Children's Understanding of the Knowledge Prerequisites of Drawing and Pretending

Developmental Psychology
Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 38, No. 6, 1004 –1015
0012-1649/02/$5.00
DOI: 10.1037//0012-1649.38.6.1004
Children’s Understanding of the Knowledge Prerequisites
of Drawing and Pretending
Rebekah A. Richert and Angeline S. Lillard
University of Virginia
Many young children will claim that someone is pretending to be something even when the person does
not know what that something is. To examine whether children’s failure to take knowledge prerequisites
into account is part of a more fundamental problem in recognizing how mental representations constrain
external ones, the authors asked children whether an artist who did not know what something was, yet
whose drawing bore resemblance to it, was drawing it. The same questions were asked regarding
pretending. Children performed similarly on pretending and drawing questions, and performance on both
questions improved when the protagonists’ point of view was emphasized. Performance for drawing
improved somewhat when alternative goals were stated. Further, cross-sectional data indicated that
understanding how knowledge relates to producing external representations increases gradually from
age 4 to age 8, suggesting that experiential factors may be crucial to this understanding.
Ever since Wimmer and Perner’s (1983) landmark study involv-
(Beardsley & Wimsatt, 1976) and the idea that there are more
ing the false belief task, there has been considerable interest in the
components to interpretation of art than simply the artist’s inten-
field of cognitive development in how and when children under-
tion (Dobie, 1998). This approach aligns with deconstructionism.
stand mental representations and their links to actions. One of the
In this view, multiple components (the viewer, the image, the
key areas of study and debate has been in the arena of children’s
context, and the author’s intention) allow for multiple interpreta-
understanding of pretending. There is some disagreement in the
tions of the same image. In fact, some philosophers even argue that
literature concerning when the majority of children understand the
the artist’s intention is irrelevant to interpretation altogether (Tol-
mentalistic components of pretense, with estimates ranging any-
hurst, 1979). Thus, the role that mental states actually play in the
where from 3 years old (e.g., Hickling, Wellman, & Gottfried,
interpretation of representational events is not settled. Evidence
1997) to 6 years old (e.g., Lillard, 2001). In fact, the nature of the
from research on children’s understanding of the interpretation of
link between mental representations and external representations
pretense actions and drawings aligns with this particular philo-
has been debated amongst philosophers for some time. The two
sophical debate.
sides of this debate offer helpful descriptors for the nature of
Pretense is a representational act in that the pretense outcome is
children’s responses regarding external representations.
dependent on a specific mental representation. The mental repre-
There are two general schools of thought concerning how men-
sentation is projected onto the self or other aspects of the world
tal representations relate to external representational productions
and determines the pretense outcome (Lillard, 1998). For example,
(like drawings or novels). The traditional romantic interpretation is
to pretend to be a kangaroo, a child has to know what a kangaroo
that the artists’ intentions determine the meaning of the production
is and to think of himself or herself as a kangaroo at the moment
(Taylor, 1998). In other words, a picture is not “of an apple” unless
of pretending (Lillard, 1993a). Pretending is thus governed by the
the artist meant it to be an apple. In analyzing literature, roman-
romanticist view. Adults do not claim that a character is pretending
ticists consider what the author meant in writing it. The other side
to be something unless he or she intends to be that thing, no matter
of this philosophical debate has centered on the intentional fallacy
how much he or she is hopping like it (Lillard, 1999). This
romanticist interpretation of pretense, however, may be something
that develops, according to research by Lillard (1993b). In the Moe
paradigm, a troll from the Land of the Trolls was hopping like a
Rebekah A. Richert and Angeline S. Lillard, Department of Psychology,
kangaroo hops. But he himself did not know that kangaroos
University of Virginia.
hop—in fact, he had never heard of a kangaroo and had never seen
This research was supported by National Science Foundation Grant
DGE-9550152 to Angeline S. Lillard and National Institute of Mental
one either. Children were asked whether Moe was pretending to be
Health Training Grant 5T32MH18242-15 to the University of Virginia. We
a kangaroo. Over many experiments in Lillard’s and other labo-
thank Michael Kubovy for insightful discussion about this line of work. We
ratories, about 35% of 4-year-olds claimed he was not pretending,
are also grateful to the children who participated and their parents and
and the other 65% consistently claimed that he was (see summary
teachers, as well as research assistants and students involved with the
in Lillard, 2001). Young children thus offer deconstructionist
experiments: Ali Zeljo, Stephanie van Patten, Jeanine Dick, Stephanie
responses with regard to pretense, privileging the action outcome,
Curenton, Martin Ho, Andria Hollis, and Brandi Sharpe.
as opposed to the mental state, as crucial to the interpretation of the
Correspondence concerning this article should be addressed to Rebekah
external pretense representation.
A. Richert, Department of Psychology, University of Virginia, 102 Gilmer
Hall, P.O. Box 400400, Charlottesville, Virginia 22904-4400. E-mail:
Although several studies have suggested that young children
rar8z@virginia.edu
may appreciate other mental aspects of pretense (e.g., Aronson &
1004

---

CHILDREN’S KNOWLEDGE OF DRAWING AND PRETENDING
1005
Golomb, 1999; Bruell & Woolley, 1998; Hickling et al., 1997),
time of the drawing to label the pictures at a later point in time,
many young children apparently believe that ignorance does not
despite the similar (objective) appearance of the drawings. This is
prevent one from pretending something. Adults do not share that
better performance than on a similar task in a pretense scenario
belief and instead respond with a romanticist stance that indicates
concerning their own behaviors in a study by Mitchell (2000). He
the importance of the mental state when interviewed on this task
asked 3- to 6-year-old children to reach for some keys, told them
(Lillard, 1999).
they looked like a cat when they reached, and then asked if they
Children’s difficulty with the Moe task can be seen as stemming
had been pretending to be a cat. There was no effect of age overall,
from two possible sources. One possibility is that it is specific to
but only 55% of the children appeared to attend to their own
pretense. Children might think that pretending is solely external
intention to get keys by not accepting that they were pretending to
actions because they learned about pretending before they knew
be a cat. Children thus appear more proficient in respecting their
about thoughts and knowledge (Lillard, 1993b). Children tend to
own intentions regarding drawing than regarding pretense.
be exposed to pretending by the end of their first year of life
Children also appear to use mental representational information
(Lillard, 2001), and many demonstrate their first pretend acts
when naming pictures that someone else has produced. Gelman
before age 2 (e.g., Harris & Kavanaugh, 1993). Such young
and Ebeling (1998) showed 2- and 3-year-olds pictures that were
children have limited understanding of the representational nature
shaped like familiar objects, such as a flower, and told a story
of the mind. In fact, despite the ability to correctly use desire terms
about how each picture was produced. In the intentional condition,
at 2 years of age (Bartsch & Wellman, 1995), children do not seem
the picture was drawn intentionally (e.g., for a teacher), and in the
to understand the connections among desires, intentions, and ac-
accidental condition, it was produced accidentally (e.g., by spilled
tion outcomes until age 3 (Flavell, 1999). Thus, it could be that
paint). Children were asked what each picture was. In the inten-
children initially respond in a deconstructionist manner when
tional condition, 3-year-old children named the apparent object
interpreting pretense, conceptualizing pretense only as action, be-
(flower) on 86% of trials; in the accidental condition, they did so
cause they cannot conceive of the intentions that drive pretense.
on only 44% of trials.
This misunderstanding may persist until middle childhood for
Taken together, these studies suggest that 3-year-olds frequently
many children because it works: Pretense mental representations
consider the artist’s mental state, not just the appearance of the
generally do align with the external ones.
picture, when asked to name a picture. Children might therefore
On the other hand, children’s difficulty with pretense under-
understand the mental representational aspects of drawing earlier
standing might stem from a more general problem with under-
than those of pretense, or, in other words, be romanticist in their
standing how knowledge representations are related to a variety of
understanding of drawings.
external representational outcomes. If so, children might well tend
In summary, past research has suggested that children have a
toward deconstructionist responses when interpreting other repre-
romanticist interpretation of drawings by age 3; they may later take
sentational actions as well. As was stated above, pretense is
romanticist or deconstructionist viewpoints, as adults do. For pre-
considered a representational action because a mental representa-
tense, the evidence has suggested a different pattern: Adults clearly
tion is projected onto some aspect of the world, and that mental
take a romanticist stance, whereas children initially tend toward a
representation determines the pretense outcome (Lillard, 1998).
deconstructionist interpretation, often responding on the basis of
Drawing is also a representational act: An artist can create a
external appearance rather than the actor’s a priori mental repre-
drawing of what he or she knows about, and is projecting onto, the
sentation. By about age 8, most children take the romanticist view
drawing. In drawing an apple, one must know what an apple is or
of pretense.
looks like in order to project that representation onto the paper.
Past research on drawing suggests then that children’s decon-
In contrast to pretense, adults can view drawings in both roman-
structionist understanding of pretense is specific to pretense. This
ticist and deconstructionist manners. On the one hand, adults might
statement cannot be made definitively, however, because pretense
say that no matter how much something looks like an apple, it is
and drawing have not been compared in parallel tasks. In addition,
not one unless the artist meant to draw one. On the other hand, one
children in the drawing and pretense studies were of different ages.
might imagine the work coming to be referred to as “The Apple
It may be that children’s understanding of drawings initially re-
Drawing” despite the artist’s lack of intention. Indeed, the Little
flects romanticist interpretations, but then becomes deconstruc-
Prince (de Saint-Exupe´ry, 1943) had just such a problem when he
tionist for a time. In the present study, we sought to determine
tried to draw a cobra who had eaten an elephant: The adults in his
whether children’s deconstructionist tendencies in pretense—the
midst repeatedly insisted that he had drawn a hat. The appearance
tendency to deny the importance of the protagonists’ mental rep-
of the drawing, not the boy’s intention, was considered paramount
resentation in judging representational outcomes—applies simul-
to determining what the drawing was.
taneously to drawing when the circumstances leading to the two
Two studies have suggested that young children tend toward
representational outcomes are the same. Because the two drawing
romanticist responses, not deconstructionist ones, in their interpre-
studies reviewed above (Bloom & Markson, 1998; Gelman &
tation of drawings. Bloom and Markson (1998) asked 3- and
Ebeling, 1998) were conducted with children younger than those
4-year-olds to draw a lollipop and a balloon, which resulted in two
given the Moe tasks, and because the Moe task may not be
quite similar drawings of a straight line with a circle on top. From
appropriate for younger children (because they may not understand
a purely objective standpoint, each drawing looked no more like a
very much about knowing), Experiment 1 examined a large cross-
lollipop than like a balloon. When later asked to identify which
section of children, from 4 to 8 years of age. This considerably
figure was which, 3-year-olds correctly identified 76% of the
extends prior work on the Moe pretense understanding task.
pictures and 4-year-olds correctly identified 87% of the pictures.
A third extension of Experiment 1 was to examine whether
Children apparently used their own mental representation at the
children’s performance improved when the question about the

---

1006
RICHERT AND LILLARD
representational outcome was rephrased to emphasize the protag-
Procedure
onist’s point of view. An adult might claim that a certain drawing
was “of an apple” even if the artist did not know what an apple
All children received four trials, two pretense and two draw. Trials were
blocked by condition (pretense or draw) and half of the children in each age
was; drawings, once complete, are permitted objective interpreta-
group received each block first.
tion. But when asked if, to the artist, the apple-like drawing was an
Pretense condition.
The experimenter introduced the children to a troll
apple, adults would be expected to understand it was not. For this
doll named Moe and told the following story: “This is Moe. Moe is from
reason children were asked, in addition to what Moe was drawing
a far away place called the Land of Trolls. The Land of Trolls is very far
(or pretending), what Moe thought he was drawing (or pretending).
away and very different from earth. There are no animals there, only trolls
Children were also asked what Moe would say he was drawing or
and trees.” The experimenter moved Moe up and down as if he was
pretending. Different responses to the second and third questions
hopping. Continuing to move the troll up and down, the experimenter said,
“Moe is moving like a rabbit. Rabbits move like that, but Moe doesn’t
were not expected, because other work shows that say and think
know that rabbits move like that. Moe’s never seen a rabbit, and he’s never
questions do not elicit different responses (Wellman, Cross, &
heard of one either. So Moe doesn’t know anything about rabbits. But Moe
Watson, 2001), but asking the additional question ensured children
is moving like a rabbit. That’s just how rabbits move.”
would have a second opportunity to reveal understanding if they
While the experimenter was still moving the troll doll, control questions
had it.
were asked to verify the child’s understanding of Moe’s action and knowl-
edge. The control questions were, “Is Moe moving like a rabbit?” and
“Does Moe know that rabbits move like that?” These were asked in
Experiment 1
counterbalanced order across and within children. Children who answered
a control question incorrectly were reminded of the premises and asked the
As just discussed, Experiment 1 had three purposes: to examine
question again. Finally, the child was asked the experimental questions, “Is
whether children’s difficulty in recognizing Moe’s inability to
Moe pretending to be a rabbit?” (hereafter referred to as the pretend-
be pretending, on the basis of his lack of knowledge, was specific
objective question), “Does Moe think he is pretending to be a rabbit?”
to the pretense scenario or extended to other representational
(pretend-think, or the subjective pretense question), and “What if you
asked Moe what he was doing, what would he say?” (pretend-say). The
actions (i.e., drawing); to examine responses to the Moe test across
second pretend task was the same, but Moe was moving like a snake.
a wide age range; and to examine whether specifying the perspec-
Drawing condition.
In the draw condition, children were introduced to
tive from which the question should be answered would improve
a troll doll named Luna and were told a story that closely paralleled that of
performance.
the pretense condition. They were told, “This is Luna. Luna is from a far
away place called the Land of Trolls. The Land of Trolls is very far away
and very different from earth. There are no animals there, only trolls and
Method
trees.” Children were then shown Luna drawing something that looked like
a fish. Children were reminded that Luna had never seen or heard of a fish
Participants
and did not know what fish looked like. With Luna still drawing, control
questions were asked to verify children’s understanding of the appearance
Participants were 88 children recruited from public and private pre-
of the drawing (“Does this look like a fish?”) and Luna’s knowledge
schools and elementary schools in a small metropolitan area. They were
(“Does Luna know that fish look like that?”). These questions were
divided into five groups: 4-year-olds (n
16; mean age
4 years 7
counterbalanced across and within participants. Finally, they were asked,
months; range
4 years 3 months to 4 years 11 months), 5-year-olds (n
“Is Luna drawing a fish?” (draw-objective), “Does Luna think she is
17; mean age
5 years 9 months; range
5 years 3 months to 5 years 11
drawing a fish?” (draw-think), and “What if you asked Luna what she was
months), 6-year-olds (n
18; mean age
6 years 8 months; range
6
doing, what would she say?” (draw-say). The second draw task concerned
years 2 months to 6 years 11 months), 7-year-olds (n
19; mean age
7
a worm-like drawing.
years 5 months; range
7 years 0 months to 7 years 11 months), and
8-year-olds (n
18; mean age
8 years 8 months; range
8 years 1
month to 9 years 2 months). Forty-six of the children were girls and 42
Results and Discussion
were boys; there were approximately equal numbers of boys and girls in
Children responded correctly on 330 of 336 control questions;
each age group. The majority of the participants were Caucasian, and all
all six incorrect responses were correctly revised once children
children were from middle-class families.
were reminded of the premises. Responses for the experimental
A female experimenter visited the schools and explained to the classes
what research is and that she needed their help with some research. Parental
questions were scored as 1 if correct and 0 if incorrect. Because
consent forms were sent home with the children. The experimenter re-
pretend-say and draw-say were open-ended questions, responses
turned approximately a week later for a day at the school. Those children
were coded as incorrect if the child said the character would
who returned signed consent forms were interviewed one at a time for
mention the animal (“[Moe] would say he is being like a rabbit”)
approximately 10 min during the school day. All children were interviewed
and correct otherwise. Responses for each question type were
in a small room or quiet hallway of the school. Each child was asked not
summed for a total score ranging from 0 to 2 for each type of
to talk about the interview until all children had a turn to participate. The
question.
experimenter obtained verbal consent from children to tape the interview.
The numbers of children responding correctly to zero, one, or
One child did not want to be taped, so his responses were recorded
two of the questions for each of the variables are listed in Table 1.
manually.
Spearman correlational analysis revealed that responses to the
second and third questions (think and say) were highly correlated
for both pretend (r
.78, p
.01) and draw (r
.67, p
.01).
Materials
Because the say questions did not provide any additional informa-
Two troll dolls, a 5-in. doll with blue hair and an 8-in. doll with yellow
tion beyond that provided by the think ones, they were dropped
hair, and paper and pencils for drawing were used.
from the analysis. First, chi-square goodness of fit tests were

---

CHILDREN’S KNOWLEDGE OF DRAWING AND PRETENDING
1007
Table 1
Number and Percentages of Children in Each Group Responding Correctly on Zero, One, or
Two Questions, and Chi-Square Goodness of Fit Values, for Variables in Experiment 1
Condition
P-O
P-T
D-O
D-T
P-S
D-S
Age group, no. of
questions, and 2 values
n
%
n
%
n
%
n
%
n
%
n
%
4-year-olds
Zero
12
75
9
56
12
75
8
50
8
50
9
56
One
1
6
2
13
2
13
0
0
3
19
1
6
Two
3
19
5
31
2
13
8
50
5
31
6
38
2(2, N
16)
22.38**
11.00*
21.50**
16.00**
5-year-olds
Zero
11
65
6
35
15
88
7
41
4
24
7
41
One
0
0
3
18
2
12
1
6
5
31
3
18
Two
6
35
8
47
0
0
9
53
8
47
7
41
2(2, N
17)
18.86**
6.93†
34.87**
12.81*
6-year-olds
Zero
10
56
4
22
17
94
5
29
2
11
5
29
One
0
0
1
6
1
6
2
11
2
11
0
0
Two
8
44
13
72
0
0
11
61
14
78
13
72
2(2, N
18)
18.44**
23.22**
46.33**
14.89*
7-year-olds
Zero
6
32
3
16
18
95
3
16
4
21
4
21
One
1
5
3
16
1
5
0
0
2
11
1
5
Two
12
63
13
68
0
0
16
84
13
68
14
74
2(2, N
19)
19.00**
19.42**
49.32**
36.79**
8-year-olds
Zero
2
11
0
0
10
56
1
6
0
0
1
6
One
1
6
0
0
4
22
0
0
1
6
0
0
Two
15
83
18
100
4
22
17
94
17
94
17
94
2(2, N
18)
33.00**
54.00**
9.56*
46.44**
Note.
P-O
pretend-objective; P-T
pretend-think; D-O
draw-objective; D-T
draw-think; P-S
pretend-say; D-S
draw-say.
† p
.05.
* p
.01.
** p
.001.
conducted to test for chance responding on the objective and think
there was a gradual increase in mean responses of children in each
questions. All responses were significantly different from chance
age group to the objective pretense questions, with most children
(see Table 1).
in the 8-year-old group stating that Moe was not pretending to be
Figure 1 shows the mean responses (out of two) of children in
a rabbit or snake; answers to the subjective pretense questions
each age group for the objective pretend questions. As can be seen,
were slightly elevated and followed the same trend.
As indicated in Figure 2, the developmental trend was absent for
the objective draw question, with virtually none of the children
giving romanticist answers and instead stating that Luna was
drawing that animal. Children’s performance on the draw-think
question, which assessed the ability to take a romanticist stance,
was similar to their performance on the parallel pretend question.
Possible reasons for the discrepancy in children’s responses to the
two drawing questions are discussed later.
Wilcoxon matched-pairs signed ranks tests were first conducted
by collapsing across age groups to determine if there was a general
difference between the responses for subjective and objective
questions and between drawing and pretending questions. Overall,
the differences between the subjective and objective questions
were significant for both drawing (Z
6.92, p
.01) and
pretending (Z
3.71, p
.01). In addition, there were significant
differences between responses for the pretending and drawing
questions for the objective questions (Z
5.47, p
.01) but not
for the think questions. The significant difference in the objective
Figure 1.
Mean response (out of 2) and standard error of children in
each age group for pretend-objective and pretend-think questions for
questions was apparently due to the generally poor performance of
Experiment 1.
children on the objective draw question.

---

1008
RICHERT AND LILLARD
p
.05), draw-objective and draw-think (Z
2.92, p
.01), and
draw-objective and pretend-objective (Z
2.20, p
.05). The
difference between draw-think and pretend-think was not signifi-
cant. This indicates that 5-year-olds responded better on the sub-
jective than the objective questions in general. In comparing the
subjective questions, we found that 5-year-olds performed simi-
larly for the drawing and pretending questions.
The same pattern of significant differences emerged for the
6-year-olds. The differences between pretend-objective and
pretend-think (Z
1.96, p
.05), draw-objective and draw-think
(Z
3.36, p
.01) and pretend-objective and draw-objective
(Z
2.76, p
.01) were again significant, but the difference
between draw-think and pretend-think remained nonsignificant.
This suggests the same pattern was true for the 6-year-olds as for
the 5-year-olds: better performance on the subjective questions and
similar performance on drawing and pretending when comparing
the subjective questions.
Figure 2.
Mean response (out of 2) and standard error of children in each
The pattern of significant differences changed for the 7- and
age group for draw-objective and draw-think questions for Experiment 1.
8-year-olds and appears to reflect the poor performance for the
draw-objective question. More specifically, for the 7- and 8-year-
olds, respectively, the differences between pretend-objective and
To analyze children’s responses across age groups, we con-
pretend-think and pretend-think and draw-think were not signifi-
ducted a Kruskal-Wallis chi-square test. This revealed significant
cant. However, the differences between draw-objective and
differences on pretend-objective, 2(4, N
88)
18.35, p
.01;
draw-think (Z
3.90, p
.01 and Z
3.24, p
.01, for 7-
pretend-think,
2(4, N
88)
20.66, p
.01; draw-objective,
and 8-year-olds, respectively) and draw-objective and pretend-
2(4, N
88)
14.18, p
.01; and draw-think,
2(4, N
objective (Z
3.42, p
.01 and Z
3.07, p
.01, for 7- and
88)
12.52, p
.05. Mann–Whitney U tests of specific differ-
8-year-olds, respectively) were significant. This indicates that in
ences were conducted between the 8-year-olds and the other age
general, the 7- and 8-year-olds responded similarly for the subjec-
groups to test for a developmental trend. The 4-year-olds were
tive and objective questions in the pretend condition, and similarly
significantly different from the 8-year-olds on the pretend-
on the subjective drawing and pretending questions, indicating that
objective (U
46.00, p
.01), pretend-think (U
45.00, p
the understanding of drawing and pretending is fairly sophisticated
.01), and draw-think (U
80.00, p
.05) questions. The 5-year-
by 7 years of age. However, performance on the draw-objective
olds were significantly different from the 8-year-olds on the same
question is quite poor if one has a romanticist interpretation of
questions (U
74.00, 72.00, and 90.00, p
.01, .01, and .05,
drawing.
respectively). The 6-year-olds were only different from the 8-year-
olds on the pretend-objective (U
94.00, p
.05) and draw-
Summary
objective (U
97.00, p
.05) questions. The 7- and 8-year-olds
were significantly different only on the draw-objective question
To review, this experiment had three aims. The first aim was to
(U
102.00, p
.05). Thus, except for responses to the draw-
compare children’s responses on a Moe test concerning pretense
objective question, significant differences disappeared with age,
with their responses on a parallel test concerning another repre-
suggesting a developmental trend in ability to respond correctly.
sentational act, that of drawing. The second was to do a full-scale
Also important to note, as shown in Table 1, is that few children
cross-sectional study of children’s responses to the Moe procedure
at any age received a score of 1. In other words, children were
to determine percentage passing at each age, and the third aim was
consistent in their responding. A score of 1 could indicate either
to examine whether children’s responses would change when
confusion and guessing or a transition phase between response
asked to answer the questions from Moe’s perspective.
strategies. The consistency of responding indicates that over the
Results clarified the developmental course of children’s ability
course of development, children shift from responding in com-
to pass the Moe task: Only about 20% passed at age 4 and this
pletely deconstructionist ways, emphasizing external appearance,
percentage gradually increased until age 8, when 85% of children
to completely romanticist ways, emphasizing underlying mental
passed. When subjectivity was highlighted by asking what Moe or
states.
Luna would think, performance improved: 31% of 4-year-olds and
Wilcoxon matched-pairs signed ranks tests within each age
100% of 8-year-olds answered correctly on both pretend-think
group were conducted for question type within each condition
questions. This effect was even stronger for draw: Only 12% of
(draw-objective vs. draw-think and pretend-objective vs. pretend-
4-year-olds responded correctly to both draw-objective questions,
think) and for condition by question (draw-objective vs. pretend-
compared to 50% who answered correctly on both of the draw-
objective and draw-think vs. pretend-think). There were no signif-
think questions. In addition, only 22% of 8-year-olds answered
icant differences in the 4-year-olds’ responses for any of the
correctly on both draw-objective questions, but 94% responded
questions, indicating that performance was poor in general for all
correctly on both draw-think questions.
questions. For the 5-year-olds, there were significant differences
In regard to the first aim, there were no significant differences
for responses on pretend-objective and pretend-think (Z
2.07,
at any age between responses on the pretend-think and draw-think

---

CHILDREN’S KNOWLEDGE OF DRAWING AND PRETENDING
1009
questions, suggesting that development in understanding the
condition, children named the picture 44% of the time and referred
knowledge requirements for these representational actions devel-
to the literal contents 48% of the time, a nonsignificant difference.
ops in parallel. The expectation was that children would under-
Thus, when a picture was produced accidentally, children identi-
stand the knowledge prerequisites for drawings sooner than they
fied the picture essentially equally by the literal contents or the
would demonstrate such an understanding for pretense, given their
appearance, and thus it becomes unclear from this study whether
seeming understanding of the mental underpinnings of drawings in
we should expect children to respond in a deconstructionist or
other studies (Bloom & Markson, 1998; Gelman & Ebeling, 1998).
romanticist manner on our task. Furthermore, adults named the
The implications of this particular finding are discussed in more
accidental picture about 70% of the time, suggesting people be-
detail below.
come more deconstructionist about drawings with age. Comparing
First, however, it is important to address the fact that children
the percentage of 4-year-old children correctly responding to our
did not perform as well as expected on the objective draw question.
draw-think question (50%) to the percentage of children correctly
One possible reason for this is that children might have interpreted
responding to Gelman and Ebeling’s accidental condition (48%)
the question, “Is Luna drawing a fish/worm?” as “Does Luna’s
reveals virtually no difference.
drawing look like a fish/worm?” If children were answering the
A third important difference between our study and the Bloom
question in this way, then the test was not about their understand-
and Markson (1998) and Gelman and Ebeling (1998) studies is that
ing of the implications of Luna’s knowledge, but rather it was
in those studies, the artist’s intention was likely highlighted more
about their own interpretation of the appearance of the drawing. In
strongly than in our scenario. Experiment 2 aimed to test whether
other words, perhaps the question begged for a deconstructionist
children’s performance might improve further when the protago-
response. Certainly the drawings did look like fish and worms, and
nist’s intention was emphasized within the story itself. One way to
perhaps asking children “Is this a fish/worm” instead of “What is
emphasize intention is to have the child perform the task himself
this?” (as in Gelman & Ebeling, 1998) emphasized the appearance
or herself, as in Bloom and Markson’s (1998) study. Another is to
too strongly for children to consider the role of the (inconsistent)
state the reason why an act is performed, as in the Gelman and
mental state. The issue of different interpretations for the draw-
Ebeling (1998) study, in which artists’ purpose in making the
objective and draw-think questions is addressed more specifically
picture was explicitly stated. For Experiment 2, we chose also to
in Experiment 3. Given the ambiguity of this question in Experi-
use the Gelman and Ebeling approach by emphasizing reasons, in
ment 1, the clearest test of the development of children’s under-
the form of explanations, for the protagonist’s actions. For pretend,
standing of the role of mental states is their responses on the
this change also served to address a concern that children may
draw-think questions, which emphasized the mental state. Even on
answer incorrectly on the Moe task because they cannot think of
this question, however, children did not perform as well as
any other reason for his hopping except that he must be pretending
expected.
(Aronson & Golomb, 1999; Custer, 1996; Hickling et al., 1997;
We had expected that children would respond on the basis of the
Joseph, 1998).
character’s mental states, given prior findings suggesting that
children initially have a romanticist approach to drawings (Bloom
& Markson, 1998; Gelman & Ebeling, 1998). Closer comparisons
Experiment 2
between the studies, however, revealed that our findings are not
necessarily inconsistent with previous research. There are three
In Experiment 2 we examined the effects of offering children an
important differences to consider. First, the Bloom and Markson
alternative explanation (in effect, an intention) for the pretense and
(1998) study asked children about their own drawings. Other
drawing actions. Other drawing studies have highlighted intention,
research with pretense has suggested that children perform better
and children have shown more romanticist interpretations (Bloom
on tasks when asked about themselves, as opposed to another
& Markson, 1998; Gelman & Ebeling, 1998). Other pretense
(Mitchell, 2000). Because our study asked children about the
studies have also shown that children perform better when the
drawings that another character produced, it may have been more
protagonist’s intention is clear (Ganea, Lillard, & Turkheimer,
difficult for children to take into account the underlying mental
2002; Joseph, 1998). In fact, regarding the basic Moe procedure,
state.
the criticism has been made that although children answered the
Second, in Gelman and Ebeling’s (1998) study, children were
control questions correctly (e.g., agree that Moe is hopping like a
told that the line drawings were made either intentionally or
rabbit, and that Moe does not know that rabbits hop), they may
accidentally. Children’s strong tendency to provide names for the
have ignored that information in deciding that Moe was pretending
pictures produced intentionally (as opposed to referring to the
to be a rabbit (Aronson & Golomb, 1999). If this criticism were
literal contents of the drawing, like paint) was taken as evidence
correct, then perhaps if children were given some other logical
that children understood the mental underpinnings of drawing.
reason why Moe is hopping and Luna is drawing, they would avoid
However, there was also a parallel condition where the pictures
the error. To this end, in Experiment 2 children were told why Moe
were produced accidentally (e.g., by spilling paint). This acciden-
and Luna were doing what they were doing, for example, that Moe
tal condition is actually the closest condition to the scenario given
was hopping to avoid burning his feet on the very hot pavement
to children in this experiment. It would be unlikely that Luna
and that Luna wanted to draw a red balloon. Care was taken to
would intentionally produce a picture resembling an animal if she
make the pretend and draw conditions as similar as possible.
did not know what an animal was. In other words, if Luna does not
Children were then asked the same objective questions as were
know about animals and yet is drawing pictures of animals, the
asked in Experiment 1 to test whether emphasizing the mental state
similarity between the picture and a particular animal likely oc-
in the story would serve the same function as emphasizing the
curred by accident. In Gelman and Ebeling’s (1998) accidental
mental state in the test question.

---

1010
RICHERT AND LILLARD
For half of the trials, a behavior was shown and described with
pretending to be a monkey?” The no-explanation tasks were exactly the
its explanation in terms of the protagonist’s intention. Following
same but omitted the first two sentences, and the explanation question
the explanation, the resemblance of the representational outcome
(“Why was Moe climbing the tree?”) was asked after the test question.
of that behavior to something else was noted, along with the
In addition to the monkey climbing task, children in the pretend condi-
tion were given a rabbit hopping (explanation: hot pavement), a dog
protagonist’s ignorance of that something else. Then the same
digging (explanation: to find treasure), and a pig rolling (explanation: to
objective questions were asked as in Experiment 1, but a third
itch his back) task. Each task was given to half of the children without an
control question regarding the protagonist’s intention was asked.
explanation. The two explanation and the two no-explanation tasks were
Each child participated in both the explanation and the no-
blocked; half of the children received each type of task first.
explanation conditions, but drawing and pretend were between-
Draw condition.
The draw preamble was also similar to that of Ex-
subjects conditions because presenting all possible variations to all
periment 1, except in this case children were told there were no lollipops,
children created too long a procedure. Only preschoolers were
worms, boats, or cats. After the preamble, the experimenter held a pencil
tested because they are of focal interest regarding the issue at hand.
by the troll’s hand and manipulated the troll to draw a picture of a circle
with a straight line at the bottom. The experimenter then began to color in
the circle and told the child, “This is Luna’s picture. She wants to make a
Method
red balloon, so she is coloring it red (explanation). Her picture looks like
a lollipop. Lollipops look like that. Luna doesn’t know what lollipops look
Participants
like. Luna doesn’t know anything about lollipops.” The child was then
Thirty-three children ranging in age from 3 years 3 months to 5 years 4
asked three control questions: “Why is Luna coloring red?”, “Does Luna’s
months (mean age
4 years 6 months) participated in a between-subjects
picture look like a lollipop?”, and “Does Luna know that lollipops look like
design, with 17 children (mean age
4 years 6 months; range
4 years 0
that?” If children responded incorrectly to either control question (11 of
months to 5 years 0 months) in the pretend condition and 16 children (mean
128 questions), they were reminded of the premises. Once children cor-
age
4 years 6 months; range
3 years 3 months to 5 years 4 months)
rectly responded to the control questions, the experimental question was
in the draw condition. One child in the pretend condition was dropped
asked, “Is Luna drawing a lollipop?” The explanation control question
because it became apparent during the procedure that he was not fluent in
(“Why is Luna coloring red?”) was asked before the other control questions
English. In the final sample, there were 10 boys and 6 girls in the pretense
in the explanation condition and after the experimental question in the
condition and 10 girls and 6 boys in the draw condition. Children were
no-explanation condition. In the no-explanation condition, the explanation
mostly from White, middle-class families and were recruited from pre-
sentence (“She wants to make a red balloon, so she is coloring it red.”) was
schools and day-care centers in a small U.S. city. All but 2 children gave
removed from the story.
verbal consent to having their interviews taped. Responses were recorded
Children in the draw condition were told four such stories. The other
manually for those children who did not want to be taped.
stories were about a snake that also looked like a worm, a watermelon that
also looked like a boat, and a dog that also looked like a cat. The
explanations for Luna’s drawings were to draw (a) stripes on the snake, (b)
Materials
seeds on the watermelon, and (c) whiskers on the dog. Each task was given
to half of the children without an explanation, and half of the children
The same two troll dolls as in Experiment 1, a 5-in. doll with blue hair
received each type of task first. As with the pretense condition, two
and an 8-in. doll with yellow hair, were used. In addition, a cardboard
explanation and two no-explanation tasks were blocked, and half of the
replica of a tree with apples was used as materials for the pretense
children received each type of task first.
condition. For the draw condition, materials were a piece of plain white
paper and colored pencils used to draw the pictures.
Results and Discussion
Procedure
Children participated in either a draw or pretense condition.
Children were tested individually in a small room by a trained experi-
Within each condition they had two trials of each type (explanation
menter. Once children were comfortable, the experimenter began the
or no explanation) and were scored 1 for each correct answer to a
interview.
test question. Scores for each variable ranged from 0 to 2, with 0
Pretense condition.
The preamble for the pretense condition was sim-
indicating incorrect responses on both questions and 2 indicating
ilar to that of Experiment 1, except in this case children were told there
correct responses on both questions. Table 2 shows the number of
were no monkeys, rabbits, dogs, or pigs in the Land of Trolls. Four pretend
children responding correctly on zero, one, or two of the explana-
tasks were administered, followed by a series of follow-up questions. The
tasks were of two different types: explanation and no explanation. Each
tion and no-explanation questions by condition. Chi-square good-
child received two of each type, blocked. As an example of a task in which
ness of fit tests were conducted to test for chance responding. All
an explanation was provided, children were shown the cardboard tree and
responses were significantly different from chance (see Table 2).
were told, “This tree has apples. Moe wants to pick an apple so he’s
Figure 3 shows the mean number of correct responses for
climbing the tree. [Moe was made to climb.] Moe looks like a monkey.
children in each condition. For the pretense condition, the overall
Monkeys climb like that. Moe doesn’t know that monkeys climb like that.
mean correct was 0.50 out of 4. The means for the explanation and
Moe doesn’t know anything about monkeys.” For one item of each type,
no-explanation conditions were both 0.25 out of 2. For the draw
the order of the looks like and know information was reversed, with Moe’s
condition, the overall mean was 1.38 out of 4. The means for the
knowledge about the animal stated first. Children were then asked control
explanation and no-explanation conditions were 0.75 and 0.63 out
questions: “Why was Moe climbing the tree?”, “Does Moe know what a
of 2, respectively. Mann–Whitney U tests conducted to compare
monkey is?”, and “Does Moe look like a monkey?” (with the latter two
asked in counterbalanced order). When children incorrectly answered a
the responses (combining explanation and no-explanation condi-
control question about Moe’s knowledge or appearance (15 of 128 times,
tions) for the drawing and pretending conditions revealed a non-
combining conditions), the premises were repeated and the questions were
significant trend in better responding on the draw questions
asked again. The experimenter next asked the test question, “Is Moe
(U
87.00, p
.07). There was also a trend toward better

---

CHILDREN’S KNOWLEDGE OF DRAWING AND PRETENDING
1011
Table 2
kers”). Children were correct on 38% of these trials, as compared
Number and Percentage of Children Responding Correctly on
with 37% of trials for which they supplied a different explanation
Zero, One, or Two Questions, and Chi-Square Goodness of Fit
(e.g., “because cats have whiskers” and “some lollipops are red”).
Values, by Condition in Experiment 2
These percentages reflect that children were in general better at the
drawing task, but they were not significantly different from each
Explanation
No explanation
other, suggesting that simply registering the explanations may not
Condition, no. of
actually influence children to take a romanticist view of drawings.
questions, and 2 values
n
%
n
%
In addition, the 38% of children in the draw condition who
Pretend
responded correctly to both the “why” question and the test ques-
Zero
13
81
13
81
tion was not significantly different from the 25% of children in the
One
2
13
2
13
comparable group in the pretend condition. This suggests that
Two
1
6
1
6
2
understanding the intention of the action assists children in their
(2, N
16)
27.00**
27.00**
Draw
interpretation of both drawing and pretense outcomes.
Zero
8
50
9
56
In summary, this experiment provided children with explana-
One
4
25
4
25
tions for the troll’s behavior on half of their trials, to see whether
Two
4
25
3
19
2
children’s performance would improve when the troll’s intention
(2, N
16)
6.00*
8.50*
was made clear. Regardless of whether an alternative explanation
* p
.05.
** p
.001.
for the resembling action was provided, and even when children
correctly recalled that explanation, most young children still did
very poorly on pretense trials, claiming that the doll was pretend-
responding in the draw condition when an explanation was pro-
ing to be something that he did not know anything about. There
vided (U
86.00, p
.06). A nonsignificant trend for an order
were, however, nonsignificant trends toward better responding in
effect was also observed, toward better responding when children
the drawing condition and the explanation first condition. The best
received an explanation in the first story they heard (U
86.00,
overall performance, 44% correct, was obtained for the drawing
p
.06).
condition when an explanation was provided for the first two tasks
In this experiment, we attempted to make the procedure easier
presented. These findings are again consistent with those of
for 4-year-old children by including an alternative explanation for
Gelman and Ebeling (1998), who found that 3-year-olds correctly
the actions, thereby emphasizing intention. This manipulation
identified pictures on the basis of their literal components on 48%
appeared to assist children in formulating a romanticist interpre-
of trials when the picture resulted from an accidental spilling of
tation of the drawings. Children persisted in below chance re-
paint. Thus, providing the explanation in the initial story influ-
sponding for both drawing and pretending, however, continuing to
enced children to be somewhat less deconstructionist in their
respond in a deconstructionist manner about drawing and pretense
responses about the identity of a drawing. This was not a signif-
by claiming the external appearance constituted the identity of the
icant result, however, and Experiment 3 attempted to further
representation.
clarify this finding and extend the age range upwards.
Crucial to the interpretation of this experiment is whether chil-
dren actually heeded the explanations we provided. Children’s
Experiment 3
responses to the “why” question suggest that they did register the
explanations provided. In the pretense condition, when asked why
In Experiment 2, we had expected that emphasizing the inten-
Moe was engaged in his behavior, 16 of 32 replies stated the given
tion of the actor would lead to more romanticist interpretations of
reason, for example, that he was climbing the tree in order to get
apples. Of the remaining responses, on 7 (of the original 32)
occasions children appeared to overwrite the given reason and
apply one consistent with pretense (e.g., he was climbing in order
to pretend to be a monkey). This offers limited support to Aronson
and Golomb’s (1999) claim of this problem for the Moe-pretense
procedure. Given this, one must ask if children performed better
when they heeded the explanation than when they overwrote it.
They apparently did: When they bore in mind the given explana-
tion for his behavior, performance was 25% correct, and when they
gave a different explanation no children responded correctly. This
difference was significant (Z
3.46, p
.01), suggesting that
having a clear idea of the character’s nonpretense intention seems
to influence children’s interpretations of the character’s behavior,
perhaps serving the same function as emphasizing the mental state
in the experimental question. However, performance was still very
poor.
In the draw condition, 13 of the 32 responses to the why
Figure 3.
Overall mean response and standard error of children on the
question in the explanation condition concerned the explanation
pretend and draw questions in the explanation and no explanation condi-
given (e.g., “because it’s a watermelon” and “dogs have whis-
tions for Experiment 2.

---

1012
RICHERT AND LILLARD
drawings. There was only very limited support for that possibility.
The goals of Experiment 3 were to see if results could be clarified
by including an older group of children and by providing children
with an even more straightforward procedure. The procedure was
clarified in three ways. First, as in Experiment 1, children received
subjective questions that highlighted the mental state of the troll
(e.g., “Would Moe say he is pretending to be a monkey?”). The
same objective questions (e.g., “Is Luna drawing a cat?”) were
included as well for comparison. Second, children were shown
pictures of the trolls either drawing or pretending (see Figures 4
and 5 for examples), instead of the experimenter holding the dolls
and performing the movements. This was to correct for the pos-
sibility that some children were attributing an intention to the
experimenter when they named the drawing. Third, children were
always provided with an alternative explanation for the trolls’
actions or drawings, because the findings of Experiment 2 sug-
gested this may make the task easier for children. Both 4- and
5-year-olds were interviewed in this study to further probe the
Figure 5.
Sample stimuli for Experiment 3: Luna drawing a snake.
possible age trends for the preschoolers revealed in Experiment 1.
Method
children were comfortable. All children were told four stories (two pretense
and two draw) with the help of pictures. At the end of the story, they were
Participants
asked questions about the pictures. Stories were blocked by the pretense
Sixty children were recruited from day-care centers, preschools, and
and draw conditions, and which type children received first was counter-
elementary schools in a small U.S. city. Children were from White,
balanced across participants.
middle-class families and were separated into two groups by age. There
Pretense condition.
The preamble for the pretense condition was sim-
were 30 children (18 girls and 12 boys) in the 4-year-old group (mean
ilar to that of Experiment 2. Children were told that Moe was from the
age
4 years 4 months; range
3 years 8 months to 4 years 11 months)
Land of Trolls where there are no monkeys and rabbits. They were then
and 30 children (13 girls and 17 boys) in the 5-year-old group (mean
told the monkey and rabbit stories from the explanation condition of
age
5 years 7 months; range
5 years 0 months to 6 years 6 months).
Experiment 2. For example, children were shown the picture in Figure 4
and told, “This tree has apples. Moe wants to pick an apple, so he’s
climbing the tree. Moe looks like a monkey. Monkeys climb like that. Moe
Materials
doesn’t know that monkeys climb like that. Moe doesn’t know anything
Materials were four pictures, one each of a troll hopping and a troll
about monkeys.” The experimenter counterbalanced whether children were
climbing a tree and two pictures of trolls drawing. Examples of the pictures
given the information about the action or Moe’s knowledge first. Following
are displayed in Figures 4 and 5.
the story, children were asked two experimental questions: (a) “Is Moe
pretending to be a monkey?” (pretend-objective) and (b) “Would Moe say
he is pretending to be a monkey?” (pretend-say). For the other pretend
Procedure
story, Moe was hopping so as not to burn his feet on hot pavement, and he
Children were tested individually in a small room or quiet hallway of the
looked like a rabbit.
school by an experimenter. The experimenter began the interview after
Draw condition.
The preamble for the draw condition was similar to
the preamble for the draw condition in Experiment 2, in which children
were told that Luna was from the Land of Trolls, where there are no worms
and cats. They were then told the worm and cat stories from the explanation
condition of Experiment 2. For example, in the worm story children were
shown the picture in Figure 5 and told the following story, “This is Luna’s
picture. She wants to make a striped snake, so she is drawing stripes. Luna
doesn’t know what worms look like. Luna doesn’t know anything about
worms. Her picture looks like a worm. Worms look like that.” Again, the
knowledge and action information were counterbalanced. The experimen-
tal questions were the following: (a) “Is Luna drawing a worm?” (draw-
objective) and (b) “Would Luna say she is drawing a worm?” (draw-say).
For the second draw story, Luna was drawing a dog that also looked like
a cat.
Results and Discussion
All children were told each story and were given a score of 1 for
each correct response. Thus, scores for each variable (pretend-
objective, pretend-say, draw-objective, and draw-say) ranged
from 0 to 2, with 0 indicating the child did not respond correctly
Figure 4.
Sample stimuli for Experiment 3: Moe climbing a tree.
on the question for either story and 2 indicating the child re-

---

CHILDREN’S KNOWLEDGE OF DRAWING AND PRETENDING
1013
sponded correctly on the question for both stories. The numbers of
children responding correctly to zero, one, or two of the questions
for each of the variables (objective and say questions) are listed in
Table 3. Chi-square goodness of fit tests were conducted to test for
chance responding. All responses were significantly different from
chance (see Table 1).
Figure 6 illustrates the mean response (out of 2) of children in
each age group. Responses were similar to responses in the re-
spective explanation conditions in Experiment 2. On the draw
questions, 4-year-olds had means of 0.63 (SD
0.81) and 0.77
(SD
0.82) on draw-objective and draw-say, respectively. The
5-year-olds had means of 0.43 (SD
0.77) and 0.80 (SD
0.89)
on draw-objective and draw-say, respectively. On the pretense
questions, 4-year-olds had means of 0.47 (SD
0.73) and 0.73
(SD
0.87) on pretend-objective and pretend-say, respectively.
The 5-year-olds had means of 0.57 (SD
0.86) and 0.77
Figure 6.
Mean response (out of 2) and standard error of children in each
(SD
0.90) on pretend-objective and pretend-say, respectively.
age group in Experiment 3.
Mann–Whitney U tests were conducted to compare the re-
sponses of the age groups on the different variables. There were no
significant differences by age, indicating that the 5-year-olds did
performance in general on the draw-objective question was likely
not perform differently than the 4-year-olds on any of the
reflective of the fact that participants interpreted the question as
questions.
asking about the appearance of the drawing and not the artist’s
Wilcoxon signed-ranks tests for matched pairs were conducted
intention.
on the within-subjects variables (draw-objective vs. draw-say,
The results from Experiment 3 indicate that despite numerous
pretend-objective vs. pretend-say, draw-objective vs. pretend-
attempts to make the mental states of the protagonists as salient as
objective, and draw-say vs. pretend-say). There were no significant
possible for children, most 4- and 5-year-olds continued to respond
differences among the 5-year-olds’ responses. For the 4-year-olds,
as deconstructionists, denying a role to the artists’ and pretenders’
the only significant difference was between pretend-objective and
mental representations in interpreting the outcomes of representa-
pretend-say (Z
1.99, p
.05), indicating that 4-year-olds
tive actions.
performed significantly better on the pretense questions when the
mental state was highlighted in the questions.
The discrepancy in responding for the draw-objective and draw-
General Discussion
think/say questions evidenced in Experiment 1 disappeared with
the adjusted procedure. This suggests that highlighting the mental
Three experiments examined whether preschoolers’ difficulty in
state in the procedure with an explanation serves the same function
understanding that mental representations underpin pretense exter-
as highlighti