Unlike unimodal interactive agents (e.g., text-based chatbots or voice-based personal digital assistants), multimodal HCI and HRI inherently require a notion of embodiment, or an understanding of the agent’s placement within the environment and that of its interlocutor. We present a dynamic semantics of the language, VoxML, to model human-computer, human-robot, and human-human interactions by creating multimodal simulations of both the communicative content and the agents’ common ground, and show the utility of VoxML information that is reified within the environment on computational understanding of objects for HOI.
Young children have a nascent capacity to understand how to control variables to uncover causal structure. However, they do not always explore their environment in a way that generates appropriate data for causal learning. Here, we consider a potential reason that children do not explore systematically: whether the nature of their first action influences their subsequent play. Four- to 7-year-olds (N = 105) saw ambiguous data about a novel causal system and were then allowed to explore this system through play. The nature of their first action – specifically whether it varied only one potential cause and whether it was efficacious – affected their subsequent exploration, particularly whether they generated the data necessary to learn the causal structure. But generating these data, independent of their first action, related to whether they disambiguated the data they had initially observed. These results suggest that the extent to which children can learn from their own exploration in play might depend on how that play unfolds.
Optimism, a bias to overestimate positive and underestimate negative outcomes, may shape how children learn, confront challenges, and overcome setbacks. Although approximately 80% of adults are optimistic, childhood optimism is understudied. A racially and socioeconomically diverse community sample of 152 three- to six-year-old children participated in two experiments (one story-based, one numeric probability-based) that assessed expectations of event outcomes when the likelihood of the outcome occurring either matched or conflicted with the most desirable outcome. The results systematically demonstrate that children are optimistic, even more optimistic for themselves than others, and increasingly integrate probabilistic information into their predictions with age. Differences in optimism were found in children from different socioeconomic backgrounds and those with different levels of depressive symptoms. These findings provide insight into how children reason about the future and elucidate key factors that impact optimistic predictions in childhood.
Humans constantly search for and use information to solve a wide range of problems related to survival, social interactions, and learning. While it is clear that curiosity and the drive for knowledge occupies a central role in defining what being human means to ourselves, where does this desire to know the unknown come from? What is its purpose? And how does it operate? These are some of the core questions this book seeks to answer by showcasing new and exciting research on human information-seeking. The volume brings together perspectives from leading researchers at the cutting edge of the cognitive sciences, working on human brains and behavior within psychology, computer science, and neuroscience. These vital connections between disciplines will continue to lead to further breakthroughs in our understanding of human cognition.
People are sometimes drawn to novel items, but other times prefer familiar ones. In the present research we show, though, that both children’s and adults’ preferences for novel versus familiar items depend on their goals. Across four experiments, we showed 4- to 7-year-olds (total N = 498) and adults (total N = 659) pairs of artifacts where one was familiar and the other was novel (e.g., a four-legged chair and ten-legged chair). In Experiment 1, children wanted to have familiar artifacts, but to learn about novel ones. Experiment 2 replicated this pattern using a simpler procedure, and found the same pattern in adults. In Experiment 3, 4- to 6-year-olds and adults more strongly preferred familiar items when choosing which they would rather have than when choosing which they would rather try using. Finally, Experiment 4 replicated adults’ preferences to have familiar items and learn about novel ones with an additional set of items. Together these findings show that preferences for novelty depend on people’s goals. We suggest these effects arise because children and adults are motivated both by the promise of information and the desire for safe options in high commitment decisions that entail risk.
Full-text available
Abstract The conditional intervention principle is a formal principle that relates patterns of interventions and outcomes,to causal structure. It is a central assumption of the causal Bayes net formalism. Four experiments suggest that preschoolers can use the conditional intervention principle both to learn complex,causal structure from patterns of evidence and to predict patterns of evidence from knowledge,of causal structure. Other theories of causal learning do not account for these results. CONDITIONAL INTERVENTIONS Page 1 Preschool children learn causal structure from conditional interventions By the time children are five years old, they understand a great deal about the causal structure of the world. Research on children’s intuitive theories suggests that preschool children know some of the causal principles of physics, biology and psychology (Flavell, Green, & Flavell, 1995; Gelman & Wellman, 1991; Gopnik & Meltzoff; 1997; Gopnik & Wellman, 1994; Kalish, 1996). Across domains, young children are able to generate appropriate causal predictions, explanations and even
Error is a pervasive and inescapable aspect of empirical science, and it often plays a causal role in experimental outcomes. But little is known about children's understanding of the causes and consequences of experimental error. In this article, we propose a new framework for characterizing experimental error and we use that framework to guide an empirical assessment of elementary school children's understanding of error, their use of theory and evidence in guiding this understanding, and the role of context in reasoning about error. We found that 2nd- and 4th-grade children could both propose and recognize potential sources of error before they could design unconfounded experiments. They used evidence to guide their reasoning, making predictions and drawing conclusions based on the design of their experiments, and they were sensitive to the context of reasoning: They differentiated the role of error in relative and absolute measurements. Long before children have acquired the formal procedures necessary to control error, they have a surprisingly rich-albeit unsystematic-understanding of its various sources.
The purpose of this study was to investigate the exploratory behavior of young children as it relates to individual differences in curiosity, the novelty of the objects explored, and the interactive style employed by an adult experimenter. 48 high- and low-curiosity preschoolers were identified and randomly assigned to 1 of 4 conditions. 3 treatment conditions varied in how responsive and intrusive an adult was during 6 play sessions with novel-perceptual, novel-problem-solving, and conventional toys. In a control condition, only conventional toys were available to the child. A generalization session with 3 novel-perceptual toys not used in previous sessions was conducted for all children following the sixth session. Number of different manipulations, questions asked, and time exploring were recorded during the 7 sessions. High-curiosity children consistently explored more than low-curiosity children in the treatment sessions and in the generalization session but not in the conventional-toy control sessions. Within the treatment sessions, the novel-perceptual toys were explored more than the novel-problem-solving toys, and both types of novel toys tended to be explored more than the conventional toys. The adult-interactive-style conditions had little effect on children's exploration. The results are discussed in terms of the validity of the curiosity battery and the role of different types of novelty in eliciting exploration.
This study investigated the use of covariation as a principle of causal analysis in children of 3-4, 6-7, and 9-11 years of age. The covariation principle specifies that an effect is attributed to the 1 of its possible causes with which it covaries. The results indicated that children as young as 3 years were capable of using covariation information in their attributions of simple physical effects. Children of 3-4 and 6-7 years had greater difficulty identifying inhibitory causes than in identifying facilitatory causes, while 9-11-year-olds identified the 2 types of causes with equal accuracy. 3-year-olds showed a tendency to attribute effects to causes that followed the effects, whereas older children appeared to assume that causes preceded effects.